Seite - 45 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2019,12, 164 of information. Furtherdetailsoneachdomain, its involvedactors, andrespectiveapplicationscanbe foundin [3].Oneof theadvantagesof this integration iscustomerengagement,whichplaysakeyrole in theeconomiesofenergy trade. Inotherwords, theoldconceptofuni-directional energyflowis replacedbythenewandsmartconceptofbi-directionalenergyflow—transformationfromtraditional consumer toasmartprosumer [4]. Figure1.ConceptualdiagramofSG. Theresulting/newgrid, integratedwithadvancedmeteringinfrastructure, facesmanychallenges suchas [5]: (i)designingnewtechniques tomeet the loadwhilenot increasing thegenerationcapacity; and (ii) devisingnewways/policies to ensure customer engagementwithutility. When installing new technologies, utilities aim for amaximumpossible return on an investment. However, this maximizationwould require that thedaily operationsof anSGutility (suchas strategicdecisions tobridgethegapbetweendemandandsupply,andfuel resourceplanning)areproperlyconveyed. All thesedecisions are highly influencedby load forecast strategy(ies) [6]. Accurate load forecast means that both utility and prosumer can maximize their electricity price savings due to spot price establishment—one of the major reasons that utilities show growing interest towards SG implementation. The concerned utility forecasts the future price/load signal which is based on thepastactivitiesofusers’ energyconsumptionpatterns. Inresponse to the forecastprice/loadsignal, theusersadjust theirenergyconsumptionschedulessubject tominimizationofelectricitycostand/or theircomfort level[7]. Inreference[8],Hippertetal. classifyloadforecastbasedontimetobepredicted (Figure2): short-term,medium-termandlong-term. Short-termloadforecasting is furthercategorized into two types: (i) very short-term; and (ii) short-term forecasting. Thefirst onehas aprediction durationfromseconds/minutes tohoursandmodelapplications inflowcontrol. Thesecondonehas predictionhorizonfromhours toweeksandmodelapplications inadjustinggenerationanddemand, therefore,usedto launchoffers to theelectricalmarket. Theshort-termforecastingmodelsarevital inday-to-dayoperations,evaluationofnet interchange,unit commitmentandschedulingfunctions, andsystemsecurityanalysis. Inmediumtermforecasting, thepredictionhorizon is typicallybetween months. Thesemodels areusedbyutilities for fuel scheduling,maintenanceplanning, andhydro reservoirmanagement. In long-termforecasting, thepredictionhorizon is foryears.Utilitiesuse these typesofmodels forplanningcapacityof thegridandmaintenancescheduling. Sinceaccurate load forecast isneededbyutilities toproperlyplantheongoinggridoperations forefficientmanagement of their resources, this paper aims at an accurate load-forecastingmodel. However, the scope of this paper is limited to short-term load forecastingwith aday-aheadprediction horizon only. In the literature, twotypesofday-ahead loadforecasting (DALF)modelshavebeenpresented: linear andnon-linear [9].Also, [10]hashighlightedtherelative limitation(s)of linearmodelsascompared to non-linearmodels. In reference [9], the non-linearmodels are investigated in five classes: (i) supportvectormachine-basedmodels; (ii)Markovchain-basedmodels; (iii) artificialneuralnetwork (ANN)-basedmodels; (iv) fuzzyANN-basedmodels; and(v) stochasticdistribution-basedmodels. Thesupportvectormachine-basedmodels [11–13]achieverelativelymoderateaccuracy,butat the 45