THEME [ENERGY.2012.7.1.1] Integration of Variable
Distributed Resources in Distribution Networks
Deliverable 4.2
Methodology for optimising QoS mitigation
infrastructure based on differentiated customer
requirements
Lead Beneficiary:
UNIMAN
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
2/83
Table of Contents
TABLE OF CONTENTS ......................................................................................................... 2
LIST OF TABLES .................................................................................................................. 5
LIST OF FIGURES ................................................................................................................ 6
LIST OF ABBREVIATIONS .................................................................................................... 8
AUTHORS ......................................................................................................................... 10
EXECUTIVE SUMMARY ...................................................................................................... 11
1 BACKGROUND ........................................................................................................... 13
1.1 Summary of the task ..................................................................................................... 13
1.2 PQ phenomena and evaluation methodologies ........................................................... 13
1.2.1 Voltage sags ....................................................................................................... 14
1.2.2 Harmonics .......................................................................................................... 15
1.2.3 Voltage unbalance ............................................................................................. 16
1.3 Provision of differentiated PQ levels ............................................................................ 17
1.4 PQ mitigation techniques .............................................................................................. 18
1.4.1 Prevention strategy ........................................................................................... 19
1.4.2 Compensation strategy ..................................................................................... 20
2 DEVELOPMENT OF PQ EVALUATION INDICES .............................................................. 23
2.1 Voltage sag severity indices .......................................................................................... 23
2.2 Unified bus performance index (UBPI) ......................................................................... 24
2.3 PQ gap indices ............................................................................................................... 24
3 SPATIAL AND TEMPORAL PQ STUDY ........................................................................... 27
3.1 Assessment of voltage sags ........................................................................................... 29
3.2 Assessment of harmonics ............................................................................................. 33
3.3 Assessment of voltage unbalance ................................................................................. 34
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
3/83
3.4 Assessment of PQ using UBPI ....................................................................................... 35
3.5 Harmonic study in LV network ...................................................................................... 36
3.5.1 Modelling ........................................................................................................... 36
3.5.2 Harmonic power flow results ............................................................................ 43
3.5.3 Discussion .......................................................................................................... 44
4 MITIGATION EFFECTIVENESS ...................................................................................... 46
4.1 Based on a simple test network .................................................................................... 46
4.2 Based on a generic distribution network ...................................................................... 48
4.3 Harmonic mitigation in LV Network .............................................................................. 52
4.3.1 Passive filters ..................................................................................................... 52
4.3.2 Distributed compensation and emission reduction .......................................... 54
5 PQ MITIGATION PLANNING BASED ON TECHNO ANALYSIS ......................................... 56
5.1 Optimisation methodologies based on techno analysis ............................................... 56
5.1.1 Initialising the locations for mitigation techniques ........................................... 56
5.1.2 Optimisation using greedy algorithm ................................................................ 57
5.2 Simulation results .......................................................................................................... 59
6 PQ MITIGATION PLANNING BASED ON ECONOMIC ANALYSIS ..................................... 63
6.1 Methodologies of financial assessment ........................................................................ 63
6.1.1 Assessment of financial consequence of PQ phenomena ................................ 63
6.1.2 Assessment of financial cost of mitigation approaches .................................... 66
6.2 Optimisation methodologies based on economic analysis........................................... 68
6.3 Simulation results .......................................................................................................... 69
6.4 Economically incentivised solution to mitigate LV network harmonics ....................... 73
7 CONCLUSION ............................................................................................................. 76
REFERENCES ..................................................................................................................... 78
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
4/83
APPENDIX A: LIST OF PUBLICATIONS BASED ON THIS REPORT ........................................... 82
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
5/83
List of tables
Table 1.1: Effectiveness and cost of reducing faults, adopted from [25] ................................. 20
Table 3.1: System fault statistic for components in GDS network ............................................ 28
Table 3.2: Fault clearing time for primary and back-up bus and line protection relays ........... 28
Table 3.3: Harmonic current spectra amplitude ranges of non-linear loads ............................ 28
Table 3.4: Harmonic current spectra amplitude ranges of DG ................................................. 28
Table 3.5: Ranks of worst buses ................................................................................................ 31
Table 3.6: Sensitivity of BPIS to various parameters ................................................................. 32
Table 3.7: Load at each node in CIGRE benchmark model ....................................................... 42
Table 3.8: System losses by category. ....................................................................................... 43
Table 3.9: Voltage THD at each node with connected load. ..................................................... 43
Table 3.10: Voltage distortion at node with highest THD (R18). .............................................. 44
Table 4.1: Harmonic current injection ...................................................................................... 47
Table 4.2: THD (%) obtained with FACTS devices ...................................................................... 48
Table 4.3: Parameters for unbalance load at B2 ....................................................................... 48
Table 4.4: VUF performance by various FACTS devices ............................................................ 48
Table 4.5: THD performance by various FACTS devices ............................................................ 50
Table 4.6: VUF performance by various FACTS devices ............................................................ 50
Table 4.7: Voltage THD mitigation by use of tuned filters. ....................................................... 53
Table 4.8: Effect on system losses of harmonic mitigation equivalent to harmonic spectrum
for 100 W device with line frequency diode bridge converter grid interface in each phase. .. 54
Table 4.9: Reduction of voltage THD of harmonic mitigation equivalent to harmonic spectrum
for 100 W device with line frequency diode bridge converter grid interface in each phase.
Results for phase b are shown. ................................................................................................. 54
Table 6.1: Details of processes for different customers ........................................................... 67
Table 6.2: Financial analysis of voltage sags ............................................................................. 68
Table 6.3: Equipment operating and aging costs at THDv 3.5% ............................................... 68
Table 6.4: Reduction of voltage THD in phase b at node R18 by compensation of 1 A of
harmonic current in each phase for each node. ....................................................................... 74
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
6/83
List of figures
Figure 1.1: General concept of the PQ zones and PQ temporal and spatial thresholds ............. 18
Figure 1.2: Configuration of SVC connected to grid and SVC Model ........................................ 21
Figure 1.3: STATCOM model...................................................................................................... 21
Figure 1.4: DVR model and voltage/current controllers ........................................................... 22
Figure 2.1: Illustration of SSI based on voltage tolerance curve ............................................... 23
Figure 3.1: Single line diagram of 295-bus generic distribution network ................................. 27
Figure 3.2: Flowchart of proposed stochastic approach and calculation of BPIS ...................... 30
Figure 3.3: Heat maps of the test network based on BPIS ........................................................ 31
Figure 3.4: Normalized BPIS for different operating points ....................................................... 32
Figure 3.5: Tornado diagram of eight parameters in sensitivity analysis of BPIS ...................... 33
Figure 3.6: Heat Maps identifying the most affected areas before and after DG connections .. 34
Figure 3.7: Heat maps of the network indicating the areas affected by unbalance .................... 34
Figure 3.8: Heat maps of the annual voltage sag performance based on UBPI ........................ 35
Figure 3.9: UBPIs at all buses for 24 hours ................................................................................ 35
Figure 3.10: Heatmap indicating PQ performance of EVORA based on UBPI ........................... 36
Figure 3.11: Circuit diagram for line frequency diode bridge converter. ................................. 37
Figure 3.12: Circuit diagram for switched-mode ac-dc converter. ........................................... 37
Figure 3.13: Harmonic spectra of line frequency diode bridge converter (left) and switched-
mode ac-dc converter (right). ................................................................................................... 38
Figure 3.14: Norton equivalent circuit of line frequency diode bridge converter. ................... 38
Figure 3.15: Cable construction used in FEM modelling. Beginning with the top left conductor
and moving clockwise, the conductors are phase b, phase a, neutral, phase c. ...................... 39
Figure 3.16: Current density distribution in cable with 50mm2 conductors at a range of
frequencies. (a) 50 Hz, (b) 250 Hz, (c) 750 Hz, (d) 5 kHz. .......................................................... 40
Figure 3.17: Resistance (left) and inductance (right) for phases a and b for the 50mm2 cable
(phase c is similar to a). ............................................................................................................. 41
Figure 3.18: LV CIGRE residential benchmark topology. ........................................................... 41
Figure 3.19: Load profile applied to LV residential network for c.2035. ................................... 42
Figure 4.1: A simple test network ............................................................................................. 46
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
7/83
Figure 4.2: Dynamic response of voltage with FACTS devices .................................................. 47
Figure 4.3: Single line diagram of 295-bus generic distribution network ................................. 49
Figure 4.4: Dynamic response of voltage with FACTS devices .................................................. 50
Figure 4.5: BPI performance of all buses ................................................................................... 51
Figure 4.6: THD performance of all buses ................................................................................. 51
Figure 4.7: VUF performance of all buses ................................................................................. 51
Figure 4.8: Second order damped shunt filter. ......................................................................... 53
Figure 5.1: Flowchart of greedy algorithm ................................................................................ 58
Figure 5.2: UBPIs of all buses and convergence characteristic of the optimisation methodology
................................................................................................................................................... 59
Figure 5.3: Illustration of the performance of the proposed optimisation methodology ........ 59
Figure 5.4: The illustration of zone division .............................................................................. 60
Figure 5.5: Illustration of the performance of the proposed methodology for provisioning
differentiated QoS ..................................................................................................................... 60
Figure 5.6: Heatmaps of annual PQ performance obtained with 6 devices in case 4 with and
without mitigation ..................................................................................................................... 60
Figure 5.7: Heat maps of the concerned operating points with optimal mitigation solution
based on UBPI............................................................................................................................ 61
Figure 5.8: UBPIs obtained without and with mitigation at all buses for 24 hours’ operation 62
Figure 6.1: Layout of VoDCAT tool ............................................................................................ 67
Figure 6.2: Convergence curves of various components against the number of mitigation
techniques applied .................................................................................................................... 71
Figure 6.3: PQ performance of various buses obtained with 10 techniques ............................ 71
Figure 6.4: Heatmaps of UBPIs obtained with the application of 10 techniques when 𝛽 = 1E8
................................................................................................................................................... 72
Figure 6.5: Representative load profile for January in Germany. ............................................. 73
Figure 6.6: Reduction of system losses by harmonic compensation of 1 A in each phase
against the distance from the main bus of the node at which compensation takes place. ..... 74
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
8/83
List of abbreviations QoS Quality of Supply PQ Power Quality WP Work Package FACTS Flexible AC Transmission System MSI Magnitude Severity Index DSI Duration Severity Index MDSI Magnitude-Duration Severity Index THD Total Harmonic Distortion DNO Distribution Network Operator/Distribution System Operator VUF Voltage Unbalance Factor BPIS Bus Performance Index for Sag UBPI Unified Bus Performance Index VSC Voltage Source Converter SVC Static VAR compensator DVR Dynamic Voltage Restorers DSTATCOM Distribution Static Compensator UPS Uuninterruptible Power Supplies SVS Static Var System TCR Thyristor Controlled Reactor MSC Mechanically Switched Capacitor DG Distribution Generation EV Electric Vehicle MC Monte-Carlo GDN Generic Distribution Network IEC International Electrotechnical Commission IEEE Institute of Electrical and Electronics Engineering NEMA National Electrical Manufacturers Association PLC Programmable Logic Controller PV Photovoltaic SSI Sag Severity Index VoDCAT Voltage Disturbance Cost Assessment Tool OP Operating Point TSP Travelling Salesman Problem NPV Net Present Value SGI Sag Gap Index HGI Harmonic Gap Index UGI Unbalance Gap Index PQGI PQ Gap Index UNIMAN The University of Manchester AHP Analytical Hierarchy Process FCL Fault Current Limiters CFL Compact Flourescent Lamp FEM Finite Element Method PWM Pulse-Width Modulation ESR Equivalent Series Resistance
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
9/83
ESL Equivalent Series Inductance
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
10/83
Authors
Authors Organization E-mail
Huilian Liao University of Manchester [email protected]
Sami Abdelrahman University of Manchester [email protected]
Zhixuan Liu University of Manchester [email protected]
Jovica V. Milanović University of Manchester [email protected]
Thomas Wood TU Berlin [email protected]
Kai Strunz TU Berlin [email protected]
Access:
Project Consortium
European Commission x
Public
Status:
Draft version
Submission for approval
Final version x
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
11/83
Executive Summary
This report describes the SuSTAINABLE Functionality 8 (SF 8) “Power quality planning
for flexible distribution systems” as part of Work Package 4 (WP4) “Methodology for
optimising QoS mitigation infrastructure based on differentiated customer requirements”.
This report discusses the methodologies and indices for evaluating the severity of
power quality (PQ) phenomena including voltage sags, harmonics and unbalance in
distribution networks, as these phenomena would most likely result in PQ interruption to
equipment or industrial processes and thus cause massive financial loss to both utilities and
customers in distribution networks. This report also introduces several new PQ gap indices
which evaluate the satisfaction levels of the received PQ performance compared to the
customer specified thresholds, aiming to enable the provision of differentiated PQ levels in
different zones of the network. Based on these indices, temporal and spatial analysis of PQ
performance was performed on a large-scale generic distribution network with the presence
of stochastic and intermittent power electronics interfaced DGs.
In the harmonic investigation of the LV distribution network, a new modelling
methodology for tracking losses and other network impacts is developed. Based on this
method, an incentive scheme is proposed to promote harmonic compensation by distributed
energy resources. This is intended to form the basis for establishing a market in which
harmonic emissions are reduced through economic incentive, without the need for direct
intervention by the network operator.
A range of PQ mitigating solutions was investigated to insure cost-effective
management of PQ in the network. Flexible ac transmission system (FACTS) devices and
network/plant based mitigation techniques were tested as the potential solutions to the PQ
problems at hand. The effectiveness of these mitigation techniques were discussed based on
the aforementioned severity indices and further tested in generic distribution network.
An optimisation methodology was proposed for optimal, cost-effective, PQ mitigation
in the network with the focus on the type of mitigating solutions and the optimal level of the
mitigation. In this methodology, greedy algorithm is applied to search the optimal mitigation
scheme in order to enable the provision of differentiated PQ levels. Taking the new gap
indices as the objective functions of the proposed optimisation methodology, the mitigation
strategy is optimised to meet the zonally specified thresholds. The optimality and robustness
of the obtained solutions were validated through extensive simulations in
DigSILENT/PowerFactory, and heatmaps are applied to present the significant PQ
improvement as a result of the application of the optimal mitigation strategy obtained by the
proposed methodology.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
12/83
This report introduces the economic assessment of PQ mitigation at planning level, by
considering the benefits during the entire life span of the deployed solution. Since the
upfront investment made for a mitigation solution pays back its returns only during the life
span duration, the methodology calculates the net present value of future benefits, as well as
the net present value of future maintenance, which brings the investment cost and its benefit
to a common ground/level of comparison with planning or deployment year as the reference.
In the report, the potential cost of implementation of each mitigation technique over the life
time was identified, and the cost of remaining losses due to inadequate PQ was evaluated in
order to provide cost-effective mitigation scheme and delivery of required PQ to each
customer. Using the proposed optimisation methodology, PQ mitigation is also planned
based on the economic analysis while taking the provision of differentiated PQ levels as the
constraints. This methodology was tested in a large scale generic distribution network, and
both the financial and technical benefits of the optimal mitigation solution obtained by the
proposed methodology are discussed in the report.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
13/83
1 Background
1.1 Summary of the task
Description of the Task 4.2 “Power quality planning for flexible distribution systems” from
the DoW:
“This task will focus on optimising the QoS mitigation infrastructure based on customer
requirements and presence of stochastic and intermittent power electronics interfaced DG in
the network.”
“According to a temporal and spatial mapping of QoS zones in the network, all feasible
mitigation schemes will be considered to ensure cost-effective management of QoS with the
focus on harmonics, unbalance, voltage regulation and voltage sags.”
This task investigates the optimisation of PQ mitigation infrastructure based on
customer requirements, with the presence of stochastic and intermittent power electronics
interfaced DG and power electronics interfaced storage devices in the network. A range of PQ
mitigating solutions for different PQ zones has been investigated to insure cost-effective
management of PQ in the network, and methodologies are developed for optimal, cost-
effective PQ mitigation in the network. In this task, new indices have been proposed to assess
the PQ performance compared to customer specified requirements. PQ phenomena,
including voltage sags, harmonic and unbalance phenomena, have been studied spatially and
temporally based on the newly proposed indices. UNIMAN has developed an optimisation
methodology for optimal PQ mitigation in the network using power electronic devices and
network based mitigation techniques, aiming to ensure the provision of differentiated PQ
levels throughout the network. Heat-maps are used to present improvement of PQ
performance of the network when the optimal mitigation solution obtained by the developed
optimisation methodology.
Close interaction with T3.6 (Provision of differentiated quality of supply) was
maintained while working on this task to insure full consistency of results.
1.2 Considered PQ phenomena and evaluation methodologies
PQ issues have attracted more and more attention from both utilities and customers
due to the substantial financial loss caused by insufficient PQ performance. Furthermore, the
increasing level of renewable power penetration and electric vehicles connection introduces
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
14/83
more negative impacts on PQ supply in power systems. Typical PQ phenomena in distribution
networks include voltage sags, unbalances and harmonics. These three critical phenomena
would most likely result in PQ interruption to equipment and industrial processes, and thus
cause massive financial loss to customers and utilities in distribution networks. Considering
that there are numerous publications, books and reports describing different PQ phenomena
this report will only briefly describe the three phenomena in following sections to ensure
completeness and consistency of this report.
1.2.1 Voltage sags
Voltage sag phenomenon is defined as a decrease in the voltage or current RMS value
between 0.9 and 0.1 p.u. lasting typically between half a cycle (10ms in 50Hz system) and
several seconds. The main cause of voltage sags are the faults in the network, but they can
also occur as a result of connecting large loads, motor starting or transformer energising.
Voltage sags is one of the most critical PQ problems. It causes frequent disruptions to
industrial processes and malfunction of electronic equipment [1], and result in substantial
financial losses to many utilities and industries.
Sag severity assessment has been a focal point for many researchers in the area of PQ
in the past. A number of single-event characteristics have been proposed to assess the
severity of voltage sags in literature. Voltage sag energy was proposed based on retained
voltage and duration by calculating energy during voltage sags [2, 3]. Lost energy is used to
evaluate the severity of voltage sags by calculating the energy that was not delivered by the
system to the load during a voltage sag [4]. Voltage sag severity is defined based on residual
voltage and duration of a voltage sag by comparing their values with a reference curve [5-7].
More recently in [8], two indices Magnitude Severity Index (MSI) and Duration Severity Index
(DSI) were proposed to represent the magnitude and duration severity respectively, and
served as input parameters for different assessment approaches. Magnitude-Duration
Severity Index (MDSI) uses single numerical value, by combing MSI and DSI, to represent the
failure risk of equipment when subjected to voltage sags. The efficiency of MDSI has been
demonstrated in [9]. In [10], a sag severity index was introduced based on calculating the
missing voltage-time area of a sag. Apart from the aforementioned numerical sag indices,
voltage sag indices based on fuzzy logic approaches were also proposed [11-13]. Most
recently, a sag detection and assessment in terms of magnitude and duration was proposed
based on spectral energy of phase voltages instead of rms-based methods [14]. Although
many indices were proposed to assess the severity of voltage sag events, only a few of them
take the sensitivity of equipment to sag events into account. Incorporation of voltage
tolerance curves (especially the step-shaped curves/standards that are widely-used in
computer and semiconductor industry) into sag severity assessment still needs more
research, considering that multiple interconnected equipment is typically exposed to sags
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
15/83
occurring at a bus and more importantly having in mind that the ultimate measure of sag
severity is not electrical equipment but industrial process failure due to a sag.
