A variable structure neural network model for mid-term load forecasting of Iran national power system

Mid-term load forecasting is taken into account as one of the most important policies in the electricity market and brings about many financial, commercial and, even, political benefits. In this paper, artificial neural networks are represented for mid-term load forecasting of Iran national power system. To do so, the multi layer perceptron (MLP) neural network as well as radial basis function (RBF) networks are considered as parametric structures. Moreover, because of some problems such as a limitation on the number of data for training networks, the number of neurons and basis functions is also adjusted during the training process. The obtained optimal networks are used to forecast the electricity pick load of the next 52 weeks. Simulation results show the superiority of both proposed structures in the mid-term load forecasting of Iran national power system.