Real-Time Parameter Identification and State of Charge Estimation of Electric Vehicle Batteries

Abstract

Accurate determination of the state of charge (SOC) is crucial for carrying out a range of battery management tasks. Meanwhile, for figuring out the SOC, it is crucial to determine the battery model parameters as they can vary based on the operating conditions. This paper proposes a novel algorithm called the variable forgetting factor recursive least squares algorithm (VFFRLS) to tackle this problem. Simulations are carried out on two different battery models, specifically one RC and two RC models. The fixed forgetting factor RLS (FFRLS) algorithm is implemented with two different forgetting factor (FF) values, while the VFFRLS method utilizes different initial FF values. From the results obtained from two RC-ECM, the MSE of VFFRLS (λ₀ = 0.95) is about 2.45e-4, followed by VFFRLS (λ₀ = 1) by 2.48e-4, FFRLS (λ = 0.95) by 3.53e-04, and FFRLS (λ = 1) by 0.002, confirming the accuracy of VFFRLS over FFRLS. The simulation results clearly show that the suggested VFFRLS technique outperforms the conventional RLS. In addition, the SOC estimation has been conducted using the optimized extended Kalman filter. The suggested battery model, parameter identification algorithm, and optimized filter have been tested and validated using real-time datasets from various sources, including the NASA online battery dataset, data collections of Panasonic 18650PF and LG 18650HG2 batteries. The verification process involved both constant load conditions and the dynamic drive profile of an electric vehicle.