Short-term Wind Power Forecasting Using the Hybrid Model of Improved Variational Mode Decomposition and Maximum Mixture Correntropy Long Short-term Memory Neural Network

Abstract

With the development of emerging technology, wind power forecasting hybrid with artificial intelligence methods has become a research hotspot. Most of these methods are based on Mean Square Error (MSE) loss. However, when conducting the forecasting studies, the forecasting models built based on the traditional MSE loss have a poor effect, and the wind power data also lack the sensitivity to the nuclear parameters, make it difficult to achieve satisfactory results. Therefore, a wind power forecasting method based on Mixture Correntropy (MC) Long Short-term Memory (LSTM) neural network and Improved Variational Mode Decomposition (IVMD) is proposed in this paper. Aiming at the fact that the mixing coefficient and kernel parameters in Maximum Mixture Correntropy Criterion (MMCC) loss have an impact on its performance, Particle Swarm Optimization (PSO) algorithm is used to optimize the parameters, and PMC(PSO-MC)-LSTM model is constructed. Meanwhile, an IVMD-SE data preprocessing strategy combining Sample Entropy (SE) and IVMD is proposed. The IVMD-SE-PMC-LSTM hybrid forecasting model is constructed. Finally, four groups original data from a wind farm are simulated to verify the forecasting performance of the proposed method. The results show that the hybrid forecasting method proposed in this paper can be better applied to the forecasting with higher data complexity.