Improved LSTM Network Wind Speed Second-level Forecast Based on Error Modal Decomposition

The capability of frequency active support in wind power is essential for maintaining the frequency stability of systems with a high proportion of wind energy. However, the output of wind power is subject to stochastic fluctuations. Accurately estimating the current output and forecasting future regulation capabilities are critical for ensuring efficient and safe participation in active grid frequency regulation. To address this issue, this paper introduces a method for second-level wind speed forecasting using an improved Long Short-Term Memory (LSTM) network, which is based on the modal decomposition of forecast errors. This approach enhances the accuracy of wind speed predictions for the next few seconds by extracting valuable information from the errors and correcting the forecasted wind speed in real time. Ultimately, a second-level wind speed forecasting model is developed in MATLAB/Simulink, and the accuracy and effectiveness of the proposed method are validated using actual operational data from wind power systems. This research holds practical significance for the efficient and safe active frequency regulation of large-scale wind power installations.