tutorial 8: actual market data

ElectricityMarkets,theSimplexAlgorithm,andtheSimplexNodalapp

Tutorial8:ActualMarketData Page164

Tutorial 8: Actual Market Data Thistutoriallooksathowtomodelaportionoftheactual New Zealand electricity market, detailingwhereto locatethesource filesandhowtoextractthenecessarymodellingdata.Theresultsproducedbytheapparecomparedwiththosefromtheactualsystem.We also use these sample models to furtherinvestigate transmissionrentals, and to lookat theimpact that the ordering of the inputs has on theprogressofthesimplexalgorithm.

Hawkes Bay Model The portion of theNew Zealand electricitymarketthatwillbemodelledistheHawkesBay.Thedataisobtainedfromfreelyavailableonlinesources.

Sourceofreal-worlddataAdiagramof the networkmodel used by theNewZealand electricity market is available on theSystemOperator’swebsite:https://www.transpower.co.nz/system-operator/key-documents/maps-and-diagramsThediagramisnamedthe“SpdLocatordrawing”soyoucanalsofinditusingGoogle.



The input and output data for final pricingschedules fromtheNewZealandelectricitymarketis available by going to the Electricity Authority(EA)website…https://www.emi.ea.govt.nz/…and then following the links:WholesaleDatasets,FinalPricing,CaseFiles.

TwosamplemodelsThe market data from the case files on the EAwebsitehasbeenusedtobuildsamplemodelsthatcomepre-loadedwiththeapp.Therearetwoversionsofthesesamples.Thelargerof these more closely follows the actual marketsystem, modelling the 110kV transmission systemand the supply transformers that convert thevoltage from 110kV to the distribution voltage of33kV or 11kV. The load quantity is the valuemetered at the 33kVor 11kVbuses. Thismodel isreferred to as the 033 model (because it modelsdowntothe33kV/11kVlevel).The smallermodel uses the scheduled flow on thesupplytransformerstomodeltheloadatthe110kVlevel. This removes the need to model the supplytransformersandmakesthemodelfastertosolve.Italsoallowsustoaccountforthefixedlossesofthe



supply transformers, which is something that theactualmarketsolverincludesbuttheappdoesnot.Thisisreferredtoasthe110model.

LoadingasamplemodelSample models are loaded via the Models display,whichisaccessedviatheFolderbuttonindicatedinFigure100.

Figure100:ButtonthatleadstotheModelsdisplay

Onceon theModels display the Samplesdisplay isaccessed via the Samples button located on thelowertoolbar,asshowninFigure101.

Figure 101: The Samples button on the lower toolbar of theModelsdisplay

The Samples display is shown in Figure 102. Thesample models of the Hawkes Bay are the 033modelandthe110modelasdescribedabove.



Figure102:Samplesdisplay

Viewa screenshotof the033modelby tapping itsname, then set it as the currentmodel by tappingthe Load button on the toolbar. On an iPhone themodel is larger than thedisplay…youcanzoom inandoutusingthepinchgesture.Whenthemodeliszoomedinyoucanmoveabout(scroll)bydraggingthebackground.

Real-worldnetworkmodelThe portion of theNew Zealand electricitymarketthat ismodelled is shown in Figure 103. This is asection of the network model diagram from theSystemOperator’swebsite,asdescribedabove.The orange lines are 220kV, red lines are 110kV,green33kVandblack11kV.



Figure103:HawkesBayportionoftheNZelectricitynetwork



This portion of the electrical network covers theHawkesBayprovinceofNewZealand.Itconnectstothe rest of the power system via twointerconnecting transformers, T3 and T4 at theRedcliffe(RDF)substation.

ModellinginflowfromtherestofthesystemInthesamplemodelsweuseadummygeneratortomodel the inflow of power from the rest of thesystem via the T3 and T4 interconnectingtransformers at RDF. The offer price and offerquantity of thedummygeneratorwill bebasedonthereal-worldresultattheseinterconnectors…thesumofthescheduledflowonT3andT4willbetheofferquantity,while theofferpricewillbe thebuspriceattheRDF110kVbus.There are transmission circuits to the southof theHawkesBay.Thesearecapableofconnectingtotherest of the electrical network, but they are notconnected because there is an operational split inplace to prevent overloads on the 110kV… the110kV would effectively be in parallel with the220kVifthesplitwasclosed.