1.2.2 Harmonics
Harmonics are sinusoidal voltages or currents with frequencies which are integer
multiples of the fundamental frequency in the network [15]. The main causes of harmonics
are: i) Saturable devices (due to: physical characteristics of the iron core): transformers,
rotating machines, non-linear reactors. ii) Arcing devices (due to: physical characteristics of
the electric arc) (They can produce harmonic currents 20% of their rating): furnaces,
welders (also cause transients and phase imbalance), fluorescent lighting (4th harmonic -
about 50% of a modern building’s load). iii) Power electronics (due to: semiconductor device
switching which occurs within a single cycle of the power system fundamental frequency.)
(They can produce harmonic currents 20% of their rating): VSD, DC motor drives, electronic
power supplies, rectifiers, inverters, SVCs, HVDC transmission.
The main consequences of harmonics are [16]: i) Thermal stress - through increasing
copper, iron and dielectric losses. ii) Insulation stress - through the increase of peak voltage,
i.e., voltage crest factor. iii) Load disruption (e.g., due mal-operation of protection systems,
contactors and numerically controlled process relying on number of waveform zero-
crossings). Typical consequence of harmonics though is reduced life time of equipment due to
increased heating. Harmonics are typically characterised using harmonic distortion index,
though other indices are also in use in different applications (e.g., voltage or current crest
factor, telephone interference index, number of zero crossings, etc.).
The general methodology for calculating the system harmonics indices can be divided
into three steps; calculating the different spectra of the voltage and currents over a window
of time, calculating the required indices from the spectra for different sites, and then
calculating the total system indices from the sites indices. Several indices are developed to
describe the harmonics phenomenon; the most common indices are the Total Harmonic
Distortion (THD) for the voltages and currents and can be calculated by (1.1) and (1.2),
THDV = √∑ 𝑉ℎ
2∞ℎ=2
𝑉1⁄ (1.1)
THDI = √∑ 𝐼ℎ
2∞ℎ=2
𝐼1⁄ (1.2)
where h is the harmonic number, Vh and Ih are the harmonics voltages and currents, and V1
and I1 are the fundamental voltage and current. Other harmonics indices are developed for
more specific applications; the Total Demand Distortion (TDD) was developed to describe the
harmonics performance in case of low fundamental current, where the THD could be
misguiding. Also telephone interference factor was developed to describe the harmonics
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
16/83
performance when it affects the audio and communication system (high harmonics orders).
The K-Factor index in the USA (similar to Factor-K in Europe) to describe the de-rating of
transformers under the harmonics presence, and also the zero crossing factor to describe the
impact of harmonics in the equipment that works on the concept of waveform changing from
positive to negative or vice versa (e.g. contactors and electronic clocks).
Harmonic levels are also increasing due to increasing levels of grid connected power
electronic interfaced load [17]. Furthermore, the proportion of each type of these devices and
the design of their power electronic converters is changing. The use of line frequency diode
bridge converter connected devices is expected to increase, in particular as lighting continues
to migrate from resistive load to Compact Flourescent Lamp (CFL) or other power electronic
interfaced load type [17]. Both Electric Vehicle (EV) [18] and PV penetration [19] is expected
to increase. Connection of PV installation requires a switched-mode ac-dc converter.
Although some early EVs use line frequency diode bridge converters, in order to achieve
controllability and vehicle-to-grid operation, switched-mode ac-dc converters will be
required. The differing harmonic profiles of switched-mode ac-dc converters and line
frequency diode bridge converters are examined in section 3.5.1.1.
1.2.3 Voltage unbalance
Voltage unbalance describes the condition when the three phase voltages are of
different magnitudes and/or do not have a phase shift of 120° with respect to each other. It
incurs overheating in both power system equipment, such as transformers and motors, and
end user device, contributing to accelerated thermal ageing and therefore a reduction of
equipment lifetime [20]. With unexpected negative sequence power flowing in the same path
with positive sequence power, the capacities of online equipment are reduced, resulting in a
reduction in efficiency. Generators and induction motors are additionally required to be
derated due to safety consideration. As unbalance yields extra costs on operation,
maintenance or replacement for both Distribution Network Operators (DNO) and end
customers, it is a very timely topic for the future designs of smart grids where the presences
of single phase generation and load in particular will be noticeably increased. It should be
mentioned that in some countries, single phase feeding is adopted for domestic dwellings,
which means that unbalance phenomenon already exists in some grids. In recent years,
unbalance studies attracted more attention due to the increasing complexity of the
distribution network. The large-scale integrations of single-phase load, dual-phase load and
storage amplify the unpredictability of unbalance in the network. Therefore, the initial
network construction plan may not provide balanced working condition for customers with
the temporally and spatially varying distribution of customer loads. The distributed
generation (DG), such as single-phase renewable energy sources (e.g. PVs), may either
mitigate unbalance or even aggravate it. As the power supplied from transmission level is
regulated to be balanced, the unbalance in distribution networks is typically not emitted from
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
17/83
upstream infrastructure.
Voltage unbalance is typically defined using Voltage Unbalance Factor (VUF). The VUF
is defined as the ratio of the negative sequence voltage component to the positive sequence
voltage component” [21], as given below
𝑉𝑈𝐹 =𝑉1
𝑉2× 100 % (1.3)
The alternative approach given by NEMA (National Electrical Manufacturers
Association) defines unbalance as ratio of maximum voltage deviation from the average line
voltage over the average line voltage [22].
1.3 Provision of differentiated PQ levels
In reality, requirements on PQ vary from area to area, depending on the customers’
nature (commercial, industrial or residential) as well as the sensitivity of their processes and
equipment to the PQ phenomena. In this case, it is promising to provide differentiated levels
of PQ of electricity supply to different zones of the network while the PQ threshold of each
zone is set based on customers’ requirement in that zone. This approach requires less
mitigation effort. It improves the efficiency of electricity/energy distribution by only ensuring
the PQ performance as required. Furthermore, it not only reduces the investment cost, but
also helps utilities to price the electricity. The utilities can get additional revenue from
offering a differentiated and guaranteed PQ levels, and ultimately plan mitigation solution
based on customers’ willingness to pay in different areas. This provides a fair way to subsidize
the mitigation activity. Therefore, in PQ mitigation planning, it is necessary to address the
difference of PQ requirements among different zones. Although characterization and
grouping of customers based on quality demands have been comprehensively investigated
for forming PQ zones in the past [23], the implementation of providing differentiated levels of
PQ supply has not been properly addressed in PQ mitigation planning.
To provide differentiated PQ levels, the PQ zones and PQ (temporal and spatial)
thresholds should be defined. PQ zones, denoted as 𝑃𝑄Z,𝑖 (i=1…N), can be obtained by
demarcating a network based on customer business types and their sensitivity to PQ
problems. Figure 1.1 shows an example. The DNO in this example provides three levels of PQ
supply in each of the three zones; high power quality (HPQ), moderate power quality (MPQ)
and low power quality (LPQ). The proposed PQ levels may not necessarily be equal among
zones (LPQA =/≠ LPQB =/≠ LPQC). For the study period (day, month, year etc.) and due to
different uncertainties in different zones the PQ could vary between different levels, for
example Zone A shows higher variation in received PQ, where zones B and C show less
variation (the red solid circle shows the current PQ level, the dotted circles shows the
possibility of the level to occur). Also, different types of loads with different requirements
have variable thresholds to PQ disturbances. These thresholds may also vary temporally
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
18/83
throughout the study period. Therefore by studying the overlap between the PQ performance
variation and load thresholds variation the adequacy of the provided PQ levels can be
determined.
LPQALPQA
MPQAMPQA
HPQAHPQA
LPQCLPQC
HPQCHPQC
LPQBLPQB
MPQBMPQBPQ
Variation
PQVariation
PQVariation
PQVariation
PQVariation
PQVariation
Load Type 1 (LT1)Load Type 1 (LT1) Load Type 2 (LT2)Load Type 2 (LT2)
Load Type 3 (LT3)Load Type 3 (LT3)
HPQBHPQB
MPQCMPQC
Threshold 1Threshold 1 Threshold 1Threshold 1
Threshold 2Threshold 2
Threshold 2Threshold 2
Threshold 2Threshold 2
Threshold 1Threshold 1
ThrVariation
ThrVariationThr
Variation
ThrVariation
ThrVariation
ThrVariation
Zone CZone C
Zone AZone A Zone BZone B
Figure 1.1: General concept of the PQ zones and PQ temporal and spatial thresholds
Additionally different load types (LT) can be considered (three LTs in this example).
For the same study period, the variation of thresholds due to uncertainties in load sensitivity
was recorded. The minimum and maximum PQ thresholds/requirement are represented as
Thresholds 1 and 2 respectively. From the figure several observations can be noted:
- For loads of LT1 in Zone A, as a result of the overlap between the variation of
the received PQ and the threshold variation of LT1, the DNO must plan mitigation
solution (depending on the customers’ willing to pay) to ensure that the received PQ
performance stays at HPQ level and be better than Threshold 2 of LT1.
- For loads of LT2 in Zone B even the currently received PQ level is HPQ, there is
still probability for LT2 to be exposed to PQ problems, individual plant level solution will
be recommended for this type of customers.
- For loads of LT3 in Zone C since there is no overlap between the PQ and the
threshold variations, no PQ problems will arise for LT3 in this zone.
- LT3 will not face PQ problems in both zones B and C, as the received PQ levels
are always larger than the maximum possible threshold, where in Zone A the DNO needs
to take actions to provide LT3 with the required PQ. LT2 will face PQ problems in all the
zones unless plant level actions are taken to increase the immunity of the plant (i.e.,
lowering Threshold 2 below HPQx) before agreeing on PQ provision by the DNO.
1.4 PQ mitigation techniques
To provide sufficient quality of supply, various methodologies have been explored to
mitigate PQ phenomena in literature [24, 25]. PQ mitigation can be handled at different levels
of system hierarchy. Mitigation options and approaches vary at equipment level, process
level, plant level and network level.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
19/83
At equipment and process level, the equipment can be made more insensitive and
immune to quality of power supply, and the process design can incorporate more redundancy
and fail proof approach. In industries, the immunity capability of equipment needs to meet
certain standards. For instance, the performance of most modern electronic equipment
should follow ITI/CBEMA curve. SEMI F47 defines standard requirements that semiconductor
processing equipment needs to tolerate voltage sags connected to their AC power line. SEMI
has one of the most stringent immunity requirements in place comparing to other sensitivity
curves. Equipment and process level immunity enables only limited immunity from financial
loss, and inadequate PQ and immunity outside their tolerance limits has to be dealt on a
higher level.
PQ mitigation can be handled at plant level and network level. Offering PQ as an
additional service by a distribution network operator has multiple benefits. Individual
customer does not have to make huge upfront investments in capital costs for insulating
themselves against PQ problems. DNO gets additional revenue from offering a differentiated
quality and guaranteed levels. The benefits in having network level solution is distributed
through an overall improvement of quality levels in demarcated zones of the network, hence
there exists opportunities for DNO to ensure profits from this business. Solutions at plant
level and network level can guarantee better immunities by two different approaches.
1.4.1 Prevention strategy
The first approach follows the general principle of that prevention is better than cure.
Preventive approaches are more relevant in the case of network wide solutions since they got
a wide scope of application and overall benefit when performed at network level. Most of the
preventive strategies at network level deal with designing, planning, operating, and
maintaining aspects of networks. Power supply regulations also play important role as a
preventive approach by limiting the amount of emissions allowed by individual customers.
Voltage sags observed in distribution networks originate typically from short circuit faults in
the transmission and distribution networks. Therefore, sags can be mitigated by reducing the
faults. The faults caused by falling tree branches can be reduced by proper tree trimming
schedule, covering overhead lines with insulation or replacing overhead lines with cables.
Lightning faults can be reduced by converting overhead lines to underground cables,
installation of shield wire and surge arresters, and insulating lines. Contact faults due to wind
and animals can be reduced by converting to underground cables, insulating lines, and
installing animal guards for animal caused faults. To reduce accidental dig-ins due to
construction work, better communication and data recording system, which requires the
investment in proper data storage of cable locations and making the information available
prior to any construction work, can be adopted. The effectiveness of the aforementioned
mitigating solutions is provided in Table 1.1, in which the figures are compiled from [26-29].
Instead of reducing fault rate, the sags phenomena can also be mitigated by reducing the
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
20/83
fault severity. The severity of faults can be reduced by reducing fault clearing time, i.e., the
response time of circuit breakers, or by placing fault current limiters (FCL) around the
network. For harmonics phenomena, their severity can be mitigated by line reactors [30].
Transformers with different couplings can limit 3th harmonics and multiples. Phase shifting
transformers (quasi 12-pulse methods) can cancel 5th and 7th harmonics on the primary side
of the transformer to the extent that these currents are balanced in each of the secondary
windings of the transformer.
Table 1.1: Effectiveness and cost of reducing faults, adopted from [25]
Techniques Effect on improved feeder Assumed cost/km
(£)
undergrounding Dig-in faults remains 100,000
Shield wire 78% reduction in lightning faults 22,800
Surge arrester 78% reduction in lightning faults 8,150
Animal guard Assume 50% reduction in animal caused faults 200
Tree trimming 20% reduction in tree caused faults for every year earlier than
5 years
200/trim
Insulated line 75% reduction in lightning faults, 100% reduction in contact
faults
10,000
Communication
system
Assume 50% less dig-ins 100
1.4.2 Compensation strategy
The second approach is to provide online real-time compensation detecting PQ events
using custom power devices and harmonic filters. Flexible ac transmission system (FACTS)
devices are getting more and more popular in power systems, due to the fast development of
power-electronic-based devices and the declining prices of purchasing these devices. They
have been reasonably and widely investigated in power systems for various purposes, e.g.
restoring bus voltages locally or globally [31], enhancing transfer capability [32] and
maximising power system loadability [33], etc. With the flexibility of FACTS devices, they have
also been considered as promising solutions for mitigating PQ phenomena. Custom power
devices for voltage sag mitigation works on the basic principle of injecting power to
compensate the lost voltage. These devices take the required power from less impacted
phases or lines, or from energy storage. Most common type of custom power devices used
for voltage sag mitigation are dynamic voltage restorers (DVR), static VAR compensator (SVC),
distribution static compensator (DSTATCOM), uninterruptible power supplies (UPS). Static
Transfer Switch is considered as good option to protect from sags originated at distribution
networks. Passive filters are also selected as the potential solution for harmonic
phenomenon, as they have been, and are still being, proposed to mitigate harmonic pollution
for utilities or industrial installations. A Voltage Source Converter (VSC) based shunt active
filters are most commonly used mitigation devices for harmonic compensation of line
current. Synchronous Static compensators can be used effectively in mitigating voltage
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
21/83
unbalance. FACTS devices including SVC, STATCOM and DVR are investigated here. In the
study, commercially available DIgSILENT/PowerFacotry software is applied to perform all
dynamic simulations, and the FACTS configuration in the sequel is developed to facilitate its
use.
SVC, as a shunt device, provides rapidly controllable reactive shunt compensation for
dynamic voltage control through its utilization of high-speed thyristor switching/controlled
reactive devices. The model of SVC is given in Figure 1.2. It consists of harmonic filter and a
Static Var System (SVS) which comprises Thyristor Controlled Reactor (TCR), Thyristor
Switched Capacitor (TSC), and Mechanically Switched Capacitor (MSC). SVC regulates the
voltage by controlling the reactive power generated into (via TSC) or absorbed from (via TCR)
the power system. The TSC provides a “stepped” response and the TCR provides a “smooth”
or continuously variable susceptance.
Harmonic
filter
Static Var
System (SVS)
PCC
LoadGrid
Transformer
Voltage
measurement
Positive
SequenceDQ
transform
PLL
Voltage
controller
B Primary to
SecondaryFiring Pulse
GeneratorSVS
(a) Configuration (b) Block diagram
Figure 1.2: Configuration of SVC connected to grid and SVC Model
STATCOM, connected in shunt to the AC power system, regulates the voltage by
adjusting the amount of reactive and active power transmitted between the power system
and VSC. The model of STATCOM is illustrated in Figure 1.3, which mainly consists of a power
transformer, a VSC on the secondary side of the transformer and a DC capacitor working as
an energy storage device. The VSC provides a multifunctional topology which can be used for
various purposes, e.g., voltage regulation and compensation of reactive power, correction of
power factor, and elimination of current harmonics [34]. In the study, the first two control
strategies are investigated.
PWM based
VSC
PCC
LoadGrid
Transformer
DC Capacitor
Voltage DC
measurement PLL
PWM
converter
VDC/VAC
controller
Voltage
(PCC)
measurement
Q (PCC)
measurement
(a) Configuration (b) Block diagram
Figure 1.3: STATCOM model
DVR, a device connected in series with the grid, is capable of protecting sensitive
loads against the voltage variations or disturbances via a VSC that injects a dynamically
controlled voltage in series with the supply voltage through transformers for correcting the
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
22/83
load voltage. With proper control design, DVR can be used to mitigate key PQ disturbances
like voltage sags [35].
Energy storage
PWM VSC
PCC
LoadGrid
Filter
PLL
Current
Controller
Voltage
controller
AC Voltage
measurement
AC Current
measurement PWM
converter
(a) Configuration (b) Block diagram
1
1+sTr
1
sTP
KP (1+ )
1
1+sTr
∑
Vr_ref
yPmax
yPmin
yP
1
1+sTr
1
sTQKQ (1+ )
1
1+sTr
∑
Vi_ref
yQmax
yQmin
yQ
d-q
Transformation
Ir
Ii
cosref
sinref
1
1+sTm
1
1+sTm
Kr (1+ )∑
Pmd_max
Pmd_min
1
sTr
∑
Pmq_max
Pmq_min
d-q
Transformation
Pmr
Pmi
-
-
-
-
Vr
Vi
Voltage controller
Voltage controller
Current controller
Kr (1+ )1
sTr
(c) DVR voltage/current controller
Figure 1.4: DVR model and voltage/current controllers
The modelling of DVR is given in Figure 1.4 (a) and (b). In the study, a PI-based control
strategy is developed to compensate for voltage sag disturbance using commercially available
software PowerFactory/DIgSILENT, as shown in Figure 1.4 (c). The control structure consists
of PI-based current controller and PI-based feedback voltage controller, together with proper
time-delay function. Pmr and Pmi, the signal coming from the controller, are modulation
indices which will be used by PWM VSC to determine the real and imaginary parts of the
voltage at AC-side respectively based on the following equations:
𝑈𝐴𝐶𝑟 = 𝐾0𝑃𝑚𝑟𝑈𝐷𝐶 (1.4)
𝑈𝐴𝐶𝑖 = 𝐾0𝑃𝑚𝑖𝑈𝐷𝐶 (1.5)
where 𝐾0 is a constant that depends on the modulation method applied in PWM, and 𝑈𝐷𝐶 is
the voltage at DC-side.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
23/83
2 Development of PQ Evaluation Indices
2.1 Voltage sag severity index
Since the severity of voltage sags is strongly related to the response of equipment to
voltage sags, standard voltage tolerance curve SEMI F47, as shown in Figure 2.1 (a), is applied
to assess sag performance in the study. For given sag magnitude and duration SSI=0 if the
corresponding sub-sag is in the non-susceptibility area, i.e., the area above the voltage
tolerance curve; if the sag is in the susceptibility area, SSI is calculated as:
SSIB𝑖𝑗C𝑤 = (∑ 𝑎𝑥 ×
Vmax(T𝑥)−𝑣B𝑖𝑗
Vmax(T𝑥)−Vmin(T𝑥)
𝑦−1𝑥=1 ) + 𝑎𝑦 ×
Vmax(T𝑦)−𝑣B𝑖𝑗
Vmax(T𝑦)−Vmin(T𝑦)× (
𝑡B𝑖𝑗−Tmin(T𝑦)
Tmax(T𝑦)−Tmin(T𝑦))
𝑏𝑦−1
(2.1)
The contour colour map of SSI estimated using (2.1) is presented in Figure 2.1 (b),
where the more affected areas are marked in red, and the areas which are not subject to
potential voltage sag interference are marked in light blue. The detailed discussion on the
advantage of this newly proposed sag severity index in comparison to other existing indices
proposed in the past refers to [36].
