Intelligent state of charge estimation of battery pack based on particle swarm optimization algorithm improved radical basis function neural network. (June 2022)

• Record Type: Journal Article
• Authors: Zhang, Guanyong; Xia, Bizhong; Wang, Jiamin; Ye, Bo; Chen, Yunchao; Yu, Zhuojun; Li, Yuheng
• Is Part Of: Journal of energy storage. Volume 50(2022)
• Publication Date: 2022-06
• Subjects: Battery pack -- State of charge -- Feature extraction -- Particle swarm optimization -- Radial basis function neural network; Energy storage -- Periodicals; Energy storage -- Research -- Periodicals; 621.3126
• DOI: 10.1016/j.est.2022.104211
• Languages: English
• ISSNs: 2352-152X

Abstract

Highlights: Propose a novel method for battery pack SOC estimation using RBFNN improved by particle swarm optimization algorithm. The battery pack SOC under three different structures of series, parallel and hybrid connection are clearly defined and analyzed. The compressed data set highly related to battery pack SOC is obtained by using the feature extraction strategy based on VCA and PCA. Good performance of the proposed method is verified by the results of comparative analysis and robustness evaluation experiments. Abstract: As the global variable of the battery management system (BMS), the state of charge (SOC) of the battery pack represents the residual capacity of the whole battery system. High precision estimation of the battery pack SOC is the basis for realizing other functions of BMS. In this paper, the battery pack SOC under three different structures of series, parallel and hybrid connection are clearly defined and analyzed, and then the compressed data set highly related to battery pack SOC is obtained by using the feature extraction strategy based on variable correlation analysis and principal component analysis, which is used as the input of radial basis function neural network (RBFNN) to estimate battery pack SOC. Besides, the particle swarm optimization (PSO) algorithm is used to improve the RBFNN estimation model (PSO-RBFNN), which improves the estimated accuracy. It is verified that the PSO-RBFNN method has better estimation performance than RBFNN, the average absolute error and the root mean square error of PSO-RBFNN can be reduced to 0.23% and 0.34% respectively under the New European Driving Cycle. Finally, through the comparative analysis and the robustness evaluation experiments including four driving cycles and measurement noises test verify the good performance of the proposed method.