Date post: | 08-Jan-2017 |
Category: |
Documents |
Upload: | narayana-prasad |
View: | 213 times |
Download: | 1 times |
Electrical Power and Energy Systems 62 (2014) 450–460
Contents lists available at ScienceDirect
Electrical Power and Energy Systems
journal homepage: www.elsevier .com/locate / i jepes
SCUC problem for solar/thermal power system addressing smart gridissues using FF algorithm
http://dx.doi.org/10.1016/j.ijepes.2014.04.0610142-0615/� 2014 Elsevier Ltd. All rights reserved.
⇑ Corresponding author. Tel.: +91 9095039194.E-mail addresses: [email protected] (K. Chandrasekaran),sishajpsimon@
nitt.edu (S.P. Simon), [email protected] (N.P. Padhy).
K. Chandrasekaran a,⇑, Sishaj P. Simon b, Narayana Prasad Padhy c
a Department of Electrical and Electronics Engineering, National Institute of Technology Puducherry, Karaikal, Indiab Department of Electrical and Electronics Engineering, National Institute of Technology, Tiruchirappalli, Tamilnadu, Indiac Department of Electrical and Electronics Engineering, Indian Institute of Technology, Roorkee, India
a r t i c l e i n f o
Article history:Received 27 April 2013Received in revised form 30 March 2014Accepted 30 April 2014
Keywords:Demand responseEnergy storageReserve paymentFirefly algorithmReliabilitySecurity-constrained unit commitment
a b s t r a c t
This paper critically reviews the reliability impacts of major smart grid resources such as solar energy anddemand response (DR). It is therefore important to develop a technique for the integration of solar andthermal generation system. High solar penetration can lead to high-risk level in the power system reli-ability. In order to maintain the system reliability, solar power dispatch is usually controlled and theenergy storage is considered for smoothing out the fluctuations. This smart grid environment gives cus-tomer the opportunity to declare their own energy requirements together with their desired reliabilitylevels (DRL) to independent system operator (ISO) through demand response provider (DRP). Based onthe proposed model, firstly, a security-constrained unit commitment (SCUC) problem for the solar/ther-mal power system incorporating DR is solved using binary real coded firefly (BRCFF) algorithm. Secondly,the different reserve settlement schemes in the deregulated frame work are discussed based on theirdemand and DRL of the DRP. The effectiveness of the proposed model is demonstrated on IEEE RTS 24bus test system by comparing its performance with other methods reported in the literature.
� 2014 Elsevier Ltd. All rights reserved.
1. Introduction
The unit commitment problem (UCP) has an important role indaily operation of power system. It determines the scheduling ofgenerating units in a utility for minimizing the operating costwhile satisfying the system and unit constraints. Recently, theintroduction of renewable energy resources into conventional util-ities creates new concerns for power engineers. As the amount ofrenewable power generation increases around the world, impactson power system reliability and cost has increased. Among therenewable resources, wind and solar energy converters have beenrecognized as the most promising means of electric power genera-tion in future. In 2011, the U.S. Department of the Interior (DOI)approved a land to serve a 150 MW Rice Solar Energy Project, aconcentrating solar power (CSP) plant planned for RiversideCounty, California [1]. Also DOI approved 22 major renewableenergy projects in the past 24 months, which includes 13 commer-cial-scale solar energy projects that will produce roughly 5 GW ofelectricity during its completion. Not only USA, other countries also
understood the importance, and started the integration of renew-able resources with conventional power utilities.
In addition, DSM offers provision for incentives to customersdesigned to induce lower electricity use at times of high marketprices or when the system reliability is jeopardized [2]. DR refersto actions taken by the ISO to respond to a shortage of supply fora short duration of time in the future. Spinning reserve (SR)obtained by DR through demand response provider (DRP) togetherwith the reserve generated by the online committed units helps insatisfying the system reliability in an efficient manner.
In this paper a new conceptual model is proposed to integratethe solar and energy storage battery with the thermal generationin the UCP with DR. Due to the stochastic nature noticed duringthe integration of solar power and energy storage, the systembecomes highly nonlinear. Therefore newly developed heuristicfirefly algorithm [4–10] is suitably implemented in the schedulingof generators.
2. Problem formulation
Fig. 1 clearly shows the information flow in the smart grid envi-ronment between ISO and market entities (GENCO, TRANSCO andDRP). Here, ISO receives generation bids from generating compa-nies (GENCOs). The operational strategy while scheduling power
Nomenclature
C(Pi,k) cost function ai þ bi � Pi;k þ ci � P2i;k cost co-efficient
of ith generator unitH total number of hours consideredIi,k status of unit i at kth hour. (i.e.) 1 for ON and 0 for
OFFEENSspec
k;d specified EENS for DISCO d at kth hourEENSk,d calculated EENS for DISCO d at kth hourFSP forecasted solar powerLoadk total system demand at kth hourLOLPspec
k;d specified LOLP for DISCO dth at kth hourLOLPk,d calculated LOLP for DISCO dth at kth hourLj load curtailed at DISCO d due to generator outage jPi,k generation power output of unit i at kth hourPi,max, Pi,min maximum and minimum power output of unit iPbf–bt power flow in branch connected at bus between bf–
btPbf–bt,max maximum power flow limit for branch connected at
bus between bf–btPLk system losses at kth hourPRj PRj is the probability of availability that corresponds
to jth stateRU(i), RD(i) ramp up and down rate limit of unit iSRk system spinning reserve in MW at kth hourCost individual string’s cost of generation
Ton(i), Toff(i) minimum ON and OFF time for unit iVMb,min, VMb,max minimum and maximum limit of voltage at bus
bVMb voltage at bus bXon(i, k) time duration for which unit i is ON at kth hourXoff(i, k) time duration for which unit i is OFF at kth hour
Indexn total number of states in capacity outage probability
tablei, g thermal and solar generating unit index (subscript)k time index (subscript)d DISCO index (subscript).c reliability class index (subscript)
ObservationDRLk,d desired reliability level of dth DISCO at kth hourDDk,d demand of dth DISCO at kth hourNDC number of DISCOs associated with a classNTR number of transformersDDRk,d DR contribution of DISCO d at kth hour.LC total number of Load curtailment due to generator
contingency j
K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460 451
also checks the availability of transmission lines from TRANSCO ateach hour. ISO receives load demand and desired reliability level(DRL) from DISCOs through DRP. The maximum reliability levelat each DISCO includes the unavailability of thermal generatorsand uncertainties of forecast solar power and solar thermal gener-ator (STG). The ISO is responsible for balancing generation anddemand. It also has the authority to commit resources and curtailloads for maintaining the system reliability. The reliability level iscalculated using loss of load method [11] by the ISO and this infor-mation is sent to individual DRP. In this market model, the maxi-mum reliability level is known to a DISCO through theirrespective DRP. Also this market model gives DISCO an opportunityto declare its own energy requirements and desired reliability lev-els to the ISO. Therefore on solving SCUC problem, ISO has theresponsibility to maintain the reliability level for each dispatchperiod. During energy deficiency or violation of system reliabilityat each hour, the DR at each DISCO is evaluated by the ISO basedon its specified reliability level. Then the ISO sends the voluntarydemand reduction information or mandatory command informa-tion to the DISCO through the DRP. Thereby, the DISCO will partic-ipate by reducing certain percentage of load based on the contractwith the ISO.
