MIMO narxnet for future predictions. HOW!!

Here is my problem, I have a code that is running but I'm not sure that I do understand the concept because the results don't look like what I'm expecting. Here is my code:

x = [cost(1:24,:);Tamb24(1:24,:);SOC_old24(1:3,:);P_old24(1:3,:)];

xx = cell(1,nn);

PP = cell(1,nn);

for i = 1:nn

    xx{i} = x(:,i);

    PP{i} = P24(1:24,i);

end

net = narxnet(1:2,1:2,5);

[Xs,Xi,Ai,Ts] = preparets(net,xx,{},PP);

net = train(net,Xs,Ts,Xi,Ai);

[Y,Xf,Af] = net(Xs,Xi,Ai);

perf = perform(net,Ts,Y);

[netc,Xic,Aic]= closeloop(net,Xf,Af);

xnew = [cost(1:24,:);Tamb24(1:24,:);SOC_old24(1:3,:);P_old24(1:3,:)];

n = 100;

xxnew = cell(1,n);

for i = 1:n

    xxnew{i}  = xnew(:,i);

end

y2 = netc(xxnew,Xic,Aic);

yy = cell2mat(y2);

figure(2)

plot(1:24,P24(1:24,4))

hold on

grid on

plot(1:24,yy(:,4))

legend('Measured','Model')

hold off

Sometimes I run this code and get really very good R values and the parity plot looks nice, but the plot of the predicted data vs. the actual one is really terrible! Is there something wrong with my understanding of how this should work?

[Greg Heath]

LINE 6:

                     P24 IS UNDEFINED!

Answer 1

[Moataz Sheha]

And by the way, I have already tried a lot of variations in the input delay, feedback delay, and the number of nodes.

[Greg Heath]

PLEASE MOVE YOUR COMMENT TO A COMMENT BOX. THEN DELETE IT FROM THE ANSWER BOX

THANKS