Electrical system load forecasting with polynomial neural networks (based on combinatorial algorithm)

A polynomial neural network model for short term electrical load forecasting (STLF) is developed. Several models use past weekly and monthly system loads to forecast future electrical demands. All models are validated with actual system load data from the Azerbaijani Power Company. Combinatorial algorithm is elaborated to find efficiently the coefficients of regression type model. The paper presents the results, conclusions and points out some directions for future work.