first of all i'm using matlab as the engine..
secondly , i have data for 6 years of load data and i arranged my training data as follows :
7 inputs (BINARY for weekday)
24 24 hours of day before (load)
24 24 = = of 2 days before (L)
24 24 = = = = week before Load
i used 5 years for training and the last year of the six as a test..

i'm using a network that consists of 1 hidden Layer with (tansig) and the output layer using the function (purelin) ..
(input--->24---->24)
i scaled the input data and output data in the end by the built in function premnmx.m (and i even once scaled it manually) i used several training functions (trainscg , traingdx , traincgf ...) and i even tried to change the hidden neuron's number
BUT NOTHING SEEMS TO WORK... PLZZZZ HELP ME I'M DROWNING HERE..

This is the code i'm using..

%% putting the data in one matrix..

input(:,:) = y200018(:,:);
input(:,245:366)= y2000912(:,:);
input(:,367:609)=y200118(:,:);
input(:,610:1:731)=y2001912(:,:);
input(:,732:974)=y200218(:,:);
input(:,975:1096)=y2002912(:,:);
input(:,1097:1339)=y200318(:,:);
input(:,1340:1461)=y2003912(:,:);
input(:,1462:1705)=y200418(:,:);
input(:,1706:1827)=y2004912(:,:);
target(:,:)=input(12:35,2:1826);
target(:,1827)=y200518(12:35,1);

%%scaling input data

[norma,minp,maxp,normt,mint,maxt] = premnmx(input,target);
%% creating BPANN

net = newff(minmax(norma),[24 24],{'tansig' 'purelin'},'traingdx');
net.trainParam.show = 50;
net.trainParam.lr = 0.01;
net.trainParam.epochs = 10000;
net.trainParam.goal = 1e-5;
net.trainParam.lr_inc = 1.05;
net.trainParam.lr_dec = 0.7;
net.trainParam.mc = 0.95;

%%training data
train(net,norma,normt)

%%and then the simulation part
pnewn = tramnmx(y2005,minp,maxp);
anewn = sim(net,pnewn);

*** the error is very high and i encounter many breakdowns all the way in my comparison plots between actual and forecasted..
can anyone help me here??.. and i'll be SOOOOOOOOO thankful..
ciao