مشخصات پژوهش

This paper illustrates application of neural network to chaotic time series prediction. Electricity load time series is modeled as chaotic time series and predicted by using MLP neural network. For the sake of training NN, LM training algorithm is used that is one of the most efficient learning mechanisms for the prediction. The LM method trains a NN 10-100 times faster than the gradient descent back propagation (GDBP) algorithm. Proposed method is examined in New York electricity market with different forecasting horizons.