Usingreal-worlddataThesamplemodelshavebeenbuiltusingdatafromcase files obtained from the Electricity Authoritywebsite,asdescribedabove.The following sections describe how the case filedatarelatestotheinputdatausedbytheapp.Therearesomecaveats.Thefirstisthatasof01-April-15the actual system moved from modelling branchlosses using three flow-loss segments, i.e., three ineach direction, to using six segments in eachdirection. The app is currently only set up to usethree or four segments. Hence, while you can stilluseinputdatafromafter01-April-15,therewillbesmall differences in the results due to thedifferencesinthenumberofsegments.TheothercaveatrelatestotheHawkesBaylocation.As of 01-April-15 (coincidentally the same date asthe loss segments change) the substations atGisborne (GIS) andWairoa (WRA),which arebothin the sample models, are no longer under thecontrol of the System Operator, hence they havebeen removed from the electricity market model,andthereforeyouwon’tfindtheirdataincasefilesafterthisdate.



Detailsofreal-worlddataThe market data on the Electricity Authoritywebsite consists of zip files containing the 48individualhalfhours(tradingperiods)thatmakeupatradingdayfortheelectricitymarket.Thedateandtimeofthefirstperiodofthescheduleareinthefilename,e.g.,forthefile…MSS_91112015011100714_0X…thefirstperiodis09-JAN-201511:00.Except,thisis UTC time, so the local time is actually 10-JAN-2015 00:00. The schedule type is 111, which is afinal pricing schedule. The 714 on the end is arandomthree-digitnumber.Eachzip filecontains input filesandresult files forall of the trading periods covered by the case. Foreach trading period there is one results file andseveral input files.Thedata for thesamplemodelsisobtainedfromtheinputfilesshowninTable4.Table4:Inputfiledatatobuildasamplemodel

Fileextension

Section Modeldata

PERIOD BIDSANDOFFERS ENOF=Energyoffers

PERIOD PNODEINT MV90LOAD.



Meteredloadatthesupplybus.Usedasloadbythe033model

MSSMOD BRANCHLIMIT BranchmaxflowMSSNET BRANCHBUS Susceptanceand

ResistanceWealsousedata fromtheresults file.AsshowninTable5theresultsdataisusedtomodeltheT3andT4 interconnecting transformers as a dummygenerator,andtoprovidethescheduledflowonthesupply transformers, which we use as the load inthe110model.Table5:Resultfiledatausedtobuildsamplemodel

Fileextension

Section,Column

Modeldata

SPDSOLVED BRANCH,TO_MW

Usedasofferquantityforthedummygenthatrepresentsthe220kV-110kVtransformersT3&T4

SPDSOLVED BUS,PRICE Usedasofferpriceforthedummygenthatrepresentsthe220kV-110kV



transformersT3&T4SPDSOLVED BRANCH,

FROM_MWFlowonthesupplytransformers,usedasloadbythe110model

Theabovetablesprovideanoverviewofwherethesourcedatais located.Thefollowingsectionsshowthedetailsofhowthedataisextracted.

GeneratoroffersThe energy offers for the actual generators (asopposed to the dummy generator that models T3and T4) are from the BIDSANDOFFERS section ofthePERIODfile,asshowninFigure104.

Figure 104: Energy offers from BIDSANDOFFERS section ofPERIODfile,showingoffersforgeneratorPRIattheTUIbus

In-flowdataInourmodelthedummygeneratornamed“T3&T4”attheRDFbusrepresentsthein-flowofpowerfromthe rest of the power system to the Hawkes BayareaviainterconnectingtransformersT3andT4.



The offer quantity for the dummygenerator is thesumofthescheduledbranchflowsontheT3andT4branches.Theofferprice for thedummygeneratoris the scheduled bus price for theRDF110kVbus.Note that thescheduledquantitiesareresults fromtheactualsystem.The scheduled branch flow is from the BRANCHsection of the SPDSOLVED file as shown in Figure105.

Figure 105: Scheduled branch flow from BRANCH section ofSPDSOLVEDfile(usedasofferquantityfordummygen)

Thescheduledbusprice is fromtheBUSsectionoftheSPDSOLVEDfile,asshowninFigure106.

Figure 106: Scheduled bus price from BUS section ofSPDSOLVEDfile(usedasofferpricefordummygen)



Loaddata:frominputdataFor the033model the loaddata is fromtheMV90column of the PNODEINT section of the PERIODinputfile,asshowninFigure107.

