Short-Term Wind Speed Prediction Model Based on Hybrid Decomposition Method and Deep Learning

Abstract

Wind power, as an important component of distributed power grid integration, plays a vital role in the establishment of a robust power grid. However, the size and direction of wind speeds are random and intermittent, posing significant challenges to the integration of wind power into the grid. To address this issue, this article proposes a highly accurate hybrid optimized wind speed prediction model (HOWSPM) by combining techniques such as data noise processing methods, intelligent optimization algorithms, and deep learning models. First, HOWSPM utilizes the Rime optimization algorithm (RIME) to optimize the variational modal decomposition (VMD) and obtain the RIME-VMD data decomposition model. Second, the RIME-VMD decomposition model is employed to preprocess the nonlinear wind power data, resulting in 10 modal eigencomponents. Additionally, the fruit fly optimization algorithm (FOA) is applied to determine the optimal hyperparameters of the bidirectional long-short memory network (Bi-LSTM), leading to an optimized Bi-LSTM network. Finally, experiments are conducted using the optimized Bi-LSTM network for feature extraction and training on the 10 types of modal data. The experimental results show that the RMSE, MAE, MAPE, and R² of HOWSPM were improved by an average of 36.04%, 42.42%, 23.65%, and 3.09%, respectively, across the four sites. Experimental results indicate that the proposed HOWSPM model effectively enhances the accuracy of wind speed prediction, thereby improving the efficiency of wind power grid integration.