The payment for SR is achieved based on the market philosophypresent in different countries. The electricity market such as Aus-tria, Netherland and Singapore justifies in forcing the GENCOs forthe payment of the reserve cost [12–14]. The electricity marketin UK and Switzerland recovers this cost from both GENCOs andDISCOs [15]. Recently, Swiss grid has imposed this cost indirectlyto the electricity consumers [16]. Overall, it is understood that allpower markets in the world directly or indirectly shift the burdenof the reserve to those who are benefiting from it [17]. This paperdiscusses all the above mentioned market strategies by consideringthree scenarios.
� Scenario 1: GENCOs are responsible for the reserve paymentand DISCOs have no opportunity to choose their DRL (i.e. DRLis set by the ISO).
� Scenario 2: Both GENCOs and DISCOs are responsible for thereserve payment and DISCOs have an opportunity to declaretheir reliability level through their respective DRP.� Scenario 3: DISCOs are responsible for the reserve payment and
DISCOs have an opportunity to declare their reliability levelthrough their respective DRP.
2.1. Objective function
In deregulated market, ISO will solve the SCUC problem and hasthe responsibility to maintain the reliability level for each dispatchperiod. The objective function of SCUC problem is formulated asminimization of the total cost (TC) without compromising the sys-tem security-constraints and is given in (1).
min TC ¼XH
k¼1
XN
i¼1
CðPi;kÞ � Ii;k þ SCi;k
� �þXH
k¼1
XN
i¼1
Ri;k � rgr� �
þXND
d¼1
rdrr � DRd½ � ð1Þ
The first term of the objective function is the sum of the fuel costand the start up cost (SC) of a generating unit in the schedulinghorizon. Fuel cost of a thermal generating unit is expressed as a sec-ond order function of each of the unit’s output as below:
CðPi;kÞ ¼ ai þ bi � Pi;k þ ci � P2i;k ð2Þ
The start-up cost of a generating unit depends on the period of timeduring which the unit is previously off and whether the boilers arekept hot during the shut-down period. Thus, the start-up cost canbe defined by Eq. (3).
SCi;k ¼hci when Toff
i 6 Xoff ði; kÞ 6 Toffi þ cshi
cci when Xoff ði; kÞ > Toffi þ cshi
( )ð3Þ
The second term is the reserve cost where each of the GENCOs offerto provide Ri,k spinning reserve at the rate of rgr $/MW. The lastterm is the DR reserve cost where each of the DISCOs offer to
Solar/Thermalgenerating companies
(GENCOs)
Transmissioncompanies
(TRANSCOs)
Generationbids submittedby GENCOs
Transmissionline information submitted
by TRANSCOs
DRInformation
ISO
Federal energy regulatorycommission (FERC)
Reliabilityregulations
SCUC solutionusing FF
algorithm
DRP1 DRP2 DRPd..........
..........DISCO1 DISCO2 DISCOd
Distribution Companies (DISCOs)
Demand and DRL bids submittedby DISCOs
Fig. 1. Information flow diagram.
ISO receives generation bids,transmission availability andDRL demand fromGENCO,
TRANSCO and DISCO,respectively.
ISO solves SCUC problemand clears energy, reserve(SR) and DR based on the
load demand and DRL usingSI techniques.
SR shared by each DISCO isevaluated using the steps givenin section 5. Then the reservecost is evaluated using second
term of equation (1)
Once the DR is evaluated, then theDR contribution at each DISCO isevaluated using the steps given in
the section 4. Then the DR cost isevaluated using the third term of
equation (1).
Operating cost iscalculated using the
first term ofequation (2-3).
Fig. 2. Steps involved in the SCUC problem.
452 K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460
provide DRd MW at the rate of rdrr $/MW [3]. The steps involved inclearing the energy, reserve and DR offer at each hour using theproposed method is shown in Fig. 2.
The objective function given in Eq. (1) is a minimization func-tion subject to the following constraints.
2.2. Problem constraints
2.2.1. System constraints
1. Power balance constraints
XN
i¼1
ðPi;k � Ii;kÞ þ PinvðkÞ ¼ Loadk þ PLk; k 2 ½1;H� ð4Þ
Here the inverter is assumed to be an ideal one and therefore lossesfor the inverter is not considered. The output power limit of theinverter for the Eq. (5) is given below.
0 6 PinvðkÞ 6 Pmaxinv ð5Þ
In order to minimize the total cost (1), the output power of the bat-tery Pinv should be determined stochastically. Therefore the optimalbattery output power during the integration of the system becomeshighly nonlinear. Hence Pinv has to be considered as another vari-able in addition with N generator variables while solving SCUCusing FF algorithm.
