IETdoi

wwwietdlorg

Published in IET Renewable Power GenerationReceived on 14th July 2008Revised on 19th December 2008doi 101049iet-rpg20080064

ISSN 1752-1416

Coordinated voltage support in distributionnetworks with distributed generationand microgridsAG Madureira JAPecas LopesPower Systems Unit of INESC Porto and Faculty of Engineering of Porto University Rua Dr Roberto Frias 378 Porto 4200-465PortugalE-mail agminescportopt

Abstract This paper proposes a new methodology for coordinated voltage support in distribution networks withlarge integration of distributed generation and microgrids Given the characteristics of the LV networks it isshown that traditional control strategies using only reactive power control may not be sufficient in order toperform efficient voltage control Therefore microgeneration shedding must also be employed especially inscenarios with extreme microgeneration penetration An optimisation tool based on a meta-heuristic approachwas developed to address the voltage control problem In addition neural networks were employed in orderto decrease computational time thus enabling the use of the tool for online operation The results obtainedrevealed good performance of this control approach

1 IntroductionElectrical distribution networks have been undergoingsignificant changes in the last few years due to the growthof distributed generation (DG) Nowadays due to therestructuring process in electrical distribution systems andthe implementation of market structures in severalEuropean countries new business opportunities are arisingfor DG units Particularly the connection of these units tothe electric power system brings additional controlpossibilities to the distribution system operator (DSO) andnew technical developments enable DG participation inproviding ancillary services such as reserves and voltagesupport [1]

Usually DG units are not subject to centralis dispatch andreactive power generation is in most cases restricted by theDSO Therefore several changes are required in order tofully profit from the benefits resulting from DGintegration and voltage support emerges as one of themain services to be provided by DG units This resultsfrom the fact that a significant growth of DG penetrationwill require new operation philosophies in order to exploitreactive power generation capability of DG units with the

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

objective of optimising network operation minimisingactive power losses and maintaining voltage profiles

Several authors have developed interesting workconcerning the impact of DG on voltage control indistribution grids [2ndash5] In [6] the authors formulate analgorithm for voltage control in distribution grids with DGby solving an optimisation problem where active powerlosses are minimised subject to a set of technicalconstraints The control variables to consider are DGreactive power generation On-line tap changing (OLTC)transformer settings and capacitor bank settings

One issue that frequently results from high DGpenetration is voltage rise problems especially in weakdistribution networks [3 7] To overcome this problem itis necessary to control both active and reactive power of theDG units andor reduce the voltage at the HVMVsubstation [3]

This voltage rise effect is partly the result of DSO policiesbased on a lsquofit and forgetrsquo approach which also requires DGto operate at a fixed power factor thus limiting the

439amp The Institution of Engineering and Technology 2009

440

amp

wwwietdlorg

integration of DG and failing to take full profit of DG abilityto mitigate such effects [2 8]

In addition it is expected that a similar phenomenon willtake place at the low voltage (LV) side of the distributionsystem where microgeneration growth will develop rapidlyThe connection of microgeneration to LV networkscreating microgrids will be playing an important role infuture distribution networks The effects seen at themedium voltage (MV) level may propagate to the LV sidewith even worse consequences given the highmicrogeneration penetration

To overcome the problems resulting from high DG andmicrogeneration penetration an effective control schememust be developed Formerly in conventional systemsvoltage control was usually considered as a decentralisedcontrol problem This has mainly to do with the fact thatvoltage is predominantly a local or regional problem

However considering this new operation scenario adecentralised yet hierarchical voltage control scheme mustbe envisaged exploiting communication and controlpossibilities available for microgrid operation [9]

This paper describes a new tool to be used at thedistribution management system (DMS) level in order tomanage in an integrated way voltage control in MVLVdistribution grids in scenarios characterised by large-scaleintegration of DG connected at the MV level andmicrogeneration connected at the LV level

2 Multi-microgrid systemarchitectureA microgrid comprises an LV feeder with several microsourcesstorage devices and controllable loads connected on the samefeeder [10] These LV microgrids may be operated ininterconnected or in islanded mode (under emergencyconditions) and are managed by a microgrid central controller(MGCC) that includes several key functions such aseconomic management and control functionalities [10 11]

The new concept of a multi-microgrid is related to a multi-level structure existing at the MV distribution levelconsisting of LV microgrids and DG units connected onadjacent MV feeders This concept has been developedwithin the framework of European project More MicroGrids[12] Therefore microgrids DG units and MV loads underdemand side management control can be considered in thisnetwork as active cells for control and management purposesThe technical operation of such a system requires thetransposition of the microgrid concept to the MV levelwhere all these active cells should be controlled by a centralautonomous management controller (CAMC) to be installedat the MV bus level of an HVMV substation The CAMCwill serve as an interface to the DMS and operate under theresponsibility of the DSO [9]

The Institution of Engineering and Technology 2009

This architecture can be seen in Fig 1

Nowadays the DMS is responsible for the supervisioncontrol and management of the distribution system Infuture power systems in addition to the DMS there willbe new management levels

dagger The CAMC to be housed in HVMV substations whichwill accommodate functionalities that are normally assignedto the DMS (or other new functionalities) and will beresponsible for interfacing the DMS with lower level controllers

dagger The MGCC to be housed in MVLV substations whichwill be responsible for managing the microgrid including thecontrol of the microsources and responsive loads Voltagemonitoring in each LV grid will be performed using themicrogrid communication infrastructure

Microgrids together with DG units will have a significantimpact on the electrical distribution system and will enablethe participation of these units in providing ancillaryservices such as coordinated voltage support

3 Voltage control in multi-microgridsA hierarchical control system must then be established forvoltage control in distribution systems comprising large DGand microgeneration penetration using communication andcontrol possibilities that will become available in futuredistribution networks

The main objective of this voltage control system is toensure an optimised and coordinated strategy between the

Figure 1 Control and management architecture of a multi-microgrid system

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

several voltage levels in the distribution system namely MVand LV

As previously mentioned in extreme situations wherethere is significant voltage rise due to massive DG andmicrogeneration penetration reactive power control is notsufficient to maintain efficient system operation especiallyin LV networks where the XR index is low Consider theexample system presented in Fig 2

Given the example system presented and considering thatin LV networks the line resistance is greater than the linereactance (ie R X ) the following expression may bederived from the power flow equations

Pinj frac14V 2

2

R

V2V1

Rcos deth THORN (1)

where Pinj is the active power injected in the MV network V2

is the bus voltage at the LV network (microgrid) V1 is the busvoltage at the MV network R is the branch resistance and d

is the angle between V1 and V2

According to (1) it may be seen that in order to be able toinject active power (Pinj) from the LV side (microgrid) to theMV side in resistive networks voltage should be higher at theLV bus (ie V2 V1)

It must be stressed that the low XR ratio applies only tobranches in the LV microgrid since in this worktransformers in grid-connected mode are not modelled inthe LV microgrid

Therefore high DG and microgeneration penetration willrequire the development of an effective voltage controlscheme based on active and reactive power control sincethe decoupling between active power and voltage is notvalid in LV networks In the case of LV grids withmicrogeneration the possibility of controlling active powerinjected by microgeneration units has to be envisaged since

Figure 2 Example system

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

this is the most effective way of controling voltage at theLV level under these conditions

In this paper a new voltage control procedure is proposedthat includes optimising operating conditions by using DGinstalled at the MV level microgrid and OLTC transformercontrol capabilities In multi-microgrid systems it is necessaryto address the problem of voltage control at both the MV andthe LV levels To ensure a coordinated operation a globalvoltage control algorithm will run at the MV level and thesolution obtained will be tested at the LV microgrid level inorder to evaluate its feasibility This approach requires asequence of global problem solutions and local sub-problemsolutions in order to converge to a near-optimum solution

The voltage control system is designed to be a real-timeapplication that helps the DSO to efficiently manage thedistribution network This proposed functionality comprises apreliminary stage where several studies are performed offlinein order to evaluate the performance of the algorithm giventhe characteristics of the network and adequately select theparameters for the optimisation problem formulation Onlythen the optimisation tool may be used in real-timeoperation and made available to the operator

31 Mathematical formulation

The voltage control problem for multi-microgrid systems is anon-linear optimisation problem that can be formulated asshown in (2)

min OF(X )

subject to

V mini Vi V max

i

Sminik Sik Smax

ik

tmini ti tmax

i

Qmini Qi Qmax

i

(2)

where OF is the objective function X is the control variablesVi is the voltage at bus i V i

min V imax are the minimum and

maximum voltage at bus i Sik is the power flow in branchik Sik

min Sikmax are the minimum and maximum power flows

in branch ik ti is the transformer tap of or capacitor stepposition andti

min timax are the minimum and maximum tap

The objective function chosen aims at minimising activepower losses and microgeneration shedding and is shown in (3)

minX

Ploss thornX

mGshed (3)

where Ploss is the active power losses andmGshed is the amount ofmicrogeneration shed

Active power losses were assessed as a function of the lineresistance and the current flowing in each line after a power

441

amp The Institution of Engineering and Technology 2009

442

amp

wwwietdlorg

flow routine The total active power losses used in theobjective function are calculated as the sum of the activepower losses in each line

