Improving the state of charge estimation of reused lithium-ion batteries by abating hysteresis using machine learning technique

The accuracy of the state of charge (SoC) estimation is of great importance to the operational safety of a battery pack, especially for secondary applications with retired batteries. Here, a novel approach combining Sigma-point Kalman filter and machine learning technique based on an equivalent circuit model is proposed to improve the state of charge estimation accuracy of a reused battery pack (LiFePO4) by abating the negative effect of the hysteresis phenomenon. Compared to traditional estimation methods, this approach can reduce the root mean square error by up to 8.3%. The maximum estimation error for three experimental tests is only 0.016 being within acceptable range and demonstrating the effectiveness of the proposed approach.