Research on Combination Forecast of Ultra-short-term Wind Speed Based on CEEMDAN-PSO-NNCT Multi-model

Abstract

An ultra-short-term wind speed prediction method based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) decomposition using Particle Swarm Optimization (PSO) to optimize the No Negative Constraint Theory (NNCT) to determine the combined model weights is proposed. First, CEEMDAN is used to decompose the wind speed time series into several sub-components to reduce the non-stationary characteristics of the wind speed time series. Then calculate the sample entropy of each component, and construct a combination prediction model based on the no negative constraint theory for the components with higher sample entropy and use particle swarm optimization algorithm to determine the optimal combination weighting factor. Least Squares Support Vector Machines (LSSVM)prediction models were established for the remaining components. The predicted values of all components are superimposed to obtain the final predicted wind speed. Taking the actual wind speed data of a wind farm in Northwest China as an example, the validity of the proposed model is verified. Compared with the CEEMDAN-LSSVM prediction model, the root mean square error is reduced by 3.1%.