LSTM-based Integrated Model for Wind Power Forecasting

In this paper, an integrated prediction model based on Long Short-Term Memory (LSTM) network is proposed to address the problems of intermittency and randomness in wind power prediction. The model utilizes a large amount of real wind power generation data provided by Longyuan Power, and realizes accurate prediction of wind turbine operation results for the next 24 hours through data preprocessing, feature extraction and model training. The study demonstrates that, in comparison to a single LSTM model, the integrated model significantly improves prediction accuracy and reduces error, with a 56% decrease in the average MAE and an 80% decrease in the average MSE. This study highlights the benefits of integrated models in addressing nonlinear dynamics and intricate data structures, offering novel perspectives and approaches for wind power forecasting. Future research will focus on incorporating additional models to enhance predictive accuracy further.