Figure107:LoaddatafromPNODEINTsectionofPERIODfile

Because we are obtaining our data from a FinalPricing schedule the load data representsmeteredload,i.e.,loadthatwasactuallysupplied.Hence,itisrequired load and therefore we will assign a veryhighbidpricetoensurethatitallclears.

BranchdataThe branch limit data is from the BRANCHLIMITsection of the MSSMOD input file, as shown inFigure108.

Figure 108: Branch limit from BRANCHLIMIT section ofMSSMODfile



Branch susceptance and resistance are from theBRANCHBUSsectionoftheperiod–specificMSSNETfile,asshowninFigure109.

Figure 109: Branch susceptance and resistance fromBRANCHBUSsectionofperiod-specificMSSNETfile

There is oneof theseperiod–specificMSSNET filesfor each trading period in the schedule, not to beconfusedwiththesingleMSSNETfilethatcoversalltrading periods (we don’t use data from that filebecause it mostly contains info about theconnectivity of the network model… we manuallybuilt our model based on the System Operator’snetworkdiagram,sowedon’tneedthemodelfromtheMSSNETfile).

Percentper-unitIntheNewZealandelectricitymarkettheresistanceandsusceptancevaluesinthecasefilesareper-centper-unit, i.e.,100 times theper-unitvalue.Becausetheappexpectsper-unitvalues, the resistanceandsusceptance values from the input files must be



dividedby100toremovetheper-centfactorbeforetheyareenteredintothemodel.

Per-unitbaseAs discussed in the branch losses section, the appexpectsper-unitvaluesthatwerecalculatedusinga100MVA base. The data in the input files wascalculatedusing a 100MVAbase, sowedon’t needtomakeanyadjustmentsforthat.

ResultcomparisonFigure110displaystheresultproducedbythe033samplemodelwhileTable6andTable7presentthereal-world results, extracted from the SPDSOLVEDfile.Table6:BusresultsfromSPDSOLVEDfile

ID_ST ID_KV LOAD PRICE FHL 33 32.995 75.7776 FHL 110 0 75.4302 GIS 50 25.791 78.3108 GIS 110 0 78.2459 GIS 11 0 78.2726 RDF 33 40.785 74.9964 RDF 110 0 74.909 TUI 11 0.291 76.8425 TUI 110 0 76.6103 WRA 11 5.966 77.1013 WRA 110 0 76.949



Figure110:ResultforHawkesBay033sample



Table7:BranchresultsfromSPDSOLVEDfile

BRANCHNAME FIXED LOSS FLOW

BRANCH LOSSES

FHL FHL_RDF1 1 0 -20.775 0.06202 FHL FHL_RDF2 1 0 -20.556 0.0609 FHL FHL_TUI1 1 0 8.056 0.14706 FHL T1 T1 0.0545 12.837 0.07543 FHL T2 T2 0.059 20.258 0.08147 GIS GIS_TUI1 1 0 -13.228 0.27438 GIS GIS_TUI2 1 0 -13.194 0.2779 GIS T2 T2 0.0286 12.92 0.03328 GIS T4 T4 0.0286 12.92 0.03328 RDF RDF_TUI1 1 0 6.066 0.11852 RDF RDF_TUI2 1 0 6.066 0.11852 RDF T1 T1 0.045 20.738 0.06541 RDF T2 T2 0.0326 20.133 0.05985 TUI T15 T15 0.0112 -0.297 0.0121 TUI TUI_WRA1 1 0 3.043 0.0134 TUI TUI_WRA2 1 0 3.037 0.01337 WRA T1 T1 0.0379 -3.01 0.0438 WRA T2 T2 0.0374 -3.005 0.04336

Differences between the app and the real-worldresults can be tracked down to the app notmodelling fixed losses. Transformers have fixedlosses, which are losses that occur when thetransformer isenergised, regardlessof the flow. Intheactualsystemthesolverassignsthefixedlosses50-50 to thebusesat eachendof the transformer,



regardless of where the dynamic losses areassigned.Theappdoesnotmodelfixedlosses.

SimplifyingthemodelandimprovingtheresultsAsmentioned in the introduction, we can simplifythe model by using the scheduled in-flow of thesupply transformers as the load value. This willremovetheneedtomodelthesupplytransformersandthesupplybus. Itwillalsoallowus toaccountforthetransformerfixedlosses…becausetheactualsystem assigns fixed losses 50/50 to the buses ateitherendofthetransformer,halfofthefixedlossesare already included in the scheduled flow on thetransformer. We can model the other half of thefixedlossvaluebyexplicitlyaddingittothe110kVload.

