IEEE TRANSACTIONS ON SMART GRID 1 Smart Distribution Grid: Optimal … · 2015-07-30 · This article has been accepted for inclusion in a future issue of this journal. Content is

This article has been accepted for inclusion in a future issue of this journal Content is final as presented with the exception of pagination

IEEE TRANSACTIONS ON SMART GRID 1

Smart Distribution Grid Optimal Day-AheadScheduling With Reconfigurable Topology

Sajjad Golshannavaz Saeed Afsharnia and Farrokh Aminifar Member IEEE

AbstractmdashThis paper proposes an optimal operational schedul-ing framework to be taken in use in the distribution managementsystem (DMS) as the heart of smart active distribution net-works (ADNs) The proposed algorithm targets to optimallycontrol active elements of the network distributed generations(DGs) and responsive loads (RLs) seeking to minimize the day-ahead total operation costs The technical constraints of thecomponents and the whole system are accommodated in theac power flow fashion As an innovative point the DMS effec-tively utilizes the hourly network reconfiguration capabilitiesbeing realized by the deployment of remotely controlled switches(RCSs) Accordingly besides the optimal schedule of active ele-ments the optimal topology of the network associated with eachhour of the scheduling time horizon is determined as well Theeffect of hourly reconfiguration on the capacity release of DGsand RLs is highlighted which could be envisaged as a new trendin the reserve scheduling problem Considering practical issuesthe maximum daily switching actions of RCSs as well as switch-ing costs are judicially included The optimization procedure isformulated as a mixed-integer nonlinear problem and tackledwith the genetic algorithm To validate the satisfactory perfor-mance of the proposed framework a 33-bus ADN is thoroughlyinterrogated

Index TermsmdashActive distribution network (ADN) distributedgeneration (DG) reconfiguration remotely controlled switch(RCS) responsive load (RL)

NOMENCLATURE

Indices and Sets

t T Index and set of time intervalsi j B Nbus Indices set and total number of busesf F Nbr Index set and total number of feederss S Index and set of substationsg G Gi Index and set of DGs and set of DGs at bus il L Index and set of RLsk Index of RCSs

Parameters

ρDA Day-ahead wholesale electricity priceSU SD Start-up and shut-down cost of DGsCDG

p Cost function for DGs active power produc-tion

a b c Cost function coefficients of DGs

Manuscript received May 28 2014 accepted June 30 2014Paper no TSG-00515-2014

The authors are with the School of Electrical and Computer EngineeringCollege of Engineering University of Tehran Tehran 1439957131 Iran(e-mail sgolshannavazutacir safsharutacir faminifarutacir)

Color versions of one or more of the figures in this paper are availableonline at httpieeexploreieeeorg

Digital Object Identifier 101109TSG20142335815

ρRL Contract price of RLs participationkq

DA Coefficient for day-ahead reactive powerprice

kqRL Coefficient for RLs reactive power price

CSW Cost of each switching action for RCSskq

DG Coefficient for DGs reactive power costCDG

q Cost function for DGs reactive power pro-duction

PD QD Active and reactive power loads in each busY θ Magnitude and phase angle of feederrsquos

admittancePFRL Constant power factor for RLsSDA

max Substation maximum apparent power capac-ity

PDGmax PDG

min DGs maximum and minimum active powerlimits

QDGmax QDG

min DGs maximum and minimum reactive powerlimits

SDGmax DGs maximum apparent power capacity

PFDGmax PFDG

min DGs maximum and minimum power factorsPRL

max Maximum power reduction by RLs

Sfmax Feeder maximum apparent power flow

capacityVmax Vmin Maximum and minimum limits of bus volt-

ageNmL Number of main loops in the networkNSW

max Maximum number of daily switchingactions

Variables

PDA QDA Day-ahead active and reactive power purchasesPDG QDG DG active and reactive power dispatchesPRL QRL RL active and reactive power participationsW X Z Binary variables for DG commitment start-up

and shut-down statusI Binary variable denoting that the power factor

of a DG is beyond the mandatory regionV Bus voltageSf Apparent power flow of feeder fRCS Remotely controlled switch statusNSW

RCS Daily switching actions for each RCS

I INTRODUCTION

WORLDWIDE global warming ongoing environmen-tal pollutions and more potent energy crises have

motivated all nations to cater the utilization of renewable

1949-3053 ccopy 2014 IEEE Personal use is permitted but republicationredistribution requires IEEE permissionSee httpwwwieeeorgpublications_standardspublicationsrightsindexhtml for more information


2 IEEE TRANSACTIONS ON SMART GRID

energies instead of fossil fuels [1] In this way distributedgenerations (DGs) and storage devices are recent trends inresponding to the electrical energy requirements of distri-bution companies (discos) These technologies although areeconomicallyenvironmentally beneficial would pose a morecomplex operational situations for discos On the other handemergence of new mechanisms such as the open access tothe electricity markets and demand side programs are otherchallenging issues in the way of ensuring a sustainable energyfuture [2]

Although forgotten or maybe underemphasized before butnow there is a common sense that the sustainable energydevelopment necessitates more intelligent grid with onlinecontrol capabilities as well as bidirectional interactions withconsumers [3] Thanks to the substantial recent advances in theinformation and communication technology (ICT) along withthe rapid growth of field intelligent electronic devices (IEDs)the notion of Smart Grid has been evolved in power systemsspecifically at the distribution level Active distribution net-works (ADNs) as the leading sort of smart distribution gridsare characterized with controllable embedded generations suchas DGs demand side management programs such as respon-sive loads (RLs) and also remotely controlled switches (RCSs)to attain flexible network topologies Considering the topol-ogy changes modern bidirectional relays should be effectivelydeployed in all feeders to yield a robust protection schemein different operational horizons Distribution managementsystem (DMS) as the heart of economical and technicaloptimizations is responsible for optimal operational decisionmakings in ADNs utilizing optimal load flow algorithms [4]As well DMS is apt for monitoring the harmonic contentsof the network and alleviating the power quality concernsby optimally adjusting the set points of power electronics-based devices By this way the DMS optimally assigns thepower routings and performs an effective voltage control inthe network while keeping all constraints satisfied