2. Spinning reserve constraint
XN
i¼1
ðPi;max � Ii;kÞP Loadk þ PLk þ SRk; k 2 ½1;H� ð6Þ
2.2.2. Unit constraints
1. Generation capacity limit
Pi;min 6 Pi;k 6 Pi;max ð7Þ
2. Unit minimum ON/OFF durations
Xonði; kÞ � TonðiÞ½ � � Ii;k�1 � Ii;k
� �P 0 ð8Þ
Xoff ði; k� 1Þ � Toff ðiÞh i
� Ii;k�1 � Ii;k
� �P 0 ð9Þ
3. Unit ramp constraints
Pi;k � Pi;k�1 6 RUðiÞ ð10Þ
Pi;k�1 � Pi;k 6 RDðiÞ ð11Þ
K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460 453
2.2.3. Network security constraints
VMp;min 6 VM 6 VMp;max ð12Þ
jPbf —btj � Pbf�bt;max bf —bt 2 NB ð13Þ
2.2.4. Reliability constraintsSpinning reserve requirements in Eq. (6) can be calculated using
either deterministic criteria or probabilistic techniques. However,in the proposed reliability constrained (probabilistic) method thespinning reserve requirement is assessed according to the desiredlevel of reliability. Therefore spinning reserve requirement shouldsatisfy either one of the reliability constraints given as follows:
LOLPk;d 6 LOLPspeck;d ; d 2 ½1;ND� ð14Þ
EENSk;d 6 EENSspeck;d ; d 2 ½1;ND� ð15Þ
2.2.5. Solar power limit and energy storage operating constraintsIn this paper, the solar energy dispatch is restricted to a fixed
percentage of the system load considering system operating con-straints. The solar power dispatch is controlled depending on theability of the existing conventional units in the system to respondto the variation of the solar power. Also energy storage facilitieshave a significant positive effect on the reliability of this small iso-lated solar power generation system. Let the hourly solar energydispatch is restricted to X % of the hourly system load demand, inorder to maintain system reliability. Then the surplus solar energyabove X % of the system load is stored in the battery during charg-ing operation [18]. The charging/discharging operation in the bat-tery is subject to the following constraints.
a. Initial storage energy limit
The energy storage starts charging from any initial value of anelectric energy.
Eini ¼ w � Emin ð16Þ
b. Charge/discharge constraints
EcðkÞ ¼Z
a � PinvðkÞ � dk for charge ð17Þ
EdðkÞ ¼Z
PinvðkÞ � dk for discharge ð18Þ
Total efficiency of the battery is integrated into a, which is the com-bined efficiency of energy storage system. Where Ec(k) and Ed(k) arethe charge and discharge energy in the battery. The discrete repre-sentation for storage energy at hour k, E(k) is given as
EðkÞ ¼ Eðk� 1Þ þ EcðkÞ � EdðkÞ ð19Þ
The constraints for charge/discharge operation of the battery aregiven below.
Emin 6 EðkÞ 6 Emax for all k ð20Þ
w � Emin 6 EðkÞ 6 Emax at the end of discharge ð21Þ
Emin 6 EðkÞ 6 b � Emax at the end of charge ð22Þ
The charge operation ends when E(k) increase to a limit up tob � Emax, whereas discharge operation ends when E(k) decrease toa limit up to w � Emin. In order to ensure that sufficient reservepower is available to prevent unexpected outages, the minimumcharge stored in the battery is considered as w � Emin � b and w arecharging and discharging constant.
3. Reliability analysis of power system
Reliability analysis of the power system represents a thermalgenerator and STG as a two state model [11], according to whicha unit is either available or unavailable for generation. The reliabil-ity indices for each hour are calculated by using Eqs. (23) and (24).
LOLPk;d ¼Xj¼LCd
PRj ð23Þ
EENSk;d ¼Xj¼LCd
PRjLj MW h ð24Þ
3.1. Solar power forecast uncertainty
In the proposed method, the solar power forecast uncertainty isassumed to be normally distributed. The probability distributionfunction of the demand can be described by a four-step model(0, � 1r, � 2r, � 3r), where the standard deviation r is a percent-age (usually 5%) of the expected solar power (distribution mean).Such a model will encompass more than 99% of load uncertainty.Based on the above analysis, the LOLP and EENS indices for eachhour incorporating the solar power forecast uncertainty is calcu-lated by the following equations:
LOLPk;d ¼X4
l¼1
LOLPk;dðlÞ � PLðlÞ; k 2 ½1;H� ð25Þ
EENSk;d ¼X4
l¼1
EENSk;dðlÞ � PLðlÞ; k 2 ½1;H� ð26Þ
where PL(l) – Probability of load step l, LOLPk,d(l) – LOLP for the loadstep l of hour k, EENSk,d(l) – EENS for the load step l of hour k.
4. Evaluation DR shared by each DISCO
The most familiar and simplest way of allocation of DR is basedon DISCOs load. In the above method the desired reliability level(DRL) of the DISCO is not considered. Hence to overcome this drawback, the DR based on the DRL is proposed to motivate the custom-ers to manage and control their consumption. The contribution ofDR made by each of the DISCOs is based on the DRLd,k at kth hourwhich is calculated by the following Eq. (27).
DDRd;k ¼ LPRCd;k � TDRk ð27Þ
where TDR is the total demand response at kth hour, DDRd,k is theDR contribution of DISCO d at kth hour and LPRCd,k is the load pointreliability coefficient of DISCO d at kth hour which is calculatedusing the Eq. (28)
LPRCd;k ¼DRLd;kPNDd¼1DRLd;k
ð28Þ
4.1. Step by step demand response algorithm
Demand response at each DISCO is based on DRLd,k at kth hour.
However, it should be noted that if there is a higher reliabilitylevel, then the reliability index value will have a very small value.
Step 1: Determine TDRk (i.e. Initially TDRk is 5% of the load at kthhour).Step 2: Set DISCO count d = 1.Step 3: Calculate LPRCd,k using Eq. (28).Step 4: Calculate DDRd,k using Eq. (27).Step 5: Do demand response at the dth DISCO (i.e. DDd,k = DDd,k
� DDRd,k), where, DDd,k is the demand of dth DISCO at kth hour.
454 K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460
Step 6: Once the reliability of the DISCO is satisfied, then DR atthe DISCO d is stopped (i.e. DRLd,k = 0).Step 7: Do d = d + 1.Step 8: If d 6 ND, then go to step 3. Otherwise, go to next step.Step 9: If sum(errord,k) = 0, then go to next step. Otherwise, setTDRk = sum(errord,k) and go to step 2Step 10: End.
5. Evaluation of reserve allocation
Several approaches are available for allocating the reserve thathas to be shared among various DISCOs. The most familiar and sim-plest way of allocation is based on the DISCOs load. In this method,ISO charges an amount for the reserve equally for both the indus-trial and residential customers that are having the same load anddifferent reliability level. Therefore, this method is not attractivefor residential customers as disparity is present. Hence to over-come this draw back, the methodology presented in [19] isextended to consider not only customers power demand, but alsotheir desired reliability level in the reserve allocation. The stepby step procedure to evaluate the DISCO’s reserve contribution atkth hour is given below.
