Abstract:

According to the most crucial modeling requirements regarding accuracy, configuration effort, computational complexity, and ease of implementation, the Equivalent Circuit Model (ECM) is the preferred choice for online applications. Since the model parameters are time-variant, the Variable Forgetting Factor Recursive Least Square algorithm (VFFRLS) is proposed to identify the model parameters. The suggested VFFRLS algorithm can feed its identification results into ECM to estimate the SOC of battery using nonlinear Kalman filtering algorithms. However, the performance of the Kalman filter strongly depends on process noise covariance matrix (Q) and measurement noise covariance matrix (R) values. The improper value of these matrices reduces the convergence rate and increases the estimation error. In this work, the impact of these matrices on SOC estimation is analyzed by the trial and error method. Since the order of the system increases, manual tuning of these matrices becomes hard. As a result, filter tuning is essential to ensure the accuracy and convergence speed of the filter for SOC estimation of Li-ion batteries. To deal with this, the optimized Extended Kalman Filter (EKF) is proposed in this work.