0.8
0.7
0.02 0.2 0.5
0.5
1 Duration t (s)Ma
gn
itu
de
v (
p.u
.)
T1 T2 T3
0.1
(a) Sag susceptibility area (b) Contour map of SSI using (1) with a=3 and b=
1
2
Figure 2.1: Illustration of SSI based on voltage tolerance curve
In order to assess comprehensively bus performance with respect to voltage sags,
three key aspects of sag performance (i.e., occurrence frequency, magnitude and duration)
are taken into account and represented by a new single numerical index, namely Bus
Performance Index (BPIS). For the jth sub-sag event at bus i, the mean sag severity index SSIB𝑖𝑗M
is calculated using sag magnitude and duration, taking into account the variation/uncertainty
of voltage tolerance curve. Then single-site index BPIS is defined as:
BPIB𝑖𝑗S = ∑ 𝑓B𝑖𝑗
4𝑀𝑁𝑗=1 × SSIB𝑖𝑗
M (2.2)
The detail procedure for calculation of BPIS is given in [37].
0 0.5 1
0.2
0.4
0.6
0.8
1
Duration t (s)
Magnitude v
(p.u
.)
Colour contour map
Low
High
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
24/83
2.2 Unified bus performance index (UBPI)
To comprehensively assess PQ performance, the critical PQ phenomena including
voltage sags, harmonic and unbalance, which would most likely result in PQ interruption to
equipment or industrial processes and thus cause massive financial loss to utilities and grid
connected consumers in distribution networks, should be considered simultaneously.
Besides, from the perspective of planning, it is efficient to consider these phenomena at the
same time when planning mitigation solutions for PQ issues, as generally one mitigation
device can affect the performance of more than one phenomenon. Therefore, the
aggregation of the performance of all critical phenomena is required. In the study, a newly
proposed unified Bus Performance Index (UBPI), as defined in (2.3), was applied to represent
the overall PQ performance with respect to voltage sags, harmonic and unbalance, using
Analytical Hierarchy Process (AHP) methodology. Analytic Hierarchy Process (AHP) is one of
the common mathematical models for multi criteria decision making problems. It solves the
problem of selecting a goal from a number of alternatives based on a number of selecting
criteria. Different selection criteria will have different weights on the final decision. Also, each
selecting criterion can have a number of sub-criteria, which again can have different weights
in the main selecting criterion. Based on the different weights, each criterion has a different
priority on the final decision. The alternatives have different scores for each selecting criteria,
then based on the criteria relative priorities the final score will be given to the alternatives
and the final decision will be made. Further details and mathematical modelling can be found
in [38]. The methodology adopted here is to evaluate the overall PQ performance of network
bus based on different PQ phenomena taken separately and simultaneously using a single
index. Different from the general PQ indices, critical states which are derived from thresholds
specified in standards or user requirements are incorporated in the AHP
𝑈𝐵𝑃𝐼𝑖=AHP (𝐵𝑃𝐼𝑖 , 𝑇𝐻𝐷𝑖, 𝑉𝑈𝐹𝑖) (2.3)
where THD and VUF are the severity indices for harmonic and unbalance phenomena
respectively.
2.3 PQ gap indices
To provide differentiated levels of PQ supply, PQ zones and the associated PQ
thresholds should be defined based on customers’ requirements. PQ zones can be obtained
by demarcating a network based on customer business types and their sensitivity to PQ
problems [23]. To accurately evaluate the PQ performance from the perspective of utilities
and customers, appropriate indices should be adopted. The severity of voltage sags is
assessed using BPI, which takes into account various sag characteristics simultaneously as
well as sensitivity of equipment to voltage sags, and reflects to a good approximation the
practical consequence of voltage sags from the point of view of system/equipment operation
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
25/83
[37]. Harmonics and unbalance phenomena are evaluated using THD and VUF respectively,
which are widely used in practice [39]. For each PQ zone, the threshold with respect to each
PQ phenomenon is determined based on the sensitivity of customers’ equipment/process to
the specific phenomenon in that zone. Given the PQ zones and specific zonal PQ thresholds,
this report investigates the mitigation strategy to ensure the provision of differentiated levels
of PQ supply in different zones.
This problem is defined as an optimisation problem, which is to minimise the gap
between the received PQ performance and the zonal thresholds. To facilitate the concept of
provision of differentiated levels of PQ, five new indices are proposed here to present the PQ
gap with respect to different forms of customer requirements.
The thresholds with respect to voltage sags, harmonics and unbalance phenomena in
PQ zone i are denoted as 𝐵𝑃𝐼TH,𝑖, 𝑇𝐻𝐷TH,𝑖 and 𝑉𝑈𝐹TH,𝑖 respectively. If the PQ phenomena
are considered individually, three gap indices can be derived. Sag Gap Index (SGI), which
presents the gap between the received voltage sag performance and the imposed zonal sag
requirements, can be defined as:
𝑆𝐺𝐼 = ∑ (∑ |𝐵𝑃𝐼𝑖,𝑗 − 𝐵𝑃𝐼TH,𝑖|𝐵𝑃𝐼𝑖,𝑗>𝐵𝑃𝐼TH,𝑖
𝐵𝑖𝑗=1 )𝑁
𝑖=1 (2.4)
where 𝐵𝑗 denotes the total number of buses within PQ zone i; and 𝐵𝑃𝐼𝑖,𝑗 denotes BPI of the
jth bus in zone i. With the same principle applying to the phenomena of harmonics and
unbalance respectively, Harmonic Gap Index (HGI) and Unbalance Gap Index (UGI) can be
derived:
𝐻𝐺𝐼 = ∑ (∑ |𝑇𝐻𝐷𝑖,𝑗 − THDTH,𝑖|𝐵𝑃𝐼𝑖,𝑗>𝐵𝑃𝐼TH,𝑖
𝐵𝑖𝑗=1 )𝑁
𝑖=1 (2.5)
𝑈𝐺𝐼 = ∑ (∑ |𝑉𝑈𝐹𝑖,𝑗 − 𝑉𝑈𝐹TH,𝑖|𝐵𝑃𝐼𝑖,𝑗>𝐵𝑃𝐼TH,𝑖
𝐵𝑖𝑗=1 )𝑁
𝑖=1 (2.6)
From the perspective of mitigation efficiency, the three PQ phenomena should be
considered simultaneously, as generally one mitigation device can affect more than one PQ
phenomenon. Therefore, the performance of the concerned phenomena should be
aggregated somehow. In this report, AHP based UBPI defined in (2.3) is adopted to represent
the aggregated PQ performance of the three PQ phenomena, i.e., voltage sags, harmonic and
unbalance.
In (2.3), zonal PQ thresholds are not included yet. Given the aggregated PQ
performance and zonal PQ thresholds denoted as UBPITH, the gap between the received UBPI
and zonal PQ thresholds can be defined as:
𝑃𝑄𝐺𝐼UBPI = ∑ (∑ |𝑈𝐵𝑃𝐼𝑖,𝑗 − 𝑈𝐵𝑃𝐼TH,𝑖|UBPI𝑖,𝑗>𝑈𝐵𝑃𝐼TH,𝑖
𝐵𝑖𝑗=1 )𝑁
𝑖=1 (2.7)
In (2.7), the performance of each PQ phenomenon in comparison to its threshold is
not reflected, and the performance of different PQ phenomena can cancel each other. For
instance, for a bus, there exist the poor-performed BPI and well-performed THD and VUF. In
this case, the poor performance of BPI will be compensated by the better performance of
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
26/83
other PQ phenomena in 𝑃𝑄𝐺𝐼UBPI, and the contribution of individual PQ phenomenon on the
aggregated performance is not reflected in the gap index. If the performance of each PQ
phenomenon is expected to meet the threshold that is individually specified for this PQ
phenomenon, the PQ gap can be defined as:
𝑃𝑄𝐺𝐼IND = ∑ (∑ AHP (|𝐵𝑃𝐼𝑖,𝑗 − 𝐵𝑃𝐼TH,𝑖|BPI𝑖,𝑗>𝐵𝑃𝐼TH,𝑖, |𝑇𝐻𝐷𝑖,𝑗 − 𝑇𝐻𝐷TH,𝑖|𝑇𝐻𝐷𝑖,𝑗>𝑇𝐻𝐷TH,𝑖
, |𝑉𝑈𝐹𝑖,𝑗 −𝐵𝑖𝑗=1
𝑁𝑖=1
𝑉𝑈𝐹TH,𝑖|VUF𝑖,𝑗>𝑉𝑈𝐹TH,𝑖)) (2.8)
Between (2.7) and (2.8), the former aggregates the performance of the three PQ
phenomena first and then compares it with the zonal thresholds presented as aggregated PQ
performance, the latter compares the performance of each PQ phenomenon with its
corresponding zonal threshold first, and then aggregates the gaps of these three PQ
phenomena together.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
27/83
3 Spatial and Temporal PQ Variation
In the first part of this section that focuses on laying the fundamentals for subsequent
studies, spatial and temporal variation of voltage sags, harmonics and voltage unbalance is
studied for each phenomenon individually. In the second part of the section the global PQ
performance of the network using UBPI is presented. The last part focuses on assessment of
harmonics in LV networks. Several different test networks are used to illustrate the results of
the study.
A 295-bus generic distribution network (GDN), as shown in Figure 3.1, is used in the
study. It comprises 275 kV transmission in-feeds, 132 kV and 33 kV predominantly meshed
sub-transmission networks, and 11 kV predominantly radial distribution network. The
network consists of 276 lines including overhead lines and underground cables, 37
transformers with various winding connections, 297 loads (including 10 unbalance loads)
representing industrial, commercial and domestic loads, and 12 distributed generators
(including 2 wind turbines, 5 fuel cells and 5 photovoltaic) connected to 11 kV distribution
network. The locations of unbalanced loads and distributed generators are marked by
different labels in Figure 3.1.
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
235 236
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
249 250
266267
242
252
260
289
237
244
261
251
241 245246 243
262272270253 274271 276 275 263 264 273
240
254258 256259
277
247
278
248
280
234
279
233255257
293 292
294 295 297 296
299 298
300
77238
288
269
287 285286
56
BA C D E
F
H I J K L
ON
G
132kV
33kV
11kV
3.3kV
275kV
400kV
G
H2 O2
PV
H2 O2
H2 O2
H2 O2
PV
PV
PV
PV
H2 O2
195
196
H2 O2
PV
Wind Turbine
Photovoltaic
Fuel Cell
Unbalance Load
Zone-1 Zone-2 Zone-3
Figure 3.1: Single line diagram of 295-bus generic distribution network
The components at different voltage levels have different fault rates, and the
detailed system fault statistics in the distribution network are given in Table 3.1[37, 40]. The
mean and standard deviation of the distribution of fault clearing time is given in Table 3.2,
together with the failure probability of primary protection relays. To model the unbalanced
operation of a network, a number of loads are selected as potential sources of unbalance in
the network. For these unbalance loads, real power demand at each phase is set according to
the true load profile, while the reactive power is set based on power factors which are
generated randomly based on a preset normal distribution. In the study, 10 unbalance loads
are applied. The mean of the normally distributed power factors is set to 0.95 representing a
general load [41], and their standard deviation is set to 0.053. Ten loads are selected as fixed
non-linear loads which inject harmonic current into the grid. Besides that, 20 more loads are
randomly selected from the rest unselected loads to inject harmonic current. The ratio of the
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
28/83
magnitude of the injected harmonic current to that of the fundamental component follows
pre-set normal distributions. The means of the normal distribution used for different types of
non-linear loads and various DGs are listed in Table 3.3 and Table 3.4 respectively [42]. The
standard deviation of the aforementioned normal distributions is set to 10% of the mean.
Table 3.1: System fault statistic for components in GDS network
Components Buses Lines Lin
es
Line
s
Cabl
es
Fault rate (Number of
events/year*100km)
0.08 8.6 3.7 0.6 4.9
Type of faults SLGF LLG
F
LLF LLL
F
Percentages 73% 17% 6% 4%
Table 3.2: Fault clearing time for primary and back-up bus and line protection relays
Components Relays Mean (ms) Std (ms) Failure probability
(%)
Buses Primary 60 3 1.09%
Back-up 800 5 N/A
Lines Primary 300 13 2.22%
Back-up 800 40 N/A
Table 3.3: Harmonic current spectra amplitude ranges of non-linear loads
Harmonic
order
Type 1
(Domestic and
Commercial)
Type 2
(Industrial)
1 100% 100%
3 69% 4.7%
5 48% 32%
7 28% 16%
9 27% 0%
11 0% 6.5%
Table 3.4: Harmonic current spectra amplitude ranges of DG
Harmonic
order Wind Gen. PV
Fuel
Cells
1 100% 100% 100%
5 1.9% 0.1% 0.05%
7 0.4% 0.1% 0.1%
11 0.1% 0.2% 0.15%
13 0.1% 0.3% 0.2%
To reflect the PQ performance accurately, the variation of load profiles and network
parameters are taken into account. Annual hourly loading curves were extracted from 2010
survey of different types of loads (including commercial, industrial and residential loads), and
8760 operating points are obtained. The wind and photovoltaic generators have annual
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
29/83
hourly output curves which are extracted from the realistic outputs data based on the UK
weather [43, 44]. In each of these operating points different loading levels of loads and
outputs of DG units are considered. The maximum outputs of the wind and photovoltaic
generators are at different times of the year, and the fuel cells are assumed to have a
constant output. Since there exist similar patterns of load demand variation among loads of
the same types (e.g., industrial load, commercial load and domestic load) and similar
variation trends of the outputs of certain DGs (i.e., PV) in terms of day and season, similar
operating condition re-occurs throughout the whole year. In the study, Cluster Evaluation of
Statistics Toolbox in Matlab is used to find the representative operating condition. The
industrial load, commercial load, domestic load and PV output are taken as the input to the
classification approach of K-means. The classification is performed based on the clustering
evaluation criterions of Calinski-Harabasz, which defines the ratio between the overall
between-cluster variance and the overall within-cluster variance, and the obtained clusters
are then validated using the method of Silhouette, which is used to measure how
appropriately the data has been clustered. Using this approach, 9 representative operating
points are obtained. Additionally, further 7 operating points corresponding to the maximum
load, the maximum DG output, the maximum wind output, the maximum PV output, the
maximum industrial load, the maximum commercial load, and the maximum domestic load
are also accounted for in the simulation. So, in total there are 16 characteristic operating
points taken into account. The average of the 16 indices evaluated from the 16 operating
points respectively is taken as the objective function for optimisation.
3.1 Assessment of voltage sags
A comprehensive stochastic approach is developed to assess the performance of all
buses in distribution networks with respect to sags. It includes probabilistic nature of faults in
the network, variation of network operation condition, different reliability performance and
settings of protection system and variation in voltage tolerance curves of equipment. The
flowchart illustrating the proposed approach is given in Figure 3.2, where NFL denotes the
number of fault locations, IOP and IFC denote the indices of operating point and fault case,
respectively. This procedure can also be substantially simplified and easily applied in cases
when voltage sag monitoring results are directly available.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
30/83
Complete all 10×NFL fault cases ?
Calculate 50 SSI for each sag based on probabilistic model of voltage
tolerance curves; Calculate SSIM, the mean of 50 SSI
end
Begin
Input power network, load duration curve, NFL fault locations and their fault
rates; Select 11 operating points
Calculate BPIS for all buses
For each bus, record voltage sag occurring on this bus, occurrence
frequency of the corresponding fault, type of component that caused the sag
YesNo
Yes
Select IOP=1
Obtained final index by summing up 11 BPI
S weighted by their percentages
of representation in yearly load duration
Calculate frequencies of 10 faults for all fault locations
Run short circuit simulation
Update sag frequencies for all sags in record using failure probabilities
Derive sag duration according to distribution of clearing time of
corresponding protection relay and division of duration zones
Select IFC=1
Complete 11 operating points ?
IFC=IFC+1
IOP=IOP+1No
Figure 3.2: Flowchart of proposed stochastic approach and calculation of BPIS
The proposed stochastic approach uses the deterministic results of the short circuit
simulation and the stochastic data about the faults. This is similar to existing prediction
methods, e.g., fault position methods and the stochastic method presented in [45]. Similar to
Monte-Carlo (MC) simulation methods, the proposed approach includes numerous
uncertainty factors in the network but avoids a large number of simulations required by MC
to achieve convergence. This is one of its advantages compared to conventional MC
simulations. Furthermore, different from existing stochastic methods, the proposed method
derives the sag frequency and sag duration based on the stochastic distribution of clearing
time by protection system. The total required computation time of 31min consists of 30min
and 8s spent on performing fault simulations in DIgSILENT/PowerFactory and 52s spent to
calculate BPIS for all buses in the network. The computation time however, is provided here
for reference only as it depends on the computer processing speed and programming
efficiency.
To observe the sag performance of various buses visually, i.e., to identify areas of the
network which are most vulnerable to voltage sags, a heat map is used. The final mean BPIS is
calculated for each bus of the 11 kV section of test distribution network. The calculated mean
BPIS are used here to generate heat maps of the network, as shown in Figure 3.3 (a), though
they can be equally used to rank buses in the network with respect to exposure to voltage
sags. With the help of heat maps, it is much easier, compared to simple numerical bus ranking,
to identify the weak areas of the network with respect to exposure to voltage sags. The bus
performance is poor in the area marked in red. The heat map using BPIS obtained for
operating point 1 only is presented in Figure 3.3 (b). It can be seen that there are very small
differences between the two heat maps shown in Figure 3.3, though they have been
produced using different number of operating conditions of the network.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
31/83
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
Heat Map of BPIS
Worst Buses: 196, 194, 193 (a) Derived from 11 operating points
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
Heat Map of BPIS
Worst Buses: 196, 195, 193
(b) Derived from operating point 1
Figure 3.3: Heat maps of the test network based on BPIS
To investigate the influence of operating conditions considered in deriving BPIS , and
consequently heat maps of the network, buses are ranked according to BPIS derived from OP1
(operating point 1), OP6 and OP11, which represent heavy, medium and low load demand
respectively. The nine buses which are most affected by the voltage sags for each operating
condition (based on calculated BPIS) are listed in Table 3.5. The bottom row of the table
“Overal” lists the buses ranked based on the final mean BPIS determined from 11 operating
conditions. It can be seen that the bus ranking obtained based on OP6 (medium loading of
the network) is the same as the final ranking. As for OP1 and OP11, although their ranks are
different to the last row, the difference is small. This suggests that the operating condition
have minor influence on the final ranking of network buses.
To illustrate the effect of operating condition on BPIS, the worst performing bus (Bus
196) and the best performing bus (Bus 61) are selected for further analysis, as well as Bus 1.
The BPIS derived from 11 operating points respectively are given in Figure 3.4. It can be seen
that the variation of BPIS for all three buses obtained for operating points 2-11 is quite small,
while the BPIS obtained for OP1 (extreme loading of the network accounting for about 2% of
the time of the year) is noticeably different from the others (though very fine scale is used).