Step 1: Based on the DRL, all DISCOs are classified into NC clas-ses and are arranged in a descending order with their corre-sponding load level.
DRLk;c > DRLk;cþ1; c 2 ½1;NC�; NC 6 ND ð29Þ
where NC is the number of classes of DRL and ND is the totalnumber of DISCOs available in the isolated power system. Toillustrate, let us consider a hypothetical system that consistsof five DISCOs with different DRL. Then the number of classesof DRL is five. Similarly, if the system consists of five DISCOs,where three DISCOs having different DRL and two DISCOs hav-ing same DRL, then the total classes of DRL is four. Therefore iftwo or more DISCOs are present in a class, then the load levelfor that class will be the sum of their individual load level ofthe DISCOs. If ND is the total number of available DISCOs, thenthe total system reserve shared by them is based on the DRL andtheir corresponding load level.Step 2: Once NC classes are arranged in a descending order, thereserve required to be met by each of the NC classes (RRC) arecalculated by the following equation [19].
RRCk;c ¼ SRk �Yc�1
d¼1
1� ORLCk;d
DRLk;d
� �� ORLCk;c
DRLk;cð30Þ
where ORLCk,c in the above equation is calculated using theequation (31). ORLCk,c is the overall system reliability level.
ORLCk;c ¼PNC
d¼cDRLk;d � DDk;dPNCd¼cDDk;d
ð31Þ
Step 3: Once RRC is computed, the reserve contribution forthat class (RCC) is calculated using the following equation [19].
RCCk;c ¼ RCCk;c�1 þRRCk;c
NC � ðc � 1Þ ð32Þ
Step 4: Next the reserve contribution for each DISCO (RCD) iscalculated using the following steps.
Step 4.1: Set class count c = 1 and DISCO count d = 1.Step 4.2: If the class c consists of more than one DISCO, thengo to step 4.3. Otherwise, calculate the DISCO reserve contri-bution using the Eq. (33). Update the class count c = c + 1 andthe DISCO count d = d + 1 and then go to step 4.4.
RCDk;d ¼ RCCk;c ð33Þ
Step 4.3: If two or more DISCOs have same reliability value,then the reserve contribution of each DISCO (RCD) belongingto the same class is evaluated using the following steps.
Step 4.3a: Set count dc = 1Step 4.3b:
RCDk;d ¼ RCCk;c �DDk;dcPNDC
dc¼1DDk;dc
ð34Þ
where NDC is the total number of DISCOs associated withclass c.
Step 4.3c: Update count d = d + 1 and dc = dc + 1. Ifdc 6 NDC, then go to step 4.3b. Here NDC is the totalnumber of DISCOs available in the cth class. Otherwise,update the class count c = c + 1 and then go to next step.
Step 4.4: If c 6 NC, then go to step 4.2. Otherwise go to nextstep.
Step 5: End.
6. Implementation of BRCFF for UCP
In UCP, binary numbers 0 and 1 are used to indicate the unitstatus (i.e., OFF or ON). The firefly algorithm used in [4–10] isessentially a real-coded algorithm. Hence the modification is madeas in Ref. [23] to enable it to deal with the binary-coded optimiza-tion problem.
6.1. Initial generation of binary string population
Randomly generate a population of M initial firefly position rep-resented by a binary string. Initialize randomly an initial positionM = [X1; X2; . . . ; Xm] of m solutions or firefly positions in themulti-dimensional solution space where m represents the size ofa population. Each solution of X is represented by the D-dimen-sional vector. Here D is equal to (N + 1) � H. i.e. plus ‘1’ representsinverter as another variable in FF.
6.2. Repair strategy in binary coded FF
Whenever the commitment status for each time interval is gen-erated randomly or if the firefly position is modified, violation ofminimum up/down time constraints (6–7) and spinning reserveconstraint (12–13) has to be checked as follows. The randomlygenerated commitment status for each time interval is checkedfor the violation of minimum up/down time constraints and reli-ability constraints given in Eqs. (12) and (13).
Step 1: If the reliability level (12–13) is met, then go to step 4.Otherwise, go to next step.Step 2: Add DR reserve in addition with the reserve generatedby the online committed unit to satisfy the reliability level. Ifthe reliability level (12–13) is not met, then set DR = 0 and thengo to next step. Otherwise, go to step 4.Step 3: The less expensive units which are in the OFF state isidentified and turned ON. Then go to step 1.Step 4: If the reliability constraint is satisfied, then the mini-mum up and down time constraints (6–7) are checked for eachunit over the scheduling horizon in each interval. If there is anyviolation in the minimum up or down time constraint then therepair mechanism is used to overcome the violation. Forinstance, let us assume that the Ton and Toff for a hypotheticalunit is 4 and 5. For a scheduling interval of 12 h, if the actualoff time for unit 1 is 3 h (5th–7th hour), then it violates the Toff
constraint. In this case, the unit status before 5th hour or after7th hour can be made 0. By doing this change, if it violates theTon constraint, then the status of the units are made 1 during theviolated down time period.
K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460 455
Step 5: The repair strategy in step 4 may affect the reliabilitylevel of the system. If the reliability level is met, accept the fea-sible solution. Otherwise go to step 1.Step 6: Once reliability and unit up/down time constraints aresatisfied, the optimal power flow (OPF)/economic dispatch(ED) problem is solved using real coded FF algorithm over thescheduling horizon in each interval. If any violations persistfor specified number of iteration, then the violation is mitigatedby committing additional units based on the network violationinformation from optimal power flow problem. If any additionalunit is turned on to satisfy the security constraints, then go tostep 4. Otherwise go to next step. A minimum number of trialsshould be set for the repair mechanism. These steps are carriedout for the entire hourly load.
6.3. Initial generation of real string population
Randomly generate a population of R initial solutions repre-sented by real values for unit status of pth population withS-dimensional vector. It should be noted that the real values willbe initialized only for the unit which has ON status obtained bybinary coded FF. Hence initialize randomly an initial populationR = [Y1; Y2; . . . ; Ym] of m solutions or firefly positions in themulti-dimensional solution space where m represents the size ofpopulation. Each solution of Y is represented by the S-dimensionalvector (i.e. ON status obtained from D-dimensional vector). Here Sis equal to (Non + Vm + Tt + Ta) � H. Non is the number of generatingunit in ON status, Vm is the number of generator voltage magni-tude, Tt and Ta is the number of transformer tap setting and angle,respectively.
