Iterative Combination of Wavelet Artificial Neural Networks for Short-Term Solar Radiation Forecasting

Julio C. Royer¹; Luiz Albino Teixeira Junior²; Edgar M. C. Franco³

¹ IFPR; ² UNILA; ³ UNIOESTE

doi:10.20906/CPS/SBSE2016-0386

Resumo

An iterative Combination of Wavelet Artificial Neural Networks to produce short-term solar radiation time series forecasting is presented in this work. A decomposition of level r, defined in terms of a wavelet basis of a given solar radiation time series is performed, generating r+1 Wavelet Components; these components are individually modeled by different Artificial Neural Networks and the best forecasts of each component are combined by means of another Artificial Neural Network; last, the combined forecasts are added, generating the forecasts of the underlying solar radiation data. An iterative algorithm is developed to search for the optimal values for the method parameters. Ten real solar radiation time series of the Brazilian system were modeled, achieving significantly better forecasting performances when compared with naive predictor, a traditional Artificial Neural Network and another hybrid Wavelet-Artificial Neural Network method.

Palavras-chave: Artificial neural networks; Forecasting; Solar power generation