Onboard State of Charge Estimation for LiFePO4 Battery Packs Using a Hybrid Extended Kalman – Particle Filter Algorithm

In recent years, Lithium Iron Phosphate (LiFePO4) has emerged as a popular choice for Li-ion battery chemistry in electric vehicles and energy storage systems due to its inherent safety, long lifecycles, absence of cobalt and nickel, and reliance on common raw materials. State-of-Charge (SoC) is a derived parameter used for optimal and safe battery operation. The Kalman Filter-based SoC estimator is highly regarded for its accuracy and suitability in Battery Management systems. However, estimators within the Kalman Filter Family requires an accurate initial SoC to deliver precise estimates. Errors in this initial SoC can result in significant inaccuracies, prolonging the time required for the estimator to converge to the true value. The Open Circuit Voltage–SoC (OCV–SoC) method is commonly used to set the initial battery SoC during idle periods. This method proves ineffective in LiFePO4 chemistry due to its flat OCV-SOC profile. Given the critical nature of battery applications, where precise and timely control actions are imperative, waiting for the estimator to converge from an inaccurate initial state is impractical. To overcome this limitation, a model-based strategy employs a 2RC Equivalent Circuit model with temperature considerations. Model parameters are estimated for temperatures from 15°C to 45°C by fitting Hybrid Pulse Power Characterization (HPPC) test data using fminsearch. Then, a Hybrid Extended Kalman Filter - Particle filter (EKF-PF) based SoC estimator is developed and deployed for testing onboard a BMS fitted into a 2.88kWh LiFePO4 battery pack, under varying charging and discharging profiles. The performance of the hybrid EKF-PF algorithm is assessed against the Extended Kalman Filter (EKF) and Particle Filter (PF) in terms of Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and computation time for Modified Indian Drive Cycle (MIDC) and Worldwide Harmonized Light Vehicles Test Procedure (WLTP). The ECM with temperature considerations achieves an RMSE of 30.3 mV for MIDC and 18.4 mV for WLTP. The Hybrid EKF-PF demonstrates rapid convergence compared to EKF and reduces computation time relative to PF. This algorithm proves to be efficient for onboard SoC estimation.