6.4. Repair strategy in real coded FF
Whenever the firefly position is modified by real coded FF algo-rithm, violation of capacity limits of generating unit (5) and inver-ter capacity limit (3) has to be checked as follows.
Step 1: If the generated power (Pi) capacity limit (5) is met, thengo to step 3. Otherwise, go to next step.Step 2: If Pi > Pmax, then Pi = Pmax. If Pi < Pmin, then Pi = Pmin.Step 3: If the energy storage (Pinv) capacity limit (3) is met, thengo to step 5. Otherwise, go to next step.Step 4: If Pinv > Pmax
inv , then Pinv ¼ Pmaxinv and go to next step.
Step 5: End.
6.5. Evaluation of fitness
Evaluate the fitness value of each firefly position correspondingto the brightness using (35).
FIT ¼ 1=Cost; if Cost > 0¼ 1þ absðCostÞ i Cost < 0
ð35Þ
The step-by-step procedure for the proposed method is given as aflowchart in Fig. 3.
7. Results and discussions
All the programs are developed using MATLAB 7.01. The systemconfiguration is Pentium IV processor with 3.2 GHz speed and 1 GBRAM. In our previous work UCP [23] and reliability constrainedUCP [24] is solved using binary coded FF algorithm. It is found that,the BRCFF algorithm is very efficient and accurate in obtaining nearglobal optima with high success rates for the given constrainedoptimization problem. This motivated to solve the SCUC Problemfor solar/thermal power system addressing smart grid issues using
FF Algorithm. In this section, SCUC problem is solved using BRCFFalgorithm and smart grid issues are addressed. Different case stud-ies are carried out on IEEE RTS 24 bus system.
Case 1:
SCUC problem is solved for thermal power systemsand security issues are addressed.Case 2:
To validate the benefit of the hybrid power system,new operational strategy is developed for SCUCproblem incorporating DR.7.1. Selection of control parameters for BRCFF Algorithm
The control parameters are selected based on the statistical testas in Refs. [23,24] and the best combination of parameters for bin-ary coded FF algorithm that provided the best results is c = 1,b0 = 0.8 and firefly population size is 100. Similarly for the realcoded FF algorithm, the final combination of parameter that pro-vided the best result is c = 0.9, b0 = 0.9, n = 1 and firefly populationsize is 100.
7.2. Case 1-SCUC for thermal power system
In this section, the SCUC problem is solved for the thermalpower system without solar power and DR. The generator reliabil-ity data is adapted from [20] and generation cost coefficient istaken from Ref. [21]. Load data is adapted from [22]. Here, SCUCproblem is solved using BRCFF algorithm. The operating costobtained by BRCFF algorithm for case 1 is 747389.08$. The powerflow solution at certain time periods are infeasible as power flowsat certain time transmission lines exceeds their thermal limits. Thetransmission line between 6 and 10 is mostly overloaded duringtime intervals between 8th and 15th hours. Hence the additionalunits P23–P26 are committed as per the required level to mitigatethe violation in the network. Also, all the bus voltages are main-tained within the limits. The reliability index at each hour is calcu-lated using the steps given in the Sections 4 and 5. Here, in case 1,in all the hour reliability level of the system is maintained asLOLPspec = 1.5%.
7.2.1. Comparison of solution qualityTo validate the convergence efficiency of the proposed BRCFF,
the Case 1 Problem is solved using standard GA and PSO algorithm.Here, it is to be noted that the initial random generated populationis taken same for all the three techniques (GA, PSO and BRCFF).Table 1 gives the comparison of the total operating cost obtainedby BRCFF algorithm with respect to GA and PSO. BRCFF algorithmfinds a solution which is lesser than GA and PSO and is found tobe promising. The convergence characteristics of minimum, aver-age values are shown in Fig. 4. It is to be noted that GA and PSOexhibit premature convergence compared to BRCFF. This showsthe superiority of the proposed BRCFF algorithm.
7.3. Case 2-SCUC for hybrid power system data
This case study investigates the benefit of hybrid power systemand DR management in the SCUC problem. In this section, the SCUCproblem is solved for the thermal power system by considering asolar power and DR. As in Ref. [24], the solar generation with100 identical 2 MW STG units with Forced Outage Rate of 0.02 isintegrated with the IEEE RTS 24 bus system. The solar power pen-etration in this case is 5% of the system load from 8 to 18 h of thedispatch period. To include the uncertainty of solar power forecast,the standard deviation of the solar power forecast uncertainty istaken as 5% and is used in Eq. (24). The inverter capacity of150 MW is used to improve the system reliability. Emin and Emax
Start
Input system data, set binary coded FF
parameters
Generate random initial population as in section VI A
where,m is the number of fireflies
Set iter=1
Evaluate the degree of attractiveness of each
firefly using the equation β(r)=β0e-γr2
Do repair scheme for constraint management as given
in section VI D
Calculate Fitness (FITnew) of the modified
firefly position
Memorize the best solution at each interval
Is
Iter ≤ itermax
Yesiter=iter+1
Calculate total cost
No
End
Generate random initial real coded population
R=[Y1,Y2...Ym] for the status of pth
population as in section VIC
Modify the firefly position using the equation
V'p=Vp+β(r)*(Vp-Vq)+α (rand-0.5)
Store best solution with corresponding generator
values Then do p=p+1
Set iter=1
Evaluate the degree of attractiveness of each
firefly using theeq. given below β(r)=β0e-γr2
Modify the firefly position using the equation
V'p=Vp+β(r)*(Vp-Vq)+α (rand-0.5)
Do repair scheme for constraint management as
given in section VI B
Calculate Fitness (FITnew ) of the modified
position
Memorize the best solution at each interval
Is
Iter ≤ itermax
iter=iter+1
EDP/OPF using Real coded FF algorithm
Yes
Output the best solution
of generated UC status
Start
No
Yes
No
Cal
ling
Pr
ogra
m
Is p≤ m
No
Fig. 3. Flowchart for the proposed methodology.
Table 1SCUC solution-Case 1.
