Seite - 219 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 1900 This smallweather station is located inTianjinUniversity, China,which consists of aPC-2-T solar radiationobserver, aPC-4meteorologicalmonitoringrecorder, transducersandthemanagement software of theweather stationmonitoring system. It records theweather data every half hour by thesedevices and transfersdata to the computer viawired cables. Theweatherdata collected bythemeteorological stationmainly includedry-bulb temperature, relativehumidity,windspeed, winddirection, sunshinehours, rainfall, solar radiation intensity. In addition, among theweather parameters frommostexistingweatherwebsites, thepredictionaccuraciesof thedry-bulbtemperature and the relative humidity are relatively high,while the prediction accuracies of parameters such aswind speed, winddirection and solar radiation intensity are poor. Some cannot be predicted in advance, suchas solar radiation intensity. Most of theprevious literature selected temperature andrelativehumidityas inputs to establish thepredictionmodel [14–16]. Therefore, in thispaper, wemainlyrecordedthehourlyweather forecastdata fromtheweatherwebsite, includingdry-bulb temperatureandrelativehumidity,anddiscussedthe influenceof theuncertaintyof forecastdry-bulb temperatureandrelativehumidityonthecooling loadforecast. 3.3. TheDBModel of theOfficeBuilding Thecaseselected in thisarticle isanofficebuilding inTianjinCity, located inBinhaiNewDistrict, Tianjin,withaconstructionareaof10,723.16squaremeters,buildingheightof22.80m,5floorsabove ground,1floorundergroundandaroofsetwithskylights. ThefinalmodelcreatedbytheDesignBuildersoftwareversion4.2.0.015isshowninFigure4.DBis themostcomprehensiveGraphicalUserInterfacetotheEnergyPlussimulationenginewhichiswidely usedformodeling[35]. Parametersof thebuildingstructureareobtainedthroughresearch,andother parameters refer to“TianjinPublicBuildingEnergyEfficiencyDesignStandards” (DB29-153-2014) for setting, suchaspersonneldensity,personnelperroomrate, lightingdensity, runningtime. Theheat sourceissuppliedbythedistrictheatingpipenetworkinwinter,andtheterminaloftheairconditioning systemis thefancoil system,while ituses thesplitVariableRefrigerantVolume(VRV)airconditioning system for cooling in summer. It is difficult to obtain the hourly cooling load bymeasurement. Inaddition, theHVACsystemsof theofficebuildingarenormallyusedfromMondaytoFridayand arenotusedonweekendsandholidays. Therefore,only the loads from9a.m. to5p.m. onweekdays areconsidered in thescopeof thestudyof loadforecasting. Theerroranalysisof thesimulatedhourly heatingloadandthemeasuredheatingsupplydata from9:00to17:00for threeworkingdays iscarried out toverify thesimulation. Figure4.TheofficebuildingmodelbuiltbyDesignBuilder. Theresult is showninFigure5. Theaveragerelativeerrorbetweenthemeasureddataandthe simulateddatawas16.1%,which isacceptableconsideringof the limitationsof theon-site testsand measurement instruments. Therefore, thesimulation loadcanberegardedas thereal loadtoestablish thedatabase. 219