In this context Algarni and Bhattacharya [5] have proposeda day-ahead operational scheduling framework where DGsare judiciously treated by including their relevant goodnessfactors Results revealed remarkable reduction in operationcosts due to more decrease in power losses Niknam et al [6]have proposed a multiobjective strategy minimizing the powerlosses and operation costs for the day-ahead optimal opera-tion of a distribution network including fuel cell (FC) powerplants Cecati et al [7] have devised an efficient DMS algo-rithm for day-ahead operation scheduling of smart distributionnetworks wherein an optimal power flow finely controls theelements such as wind turbines (WTs) and DGs Meanwhilesome recent studies have proposed a two-stage frameworkfor scheduling of ADNs [8] The first stage performs aday-ahead scheduling of resources and the effect of real-time data on optimal operation strategies are subsequentlyexplored in the second stage Although demonstrating notice-able achievements some of the marvelous merits providedby ADNs such as integration of RLs and RCSs are disre-garded in these works More recently Khodayar et al [9]have benefited from the application of motorized switches inoptimal operation scheduling of microgrid of Illinois Institute

of Technology campus to increase the reliability of the sys-tem Wang and Cheng [10] and Rao et al [11]ndash[12] havetailored the reconfiguration procedure considering the pres-ence of DGs in the network In these studies the impact ofconcurrent reconfiguration and DG sizing problem on the peakload power loss reduction is well-assessed It is however sen-sible that the possibility of daily reconfiguration could bringfurther monetary savings in the utility operation cost

Based on the foregoing discussions this paper presents anintelligent DMS incorporating supervisory control and dataacquisition (SCADA) system to tackle the optimal short-termoperational scheduling of ADNs The proposed DMS under-takes the optimal control of available active elements byminimizing the total operation costs including cost of pur-chasing electricity from the wholesale market cost of DGdispatches cost of RL participations cost of reactive sup-port as well as cost of RCSs switching actions In contrastto constant pricing for reactive power ancillary services herea practical pricing model is deployed for DG reactive powerparticipations Furthermore the automatic motorized switchesnamely RCSs and available tie-lines are wisely scheduled tosettle the optimal hourly network topologies In order to estab-lish a more practical technique the maximum number of RCSswitching actions is also limited The obtained technical andeconomical improvements are thoroughly explored to assessthe effect of applying RCSs in daily scheduling of ADNThe effect of hourly reconfiguration on the capacity releaseof DGs and RLs is investigated likewise which puts forwardnew insight to the reserve scheduling problems

The rest of this paper is organized as follows Section IIaddresses the generic description of the envisaged ADNSection III formulates the proposed operational schedulingframework Section IV interrogates several test cases and pro-vides numerical results Eventually the paper is concluded inSection V

II GENERIC DESCRIPTION OF THE ENVISAGED ACTIVE

DISTRIBUTION NETWORK

The advent of modern ICT such as wireless communica-tions by general packet radio service (GPRS) are offeringexcellent online controllability on different engineering appli-cations [13] Meanwhile ADNs benefitted from a compre-hensive ICT infrastructure are in the way of rapid evolutionADN refers to a self-organized system capable of controllingand enhancing the operation of all its interconnected ingredi-ents [14] Fig 1 demonstrates an ADN wherein the followinginformation is elaborated

1) Besides the main grid connection disco also encom-passes DGs WTs RLs and RCSs as active elements

2) DMS as a set of hardwaresoftware complements isentrusted to obtain the optimal operational decisionsregarding all network ingredients To do so it requiresgathering some preliminary system information andhenceforth collaborates with system operator to receivethe required data such as market prices one day inadvance [15] Also DMS may be equipped with mar-ket price forecasting mechanism specifically for havingeffective participation in energy and reserve markets


GOLSHANNAVAZ et al SMART DISTRIBUTION GRID 3

Fig 1 Envisaged ADN and an intelligent DMSSCADA framework

The day-ahead load pattern and natural resources fore-cast comprising wind speed and solar radiation areof great importance for DMS too High-performanceforecasting approaches are hence needed in essence

3) DSO should also be fully aware of the network sta-tus Thus different sorts of distribution remote terminalunits (DRTUs) as the field IEDs are installed in the ter-ritory of ADN Substation-RTU (SRTU) feeder RTU(FRTU) distributed generation RTU (DGRTU) respon-sive load RTU (LRTU) and wind turbine RTU (WRTU)are SCADA system ingredients for both telemetry andcontrol purposes FRTUs are capable of remotely alter-ing the open or closed status of RCSs to realize flexiblereconfigurations in ADN Also each RCS is capable ofsensing the fault currents and isolating the faulty sectionof the distribution feeder By deploying different sorts ofDRTUs and remote telemetry efficient estate estimationalgorithms (ESA) is also carried out as an indispen-sible feature of the online monitoring of the networkstatus [16]

4) DSO may also sign bilateral contracts with RLs orDGs to cooperate in incentive-based programs such asload reduction or generation curtailment to effectivelyreduce the peak loads and mitigate power mismatchesThe amounts and times of participation and the payingmechanism are specified in the contracts

5) Having communicated the required data DMS executesan optimal load flow program to determine apt set-pointsof control variables for all participants

III PROPOSED OPERATION SCHEDULING FRAMEWORK

By feeding the fundamental information for DMS DSOseeks to make the most optimal operational decisions whichwill finally result in reduction of its potential risks in volatilereal-time activities The proposed framework schedules theactive elements of ADN in a 24 h Day-ahead time horizonand deeply interrogates the effect of flexible structures Here

it is assumed that the forecasts of load and natural resourcesare acceptably accurate and no real-time effort is necessitatedfor compensating forecast errors

A Objective Function

DMS would endeavor to determine the unknown controlvariables including grid power purchases DG dispatches RLparticipations and RCS open or closed status Each branchis deemed to have RCSs which allow reaching to the bestoperating configuration Otherwise any specific switch can bereadily set to be must closed or must open Reconfiguration isthen suitably included in the optimization process by introduc-ing some mixed integer variables Various objective functionscould be realized by DMS however the most important oneis minimizing total day-ahead operation costs That is

Minimize F (x u)

