SOC estimation and fault diagnosis framework of battery based on multi-model fusion modeling

Accurate estimation of the battery state characteristics can ensure the safe driving of electric vehicles(EVs). This paper aims to propose a new method for state of charge (SOC) estimation and fault diagnosis based on multiple equivalent circuit models(ECMs) fusion approach. This paper improves the accuracy of SOC estimation and fault diagnosis from the following four aspects. Firstly, the Thevenin model and second-order ECM are selected to describe the dynamic characteristics of the battery and the least square method is then used to determine the model parameters. Besides, the unscented Kalman filter(UKF) is employed to estimate the SOC from the two models and the Bayesian theorem is employed to determine the optimal weights for synthesizing the SOCs estimated from the two models. Moreover, the voltage residual innovation sequence(RIS) is introduced to adaptively adjust the size of the window width in real-time, which can promote the iterative performance of traditional UKF and the prediction ability of the models. Finally, an adaptive fault diagnosis framework based on multiple ECMs fusion is constructed to determine the current battery operation status based on the probability weight of each model, and also to achieve accurate early warning of the current sensor fault. The experimental results show that the SOC estimation errors are controlled within 1 %, and the battery status and current sensor fault can also be accurately diagnosed.