Load forecasting using wavelet fuzzy neural network

Abstract

This paper presents a Wavelet Fuzzy Neural Network (WFNN) that takes the fuzzified wavelet features as inputs to Fuzzy Neural Network. This network is constructed from the fuzzy rules which are modified form of the fuzzy rules of Takagi-Sugeno fuzzy model. The number of fuzzy rules is found from the fuzzy curve approach. As the output of the model is the forecasted demand, we need a fuzzy curve corresponding to each input-output. The model can forecast hourly load with a lead time of one hour as this work is concerned with short-term electric load forecasting. Electric Load demand data and weather variables are procured from Northern Region Load Dispatch Centre and Mausam Bhawan, Delhi (India) respectively. The results of the network are compared with ANFIS and other conventional methods. The performance of the proposed Wavelet Fuzzy Neural Network is found to be superior to all others compared.