Page - 101 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 3442 basedonthedemographicandsocio-economicvariablesviz.,GDP,population, import, exportand employmentusingregressionandANN[31]. Machine learning is one of the effective methods for pattern recognition in big data. These algorithmsfind thepatterns in the data bynature andhelpmaking better predictions and critical decisions in Energy load, peak and price forecasting, image processing, face recognition, motionandobjectdetection, tumourdetection,predictivemaintenance,natural languageprocessing. Gajowniczeketal. hasproposedadataminingtechniquetofindout theelectricitypeak loadfor the countrybyrepresentingthesameasapatternrecognitionresearchproblemrather thanatimeseries forecastingproblembyusingANN.Themaininnovationisthattheydetect96.2%ofthepeakelectricity loadaccuratelyup toadayahead [32]. Singhet al. alsopresentedadataminingmodel topredict the trendinenergyconsumptionpattern thatdescribe thedomesticdeviceusage inconnectionwith hourly,daily,weekly,monthlyyearlybasisaswellasdomesticdevice todomesticdevice linkages ina house. Theyproposedunsuperviseddataclusteringandfrequentpatternmininganalysisonenergy timeseries. Bayesiannetworkpredictionwasreferredforenergyusageforecasting. Theaccuracyof theresultsoutperformedSVMandMLP’saccuracyof81.82%,85.90%and89.58%for25%,50%and 75%of thesizeof thedatausedfor trainingrespectively [33]. Thus the literature reviewof threedecades reveals thatvarious technologiesandapplications wereusedtopredictenergyconsumption invarioussectorswhichhelpedustoutilize theproposed approach incomputingtheenergyconsumptionfor India. Themaincontributionof thisarticle is that itprovides • Apoint forecast for the total electricity consumption for the upcoming years up to 2030 is determinedwhich in turnwillhelp theenergyplanning inaholisticapproachfor thenation. • Aninsight to thepolicymakersatbridgingthegapbetweentheforecastedandtheactualdata for future. • Themajorcontributionof thearticle is that it emphasizes theresearchers toget toknowthebasic statisticalmodelsbeforeproceedingto theadvancedpackages. The goal of the study is to forecast the short-termTECof India using the basic and reliable methodologieswhichseemstobemuchbetter thantheadvancedmethods in forecastingtheenergy consumptionof India. Thecorrespondingauthorhasdoneaforecastofenergyconsumptionofastate in India,TamilNadu, inasmallerscaleduringhispost-graduation;whichactually is themotivationof theresearch.Apart fromthat theauthorsreviewedmanystudiespertainingtoenergyconsumption ofTurkey, JordanandChinaandso forth. whichmotivated themtoundertake thestudy for India. Dr. Iniyan is theResearch supervisor of the correspondingauthor and is aveteran in thisfieldof energyplanning,whohas takenupvariousprojectsandisalsoavoraciouspublisherandisoneof themajorsourcesof inspiration. Thedatausedfor theanalyses is sometimescarriedover fromthe year1970. Forecastedoutcomereveal that itholdsgoodonthehistoricaldata takenfor theanalysis. With the interventionofnewmethods, thereareareas forprobablepotentialenhancement.Anadded region forprogresswouldbe tooptimize the forecast further. For India theenergyconsumption is forecastedfor theyear2030andthisshallbedoneevenforyearsdownthe lanefromthenonthat is, longtimeforecasting. 2.MaterialsandMethods Data drivenmodels are those which use available prior data to forecast energy behaviour. Toperformthis, adatabase isestablishedto train themodels,bycombiningdissimilar techniques for predictingtheenergyconsumption.Amongthedatadrivenmodels themostpopularareblack-box basedapproacheswhichshallbeusedforenergypredictionandforecastinginwhichregressionmodel, multiple linear regressionmodel, decision trees,ANN, support vectormachineandvariousother optimization techniquesshall alsobeemployed. Byutilizing theblack-boxapproachthepresentstudy isperformedwiththemajorobjectiveofpredictingtheTotalElectricityConsumption(TEC)inindustry, 101