=sum

tisinT

⎛

⎝sum

sisinS

ρDAt PDA

ts +sum

gisinG

XgtSUg

+sum

gisinG

CDGgpt +

sum

gisinG

ZgtSDg +sum

lisinL

ρRLt PRL

ts

+sum

gisinG

kDAq ρDA

t QDAts +

sum

gisinG

kRLq ρRL

t QRLtl +

sum

gisinG

CDGgqt

⎞

⎠

+sum

k

CSWNSWRCSk

(1)

In (1) associated with each time interval t vectors x and urespectively comprise control and dependent variables as

x =[PDA

s QDAs PDG

g QDGg PRL

l QRLl RCS

](2)

RCS = [RCS1 RCS2 RCSNbr

](3)

u =[|V| |Sf |

](4)

|V| = [|V1| |V2| ∣∣VNbus

∣∣] (5)∣∣Sf

∣∣ = [|S1| |S2| ∣∣SNbr

∣∣] (6)



The objective function (1) consists of different cost termsThe cost of purchasing active power from day-ahead wholesalemarket is represented by the first term The start-up cost forDGs is given in the second term DG production costs areincluded by the third term whose mathematical relation is

CDGgpt = Wgtag + bgPDG

gt + cg

(PDG

gt

)2 (7)

Continued in (1) the fourth term accounts for DG shut-down costs Herein it is assumed that DGs are utility-ownedand are centrally scheduled by DSO Though the presentedmathematical model could be slightly modified to consider theinvestor-owned DGs as well For this kind of DGs DMS isnot permitted to schedule the generation patterns However itconsiders the offered generation bids by the owners in the exe-cuting of the day-ahead scheduling process If any violationis seen the offered bids may be subjected to the curtailmentThe cost of discorsquos contracting with RLs in supporting activepower is accounted for by the fifth term The cost of reactivepower purchased from day-ahead market is given in the sixthterm Some literature such as [5] discussed that the reactivepower price provided via upstream network is approximatelybetween 5 and 10 of its active counterpart although thereactive power pricing is itself a complex issue in emergingelectricity markets which is beyond the scope of this paperThus this paper assumes kDA

q to be equal to 5 The sev-enth term similarly represents the reactive power cost posedby RLs participation In order to encourage RLs cooperationkRL

q is assumed to be 10 here Regulatory rules and gridcode requirements oblige DGs to provide mandatory reactivesupport within a power factor (PF) ranging from 095 lag-ging to 095 leading without any financial compensation [17]However if the network status forces DGs to operate atPFs out of the designated range monetary reimbursementsare necessary which is denoted by eighth term and derivedbased on (8) [18] This cost term is primarily activated forinvestor-owned DGs

Cgqt = kDGq Igt

[Cgpt

(SDG

gmax

)

minus Cgpt

(radic(SDG

gmax)2 minus (QDG

gt )2

)] (8)

kDGq is the profit rate of active generation taken between 5

and 10 Herein it is set at 10 to value the DGs cooper-ation in reactive power provisions Eventually the ninth termaccounts for the daily switching cost of RCSs

B Constraints

1) Load Flow Equations The active and reactive powerload flow equations are properly amended to contemplate theinfluential support of DGs and RLs in active and reactivepower generation

PDAts +

sum

gisinGi

PDGtg + PRL

ti minus PDti

minussum

f isinFampjisinB

Pftij

(Vi Vj Yij θij

) = 0 (9)

QDAts +

sum

gisinGi

QDGtg + QRL

ti minus QDti

minussum

f isinFampjisinB

Qftij

(Vi Vj Yij θij

) = 0 (10)

2) Grid Purchase Constraint This constraint ensures thatthe imported active and reactive power from wholesale marketto the disco is limited in the range of transformer capacities

[(PDA

ts

)2 +(

QDAts

)2]12

le SDAsmax foralls isin S (11)

3) DG Generation Limits These constraints are to scheduleDGs keeping their active and reactive powers within their ratedcapacities Meanwhile by applying DGs in adaptive PF modethey should also satisfy the permissible range for PF

PDGgmin le PDG

tg le PDGgmax forallg isin G (12)

QDGgmin le QDG

tg le QDGgmax forallg isin G (13)

[(PDG

tg

)2 +(

QDGtg

)2]12

le SDGgmax forallg isin G (14)

PFDGtg = PDG

tg(PDG

tg + QDGtg

)12forallg isin G (15)

PFDGgmin le PFDG

tg le PFDGgmax forallg isin G (16)

4) RL Constraints This constraint guarantees that theinvoked amount of RLs by DSO would be bounded in the pre-specified ranges appointed in the contract for hour t RLs areassumed to have a constant power factor as indicated in (18)

0 le PRLti le PRL

tlmax foralll isin L (17)

QRLti = tan

(cosminus1 (

PFRLi

)) times PRLti foralll isin L (18)

5) Flow Limits on Feeders The power carrying capacityin each distribution feeder is capped That is

[(Pf

tij

)2 +(

Qftij

)2] 1

2 le Sfmax forallf isin F (19)

6) Bus Voltage Limits The following constraints confirman acceptable voltage profile for all buses For substationbuses the voltage magnitude is kept constant and equal tothe unity

Vmin le ∣∣Vti∣∣ le Vmax foralli isin B (20)∣∣Vts

∣∣ = 1 pu foralls isin S (21)

7) Network Radialilty and Switching Limitation for RCSsIn the reconfiguration process of ADN through altering openand closed status of RCSs it is necessary to meet the followingconstraints

NmL = Nbr minus Nbus + 1 forallt isin T (22)

NSWRCSk

=sum

t

abs(RCSkt minus RCSktminus1

)(23)

NSWRCSk

le NSWmax (24)

The constraint (22) imposes a radial structure for the equiv-alent graph of the network maintaining total number of mainloops constant It means when one RCS is made closed in



Fig 2 Structure of the proposed chromosome

one specific loop exactly one RCS from that loop should bemade open to keep the network radial Different approachescould be established to avoid the loops structure in the recon-figuration process Depth first search (DFS) algorithm hasbeen effectively included to guarantee a radial structure forthe ADN [19] Also constraints (23) and (24) ensure that theswitching actions of each RCS will not violate the maximumnumber of possible switchings