32 Optimisation algorithm

Given the characteristics of the problem under analysis with alarge dimension and a mixed continuousinteger natureconcerning the control variables (possibility of controllingactive and reactive generation levels ndash continuous variablesand transformer tapscapacitor banks ndash integer variables) ameta-heuristic approach was chosen The optimisationalgorithm used in this work was evolutionary particle swarmoptimization (EPSO) This algorithm is a combination oftraditional particle swarm intelligence developed by Kennedyand Eberhardt [13] and evolutionary strategies developed bySchwefel [14] and has been used extensively in optimisationproblems for electrical power systems More details on thealgorithm can be found in [15 16]

The variables or parameters in EPSO are divided into objectparameters (the X variables ndash control parameters of the problem)and strategic parameters (the weights w) The algorithmconsiders a set of solutions or alternatives that are calledparticles The X variables include all the control variables usedin the voltage and reactive power control optimisationproblem as described in Section 31 Strategic parameters (w)are used to control the behaviour of the optimisation algorithm

In EPSO each particle (or solution at a given stage) isdefined by its position Xi

k and velocity vik for the coordinate

position i and particle k

The general scheme of EPSO is presented next

dagger Replication ndash each particle is replicated r times

dagger Mutation ndash each particle has its weights w mutated

dagger Reproduction ndash each particle generates an offspringaccording to the particle movement rule

dagger Evaluation ndash each offspring has its fitness evaluated

dagger Selection ndash the best particles are selected by stochastictournament

The particle movement rule is the following given aparticle Xi a new particle Xi

new can be obtained from (4)

X newi frac14 Xi thorn vnew

i (4)

with

vnewi frac14 wi0Vi thorn wi1(bi Xi)thorn wi2(bg Xi) (5)

where Xi is the position of the particle vi is the velocity of theparticle wik is the strategic parameter (weight) bi is the best

The Institution of Engineering and Technology 2009

solution of each particle and bg is the best solution among allparticles

The weights (wik) are mutated as shown in (6)

wik frac14 wik thorn tN (0 1) (6)

where t is the fixed learning parameter and N(0 1) is therandom variable with Gaussian distribution 0 mean andvariance 1

The global best (bg) is also mutated as shown in (7)

bg frac14 bg thorn t0N (0 1) (7)

where t0 is the fixed learning parameter

The control variables used in this work were

dagger Reactive power from DG sources

dagger Active power from microgrids by means ofmicrogeneration shedding (assuming that all microsourcesare curtailed in the same proportion)

dagger Reactive power from microgrids by means of capacitorbanks located at the MVLV transformer substation ofeach microgrid

dagger OLTC transformer settings at distribution transformers

These control variables are used as set-points sent to eachdevice (DG unit microgrid or OLTC transformer) using thecontrol scheme presented in Fig 1 For instance consideringthe active power of microgeneration a set-point is sent to theMGCC indicating that microgeneration should be reducedaccording to that set-point This information is then sentto individual microsource controllers (as defined in thecontrol structure presented in [10]) that receive anindividual set-point in order to lower their generation

In this work the constraints were dealt with in traditionalevolutionary strategies that is using a penalty approach

33 Artificial neural network

As previously mentioned the voltage control schemepresented is intended to be used as an online functionmade available to the DSO Therefore and in order tospeed up the control algorithm an artificial neural network(ANN) able to emulate the behaviour of the LV network(or microgrid) was included This option enables the use ofthe meta-heuristic tool employed in real-time operationreducing the long simulation times that were required inorder to calculate consecutive LV power flows In fact theANN is used to provide a steady-state equivalent of themicrogrid avoiding in this way the extension of the loadflow analysis to the LV microgrid level Using the ANNthe computational burden and consequently thecomputational time have significantly decreased

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

The ANN to be employed has the following inputs

dagger Voltage at the MVLV substation

dagger Active power generated by each microgenerator unit

dagger Total load of the microgrid

The two outputs are the active power losses and themaximum voltage inside the microgrid since one of themost critical problems to be addressed is related withovervoltage problems that may occur

To achieve good performance different ANN topologies(different number of layers and number of neurons in eachhidden layer) were tested The ANN with the bestperformance has eight inputs one hidden layer with 24neurons and two outputs

To generate the data set corresponding to the inputschosen to train the ANN a large number of power flowswere computed considering different combinations of theinputs (ie several values for the voltage reference at theMVLV substation for the power generated by eachmicrosource and for the load) in order to calculate theactive power losses and assess the voltage profiles in eachscenario

The generated data set contains 11 664 operating pointswhich were split into a training set and a test setcontaining 23 and 13 of the total operating points

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

respectively The training set was used for training theANN and the test set for performance evaluation purposesThe MATLAB Neural Network Toolbox using theLevenbergndashMarquart back-propagation algorithm was usedto identify the ANN parameters

The performance obtained can be evaluated in terms of themean absolute error (MAE) The MAE obtained for one ofthe outputs (active power losses) was 101 1023 Despitethe simple structure adopted for the ANN theperformance parameters illustrated the quality of the toolfor emulating the behaviour of the microgrid

Nevertheless in case of microgrid reconfiguration (due toadding of a new microgenerator for instance) a new ANNmust be computed This means that it is necessary toupdate the data for the power flow and then generate a newdata set and train a new ANN that emulates the LVmicrogrid steady-state behaviour

4 Test case networks andscenariosTo test the voltage control approach developed two realdistribution networks were used an MV distributionnetwork and an LV distribution network shown in Figs 3and 4 respectively

The MV network used is a rural network with two distinctareas with different voltage levels 30 kV at node NO119 and

Figure 3 MV test network

443

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

15 kV after the 30 kV15 kV transformer at node NO206 Itcomprises six microgrids (marked as mG in Fig 3) all with asimilar topology and 3 DG units (marked as DG in Fig 3)based on Combined Heat and Power ndash CHP (NO3 inFig 3) and wind (NO16 and N061 in Fig 3)

The LV network (microgrid) used is also a rural networkwith a radial structure It was considered that all microgridshave the same structure (shown in Fig 4) As previouslystated this LV network was used to train the ANN thatwas included in the optimisation algorithm

Data on the test networks used in this work are presentedin Tables 2ndash5

Each microgrid is supposed to comprise sixmicrogenerators all photovoltaic units (marked as PV inFig 4) The ANN is however capable of dealing withdifferent levels of microgeneration penetration as well aswith different locations for the microsources

Figure 4 LV test network

4The Institution of Engineering and Technology 2009

In this approach from the MV point of view each LVmicrogrid was considered as a single bus with an equivalentgenerator (corresponding to the sum of all micro-sourcegenerations) and equivalent load (corresponding to the sumof all LV loads)

Typical generation curves for each generating technology(CHP Wind and PV) were used as well as typical loadcurves for a 24 h period

The total generation installed capacity for the scenariosused is presented in Table 1

The 24 h profile of the total load (a mix of residential andcommercial consumers) in the MV network is presented inFig 5

Table 1 Test case scenarios for generation

Testnetwork

Distributed generationmicrogeneration

Installed capacityMW

Percentage peakload

MV 36 60

LV 01 100

Figure 5 MV test network total load

Figure 6 Maximum voltage values in microgrid at node 64

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Figure 8 Active power losses in microgrid at node 64

Figure 9 Total microgeneration in microgrid at node 64

Figure 7 Active power exported to the MV network by the microgrid at node 64

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 445 101049iet-rpg20080064 amp The Institution of Engineering and Technology 2009

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

440

amp

wwwietdlorg

integration of DG and failing to take full profit of DG abilityto mitigate such effects [2 8]

In addition it is expected that a similar phenomenon willtake place at the low voltage (LV) side of the distributionsystem where microgeneration growth will develop rapidlyThe connection of microgeneration to LV networkscreating microgrids will be playing an important role infuture distribution networks The effects seen at themedium voltage (MV) level may propagate to the LV sidewith even worse consequences given the highmicrogeneration penetration

To overcome the problems resulting from high DG andmicrogeneration penetration an effective control schememust be developed Formerly in conventional systemsvoltage control was usually considered as a decentralisedcontrol problem This has mainly to do with the fact thatvoltage is predominantly a local or regional problem

However considering this new operation scenario adecentralised yet hierarchical voltage control scheme mustbe envisaged exploiting communication and controlpossibilities available for microgrid operation [9]

This paper describes a new tool to be used at thedistribution management system (DMS) level in order tomanage in an integrated way voltage control in MVLVdistribution grids in scenarios characterised by large-scaleintegration of DG connected at the MV level andmicrogeneration connected at the LV level

2 Multi-microgrid systemarchitectureA microgrid comprises an LV feeder with several microsourcesstorage devices and controllable loads connected on the samefeeder [10] These LV microgrids may be operated ininterconnected or in islanded mode (under emergencyconditions) and are managed by a microgrid central controller(MGCC) that includes several key functions such aseconomic management and control functionalities [10 11]

The new concept of a multi-microgrid is related to a multi-level structure existing at the MV distribution levelconsisting of LV microgrids and DG units connected onadjacent MV feeders This concept has been developedwithin the framework of European project More MicroGrids[12] Therefore microgrids DG units and MV loads underdemand side management control can be considered in thisnetwork as active cells for control and management purposesThe technical operation of such a system requires thetransposition of the microgrid concept to the MV levelwhere all these active cells should be controlled by a centralautonomous management controller (CAMC) to be installedat the MV bus level of an HVMV substation The CAMCwill serve as an interface to the DMS and operate under theresponsibility of the DSO [9]

