Short-Term Load Forecasting Based on NARX and Radial Basis Neural Networks Approaches for the Jordanian Power Grid
Mohammed A. Momani, Wasseem H. Alrousan, Amin T. Alqudah |Pages: 81-93|

Abstract— This paper presents two techniques for short-term load forecasting (STLF) based on Artificial Neural Networks method (ANN). These techniques are the nonlinear auto regressive with external input (NARX) and radial basis function (RBF). The results from both methods are compared in order to attain minimum percentage errors. Input data implies weather factors such as temperature and humidity. A comparison between the two techniques shows that RBF method has a better performance that NARX method in short periods training whereas NARX has the advantage in long periods training. The comparison between hourly actual and forecasted load readings shows a reasonable normalized mean square error (NMSE) with minimum values in summer: 3.9 % for NARX and 3.5% for RBF, and in winter: 3.5% for NARX and 3.47% for RBF. Results show that the minimum error is achieved by using five training days for summer and nine days for winter.