C Optimization Technique

The proposed day-ahead operational scheduling frameworkis a mixed-integer nonlinear problem which is solved usingmixed-integer genetic algorithm (GA) GA as a global searchtechnique mimics the biological selection process in which themost qualified parents would survive more likely and dupli-cate themselves This process is recognized as the evolutionprocess performed by specific operators namely recombinationand mutation The iterative generation and evaluation processmakes GA apt to carefully scan the search space and then findthe optimal solutions [20]

1) Problem Codification The codification process refers toestablishing a candidate solution in the form of one chromo-some The set of unknown variables as the imagined genescomprise a chromosome in GA The proposed coding strategyfor optimal operation scheduling in each time interval is shownin Fig 2 wherein the proposed chromosome is composed offour strings The first and second strings represent DG activeand reactive powers respectively The third string relates tothe percent of RLs participation and finally the fourth binarystring determines open or closed status of each RCS in thereconfiguration process

2) Recombination and Mutation In this stage some ofthe chromosomes are randomly selected and from stochasti-cally selected crossover points (CPs) are combined together toproduce new offsprings Next some of genes in generated off-springs are mutated randomly to maintain the stochastic natureof reproduction course Fig 3 demonstrates the multipointrecombination and mutation process

IV CASE STUDIES AND NUMERICAL RESULTS

This section is dedicated to present several numerical studiesto assess the efficiency of the proposed day-ahead operationalscheduling framework

Fig 3 (a) Selected parents (b) Recombination process (c) Mutation process

TABLE ICOST COEFFICIENTS AND TECHNICAL DATA FOR DG UNITS

A System Specifications

Fig 4 demonstrates the modified single line diagram ofthe well-known IEEE 33-bus 1266 kV radial distributiontest system as the envisaged ADN The basic data regard-ing this system could be found in [21] As well this systemhas been determined as the test case in some similar stud-ies such as [22] and [23] In [24] an expansion planningapproach for the IEEE 33-bus system has been presentedwhich is able to accommodate an acceptable penetration ofDGs and WTs Hence the network has been supplemented byfour diesel-based DGs at nodes 8 13 16 and 25 whose char-acteristics are presented in Table I For improving DGs overallefficiencies they should be operated approximately over than25 of their ratings [25] Otherwise they will not be switchedon Furthermore the leading or lagging minimum PF for DGsis considered to be 085

Also three 3 MW WTs are included in the network atnodes 14 16 and 31 Five largest loads at nodes 8 14 24 30and 32 are considered as RLs with constant PF that could bedecreased by DSO up to 10 during hours 8 to 24 The pricefor 1 MW decrease by RLs is determined as $115MW asdeclared in the contract The network includes five maneu-ver tie-lines Each feeder branch is enriched with RCSs toprovide flexible topologies in optimal operation scheduling ofthe network by DMS Node 1 as the main substation is theonly connection point of disco to the wholesale market Alsoas it is observed different DRTUs are installed in ADN toyield an efficient online communication between DMS and



Fig 4 Single line diagram of the IEEE 33-bus ADN

Fig 5 Day-ahead wholesale electricity prices

all other active elements Based on an effective lifetime of15 years for each RCS [26] the maximum daily switchingoperations of each RCS is computed as six among which fouroperations are assumed to be devoted to the network reconfig-uration and two operations are allocated for the fault detectionand isolation duty maintenance outages or other tasks Alsothe initial capital investment and maintenance costs for eachRCS introduce $1 cost for each switching action [27] Theday-ahead scaled wholesale market prices and forecasted loadon July 16 2013 at NYISOs PJM have been utilized to assessthe proposed scheduling framework [28] Fig 5 demonstratesthe day-ahead electricity prices while the network total loadis illustrated in Fig 6 Wind speed forecasts and power gen-eration by WTs are respectively shown in Figs 7 and 8 Itis supposed that the three WTs are faced with similar geo-graphical conditions and hence the wind speed profile is takenanalogous

B Main Assumptions

Although partially stated before the most importantassumptions made in the test cases are as follows

1) By a fairly accurate forecast of wind speed and loadprofiles the prediction errors are considered negligible

Fig 6 Forecasted day-ahead load profile of the ADN

Fig 7 Forecasted hourly wind speed profile

2) WTs are evoked only to inject active power to thenetwork without reactive power support capabilitiesAlso the operation costs of WTs are very small andhence it can be safely assumed to be zero

3) Modern bidirectional digital relays are evolved in thenetwork to ensure a safe protection scheme

C Test Cases and Comparative Discussions

Two different cases are devised to examine the efficiency ofthe proposed framework Afterwards comparative discussionsare made to highlight the promising improvements



Fig 8 Forecasted hourly power generation by WTs

Fig 9 DG1 optimal active power schedule

1) Optimal Operation Scheduling Excluding RCSs Thiscase is referred to as Case-I where DSO seeks to optimallyschedule the ADN without possibility of remotely modify-ing the network topology ie the network is not equippedwith RCSs In this case DMS determines the optimal gridpurchases DG dispatches and RL participations

2) Optimal Operation Scheduling Including RCSs The sec-ond case denoted by Case-II benefits from RCSs DSOstrongly strives to optimally schedule active elements ofthe network and simultaneously assigns optimal switchingsequences for the network topology In this case DMSwould determine optimal grid purchases DG dispatchesRL participations as well as optimal status for each RCS

3) Comparative Discussions and Promising ImprovementsThe performance of the proposed operation scheduling frame-work in optimal dispatch of DGs and grid purchases in thebasic structure of the ADN namely Case-I is demonstrated inFigs 9ndash13 It is seen that as there is no possibility of alteringthe network topology all DG units are made on specificallyat peak hours to cover the network demand requirements andto meet the technical constraints Although DG2 has a higherproduction cost rather than DG4 its settlement in a strate-gic part of the network leads to a more generation outputand impressive effect in the network performance DG1 withthe lowest production cost and the higher technical impactis the most dispatched unit throughout the day In off-peakhours namely between 4 and 7 as the load of the network isminimum and the grid purchase price is lower than the DGproduction costs DMS does not switch DGs on and DSOpurchases all its requirements from the wholesale market