The Institution of Engineering and Technology 2009

This architecture can be seen in Fig 1

Nowadays the DMS is responsible for the supervisioncontrol and management of the distribution system Infuture power systems in addition to the DMS there willbe new management levels

dagger The CAMC to be housed in HVMV substations whichwill accommodate functionalities that are normally assignedto the DMS (or other new functionalities) and will beresponsible for interfacing the DMS with lower level controllers

dagger The MGCC to be housed in MVLV substations whichwill be responsible for managing the microgrid including thecontrol of the microsources and responsive loads Voltagemonitoring in each LV grid will be performed using themicrogrid communication infrastructure

Microgrids together with DG units will have a significantimpact on the electrical distribution system and will enablethe participation of these units in providing ancillaryservices such as coordinated voltage support

3 Voltage control in multi-microgridsA hierarchical control system must then be established forvoltage control in distribution systems comprising large DGand microgeneration penetration using communication andcontrol possibilities that will become available in futuredistribution networks

The main objective of this voltage control system is toensure an optimised and coordinated strategy between the

Figure 1 Control and management architecture of a multi-microgrid system

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

several voltage levels in the distribution system namely MVand LV

As previously mentioned in extreme situations wherethere is significant voltage rise due to massive DG andmicrogeneration penetration reactive power control is notsufficient to maintain efficient system operation especiallyin LV networks where the XR index is low Consider theexample system presented in Fig 2

Given the example system presented and considering thatin LV networks the line resistance is greater than the linereactance (ie R X ) the following expression may bederived from the power flow equations

Pinj frac14V 2

2

R

V2V1

Rcos deth THORN (1)

where Pinj is the active power injected in the MV network V2

is the bus voltage at the LV network (microgrid) V1 is the busvoltage at the MV network R is the branch resistance and d

is the angle between V1 and V2

According to (1) it may be seen that in order to be able toinject active power (Pinj) from the LV side (microgrid) to theMV side in resistive networks voltage should be higher at theLV bus (ie V2 V1)

It must be stressed that the low XR ratio applies only tobranches in the LV microgrid since in this worktransformers in grid-connected mode are not modelled inthe LV microgrid

Therefore high DG and microgeneration penetration willrequire the development of an effective voltage controlscheme based on active and reactive power control sincethe decoupling between active power and voltage is notvalid in LV networks In the case of LV grids withmicrogeneration the possibility of controlling active powerinjected by microgeneration units has to be envisaged since

Figure 2 Example system

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

this is the most effective way of controling voltage at theLV level under these conditions

In this paper a new voltage control procedure is proposedthat includes optimising operating conditions by using DGinstalled at the MV level microgrid and OLTC transformercontrol capabilities In multi-microgrid systems it is necessaryto address the problem of voltage control at both the MV andthe LV levels To ensure a coordinated operation a globalvoltage control algorithm will run at the MV level and thesolution obtained will be tested at the LV microgrid level inorder to evaluate its feasibility This approach requires asequence of global problem solutions and local sub-problemsolutions in order to converge to a near-optimum solution

The voltage control system is designed to be a real-timeapplication that helps the DSO to efficiently manage thedistribution network This proposed functionality comprises apreliminary stage where several studies are performed offlinein order to evaluate the performance of the algorithm giventhe characteristics of the network and adequately select theparameters for the optimisation problem formulation Onlythen the optimisation tool may be used in real-timeoperation and made available to the operator

31 Mathematical formulation

The voltage control problem for multi-microgrid systems is anon-linear optimisation problem that can be formulated asshown in (2)

min OF(X )

subject to

V mini Vi V max

i

Sminik Sik Smax

ik

tmini ti tmax

i

Qmini Qi Qmax

i

(2)

where OF is the objective function X is the control variablesVi is the voltage at bus i V i

min V imax are the minimum and

maximum voltage at bus i Sik is the power flow in branchik Sik

min Sikmax are the minimum and maximum power flows

in branch ik ti is the transformer tap of or capacitor stepposition andti

min timax are the minimum and maximum tap

The objective function chosen aims at minimising activepower losses and microgeneration shedding and is shown in (3)

minX

Ploss thornX

mGshed (3)

where Ploss is the active power losses andmGshed is the amount ofmicrogeneration shed

Active power losses were assessed as a function of the lineresistance and the current flowing in each line after a power

441

amp The Institution of Engineering and Technology 2009

442

amp

wwwietdlorg

flow routine The total active power losses used in theobjective function are calculated as the sum of the activepower losses in each line

32 Optimisation algorithm

Given the characteristics of the problem under analysis with alarge dimension and a mixed continuousinteger natureconcerning the control variables (possibility of controllingactive and reactive generation levels ndash continuous variablesand transformer tapscapacitor banks ndash integer variables) ameta-heuristic approach was chosen The optimisationalgorithm used in this work was evolutionary particle swarmoptimization (EPSO) This algorithm is a combination oftraditional particle swarm intelligence developed by Kennedyand Eberhardt [13] and evolutionary strategies developed bySchwefel [14] and has been used extensively in optimisationproblems for electrical power systems More details on thealgorithm can be found in [15 16]

The variables or parameters in EPSO are divided into objectparameters (the X variables ndash control parameters of the problem)and strategic parameters (the weights w) The algorithmconsiders a set of solutions or alternatives that are calledparticles The X variables include all the control variables usedin the voltage and reactive power control optimisationproblem as described in Section 31 Strategic parameters (w)are used to control the behaviour of the optimisation algorithm

In EPSO each particle (or solution at a given stage) isdefined by its position Xi

k and velocity vik for the coordinate

position i and particle k

The general scheme of EPSO is presented next

dagger Replication ndash each particle is replicated r times

dagger Mutation ndash each particle has its weights w mutated

dagger Reproduction ndash each particle generates an offspringaccording to the particle movement rule

dagger Evaluation ndash each offspring has its fitness evaluated

dagger Selection ndash the best particles are selected by stochastictournament

The particle movement rule is the following given aparticle Xi a new particle Xi

new can be obtained from (4)

X newi frac14 Xi thorn vnew

i (4)

with

vnewi frac14 wi0Vi thorn wi1(bi Xi)thorn wi2(bg Xi) (5)

where Xi is the position of the particle vi is the velocity of theparticle wik is the strategic parameter (weight) bi is the best

The Institution of Engineering and Technology 2009

solution of each particle and bg is the best solution among allparticles

The weights (wik) are mutated as shown in (6)

wik frac14 wik thorn tN (0 1) (6)

where t is the fixed learning parameter and N(0 1) is therandom variable with Gaussian distribution 0 mean andvariance 1

The global best (bg) is also mutated as shown in (7)

bg frac14 bg thorn t0N (0 1) (7)

where t0 is the fixed learning parameter

The control variables used in this work were

dagger Reactive power from DG sources

dagger Active power from microgrids by means ofmicrogeneration shedding (assuming that all microsourcesare curtailed in the same proportion)

dagger Reactive power from microgrids by means of capacitorbanks located at the MVLV transformer substation ofeach microgrid

dagger OLTC transformer settings at distribution transformers

These control variables are used as set-points sent to eachdevice (DG unit microgrid or OLTC transformer) using thecontrol scheme presented in Fig 1 For instance consideringthe active power of microgeneration a set-point is sent to theMGCC indicating that microgeneration should be reducedaccording to that set-point This information is then sentto individual microsource controllers (as defined in thecontrol structure presented in [10]) that receive anindividual set-point in order to lower their generation

In this work the constraints were dealt with in traditionalevolutionary strategies that is using a penalty approach

33 Artificial neural network

As previously mentioned the voltage control schemepresented is intended to be used as an online functionmade available to the DSO Therefore and in order tospeed up the control algorithm an artificial neural network(ANN) able to emulate the behaviour of the LV network(or microgrid) was included This option enables the use ofthe meta-heuristic tool employed in real-time operationreducing the long simulation times that were required inorder to calculate consecutive LV power flows In fact theANN is used to provide a steady-state equivalent of themicrogrid avoiding in this way the extension of the loadflow analysis to the LV microgrid level Using the ANNthe computational burden and consequently thecomputational time have significantly decreased

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

The ANN to be employed has the following inputs

dagger Voltage at the MVLV substation

dagger Active power generated by each microgenerator unit

dagger Total load of the microgrid

The two outputs are the active power losses and themaximum voltage inside the microgrid since one of themost critical problems to be addressed is related withovervoltage problems that may occur

To achieve good performance different ANN topologies(different number of layers and number of neurons in eachhidden layer) were tested The ANN with the bestperformance has eight inputs one hidden layer with 24neurons and two outputs

To generate the data set corresponding to the inputschosen to train the ANN a large number of power flowswere computed considering different combinations of theinputs (ie several values for the voltage reference at theMVLV substation for the power generated by eachmicrosource and for the load) in order to calculate theactive power losses and assess the voltage profiles in eachscenario

The generated data set contains 11 664 operating pointswhich were split into a training set and a test setcontaining 23 and 13 of the total operating points

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

respectively The training set was used for training theANN and the test set for performance evaluation purposesThe MATLAB Neural Network Toolbox using theLevenbergndashMarquart back-propagation algorithm was usedto identify the ANN parameters

The performance obtained can be evaluated in terms of themean absolute error (MAE) The MAE obtained for one ofthe outputs (active power losses) was 101 1023 Despitethe simple structure adopted for the ANN theperformance parameters illustrated the quality of the toolfor emulating the behaviour of the microgrid

Nevertheless in case of microgrid reconfiguration (due toadding of a new microgenerator for instance) a new ANNmust be computed This means that it is necessary toupdate the data for the power flow and then generate a newdata set and train a new ANN that emulates the LVmicrogrid steady-state behaviour

