Error theory study on EKF-based SOC and effective error estimation strategy for Li-ion batteries

Abstract

Accurate state-of-charge (SOC) estimation is crucial for ensuring the safe and reliable operation of battery management systems (BMS). Among the various algorithms used for SOC estimation in real-vehicle BMS, the extended Kalman filter (EKF) algorithm holds significance due to its adherence to optimal estimation principles and its property of insensitivity to initial values. By studying the relationship between error sources and SOC estimation errors, it becomes possible to develop targeted measures for enhancing the accuracy of SOC estimation based on the EKF. From a probabilistic perspective, this paper derives theoretical equations that establish the connection between SOC estimation errors and various error sources, including measured voltage, measured current, open circuit voltage curve, capacity, ohmic internal resistance, and polarization resistance. Furthermore, the paper analyzes the relationship among multiple error sources in generating SOC estimation errors. Building upon the outcomes of this theoretical analysis, a joint SOC estimation method that combines the EKF with Ampere-hour counting (AHC) is employed to identify errors. In scenarios where simultaneous faults occur in the current and voltage sensors, they are identified based on the slope and Euclidean distance of the two SOC trajectories, respectively. Subsequently, sensor faults correction and SOC compensation are implemented by leveraging simplified equations involving capacity and SOC increments and measured voltage and SOC estimation errors. In addition to addressing sensor faults, the paper also considers battery model parameter errors. By incorporating customized current pulse profile and theoretical equations relating to error sources and SOC estimation errors, a comprehensive error estimation of model parameters is achieved, capable of handling single and multiple errors. The derived simplification bridges the gap between error sources and SOC estimation errors, offering a novel approach for parameter sensitivity analysis and a theoretical foundation for quantifying these errors. The experimental results demonstrate the effectiveness and rapidity of the proposed method for identifying and correcting sensor faults and model parameter errors.