A Comprehensive Multivariate Wind Speed Forecasting Model Utilizing Deep Learning Neural Networks

Predicting wind speed accurately is essential for the efficient generation of wind power. To enhance the precision of wind speed forecasting, this paper proposes a multivariate combinatorial model based on a deep learning neural network, which incorporates not only historical wind speed data but also relevant meteorological features. Initially, the feature extraction of meteorological features related to wind speed is first performed using an autoencoder and singular value decomposition. Subsequently, the complementary ensemble empirical mode decomposition and wavelet transform method is utilized to mitigate noise in the wind speed series. Finally, this paper utilizes a gated recurrent unit (GRU) deep learning neural network for predicting the wind speed series. The optimization of the GRU's four hyperparameters is accomplished through the implementation of the improved gray wolf algorithm. This paper evaluates and validates the predictive performance of the model using two datasets. The experimental results demonstrate that the mean absolute percentage error of the proposed model's 1-step predictions on the two datasets is 0.7532% and 0.5263%, with corresponding root mean square error values of 0.0283 and 0.0227, respectively. These results indicate a significant improvement over those achieved by other models under comparison.