Table 3.5: Ranks of worst buses
Ranking 1 2 3 4 5 6 7 8 9
OP1 196 195 193 194 210 200 197 198 199
OP6 196 194 193 195 210 200 197 199 198
OP11 196 194 193 195 210 200 197 198 199
Overal 196 194 193 195 210 200 197 199 198
To be specific, at Bus 1, the BPIS obtained for OP1 is different from the mean of BPIS
obtained from other OPs by 1.14%; at Bus 61, the BPIS obtained for OP1 is different from the
mean of others by 0.82%; while at Bus 196, the difference is only 0.46%. It can be seen that
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
32/83
the influence of different OPs on the variation of BPIS is very small.
(a) For Bus 1 (b) For Bus 61 (c) For Bus 196
Figure 3.4: Normalized BPIS
for different operating points
The sensitivity of BPIS, and consequently the area of vulnerability of the network, to
various parameters, including load demand, number of considered voltage tolerance curves,
number of considered duration zones, fault rates of various components and distribution of
the clearing time required by various protection relays, is analysed here based on Bus 1. Each
parameter is set in turn to a number of different values with all other parameters fixed, and a
corresponding set of BPIS values calculated. The standard deviation of calculated BPIS values is
used to present the sensitivity of BPIS to the variation of that parameter. The setting range of
various parameters and the derived standard deviation of BPIS are given in Table 3.6. It
should be noticed that the range of parameter settings would impact the standard deviation
of BPIS.
Table 3.6: Sensitivity of BPIS to various parameters
No Parameters Settings Std
1 Fault rate of buses [0.0720, 0.0880] 0.1308
2 Fault rate of lines [6.1200, 7.4800] 27.9004
3 Fault rate of primary protection relays on buses
[0.0107, 0.0131] 0.0979
4 Fault rate of primary protection
relays on lines
[0.0200, 0.0244] 1.6154
5 Mean of clearing time of primary protection relays on buses
[54, 66] 0.1068
6 Mean of clearing time of backup
protection relays on buses
[720, 880] 0.1029
7 Mean of clearing time of primary
protection relays on lines
[270, 330] 28.7946
8 Mean of clearing time of backup
protection relays on lines
[720, 880] 0.9508
9 Number of voltage tolerance
curves
[ 50, 500] 0.0527
10 Number of divided duration zones [ 15, 49] 0.0110
11 Load demand N/A 1.3776
12 Selection of operating points N/A 0.6471
It can be seen from the last column of Table 3.6 that BPIS is by far the most affected
by the fault rate of lines and the mean value of clearing time required by primary line
protection relays. These two parameters impact the frequency and duration of sags,
respectively. Fault rates o have larger influence on f lines and the clearning time of the
associated primary protection relays BPIS compared to the fault rate of buses. It can be also
0 5 100.138
0.14
0.142
BP
I
0 5 100.0495
0.05
0.0505
BP
I
0 5 100.945
0.95
0.955
BP
I
Index of OP Index of OP Index of OP
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
33/83
seen that BPIS is not very sensitive to variation in other parameters for the settings given in
Table 3.6.
Figure 3.5: Tornado diagram of eight parameters in sensitivity analysis of BPI
S
The sensitivity analysis of the first eight parameters, which are related to the network
components, is represented using a Tornado diagram, which depicts the most sensitive
precedent parameter along with the impact on the overall result [46]. Tornado analysis
determines the effect on the overall result by changing one variable at a time. The Tornado
diagram is shown in Figure 3.5, and it indicates the range (both minimum and maximum) of
the obtained BPIS when changing the corresponding parameter setting. The vertical solid line
marks the base value of BPIS, i.e., the BPIS obtained using the default parameter settings. It
can be seen that BPIS is mostly affected by parameters 2 and 7, i.e., the fault rate of lines and
the mean value of clearing time required by primary line protection relays, which is in line
with the standard deviation analysis above.
3.2 Assessment of harmonics
To study harmonic phenomenon, a probabilistic methodology is developed to model
and study harmonic propagation through the power network over specified time period by
taking into account uncertainty factors including the harmonic injections from diverse, but
fixed location sources (renewable generation, i.e., wind, PV and other converter connected
generation) and variable location sources, e.g., non-linear loads, electric vehicles (EV), etc.
This was assessed using the harmonic injection levels of DG and loads as defined in tables 3.3
and 3.4. The assessment was performed for the MV level only (11 kV); however some of the
loads connected to higher voltage level (132 kV) were also considered as distorting loads.
Grid supply points are assumed to be purely sinusoidal and no background distortion is
considered. Heat maps of annual harmonic performance obtained with and without DG are
presented in Figure 3.6 (a) and (b) respectively, where the most affected areas can be
identified easily (Please note the different levels of color bar codes due to different THD
maximum values in each case). It can be seen that before the connection of DGs, all 11 kV
buses have very similar performance, with the average THD ranged between 0.5% and 1%.
Following the connection of DGs the most affected area remains the same. The THD in this
case, however, is higher and ranges from 3.5% to 5.7%, while the maximum average THD for
buses connected to other substations was 0.7%.
419.9539
377.9807
420.0558
417.8003
420.0602
419.9987
373.1393
418.6426
420.3532
462.1291
420.3498
422.6811
420.3536
420.3418
460.1282
421.5197
340 390 440
1
3
5
7
BPIS
Par
amet
er N
o
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
34/83
(a) Harmonic performance without DGs (max THD = 1%) (b) The most affected phase following DG connections (max THD = 5.7%)
Figure 3.6: Heat Maps identifying the most affected areas before and after DG connections
3.3 Assessment of voltage unbalance
For unbalance study, three-phase full Newton-Raphson load flow method and Monte
Carlo simulations are used to probabilistically estimate the voltage unbalance level at all
buses in the network under unbalanced loading conditions. In the study, variation of
operation condition and uncertainty of unbalance levels are considered. To model the
unbalanced loads in the load flow loops, the power factors of three phases of each load vary
assuming constant real power demand of each phase. Three base values of power factor, 1.0,
0.95 and 0.8, corresponding to lighting load, general load and induction motor load
respectively, are selected to model load asymmetry. Furthermore, in order to create large
amount of unbalanced loading scenarios probabilistically, a normal distribution is employed
to estimate the possible variation of power factor in each phase. By taking the 1, 0.95 and 0.8
as the mean values for the normal distributions, random values are created within
corresponding ±20% (±3σ) ranges. The unbalance performance was studied on a 24-bus test
network, which is part of real UK distribution network. A day is divided into 8 time zones, and
the unbalance performance at time zones 2 and 7 is presented in Figure 3.7, which presents
the temporal variation of unbalance performance. With this study, the areas with high VUF
can be easily identified and locations where mitigation measures should be applied can be
pointed out to network operators.
(a) Time zone 2 (b) Time zone 7
Figure 3.7: Heat maps of the network indicating the areas affected by unbalance
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
35/83
3.4 Global assessment of PQ using UBPI
In this sub-section the three aforementioned phenomena are studied simultaneously
based on UBPI. The UBPI performance of the operating point which is corresponding to the
maximum cluster is presented in Figure 3.8 (a), and that of the operating point which is
corresponding to the maximum load during the whole year is presented in Figure 3.8 (b). It
can be seen that with the increased load demands in the network, the PQ performance is
degraded obviously.
(a) maximum cluster
(b) maximum load
Figure 3.8: Heat maps of the annual voltage sag performance based on UBPI
Time series simulations are adopted to present the temporal performance of UBPI at
all buses. 24 hours simulations for a selected day were performed, the UBPI was calculated
for every bus under evaluation at every hour of the day. The results are shown in Figure 3.9,
in which the thresholds set based on customers’ requirement is marked in dash bold line. It
can be seen most buses violate the requirements of the zone for most of the time.
Figure 3.9: UBPIs at all buses for 24 hours
0 50 100 150 2000
0.2
0.4
0.6
0.8
1
Bus index
UB
PI
0 1002000
0.5
1
Without mitigation
Thresholds
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
36/83
PQ performance is also studied spatially in EVORA, which was modelled in DIgSILENT
with provided parameters regarding load and lines. UBPI was calculated with the assumed
locations of unbalance loads, fixed and randomly selected nonlinear loads. The premier
results are presented in Figure 3.10, and it can be seen that the PQ performance around
buses 14 and 30 is worse than other areas.
Figure 3.10: Heatmap indicating PQ performance of EVORA based on UBPI
3.5 Harmonic study in LV network
In this section, the propagation of harmonics in the LV distribution network is studied
in detail. In general, the amount and level of detail of information on harmonics in such
networks is limited. The modelling provides an insight into factors that influence harmonic
performance in the network, particularly for voltage distortion and losses.
A modelling method is developed and a case study carried out using the CIGRE
benchmark LV domestic distribution network. New understanding gained from the modelling
is used to inform the work in section 4.3 on mitigation.
3.5.1 Modelling
The modelling is broken down into three sections. Power electronic converters are
modelled using Simulink and PSCAD, LV distribution cables are modelled using the Finite
Element Method (FEM) software FEMM and the models are integrated into the CIGRE
benchmark LV network in Simulink. Integration of these models presents a particular
challenge because FEM results are in the frequency domain. To overcome this, the converter
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
37/83
models are represented in the frequency domain using their Norton equivalent circuit, as
described in [47, 48]. Further details of these methods are outlined in the following sections.
3.5.1.1 Converter modelling
In order to investigate the behaviour of a representative range of harmonics, the
types of converter are broken down into two categories: line frequency diode bridge
converters and switched-mode ac-dc converters. Line frequency diode bridge converters are
represented by a single-phase line frequency diode bridge converter and an inductor for
power factor correction. The circuit diagram is shown in Figure 3.11. Typically, such
converters are used for low power devices such as PCs, CFLs or televisions.
Figure 3.11: Circuit diagram for line frequency diode bridge converter.
Switched-mode ac-dc converters are typically used for devices with higher power
ratings and are typically required in cases where power is fed into the grid (e.g. distributed
generation). The principle examples are EVs and PV installations. Three-phase connection is
assumed for these devices. The circuit diagram used for the switched-mode ac-dc converter is
shown in Figure 3.12. The voltage waveform in this circuit is generated using 2-level Pulse-
Width Modulation (PWM) with a switching frequency of 5kHz.
Figure 3.12: Circuit diagram for switched-mode ac-dc converter.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
38/83
The harmonic spectra from these devices were obtained using the Fourier transform
and a selection of frequencies is shown in Figure 3.13.
Figure 3.13: Harmonic spectra of line frequency diode bridge converter (left) and switched-mode ac-dc
converter (right).
Norton equivalent circuits
In order to integrate the converters into a model which includes the frequency
domain cable parameters, Norton equivalent circuits are required at a suitable range of
frequencies. Frequencies regulated in [49] are from the 3rd to the 40th harmonics. In addition,
the most significant harmonics generated by the switched-mode ac-dc converter are
considered, namely the 97th to 102nd, 197th to 202nd and 297th to 302nd.
The Norton equivalent circuit for a line frequency diode bridge converter is shown in
Figure 3.14.
Figure 3.14: Norton equivalent circuit of line frequency diode bridge converter.
The magnitude and phase angle of the current source at each frequency was obtained
using the Fourier transform on the converter’s AC current waveform. The Norton equivalent
impedance was derived using the method described in [48]. In this method, the pure 50Hz
sine wave voltage input from the AC system is altered by the addition of a voltage component
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
39/83
at a single harmonic. The change in the harmonic current at the same frequency is observed
and the equivalent impedance calculated (using Z = V/I). This is performed at every harmonic
of interest with harmonic voltage components of 0o and 90o.
This method provides frequency domain equivalents of the converter models which
account for the effect that system voltage distortion has on harmonic emissions. Measuring
the harmonic impedance at 0o and 90o also accounts for the effect of voltage distortion phase
angle on Norton equivalent harmonic impedance.
At the higher frequency range (i.e. 97th harmonic and higher), the harmonic
impedance is well represented by the device filters. In the case of the line frequency diode
bridge converter, this means the filter inductance and resistance in series with the DC side
filter capacitor (Lf, Rf and Cin respectively in Figure 3.11). In the case of the switched-mode ac-
dc converter, the filter capacitor (Cf in Figure 3.12) was used as the Norton equivalent
impedance (including capacitor ESR and ESL) at all frequencies.
3.5.1.2 Cable modelling
Two LV cable constructions, used in the CIGRE benchmark LV distribution network,
were modelled. One type has 50mm2 and the other 240mm2 conductor size. Both have
aluminium conductor, XLPE insulation and PVC outer sheath [50].The cable construction is
shown in Figure 3.15.
Figure 3.15: Cable construction used in FEM modelling. Beginning with the top left conductor and moving
clockwise, the conductors are phase b, phase a, neutral, phase c.
The FEM provides a flexible and accurate method to calculate cable parameters at a
range of frequencies [51]. The cables were modelled at the same range of frequencies as
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
40/83
described in section 3.5.1.1. By way of example, current density plots for the 50mm2 cable
are shown in Figure 3.16.
(a) (b)
(c) (d)
Figure 3.16: Current density distribution in cable with 50mm2 conductors at a range of frequencies. (a) 50 Hz, (b) 250 Hz, (c) 750 Hz, (d) 5 kHz.
Important observations from Figure 3.16 that affect cable parameters are:
Skin effect and proximity effect are both clearly seen at the 15th and 100th harmonics
(750 Hz, 5 kHz), but are less pronounced at low order harmonics. This affects both
resistance and inductance.
The 15th harmonic (750 Hz) is a zero-sequence harmonic and therefore the phase
currents are in phase and the neutral acts as a return path, carrying approximately 3x
the phase current. This increases losses by approximately a factor of 4.
The resulting resistance and inductance for the 50mm2 cable are shown in Figure 3.17.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
41/83
Figure 3.17: Resistance (left) and inductance (right) for phases a and b for the 50mm2 cable (phase c is
similar to a).
The inductance for phase b is different to that for phases a and c, particularly for the
zero sequence harmonics. This is explained by the position of phase b in relation to the
neutral, which acts as the return conductor for the zero-sequence harmonics.
3.5.1.3 CIGRE benchmark LV network modelling
The cable and converter models were used in the CIGRE benchmark model. The
topology is shown in Figure 3.18 [52]. Zero residual harmonic distortion is assumed. The short
circuit level of the MV bus is assumed to be infinite.
Figure 3.18: LV CIGRE residential benchmark topology.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
42/83
Load Profile
The load connected at each node in the benchmark model is shown in Table 3.7. The
power of each type of load connected (resistive, inductive, line frequency diode bridge
converter, switched-mode ac-dc converter) is estimated according to data from [17-19]. In
line with the SuSTAINABLE concept, a significant amount of connected EVs and PV
installations is included. The basis for the load profile is that of a typical UK house for 2035.
The load profile is shown in Figure 3.19.
Table 3.7: Load at each node in CIGRE benchmark model
Figure 3.19: Load profile applied to LV residential network for c.2035.
In addition, PV share is modelled as 10% of load. EVs are assumed to be interfaced to
the grid through switched-mode ac-dc converters – this is consistent with the level
controllability inherent in the SuSTAINABLE concept. A simplifying assumption is made that
loads connected to the network are balanced. The modelling focusses on one point in time,
with peak load conditions, in order to measure the worst case for voltage THD in the system.
Temporal variation is considered in section 6.4, where loss analysis is based on an
extrapolation from this result using standard load curves [53].
3.5.1.4 Iterative process
The model is run at each frequency in the considered range. For each node and at
each harmonic frequency, the phase of the voltage distortion is recorded and used to adjust
the Norton equivalent impedance of the line frequency diode bridge converter models at that
Node Load (kW)
R1 200
R11 15
R15 52
R16 55
R17 42
R18 47
Lighting 27%
PCs 12%
TV 20%
Routers and Phones 1%
Wet Appliances 1%
Cooking 8%
Cold Appliances 13%
EVs 18%
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
43/83
node. This results in an iterative process which gives the final result of the harmonic power
flow at each frequency.
3.5.2 Harmonic power flow results
The key results are the system losses and total harmonic distortion. The total losses
by category are shown in Table 3.8. The total harmonic distortion at each node to which load
is connected is shown in Table 3.9.
Table 3.8: System losses by category.
Current Losses (kW)
All 11.00
Load (50 Hz) 7.44
Harmonics 3.56
3rd-41st Harmonic 3.56
97th-302nd Harmonic 7.85E-05
3rd Harmonic 2.42
5th Harmonic 0.83
7th Harmonic 0.23
3rd Harmonic Neutral 1.82
Table 3.9: Voltage THD at each node with connected load.
Node THD phase a THD phase b
R1 0.00% 0.00%
R11 1.54% 1.77%
R15 3.62% 3.88%
R16 3.47% 3.95%
R17 4.32% 4.94%
R18 4.51% 5.15%
The breakdown of the voltage distortion for phases a and b (phase c is almost
identical to phase a) for a selection of frequencies for the node with the highest THD (node
R18) are shown in Table 3.10.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
44/83
Table 3.10: Voltage distortion at node with highest THD (R18).
Harmonic Order Phase a Voltage Distortion as % of
Fundamental Phase b Voltage Distortion as % of
Fundamental
3 1.73% 3.15%
5 3.31% 3.21%
7 2.05% 2.00%
9 0.28% 0.59%
11 0.85% 0.85%
13 0.90% 0.84%
15 0.17% 0.22%
17 0.32% 0.31%
19 0.46% 0.47%
3.5.3 Discussion
Losses caused by harmonics are 48% of the losses caused by the 50Hz current. The
losses caused by the 3rd harmonic are 68% of the total harmonic losses. This results from the
loading of the neutral cable by the 3rd harmonic. The 3rd harmonic losses in the neutral alone
are more than all the other harmonic losses combined.
Losses at frequencies from the 97th harmonic and higher are minimal, in spite of the
higher cable resistance at these frequencies. There are two reasons. First, emissions are
lower at these higher order harmonics than for low order harmonics. Secondly, the inductive
reactance of the cables in the network is very large at these frequencies and the currents
therefore circulate in the lower impedance paths provided by the capacitances connected
locally at each node.
Voltage distortion increases with distance from the point of connection to the MV
network (main bus). This is to be expected as harmonic current flowing through the network
impedance results in the voltage distortion and the network impedance between node and
main bus increases with distance.
Considering the voltage distortion in phase a and c, it is highest for the 5th harmonic,
although the system current is lower for the 5th than the 3rd harmonic. This results from the
higher inductive reactance of the cables at 250 Hz compared to 150 Hz.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
45/83
In phase b, the 3rd harmonic voltage distortion is significantly higher than for phases a
and c. This is true for all zero-sequence harmonics and results in a significantly greater THD in
phase b. The reason is that the zero-sequence inductance in phase b is larger for the third
harmonic than that in phase a and c. This can be seen in Figure 3.17 and is caused by the
geometry of the cable in which phase b is located further from the return conductor (in this
case the neutral) than phases a and c (which have near identical results).
Assuming a voltage THD limit of 5%, the voltage THD must be mitigated, while losses
are high and should also be addressed. Sections 4.3 and 6.4 address mitigation.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
46/83
4 Effectiveness of PQ Mitigation
The first part of this section investigates the mitigation effect of FACTS devices on
various PQ phenomena (including voltage sags, harmonics and unbalance) Evaluation
methodologies/indices BPI, THD and VUF are applied in the study. The detailed analysis of the
mitigation effectiveness of network-based solutions can be found in [25]. In the second part
of the section harmonic mitigation and suppression of harmonic emission in LV networks is
presented and discussed.
