Time Series Narnet N points ahead Prediction or multi-step ahead

I am inexperience in NARNET or NARXNET. From online I am unable to find an example for my own task. All the examples are mainly focus on to predict only 1 point ahead. So I need help.

I am having trouble in implementing neural network to predict N points ahead. My only feature is previous time. My time series data is formed by seconds. The values are in between 30-90. They usually move 30 to 90 and 90 to 30, as seen on the example picture below. This figure is snapshot from my training set, blue is the original time series and red line is the predicted. It doesn't depends on a equation but it always acts on similar behavior.

My data set is attached.

load('17HourTrace.mat'); %myKalman17
set_size = 1000;
targetSeries = myKalman17(1:set_size);  %it can be larger
targetSeries = targetSeries';

My main goal is to predict 3 minutes ahead, which would be 180 points ahead. Because I compressed my data into 2 seconds, I have to predict 90 points ahead.

    N=90
    P(:, :)         T(:)     
   [1,2,3,4,5]     [5+N]  
   [2,3,4,5,6]     [6+N] 
%Values inside array are just indexes to continues array

Mostly I was focus on to implement neural network(gradient descent function) in octave and also use the recurrent neural network function(newelm, newcf) in the matlab. Both of those implementations didn't give me a good results. I have been told to experiment it on narnet().

My questions is that; while using narnet() how can I form my input and output data set to predict (N=90) points ahead. In default narnet() predicts only one point ahead. Since multip step ahead predictions is not efficient for long term prediction I want directly N point ahead. And how should I form my input_data set, which is P.

Any example code or suggestions for me about narnet() to prepare my input and output from my own data set to predict N points ahead will be very valuable.

[Insos]

Hi. I'm facing the same problem. Do you have solved your problem? If so, can you share the solution? Thanks!