Wind speed short-term prediction using recurrent neural network GRU model and stationary wavelet transform GRU hybrid model

This study aims to evaluate the application of the wavelet transform (WT) as a pre-processing and hybridization technique for Recurrent Neural Networks (RNN). The modeling approach presented here aims to enhance hourly wind forecasting by improving its accuracy. For this strategy of study, a model based on the Gated Recurrent Unit (GRU) was employed. We propose a methodology for integrating wavelet transforms with RNNs, along with an analysis of the potential errors arising from incorrect partition and processing of training and validation data. Ultimately, our observations suggest that employing WT as a pre-processing step for GRU input data does not yield improvements that would justify its use.