State of Charge Estimation of LiFePO₄ Battery Based on a Gain-classifier Observer

A state of charge (SOC) estimation is required to work well over a wide range of conditions. In this study, a multi-gain observer based on classified conditions is proposed in order to estimate SOC efficiently. The feedback gain in the observer is switched by a proposed geometry classifier to categorize the voltage error into different groups so that the observing strategies can be designed for different error sources, and thereby robust and accurate SOC estimations. Different load conditions (including QC/T 897-2011 standard suggested condition) are tested to verify the proposed method. The results show that the proposed method is effective and accurate. It is possible to solve local model inaccuracies and data saturation problems in a computationally efficient way.