When change the Multi-dimentional input, how to show an offset on the output in neural network?

Hi,

I am making a neural network for solar panel. I have two input: solar irradiance([200,400,600,800,1000]) and voltage(from 0 to 5V, 32 average values like a step input). The output is the Current(also 32 values.which follow the input voltage, can plot an I-V curve with the input voltage).

I have made a self-definition nueral network with two inputs, one is irradiance (200, only one value not a 5 value vector), the other is the voltage as described above. And I can get the 32-values vector current output. But I want to make a network when I change the irradiance, for example, I change the irradiance to 400, the current will also varied with it. The current will show an offset between different input irradiance. The target is I train the network with this 5 irradiance level, when I want to simulate the model, I give an input such as 900, it will give out an output curve between 800 and 1000. In the whole process, the other input "Voltage" is fixed to the 32-values vector described above.

How can I make such a model or do you have some other ideas for this problem?I am blocked by this problem for two weeks, really appreciate for your help!

Thank you !!!

Sindy

Answer 1

[Greg Heath]

Pair each of your 32 voltage values with a corresponding irradiance value to get an input matrix of size [2 32]. The corresponding target matrix of size [1 32] contains the resulting current values.

Use fitnet to represent a function which should interpolate between the mins and maxes of the input variables.

Hope this helps.

Thank you for formally accepting my answer.

Greg

[Sindy]

Hi,Greg,

You means I can make a network on Neural network fitting toolbox? And I can use different group of input data ([2 32]) to train a function? Sorry I am new in this field and dont so clear with your meaning by "fitnet".

Thank you for answering!

Sindy

[Sindy]

Thank you I think I got it. fitnet(hiddenSizes,trainFcn). Thank you very much!

[Greg Heath]

You didn't formally accept my answer.

Greg