Short Term Wind Power Prediction Based on Empirical Mode Decomposition and Kernel Principal Component Analysis

The randomness, intermittency, and volatility of wind power generation add a lot of pressure to grid scheduling. In order to predict the power of wind turbines and improve the accuracy of prediction, this paper proposes a prediction method that uses Empirical Mode Decomposition (EMD) combined with Kernel Principal Component Analysis (KPCA) to optimize Long Short Term Memory Network (LSTM). Firstly, fully considering the six factors that constrain the power of the wind turbine, EMD is used to decompose the six factors into a series of intrinsic mode functions, each of which represents the characteristic components at different scales. Therefore, the changes in data signals at different time scales can be obtained; Secondly, using the KPCA method to extract the key influencing factors of the feature sequence can remove the redundancy of the original sequence and reduce the dimensionality of the model input; Finally, the LSTM network is used to dynamically model the multivariate feature sequence and achieve the prediction of wind turbine power. Using historical data from a wind farm in Shanghai for sample measurement and analysis, the experimental results show that the prediction model has higher accuracy than traditional wind turbine power prediction methods.