By applying RCSs in the optimal scheduling problemsome appealing observations are achieved as illustrated inFigs 9ndash13 and denoted by Case-II Remarkable changes areoccurred at hours with normal load where the wholesale mar-ket prices are comparable with DG production costs For thesehours including 1 to 3 8 to 12 and 20 to 24 reconfiguringthe network topology enables DSO to switch off the interior




Fig 13 Apparent power purchased from grid

expensive production units and purchase the required powerfrom the wholesale market with lower prices As Fig 13 showsassociated with these hours the imported apparent power fromthe substation has increased and consequently DGs are totallymade off This strategy has resulted in more economic savingsas will be discussed next However for the peak hours namely13 to 19 as the total DG units are switched on in both Cases-Iand -II there is no remarkable differences between them

Fig 14 depicts the share of grid purchases and DG dis-patches in accommodating the network load With respect tothis figure it is revealed that hourly modification of the net-work topology by RCSs not only decreases day-ahead opera-tion costs but also brings about a more 1208 free capacityfor DGs This released capacity would provide a good optionto overwhelm the uncertainties regarding wind speed and loadforecasts as well as to participate in the reserve markets



Fig 14 Daily overall optimal active power purchased from wholesale marketand dispatched by DGs

Fig 15 Total reactive power procurement by DGs

Fig 16 Worst bus voltage magnitudes in pu

Fig 15 presents similar considerations for the reactivepower procurement by DGs It can be observed that by activat-ing hourly reconfigurations the reactive power support by DGshas decreased saved at the peak hours to reduce the total oper-ation costs DMS has tried to optimize the ADN by purchasingreactive power from the wholesale market in lower prices It isworth noting that although the share of reactive power supportby DGs has decreased but the voltage profile for the worstbus has improved significantly as illustrated in Fig 16 Thissalient benefit is due to the physical variations of the networkstructure which has resulted in suitable power flows in thenetwork

Fig 17 explores the RL participation in the optimal day-ahead scheduling framework As it is seen for Case-Ialthough RLs are partially dispatched for the hours sapping8 to 24 they are mainly scheduled for peak hours includ-ing 14 to 18 The scheduling process in Case-II has loweredthe application of RLs by 2714 due to the network topol-ogy changes and the other dispatched elements The reducedamount of RLs can be envisaged as nonspinning standbycapacity to be offered to reserve or frequency control markets

The other remarkable but maybe unexpected observationregards the total power losses in the network The networkhourly losses associated with Cases-I and II are depicted inFig 18 It should be noted that the majority of existing studiesfocusing on application of reconfiguration have been mainly on

Fig 17 Apparent power reduction by RLs

Fig 18 Total active power losses in the network

TABLE IIHOURLY COST ANALYSIS IN CASES-I AND II

power loss minimization In contradiction here as the DMSturns off its DGs at the hours with low market prices such as8 to 12 and 21 to 24 it increases the amount of purchasedpower from the wholesale market and the total power lossesslightly rises at these hours However the total economic sav-ings are more dominant Accordingly DSO may compromisebetween increase in losses and more economic savings in day-ahead operations Effectively at peak hours during 13 to 20and at off-peak hours during 3 to 7 the implicit impact ofRCSs on lowering the power losses is evidently observable

Table II demonstrates the optimal hourly operation costs inboth cases along with optimal opening sequences of RCSsdetermined by DMS It can be observed that having activated



the hourly reconfiguration there will be economic savings atall hours throughout the day Specifically speaking the eco-nomic improvements are more remarkable at hours with lowwholesale electricity prices and normal load of the networkAt these hours DMS switches off the expensive interior DGunits and by changing the network topology provides suitablepatches for more purchase of electricity from the market withlower prices The total operation cost for Cases-I and II areachieved as $56 13437 and $53 03225 respectively It can beseen that by applying RCSs in the optimal operation schedul-ing of ADN there will be 553 decrease in the total operationcost which is an acceptable economic achievement for DSOAs the huge number of discos makes up the main buildingblocks of the electricity consumption in power systems anyreduction in the cost is of a vital importance

V CONCLUSION

This paper targeted to address an optimal operation schedul-ing framework for ADNs The main contributions of this paperare as follows

1) considering RCSs and flexible network topologies inoptimal scheduling of ADNs

2) considering the maximum switching operations forRCSs and hence confining the possible network topolo-gies

3) interrogating the effect of network reconfiguration inreserve scheduling problems

Through the extensive simulation studies it can be deducedthat considering RCSs in scheduling of the ADN wouldhave considerable effects in optimal daily dispatches of DGsand RLs Associated with hours with lower wholesale elec-tricity prices DMS has turned off its interior expensive unitsThen it has determined the optimal switching sequences forRCSs and by altering the network topology it has providedmore suitable patches for higher electricity delivery fromexternal grid This trend has declined the share of DGs andRLs in accommodating the networkrsquos electrical demand when-ever unprofitable and has resulted in released capacity ofDGs and RLs The unused capacities can be envisaged as thespinning and nonspinning reserve resources which could bespeculated for covering renewable energy uncertainties suchas those associated whit WTs Meanwhile implementation ofhourly flexible topologies has a significant improvement ineconomics of the ADN with a record of 553 decrease inthe total daily operation cost of DSO Regarding the networkpower losses although there is a slight increase associated withhours with normal loads hourly optimal reconfigurations havedecreased the total power losses for the hours both with peakand off-peak loads Results certify the recognizable meritsof integrating network reconfiguration capabilities in modernDMSs for optimal operation planning of emergent ADNs

REFERENCES

[1] G Martin ldquoRenewable energy gets the lsquogreenrsquo light in Chicagordquo IEEEPower Energy Mag vol 1 no 6 pp 34ndash39 NovDec 2003

[2] B Ramachandran S K Srivastava C S Edrington and D A CartesldquoAn intelligent auction scheme for smart grid market using a hybridimmune algorithmrdquo IEEE Trans Ind Electron vol 58 no 10pp 4603ndash4612 Oct 2011

[3] B Hamilton and M Summy ldquoBenefits of the smart grid [in my view]rdquoIEEE Power Energy Mag vol 9 no 1 pp 102ndash104 JanFeb 2011

