Page - 90 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2019,12, 57 Weappliedthe three typical loads to investigate their impactsontheoptimizationof thehybrid powersystembytuningthecomponentsizesandpowermanagementstrategies. 3.1.HouseholdLoad Applyingthehousehold load(seeFigure5) to theoriginal systemlayout (b, s,w)= (8,10,2)and management settingsof (SOClow,SOChigh)= (40%, 50%)gave the system’s referenceplot shown in Figure6a,where the systemcostwasestimatedas J=1.300$/kWhwithLPSP=4.89%(seeStep1 of Table 5). FromFigure 6a, the systemcost can be reduced to J=1.169 $/kWhbyadjusting the componentsas (b, s,w)= (18,9,2)butwithapossiblepowercut (LPSP=2.61%,seeStep2ofTable5). If the requirementwasLPSP=0, then theoptimal systemcostwas J=1.189 $/kWh, achievedby setting (b, s,w)= (18,10,2) (seeStep3ofTable5). That is,wecanreduce thesystemcost from J=1.300 to1.189$/kWh,while improvingthesystemreliability fromLPSP=4.89%to0. ȱ (a)ȱwȱ=ȱ2ȱandȱ(SOClow,ȱSOChigh)ȱ=ȱ(40%,ȱ50%)ȱ (b)ȱwȱ=ȱ0ȱandȱ(SOClow,ȱSOChigh)ȱ=ȱ(30%,ȱ40%)ȱ Figure6.Thereferenceplots for thehousehold load. Table5.Theoptimaldesignprocedure for thehouse load. (b,s,w) (SOClow,SOChigh) LPSP (%) J ($/kWh) Step1 (8,10,2) (40%,50%) 4.89% 1.300 Step2 (18,9,2) (40%,50%) 2.61% 1.169 Step3 (18,10,2) (40%,50%) 0% 1.189 Step4 (15,15,0) (40%,50%) 0% 0.822 Step5 (15,15,0) (30%,40%) 0% 0.810 Step6 (23,15,0) (30%,40%) 0% 0.794 Step7 (23,15,0) (30%,40%) 0% 0.794 Optimal (23,15,0) (30%,40%) 0% 0.794 Because thecostofwindenergywasmuchhigher thanthecostof solarenergy(seeTable2)and thecompensationeffectswerenotsignificant(seeFigure4), theuseofsolarandaPEMFCwithchemical hydrogenproductionwasviewedas economically efficient for thegreenbuilding [32]. Therefore, wesetw=0andtheresultingoptimizationshowedthat thesystemcostcanbesignificantlyreduced to J=0.822$/kWhbysetting (b, s,w)= (15, 15, 0), as illustrated inStep4ofTable5. Furthermore, whenwefixed the component settings of (b, s,w) = (15, 15, 0) and tuned thepowermanagement strategies (SOClow,SOChigh)= (30%,40%), thesystemcostwasfurtherdecreasedto J=0.810$/kWh (see Step 5 of Table 5). Steps 6 and 7 illustrate the iterative tuning of component size andpower management, respectively. Theresults indicatedthat thesystemcostconvergedto J=0.794$/kWh with(b, s,w)= (23,15,0)and(SOClow,SOChigh)= (30%,40%).Comparedwiththeoriginalcost, thecost wasreducedby38.9%,whilemaintainingcompletesystemreliability.Note that the iterativemethod cangreatlyreduce thecomputationtimebecause thesimultaneousoptimizationof fourparameters (b, s,SOClow,SOChigh) tookmuchlonger than iterativeoptimization,as indicated in [45]. Therefore, 90