Short-term electrical load forecasting using radial basis function neural networks considering weather factors

In the recent years, the demand for electricity is growing rapidly and the accuracy of load demand forecast is crucial for providing the least cost and risk management plans. In the competitive power market, utilities tend to maintain their generation reserve close to the minimum required by the system operator. Load forecasting has many applications including energy purchasing and generation, load switching, contract evaluation, and infrastructure development. This creates the requirement for an accurate day-ahead instantaneous load forecast. Therefore, in this paper a novel methodology is proposed for solving the short-term load forecasting problem using the radial basis function neural networks (RBFNNs) considering the weather factors such as temperature and humidity. The RBFNN has the advantage of handling augment new training data without requiring the retraining. The hidden layer and linear output layer of RBFNNs has the ability of learning the connection weights efficiently without trapping in the local optimum. The simulation results are performed on Pennsylvania–New Jersey–Maryland (PJM) interconnection and the obtained results are promising and accurate. The simulation studies show that the forecast results are reliable, specifically when weather factors are included in the training data.