[4] CD Adamo etal ldquoDevelopment and operation of active distributionnetworksrdquo CIGRE Working Group C611 Technical Brochure No 457Apr 2011

[5] A A S Algarni and K Bhattacharya ldquoDisco operation consideringDG units and their goodness factorsrdquo IEEE Trans Power Syst vol 24no 4 pp 1831ndash1840 Nov 2009

[6] T Niknam H Zeinoddini-Meymand and H Doagou-Mojarad ldquoA prac-tical multi-objective PSO algorithm for optimal operation managementof distribution network with regard to fuel cell power plantsrdquo RenewEnergy vol 36 no 5 pp 1529ndash1544 May 2011

[7] C Cecati C Citro A Piccolo and P Siano ldquoSmart operation of windturbines and diesel generators according to economic criteriardquo IEEETrans Ind Electron vol 58 no 10 pp 4514ndash4525 Oct 2011

[8] A Safdarian M Fotuhi-Firuzabad and M Lehtonen ldquoA stochasticframework for short-term operation of a distribution companyrdquo IEEETrans Power Syst vol 28 no 4 pp 4712ndash4721 Nov 2013

[9] M E Khodayar M Barati and M Shahidehpour ldquoIntegration ofhigh reliability distribution system in microgrid operationrdquo IEEE TransSmart Grid vol 3 no 4 pp 1997ndash2006 Dec 2012

[10] C Wang and H Z Cheng ldquoOptimization of network configurationin large distribution systems using plant growth simulation algorithmrdquoIEEE Trans Power Syst vol 23 no 1 pp 119ndash126 Feb 2008

[11] R S Rao S V L Narasimham M R Raju and A S Rao ldquoOptimalnetwork reconfiguration of large-scale distribution system using har-mony search algorithmrdquo IEEE Trans Power Syst vol 26 no 3pp 1080ndash1088 Aug 2011

[12] R S Rao K Ravindra K Satish and S V L Narasimham ldquoPower lossminimization in distribution system using network reconfiguration in thepresence of distributed generationrdquo IEEE Trans Power Syst vol 28no 1 pp 317ndash325 Feb 2013

[13] A Timbus M Larsson and C Yuen ldquoActive management of distributedenergy resources using standardized communications and modern infor-mation technologiesrdquo IEEE Trans Ind Electron vol 56 no 10pp 4029ndash4037 Oct 2009

[14] C Cecati C Citro and P Siano ldquoCombined operations of renewableenergy systems and responsive demand in a smart gridrdquo IEEE TransSustain Energy vol 2 no 4 pp 468ndash476 Oct 2011

[15] M Parvania M Fotuhi-Firuzabad and M Shahidehpour ldquoISOrsquosOptimal Strategies for Scheduling the Hourly Demand Response inDay-Ahead Marketsrdquo IEEE Trans Power Syst to be published

[16] G Han B Xu K Fan and G Lv ldquoAn open communication architecturefor distribution automation based on IEC 61850rdquo Int J Elect PowerEnergy Syst vol 54 pp 315ndash324 Jan 2014

[17] J Billo (2012) ldquoReactive power resourcesrdquo Federal EnergyRegulatory Commission AD12-10-000 [Online] Available httpwwwfercgoveventcalendarFiles20120420161518-AD12-10-000pdf

[18] A Kumar S C Srivastava and S N Singh ldquoA zonal congestion man-agement approach using real and reactive power reschedulingrdquo IEEETrans Power Syst vol 19 no 1 pp 554ndash562 Feb 2004

[19] B Moradzadeh and K Tomsovic ldquoMixed integer programming-basedreconfiguration of a distribution system with battery storagerdquo in ProcNorth Amer Power Symp Champaign IL USA pp 1ndash6 Sep 2012

[20] D E Goldberg Genetic Algorithms in Search Optimization andMachine Learning Boston MA USA Addison-Wesley 1989pp 65ndash90

[21] M E Baran and F Wu ldquoNetwork reconfiguration in distribution systemfor loss reduction and load balancingrdquo IEEE Trans Power Del vol 4no 2 pp 1401ndash1407 Apr 1989

[22] T Sousa H Morais Z Vale P Faria and J Soares ldquoIntelligent energyresource management considering vehicle-to-grid A simulated anneal-ing approachrdquo IEEE Trans Smart Grid vol 3 no 1 pp 535ndash542Mar 2012

[23] J Soares H Morais T Sousa Z Vale and P Faria ldquoDay-aheadresource scheduling including demand response for electric vehiclesrdquoIEEE Trans Smart Grid vol 4 no 1 pp 596ndash605 Mar 2013

[24] S Wong K Bhattacharya and J D Fuller ldquoElectric power distributionsystem design and planning in a deregulated environmentrdquo IET GenerTransmiss Distrib vol 3 no 12 pp 1061ndash1078 Dec 2009

[25] I Ziari G Ledwich A Ghosh and G Platt ldquoOptimal distribution net-work reinforcement considering load growth line loss and reliabilityrdquoIEEE Trans Power Syst vol 28 no 2 pp 587ndash597 May 2013



[26] (2013 Dec) Recloser Based Automation Solutions for SmartGrid up to 405 KV [Online] Available httpwwwtavridaeuaction=tavridapageampmenuid=79

[27] (2013 Dec) Outdoor Medium Voltage Automatic CircuitReclosers [Online] Available httpwwwalibabacomproduct-detail33KV-automatic-circuit-recloser ACR_1673941121html

[28] (2013 Jul) New York Independent System Operator [Online] Availablehttpwwwnyisocom

Sajjad Golshannavaz received the BSc (Hons) and MSc (Hons) degreesin electrical engineering from Urmia University Urmia Iran in 2009 and2011 respectively He is currently pursuing the PhD degree from the Schoolof Electrical and Computer Engineering University of Tehran Tehran Iran

His current research interests include smart grid technologies electric dis-tribution and energy management systems renewable energies and distributedgeneration

Saeed Afsharnia received the BSc and MSc degrees in electrical engi-neering from the Amirkabir University of Technology Tehran Iran in 1987and 1990 respectively and the PhD degree in electrical engineering fromthe Institute National Polytechnique de Lorraine (INPL) Lorraine Francein 1995

