A fuzzy inference neural network based method for short-term load forecasting

This work proposes a fuzzy inference neural network (FINN) based method for short-term load forecasting in electric power systems. FINN focuses on the classification of input and output variables and optimizes the fuzzy membership function and consequence parameter. As the classification technique, FINN makes use of the Kohonen self-organization map. Unlike the conventional methods with the classification of input variables, FINN has better performance of extracting the features of input variables due to the addition of the output variable. The proposed method is successfully applied to one-step ahead daily maximum load forecasting. To demonstrate the effectiveness, it is tested for real data of daily maximum load forecasting in electric power systems.