WAVELET DECOMPOSITION AND NEURO-FUZZY HYBRID SYSTEM APPLIED TO SHORT-TERM WIND POWER FORECASTING

摘要

This paper presents a new statistical short-term wind power forecasting model based on wavelet decomposition and neuro-fuzzy systems optimized with a genetic algorithm. The forecasted variable is the mean electric power production in a wind farm corresponding to half hour intervals. The forecasting horizons range from 0.5 to 4 hours. The optimization process, ruled by the genetic algorithm, selects the proper inputs and the parameters needed by a clustering algorithm to obtain after training, the neuro-fuzzy system with the lowest forecasting errors. The forecasting results obtained with the final models have been compared to those obtained with traditional forecasting models showing a better performance for all the forecasting horizons.