4.1 PQ mitigation using FACTS devices - small test network
The impact of FACTS devices on various PQ phenomena is studied on a simple test
network first. The test network is given in Figure 4.1, which consists of an external grid
modeled as a PV bus type, two buses and one sensitive load modeled as an impedance with
2MW of rated power.
B1 B2
External Grid
Sensitive
load
Line
Figure 4.1: A simple test network
1) Mitigation of voltage sags
In the dynamic simulation of voltage sags, a 3-phase short-circuit is applied to B1 at
0.1s and cleared at 0.3s. In this case, B2 is exposed to a voltage sag. SVC, STATCOM and DVR
are connected to the network individually to test the dynamic response of voltage at B2
during 0-2s, which includes the periods of pre-fault, during-fault and post-fault. The control
parameters of these devices are tuned to be optimal based on trial-and-error method.
STATCOM used for voltage regulation is denoted as STATCOM-V, and that used for reactive
power compensation is denoted as STATCOM-Q. The compensation ability of SVC, STATCOM-
Q and DVR during the sag event is presented in Figure 4.2 (a), (b) and (c) respectively. Since a
sag is defined as the decrease in voltage magnitude between 0.1 and 0.9 p.u., the threshold
of 0.9 p.u. is marked as a solid black line in Figure 4.2. Under initial condition, the voltage
magnitude at B2 is 1.p.u. Given the ideal initial voltage at B2, STATCOM-V and STATCOM-Q
present similar performance. For SVC and STATCOM, relatively large rating is required to
compensate the voltage to 0.9 p.u., as seen from Figure 4.2 (a) and (b), due to the factor that
B2 is connected to a strong bus which is highly affected by the external grid that is modeled
as PV bus. SVC and STATCOM have similar compensation performance during sag event
providing the same rating is applied. As seen from Figure 4.2, the three FACTS devices present
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
47/83
different capabilities of post-fault voltage recovery. As the rating is increased, STATCOM has
severer post-fault voltage oscillation compared to SVC. DVR outperforms SVC and STATCOM
in terms of the obtained dynamic response of voltage. DVR not only compensates the during-
fault voltage as expected, and also recovers the post-fault voltage quickly without
experiencing voltage oscillation.
(a) SVC (b) STATCOM-Q & STATCOM-V
(c) DVR
Figure 4.2: Dynamic response of voltage with FACTS devices
2) Mitigation of harmonics
The mitigation of harmonics using various FACTS devices is studied by injecting
harmonic current to the sensitive load. The parameters of the injected harmonic current are
given in Table 4.1. The THD performance obtained with various FACTS devices is given in
Table 4.2. It can be seen that STATCOM has the best performance in terms of harmonic
mitigation effect among the three FACTS devices. SVC-5MVA has a magnified THD compared
to the case of No FACTS, due to the connection of SVC of 5MVA causing resonance at around
3-order. When the rating of SVC is increased to 150MVA, the corresponding THD
performance is significantly improved compared to the case of SVC-5MVA. As for STATCOM
and DVR, the variation of device rating does not appreciably impact THD.
Table 4.1: Harmonic current injection
Harmonic
Order
IAh/IA
(%)
IBh/IB
(%)
ICh/IC
(%)
θAh-θA
(deg.)
θBh-θB
(deg.)
θBh-θB
(deg.)
3 3.849 5.269 4.766 79.353 123.649 121.960
5 25.286 26.130 15.235 24.325 43.858 169.577
7 10.360 7.527 2.134 38.968 157.14 115.175
9 1.206 3.010 5.576 162.417 89.596 26.130
0 0.2 0.4 0.6 0.8 10.5
1
1.5
2
2.5
Time (s)
Voltage m
agnitude (
p.u
.)
0MVA
5MVA
50MVA
100MVA
150MVA
>200MVA
0.9
0 0.2 0.4 0.6 0.8 10.5
1
1.5
2
2.5
Time (s)
Voltage m
agnitude (
p.u
.)
0MVA
5MVA
50MVA
100MVA
>150MVA
0.9
0 0.2 0.4 0.6 0.8 10.5
1
1.5
2
2.5
Time (s)
Voltage m
agnitude (
p.u
.)
0MVA
0.05MVA
0.25MVA
0.5MVA
0.75MVA
>1MVA
0.9
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
48/83
Table 4.2: THD (%) obtained with FACTS devices
No
FACTS
SVC-
15MVA
SVC-
150MVA
STATCOM DVR
THDA at B1 0.95 15.43 1.04 0.08 0.11
THDB at B1 1.19 14.24 1.05 0.10 0.06
THDC at B1 0.63 16.71 1.04 0.05 0.14
THDA at B2 1.11 18.87 1.22 0.09 0.18
THDB at B2 1.39 17.65 1.23 0.12 0.10
THDC at B2 0.73 19.20 1.21 0.06 0.24
3) Mitigation of unbalance
The correction effect of various FACTS devices on unbalance phenomenon is studied
by connecting an unbalance load at B2. The rated power of the three phases of the sensitive
load (connected as 3PH-Delta) is given in Table 4.3. In the study, the SVS component of SVC is
equipped with unbalanced controller. PWM is modeled as three-phase converter. The results
are given in Table 4.4. DVR does not impact the VUF at upstream buses. However it increases
the VUF slightly at downstream buses, due to the positive-sequence voltage is used as the
reference for the controller in DVR. To improve this, independent phase controller (i.e., phase
voltage compensation) can be applied. However this approach is not preferred in practice,
due to this requires constantly withdrawing unbalance current from the DC capacitor which
will cause the reduced working life of the capacitor. Besides that, cost will be increased due
to the adoption of more complicated three relatively independent full bridges in PWM [54]. In
the study, DVR is not developed for the purpose of unbalance mitigation.
Table 4.3: Parameters for unbalance load at B2
Phase A Phase B Phase C
Active power (MW) 0.9 0.6 0.8
Power factor 0.3 0.8 0.9
Table 4.4: VUF performance by various FACTS devices
None SVC STATCOM-
50MVA
STATCOM-
150MVA
DVR
VUF at B1 (%) 0.98 0 0.23 0.13 0.98
VUF at B2 (%) 1.18 0 0.28 0.16 1.66
4.2 PQ mitigation using FACTS devices - generic distribution
network
The impact of FACTS devices on various PQ phenomena is further tested in a relatively
practical large-scale distribution network, a 295-bus generic distribution network (GDN), as
shown in Figure 4.3.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
49/83
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
235 236
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
249 250
266267
242
252
260
289
237
244
261
251
241 245246 243
262272270253 274271 276 275 263 264 273
240
254258 256259
277
247
278
248
280
234
279
233255257
293 292
294 295 297 296
299 298
300
77238
288
269
287 285286
56
BA C D E
F
H I J K L
ON
G
132kV
33kV
11kV
3.3kV
275kV
400kV
G
H2 O2
PV
H2 O2
H2 O2
H2 O2
PV
PV
PV
PV
H2 O2
195
196
H2 O2
PV
Wind Turbine
Photovoltaic
Fuel Cell
Unbalance Load
Zone-1 Zone-2 Zone-3
Figure 4.3: Single line diagram of 295-bus generic distribution network
STATCOM and SVC devices are placed at bus 217, and DVR is connected on the line
between buses 216 and 217. For the convenience of comparison, only one representative
operating point is used, and only one device is activated at one time. The parameters of
FACTS devices are tuned to be optimal. In Figure 4.3, the two branches following bus 217 are
denoted as feeders 1 and 2 respectively.
The compensation effect of various devices on voltage sags is studied by creating a
fault at bus 178. The dynamic response of voltage at bus 217 during simulation is given in
Figure 4.4. Without the connection of any FACTS device, the steady state voltage at bus 217 is
larger than 1 p.u. With the connection of STATCOM-V, SVC or DVR, the pre-fault voltage is 1
p.u. When STATCOM-Q is connected, the pre-fault voltage is similar to that obtained without
the connection of any device, as seen from Figure 4.4 (b). Although STATCOM-Q and
STATCOM-V perform similarly during fault, they present different post-fault voltage recovery
ability. In the post-fault period, the performance of STATCOM-Q and STATCOM-V is similar
when the rating is small. However, as the rating increases, STATCOM-V results in more
serious voltage oscillation compared to STATCOM-Q. SVC also suffers from post-fault voltage
oscillation when the rating is increased. In the case of generic distribution network, SVC,
STATCOM-Q and STATCOM-V have the limitation of compensation cap, and they are not able
to compensate the voltage up to 0.9 p.u. In this case, DVR provides much better voltage
dynamic response at B217: it is able to compensate the voltage up to 1 p.u.; and it does not
lead to any post-fault voltage oscillation.
(a) SVC (b) STATCOM-Q
216
217
218 171
219
220
172
173
174
175
176
177
178
0 0.2 0.4 0.6 0.8 10.6
0.8
1
1.2
1.4
1.6
1.8
Time (s)
Voltage m
agnitude (
p.u
.)
0MVA
1MVA
5MVA
10MVA
12MVA
0.9
0 0.2 0.4 0.6 0.8 10.6
0.8
1
1.2
1.4
1.6
1.8
Time (s)
Voltage m
agnitude (
p.u
.)
0MVA
1MVA
5MVA
10MVA
12MVA
0.9
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
50/83
(c) STATCOM-V (d) DVR
Figure 4.4: Dynamic response of voltage with FACTS devices
The mitigation effect of various FACTS devices on harmonics and unbalance
phenomena is given in Table 4.5 and Table 4.6 respectively. STATCOM-Q outperforms other
FACTS devices in terms of harmonics mitigation, and SVC has the best performance in terms
of unbalance mitigation, which is in line with the discussion on the simple test network.
Table 4.5: THD performance by various FACTS devices
None SVC STA-Q STA-V DVR
THDA at B216 2.143 0.041 0.046 0.045 0.536
THDB at B216 0.636 0.037 0.026 0.028 0.437
THDC at B216 1.974 0.064 0.078 0.087 0.759
THDA at B217 2.154 0.024 0.007 0.007 0.108
THDB at B217 0.639 0.011 0.001 0.001 0.018
THDC at B217 1.983 0.028 0.011 0.013 0.206
THDA at B178 2.183 0.024 0.007 0.008 0.109
THDB at B178 0.647 0.011 0.001 0.001 0.018
THDC at B178 2.005 0.028 0.011 0.013 0.209
Table 4.6: VUF performance by various FACTS devices
None SVC STA-Q STA-V DVR
VUF at B216 (%) 4.1 0.2 0.6 0.5 4.0
VUF at B217 (%) 4.1 0 0.5 0.4 4.4
VUF at B178 (%) 4.1 0 0.5 0.4 4.4
To see the influence of each device on its neighbouring buses, BPIs at all buses are
given in Figure 4.5. SVC, STATCOM-Q and STATCOM-V have similar performance. DVR
improves the sag performance at downstream buses significantly, i.e., the buses on feeders 1
and 2. DVR also improves the sag performance at the upstream buses, by preventing the sag
propagating from downstream buses (the fault is simulated at the downstream) to upstream
buses.
0 0.2 0.4 0.6 0.8 10.6
0.8
1
1.2
1.4
1.6
1.8
Time (s)
Voltage M
agnitude (
p.u
.)
0MVA
1MVA
5MVA
10MVA
12MVA
0.9
0 0.2 0.4 0.6 0.8 10.6
0.8
1
1.2
1.4
1.6
1.8
Time (s)
Voltage m
agnitude (
p.u
.)
0MVA
1MVA
2.5MVA
>5.5MVA
0.9
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
51/83
Figure 4.5: BPI performance of all buses
Harmonics performance of all buses due to the connection of various FACTS devices is
presented in Figure 4.6. It can be seen that the performance of SVC, STATCOM-Q and
STATCOM-V is similar. For DVR, the obtained THDs at the downstream buses are better than
those at the upstream buses.
Figure 4.6: THD performance of all buses
The unbalance performance of all buses by various FACTS devices is given in Figure 4.7.
SVC outperforms other FACTS devices. When SVC is installed, the obtained VUFs at the buses
which are close to the installation location are greatly improved. Between STATCOM-V and
STATCOM-Q, the former provides better performance at the buses which are close to B217,
while the latter provides better mitigation effect at the buses which are relatively further
away from B217. DVR has minor impact on the upstream buses while causing slightly higher
VUFs at the downstream buses.
Figure 4.7: VUF performance of all buses
50 100 150 2000
1
2
3
4
Bus indexB
PI
SVC
STATCOM-Q
STATCOM-V
DVR
None
Buses infeeder 2
Buses infeeder 1
50 100 150 2000
1
2
3
4
Bus index
TH
D
SVC
STATCOM-Q
STATCOM-V
DVR
None
Buses infeeder 2
Buses infeeder 1
50 100 150 2000
2
4
6
8
10
Bus index
VU
F
SVC
STATCOM-Q
STATCOM-V
DVR
None
Buses infeeder 1
Buses infeeder 2
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
52/83
4.3 Harmonic mitigation in LV Network
Following the modelling in section 3.5, harmonic mitigation in the case study of the
CIGRE LV Benchmark Network is carried out.
Three mitigation methods are considered. The first, which is the lowest cost existing
solution and primarily used as a benchmark, is the addition of tuned passive filters to the
network to reduce voltage THD. The second is to consider the effect of distributed harmonic
compensation, which can be provided as an ancillary service by distributed energy resources
that are interfaced to the network with a switched-mode ac-dc converter [55-57]. The
strategy is to use the switched-mode ac-dc converters to produce harmonics in anti-phase to
emissions from line frequency diode bridge converters, effectively absorbing those emissions.
The method has the same principle as active filter operation, but avoids the cost of a
dedicated filter. The third method is to replace line frequency diode bridge converters with
switched-mode ac-dc converters.
The intention is to establish an economic framework in which the costs of installing
filters are avoided and losses reduced by providing incentives to owners of grid connected
equipment to either reduce their harmonic emissions or compensate harmonic emissions;
this is examined further in section 6.4.
4.3.1 Passive filters
A limit of 5% for voltage THD is assumed. Second order passive damped tuned filters
are considered. These have the following strengths [50]:
A filter at a single location provides benefits throughout the network.
Low level complexity leads to high reliability, no control requirement and relatively
low cost.
One major disadvantage of the tuned filter is that its performance is frequency
sensitive and can be affected by deviation from nominal transmission frequency.
The filter topology is shown in Figure 4.8.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
53/83
Figure 4.8: Second order damped shunt filter.
It was found that the most effective placement for the filter(s) was at the node with
highest voltage THD (node R18). Installation of a 3rd harmonic filter was found to be
problematic. The reduction in the 3rd harmonic voltage leads to an increase in the 3rd
harmonic current drawn by devices. This is because the voltage distortion at a given harmonic
acts to reduce the current drawn at that frequency. In the case of the 3rd harmonic, the extra
current drawn returns in the neutral path and further increases the largest source of
harmonic losses.
A filter for the 5th harmonic was therefore modelled. It was found that a tuned 5th
harmonic filter was adequate to reduce the voltage THD in phase b to below 5%. In addition,
losses at the 5th harmonic are reduced because a large proportion of these currents are now
drawn from the filter rather than the connection point to the MV network and therefore flow
through a shorter cable length. The voltage THD at each node for phases a and b, before and
after filtering are shown in Table 4.7.
Table 4.7: Voltage THD mitigation by use of tuned filters.
Base Case 5th Harmonic Filter
Only 5th and 7th Harmonic
Filters
Node Phase a Phase b Phase a Phase b Phase a Phase b
R1 0.00% 0.00% 0.00% 0.00% 0.00% 0.00%
R11 1.54% 1.77% 1.19% 1.50% 1.03% 1.37%
R15 3.62% 3.88% 3.09% 3.44% 2.87% 3.26%
R16 3.47% 3.95% 2.61% 3.27% 2.19% 2.97%
R17 4.32% 4.94% 3.04% 3.94% 2.37% 3.48%
R18 4.51% 5.15% 3.06% 4.04% 2.28% 3.52%
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
54/83
4.3.2 Distributed harmonic compensation and reduction of harmonic
emission
In order to assess a fair price for a scheme to incentivise customers to reduce or
mitigate harmonics, it is necessary to establish what level of loss and voltage THD mitigation
is achieved for a given amount of harmonic mitigation. This depends on spatial
considerations, i.e. reduction of harmonics at certain nodes is more effective than at others.
In order to do this, the marginal effect on voltage THD throughout the system and system
losses of reducing harmonics at each node is measured. The results are shown in Table 4.8
and Table 4.9. In order to normalise the results, the mitigation is assumed to be equivalent to
the harmonic spectrum of a device drawing 100 W, with harmonic spectrum as shown in
Figure 3.13, in each phase.
Table 4.8: Effect on system losses of harmonic mitigation equivalent to harmonic spectrum for 100 W device with line frequency diode bridge converter grid interface in each phase.
Node with Mitigation
Distance from Main Bus (m)
Loss Mitigation for Harmonic Load Reduction (W)
Loss Mitigation for Harmonic
Compensation (W)
R1 0 0 0
R11 90 1.04 1.13
R15 240 2.21 2.70
R16 205 1.99 2.28
R17 310 2.14 2.48
R18 345 2.59 2.94
Table 4.9: Reduction of voltage THD of harmonic mitigation equivalent to harmonic spectrum for 100 W device with line frequency diode bridge converter grid interface in each phase. Results for phase b are shown.
THD Measured at
Node
Compensated Node
R1 R11 R15 R16 R17 R18
R1 0.0000% 0.0000% 0.0000% 0.0000% 0.0000% 0.0000%
R11 0.0000% 0.0115% 0.0052% 0.0055% 0.0053% 0.0053%
R15 0.0000% 0.0045% 0.0283% 0.0068% 0.0065% 0.0064%
R16 0.0000% 0.0045% 0.0058% 0.0182% 0.0124% 0.0123%
R17 0.0000% 0.0041% 0.0052% 0.0116% 0.0260% 0.0203%
R18 0.0000% 0.0038% 0.0048% 0.0107% 0.0244% 0.0186%
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
55/83
An example in order to explain Table 4.9 follows. Consider node R15. The voltage THD
at this node before any compensation is 3.88%. Compensation can be provided anywhere in
the network (compensating one node affects all nodes). Providing compensation at node R11,
equivalent to mitigating emissions from a 100 W line frequency diode bridge converter,
reduces the voltage THD at node R15 by 0.0052%. Similar results are true when
compensation is provided at nodes R16, R17 or R18. Providing the same compensation at
node R15 itself reduces voltage THD at the node by 0.0283%, i.e. it is much more effective.
Voltage distortion mitigation is most effective when carried out at or close to nodes
that have the greatest voltage THD. It can be seen that harmonic reduction at the load
connected directly to the main bus does little to mitigate harmonic effects. Loss mitigation is
most effective when carried out far from the main bus.
The methods of compensation and reduction have the additional advantage that they
are not affected by deviations from nominal system frequency. In the case of compensation,
the switched-mode ac-dc converter automatically responds to changes in system frequency.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
56/83
5 PQ Mitigation Planning based on Technical
performance of the network
5.1 Optimisation methodology
With mitigation options (both FACTS devices and network-based solutions), an
optimal deployment methodology for mitigation solutions has to be applied to identify the
most appropriate type of mitigation, rating of devices, location of their placement in network
based on the objectives of minimising PQ severity indices (or minimising deviation from
guaranteed quality levels for each category of customers, which will be introduced later). In
the study, the objective function, F, which is to evaluate the mitigation solutions, is defined
as:
𝐹(𝑥) = ∑ (𝑈𝐵𝑃𝐼𝑖(𝑥))𝑖 subject to: 𝑔(𝑥) = 0, ℎ(𝑥) ≥ 0 (5.1)
where x denotes the variables to be optimised, i.e. the mitigation solutions, which consist of
the location, type and size of various FACTS devices and network-based solutions; 𝑔(𝑥) = 0
and ℎ(𝑥) ≥ 0 denote the equality and inequality constraints respectively. It can be seen that
with the sum of the unified indices of all buses, objective function F reflects the severity of
three phenomena throughout the whole network.
