Seite - 225 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1900 is10.74%,which includes theuncertaintyof themodel itself. Then,usingtheweather forecastdata beforeandafterprocessingwithMCMtoforecast loadseparately,weobtaintheforecast resultsP2and P3. Theaccuracyof loadforecastingusingthemeteorological forecastdatadirectlyandthatof thedata processedbytheMCMare11.54%,10.92%,respectively. In termsofMAEandRMSE, thevaluesofP2 are74.3807kWand90.8474kW,respectively,while thevaluesofP3are67.0291kWand85.4057kW. It is clear that theaccuracyofP3 isbetter thanthatofP2. Table5.ComparisonbetweenP1,P2,andP3. PredictionLoad MAPE(%) MAE(kW) RMSE(kW) P1 10.74% 67.8305 84.4138 P2 11.54% 74.3807 90.8474 P3 10.92% 67.0291 85.4057 3.5.2. SensitivityAnalysis The SRCs for the case study are shown in Figure 11. The uncertainty in the previous seven factors explainsmostof thevariance in the cooling load forecasting that isobserved inFigure10b. Theremainingoneuncertain factorRH(h–2)has littleornoeffectonthe loadforecasting. 0.290 -0.425 -0.611 -0.217 0.256 -0.253 -0.200 -0.064 -0.700 -0.600 -0.500 -0.400 -0.300 -0.200 -0.100 0.000 0.100 0.200 0.300 0.400 X1 X2 X3 X4 X5 X6 X7 X8 SRC Figure 11. The SRCsof input parameters of the case study. X1=L(d–1, h); X2=T(h);X3=T(h–1); X4=T(h–2);X5=T(h–3);X6=RH(h);X7=RH(h–1);X8=RH(h–2). It canbe found that fromthe inputparameterswe selected, the factors thathave thegreatest impactonthe loadforecastareT(h–1)>T(h)>L(d–1,h)>T(h–3)>RH(h)>T(h–2)>RH(h–1)>RH(h–2) in turn.Obviously, the loadat thepredictedmoment ismostlyaffectedbytheoutdoor temperatureat thepreviousmoment.Duetothethermal inertiaoftheenclosure, thedisturbancecausedbythechange of theoutdoor temperaturewillnot immediatelyaffect the indoor temperature.Heat is transferred betweentheenvelopeswithdetentionandattenuation,whichaffects the loadof thebuilding inthe nextmoments. 4.Discussion Overviewing theprevious research inSection1,most focusedon theoptimizationof the load predictionmodel itself to improvetheaccuracyof loadprediction,andfewstudieshavepaidattention to the influenceofuncertaintyof theweather forecastdataonthe loadforecasting. Thispaperfillsa gapin thisaspectbyuncertaintyanalysisofweather forecastdata forcooling loadforecastingbased onMCM.Threeevaluation indexesareusedtocompare thepredictionresultsbetweenP1,P2andP3 inSection3.4. Theresults illustrate that theevaluationof the loadforecastingwith thedataprocessed bytheMCMisbetter thanthatof the loadforecastingusingthemeteorological forecastdatadirectly 225