Seite - 377 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1138 Figure3. Primaryelectrical systeminWanjiangareaabove110kv, includingelectricpowerplants, transmission buses, converting stations, and user loads. The feeders are marked under their corresponding loadsides. 6DPSOLQJ 3RLQWV 6DPSOLQJ 3HULRG PLQXWHV Figure4.LoadcurveofCustomer53990001fromFeeder2duringamonth,wherethesamplingperiodis 15min. Besides thehistorical loadcurves, short-termloadforecasting is influencedbythe factorsofdate, weatherandtemperature. Therealhistoricaldataofweatherandtemperature in thecorresponding area inDongguanCitywereobtainedonline fromtheweather forecastwebsites. Thecategoriesof weather includesunny,cloud,overcast, light rain, shower,heavyrain, typhoonandsnow.Thedate featurescanbefoundincalendars. 2.3. ClusteringandQuantization Thecustomofelectricityuseandcharacteristicsof loadcurvearedifferentamongthedifferent categoriesofcustomerssuchas industrial customers, residential customersandinstitutioncustomers. Thedifferentcharacteristicswouldaffect theperformanceof forecasting. Trainingforecastingnetworks with each customer separatelywouldbe ahuge computation and storageproblem. Therefore, in 377