BAYESIAN ENHANCED MODIFIED FORECASTING APPROACH: APPLICATION TO WIND POWER SERIES

J.A. PUCHETA; RODRIGUEZ RIVEROS, CRISTIAN M.; PATIÑO, H. DANIEL

Buenos Aires, Argentina

Congreso; 25º Congreso Argentino de Control Automático; 2016

AADECA - Asociación Argentina de Control Automático

In this paper, we propose the Bayesian Enhanced modified (BEMA) predictor filter to forecast wind power series. Wind power forecasting is a complex, multi-dimensional, and highly non-linear system. Artificial Neural networks are able to learn the relationship between system inputs and outputs without mathematical conversion, and perform complex non-linear mapping, data classification and prediction. The goal of this work it to implement the BEMA approach to design a wind power forecasting system, with particularly interest in short-term prediction by using the data model selected, which is used to extract information to make prediction. The conducted results show that this method can be used to improve the predictability of short-term wind time series with a suitable number of neural nets parameters using a heuristic method based on Kullback?Leibler divergence compared to that of reported in the literature.