the Estimation of State of Charge for 4S2P Lithium-Ion Battery Using Kalman Filter and Coulomb Counting

Comparative Simulation-Based Study with Evaluation of RMSE, MAE, MAPE, and R² Metrics

Abstract

State of Charge (SoC) estimation is crucial for the performance and safety of Battery Management Systems (BMS). This study evaluates and compares two SoC estimation methods—Kalman Filter and Coulomb Counting—based on numerical simulation of a 4S2P lithium-ion battery charging process using MATLAB. The methods are assessed using statistical metrics: RMSE, MAE, MAPE, and R², and are compared against both current-based reference calculations and normalized actual voltage. Kalman Filter consistently demonstrates superior performance, achieving lower RMSE (0.00067) and MAE (0.00045) against SoC reference, and RMSE (0.0376), MAE (0.0312), R² (0.978) against voltage reference. In contrast, Coulomb Counting shows increased error accumulation and lower correlation with system behavior. This confirms Kalman Filter's robustness in dynamic conditions, owing to its real-time correction mechanism and noise tolerance. The study highlights Kalman Filter as a more accurate and reliable method for modern BMS applications. Recommendations for future development include real-world testing and hybrid algorithm implementation.