Short Term Load Forecasting based on An Optimized Architecture of Hybrid Neural Network Model

Abstract

Nowadays the predication of electricity demand is consdired as a main criteria for estimating the electricity generation costs. Companies always sought to create balancing between the supply and demand of power. We are interested in this paper is short-term load forecasting (1 hour -1 week). The important factors that are considered, in this paper, for the electricity supply are the temperatures and time. Then, different neural network models (NN) are used and developped for increasing the accuracy of the short term load forecasting such as backpropagation, Adaptive Neuro-Fuzzy Inference System (ANFIS), wherease Wavelets Neural Network (WNN) is used as a comparative model. An optimized neural network is found by defining the suitable number of hidden layers and their nodes and an optimized membership functions number associated to each input unit is proposed in ANFIS model. A comparison between these models shows that the optimised ANFIS model outperfroms the other methods.