Solutiontechnique
Minimum operating cost ($)
GA 748411.10PSO 748316.10BRCFF 747389.08
747000
748000
749000
750000
751000
752000
0 50 100 150 200 250 300Generation number
Ope
ratin
g co
st ($
)
BRCFF-min
GA-min
PSO-min
BRCFF-avg
GA-avg
PSO-avg
Fig. 4. Convergence graph-case 1.
456 K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460
of the energy storage battery are 20 MW h and 300 MW h respec-tively. The battery co-efficient a, b and w are set as 0.9, 1 and 1.5respectively.
The IEEE RTS 24-bus system network consists of 17 DISCOs and11 GENCOs (total 26 generating units are present) and 38 trans-mission lines of which 4 lines consist of tap setting transformer.The 17 DISCO of IEEE RTS 24 bus system is shown in Fig. 5. Trans-mission line data is adapted from Ref. [25]. Here, the UCP is solvedusing the proposed binary coded FF and OPF problem is solvedusing real coded FF. In this market model, 17 DISCOs have power
contract with ISO through its respective DRP. The contribution ofDR by each DISCO is 3% of their hourly load. Table 2 gives thedemanded DRL of the DISCOs at each hour of the schedulingperiod. Here, for simplicity, it is assumed that the DISCO demandsthe same DRL in all the hours for different load. The ISO has the
1 27
4
9
5
6
013
11 1224
1514
13
201916
1718 21
22
23
DISCO 1DISCO 2
DISCO 3
DISCO 4
DISCO 5
DISCO 6
8
DISCO 7
DISCO 8
DISCO 10DISCO 9
DISC
O 11DISCO 13
DISC
O 12
DISCO 14
DISCO 15 DISCO 16
DISCO 17
Fig. 5. IEEE RTS 24 bus system.
K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460 457
responsibility for balancing generation and demand thereby main-taining DRL while solving SCUC problem. The capacity cost of DR atDISCO is considered as 5 $/MW. The capacity cost offered for thespinning reserve is 30 $/MW.
7.3.1. Case 2-SCUC solutionThe SCUC solution for the thermal power system without solar
power and DR is given in Table 3. The forecast solar power for 8–18 h is given in Table 4. The total operating cost obtained by BRCFFalgorithm is 1075769.08 $ without considering the solar powerand DR. The transmission lines between bus 6 and bus 10 are gen-erally overloaded during time intervals between 8th and 15th
Table 2DISCOs load and their corresponding DRL.
Disco no. DRL, MW h
1 0.00262 0.01833 0.00844 0.00885 0.00566 0.00987 0.01158 0.00679 0.0118
10 0.008511 0.007312 0.006913 0.019614 0.008515 0.009116 0.005117 0.0075
hours. Hence the additional units P23–P26 are committed to miti-gate the constraint violation in the network. It should be noted thatall bus voltages are maintained within the limits. From Table 3, thetotal operational cost of SCUC with solar power and DR(1048447.79$) has been significantly reduced which shows theadvantage of smart grid environment. The detailed energy dispatchis given in Table 4. Figs. 6 and 7 show the charge/discharge powerof the energy storage and DR management for 24 h time horizon.The acceptance of DR offers and the penetration of solar power pre-vent a new unit to be committed between 1st–4th, 7th–18th,20th–21st and 24th respectively. Therefore in the hybrid powersystem, the fuel cost is reduced simultaneously maintaining therequired DRL at each of the DISCOs. Suppose, if an additional unitis committed, the cost saving is reduced significantly. This scenariohelps us to understand that the additional cost incurred in thehybrid system and DR management is less than the cost involvedin system without considering the DR and solar power. The reservecontribution of each DISCO from the total system reserve at eachhour is discussed in the next section (scenario 3).
7.3.2. Unit reserve settlement schemeTo demonstrate the unit reserve settlement scheme, for the sys-
tem load of 1700 MW (1st hour load), three different scenariosbased on different market surveys are discussed.
Scenario 1:In this scenario, DISCO has no opportunity to choose their
desired reliability level and ISO will set the DRL for the DISCO. AlsoGENCOs are responsible for the reserve payment in the power mar-ket. Here DISCOs’ maximum participation is 3% of their system loadin the DR program. The declared DRL by the ISO based on the
Table 3SCUC solution.
Without considering DRand solar power-Case 1
With considering DR andsolar power-Case 2
Operating cost, $ 747389.08 742978.29Generating unit
reserve cost, $328380.00 303750.00
DR cost, $ – 1719.50Total cost, $ 1075769.08 1048447.79Cost responsible
for GENCO, $1075769.08 744697.79
Cost responsiblefor DISCO, $
– 303750.00
Cost saving, $ 27321.29
Tabl
e4
Gen
erat
ion
sche
dule
-cas
e2-
Hyb
rid
pow
ersy
stem
.
Un
itn
o.H
our
12
34
56
78
910
1112
1314
1516
1718
1920
2122
2324
P140
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
040
0P2
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
400
P335
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
035
0P4
138.
114
5.8
135.
513
8.1
150.
815
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
5P5
132.
914
0.5
130.
413
2.9
145.
315
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
5P6
127.
813
5.3
125.
412
7.9
140
155
155
155
155
155
155
155
155
155
155
155
155
155
155
155
155
155
155
155
P712
3.1
130.
412
0.7
123.
113
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
515
5P8
15.2
15.2
15.2
15.2
15.2
21.0
534
.09
7676
7676
7676
7676
7676
7676
7676
7667
.536
20.1
31P9
15.2
15.2
15.2
15.2
15.2
23.4
241
.93
7676
7676
7676
7676
7676
7676
7676
7676
22.1
21P1
015
.215
.215
.215
.215
.219
.29
32.1
976
7676
7676
7676
7676
7676
7676
7676
65.2
6118
.388
P11
15.2
15.2
15.2
15.2
15.2
17.3
830
7676
7676
7676
7676
7676
7676
7676
7662
.343
16.5
P12
2525
2525
2525
2596
.19
100
100
100
100
100
100
100
100
100
100
96.6
710
010
090
.239
2525
P13
00
00
025
2589
.57
100
100
100
100
100
100
100
100
100
100
90.0
610
010
083
.521
2525
P14
00
00
00
2582
.73
100
100
100
100
100
100
100
100
100
99.4
683
.23
100
100
76.5
825
0P1
50
00
00
068
.95
68.9
579
.611
1.3
134.
410
810
888
.54
118.
