A wavelet based prediction of wind and solar energy for Long-Term simulation of integrated generation systems

The wavelet analysis give us a power tool to achieve major improvements on neural networks design, especially on predictive models for semi-periodic signals, as for wind speed survey or solar radiation prediction. The compressed signal coefficients set can be used to properly modify the adaptive amplitude structure of the recurrent learning algorithm for a predictive neural network. In this paper a biorthogonal wavelet decomposition was used to extract a shortened number of non-zero coefficients from a signal representative of wind speed and solar radiation sampled trough time.