神經網路辨識系統(二)

clc;close all;clear;y(1)= 0;k=1;u(k)= 2*rand-1;f(k)= 0.6*sin(pi*u(k))+0.3*sin(3*pi*u(k))+0.1*sin(5*pi*u(k));y(k+1)= 0.3*y(k)+f(k);for k=1:100    u(k)= 2*rand-1;    f(k)= 0.6*sin(pi*u(k))+0.3*sin(3*pi*u(k))+0.1*sin(5*pi*u(k));    y(k+1)= 0.3*y(k)+0.6*(k-1)+f(k); enduP= u(1:100);uT= 0.6*sin(pi*uP)+0.3*sin(3*pi*uP)+0.1*sin(5*pi*uP);netg= newff(minmax(uP),[20,20,1],{'tansig','tansig','purelin'},'trainlm','learngdm','msereg');[netg,tr] = train(netg,uP,uT);uQ= -1:0.1:1;simu= sim(netg,uQ);figure;plot(uQ,0.6*sin(pi*uQ)+0.3*sin(3*pi*uQ)+0.1*sin(5*pi*uQ),'b',uQ,simu,'r');legend('系統輸出','系統辨識');k=1;u(k)= 2*rand-1;f(k)= 0.6*sin(pi*u(k))+0.3*sin(3*pi*u(k))+0.1*sin(5*pi*u(k));y(k+1)= 0.3*y(k)+f(k);for k=2:100    u(k)= 2*rand-1;    f(k)= 0.6*sin(pi*u(k))+0.3*sin(3*pi*u(k))+0.1*sin(5*pi*u(k));    y(k+1)= 0.3*y(k)+0.6*y(k-1)+f(k); enda0= rand;a1= rand;Eta= 0.5;T = 0;while (T<10000)     E= 0;    k= round(98*rand+2);    simy(k+1)= a0*y(k)+a1*y(k-1);    e(k+1)= 0.3*y(k)+0.6*y(k-1)-simy(k+1);    Delta0= -1*Eta*e(k+1)*y(k);    Delta1= -1*Eta*e(k+1)*y(k-1);    a0= a0- Delta0;    a1= a1- Delta1;              E= 0.5*(1/99)*sum(e.^2);    T= T+1;endk=1;u(k)= sin(2*pi*k/250);f(k)= 0.6*sin(pi*u(k))+0.3*sin(3*pi*u(k))+0.1*sin(5*pi*u(k));y(k+1)= 0.3*y(k)+f(k);for k=2:100    u(k)= sin(2*pi*k/250);    f(k)= 0.6*sin(pi*u(k))+0.3*sin(3*pi*u(k))+0.1*sin(5*pi*u(k));    y(k+1)= 0.3*y(k)+0.6*y(k-1)+f(k); endsimu= sim(netg,u(2:100));simy= a0*y(2:100)+a1*y(1:99)+simu;figure;plot(1:101,y(1:101),'b',3:101,simy,'r',1:100,f(1:100),'g');legend('系統輸出','系統辨識','f函數');text(10,5.5,{strcat('a0= ',num2str(a0),', a1=',num2str(a1))});