Short-term wind power prediction model based on CEEMDAN and an improved time convolutional network
DOI:DOI: 10.19783/j.cnki.pspc.210252
Key Words:wind power prediction  complete total empirical mode decomposition based on adaptive white noise  time convolution network  time mode attention mechanism
Author NameAffiliation
ZHAO Lingyun 1. College of Electrical Engineering, Sichuan University, Chengdu 610065, China
2. Deyang Power Supply Company, State Grid Chengdu Electric Power Company, Deyang 618000, China
3. Chengdu Power Supply Company, State Grid Chengdu Electric Power Company, Chengdu 610000, China 
LIU Youbo 1. College of Electrical Engineering, Sichuan University, Chengdu 610065, China
2. Deyang Power Supply Company, State Grid Chengdu Electric Power Company, Deyang 618000, China
3. Chengdu Power Supply Company, State Grid Chengdu Electric Power Company, Chengdu 610000, China 
SHEN Xiaodong 1. College of Electrical Engineering, Sichuan University, Chengdu 610065, China
2. Deyang Power Supply Company, State Grid Chengdu Electric Power Company, Deyang 618000, China
3. Chengdu Power Supply Company, State Grid Chengdu Electric Power Company, Chengdu 610000, China 
LIU Daiyong 1. College of Electrical Engineering, Sichuan University, Chengdu 610065, China
2. Deyang Power Supply Company, State Grid Chengdu Electric Power Company, Deyang 618000, China
3. Chengdu Power Supply Company, State Grid Chengdu Electric Power Company, Chengdu 610000, China 
LÜ Shuang 1. College of Electrical Engineering, Sichuan University, Chengdu 610065, China
2. Deyang Power Supply Company, State Grid Chengdu Electric Power Company, Deyang 618000, China
3. Chengdu Power Supply Company, State Grid Chengdu Electric Power Company, Chengdu 610000, China 
Hits: 2261
Download times: 403
Abstract:In recent years, wind power has gradually become a key part of renewable energy generation. In this paper, an effective short-term wind power forecasting combination model is proposed by combining complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and an improved temporal convolutional network (TCN) to improve the accuracy of short-term wind power prediction. First, CEEMDAN is used to decompose the wind power series to obtain the subsequence components, and the subsequence components are combined with the data of key meteorological variables to form the training set. Then, the time convolution network based on temporal pattern attention (TPA) is used to predict the subsequence components, and the final prediction value is obtained after reconstructing the prediction results. The whole prediction process helps to accurately describe the component characteristics of wind power, and capture the correlation between variables through the TPA mechanism, and this effectively improves the prediction accuracy of wind power. This work is supported by the National Natural Science Foundation of China (No. 51977133).
View Full Text  View/Add Comment  Download reader