IFAC PapersOnLine 51-31 (2018) 279–284 ScienceDirect ScienceDirect Available online at www.sciencedirect.com 2405-8963 © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Peer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2018.10.050 © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. 1. INTRODUCTION Electric vehicles have positive social impact and greatly increased potential to revive as the dominant means of transportation in the coming decades (G¨ unther et al., 2015). Batteries and battery management system (BMS) play an important role in the performance of these vehicles, such as acceleration, driving range (Lu et al., 2013). One crucial task of the BMS is to evaluate the current amount of energy stored in the battery, its power capa- bility, and health for an optimal battery performance and reliable operation of electric vehicles. These all require a reliable method to obtain state of charge (SOC) in real time (He et al., 2012). Battery is a complex electrochem- ical system. The SOC involves in intrinsic electrochemical reaction process, and it cannot be measured directly. This means that it has to be estimated making use of available measurements of the BMS, such as voltage and current. Estimation of the SOC has been widely discussed in the literature, for example, see Rahimi-Eichi et al. (2014), Xiong et al. (2014), Wang et al. (2007), and references therein. One of the most straightforward methods to obtain the SOC is the Coulomb counting (Ng et al., 2009). It is simple, but an accurate estimation of SOC is difficult to be achieved just by the Coulomb counting method, ⋆ This work is supported by China Automobile Industry Innovation and Development Joint Fund (No. U1564213) and China Scholarship Council (No. 201706125075). Correspondence: Linhui Zhao (e-mail: zhaolinhui@hit.edu.cn). mainly because of initialization and accumulated errors. One solution for the initialization error is to establish a relationship between SOC and open circuit voltage (OCV) in advance. The OCV can be obtained by measuring the terminal voltage of a battery after a long time resting or by real time estimation algorithms (Zhang et al., 2014), and then the initial SOC is available. However, this requires long resting periods and high precision of the measured voltage. The accumulated errors are mainly caused by current sensor biases and noises. Due to the open loop characteristic of these methods, the estimated SOC may be unreliable if some of the measured signals are inaccurate. In order to improve the estimation accuracy of the SOC, many different model-based methods have been devel- oped and published in recent years. Kalman filter (KF), which is a classical state estimation approach, has been applied to estimate the SOC (Mastali et al., 2013), and some extended Kalman filter (EKF) techniques based on nonlinear equivalent circuit models have been employed (Sepasi et al., 2014) to improve the estimation accura- cy further. These methods can enhance the robustness and estimation accuracy of the SOC. Nevertheless, they suppose that both covariance matrices of the process and measurement noises are prior known. In addition, it needs to transform a nonlinear battery model into a linear one by linearizing. As a battery system is significantly nonlinear and its electrochemistry characteristics are very complex, model errors related to currently estimated state may be introduced by linearization, and the convergence is Keywords: Electric vehicles; Batteries; State of charge; Sliding mode observer; Input-to-state stability analysis. Abstract: Knowledge of state of charge (SOC) is extremely important for electric vehicle batteries. This paper proposes a sliding mode observer for reliable and real-time estimation of the SOC. Nonlinear dynamics of the battery is considered, and parametric and modelling uncertainties are modelled as additive disturbances. Robustness performance of the proposed observer is guaranteed using input-to-state stability (ISS) theory. This indicates that the proposed observer is robust against parametric and modelling uncertainties, and can guarantees an upper bounded estimation errors. A designed method of the observer gains is presented following the stability analysis result. The proposed observer is implemented in an embedded hardware based on Freescale MPC5554, and is validated using datasets from a lithium-ion battery under different temperatures. The robustness of the observer against model parameter uncertainties, sensors biases, and measurement noises, which may appear in real vehicles, is evaluated. The experimental and robustness testing results confirm that the proposed observer achieves good performance on estimation accuracy, real time, and robustness. * Department of Control Science and Engineering, Harbin Institute of Technology, Harbin 150001 China (e-mail: zhaolinhui@hit.eud.cn); ** China First Automotive Works (FAW) Group Corporation New Energy Vehicle Branch, Changchun 130122 China (e-mail: jgh xny@faw.com.cn). Linhui Zhao * , Huihui Li * , Guohuang Ji ** , and Zhiyuan Liu * A Robust Estimation of State of Charge for Electric Vehicle Batteries