A Model-Based Approach for Voltage and State-ofCharge Estimation of Lithium-ion Batteries

Electric vehicles are equipped with a large number of lithium-ion battery cells. To achieve superior performance and guarantee safety and longevity, there is a fundamental requirement for a Battery Management System (BMS). In the BMS, accurate prediction of the State- of-Charge (SOC) is a crucial task. The SOC information is needed for monitoring, controlling, and protecting the battery, e.g. to avoid hazardous over-charging or over-discharging. Nonetheless, the SOC is an internal cell variable and cannot be straightforwardly obtained. This paper presents a Kalman Filter (KF) approach based on an optimized second-order Rc equivalent circuit model to carefully account for model parameter changes. An effective machine learning technique based on Proximal Policy Optimization (PPO) is applied to train the algorithm. The results confirm the high robustness of the proposed method to varying operating conditions.

Automated summary

This paper presents a Kalman Filter approach based on an optimized second-order Rc equivalent circuit model to accurately predict the State-of-Charge (SOC) of electric vehicle batteries. The algorithm is trained using a machine learning technique and shows high robustness under varying operating conditions.