PSO-Tuned Variable Forgetting Factor Recursive Least Square Estimation of 2RC Equivalent Circuit Model Parameters for Lithium-Ion Batteries

To improve the accuracy of Equivalent Circuit Models (ECM) for Electric Vehicles (EV) applications, parameter identification approaches based on Recursive Least Squares (RLS) filters have been proposed. The Variable Forgetting Factor RLS (VFFRLS) algorithms are regarded as one of the most accurate parameter identification methods for lithium-ion batteries. Commonly, a fixed forgetting factor is utilized with a conventional RLS to improve the accuracy, however, this algorithm fails to keep up with a real-time deviation in the environment. In this paper, a Particle Swarm Optimization (PSO) technique is presented for identifying the optimal VFFRLS parameters to compromise between identification accuracy and stability. The proposed method is applied on a second-order Thevenin model that is extensively utilized for Battery Management Systems (BMS) using MATLAB/Simulink software for a commercially available grade pouch cell.