5.1.1 Identification of feasible locations for deployment of mitigating
solutions
In order to optimally deploy mitigating solutions , potential and effective locations are
initially identified. The locations selected for the application of network-based solutions are
mainly based on the ranking of buses according to BPI, VUF and THD, together with the
network resources and geographical constraints. The locations for installing FACTS devices
are selected based on PQ performance and sensitivity analysis. Firstly, buses are sorted
according to BPI, VUF, THD, ∑ |𝜕𝑉𝑗
𝜕𝑄|
𝑁𝐵𝑗=1 and ∑ |
𝜕𝑉𝑗
𝜕𝑃|
𝑁𝐵𝑗=1 in descending order, respectively. The
ranking index of bus i with respect to BPI is denoted as RBPI(Bi), and the same applies to other
variables. Then RBPI(Bi)=1 suggests that bus i is experiencing the worst sag performance, and
𝑅𝜕𝑉/𝜕𝑄(𝐵𝑖) = 1 that voltage of bus i is the most sensitive to the injection of reactive power.
With these rankings, the potential locations are chosen globally (i.e., based on the whole
network) and zonally (i.e., based on zones information) respectively.
The buses having RBPI=1, RVUF=1, the smallest 𝑅BPI + 𝑅𝜕𝑉/𝜕𝑄, the smallest 𝑅VUF +
𝑅𝜕𝑉/𝜕𝑄 and the smallest RBPI+𝑅VUF are selected as potential locations for installing SVC.
Following this appropriate FACTS devices are preliminarily paced at the selected potential
locations, the same selection procedure is then performed again to select the second set of
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
57/83
potential placements. The aforementioned procedure is applied to select the potential
locations for installing STATCOM and DVR, while in this case the buses having RBPI=1, RVUF=1,
the smallest 𝑅BPI +𝑅𝜕𝑉/𝜕𝑃 +𝑅𝜕𝑉/𝜕𝑄
2, the smallest 𝑅VUF +
𝑅𝜕𝑉/𝜕𝑃 +𝑅𝜕𝑉/𝜕𝑄
2 and the smallest
RBPI+𝑅VUF are selected. Instead of using 𝑅𝜕𝑉/𝜕𝑄, 𝑅𝜕𝑉/𝜕𝑃 +𝑅𝜕𝑉/𝜕𝑄
2 is used in this case as both
STATCOM and DVR can transmit both active and reactive power between devices and the
grid.
To ensure the capability of providing specific quality of supply required in a particular
zone, the potential locations should be also selected zonally. In the zonal selection, the
procedure is the same as the global selection, except that the ranking procedure is performed
within the zones rather than within the whole network.
Geography feasibility could be also taken into account during the process of selecting
potential locations. The same aforementioned selection procedure is performed to select the
potential locations for installing passive filters, while in this case the buses are ranked based
on RTHD. Besides, the intersection of two branches which have more than three buses in
downstream branches is also initially made available for placement of passive filters, as the
passive filters located at the intersection of two branches can stop the harmonic current
flowing to other branches.
5.1.2 Optimisation methodology
With these pre-selected locations, greedy algorithm is used to search the potential
solutions (i.e., optimal placement of FACTS devices and their optimal rating settings) to
minimise the gap between PQ performance achieved and the PQ thresholds. Greedy
algorithm is chosen due to its simplicity of implementation and computation efficiency. It
divides the problem into different consecutive stage and solves the problem heuristically by
making the local optimal choice ‘greedily’ at each stage. The greedy algorithm has been
applied successfully for device placement in large scale power systems.
Before applying the greedy algorithm, a pool of potential solutions, denoted as set U,
should be determined based on the initial placement/locations and rating constraints. For
FACTS devices, their initial placements including their location and type have been decided.
Assume there are MD potential devices. For each potential device, an extra variable needs to
be determined, i.e. the rating of the device. The rating range of each device can be divided
into MI intervals, and for each interval, a rating is chosen by randomly selecting a value within
the interval. Thus, a pool of MD×MI potential devices, which consists of locations, types of
devices and ratings, and a number of network-based solutions, denoted as set U are made
available initially for optimisation process.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
58/83
Figure 5.1: Flowchart of greedy algorithm
After the solution pool U and objective function are defined, a conventional
methodology, Greedy Algorithm, which is widely applied to solve the travelling salesman
problem (TSP), is applied to optimise the mitigation variables of FACTS devices. The greedy
algorithm solves the problem heuristically by making the local optimal choice ‘greedily’ at
each stage, until the stop criteria is reached. The choice made by the algorithm depends on
the choices so far instead of the future choices. In other words, it chooses the optimal
solutions of sub-problems derived along the selection process. In our study, a certain number
of stages are applied for the placement of FACTS devices. The Greedy Algorithm is applied at
each stage where only one FACTS device which has the best mitigation effect is chosen. The
choice is made after all unselected FACTS devices with various proposed ratings are tested
individually. The reason for repeatedly testing all the unselected FACTS devices is that the
operation condition varies due to the integration of selected devices, and the mitigation
effect of mitigation devices would vary as well, even though the same device with the same
rating is applied. The repeating device selection stops if the improvement to the objective
function between two neighbouring stages is smaller than a pre-set value. The flowchart of
the application of greedy algorithm for this allocation problem is given in Figure 5.1, where s is
the chosen solution which is corresponding to the minimum SGI at each stage; Γ denotes the
devices selected so far; and X is the updated pool of potential solutions at each stage. At each
stage, X is updated by removing its elements which have the same location and type of devise
as the selected s. The optimisation procedure can be terminated if the size of Γ reaches the
preset maximum number of devices allowed to be installed, or if the improvement of PQ
performance between two sequential stages is smaller than a preset threshold. Set Γ is
selected as the final optimal mitigation solutions.
To test this methodology, UBPI is taken as the objective function to be minimised by
Begin
X=U; Γ=Φ;
End
Select sϵX that minimizes
objective function;
X=X-{all elements in X which have the
same location and type of device as s};
Γ=Γ∪{s};
Reach stop criteria?
Input set U of
potential devices
Yes
No
Install covered devices Γ;
Update X by regenerate rating randomly
within corresponding interval
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
59/83
the proposed methodology. The convergence characteristic of the optimisation methodology
is presented in Figure 5.2 (a), which shows that the sum of UBPIs in the network is greatly
reduced when one mitigation technique is applied, and the performance is steadily improved
when the number of installed devices>4. The UBPIs obtained at all buses are given in Figure
5.2 (b), which shows that the UBPI is significantly improved with mitigation solutions. To
visually present the improvement of PQ performance, the heat-maps of UBPIs obtained
without and with mitigation are given in Figure 5.3 (a) and (b) respectively. It can be seen that
the critical areas which are exposed to the PQ disturbance are greatly improved.
(a) Convergence curves (b) UBPI obtained with and without mitigation
Figure 5.2: UBPIs of all buses and convergence characteristic of the optimisation methodology
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
H2 O2
Non-linear load (fixed)
Photovoltaic
Fuel Cell
Wind Turbine
Unbalance load (fixed)
(a) without mitigation
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
H2 O2
Non-linear load (fixed)
Photovoltaic
Fuel Cell
Wind Turbine
Unbalance load (fixed)
Non-linear load (random)
FACTS_based mitigation
(b) with mitigation
Figure 5.3: Illustration of the performance of the proposed optimisation methodology
5.2 Simulation results
The developed optimisation methodology introduced above is used to search the
optimal mitigation solutions to provide differentiated levels of PQ in the 295-GDS distribution
network. The GDS is divided into three zones, as shown in Figure 5.4. The comparison
between UBPIs obtained with and without mitigation solutions is presented in Figure 5.5 (a),
0 2 4 6 8 10 120
20
40
60
80
100
Number of installed Devices
Sum
(UB
PI i)
(H
arm
onic
, S
ag, U
nbala
nce)
0 50 100 150 200 2500
0.5
1
1.5
2
2.5
Bus No
UB
PI (H
arm
onic
, S
ag, U
nbala
nce)
Without mitigation with H5S3U2
With mitigation with H5S3U2
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
60/83
and the convergence characteristic of UBPI against the number of installed devices is given in
Figure 5.5 (b). Figure 5.6 presents the heatmaps of annual PQ performance obtained without
and with mitigation. It can be seen that the performance is significantly improved with the
optimal mitigation solutions found by the developed methodology. Furthermore, the
heatmaps obtained from the concerned operating points (corresponding the maximum load
and maximum cluster) are given in Figure 5.7. Compared to Figure 3.8, PQ performance
presented in Figure 5.7 is greatly improved.
Figure 5.4: The illustration of zone division
(a) UBPIs obtained with and without mitigation (b) Convergence curves
Figure 5.5: Illustration of the performance of the proposed methodology for provisioning differentiated QoS
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
H2 O2
Non-linear load (fixed)
Photovoltaic
Fuel Cell
Wind Turbine
Unbalance load (fixed)
(a) without mitigation
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
H2 O2
Non-linear load (fixed)
Photovoltaic
Fuel Cell
Wind Turbine
Unbalance load (fixed)
DVR
SVC
STATCOM
PF
(b) with mitigation
Figure 5.6: Heatmaps of annual PQ performance obtained with 6 devices in case 4 with and without mitigation
0 50 100 150 200 2500
0.2
0.4
0.6
0.8
1
1.2
Bus No
UB
PI (H
arm
onic
, S
ag, U
nbala
nce
)
Without mitigation
With mitigation
Error
Zone threshold
2 4 6 8 10 120
20
40
60
80
100
Number of devices installed
Sum
(U
BP
I i) (H
arm
onic
, S
ag, U
nbala
nce)
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
61/83
(a) Maximum cluster
(b) Maximum load
Figure 5.7: Heat maps of the concerned operating points with optimal mitigation solution based on UBPI
Time series simulations are adopted to validate the selected solutions of PQ. 24 hours
simulations for a selected day were performed. The results are shown in Figure 5.8, as it can
be seen few buses violate the requirements of the zone at some hours even after mitigation.
However, the common measures for long periods of studies are the percentiles, in particular
the 95th percentile, therefore these levels of performance can be accepted based on the
commonly adopted standards and regulations. Moreover, the requirements are also variable
with time, and the adopted thresholds are the most conservative and the tightest thresholds
of a probabilistic range of immunity level. In other words these thresholds (acceptable
immunity levels) can be more relaxed at some hours of the day (due to different load mix at
these hours and therefore different participation of sensitive customers) and the
“participating” sensitive customers could survive worse PQ performance at those hours. If the
pessimistic evaluation is to be avoided, probabilistic representation of the thresholds is
required.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
62/83
Figure 5.8: UBPIs obtained without and with mitigation at all buses for 24 hours’ operation
0 50 100 150 2000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Bus index
UB
PI
without mitigation
with mitigation
Thresholds
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
63/83
6 PQ mitigation Planning Based on Techno-Economic
Analysis
6.1 Methodology for assessment of economic viability of PQ
mitigation solutions
While performing economic assessment of PQ mitigation at planning level, it is
important to consider the benefits during the entire life span of the deployed solution. The
upfront investment made for a mitigation solution pays back its returns only during the life
span duration. This makes it important to consider the net present value of future benefits, as
well as the net present value of future maintenance. This brings the investment cost and its
benefit to a common ground/level of comparison with planning or deployment year as the
reference. This net present value approach accounts for the facts like inflation (denoted as i),
discount rate (denoted as r) and escalation rate (denoted as e) required for the time value of
money. Net present value can be calculated using the following equation:
𝑁𝑃𝑉 = 𝐶𝐼 +∑ (𝐶𝑡𝑏+𝐶𝑡𝑐)×(1+𝑒)𝑡𝑛
𝑡=0
((1+𝑟)(1+𝑖))𝑡 (6.1)
where CI denotes the initial capital investment (usually expressed as a negative amount), Ctb
denotes the benefit component occurring at the beginning of time period; and Ctc denotes
the cost component occurring at beginning of time period t (usually expressed as a negative
amount).
6.1.1 Assessment of financial consequence of PQ phenomena
Detailed analysis of economic impact due to voltage sags, unbalance and harmonics
are very important while considering the investment on mitigation solutions. Economic
impact can be assessed by identifying losses to different types of business due to various PQ
phenomena. Considering the diversity of business types that exists, customers can be
categorized based on their previous history of economic losses due to inadequate PQ issues,
similarities in their business process and the sensitivity of equipment/devices used for their
business. This categorization enables to evaluate the economic impact of a group of
customers using a common model, thus each category having a unique model for analysis.
This economic impact provides an estimate of financial losses for that particular customer for
a given level of PQ in a network. Since losses due to voltage sags, unbalance and harmonics
are due to different reasons, it requires separate financial assessment models for each of
them.
The financial losses to a certain type of customers due to sag events are mainly
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
64/83
caused by process trip. For instance,industrial processes (glass production, textile industry) are
very sensitive to sag events [1, 9]. The main elements/parameters that should be considered
when assessing process trip include sag characteristics, frequency of sags at the customer
busbar, equipment sensitivity, customer plant sensitivity, process operational cycle and the
plant’s load profile. In the study, sag characteristics and frequency are simulated using a
stochastic approach which is to comprehensive assessment of the impact of voltage sags in
large scale power networks [37]. With realistic modelling of process cycle and proper
probabilistic modelling of uncertainties associated with each of the influential parameters,
the approach of risk analysis is applied to assess possibility of having industrial process trip.
The financial loss due to a sag event can be assessed by [1]:
(Financial
loss) = (
Process failure risk
) × (Loss due to process trip
) (6.2)
where the loss due to process trip is derived through survey or from customers directly.
Economic losses due to unbalance are mainly caused by power and energy losses and the
shortened life of equipment, according to the general classification of unbalance cost
proposed by CIGRE/CIRED C4.107 report in the context of customers [58]. As far as customers
are concerned the major impact happens in the case of induction motors, and can be
considered as one of the major source of economic losses for them. These losses happens
both due to de-rating of induction motors and loss of useful life. As per NEMA guidelines
induction motors are not expected to operate under condition of unbalance power supply
percentage of above 5%. The guideline recommends de-rating of induction motors in order to
avoid damage of motor. NMEA MG-1 suggests the de-rating curves, which combines
relationships between winding temperature rise with percentage load and winding
temperature rise with voltage unbalance, for calculating power loss in induction motors [59].
The difference in rated power and de-rated power can be considered as loss of power due to
unbalance and can be converted to financial cost. Using the rate of unit cost of energy over a
period of time and applying the NPV method it is possible to identify the accumulated cost of
energy losses over the period of system study. These costs are considered as operating costs
since this gets accounted into the operating expense incurred from payment of electricity
charges by a business.
Customer with business types having substantial installation of induction motors
suffers the most from equipment ageing issues. Equipment ageing due to thermal stress can
modeled as [60, 61]:
𝐿 = 𝐿0′ 𝑒−(𝐵𝑐𝜃) (6.3)
where 𝐿0′ refers to the reference life of the equipment for a reference temperature 𝜃𝑜; B is a
constant for a material represented as B = E/K, where E is the activation energy of the aging
reaction in Joule/mol and K is the gas constant in JK-1mol-1; 𝑐𝜃 =1
𝜃0−
1
𝜃, where 𝜃0 is the
nominal operating temperature of the insulation in sinusoidal regime while 𝜃 is the operating
temperature in the presence of harmonics.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
65/83
As per NMEA-1 guideline the approximate increase in winding temperature ∆𝜃 of
induction motor due to percentage voltage unbalance 𝑉𝑎𝑠 can be calculated using ∆𝜃 = 2𝑉𝑎𝑠2 .
A decrease in useful life of the motor results in economic loss due to equipment replacement
and process disruption due to equipment damage. The financial cost due to equipment age
can be approximated by 𝐿0
′ −𝐿
𝐿× 𝐶𝑟𝑒, where 𝐶𝑟𝑒denotes the cost of replacing the motor.
Financial losses from voltage harmonics and current harmonics can be classified into
border category of energy losses, losses due to premature ageing and losses from equipment
malfunction. CIGRE/CIRED C4.107 report [58] presents methodologies for evaluating
economic losses arising from wave form distortion due to harmonics. The additional power
losses due to harmonics voltages and currents occur as core loss, copper loss and dielectric
losses in electrical equipment. In order to analyse overall costs of an industrial system each of
equipment type has to be considered separately for calculating power losses from individual
type of equipment. The harmonic losses for induction motors can be calculated using [58]:
𝑃𝑀 = 3 ∑ (𝑉ℎ
𝑍𝑀ℎ )
2
𝑅𝑀ℎℎmax
ℎ=ℎ1+ 𝑃𝑐𝑜
1 ∑ (𝑉ℎ
𝑍𝑀ℎ )
𝑚𝑀 1
ℎ0.6
ℎmax ℎ=ℎ1
(6.4)
Where 𝑉ℎ represent the voltage harmonic of order h; 𝑍𝑀ℎ and 𝑅𝑀
ℎ denotes the equivalent
impedance and resistance of the motor at the harmonic of order h respectively; 𝑃𝑐𝑜1 denotes
the core loss at the fundamental frequency; and 𝑚𝑀 is the numerical coefficient.
Harmonic distortions can cause additional electric and thermal stresses in insulating
materials of electric equipment. Additional electric stresses are mainly due to the increased
peak factor of voltage level resulting from the presence of harmonics. Increase in thermal
stress is due to the rise in temperature from additional copper and core losses of equipment
caused by harmonic contents. Additional stresses can cause failure of the equipment
insulation, consequently reducing the equipment life span. Financial losses to customers
occur by replacing the costly equipment and production losses due to unexpected process
failure. A simple electro thermal life model recommended by CIGRE/CIRED C4.107 based on
[60, 61] can be represented as:
𝐿 = 𝐿0′ (𝐾𝑝)
−𝑛𝑝𝑒−(𝐵𝑐𝜃) (6.5)
where 𝐾𝑝 is the peak factor of the voltage waveform represented by 𝐾𝑝 =𝑉𝑝
𝑉1𝑝∗ where 𝑉𝑝 is the
value of the distorted voltage while 𝑉1𝑝∗ is the peak value of fundamental voltage; 𝑛𝑝 is the
coefficient related to the shape of the distorted waveform; and 𝑐𝜃 =1
𝜃0−
1
𝜃. The operating
temperature rise at the hottest point at operating conditions can be evaluated using the
formula of ∆𝜃 ≈ (𝑃ℎ
𝑃0) × ∆𝜃0, where ∆𝜃and ∆𝜃0 represent the operating temperature rise
from ambient in non-sinusoidal operating condition and nominal power 𝑃0 respectively; and
𝑃ℎ is the harmonics content of power.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
66/83
The aforementioned costs due to PQ phenomena can be calculated using Net Present
Value method so that present worth of future economic losses can be identified.