4 Test case networks andscenariosTo test the voltage control approach developed two realdistribution networks were used an MV distributionnetwork and an LV distribution network shown in Figs 3and 4 respectively

The MV network used is a rural network with two distinctareas with different voltage levels 30 kV at node NO119 and

Figure 3 MV test network

443

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

15 kV after the 30 kV15 kV transformer at node NO206 Itcomprises six microgrids (marked as mG in Fig 3) all with asimilar topology and 3 DG units (marked as DG in Fig 3)based on Combined Heat and Power ndash CHP (NO3 inFig 3) and wind (NO16 and N061 in Fig 3)

The LV network (microgrid) used is also a rural networkwith a radial structure It was considered that all microgridshave the same structure (shown in Fig 4) As previouslystated this LV network was used to train the ANN thatwas included in the optimisation algorithm

Data on the test networks used in this work are presentedin Tables 2ndash5

Each microgrid is supposed to comprise sixmicrogenerators all photovoltaic units (marked as PV inFig 4) The ANN is however capable of dealing withdifferent levels of microgeneration penetration as well aswith different locations for the microsources

Figure 4 LV test network

4The Institution of Engineering and Technology 2009

In this approach from the MV point of view each LVmicrogrid was considered as a single bus with an equivalentgenerator (corresponding to the sum of all micro-sourcegenerations) and equivalent load (corresponding to the sumof all LV loads)

Typical generation curves for each generating technology(CHP Wind and PV) were used as well as typical loadcurves for a 24 h period

The total generation installed capacity for the scenariosused is presented in Table 1

The 24 h profile of the total load (a mix of residential andcommercial consumers) in the MV network is presented inFig 5

Table 1 Test case scenarios for generation

Testnetwork

Distributed generationmicrogeneration

Installed capacityMW

Percentage peakload

MV 36 60

LV 01 100

Figure 5 MV test network total load

Figure 6 Maximum voltage values in microgrid at node 64

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Figure 8 Active power losses in microgrid at node 64

Figure 9 Total microgeneration in microgrid at node 64

Figure 7 Active power exported to the MV network by the microgrid at node 64

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 445 101049iet-rpg20080064 amp The Institution of Engineering and Technology 2009

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

several voltage levels in the distribution system namely MVand LV

As previously mentioned in extreme situations wherethere is significant voltage rise due to massive DG andmicrogeneration penetration reactive power control is notsufficient to maintain efficient system operation especiallyin LV networks where the XR index is low Consider theexample system presented in Fig 2

Given the example system presented and considering thatin LV networks the line resistance is greater than the linereactance (ie R X ) the following expression may bederived from the power flow equations

Pinj frac14V 2

2

R

V2V1

Rcos deth THORN (1)

where Pinj is the active power injected in the MV network V2

is the bus voltage at the LV network (microgrid) V1 is the busvoltage at the MV network R is the branch resistance and d

is the angle between V1 and V2

According to (1) it may be seen that in order to be able toinject active power (Pinj) from the LV side (microgrid) to theMV side in resistive networks voltage should be higher at theLV bus (ie V2 V1)

It must be stressed that the low XR ratio applies only tobranches in the LV microgrid since in this worktransformers in grid-connected mode are not modelled inthe LV microgrid

Therefore high DG and microgeneration penetration willrequire the development of an effective voltage controlscheme based on active and reactive power control sincethe decoupling between active power and voltage is notvalid in LV networks In the case of LV grids withmicrogeneration the possibility of controlling active powerinjected by microgeneration units has to be envisaged since

Figure 2 Example system

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

this is the most effective way of controling voltage at theLV level under these conditions

In this paper a new voltage control procedure is proposedthat includes optimising operating conditions by using DGinstalled at the MV level microgrid and OLTC transformercontrol capabilities In multi-microgrid systems it is necessaryto address the problem of voltage control at both the MV andthe LV levels To ensure a coordinated operation a globalvoltage control algorithm will run at the MV level and thesolution obtained will be tested at the LV microgrid level inorder to evaluate its feasibility This approach requires asequence of global problem solutions and local sub-problemsolutions in order to converge to a near-optimum solution

The voltage control system is designed to be a real-timeapplication that helps the DSO to efficiently manage thedistribution network This proposed functionality comprises apreliminary stage where several studies are performed offlinein order to evaluate the performance of the algorithm giventhe characteristics of the network and adequately select theparameters for the optimisation problem formulation Onlythen the optimisation tool may be used in real-timeoperation and made available to the operator

31 Mathematical formulation

The voltage control problem for multi-microgrid systems is anon-linear optimisation problem that can be formulated asshown in (2)

min OF(X )

subject to

V mini Vi V max

i

Sminik Sik Smax

ik

tmini ti tmax

i

Qmini Qi Qmax

i

(2)

where OF is the objective function X is the control variablesVi is the voltage at bus i V i

min V imax are the minimum and

maximum voltage at bus i Sik is the power flow in branchik Sik

min Sikmax are the minimum and maximum power flows

in branch ik ti is the transformer tap of or capacitor stepposition andti

min timax are the minimum and maximum tap

The objective function chosen aims at minimising activepower losses and microgeneration shedding and is shown in (3)

minX

Ploss thornX

mGshed (3)

where Ploss is the active power losses andmGshed is the amount ofmicrogeneration shed

Active power losses were assessed as a function of the lineresistance and the current flowing in each line after a power

441

amp The Institution of Engineering and Technology 2009

442

amp

wwwietdlorg

flow routine The total active power losses used in theobjective function are calculated as the sum of the activepower losses in each line

32 Optimisation algorithm

Given the characteristics of the problem under analysis with alarge dimension and a mixed continuousinteger natureconcerning the control variables (possibility of controllingactive and reactive generation levels ndash continuous variablesand transformer tapscapacitor banks ndash integer variables) ameta-heuristic approach was chosen The optimisationalgorithm used in this work was evolutionary particle swarmoptimization (EPSO) This algorithm is a combination oftraditional particle swarm intelligence developed by Kennedyand Eberhardt [13] and evolutionary strategies developed bySchwefel [14] and has been used extensively in optimisationproblems for electrical power systems More details on thealgorithm can be found in [15 16]

The variables or parameters in EPSO are divided into objectparameters (the X variables ndash control parameters of the problem)and strategic parameters (the weights w) The algorithmconsiders a set of solutions or alternatives that are calledparticles The X variables include all the control variables usedin the voltage and reactive power control optimisationproblem as described in Section 31 Strategic parameters (w)are used to control the behaviour of the optimisation algorithm

In EPSO each particle (or solution at a given stage) isdefined by its position Xi

k and velocity vik for the coordinate

position i and particle k

The general scheme of EPSO is presented next

dagger Replication ndash each particle is replicated r times

dagger Mutation ndash each particle has its weights w mutated

dagger Reproduction ndash each particle generates an offspringaccording to the particle movement rule

dagger Evaluation ndash each offspring has its fitness evaluated

dagger Selection ndash the best particles are selected by stochastictournament

The particle movement rule is the following given aparticle Xi a new particle Xi

new can be obtained from (4)

X newi frac14 Xi thorn vnew

i (4)

with

vnewi frac14 wi0Vi thorn wi1(bi Xi)thorn wi2(bg Xi) (5)

where Xi is the position of the particle vi is the velocity of theparticle wik is the strategic parameter (weight) bi is the best

The Institution of Engineering and Technology 2009

solution of each particle and bg is the best solution among allparticles

The weights (wik) are mutated as shown in (6)

wik frac14 wik thorn tN (0 1) (6)

where t is the fixed learning parameter and N(0 1) is therandom variable with Gaussian distribution 0 mean andvariance 1

The global best (bg) is also mutated as shown in (7)

bg frac14 bg thorn t0N (0 1) (7)

where t0 is the fixed learning parameter

The control variables used in this work were

dagger Reactive power from DG sources

dagger Active power from microgrids by means ofmicrogeneration shedding (assuming that all microsourcesare curtailed in the same proportion)

dagger Reactive power from microgrids by means of capacitorbanks located at the MVLV transformer substation ofeach microgrid

dagger OLTC transformer settings at distribution transformers

These control variables are used as set-points sent to eachdevice (DG unit microgrid or OLTC transformer) using thecontrol scheme presented in Fig 1 For instance consideringthe active power of microgeneration a set-point is sent to theMGCC indicating that microgeneration should be reducedaccording to that set-point This information is then sentto individual microsource controllers (as defined in thecontrol structure presented in [10]) that receive anindividual set-point in order to lower their generation

In this work the constraints were dealt with in traditionalevolutionary strategies that is using a penalty approach

33 Artificial neural network

As previously mentioned the voltage control schemepresented is intended to be used as an online functionmade available to the DSO Therefore and in order tospeed up the control algorithm an artificial neural network(ANN) able to emulate the behaviour of the LV network(or microgrid) was included This option enables the use ofthe meta-heuristic tool employed in real-time operationreducing the long simulation times that were required inorder to calculate consecutive LV power flows In fact theANN is used to provide a steady-state equivalent of themicrogrid avoiding in this way the extension of the loadflow analysis to the LV microgrid level Using the ANNthe computational burden and consequently thecomputational time have significantly decreased