112
7.9
88.3
368
.95
68.9
588
.24
111.
368
.95
68.9
50
P16
00
00
00
068
.95
68.9
592
.07
115.
288
.73
88.7
369
.31
98.9
108.
769
.168
.95
68.9
569
.01
92.0
668
.95
68.9
50
P17
00
00
00
00
68.9
572
.84
95.9
769
.569
.568
.95
79.6
789
.45
68.9
568
.95
68.9
568
.95
72.8
368
.95
00
P18–
P26
00
00
00
00
00
00
00
00
00
00
00
00
FSP,
MW
00
00
00
063
7010
014
015
519
519
017
514
060
400
00
00
0
Ope
rati
ng
cost
,$19
,809
20,0
6219
,582
19,7
0320
,293
21,9
7025
,638
33,3
8236
,116
37,1
4138
,782
36,9
0536
,905
35,9
7937
,625
38,3
1835
,969
35,5
0234
,933
35,9
6537
,141
34,5
5728
,915
21,7
85
SR,M
W41
6.5
414.
542
6.5
416.
541
7.4
418.
151
942
3.5
433.
542
8.8
428.
442
9.8
432.
842
9.2
434.
342
8.7
444.
643
4.7
433.
243
3.8
437.
843
3.8
514
417.
1
Tota
lco
st,$
7429
78.2
9
458 K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460
system capacity and peak load is given in Table 5 (scenario 1). Thetotal reserve required to satisfy the ISO declared reliability level is316.5 MW and GENCOs are responsible for the generated reservecost.
Scenario 2:Here ISO has given the opportunity to DISCO either to demand
the DRL or to accept the DRL declared by the ISO. If DISCOs’required reliability level is higher than the reliability level specifiedby the ISO, then a DISCO can declare their DRL to ISO and partici-pate in the reserve cost settlement. Let us suppose, ISO declarethe reliability level to the hypothetical DISCO as X MW h. But ifthe DISCO requires higher reliability level than the reliability leveldeclared by the ISO, then the DISCO declares their required reliabil-ity level as Y MW h (reliability level of Y > X) to the ISO.
On sharing of reserve payment, GENCO will pay the cost for thereserve that required to satisfy the reliability level up to X MW hand the remaining reserve from X to Y MW h will be paid by therespective DISCO. In scenario 2 (Table 5), DISCO15 demands a reli-ability level of 0.0081 (indicated as bold letters) which is higherthan the reliability level (0.0851) declared by the ISO. The totalreserve required to satisfy the reliability level is 416.5 MW. Thecontribution of reserve by DISCO15 is given in Table 5 (scenario2). Hence DISCO15 should pay the reserve cost for the reserve of100 MW (416.5–316.5 MW) as per the ISO norms.
Scenario 3:In scenario 3, DISCOs are responsible for the reserve and also
have the opportunity to declare the DRL to ISO. The declared DRLand corresponding calculated DR at each of the DISCOs are givenin Table 5 (scenario 3). The steps given in Section 5 are used to cal-culate the reserve contribution for each DISCO. The number of clas-ses (NC) of DRL in the system is 17. The descending order ofdifferent classes of DRL and corresponding reserve required tomeet the reliability level of each class (RRC) is given in Table 6. Itis to be noted that the reserve to be met by any class will be sharedby the next level of available classes in the descending order. Thatis, for the reserve required to be met by first class(c = 1) of DRL (i.e.0.0196) is 184.15 MW. This is the minimum reserve required bythe system. Hence the reserve value 184.15 MW should be sharedby the DISCOs present in all the classes. Similarly, the reserverequired for the second class (c = 2) of DRL (i.e. 0.0183) is87.201 MW. In this case the reserve should be shared by all theDISCOs except the DISCO belonging to the first class (i.e. 0.0196)of DRL. The total reserve required to satisfy the demanded DRL is416.5 MW. The reserve contributed by each of the DISCOs (RCD)is given in Table 5 (scenario 3). Hence the DISCOs should pay thereserve cost to the ISO as per the ISO norms.
Nowadays all the market in the world shifts the burden of thereserve to those who are benefiting from it. Most of the generatorsfail due to grid problems for which GENCOs are not responsible.
0
10
20
30
40
50
60
70
80
1 3 5 7 9 11 13 15 17 19 21 23Hour
Pow
er d
isch
arge
MW
0
50
100
150
200
250
300
Ener
gy c
harg
ing
in M
wh
Power discharge, MW
Energy charging, MWh
Fig. 6. Charge and discharge power of energy storage.
0102030405060708090
1 3 5 7 9 11 13 15 17 19 21 23Hour
Dem
and
resp
onse
(MW
)
Fig. 7. DR management at each hour.
Table 6DRL with corresponding reserve.
Class RRC Corresponding to theDRL MW � 102
Class RRC Corresponding to theDRL MW � 102
1 1.841538238453276 10 0.0000316514485742 0.872018930194102 11 0.0000074802907973 0.696914717179336 12 0.0000012539726214 0.221171579997558 13 0.0000002052859565 0.082895378255346 14 0.0000000355656156 0.020538619143620 15 0.0000000083354587 0.003909614167124 16 0.0000000017621778 0.000822860760052 17 0.0000000003952059 0.000149424793182
K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460 459
Moreover, if the GENCOs pay for the reserve cost, then they areforced to increase their offer price to compensate their probablelosses in the market. Suppose, if the DISCOs are responsible forthe reserve cost, they have to manage and control their own con-sumption and required reliability level. Since DISCOs manage con-sumption and reliability level by themselves, the market model ofscenario 3 is found to be more beneficial compared to scenario 1and 2.
Table 5Reserve share for different scenarios.
Disco no. Scenario 1 Scenario 2
DRL RCD, MW DRL
1 0.0786 – 0.07862 0.0843 – 0.08433 0.0844 – 0.08444 0.0848 – 0.08485 0.0786 – 0.07866 0.0848 – 0.08487 0.0875 – 0.08758 0.0827 – 0.08279 0.0878 – 0.0878
10 0.0845 – 0.084511 0.0833 – 0.083312 0.0829 – 0.082913 0.0856 – 0.085614 0.0845 – 0.084515 0.0851 – 0.008116 0.0811 – 0.081117 0.0835 – 0.0835
DR, MW 31 0
SR, MW 316.5 416.5
8. Conclusion
This paper has employed a conceptual deregulated marketmodel for a hybrid power system. BRCFF algorithm inspired bysocial behavior of fireflies, through attractiveness and brightness,is developed and applied to solve SCUC problem.
