Very Short-Term Load Forecasting Using a Hybrid Neuro-fuzzy Approach

The purpose of this work is to employ the Adaptive Neuro Fuzzy Inference System for performing very short-term load forecasting in power distribution substations, which can enable the development of more efficient automatic load control of electrical power load systems. The system inputs are two load demand time series, composed of data measured in five minutes intervals up to seven days from substations located in the cities of Cordeirópolis and Ubatuba - SP, Brazil. The Adaptive Neuro Fuzzy Inference System is a universal approximator that can be used in function approximation and forecasting. The results of the Adaptive Neuro Fuzzy Inference System in this paper are promising, where the average MAPE of Cordeirópolis was 0.7264% and of Ubatuba was 0.5163%.