Journal of Computer Applications ›› 2016, Vol. 36 ›› Issue (12): 3492-3498.

### Battery SOC estimation based on unscented Kalman filtering

SHI Gang1, ZHAO Wei1, LIU Shanshan1,2

1. 1. Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang Liaoning 110016, China;
2. Faculty of Information and Control Engineering, Shenyang Jianzhu University, Shenyang Liaoning 110168, China
• Received:2016-06-03 Revised:2016-07-22 Online:2016-12-10 Published:2016-12-08
• Supported by:
This work is partially supported by the 2016 Intelligent Manufacturing Standardization Project of Ministry of Industry and Information Technology.

### 基于无迹卡尔曼滤波估算电池SOC

1. 1. 中国科学院 沈阳自动化研究所, 沈阳 110016;
2. 沈阳建筑大学 信息与控制工程学院, 沈阳 110168
• 通讯作者: 刘珊珊
• 作者简介:石刚(1978-),男,辽宁沈阳人,研究员,博士,主要研究方向:移动物联网、医疗电子与信息;赵伟(1978-),男,辽宁沈阳人,副研究员,硕士,主要研究方向:移动物联网、医疗电子与信息;刘珊珊(1990-),女,山东济宁人,硕士研究生,主要研究方向:移动物联网、电池管理。
• 基金资助:
2016工信部智能制造标准化项目。

Abstract: In order to estimate the State-Of-Charge (SOC) of automobile power lithium-ion battery online, an Unscented Kalman Filtering (UKF) algorithm was proposed combined with neural network. First of all, Thevenin circuit was treated as an equivalent circuit, the state space representation of the battery model was established and the least square method was applied to identify the parameters of model. Then on this basis, the neural network algorithm was expected to fit the functional relationships between SOC of battery and model parameters respectively. After many experiments, the convergence curve of the neural network algorithm was determined. The proposed method was more accurate than the traditional curve fitting. In addition, the Extended Kalman Filtering (EKF) principle and the UKF principle were introduced separately and some tests were designed including the validation experiment of battery equivalent circuit model, the test experiment of SOC and the convergence experiment of the algorithms. The experimental results show that, the proposed method which can be used for SOC estimation online has higher estimation precision and stronger environmental adaptability than simple extended Kalman filtering algorithm under different conditions, its maximum error is less than 4%. Finally, the proposed algorithm combining UKF and neural network has better convergence and robustness, which can be used to solve the problems of inaccurate estimation of initial value and cumulative error effectively.

CLC Number: