Wind Power Point Prediction Based on VMD-GWO-LSTM

Due to the strong volatility of wind power generation, it will have an impact on the safe operation of the power system, while accurate wind power prediction helps the stable operation of the power grid. In order to improve the accuracy of wind power prediction, this paper proposes a hybrid prediction model based on Variational Mode Decomposition (VMD) and using Grey Wolf Optimization (GWO) to optimize the hyperparameters of Long Shor-Term Memory (LSTM) network. Firstly, the initial power sequence is decomposed into multiple modal components by VMD and the sample entropy value of each decomposed sequence is calculated, and the sequences with similar sample entropy values are recombined to reduce the number of decomposed sequences; Secondly, an LSTM model is established for each component, and the GWO optimization algorithm is used to train the LSTM model of each component to search for the optimal hyper-parameters; Finally, the decomposed data is fed into individual optimized LSTM models to generate predictions, and these predicted values are then combined for each group. The actual data of a wind farm is simulated and the prediction results are compared with LSTM, GWO-LSTM and VMD-LSTM models. It is clear from the comparison results that the VMD-GWO-LSTM model is able to significantly improve the prediction performance with high accuracy and reduced prediction error compared to the above models.