Page - 218 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1900 MAE= 1 N |yˆi−yi| (20) RMSE= √ 1 N∑ N i=1(yˆi−yi)2 (21) whereyi is thereal load, yˆi is the forecasting load,andN is thenumberofsamples. MAPEnotonlyconsiders theerrorbetweenthepredictedvalueandthe truevaluebutalso the ratiobetweentheerrorandthe truevalue. It isameasureof theaccuracyof the totalprediction in the statisticalfield [32].MAEandRMSEcanamplify thevalueof the largerpredictionbias,whichcan compare thestabilityofdifferentpredictionmodels. 3.2.DataSourcesandCollection The data used in the researchwork of this papermainly include two parts: meteorological dataand loaddata. For the load forecastingwork, therearemainly twoways toobtain theenergy consumptionof theconstruction. First, it isobtainedthroughtesting. Inaddition, it is calculatedusing energysimulationsoftware. For thefirstapproach,due to the lowlevelofoperationandmanagement techniquesof thecurrentHVACsystem, it isusually regulatedbytheoperatingexperienceofworkers. Theadjustmentof theHVACsystemhasacertain lag,whichcannot immediatelybringchanges to the load even if it is adjusted according to actual conditions. In the case of fluctuations of indoor temperature, the loaddataobtainedbythismethoddonotreflect the impactof real-timechanges in meteorologicalparameters. Therefore, thispaperadopts thesecondmethod.WeuseDesignBuilder to simulate the cooling load of the building, analyze the relevant data and establish themodel. Themeteorological data used in this paper are composed of real-timeweather data collected by a small weather station shown in Figure 3 andweather forecast data from theweatherwebsite (https://www.worldweatheronline.com). Table1showsthemeasurement informationof theweather elements recordedbythestation. Figure3.Meteorological station. Table1.Measurement informationof themeteorologicalelements. MeteorologicalElement MeasuringRange ResolutionRatio Accuracy Dry-bulb temperature −50~+100 ◦C 0.1 ◦C ±0.2 ◦C Relativehumidity 0~100% 0.1% ±2%(≤80%)±5%(>80%) 218