� A detailed study has been carried out to bring out the bene-fits of integrating the solar power with thermal power gener-ation system. Due to the intermittency and unpredictabilityof the solar power generation, an energy storage facility isincorporated in the proposed model. This helps in smoothingout the fluctuating nature of solar power (reliability impactdue to climate change) and improves the continuity of powersupply from a STG.
� Also this paper presents new spinning reserve model, whichgives DISCOs the opportunity to participate in the reservemarkets through DRP. In this model, energy and spinningreserve market are cleared simultaneously using BRCFF algo-rithm. An extended methodology is developed to determinethe reserve contribution of DISCO based on their DRL andpower demand. Three spinning reserve settlement schemesare discussed based on the market philosophies present indifferent countries.
� It is observed when DR is incorporated, high reliability andeconomy can be achieved.
The comparison of the results with other methods reported inthe literature shows the superiority of the proposed method andits potential for solving nonlinear optimization problems in a
Scenario 3
RCD, MW DRL RCD, MW
– 0.0026 23.363362851159305– 0.0183 16.282696186967705– 0.0084 23.363229689981047– 0.0088 23.352945163793553– 0.0056 23.363362723529931– 0.0098 23.146248057289533– 0.0115 22.508591301479175– 0.0067 23.363362445681325– 0.0118 20.928794301496616– 0.0085 23.361173771394075– 0.0073 23.363357450821823– 0.0069 23.363361556540951– 0.0196 10.832577873254566– 0.0085 23.362834046873871100 0.0091 23.317403216819702– 0.0051 23.363362811638769– 0.0075 23.363336551278142
0
416.5
460 K. Chandrasekaran et al. / Electrical Power and Energy Systems 62 (2014) 450–460
power system. The proposed methodology can be extended tohybrid power system with the inclusion of many renewable energyresources. For large power system, it is also obvious, that the com-putation time of BRCFF algorithm is higher. However, the abovelimitations can be overcome by recent parallel computing technol-ogies. Hence parallel computing technologies can also be used inthe smart grid environment with demand side management whichdeemed to be essential for future research.
References
[1] http://www.solarserver.com/solar-magazine/solar-news/current/2011.[2] Benefits of demand response in electricity markets and, recommendations for
achieving them U.S. Department of Energy; 2006.[3] Parvania Masood, Fotuhi-Firuzabad Mahmud. demand response scheduling by
stochastic SCUC. IEEE Trans Power Syst 2010;1(1):89–98.[4] Chandrasekaran K, Simon Sishaj P. Firefly algorithm for reliable/emission/
economic dispatch multi objective problem. Int Rev Electr Eng (IREE)2012;7(1):199–210.
[5] Sayadi Mohammad Kazem, Ramezanian Reza, Ghaffari-Nasab Nader. A discretefirefly meta-heuristic with local search for makespan minimization inpermutation flow shop scheduling problems. Int J Ind Eng Comput2010;1:1–10.
[6] Apostolopoulos Theofanis, Vlachos Aristidis. Application of the fireflyalgorithm for solving the economic emissions load dispatch problem.Hindawi Publ Corpor Int J Combin 2011:1–23. http://dx.doi.org/10.1155/2011/523806.
[7] Yang X-S. Nature-inspired metaheuristic algorithm. Luniver Press; 2008.[8] Yang XS. Firefly algorithm, stochastic test functions and design optimisation.
Int J Bio-Inspired Comput 2010;2(2):78–84.[9] Yang X-S. Firefly algorithms for multimodal optimization. Stoch Algor: Found
Appl SAGA 2009, Lect Notes Comput Sci 2009;5792:169–78.[10] Yang X-S et al. Firefly algorithm for solving non-convex economic dispatch
problems with valve loading effect. Appl Soft Comput J 2012. http://dx.doi.org/10.1016/j.asoc.2011.09.017.
[11] Billinton R, Allan RN. Reliability evaluation of power system. 2nded. USA: Plenum press; 1996.
[12] Strbac G, Kirschen DS. Who should pay for reserve. Electr J 2000;13(8):32–7.[13] Kirby B, Hirst E. Allocating the costs of contingency reserves. Electr J
2003;16(10):39–47.[14] Xia LM, Gooi HB, Bai J. A probabilistic reserve with zero-sum settlement
scheme. IEEE Trans Power Syst 2005;20:993–1000.[15] National Grid Operation Manuals. <http://www.nationalgrid.com/uk/>.[16] Switzerland’s National Grid Company. <http://swissgrid.ch/>.[17] Ancillary services unbundling electricity products – an emerging market.
Report EURO electric; 2004.[18] Hu P, Karki R, Billinton R. Reliability evaluation of generating systems
containing wind power and energy storage. IET Gener Transm Distrib2009;3(8):783–91.
[19] Ahmadi-Khatir A, Fotuhi-Firuzabad M. Customer choice of reliability inspinning reserve procurement and the cost allocation using well-beinganalysis. Electr Power Syst Res 2009;79:1431–40.
[20] Billinton Roy, Fotuhi-Firuzabad Mahmud. Generating system operating healthanalysis considering stand by units, interruptible load and postponableoutages. IEEE Trans Power Syst 1994;9(3):1618–25.
[21] Wang SJ, Shahidehpour SM, Kirschen DS, Mokhtari S, Irisarri GD. Short-termgeneration scheduling with transmission and environmental constraints usingan augmented Lagrangian relaxation. IEEE Trans Power Syst1995;10(3):1294–301.
[22] Simopoulos DN, Kavatza SD, Vournas CD. Reliability constrained unitcommitment using simulated annealing. IEEE Trans Power Syst2006;21(4):1699–706.
[23] Chandrasekaran K, Simon Sishaj P. Network and reliability constrained unitcommitment problem using binary real coded firefly algorithm. Electr PowerEnergy Syst 2012;43:921–32.
[24] Chandrasekaran K, Simon Sishaj P. Demand response scheduling in SCUCproblem for solar integrated thermal system using firefly algorithm. IETRenewable Power Generation, held at Edinburg, UK; 2011.
[25] A report prepared by the reliability test system task force of the application ofprobability methods sub committee. IEEE Trans Power Syst 1999; 14(3):1010–20.