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

The ANN to be employed has the following inputs

dagger Voltage at the MVLV substation

dagger Active power generated by each microgenerator unit

dagger Total load of the microgrid

The two outputs are the active power losses and themaximum voltage inside the microgrid since one of themost critical problems to be addressed is related withovervoltage problems that may occur

To achieve good performance different ANN topologies(different number of layers and number of neurons in eachhidden layer) were tested The ANN with the bestperformance has eight inputs one hidden layer with 24neurons and two outputs

To generate the data set corresponding to the inputschosen to train the ANN a large number of power flowswere computed considering different combinations of theinputs (ie several values for the voltage reference at theMVLV substation for the power generated by eachmicrosource and for the load) in order to calculate theactive power losses and assess the voltage profiles in eachscenario

The generated data set contains 11 664 operating pointswhich were split into a training set and a test setcontaining 23 and 13 of the total operating points

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

respectively The training set was used for training theANN and the test set for performance evaluation purposesThe MATLAB Neural Network Toolbox using theLevenbergndashMarquart back-propagation algorithm was usedto identify the ANN parameters

The performance obtained can be evaluated in terms of themean absolute error (MAE) The MAE obtained for one ofthe outputs (active power losses) was 101 1023 Despitethe simple structure adopted for the ANN theperformance parameters illustrated the quality of the toolfor emulating the behaviour of the microgrid

Nevertheless in case of microgrid reconfiguration (due toadding of a new microgenerator for instance) a new ANNmust be computed This means that it is necessary toupdate the data for the power flow and then generate a newdata set and train a new ANN that emulates the LVmicrogrid steady-state behaviour

4 Test case networks andscenariosTo test the voltage control approach developed two realdistribution networks were used an MV distributionnetwork and an LV distribution network shown in Figs 3and 4 respectively

The MV network used is a rural network with two distinctareas with different voltage levels 30 kV at node NO119 and

Figure 3 MV test network

443

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

15 kV after the 30 kV15 kV transformer at node NO206 Itcomprises six microgrids (marked as mG in Fig 3) all with asimilar topology and 3 DG units (marked as DG in Fig 3)based on Combined Heat and Power ndash CHP (NO3 inFig 3) and wind (NO16 and N061 in Fig 3)

The LV network (microgrid) used is also a rural networkwith a radial structure It was considered that all microgridshave the same structure (shown in Fig 4) As previouslystated this LV network was used to train the ANN thatwas included in the optimisation algorithm

Data on the test networks used in this work are presentedin Tables 2ndash5

Each microgrid is supposed to comprise sixmicrogenerators all photovoltaic units (marked as PV inFig 4) The ANN is however capable of dealing withdifferent levels of microgeneration penetration as well aswith different locations for the microsources

Figure 4 LV test network

4The Institution of Engineering and Technology 2009

In this approach from the MV point of view each LVmicrogrid was considered as a single bus with an equivalentgenerator (corresponding to the sum of all micro-sourcegenerations) and equivalent load (corresponding to the sumof all LV loads)

Typical generation curves for each generating technology(CHP Wind and PV) were used as well as typical loadcurves for a 24 h period

The total generation installed capacity for the scenariosused is presented in Table 1

The 24 h profile of the total load (a mix of residential andcommercial consumers) in the MV network is presented inFig 5

Table 1 Test case scenarios for generation

Testnetwork

Distributed generationmicrogeneration

Installed capacityMW

Percentage peakload

MV 36 60

LV 01 100

Figure 5 MV test network total load

Figure 6 Maximum voltage values in microgrid at node 64

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Figure 8 Active power losses in microgrid at node 64

Figure 9 Total microgeneration in microgrid at node 64

Figure 7 Active power exported to the MV network by the microgrid at node 64

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 445 101049iet-rpg20080064 amp The Institution of Engineering and Technology 2009

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

442

amp

wwwietdlorg

flow routine The total active power losses used in theobjective function are calculated as the sum of the activepower losses in each line

32 Optimisation algorithm

Given the characteristics of the problem under analysis with alarge dimension and a mixed continuousinteger natureconcerning the control variables (possibility of controllingactive and reactive generation levels ndash continuous variablesand transformer tapscapacitor banks ndash integer variables) ameta-heuristic approach was chosen The optimisationalgorithm used in this work was evolutionary particle swarmoptimization (EPSO) This algorithm is a combination oftraditional particle swarm intelligence developed by Kennedyand Eberhardt [13] and evolutionary strategies developed bySchwefel [14] and has been used extensively in optimisationproblems for electrical power systems More details on thealgorithm can be found in [15 16]

The variables or parameters in EPSO are divided into objectparameters (the X variables ndash control parameters of the problem)and strategic parameters (the weights w) The algorithmconsiders a set of solutions or alternatives that are calledparticles The X variables include all the control variables usedin the voltage and reactive power control optimisationproblem as described in Section 31 Strategic parameters (w)are used to control the behaviour of the optimisation algorithm

In EPSO each particle (or solution at a given stage) isdefined by its position Xi

k and velocity vik for the coordinate

position i and particle k

The general scheme of EPSO is presented next

dagger Replication ndash each particle is replicated r times

dagger Mutation ndash each particle has its weights w mutated

dagger Reproduction ndash each particle generates an offspringaccording to the particle movement rule

dagger Evaluation ndash each offspring has its fitness evaluated

dagger Selection ndash the best particles are selected by stochastictournament

The particle movement rule is the following given aparticle Xi a new particle Xi

new can be obtained from (4)

X newi frac14 Xi thorn vnew

i (4)

with

vnewi frac14 wi0Vi thorn wi1(bi Xi)thorn wi2(bg Xi) (5)

where Xi is the position of the particle vi is the velocity of theparticle wik is the strategic parameter (weight) bi is the best

The Institution of Engineering and Technology 2009

solution of each particle and bg is the best solution among allparticles

The weights (wik) are mutated as shown in (6)

wik frac14 wik thorn tN (0 1) (6)

where t is the fixed learning parameter and N(0 1) is therandom variable with Gaussian distribution 0 mean andvariance 1

The global best (bg) is also mutated as shown in (7)

bg frac14 bg thorn t0N (0 1) (7)

where t0 is the fixed learning parameter

The control variables used in this work were

dagger Reactive power from DG sources

dagger Active power from microgrids by means ofmicrogeneration shedding (assuming that all microsourcesare curtailed in the same proportion)

dagger Reactive power from microgrids by means of capacitorbanks located at the MVLV transformer substation ofeach microgrid

dagger OLTC transformer settings at distribution transformers

These control variables are used as set-points sent to eachdevice (DG unit microgrid or OLTC transformer) using thecontrol scheme presented in Fig 1 For instance consideringthe active power of microgeneration a set-point is sent to theMGCC indicating that microgeneration should be reducedaccording to that set-point This information is then sentto individual microsource controllers (as defined in thecontrol structure presented in [10]) that receive anindividual set-point in order to lower their generation

In this work the constraints were dealt with in traditionalevolutionary strategies that is using a penalty approach

33 Artificial neural network

As previously mentioned the voltage control schemepresented is intended to be used as an online functionmade available to the DSO Therefore and in order tospeed up the control algorithm an artificial neural network(ANN) able to emulate the behaviour of the LV network(or microgrid) was included This option enables the use ofthe meta-heuristic tool employed in real-time operationreducing the long simulation times that were required inorder to calculate consecutive LV power flows In fact theANN is used to provide a steady-state equivalent of themicrogrid avoiding in this way the extension of the loadflow analysis to the LV microgrid level Using the ANNthe computational burden and consequently thecomputational time have significantly decreased

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

The ANN to be employed has the following inputs

dagger Voltage at the MVLV substation

dagger Active power generated by each microgenerator unit

dagger Total load of the microgrid

The two outputs are the active power losses and themaximum voltage inside the microgrid since one of themost critical problems to be addressed is related withovervoltage problems that may occur

To achieve good performance different ANN topologies(different number of layers and number of neurons in eachhidden layer) were tested The ANN with the bestperformance has eight inputs one hidden layer with 24neurons and two outputs

To generate the data set corresponding to the inputschosen to train the ANN a large number of power flowswere computed considering different combinations of theinputs (ie several values for the voltage reference at theMVLV substation for the power generated by eachmicrosource and for the load) in order to calculate theactive power losses and assess the voltage profiles in eachscenario

The generated data set contains 11 664 operating pointswhich were split into a training set and a test setcontaining 23 and 13 of the total operating points

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

respectively The training set was used for training theANN and the test set for performance evaluation purposesThe MATLAB Neural Network Toolbox using theLevenbergndashMarquart back-propagation algorithm was usedto identify the ANN parameters

The performance obtained can be evaluated in terms of themean absolute error (MAE) The MAE obtained for one ofthe outputs (active power losses) was 101 1023 Despitethe simple structure adopted for the ANN theperformance parameters illustrated the quality of the toolfor emulating the behaviour of the microgrid

Nevertheless in case of microgrid reconfiguration (due toadding of a new microgenerator for instance) a new ANNmust be computed This means that it is necessary toupdate the data for the power flow and then generate a newdata set and train a new ANN that emulates the LVmicrogrid steady-state behaviour

4 Test case networks andscenariosTo test the voltage control approach developed two realdistribution networks were used an MV distributionnetwork and an LV distribution network shown in Figs 3and 4 respectively

The MV network used is a rural network with two distinctareas with different voltage levels 30 kV at node NO119 and

