School of Electrical Engineering, Northeast Electric Power University, Jilin 132012, China

Wind power prediction is very important for the economic dispatch of power systems with wind power. Aiming at the problem that point prediction is difficult to describe the uncertainty of wind power, a short-term wind power interval prediction method based on combined mode decomposition and deep learning is proposed. First, the original wind power sequence is decomposed into multiple modal components by using the improved complete ensemble empirical mode decomposition with adaptive noise, and the high-frequency strong non-stationary components are decomposed again by using the variational mode decomposition. On this basis, the sample entropy is used to calculate the complexity of each component and reconstruct them into trend components, oscillation components and random components. Then, the three components are input into the Bayesian optimization bidirectional long short-term memory neural network to establish their respective prediction models, and the point prediction values of the three components are obtained. The mixed kernel density estimation method is used to estimate the error distribution of the prediction results of oscillation components and random components, and the overall interval prediction results are obtained by combining the point prediction values. Finally, the actual example analysis shows that this method has higher prediction accuracy than other models.