Loaddata:fromscheduledresultsThescheduledin-flowonthesupplytransformersisobtained from the BRANCH section of theSPDSOLVED file as shown inFigure111.The fixedlossisalsoobtainedfromthebranchresults.

Figure111:LoadfromBRANCHsectionofSPDSOLVEDfile



For example, to calculate the110kV load value forFHL, we add the branch flows on FHL T1 and T2.Thisflowwillincludehalfofthefixedlossesthattheactual solver assigned to the supply bus, the otherhalfofthefixedlossesareincludedinourmodelbyexplicitly adding them to the load value that wecalculatefromthetransformerbranchflows.ThescheduledflowandfixedlossesforFHLT1andT2areshowninTable7.Hencethevalueweuseforthe 110kV load at FHL is the scheduled flow plushalfthefixedlosses:20.258+12.837+(0.0545+0.059)/2=33.15175Theresultsofthe110kVmodelareshowninFigure112.Theseresultsagreewiththosefromtheactualsystemtowithin2decimalplaces.



Figure112:ResultforHawkesBay110samplemodel

Further investigations Wecanuse theHawkesBay110model toconductsomefurtherinvestigationsasfollows.



ChangingthenumberofbranchsegmentsScroll through the loss results for theHawkesBay110 result by using the forward and back arrowsindicated in Figure 113. Notice that most of thescheduled branch flows are in the first flow-losssegment.

Figure113:Usearrowbuttonstoscrollthroughlossresults



As explained inTutorial 4: Transmission Losses, itturns out that no loss rentals are incurred onbranches where the scheduled flow is in the firstflow-losssegment.

Figure114:Changingnumberofsegmentsfrom3to4

If we increase the number of flow-loss segmentsfromthreetofourthentheflowonanumberofthe



branches will move into the second flow-losssegment,andwecanseewhathappens.Tochangethenumberofbranchsegmentsfrom3to4, go to the Losses display for any branch and tapthe button for 4 segments. The button colour willchangetoredtoindicatethatthevaluehaschanged,asshowninFigure190.With the number of segments changed, when youleavethedisplayyouwillbealertedthatthechangewillbeappliedtoallbranches…tapOKtoapplythechange.Aftersolvingthemodelwithfourbranchsegments,theResultsdisplayisshowninFigure115.Nowthatfewerbrancheshaveflowconfinedtothefirstflow-loss segment there has been an increase in thetransmission rentals, i.e., the $Grid on the Resultsdisplay, for the reasons discussed in Tutorial 4:TransmissionLosses.



Figure115:Resultafterincreasingsegmentsfrom3to4

In contrast to the increase in transmission rentals,thelosseshavedecreasedslightlybecausethefour-segmentmodelhasresultedinalowerlossestimatefor the scheduled flows, due to the way that thesegments have adjusted in the move from 3 to 4segments.



ChangetheorderingoftheconstraintsChanging the order in which the components aresupplied to the solverwill change theorder of theconstraints in the initial tableau, and thereforechangehowthesolveprogresses.Figure 116 indicates the Solve Setting thatdeterminesthesortorder.

Figure116:Changingthesortorderoftheconstraints



Figure 117 and Figure 118 show the impact thatdifferent sort orders have on the progress of thesimplex algorithm. These plots are accessed viaResults–Iterations–Chart.Note that although the path to the solution isdifferent, the end result is the same. In Tutorial 9:Simplex Explainedwewill seewhy the sort orderchangesthepathtakenbythesimplexalgorithm.



Figure117:ObjectivevaluevsiterationcountfortheHawkesBay110modelwithascendingsortorder



Figure118:ObjectivevaluevsiterationcountfortheHawkesBay110modelwithdescendingsortorder



SummaryIn this tutorialwe saw how tomodel a portion ofthe New Zealand electricity market by sourcinginput data from the actual electricity market. Theresults produced by the app held up quite wellwhen compared with the results from the actualmarket.Weusedthesamplemodeltoinvestigatetheimpactthat changing the number of branch segments hasontransmissionrentals,asexplained inTutorial4:TransmissionLosses.We also investigated the impact that changing thesortorderoftheconstraintshasontheprogressofthe simplex algorithm. In the next section we willlookathowthesimplexalgorithmsolvesthemodel,whichwillmake itclearwhythesortorderhasanimpact.

tutorial 8: actual market data

Documents