Figure 3 MV test network

443

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

15 kV after the 30 kV15 kV transformer at node NO206 Itcomprises six microgrids (marked as mG in Fig 3) all with asimilar topology and 3 DG units (marked as DG in Fig 3)based on Combined Heat and Power ndash CHP (NO3 inFig 3) and wind (NO16 and N061 in Fig 3)

The LV network (microgrid) used is also a rural networkwith a radial structure It was considered that all microgridshave the same structure (shown in Fig 4) As previouslystated this LV network was used to train the ANN thatwas included in the optimisation algorithm

Data on the test networks used in this work are presentedin Tables 2ndash5

Each microgrid is supposed to comprise sixmicrogenerators all photovoltaic units (marked as PV inFig 4) The ANN is however capable of dealing withdifferent levels of microgeneration penetration as well aswith different locations for the microsources

Figure 4 LV test network

4The Institution of Engineering and Technology 2009

In this approach from the MV point of view each LVmicrogrid was considered as a single bus with an equivalentgenerator (corresponding to the sum of all micro-sourcegenerations) and equivalent load (corresponding to the sumof all LV loads)

Typical generation curves for each generating technology(CHP Wind and PV) were used as well as typical loadcurves for a 24 h period

The total generation installed capacity for the scenariosused is presented in Table 1

The 24 h profile of the total load (a mix of residential andcommercial consumers) in the MV network is presented inFig 5

Table 1 Test case scenarios for generation

Testnetwork

Distributed generationmicrogeneration

Installed capacityMW

Percentage peakload

MV 36 60

LV 01 100

Figure 5 MV test network total load

Figure 6 Maximum voltage values in microgrid at node 64

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Figure 8 Active power losses in microgrid at node 64

Figure 9 Total microgeneration in microgrid at node 64

Figure 7 Active power exported to the MV network by the microgrid at node 64

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 445 101049iet-rpg20080064 amp The Institution of Engineering and Technology 2009

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

The ANN to be employed has the following inputs

dagger Voltage at the MVLV substation

dagger Active power generated by each microgenerator unit

dagger Total load of the microgrid

The two outputs are the active power losses and themaximum voltage inside the microgrid since one of themost critical problems to be addressed is related withovervoltage problems that may occur

To achieve good performance different ANN topologies(different number of layers and number of neurons in eachhidden layer) were tested The ANN with the bestperformance has eight inputs one hidden layer with 24neurons and two outputs

To generate the data set corresponding to the inputschosen to train the ANN a large number of power flowswere computed considering different combinations of theinputs (ie several values for the voltage reference at theMVLV substation for the power generated by eachmicrosource and for the load) in order to calculate theactive power losses and assess the voltage profiles in eachscenario

The generated data set contains 11 664 operating pointswhich were split into a training set and a test setcontaining 23 and 13 of the total operating points

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

respectively The training set was used for training theANN and the test set for performance evaluation purposesThe MATLAB Neural Network Toolbox using theLevenbergndashMarquart back-propagation algorithm was usedto identify the ANN parameters

The performance obtained can be evaluated in terms of themean absolute error (MAE) The MAE obtained for one ofthe outputs (active power losses) was 101 1023 Despitethe simple structure adopted for the ANN theperformance parameters illustrated the quality of the toolfor emulating the behaviour of the microgrid

Nevertheless in case of microgrid reconfiguration (due toadding of a new microgenerator for instance) a new ANNmust be computed This means that it is necessary toupdate the data for the power flow and then generate a newdata set and train a new ANN that emulates the LVmicrogrid steady-state behaviour

4 Test case networks andscenariosTo test the voltage control approach developed two realdistribution networks were used an MV distributionnetwork and an LV distribution network shown in Figs 3and 4 respectively

The MV network used is a rural network with two distinctareas with different voltage levels 30 kV at node NO119 and

Figure 3 MV test network

443

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

15 kV after the 30 kV15 kV transformer at node NO206 Itcomprises six microgrids (marked as mG in Fig 3) all with asimilar topology and 3 DG units (marked as DG in Fig 3)based on Combined Heat and Power ndash CHP (NO3 inFig 3) and wind (NO16 and N061 in Fig 3)

The LV network (microgrid) used is also a rural networkwith a radial structure It was considered that all microgridshave the same structure (shown in Fig 4) As previouslystated this LV network was used to train the ANN thatwas included in the optimisation algorithm

Data on the test networks used in this work are presentedin Tables 2ndash5

Each microgrid is supposed to comprise sixmicrogenerators all photovoltaic units (marked as PV inFig 4) The ANN is however capable of dealing withdifferent levels of microgeneration penetration as well aswith different locations for the microsources

Figure 4 LV test network

4The Institution of Engineering and Technology 2009

In this approach from the MV point of view each LVmicrogrid was considered as a single bus with an equivalentgenerator (corresponding to the sum of all micro-sourcegenerations) and equivalent load (corresponding to the sumof all LV loads)

Typical generation curves for each generating technology(CHP Wind and PV) were used as well as typical loadcurves for a 24 h period

The total generation installed capacity for the scenariosused is presented in Table 1

The 24 h profile of the total load (a mix of residential andcommercial consumers) in the MV network is presented inFig 5

Table 1 Test case scenarios for generation

Testnetwork

Distributed generationmicrogeneration

Installed capacityMW

Percentage peakload

MV 36 60

LV 01 100

Figure 5 MV test network total load

Figure 6 Maximum voltage values in microgrid at node 64

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Figure 8 Active power losses in microgrid at node 64

Figure 9 Total microgeneration in microgrid at node 64

Figure 7 Active power exported to the MV network by the microgrid at node 64

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 445 101049iet-rpg20080064 amp The Institution of Engineering and Technology 2009

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

44

amp

wwwietdlorg

15 kV after the 30 kV15 kV transformer at node NO206 Itcomprises six microgrids (marked as mG in Fig 3) all with asimilar topology and 3 DG units (marked as DG in Fig 3)based on Combined Heat and Power ndash CHP (NO3 inFig 3) and wind (NO16 and N061 in Fig 3)

The LV network (microgrid) used is also a rural networkwith a radial structure It was considered that all microgridshave the same structure (shown in Fig 4) As previouslystated this LV network was used to train the ANN thatwas included in the optimisation algorithm

Data on the test networks used in this work are presentedin Tables 2ndash5

Each microgrid is supposed to comprise sixmicrogenerators all photovoltaic units (marked as PV inFig 4) The ANN is however capable of dealing withdifferent levels of microgeneration penetration as well aswith different locations for the microsources

Figure 4 LV test network

4The Institution of Engineering and Technology 2009

In this approach from the MV point of view each LVmicrogrid was considered as a single bus with an equivalentgenerator (corresponding to the sum of all micro-sourcegenerations) and equivalent load (corresponding to the sumof all LV loads)

Typical generation curves for each generating technology(CHP Wind and PV) were used as well as typical loadcurves for a 24 h period

The total generation installed capacity for the scenariosused is presented in Table 1

The 24 h profile of the total load (a mix of residential andcommercial consumers) in the MV network is presented inFig 5

Table 1 Test case scenarios for generation

Testnetwork

Distributed generationmicrogeneration

Installed capacityMW

Percentage peakload

MV 36 60

LV 01 100

Figure 5 MV test network total load

Figure 6 Maximum voltage values in microgrid at node 64

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Figure 8 Active power losses in microgrid at node 64

Figure 9 Total microgeneration in microgrid at node 64

Figure 7 Active power exported to the MV network by the microgrid at node 64

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 445 101049iet-rpg20080064 amp The Institution of Engineering and Technology 2009

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Figure 8 Active power losses in microgrid at node 64

Figure 9 Total microgeneration in microgrid at node 64

Figure 7 Active power exported to the MV network by the microgrid at node 64

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 445 101049iet-rpg20080064 amp The Institution of Engineering and Technology 2009

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

446

amp T

wwwietdlorg

Figure 10 Maximum voltage values in the MV network

Figure 11 Active power losses in the MV network

Figure 12 Total reactive power provided by DG and microgeneration

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454he Institution of Engineering and Technology 2009 doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 2 LV test network line data

From bus To bus Length m R ohmkm L mHkm

NO1 NO2 16 0164 022

NO2 NO3 23 032 023

NO3 NO4 25 0443 025

NO4 NO5 3 308 044

NO4 NO6 24 0443 025

NO6 NO7 22 12 027

NO7 NO8 5 308 044

NO7 NO9 25 126 032

NO6 NO10 22 0443 025

NO10 NO11 15 308 044

NO10 NO12 12 308 044

NO10 NO13 13 308 044

NO10 NO14 29 0443 025

NO14 NO15 9 308 032

NO14 NO16 27 12 027

NO16 NO17 3 308 044

NO14 NO18 2 308 044

NO18 NO19 3 308 044

NO19 NO20 4 308 044

NO14 NO21 19 0443 025

NO21 NO22 17 0443 025

NO22 NO23 12 0868 024

NO22 NO24 34 0443 025

NO24 NO25 26 0868 024

NO24 NO26 6 308 044

NO24 NO27 15 308 044

NO24 NO28 24 12 027

NO27 NO29 13 308 044

NO22 NO30 3 0443 025

NO30 NO31 2 308 044

NO30 NO32 12 0443 025

NO32 NO33 24 0443 025

NO33 NO34 22 308 032

NO34 NO35 54 308 032

NO35 NO36 61 308 032

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

5 Results and discussionThis control algorithm was implemented in the MATLABenvironment using MATPOWER [17] to calculate thepower flows and MATLAB Neural Network Toolbox todesign the ANN

