Medium-term wind power prediction based on LSTM classification aided Pelt-Neuralprophet HHO-SVM.

The article focuses on developing a hybrid wind power prediction model called the LSTM aided Pelt-Neuralprophet HHO-SVM, designed to improve the accuracy and real-time performance of wind energy forecasting. This model integrates the Neuralprophet algorithm for seasonal regression analysis, the Pruned Exact Linear Time (PELT) method for detecting change points in seasonal patterns, and the Harris Hawk Optimization (HHO) algorithm to optimize Support Vector Machine (SVM) hyperparameters for segmented data clusters. Additionally, a Long Short-Term Memory Network (LSTM) is employed to classify incoming real-time wind data, enhancing the model’s adaptability and prediction precision. Validation using datasets from the National Renewable Energy Laboratory and multiple wind farms in China demonstrates that this hybrid approach outperforms existing models in forecasting accuracy and robustness, particularly in capturing seasonal dynamics and enabling real-time predictions.