Adaptive power signal prediction by non-fixed neural network model with modified fuzzy back-propagation learning algorithm

The authors present a nonfixed artificial neural network (ANN) model with a modified fuzzy back-propagation (BP) learning rule for power signal prediction. This model is designed to avoid the ill-learning of ANN training caused by improper information. Taipower 1990-1993 loads and relevant weather data are implemented. The experiments of next day peak load forecasting and one-day-ahead hourly load forecasting are made in this study. Some experiments using conventional BP ANN approach are also performed as a comparison with the proposed model.