Peak load forecasting using a fuzzy neural network

This paper describes electric load forecasting using a fuzzy neural network. Neural networks, though accurate in weekday load forecasting, are poor at forecasting peak loads and holiday loads. A decision system for load forecasting requires detailed analysis of data and the rule base has to be fixed heuristically for each season. The rules fixed in this way do not always yield the best forecast. This necessitates the development of a robust forecasting technique to complement the neural network to achieve a reliable forecast with improved overall accuracy. The fuzzy neural network proposed generates the fuzzy rules from historical data while learning. The adaptive rules formed this way are capable of approximating any continuous load profile on a compact set to good accuracy. In order to evaluate the performance of the fuzzy neural network model, load forecasting is performed on a utilityʹs data susceptible to large and sudden changes in the environmental conditions. The fuzzy neural network is compared with a neural network model on two-year utility data to obtain a one-day-ahead peak load forecast and forecast results for the months of December and May are shown to validate the effectiveness of the above approach.