The objective function for the optimisation problem aimedat reducing active power losses while maintaining voltageprofiles within admissible limits (+5)

For the optimisation algorithm a maximum of 100iterations were used for each hour of the day for a totaltime horizon of one day Typical 24 h curves for eachgeneration technology and for the load (considered similarin all load nodes) have been developed for that purpose

The control variables used are presented in (8)

PmG1j jPmG6jQmG1j jQmG6jQDG1j jQDG3jttap (8)

where PmGi is the active power generated by microgrid iQmGi is the reactive power generated by microgrid i QDGi

is the reactive power generated by DG unit i and tltap is thetap value in the transformer at node 206

The main results obtained are presented in the followingfigures

Table 2 Continued

From bus To bus Length m R ohmkm L mHkm

NO36 NO37 7 308 032

NO32 NO38 30 0443 025

NO38 NO39 12 12 027

NO38 NO40 23 0443 025

NO40 NO41 33 0443 025

NO41 NO42 51 12 027

NO41 NO43 126 12 027

NO41 NO44 90 0443 025

NO44 NO45 15 308 032

NO44 NO46 33 0443 025

NO46 NO47 54 0443 025

NO47 NO48 33 308 032

NO47 NO49 40 0443 025

NO49 NO50 30 0443 025

NO50 NO51 30 12 027

NO50 NO52 188 0641 026

447

amp The Institution of Engineering and Technology 2009

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

44

amp

wwwietdlorg

Table 3 LV test network bus data

Bus no Load kVA

NO3 69

NO5 3105

NO6 69

NO7 1035

NO8 69

NO9 1035

NO10 345

NO11 276

NO12 69

NO13 69

NO15 69

NO16 1035

NO17 69

NO18 345

NO19 345

NO20 69

NO21 345

NO23 138

NO25 207

NO26 138

NO27 69

NO28 138

NO29 2415

NO30 1035

NO31 138

NO32 345

NO33 1725

NO34 1035

NO35 207

NO36 345

NO37 69

NO39 207

NO40 1035

NO41 1035

NO42 207

Continued

8The Institution of Engineering and Technology 2009

Fig 6 compares the base situation without the voltagecontrol functionality (Initial) and the result obtained afterthe application of the control algorithm (Final) in the LVmicrogrid It can be seen that the voltage values were outof the admissible range of +5 due to the massivepenetration of PV-based microgeneration that generatedpower outside peak demand hours

Fig 7 shows that the microgrid exports power to theupstream network between 10 and 16 h (sunny hours) andimports in the remaining hours including during peak loadat 21 h Nevertheless the voltage control functionalitysucceeded in bringing the voltages back into the admissiblerange either during daytime when voltages were too highor during night-time when voltages were low

In addition active power losses (Fig 8) in the microgridare reduced since some microgeneration shedding (Fig 9)was required during the sunniest hours in order to bringvoltage profiles back within admissible limits The highlosses in the base case (Initial in Fig 8) are due to the factthat microgeneration penetration was extremely highregarding the local load level and the location of themicrogeneration was not ideal

Concerning the MV network it may be observed inFig 10 that voltage values were inside admissible valuesdespite the fact that the control algorithm had to raisevoltages during night-time in order to be able to also raisevoltage within the microgrids (Fig 6)

Fig 11 shows that the MV network losses have increasedslightly which is natural given that the main concern in thisnetwork was poor voltage profiles

Fig 12 presents the evolution of reactive power generated(positive values) or absorbed (negative values) by the DGsources and the microgrids in order to aid voltage controlThe base case (Initial) is not shown since it considered aunity power factor for all generating sources

Table 3 Continued

NO43 415

NO45 1035

NO46 1035

NO47 1035

NO48 69

NO49 1035

NO50 1035

NO51 1035

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdo

wwwietdlorg

Table 4 MV test network line data

From bus To bus Length m R ohmkm L mHkm

NO206 NO202 68 0324 010

NO200 NO206 666 0521 038

NO206 NO201 71 0324 010

NO206 NO203 50 0324 010

NO118 NO119 842 0521 038

NO176 NO174 490 0324 010

NO175 NO176 90 0324 010

NO176 NO182 262 0627 011

NO131 NO133 27 0731 039

NO132 NO122 756 0731 039

NO29 NO30 323 0731 039

NO62 NO58 604 0731 039

NO79 NO77 46 0731 039

NO57 NO56 481 1181 040

NO189 NO192 100 1608 041

NO184 NO182 236 0324 010

NO44 NO45 661 1181 040

NO111 NO104 474 0324 012

NO134 NO135 119 1181 040

NO201 NO190 1349 0731 039

NO190 NO187 183 0463 010

NO162 NO163 105 0731 039

NO194 NO208 1364 1643 041

NO110 NO114 2 0561 040

NO175 NO171 210 0627 011

NO168 NO171 325 1608 041

NO185 NO191 250 0568 011

NO193 NO191 54 0463 010

NO75 NO74 217 0731 039

NO183 NO178 64 1181 040

NO147 NO149 20 0733 041

NO138 NO139 35 0568 011

NO90 NO86 163 0731 039

NO23 NO24 539 1608 041

NO65 NO73 528 1181 040

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454i 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO112 NO105 406 1608 041

