Short-term electric load forecasting using neural network with fuzzy set based classification

This research studies a short-term electric load forecasting technique using a multi-layer feedforward Artificial Neural Network with a fuzzy set-based classification algorithm. Based on the fact that the power system load strongly depends on the weather of the serving area, the hourly data is classified into different classes of weather condition using the concept of fuzzy set representation of weather variables. Then the set of artificial neural networks for these classes of weather condition is trained and used to perform the forecasting. The load forecasting index is also developed from the application of the fuzzy logic system. The presented technique is tested with the utility's data for various lead times ranging from 24 to 120 hours. The results indicate that the technique is able to forecast the system load with excellent accuracy and its performance does not deteriorate as the lead time becomes longer.