神经网络内模控制

%load trule %clc;clear all; load inv_NET global percent1 percent2 global r percent1=0.15; percent2=-0.05; n=24;ff=13;landa=30001.1;alfa=2.317; %操作条件 UA=9803*3.6; Ps=0.1013; F=100; %Zf=0.5; flag=1; beta=0.75; options=odeset('RelTol',1e-6,'AbsTol',1e-9); X0=[0.89135 0.87175 0.84764 0.81825 0.78294 0.74141 0.69391 0.64145 0.58582 0.52937 0.47458 0.42354 0.37765 0.3323 0.28676 0.2429 0.20245 0.16661 0.13599 0.11057 0.08992 0.073403 0.060311 0.049978 0.5 0.3195]; [t,X2]=ode15s('m_dync2a',0:0.01:8.02,X0,options);%step change Xt=X2(:,1);%save the change data of the top level of ITCDIC Xb=X2(:,n);%save the change data of the bottom level of ITCDIC %figure(1) %plot(t,Xt); %figure(2) %plot(t,Xb,'r'); for k=150:1:length(X2) X2(k,n+2)=0.3195*(1+percent2); end Pr=X2(:,n+2); for k=150:1:length(X2) X2(k,n+1)=0.5*(1+percent1); end q=X2(:,n+1); figure(11) plot(t,Pr,'g',t,q,'k'); Y0(1)=alfa*X0(1)/((alfa-1)*X0(1)+1); output(:,1)=[Y0(1);X0(n)]; y(:,1)=output(:,1);%+rand(1)*10^-4; for a=2:1:800 u(:,a+1)=[Pr(a+1);q(a+1)];%预测的输出 if a==2 X=X0; end if a==2 Yt(a-1)=alfa*Xt(a-1)/((alfa-1)*Xt(a-1)+1); P_output(:,a-1)=[Yt(a-1);Xb(a-1)]; g(:,a-1)=P_output(:,a-1); end u(:,a)=[Pr(a);q(a)]; X(n+1)=u(2,a); X(n+2)=u(1,a); if a==2 [t3,X3]=ode15s('tmodel',[0,0.01*(a-1)],X,options); else [t3,X3]=ode15s('tmodel',[0.01*(a-2),0.01*(a-1)],X,options); end X=X3(length(t3),:); Y(1)=alfa*X(1)/((alfa-1)*X(1)+1); output(:,a)=[Y(1);X(n)]; y(:,a)=output(:,a);%+rand(1)*10^-4; %[t,X2]=ode15s('tmodel',0.01*a:0.01:0.01*(a+1),X,options); %X=X2(length(t),:); %output(:,a)=[X(1);X(n)]; %y(:,a)=output(:,a)+rand(1)*10^-4; load NET %load datan inputf(:,a)=[Pr(a-1);q(a-1);u(1,a);u(2,a);output(1,a-1);output(2,a-1);output(1,a);output(2,a);abs(output(1,a-1)-P_output(1,a-1));abs(output(2,a-1)-P_output(2,a-1));abs(y(1,a-1)-r(1));abs(y(2,a-1)-r(2))]; %series structure %inputn_test=mapminmax('apply',inputf(:,a),inputps); %an=sim(net,inputn_test); an=sim(net,inputf(:,a)); %P_output(:,a)=mapminmax('reverse',an,outputps); P_output(:,a)=an; %P_output(:,a+1)=[0;0]; g(:,a)=g(:,a-1)+(1-beta)*[r(:)-y(:,a)]; input(:,a)=[Pr(a);q(a);y(1,a-1);y(2,a-1);y(1,a);y(2,a);r(1);r(2);g(1,a);g(2,a)]; end %for a=3:1:800 %output(:,a)=[Pr(a);q(a)]; %P_output(:,a)=[0;0]; %%r(:,a)=[0.98;0.02]; %r=[0.98;0.02]; %y(:,a)=output(:,a); %input(:,a-1)=[Pr(a);q(a);Xt(a-1);Xb(a-1);Xt(a);Xb(a);r(1);r(2);r(1)-(y(1,a)-P_output(1,a));r(2)-(y(2,a)-P_output(2,a))]; %end inputdata=input(:,2:a); outputdata=u(:,3:a+1); input_train=inputdata; output_train=outputdata; input_test=inputdata; output_test=outputdata; %[inputn,inputpsb]=mapminmax(input_train); %[outputn,outputpsb]=mapminmax(output_train); %inputn_test=mapminmax('apply',input_test,inputpsb); %an=sim(invnet,inputn_test); bn=sim(invnet,inputdata); %P_output=mapminmax('reverse',an,outputpsb); P_output1=bn; error=abs(P_output1-output_test); figure(7) plot(P_output1(1,:),'.-r','Linewidth',2) hold on plot(output_test(1,:),'Linewidth',2) hold off figure(8) plot(P_output1(2,:),'.-g','Linewidth',2) hold on plot(output_test(2,:),'Linewidth',2) hold off figure(9) plot(error(1,:)) figure(10) plot(error(2,:))