He is currently an Associate Professor with the School of Electrical andComputer Engineering University of Tehran Tehran His current researchinterests include smart distribution grids and FACTS applications in powersystems

Farrokh Aminifar (Srsquo07ndashMrsquo11) received the PhD degree in electri-cal engineering from the Sharif University of Technology Tehran Iranin 2010

He is currently an Assistant Professor with the School of Electrical andComputer Engineering the University of Tehran Tehran His current researchinterests include wide-area measurement systems reliability modeling andassessment and smart grid technologies He has been collaborating withthe Robert W Galvin Center for Electricity Innovation Illinois Institute ofTechnology Chicago IL USA since 2009

Dr Aminifar has served as the Guest Editor-in-Chief and Editor of threeSpecial Issues of the IEEE TRANSACTIONS ON SMART GRID He receivedthe IEEE Best PhD Dissertation Award from the Iran Section for his researchon the probabilistic schemes for the placement of pharos measurement unitsHe is the recipient of the 2013 IEEEPower System Operation TransactionsPrize Paper Award



energies instead of fossil fuels [1] In this way distributedgenerations (DGs) and storage devices are recent trends inresponding to the electrical energy requirements of distri-bution companies (discos) These technologies although areeconomicallyenvironmentally beneficial would pose a morecomplex operational situations for discos On the other handemergence of new mechanisms such as the open access tothe electricity markets and demand side programs are otherchallenging issues in the way of ensuring a sustainable energyfuture [2]

Although forgotten or maybe underemphasized before butnow there is a common sense that the sustainable energydevelopment necessitates more intelligent grid with onlinecontrol capabilities as well as bidirectional interactions withconsumers [3] Thanks to the substantial recent advances in theinformation and communication technology (ICT) along withthe rapid growth of field intelligent electronic devices (IEDs)the notion of Smart Grid has been evolved in power systemsspecifically at the distribution level Active distribution net-works (ADNs) as the leading sort of smart distribution gridsare characterized with controllable embedded generations suchas DGs demand side management programs such as respon-sive loads (RLs) and also remotely controlled switches (RCSs)to attain flexible network topologies Considering the topol-ogy changes modern bidirectional relays should be effectivelydeployed in all feeders to yield a robust protection schemein different operational horizons Distribution managementsystem (DMS) as the heart of economical and technicaloptimizations is responsible for optimal operational decisionmakings in ADNs utilizing optimal load flow algorithms [4]As well DMS is apt for monitoring the harmonic contentsof the network and alleviating the power quality concernsby optimally adjusting the set points of power electronics-based devices By this way the DMS optimally assigns thepower routings and performs an effective voltage control inthe network while keeping all constraints satisfied

In this context Algarni and Bhattacharya [5] have proposeda day-ahead operational scheduling framework where DGsare judiciously treated by including their relevant goodnessfactors Results revealed remarkable reduction in operationcosts due to more decrease in power losses Niknam et al [6]have proposed a multiobjective strategy minimizing the powerlosses and operation costs for the day-ahead optimal opera-tion of a distribution network including fuel cell (FC) powerplants Cecati et al [7] have devised an efficient DMS algo-rithm for day-ahead operation scheduling of smart distributionnetworks wherein an optimal power flow finely controls theelements such as wind turbines (WTs) and DGs Meanwhilesome recent studies have proposed a two-stage frameworkfor scheduling of ADNs [8] The first stage performs aday-ahead scheduling of resources and the effect of real-time data on optimal operation strategies are subsequentlyexplored in the second stage Although demonstrating notice-able achievements some of the marvelous merits providedby ADNs such as integration of RLs and RCSs are disre-garded in these works More recently Khodayar et al [9]have benefited from the application of motorized switches inoptimal operation scheduling of microgrid of Illinois Institute

of Technology campus to increase the reliability of the sys-tem Wang and Cheng [10] and Rao et al [11]ndash[12] havetailored the reconfiguration procedure considering the pres-ence of DGs in the network In these studies the impact ofconcurrent reconfiguration and DG sizing problem on the peakload power loss reduction is well-assessed It is however sen-sible that the possibility of daily reconfiguration could bringfurther monetary savings in the utility operation cost

Based on the foregoing discussions this paper presents anintelligent DMS incorporating supervisory control and dataacquisition (SCADA) system to tackle the optimal short-termoperational scheduling of ADNs The proposed DMS under-takes the optimal control of available active elements byminimizing the total operation costs including cost of pur-chasing electricity from the wholesale market cost of DGdispatches cost of RL participations cost of reactive sup-port as well as cost of RCSs switching actions In contrastto constant pricing for reactive power ancillary services herea practical pricing model is deployed for DG reactive powerparticipations Furthermore the automatic motorized switchesnamely RCSs and available tie-lines are wisely scheduled tosettle the optimal hourly network topologies In order to estab-lish a more practical technique the maximum number of RCSswitching actions is also limited The obtained technical andeconomical improvements are thoroughly explored to assessthe effect of applying RCSs in daily scheduling of ADNThe effect of hourly reconfiguration on the capacity releaseof DGs and RLs is investigated likewise which puts forwardnew insight to the reserve scheduling problems

The rest of this paper is organized as follows Section IIaddresses the generic description of the envisaged ADNSection III formulates the proposed operational schedulingframework Section IV interrogates several test cases and pro-vides numerical results Eventually the paper is concluded inSection V

II GENERIC DESCRIPTION OF THE ENVISAGED ACTIVE

DISTRIBUTION NETWORK

The advent of modern ICT such as wireless communica-tions by general packet radio service (GPRS) are offeringexcellent online controllability on different engineering appli-cations [13] Meanwhile ADNs benefitted from a compre-hensive ICT infrastructure are in the way of rapid evolutionADN refers to a self-organized system capable of controllingand enhancing the operation of all its interconnected ingredi-ents [14] Fig 1 demonstrates an ADN wherein the followinginformation is elaborated

1) Besides the main grid connection disco also encom-passes DGs WTs RLs and RCSs as active elements