6.1.2 Assessment of financial cost of PQ mitigation
The capital cost of SVC, DVR and PF can be obtained based on curve fitting approach
using available cost records. The costs of DVR and STATCOM are considered as the same. The
capital cost including construction cost of these devices can be defined as [62, 63]:
𝐶𝑆𝑇𝐴𝑇+𝐶𝑜𝑛 = 553(−0.0008𝑆𝑆𝑇𝐴𝑇2 + 0.155𝑆𝑆𝑇𝐴𝑇 + 120) (
£
𝑀𝑉𝐴𝑟) (6.6)
𝐶𝐷𝑉𝑅+𝐶𝑜𝑛 = 553(−0.0008𝑆𝐷𝑉𝑅2 + 0.155𝑆𝐷𝑉𝑅 + 120) (
£
𝑀𝑉𝐴𝑟) (6.7)
𝐶𝑆𝑉𝐶+𝐶𝑜𝑛 = 553(0.0003𝑆𝑆𝑉𝐶2 − 0.3051𝑆𝑆𝑉𝐶 + 127.38) (
£
𝑀𝑉𝐴𝑟) (6.8)
where 𝑆𝑆𝑇𝐴𝑇 , 𝑆𝐷𝑉𝑅 and 𝑆𝑆𝑉𝐶 denote the sizes of STATCOM, DVR and SVC respectively.
Continuous maintenance costs incurred every year during the life time are assumed to be 5%
for SVC, and 10% for STATCOM and DVR, of their capital cost.
A conservative reactance of 0.1p.u. is used for each FCL, and the cost of owning and
installing FCL is based on the reference given in [25], whose cost model is the ten-year
owning costs based on the cost of an IS LIMITER produced by ABB [64], with annual operation
and maintenance costs at 5% of initial cost. The cost of PF is based on [65], with the
continuous maintenance costs incurred every year assumed to be 5%. The cost of placing an
isolated transformer is based on [30], with annual maintenance costs at 5% of initial cost. The
cost of rewinding transformers is based on [66], assuming that the copper wire is 3% of the
weight of the distribution transformers, which is derived based on the experience of an
experienced rewinding worker, and the weight of distribution transformers is based on ABB
oil distribution transformer catalogue. The investment costs of other network-based
mitigation techniques are given in Table 1.1, with annual maintenance costs at 5% of the
initial cost. Tree trimming is scheduled for every 5 years.
In this task, a Voltage Disruption Cost Analysis Tool (VoDCAT) tool [25], as shown in
Figure 6.1, is applied to identify the financial impact of PQ phenomena including voltage sags,
unbalance and harmonics for the category of customers.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
67/83
Figure 6.1: Layout of VoDCAT tool [25]
Table 6.1 shows the details of customers and their processes for which calculation of
customer customized financial loss is performed. A customer process model is built for each
customer based on their grouping. Sensitive equipment is identified based on the process
types and industry. Number of sub processes, connected power supply phase for equipment,
level of sensitivity, process dependency, process immunity time, process restart time, and
process cost factor were selected randomly from reasonable set of possible values. The
detailed inputs selected for evaluated customers can refer to reference [23]. Financial
analysis of sag for each customer type is performed using VoDCAT software. This software
first evaluates a customer damage function based on the customer business type identified
by NACE code. This evaluation is by using customer survey information on financial loss due
to power supply interruption updated into the tool database. Result of this evaluation is the
customer damage function for total process interruption caused due to sags. The tool then
calculates total financial losses for the customer sag profile which is given as input. Customer
process model, equipment sensitivity model and the customer damage function is used to
calculate the customized customer loss per year for the given sag profile. Using the Net
Present Value method mentioned above the losses can be estimated for the study period of
40 years. NPV of losses were normalized based on the peak KW demand of customer for
comparison. Table 6.2 [25], provides the sensitivity of each customer’s process to sags. It can
be seen that customer process with high sag sensitivity has more sensitive equipment and
dependent processes.
Table 6.1: Details of processes for different customers[25]
Cus PQ class. Sensitive equipment Sub
Process
Process dependency Matrix
A Sensitive PLC, ASD 1,2,3,4 1111,1101,1010,0111
B Essential PLC, ASD 1,2,3,4 1111,1101,0010,0111
C Essential PLC, ASD, Contactor 1,2,3,4 1111,1101,0010,1001
D Important PC, ASD 1,2,3,4 110,110,001
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
68/83
Table 6.2: Financial analysis of voltage sags [25]
For identifying the impact of harmonics on the operating and aging cost an example
industrial system connected to the power network at 20KV/130MVA is considered. The inputs
were taken based on the information available from [60, 61].Operating and aging cost for
Transformers, Induction Motors and Cables are evaluated for a study period of 40 years for
the system being exposed to 5,7,11,13,17,19 order harmonics with THD level of 3.5% and
peak factor of 1.01. Also the impact of voltage peak factor on aging is analysed for various
levels of peak factor. Table 6.3 shows the list of equipment, their ratings and quantity with
the losses presented as percentage of operating cost due to losses at the fundamental values.
For simplification of the analysis it is assumed that all equipment operates at maximum
capacity all the time. Further details of equipment are available in [23]. This exercise
demonstrates that the total operational cost due to losses at harmonics is substantial
percentage of losses at fundamental loss values. The losses are more prominent for induction
motors and transformers, while negligible in cables. This example calculation indicates that
the operating costs due to harmonics are not negligible for an industrial system. Also this cost
increases with THDv levels. The table also shows a considerable loss of life equipment life for
induction motors and transformers. Aging costs becomes substantial as the level of power
losses or voltage peak factor increases.
Table 6.3: Equipment operating and aging costs at THDv 3.5% [25]
6.2 Optimisation based on economic analysis
The problem can be defined as an optimisation problem, which places the mitigation
techniques in the network optimally in order to minimise the overall financial cost that
includes the investment cost, operation/maintenance cost and the cost caused by various PQ
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
69/83
phenomena, and to maximise the benefits as a result of the application of mitigation
techniques. Simultaneously, in planning PQ mitigation, the provision of differentiated PQ
levels should be facilitated among zones of the network based on customers’ requirement.
The provision of differentiated PQ levels is considered as the technical requirement, and
treated as a constraint to be imposed in the optimisation process. In the study, it is included
in the objective function using the approach of Lagrangian relaxation. The present value of
annual operation/maintenance cost and cost due to various PQ phenomena during the entire
life span of the deployed solution is calculated using NPV method. The overall objective
function (F) to be minimised in the optimisation problem is defined as:
𝐹 = 𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 − 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡 + 𝛽 × 𝑃𝑄𝐺𝐼UBPI (6.9)
𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑖𝑡𝑜𝑛 = 𝐶𝐼𝐶𝐼 +∑ (𝐶𝐴𝑛𝑛𝑂𝑝𝑒𝑀𝑎𝑖
𝑡 )×(1+𝑒)𝑡𝑛𝑡=0
((1+𝑟)(1+𝑖))𝑡 (6.10)
𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡 =∑ (𝐶𝑃𝑄
1𝑡 −𝐶𝑃𝑄2𝑡 )×(1+𝑒)𝑡𝑛
𝑡=0
((1+𝑟)(1+𝑖))𝑡 (6.11)
where 𝐶𝑃𝑄1𝑡 and 𝐶𝑃𝑄
2𝑡 denotes the cost of PQ phenomena without and with mitigation at the
beginning of the time period respectively. To avoid confusion, all cost variables in (6.9)-(6.11)
are expressed as positive values (£). It can be seen that the smaller 𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 is, less
investment cost is required. The financial benefit of placing the mitigation techniques,
denoted as 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡, can be calculated by (𝐶𝑃𝑄1𝑡 − 𝐶𝑃𝑄
2𝑡 ). In (6.9), negative sign is applied to
𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡 so that the optimisation procedure will attempt to maximise the benefit.
(𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 − 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡) < 0 suggests that the subsequent financial benefits resulting from
the application of mitigation techniques will cover the initial capital investment and
maintenance cost, and placing the selected mitigation scheme is beneficial in the long run. In
(6.9), 𝛽 is a Lagrange multiplier which imposes the penalty to the selected mitigation scheme
if the technical constraints are violated.
6.3 Simulation results
To present the impact of Lagrange multiplier 𝛽 in (6.9) on the final mitigation scheme,
𝛽 is set to 1E+8 and 1E+3 respectively in the simulation. The convergence curves of various
components in (6.9), including F, (𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 − 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡), 𝑃𝑄𝐺𝐼UBPI , 𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 , and 𝐶𝑃𝑄
which denotes the NPV of 𝐶𝑃𝑄2𝑡 , are presented in Figure 6.2. Without mitigation, only the
Lagrange component, i.e., the technical constraints of (UBPI-UBPITH), contributes to the
objective evaluation. At the leftmost points in Figure 6.2 (a), F=2.41E9 when 𝛽 = 1E8, and
F=2.41E4 when 𝛽 = 1E3. For both settings of 𝛽, the objective value F is reduced significantly
after one technique is applied to the grid. When the number of mitigation techniques>5, F
tends to converge steadily, though its improvement is not as significant as that when
mitigation techniques<5. When 10 mitigation techniques are applied, the objective value F
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
70/83
obtained with 𝛽 = 1E3 is smaller than that obtained when 𝛽 = 1E8 by 6.86E5. As seen from
(6.9), F is composed of (𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 − 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡) and 𝛽 × 𝑃𝑄𝐺𝐼UBPI. The two components are
presented in Figure 6.2 (b) and (c) respectively. Without any mitigation activity, both
𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 and 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡 are zero. Thus it can be seen from Figure 6.2 (b) that (𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 −
𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡) = 0 when the number of techniques applied is zero. Afterwards, the obtained
(𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 − 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡) is smaller than zero constantly, which suggests that the financial
benefits resulting from the application of mitigation techniques at the network level will
cover the initial capital investment and maintenance cost. In Figure 6.2 (b), for the whole
convergence curves except for the leftmost points, (𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 − 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡) obtained when
𝛽 = 1E3 is always slightly smaller than that obtained when 𝛽 = 1E8; however, in Figure 6.2
(c), 𝑃𝑄𝐺𝐼UBPI obtained with 𝛽 = 1E3 is always slightly larger than that obtained when
𝛽 = 1E8. It can be seen that when 𝛽 is set to a larger number, i.e., 𝛽 = 1E8 in this case, the
optimisation will favour the mitigation schemes which are able to meet the technical
constraints well especially at the early stage of the optimisation process, compared to the
case of setting 𝛽 to a small number. Otherwise, the financial cost is the main reference for
the optimisation algorithm to select the mitigation solution. The costs of PQ phenomena and
PQ mitigation are also presented in Figure 6.2 (d) and (e) respectively, and it can be seen that
𝐶𝑃𝑄 follows the same convergence trends as presented in Figure 6.2 (b). It can be concluded
that with a large 𝛽, the technical PQ requirements (minimizing the violation of set thresholds
for considered PQ phenomena) have more influence on the selection of the final solution
compared to the case of using small 𝛽 which places more influence on final cost of the
solution, i.e., payback period.
(a) F
(b) 𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛 − 𝐶𝑏𝑒𝑛𝑒𝑓𝑖𝑡 (c) 𝑃𝑄𝐺𝐼UBPI
0 5 10-0.5
0
0.5
1
1.5
2
2.5x 10
9
No. of techs
F
0 5 10-8
-6
-4
-2
0
x 107
No. of techs
F
5 10-7.5
-7
x 107
=1E8
5 10-7.5
-7
x 107
=1E3
0 5 10-8
-6
-4
-2
0x 10
7
No. of techs
Cm
itig
atio
n-C
be
ne
fit
=1E8
=1E3
0 5 10
0
10
20
No. of techs
PQ
GI U
BP
I
=1E8
=1E3
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
71/83
(d) 𝐶𝑃𝑄 (e) 𝐶𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑖𝑜𝑛
Figure 6.2: Convergence curves of various components against the number of mitigation techniques applied
To present the PQ performance of each bus in comparison to its threshold, UBPIs
evaluated at all buses together with the thresholds set based on customers’ requirements are
provided in Figure 6.3 (a). It can be seen that without mitigation, most buses violate the PQ
thresholds. With the application of the mitigation solution obtained with 𝛽 = 1E8, almost all
buses meet the PQ thresholds. If the mitigation solution obtained with 𝛽 = 1E3 is applied to
the grid, there is still a number of UBPIs larger than the PQ thresholds. To present the gap
between the received UBPIs and the thresholds, (UBPI-UBPITH) evaluated at all buses are also
given in Figure 6.3 (b). With 𝛽 = 1E8, only UBPI obtained at bus 36 is slightly larger than the
threshold, with UBPI-UBPITH=0.0027. However, if 𝛽 = 1E3, in total there are 64 buses slightly
violating the requirements.
(a) UBPI
(b) UBPI-UBPITH
Figure 6.3: PQ performance of various buses obtained with 10 techniques
To present the PQ performance visually, and for the convenience of comparing the
aggregated UBPI obtained with and without mitigation, the heatmaps of UBPIs obtained
without mitigation and with 10 devices (when 𝛽 = 1E8) are plotted in Figure 6.4 (a) and (b)
respectively. The critical area marked in red in Figure 6.4 (a) is exposed to severe PQ
disruption, and it is greatly improved by applying the optimal mitigation solution obtained
with the proposed mitigation methodology, as shown in Figure 6.4 (b).
0 5 100
5
10x 10
7
No. of techs
CP
Q
=1E8
=1E3
0 5 100
2
4
6
8x 10
6
No. of techs
Cm
itig
atio
n
=1E8
=1E3
0 50 100 150 2000
0.2
0.4
0.6
0.8
1
1.2
Bus Index
UB
PI
Without mitigation
Thresholds
=1E8
=1E3
0 50 100 150 200
0
0.02
0.04
0.06
0.08
0.1
Bus Index
UB
PI-
UB
PI T
H
=1E8
=1E3
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
72/83
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
H2 O2
Non-linear load (fixed)
Photovoltaic
Fuel Cell
Wind Turbine
Unbalance load (fixed)
Non-linear load (random)
(a) without mitigation
149
147
154
155
150 153 156
148 146
145
141
143
142
144
140
139
129
130 133
131
135
137 136
138
134
157
161
158
186
184
132
160
165
162
163
164
166
167
168169
170
180
181
182 183
185
187
188
189
190 191
192
193
194
197
198
200 199
201
202
203 204
205
206
207
208209
151
152
224
232
77
159
225
215
216
211
212210
213
214
217
218 171
219
220
172
173
174
175
176
177
178
265
195
196
L
24
46
17
28
16
14
12
26
21 19
23
223 22
18
15
54
52
53
230
50
75
228
74
229
20221
51 76
13
268
87
48 47
49222
43
42
41
40
39
38
37
269
231
78
79
80
81
85
88
290
82
83
84
291
91
92
93
95
94
96
97 98
101 99
100
102
103 104
105
106 107
108
109
110
111
112
113
114 115
116
121
122
123124
125
126
127
128
117
118
119 120
86
11
10
8
9
7
6
5
4
55
1 65
64
66 63
57
226 227
58
62
60
59
61
2
3
72
70
179
25
27
29
30
32
31 33
3435
36
7168
67 69
73
89
45
44
266267
289
238
288
269
287 285286
56
BA C D E H I J K
26426326033kV
11kV
H2 O2
Non-linear load (fixed)
Photovoltaic
Fuel Cell
Wind Turbine
Unbalance load (fixed)
Non-linear load (random) (b) with mitigation
Figure 6.4: Heatmaps of UBPIs obtained with the application of 10 techniques when 𝜷 = 𝟏𝐄𝟖
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
73/83
6.4 Economically incentivised solution to mitigate LV network
harmonics
Based on the network modelling method developed in section 3.5 and the mitigation
methods analysed in section 4.3, a framework to incentivise customers to use switched-mode
ac-dc converters to offer harmonic compensation is proposed. Compensation is favoured
over reduction of emissions because incentives are based on spatial factors. In order to allow
fair access to the grid, it would not be appropriate to penalise users who connect line
frequency diode bridge converter interfaced devices because of where they are in the
network.
The ability to provide and the costs of compensation are dependent on technology
development. Therefore it is proposed to develop economic incentives designed to help
stimulate this development. In order to do that, a fair value is calculated for the benefit
gained by the system operator from these solutions.
The costs of the filter considered in section 4.3.1 is estimated to be 935 € [65] per
phase. The value attributable to system losses is calculated using the following method. The
temporal variation of losses is based on standard domestic load curves, such as shown in
Figure 6.5 [53].
Figure 6.5: Representative load profile for January in Germany.
The proportion of harmonics in this load curve is assumed to be constant. The load
curves cover all months of the year and therefore an average annual loss can be calculated. In
addition, the annual monetary savings from the filters and from the harmonic reduction or
distributed compensation at each node can be calculated.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
74/83
The calculation assumes an energy price of 40 €/MWh [67]. Compensation of the 3rd,
5th and 7th harmonics are considered (these cause the large majority of voltage THD and
system losses). The results are normalised to a compensated current of 1 A in each phase. To
place this in context, a 100 W device with line frequency diode bridge converter, fed by a
pure 50 Hz sine wave, operating at full power, draws 2.6 A, 1.8 A and 0.9 A at the 3rd, 5th and
7th harmonics respectively. The results are plotted showing the effect on system losses
against distance from the main bus at which compensation is provided are shown in Figure
6.6. The effect on the worst case of voltage THD in the system (node R18) for compensation
at each node is shown in
Table 6.4.
Figure 6.6: Reduction of system losses by harmonic compensation of 1 A in each phase against the distance from the main bus of the node at which compensation takes place.
Table 6.4: Reduction of voltage THD in phase b at node R18 by compensation of 1 A of harmonic current in each phase for each node.
Harmonic Order
Compensated Node
R1 R11 R15 R16 R17 R18
3 0% 0.0156% 0.0169% 0.0358% 0.0605% 0.0554%
5 0% 0.0165% 0.0219% 0.0386% 0.0622% 0.0561%
7 0% 0.0065% 0.0085% 0.0151% 0.0246% 0.0225%
The value of the THD reduction from the compensation can then be calculated:
𝐶𝑜𝑚𝑝𝑒𝑛𝑠𝑎𝑡𝑖𝑜𝑛 𝑉𝑎𝑙𝑢𝑒 𝑝𝑒𝑟 𝑎𝑚𝑝 =𝐹𝑖𝑙𝑡𝑒𝑟 𝐶𝑜𝑠𝑡𝑠× 𝑇𝐻𝐷 𝑟𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑝𝑒𝑟 𝑎𝑚𝑝
𝑈𝑛𝑚𝑖𝑡𝑖𝑔𝑎𝑡𝑒𝑑 𝑇𝐻𝐷−𝑃𝑒𝑟𝑚𝑖𝑡𝑡𝑒𝑑 𝑇𝐻𝐷 (6.12)
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
75/83
From
Table 6.4, it can be seen that the most effective way to reduce THD at node R18 is by
compensation of the 5th harmonic at node R17. As an example, based on filter costs of 3
times 935 € (1 filter in each phase) and an unmitigated voltage THD at node R18 of 5.15%
(Table 3.9), the value of compensation at node R17 for the 5th harmonic is 1163 € per amp
(provided to all phases). This compensation would only be called upon when required but
would need to be made available at all times over the estimated lifetime of a filter.
For both voltage THD and loss reduction, it should be noted that these values are
based on a small initial level of compensation. Only a finite level of harmonic compensation is
required to mitigate voltage THD and loss reduction is not linear with current reduction. In
the event that more compensation became available from distributed energy resources than
required, market forces would ultimately dictate the value of incentives.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
76/83
7 Conclusions
The report presents the concept of provision of differentiated quality of electricity
supply based on customers’ requirements in distribution networks. To fulfil this concept, five
new gap indices are introduced to reflect the satisfaction of the received PQ performance
compared to the thresholds which are set based on customers’ requirement regarding the
performance of individual PQ phenomenon or the aggregated PQ performance.