NO48 NO51 758 0731 039

NO97 NO80 1066 0731 039

NO44 NO34 634 0731 039

NO88 NO89 13 0731 039

NO88 NO64 1552 0731 039

NO31 NO36 1177 0731 039

NO106 NO92 1058 1181 040

NO78 NO63 1404 1608 041

NO204 NO209 613 0731 039

NO204 NO205 41 0731 039

NO14 NO13 10 1181 040

NO156 NO158 1310 0731 039

NO15 NO16 21 0731 039

NO38 NO37 199 1643 041

NO68 NO66 669 0731 039

NO68 NO61 555 0731 039

NO107 NO109 10 1181 040

NO196 NO199 480 0568 011

NO120 NO121 9 1608 041

NO127 NO130 172 1181 040

NO40 NO42 12 0731 039

NO46 NO49 696 0731 039

NO161 NO180 812 1181 040

NO161 NO160 10 1181 040

NO151 NO150 64 0568 013

NO69 NO60 1227 0731 039

NO166 NO169 38 1181 040

NO55 NO53 1217 0731 039

NO75 NO81 399 0731 039

NO129 NO128 275 1181 040

NO167 NO170 56 1181 040

NO185 NO174 536 0532 011

NO90 NO96 682 0731 039

NO31 NO28 594 0731 039

Continued

449

amp The Institution of Engineering and Technology 2009

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

450

amp

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO100 NO101 11 1181 040

NO100 NO84 1006 0731 039

NO94 NO85 711 0731 039

NO12 NO4 2234 0731 039

NO12 NO2 1561 0731 039

NO10 NO17 638 0731 039

NO162 NO144 1073 1608 041

NO103 NO91 649 1608 041

NO8 NO7 42 1181 040

NO11 NO9 681 0731 039

NO153 NO164 699 0731 039

NO153 NO155 245 1181 040

NO112 NO76 2286 1608 041

NO83 NO82 1833 0731 039

NO148 NO141 343 1181 040

NO140 NO154 821 1181 040

NO198 NO207 1218 0731 039

NO173 NO172 136 1181 040

NO41 NO39 176 1608 041

NO18 NO19 924 0731 039

NO8 NO1 2520 0731 039

NO26 NO25 3 0324 012

NO126 NO123 167 0731 039

NO32 NO33 257 0731 039

NO157 NO167 703 0731 039

NO197 NO195 10 0731 039

NO6 NO5 394 1181 040

NO52 NO54 272 1181 040

NO3 NO6 436 0731 039

NO69 NO71 82 0731 039

NO181 NO184 154 0324 010

NO186 NO185 94 0463 010

NO145 NO143 63 1181 040

NO22 NO20 407 0919 039

NO125 NO124 30 1181 040

Continued

The Institution of Engineering and Technology 2009

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO6 NO15 1188 0411 038

NO10 NO8 515 1181 040

NO11 NO10 302 1181 040

NO18 NO11 1114 1181 040

NO14 NO12 330 0731 039

NO21 NO14 977 0731 039

NO15 NO29 2412 0411 038

NO21 NO18 594 1181 040

NO27 NO21 1114 1181 040

NO27 NO22 1394 0919 039

NO22 NO23 610 1608 041

NO41 NO26 2688 0731 039

NO29 NO32 518 0411 038

NO48 NO31 3460 0731 039

NO32 NO42 952 0411 038

NO38 NO35 138 0731 039

NO43 NO38 320 0731 039

NO43 NO44 735 1181 040

NO50 NO43 1536 0731 039

NO47 NO27 3422 0919 039

NO47 NO41 1460 1608 041

NO55 NO47 2024 0919 039

NO50 NO48 1798 0731 039

NO52 NO50 204 0731 039

NO57 NO52 879 0731 039

NO70 NO55 1589 0919 039

NO78 NO57 1476 0731 039

NO49 NO59 2060 0411 038

NO59 NO83 1449 0411 038

NO59 NO62 454 0731 039

NO62 NO65 550 0731 039

NO65 NO67 379 0731 039

NO67 NO68 175 0731 039

NO75 NO69 847 0731 039

NO79 NO70 605 0919 039

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO70 NO75 466 0731 039

NO99 NO78 1380 0731 039

NO93 NO79 1377 0919 039

NO83 NO87 536 0411 038

NO87 NO90 395 0731 039

NO87 NO131 3561 0411 038

NO93 NO94 1177 0521 038

NO118 NO93 5019 0521 038

NO94 NO97 1100 0521 038

NO97 NO98 60 0521 038

NO98 NO110 872 0919 039

NO98 NO102 2643 0521 038

NO99 NO103 1494 0731 039

NO106 NO99 712 0731 039

NO102 NO100 102 0731 039

NO102 NO107 839 0521 038

NO103 NO108 954 0731 039

NO115 NO106 935 0731 039

NO107 NO117 1521 0521 038

NO108 NO134 1714 1608 041

NO108 NO116 2589 0731 039

NO118 NO111 738 0919 039

NO132 NO112 1891 1608 041

NO125 NO115 1261 0731 039

NO116 NO140 1559 1181 040

NO116 NO120 1698 0731 039

NO117 NO165 5165 0521 038

NO117 NO127 745 1181 040

NO120 NO126 913 0731 039

NO166 NO125 2295 0731 039

NO126 NO129 457 1608 041

NO127 NO152 1511 1181 040

NO129 NO136 268 1608 041

NO131 NO137 669 0411 038

NO137 NO132 517 1608 041

Continued

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

Table 4 Continued

From bus To bus Length m R ohmkm L mHkm

NO134 NO142 494 1608 041

NO136 NO156 1363 1608 041

NO136 NO113 3080 0731 039

NO173 NO137 3436 0411 038

NO146 NO138 634 0731 039

NO140 NO148 405 1181 040

NO142 NO153 1009 0731 039

NO142 NO146 180 1608 041

NO148 NO145 61 1181 040

NO146 NO147 68 1608 041

NO152 NO151 231 1181 040

NO152 NO161 536 1181 040

NO156 NO189 1611 1608 041

NO179 NO159 1117 0731 039

NO159 NO168 364 1608 041

NO159 NO162 161 0731 039

NO165 NO204 2069 0731 039

NO165 NO197 2090 0521 038

NO177 NO166 453 0731 039

NO168 NO167 516 1181 040

NO179 NO173 364 0411 038

NO183 NO177 195 0731 039

NO194 NO179 625 0411 038

NO203 NO183 1872 0731 039

NO187 NO186 13 0532 011

NO189 NO196 548 1181 040

NO188 NO193 482 0870 011

NO202 NO194 1144 0411 038

NO197 NO198 97 0521 038

NO198 NO200 817 0521 038

NO113 NO95 988 0731 039

NO95 NO88 387 0731 039

NO42 NO49 2311 0411 038

NO181 NO188 152 1376 012

NO67 NO72 243 0731 039

451

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

45

amp

wwwietdlorg

Table 5 MV test network bus data

Bus no Load kVA

NO176 630

NO133 100

NO122 50

NO30 100

NO58 100

NO77 50

NO56 250

NO192 100

NO182 125

NO45 25

NO104 0

NO135 50

NO190 630

NO163 100

NO208 30

NO114 1335

NO171 100

NO191 630

NO74 160

NO178 50

NO149 500

NO139 1000

NO86 25

NO24 100

NO73 100

NO105 100

NO51 160

NO80 100

NO34 100

NO89 100

NO64 50

NO36 100

NO92 250

NO63 160

NO209 100

Continued

2The Institution of Engineering and Technology 2009

Table 5 Continued

Bus no Load kVA

NO205 250

NO13 100

NO158 100

NO16 100

NO37 15

NO66 200

NO61 160

NO109 100

NO199 160

NO121 100

NO130 100

NO40 50

NO46 100

NO180 100

NO160 100

NO150 630

NO60 100

NO169 50

NO175 160

NO53 160

NO81 250

NO128 100

NO170 200

NO174 400

NO96 75

NO28 50

NO101 50

NO84 50

NO85 100

NO4 100

NO2 100

NO17 50

NO144 100

NO91 100

NO7 100

Continued

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

IETdoi

wwwietdlorg

6 ConclusionVoltageVAR control in distribution systems integratingmicrogrids can be treated as a hierarchical optimisationproblem that must be analysed in a coordinated waybetween LV and MV levels

Given the characteristics of the LV networks both activeand reactive power control is needed for an efficient voltagecontrol scheme

An optimisation algorithm based on a meta-heuristic wasadopted in order to deal with the voltageVAR control

Table 5 Continued

Bus no Load kVA

NO9 100

NO164 100

NO155 0

NO76 100

NO82 100

NO141 50

NO154 100

NO207 100

NO172 100

NO39 25

NO19 100

NO1 100

NO25 630

NO123 160

NO33 100

NO157 100

NO195 25

NO5 160

NO54 250

NO3 160

NO71 160

NO184 400

NO185 250

NO143 500

NO20 1200

NO124 100

NO72 50

Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454 101049iet-rpg20080064

problem at the MV level The algorithm has proved to beefficient in achieving the main objective function ndash voltageprofile control and active power losses reduction

The ANN used to emulate the behaviour of the LVmicrogrid proved to have good performance enablingreduction of the computational time considerably Thecombination of the meta-heuristic optimisation motortogether with an ANN equivalent representation of themicrogrid allows the use of this approach for real timeunder DMS environments

7 AcknowledgmentsThis work was supported in part by the EuropeanCommission within the framework of the MoreMicroGrids project Contract no 019864 ndash (SES6) Theauthors would like to thank the research team of the MoreMicroGrids project for valuable discussions and feedback

A G Madureira wants to express his gratitude to Fundacaopara a Ciencia e a Tecnologia (FCT) Portugal for supportingthis work under grant SFRHBD294592006

8 References

[1] JOOS G OOI BT MCGILLIS D GALIANA FD MARCEAU R lsquoThepotential of distributed generation to provide ancillaryservicesrsquo Proc IEEE PES Summer Meeting Seattle USAJuly 2000 pp 1762ndash1767

[2] VOVOS PN KIPRAKIS AE WALLACE AR HARRISON GPlsquoCentralized and distributed voltage control impact ondistributed generation penetrationrsquo IEEE Trans PowerSyst 2007 1 (22) pp 476ndash483

[3] KULMALA A REPO S JARVENTAUSTA P lsquoActive voltage levelmanagement of distribution networks with distributedgeneration using on load tap changing transformersrsquoProc IEEE Power Tech Lausanne Switzerland July 2007pp 455ndash460

[4] CALDON C TURRI R PRANDONI V SPELTA S lsquoControl issues inMV distribution systems with large-scale integration ofdistributed generationrsquo Proc Bulk Power SystemDynamics and Control ndash VI Cortina drsquoAmpezzo ItalyAugust 2004 pp 583ndash589

[5] NIKNAM T RANJBAR A SHIRANI A lsquoImpact of distributedgeneration on voltvar control in distribution networksrsquoProc IEEE Power Tech Bologna Italy June 2003 pp 1ndash6

[6] PECAS LOPES JA MENDONCA A FONSECA N SECA L lsquoVoltageand reactive power control provided by DG unitsrsquo ProcCIGRE Symp Power Systems with Dispersed GenerationAthens Greece April 2005 pp 13ndash16

453

amp The Institution of Engineering and Technology 2009

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064

45

amp

wwwietdlorg

[7] JENKINS N ALLAN R CROSSLEY P KIRSCHEN D STRBAC GlsquoEmbedded generationrsquo (IEE Power and Energy Series 31Inst Elect Eng 2000)

[8] MASTERS CL lsquoVoltage rise the big issue whenconnecting embedded generation to long 11 kV overheadlinesrsquo Power Eng J 2002 1 (16) pp 5ndash12

[9] MADUREIRA AG PECAS LOPES JA lsquoVoltage and reactivepower control in MV networks integrating microGridsrsquoProc Int Conf Renewable Energy Power Quality SevilleSpain March 2007 ICREPQ Webpage httpwwwicrepqcomicrepq07386-madureirapdf

[10] PECAS LOPES JA MOREIRA CL MADUREIRA AG lsquoDefiningcontrol strategies for microGrids islanded operationrsquo IEEETrans Power Syst 2006 2 (21) pp 916ndash924

[11] PECAS LOPES JA MOREIRA CL MADUREIRA AG ET AL lsquoControlstrategies for microGrids emergency operationrsquo Proc IntConf Future Power Syst Amsterdam The NetherlandsNovember 2005 pp 1ndash6

4The Institution of Engineering and Technology 2009

[12] MicroGrids Webpage httpmicrogridspowerecentuagrmicroindexphp accessed June 2008

[13] KENNEDY J EBERHART RC lsquoParticle swarm optimizationrsquoIEEE Int Conf Neural Networks Perth AustraliaNovember 1995 pp 1942ndash1948

[14] SCHWEFEL HP lsquoEvolution and optimum seekingrsquo (Wiley1995)

[15] MIRANDA V FONSECA N lsquoEPSO ndash Evolutionary particleswarm optimization a new algorithm with applications inpower systemsrsquo Proc IEEE Trans Distribution ConfExhibition October 2002 pp 6ndash10

[16] MIRANDA V FONSECA N lsquoNew evolutionary particle swarmalgorithm (EPSO) applied to voltageVAR controlrsquo ProcPower Syst Comput Conf Seville Spain June 2002pp 1ndash6

[17] MATPOWER Webpage httpwwwpserccornelledumatpower accessed June 2008

IET Renew Power Gener 2009 Vol 3 Iss 4 pp 439ndash454doi 101049iet-rpg20080064