BP神经网络PID控制在空调系统中的应用

:2010-03-1228420114:1006-9348(2011)04-0149-03BPPID张义1,左为恒1,林涛2(1.,400030;2.,610081):,,,BPPID,BPBP,PIDBPPIDBPPID,,,:;;:TP273:AApplicationofBPNeuralNetworkPIDControlinAir-ConditioningSystemZHANGYi1,ZUOWei-heng1,LINTao2(1.StateKeyLaboratoryofPowerTransmissionEquipment&SystemSecurityandNewTechnologyofChongqingUniversity,Chongqing400030,China;2.SouthwestMunicipalEngineeringDesign&ResearchInstituteofChina,ChengduSichuan610081,China)ABSTRACT:Fortheproblemofcontrollinganair-conditionedroomwithmulti-interference,largeinertiaandhighlynonlinear,thetraditionalcontrolstrategyhasdisadvantagesincontrolprecision,sensitivityandsystemstability,andalsohaslargeenergyconsumption.InthispaperanewPIDcontrolschemebasedonBPneuralnetworkwasproposed.TheweightsoftheBPnetworkareconditioned,andthenthePIDparametersonlineadjustmentistobeachieved.Thesimulationresultsshowthattheschemeismorestability,ofsmallerovershootsandshortermodelingtimethanBPneuralnetworkorPIDcontrolsystem.TheBPneuralnetworkPIDcontrolsystemismoresuitableforthecomplexair-conditioningsystemcontro.lKEYWORDS:Variableairvolumeair-conditioningsystem;Neuralnetwork;Neuralnetworkself-adjustingPID1,,,,[2],,,,,,,,[3-5]PID,,,PID,PIDKpKiKd,BP,,BPPID,1492,,,,,a=36m,b=30m,c=45m,20mm,370mm,20mm,1/R=1820W/,26,,175,GS01678m3/s,N=10/h1,,:mcdtnd=Gssc(ts-tn)+qn+q1+q2+q3(1):dtnd;m,c(kg)(J/(kg));Gs(m3/s);s(kg/m3);ts,tn;q1q2q3(kW),:qi=t0-tnRi,(i=1,2,3;R;t0);qn(kW)1R=1R1+1R2+1R3,(1):mcdtnd=Gssc(ts-tn)+t0-tnR+qn(2),dtnd=0,,:Gs0sc(ts-tn0)+to0-tn0R+qn0=0(3)0:tn=tn0+tnGs=Gs0+Gst0=to0+t0qn=qn0+qn(2),:mcdtnd+Gs0sctn+tnR=scGs(ts-tn0)+t0R+qn(4),:T1stn(s)+tn(s)=K1Gs(s)+K2qf(s)(5)qf=qn+t0R,(kW):W1(s)=tn(s)Gs(s)=K1Ts+1(6):W2(s)=tn(s)qf(s)=K2Ts+1(7),T=Rmc1+RGs0sc,s;K1,/(m3/s);K2,/kW:K1=ts-tn0Gs0K,K2=1Gs0scsK,KK=11+52N1a+1b+1c,a,b,c,N,:T=540s,K1=-064/(m3/s)),K2=002/kW3BPPIDPID:u(k)=u(k-1)+Kpe(k)-e(k-1)+Kie(k)+kd[e(k)-2e(k-1)+e(k-2)](8)KpKiKd,(8):u(k)=fu(k-1),Kp,Ki,Kd,e(k),e(k-1),e(k-2)(9),fKpKiKd,u(k-1),e(k),BPBPPID2,:1)PID:;2),150PIDKpKiKd,2BPPIDBPPIDBP,M,Q3,33BPPID:xii=1,2,M,M:netj(k)=iwjixiyj(k)=f(iwjixi)=f(netj(k))j=1,2,Q(10),:netl(k)=jvljyjzl(k)=f(jvljyj)=f(netl(k))l=1,2,3(11)BPPID:Kp(k)=z1(k)Ki(k)=z2(k)Kd(k)=z3(k)(12):E(k)=12r(k)-q(k)2=12e2(k)(13),E,,:vlj(k)=-E(k)vlj+vlj(k-1)(14),,E(k)vlj=E(k)q(k)q(k)u(k)u(k)zl(k)zl(k)net1(k)netl(k)vlj(15)netl(k)vlj=yj(k)(16)q(k)u(k),sgnq(k)u(k),(8)(12)u(k)z1(k)=e(k)-e(k-1)u(k)z2(k)=e(k)u(k)z3(k)=e(k)-2e(k-1)+e(k-2)(17),BP:vlj(k)=lyj(k)+vlj(k-1)l=e(k)sgnq(k)u(k)u(k)zl(k)gnetl(k)(18),:wji(k)=jxi(k)+wji(k-1)j=fnetj(k)3l=1lvlj(k)(19),g=g(x)1-g(x),f=1-f2(x)/2BPPID,:1)BP,MQ;2)r(k)q(k),e(k)=r(k)-q(k);3)e(i),r(i),q(i)u(i-1)(i=k,k-1,,k-p),;4)BP,BPPIDKp,Ki,Kd;5)PIDu(k),;6)(13),,,;7)vlj(k),wji(k);8)k=k+1,4BPPID(10)(19)MATLABm,BP,4,103,Kp,Ki,Kd,(220)151:[1],.[M].:,200159-202.[2],.PID[J].,1999,2(13):57-60.[3].[J].,2009,1(36):186-188.[4],,.[J].,2003,5(30):113-114.[5],.PID[J].,2004,9(20):16-18.[](1952-),(),,,(151)Sigmoid,SigmoidPIDPID;32,+5,-5,44BPPID,555BPPID,2:1500s,0,0,0BPPID,PID,,,,,,BPPID:[1],.[J].,2007,(11):115-117.[2],,.[M].:,2005.[3].[M].:,2002.[4],.VAV[J].,2006,(1):12.[5].[M].:,2004.[6].MATLAB[M].:,2009.[](1983-),(),,,(1961-),(),,,,(19833-),(),,,220