A range of PQ mitigating solutions was investigated to insure cost-effective
management of PQ in the network. FACTS devices including SVC, STATCOM and DVR were
investigated for PQ mitigation. Besides, network/plant based mitigation techniques were also
tested as the potential solutions to the PQ problems at hand. The effectiveness of these
mitigation techniques were evaluated using the selected severity indices and validated in
295-bus GDS.
Using the new gap indices as objective functions, an optimisation-based mitigation
strategy was proposed to carry out the strategic placement of potential FACTS devices based
on the analysis of PQ performance and sensitivity analysis. In this methodology, greedy
algorithm is applied to search the optimal mitigation scheme in order to enable the provision
of differentiated PQ levels. The feasibility of the proposed mitigation methodology was
demonstrated in a large scale generic distribution network. The pros and cons of using the
proposed indices as the optimisation objective functions are analysed in the report.
This report also presents an optimisation-based PQ mitigation methodology which
accounts for the comprehensive analysis of the financial losses due to critical PQ phenomena
as a result of industrial process trips, equipment aging issues and power losses, etc. This
report introduces the methodologies of assessing the financial investment of various
mitigation techniques (including network-based and FACTS devices-based mitigation
solutions) during the entire life span of the deployed solution. The proposed mitigation
methodology facilitates the provision of differentiated levels of PQ supply as required by
customers in different zones, by integrating the technical requirements in the optimisation
process as constraints using the approach of Lagrangian relaxation. The proposed
methodology was tested in a large scale 295-bus generic distribution network taking into
account a number of uncertainty factors. The simulation results demonstrated that placing
the mitigation scheme obtained by the proposed methodology is beneficial in the long run, as
the financial benefits of applying the mitigation techniques are much larger than the initial
capital investment and maintenance cost of the PQ mitigation.
The developed modelling method for the LV network highlighted the need for
detailed analysis of cable performance and showed that accurate cable parameters are
essential in planning harmonic mitigation. Of the mitigation methods considered, it was
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
77/83
shown that harmonic compensation is an effective method and that meaningful economic
incentives are possible based on the value offered to the network.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
78/83
References
[1] J. Y. Chan, J. V. Milanović, and A. Delahunty, "Risk-based assessment of financial losses due to voltage sag," IEEE Trans. Power Del., vol. 26, pp. 492-500, 2011.
[2] R. S. Thallam, "Power quality indices based on voltage sag energy values," in Proc. Power Quality Conf. Expo., Chicago, IL, 2001.
[3] M. H. J. Bollen, D. D. Sabin, and R. S. Thallam, "Voltage-sag indices - recent developments in IEEE P1564 task force," in Proc. IEEE PES Int. Symp. Quality and Security of Electric Power Delivery Systems (CIGRE), 2003, pp. 34-41.
[4] R. S. Thallam and G. T. Heydt, "Power acceptability and voltage sag indices in the three phase sense," in Proc. IEEE Power Eng. Soc. Summer Meeting, Seattle, WA, 2000, pp. 905-910.
[5] M. H. J. Bollen and I. Y. H. Gu, Signal Processing of Power Quality Disturbances. New York, NY, USA: IEEE Press, 2006.
[6] A. Robert and E. D. Jaeger, "Final Rep. round table on power quality at the interface T&D CIRED 2003," May. 14, 2003.
[7] M. H. J. Bollen and D. D. Sabin, "International coordination for voltage sag indices," in Proc. IEEE PES Transmission and Distribution Conference and Exhibition, 2006, pp. 229-234.
[8] J. Y. Chan and J. V. Milanović, "Severity indices for assessment of equipment sensitivity to voltage sags and short interruptions," in Proc. IEEE Power Eng. Soc. Gneral Meeting, Tampa, FL, 2007, pp. 1-7.
[9] J. Y. Chan, J. V. Milanović, and A. Delahunty, "Generic failure-risk assessment of industrial processes due to voltage sags," IEEE Trans. Power Del., vol. 24, pp. 2405-2414, 2009.
[10] A. M. Dan, "Introducing a voltage dip severity index (a proposal)," in Proc. Int. Conf. on Harmonics Quality of Power, 2010, pp. 1-5.
[11] C. N. Lu and C. C. Shen, "A voltage sag index considering compatibility between equipment and supply," IEEE Trans. Power Del., vol. 22, pp. 996-1002, 2007.
[12] C. N. Lu and C. C. Shen, "Voltage sag immunity factor considering severity and duration," in Proc. IEEE Power Eng. Soc. Gneral Meeting, 2004, pp. 626-630
[13] C. N. Lu and C. C. Shen, "Estimation of sensitive equipment disruptions due to voltage sags," IEEE Trans. Power Del., vol. 22, pp. 1132-1137, 2007.
[14] F. B. Costa and J. Driesen, "Assessment of voltage sag indices based on scaling and wavelet coefficient energy analysis," IEEE Trans. Power Del., vol. 28, pp. 336-346, 2013.
[15] G. J. Wakileh, Power Systems Harmonics: Fundamentals, Analysis and Filter Design. New York: Springer, 2001.
[16] Spec. Semicon. Process, Equipment Voltage Sag Immunity, SEMI-F47-0706. Available: www. Semi.org.
[17] G. Tsagarakis, A. J. Collin, and A. E. Kiprakis, "Modelling the electrical loads of UK residential energy users," in Proc. 47th Int. Universities Power Eng. Conf., London, 2012, pp. 1-6.
[18] "Technology roadmap: Electric and plug-in hybrid electric vehicles," International Energy Agency, 2011.
[19] "Technology roadmap: solar photovoltaic energy," International Energy Agency, 2010. [20] N. C. Woolley and J. V. Milanović, "Statistical estimation of the source and level of voltage
unbalance in distribution networks," IEEE Trans. Power Del., vol. 27, pp. 1450-1460, 2012. [21] A. Robert and J. Marquet, "“Assessing Voltage Quality with relation to Harmonics, Flicker
and Unbalance,”" WG 36.05, Paper 36-203, CIGRE 92.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
79/83
[22] W. H. Kersting, "Causes and effects of unbalanced voltages serving an induction motor," in Rural Electric Power Conference, 2000, 2000, pp. B3/1-B3/8.
[23] J. Thomas, "Development of methodology for online provision of differentiated quality of electricity supply," MSc. Electrical Power Systems, Fac. of Eng. and Phys. Sciences, The University of Manchester, 2012.
[24] T. G. More, P. R. Asabe, and C. S., "Power quality issues and It’s mitigation techniques," Int. J. of Eng. Research and App., vol. 4, pp. 170-177, 2014.
[25] J. Y. Chan, "Framework for assessment of economic feasibility of voltage sag mitigation solutions," Ph.D., Ph.D. dissertation, Dep. Electr. And Electro. Eng., University of Manchester, Manchester, U.K., 2010.
[26] E. Cinieri and F. Muzi, "Lightning induced overvoltages, improvement in quality of service in MV distribution lines by addition of shield wires," IEEE Trans Power Del. , vol. 11, pp. 361-372, 1996.
[27] T. A. Short, Electric Power Distribution Handbook. New York: CRC Press, 2004. [28] W. A. Chisholm, "Outline of guide for application of transmission line surge arresters 42 to
765 kV," Electric Power Reserch Institue, 1012313, 2006. [29] J. Sanders and K. Newman, "Polymer arresters as an alternative to shield wire," Hubbell
Power Systems Inc., EU1413-H, 1992. [30] T. C. Sekar and B. J. Rabi, "A review and study of harmonic mitigation techniques," in Proc.
Int. Emer. Trends in Elec. Eng. and Ener. Manag., Chennai, 2012. [31] H. Masdi, N. Mariun, S. Mahmud, A. Mohamed, and S. Yusuf, "Design of a prototype D-
STATCOM for voltage sag mitigation," in Proc. National Power and Energy Conf., Kuala Lumpur, Malaysia, 2004, pp. 61-66.
[32] Y. Xiao, Y. H. Song, C. C. Liu, and Y. Z. Sun, "Available transfer capability enhancement using FACTS devices," IEEE Trans. Power Syst., vol. 18, pp. 305-312, 2003.
[33] E. Ghahremani and I. Kamwa, "Optimal placement of multiple-type FACTS devices to maximize power system loadability using a generic graphical user interface," IEEE Trans. Power Syst., vol. 28, pp. 764-778, 2013.
[34] H. Hatami, F. Shahnia, A. Pashaei, and S. H. Hosseini, "Investigation on D-STATCOM and DVR operation for voltage control in distribution networks with a new control strategy," in Proc. IEEE Lau. Power Tech., 2007, pp. 2207-2212.
[35] R. Asati and N. R. Kulkarni, "A review on the control strategies used for DSTATCOM and DVR," Int. J. of Elec., Electr. Comp. Eng., vol. 2, pp. 59-64, 2013.
[36] H. L. Liao, S. Abdelrahman, and J. V. Milanović, "Identification of weak areas of power network based on exposure to voltage sags — part I: development of sag severity index for single-event characterization," IEEE Trans. Power Del. (accepted), 2014.
[37] H. L. Liao, S. Abdelrahman, Y. Guo, and J. V. Milanović, "Identification of weak areas of power network based on exposure to voltage sags—Part II: assessment of network performance using sag severity index," IEEE Trans. Power Del. (accepted), 2014.
[38] L. G. Vargas and T. L. Saaty, Models, Methods, Concepts & Applications of the Analytic Hierarchy Process. New York: Springer Science and Business Media, 2012.
[39] R. Dugan, M. F. McGranaghan, S. Santoso, and H. W. Beaty, Electrical Power Systems Quality (2 Ed.). New York: McGraw-Hill, 2002.
[40] M. T. Aung and J. V. Milanović, "Stochastic prediction of voltage sags by considering the probability of the failure of the protection system," IEEE Trans Power Del., vol. 21, pp. 322-329, 2006.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
80/83
[41] Z. Liu and J. V. Milanovic, "Probabilistic estimation of propagation of unbalance in Distribution Network with asymmetrical loads," in Proc. 8th Medi. Conf. on Power Gen., Tran., Distri. and Ene. Conver., 2012, pp. 1-6.
[42] S. Abdelrahman, H. Liao, J. Yu, and J. V. Milanović, "Probabilistic assessment of the impact of distributed generation and non-linear load on harmonic propagation in power systems," in Proc. 18th Pow. Syst. Com. Conf. , Wroclaw, Poland, 2014.
[43] CM SAF JRC EUROPEAN COMMISSION. Photovoltaic Geographical Information System - Interactive Maps [Online]. Available: http://re.jrc.ec.europa.eu/pvgis/apps4/pvest.php.
[44] "The Renewable Energy Review," Committee on Climate Change, 2011. [45] G. Olguin and M. H. J. Bollen, "Stochastic assessment of unbalanced voltage dips in large
transmission systems," in Proc. IEEE Bologna Power Tech 2003, p. 8 pp. Vol.4. [46] A. A. Groppelli and E. Nikbakht, Finance: Barrons Educational Series Inc, 2000. [47] C. A. Canesin, L. C. O. d. Oliveira, J. B. Souza, D. D. O. d. Lima, and R. P. Buratti, "A time-
domain harmonic power-flow analysis in electrical energy distribution networks, using Norton models for non-linear loading," in Proc. IEEE 16th Int. Con. on Harmonics and Quality of Power, Bucharest, 2014, pp. 778-782.
[48] E. Thunberg and L. Soder, "A Norton approach to distribution network modeling for harmonic studies," IEEE Trans. Power Deli., vol. 14, pp. 272–277, 1999.
[49] IEC 61000-3-2:2002, "Limits for Harmonic Current Emissions (Equipment Input Current up to and including 16A Per Phase)," ed, 2003.
[50] J. Arrillaga and N. R. Watson, Power System Harmonics (2nd Edition): Wiley, 2003. [51] T. B. Wood, D. E. Macpherson, D. Banham-Hall, and S. J. Finney, "Ripple current
propagation in bipole HVDC cables and applications to DC grids," IEEE Trans. Power Deli., vol. 29, pp. 926–933, 2014.
[52] K. Strunz etc., "Benchmark systems for network integration of renewable and distributed energy resources, Cigre Task Force C 6," 2009.
[53] EWE Netz GmbH, "Lastprofile 2014, available: https://www.ewe-netz.de/strom/1988.php," 2014.
[54] Y. Zhao, C. Wang, and J. Yan, "Analysis on Topology of DVR Used in Low-voltage Distribution Grid," Int. J. of Eng. Sci. and Research Tech., vol. 2, pp. 3630-3633, 2013.
[55] M. S. Munir and Y. W. Li, "Residential Distribution System Harmonic Compensation Using PV Interfacing Inverter," IEEE Trans. Smart Grid, vol. 4, pp. 816-827, 2013.
[56] M. Abinaya, N. Senthilnathan, and M. Sabarimuthu, "Harmonic Compensation as Ancillary Service in PV Inverter Based Residential Distribution System," in Proc. Int. Con. on Circuit, Power and Comp. Tech., Nagercoil, 2014, pp. 490-495.
[57] H. Geng, L. V. Qing, and G. Yeng, "Harmonic mitigation performance of the power inverter based DG in a AC microgrid," in Proc. IEEE 23rd Int. Symp. on Ind. Electron., Istanbul, 2014, pp. 2379-2384.
[58] J. G. Lglesias etc., "Economic framework for power quality, JWG CIGRE-CIRED C4.107," 2011.
[59] N. E. M. Association, "NEMA Standards Publication MG 1-1998 Motors and Generators," ed, 2002.
[60] P. Caramia, G. Carpinelli, P. Verde, G. Mazzanti, A. Cavallini, and G. C. Montanari, "An approach to life estimation of electrical plant components in the presence of harmonic distortion," in Proc. 9th Int. Conf. on Harm. and Quality of Power, 2000, pp. 887-892 vol.3.
[61] G. C. Montanari and L. Simoni, "Aging phenomenology and modeling," IEEE Trans. Elec. Insul., vol. 28, pp. 755-776, 1993.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
81/83
[62] J. V. Milanović and Z. Yan, "Global minimization of financial losses due to voltage sags with FACTS based devices," IEEE Trans. on Power Del., vol. 25, pp. 298-306, 2010.
[63] Y. Zhang, "Techno-economic Assessment of Voltage Sag Performance and Mitigation," Ph.D., Ph.D. dissertation, Dep. Electr. And Electro. Eng., University of Manchester, Manchester, U.K., 2008.
[64] A. Cali, S. Conti, F. Santonoceto, and G. Tina, "Benefits assessment of fault current limiters in a refinery power plant: a case study," in Proc. Int. Conf. Power System Tech., 2000, pp. 1505-1510 vol.3.
[65] M. Ghiasi, V. Rashtchi, and S. H. Hoseini, "Optimum location and sizing of passive filters in distribution networks using genetic algorithm," in Proc. Int. Conf. on Emer. Tech., 2008, pp. 162-166.
[66] M. Mohan, "Amorphous-Core Transformer with Copper Winding Versus Aluminium Winding-A Comparative Study," ARPN Journal of Science and Technology, vol. 2, pp. 297-301, 2012.
[67] "Electricity Production and Spot-Prices in Germany 2014, available: https://www.ise.fraunhofer.de/de/downloads/pdf-files/data-nivc-/folien-electricity-spot-prices-and-production-data-in-germany-2014-engl.pdf," Fraunhofer Institute for Solar Energy System, 2014.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
82/83
Appendix A: List of publications based on this report
Journal papers
[D1] H.L. Liao, S. Abdelrahman, Y. Guo, and J.V. Milanovic, "Identification of Weak Areas of Power
Network Based on Exposure to Voltage Sags -Part I: Development of Sag Severity Index for Single-
Event Characterization," accepted for publication in the IEEE Trans. Power Del., DOI:
10.1109/TPWRD.2014.2362965
[D2] H.L. Liao, S. Abdelrahman, Y. Guo, and J.V. Milanovic, "Identification of Weak Areas of
Network Based on Exposure to Voltage Sags -Part II: Assessment of Network Performance Using
Sag Severity Index," accepted for publication in the IEEE Trans. Power Del., DOI:
10.1109/TPWRD.2014.2362957
[D3] Z.Liu and J.V.Milanović, “Probabilistic estimation of voltage unbalance in mv
distribution networks with unbalanced load”, IEEE Transactions on Power Delivery, Vol.
30, No 2, 2015, pp. 693 – 703
[D4] Huilian Liao, Zhixuan Liu, J. V. Milanović, and Nick C. Woolley, “Optimisation
Framework for Development of Cost-effective Monitoring in Distribution Networks“,
accepted for publication in the IET Generation, Transmission and Distribution, GTD-2015-
0757
Conference papers
[D5] S. Abdelrahman, H.L. Liao, J. Yu and J.V. Milanović, "Probabilistic assessment of the impact of
distributed generation and non-linear load on harmonic propagation in power systems”, in Proc.
18th Power Systems Computation Conference (PSCC), Wroclaw, Poland, 2014.
[D6] S. Abdelrahman, H.L. Liao and J.V. Milanović, “The Effect of Temporal and Spatial Variation of
Harmonic Sources on Annual Harmonic Performance of Distribution Networks”, in Proc. 5th IEEE
PES Innovative Smart Grid Technologies Europe (ISGT), Istanbul, Turkey, 2014.
[D7] H.L. Liao, S. Abdelrahman and J.V. Milanović, “Provision of Differentiated Voltage Sag
Performance Using FACTS Devices”, in Proc. 23rd International Conference and Exhibition on
Electricity Distribution (CIRED 2015), France, June 2015.
Deliverable 4.2
Methodology for optimising QoS mitigation infrastructure based on differentiated customer requirements
83/83
[D8] H.L. Liao and J.V. Milanović, “Comparative Analysis of Different Voltage Sag
Characterisation Indices”, in Proc. 23rd International Conference and Exhibition on Electricity
Distribution (CIRED 2015), France, June 2015.
[D9] H.L. Liao, J. Saif and J.V. Milanović, “Provision of Differentiated Power Quality Using
Network Based Mitigating Solutions”, in Proc. 23rd International Conference and Exhibition
on Electricity Distribution (CIRED 2015), France, June 2015.
[D10] S. Abdelrahman, H.L. Liao and J.V. Milanović “Characterisation of Power Quality
Performance at Network Buses Using Unified Power Quality Index", in Proc. 23rd
International Conference and Exhibition on Electricity Distribution (CIRED 2015), France, June
2015.
[D11] S. Abdelrahman, H. Liao, T. Guo, Y. Guo and J.V. Milanovic “Global Assessment of Power
Quality Performance of Networks using the Analytic Hierarchy Process Model”, POWERTECH,
Netherlands, June 2015.
Submitted Journal papers
[D12] J. V. Milanović, Sami Abdelrahman and Huilian Liao, “Compound Index for Power
Quality Evaluation and Benchmarking“, submitted to the IEEE Transactions on Power
Delivery, TPWRD-00141-2015, (3/2/15;8/5/15;1/9/15)
[D13] Huilian Liao and J. V. Milanović, “Analysis of Propagation of Voltage Unbalance in
Distribution Networks with Distributed Generation “, submitted to the IEEE Transactions
on Power Delivery, TPWRD-01272-2015, (08/09/15)
[D14] Huilian Liao and J. V. Milanović, “Merits of Different FACTS Devices for Mitigation
of Power Quality Phenomena“, submitted to the IEEE Transactions on Power Delivery,
TPWRD-01382-2015, (08/10/15)