http://www.ijoart.org/docs/Neural-network-based-control-of-Doubly-Fed-Induction-Generator-in-wind-power-generation.pdf

Dear Ibtissam,

For this purpose, you should determine the target and try to find the optimum parameters for the MLP model. For example, you should find the optimum data training algorithm, the number of hidden layers, the number of neurons for each hidden layer, the best transfer function for the hidden layer and output layer. Also, the most important parameter to construct an MLP neural network is the number of data for training and testing phases of the model. For more details please see my profile or contact with [email protected]

Dear researchers,

Thank you all for your feedbacks, well my question is about how can i determine the inputs/targets to train my ANN

Dear Ibtissam

What are your parameters? you should define the target, for example, the target is the energy output from the wind turbine. For this purpose, you can control the wind turbine using wind speed, wind direction, ambient pressure, ambient temperature, relative humidity and etc. Sent your project to me. After that, I can help you for developing a network.

what dataset are you using? can you add a sample?

Hello

From what I have understand, you're trying to use a Perceptron as universal approximator. You should first write your approximation as a function of a know parameters but the function is unknown (relationship between these parameters are unknown). this will give you in the end an idea about what inputs you must use. simple example: trying to approximate this function:

f(a, b) = a * sin (b) = y --------(*)

here you're inputs are two : " a " and " b " and your output is one " y ", the cost function is the error here defined maybe as your " NN output - y ", and here you will be teaching the NN to imitate the function (* ) response.

Yours

Salah Eddine Ghamri