2) DMS as a set of hardwaresoftware complements isentrusted to obtain the optimal operational decisionsregarding all network ingredients To do so it requiresgathering some preliminary system information andhenceforth collaborates with system operator to receivethe required data such as market prices one day inadvance [15] Also DMS may be equipped with mar-ket price forecasting mechanism specifically for havingeffective participation in energy and reserve markets













Minimize F (x u)

=sum

tisinT

⎛

⎝sum

sisinS

ρDAt PDA

ts +sum

gisinG

XgtSUg

+sum

gisinG

CDGgpt +

sum

gisinG

ZgtSDg +sum

lisinL

ρRLt PRL

ts

+sum

gisinG

kDAq ρDA

t QDAts +

sum

gisinG

kRLq ρRL

t QRLtl +

sum

gisinG

CDGgqt

⎞

⎠

+sum

k

CSWNSWRCSk

(1)


x =[PDA

s QDAs PDG

g QDGg PRL

l QRLl RCS

](2)


](3)

u =[|V| |Sf |

](4)

|V| = [|V1| |V2| ∣∣VNbus

∣∣] (5)∣∣Sf

∣∣ = [|S1| |S2| ∣∣SNbr

∣∣] (6)





gt + cg

(PDG

gt

)2 (7)




Cgqt = kDGq Igt

[Cgpt

(SDG

gmax

)

minus Cgpt

(radic(SDG

gmax)2 minus (QDG

gt )2

)] (8)



B Constraints


PDAts +

sum

gisinGi

PDGtg + PRL

ti minus PDti

minussum

f isinFampjisinB

Pftij

(Vi Vj Yij θij

) = 0 (9)

QDAts +

sum

gisinGi

QDGtg + QRL

ti minus QDti

minussum

f isinFampjisinB

Qftij

(Vi Vj Yij θij

) = 0 (10)


[(PDA

ts

)2 +(

QDAts

)2]12



PDGgmin le PDG


QDGgmin le QDG


[(PDG

tg

)2 +(

QDGtg

)2]12


PFDGtg = PDG

tg(PDG

tg + QDGtg


PFDGgmin le PFDG



0 le PRLti le PRL


QRLti = tan

(cosminus1 (

PFRLi



[(Pf

tij

)2 +(

Qftij

)2] 1







NSWRCSk

=sum

t


)(23)

NSWRCSk

le NSWmax (24)






















B Main Assumptions







































V CONCLUSION






REFERENCES


















































Minimize F (x u)

=sum

tisinT

⎛

⎝sum

sisinS

ρDAt PDA

ts +sum

gisinG

XgtSUg

+sum

gisinG

CDGgpt +

sum

gisinG

ZgtSDg +sum

lisinL

ρRLt PRL

ts

+sum

gisinG

kDAq ρDA

t QDAts +

sum

gisinG

kRLq ρRL

t QRLtl +

sum

gisinG

CDGgqt

⎞

⎠

+sum

k

CSWNSWRCSk

(1)


x =[PDA

s QDAs PDG

g QDGg PRL

l QRLl RCS

](2)


](3)

u =[|V| |Sf |

](4)

|V| = [|V1| |V2| ∣∣VNbus

∣∣] (5)∣∣Sf

∣∣ = [|S1| |S2| ∣∣SNbr

∣∣] (6)





gt + cg

(PDG

gt

)2 (7)




Cgqt = kDGq Igt

[Cgpt

(SDG

gmax

)

minus Cgpt

(radic(SDG

gmax)2 minus (QDG

gt )2

)] (8)



B Constraints


PDAts +

sum

gisinGi

PDGtg + PRL

ti minus PDti

minussum

f isinFampjisinB

Pftij

(Vi Vj Yij θij

) = 0 (9)

QDAts +

sum

gisinGi

QDGtg + QRL

ti minus QDti

minussum

f isinFampjisinB

Qftij

(Vi Vj Yij θij

) = 0 (10)


[(PDA

ts

)2 +(

QDAts

)2]12



PDGgmin le PDG


QDGgmin le QDG


[(PDG

tg

)2 +(

QDGtg

)2]12


PFDGtg = PDG

tg(PDG

tg + QDGtg


PFDGgmin le PFDG



0 le PRLti le PRL


QRLti = tan

(cosminus1 (

PFRLi



[(Pf

tij

)2 +(

Qftij

)2] 1







NSWRCSk

=sum

t


)(23)

NSWRCSk

le NSWmax (24)






















B Main Assumptions







































V CONCLUSION






REFERENCES










































gt + cg

(PDG

gt

)2 (7)




Cgqt = kDGq Igt

[Cgpt

(SDG

gmax

)

minus Cgpt

(radic(SDG

gmax)2 minus (QDG

gt )2

)] (8)



B Constraints


PDAts +

sum

gisinGi

PDGtg + PRL

ti minus PDti

minussum

f isinFampjisinB

Pftij

(Vi Vj Yij θij

) = 0 (9)

QDAts +

sum

gisinGi

QDGtg + QRL

ti minus QDti

minussum

f isinFampjisinB

Qftij

(Vi Vj Yij θij

) = 0 (10)


[(PDA

ts

)2 +(

QDAts

)2]12



PDGgmin le PDG


QDGgmin le QDG


[(PDG

tg

)2 +(

QDGtg

)2]12


PFDGtg = PDG

tg(PDG

tg + QDGtg


PFDGgmin le PFDG



0 le PRLti le PRL


QRLti = tan

(cosminus1 (

PFRLi



[(Pf

tij

)2 +(

Qftij

)2] 1







NSWRCSk

=sum

t


)(23)

NSWRCSk

le NSWmax (24)






















B Main Assumptions







































V CONCLUSION






REFERENCES


























































B Main Assumptions







































V CONCLUSION






REFERENCES











































B Main Assumptions







































V CONCLUSION






REFERENCES




































































V CONCLUSION






REFERENCES






















































V CONCLUSION






REFERENCES









































V CONCLUSION






REFERENCES


















































Documents

IEEE TRANSACTIONS ON SMART GRID 1 Smart Distribution Grid: Optimal … · 2015-07-30 · This article has been accepted for inclusion in a future issue of this journal. Content is