Open-Circuit Voltage-Based State of Charge Estimation of Lithium-ion Battery Using Dual Neural Network Fusion Battery Model. (10th January 2016)

• Record Type: Journal Article
• Authors: Dang, Xuanju; Yan, Li; Xu, Kai; Wu, Xiru; Jiang, Hui; Sun, Hanxu
• Is Part Of: Electrochimica acta. Volume 188(2016)
• Page Start: 356
• Page End: 366
• Publication Date: 2016-01-10
• Subjects: State of Charge -- Dual neural network fusion battery model -- Lithium-ion battery -- Second-order battery model -- Open circuit voltage; Electrochemistry -- Periodicals; Electrochemistry, Industrial -- Periodicals; 541.37
• DOI: 10.1016/j.electacta.2015.12.001
• Languages: English
• ISSNs: 0013-4686

Abstract

Graphical abstract: In this article, the estimation method for SOC by the dual neural network fusion battery model is obtained by combining the linear neural network battery model with BP neural network. The constructed dual neural network fusion battery model consists of two neural network models connected in series. The first part is a linear neural network battery model which can be used to identify parameters of the first-order electrochemical model or second-order electrochemical model for the battery, the second part is a BP (Back of Prorogation) neural network used for capturing the relationship between OCV and SOC. Highlights: The model for the battery is described by the linear neural network battery model. BP neural network is employed for capturing the relationship between the OCV and SOC. OCV-based method for SOC estimation by using dual neural network fusion battery model is proposed. ABSTRACT: The OCV (open circuit voltage)-based method for SOC (state of charge) estimation by using the dual neural network fusion battery model is proposed in this paper. The weights of the constructed dual neural network fusion battery model can be used to describe the characteristics of the corresponding parameters of electrochemical model for the battery. The constructed dual neural network fusion battery model consists of two neural network models connected in series. The first part is a linear neural network battery model which can be used to identify parameters of the first-order electrochemical model or second-order electrochemical model for the battery, the second part is a BP (Back of Prorogation) neural network used for capturing the relationship between OCV and SOC. The DST (Dynamic Stress Test) data is adopted for training the dual neural network fusion battery model, by which the relationship between OCV and SOC is offline obtained. Under FUDS (Federal Urban Driving Schedule) condition, the experimental results show that the dual neural network fusion battery model can effectively estimate SOC based on the first-order electrochemical model or second-order electrochemical model.