Enhancing short-term chaotic wind speed time-series prediction using hybrid approach with multiple data sets

The rapid growth of renewable energy sources, particularly wind energy, has accentuated the importance of accurate short-term wind speed predictions for efficient energy management. This study proposes a novel approach to enhance the Short-Term prediction of chaotic wind speed time series by leveraging hybrid neural network models such as CNN (Convolutional Neural Network) and RNN (Recurrent Neural Network) across two groups of original wind speed data sets with different sampling locations for simulation experiments. The utilization of multiple data sets, representing various meteorological conditions and geographical locations, aims to enhance the model’s adaptability and robustness. The proposed method is based on Phase Space Reconstruction for short-term prediction. The wind speed time series is divided according to seasons such as JUN–AUG (summer), SEP–NOV(Autumn), DEC–FEB(Winter), and MAR–MAY (Spring) seasons for both sites. A hybrid model CNN-RNN is trained on data of each season, 70% of samples are used for training and 30% are used for testing. Furthermore, four commonly used assessment indicators are applied to evaluate the predictive performance of different models such as MSE (Mean Squared Error), MAE (Mean Absolute Error), RMSE (Root Mean Squared Error), and SMAPE (Symmetric Mean Absolute Percentage Error). Simulation results show that the proposed hybrid model performs better than the basic FFNN (Feed Forward Neural Network) and CNN (Convolution Neural Network) models.