Adaptive and Fast State of Health Estimation Method for Lithium-ion Batteries Using Online Complex Impedance and Artificial Neural Network

Abstract

This paper presents an adaptive state-of-health (SOH) estimation method that utilizes artificial neural network (ANN) and online AC complex impedance. The zero crossing frequency of battery impedance phase can reflect the aging status of battery based on the observation from the aging data. However, the relationship between the zero crossing frequency and SOH is nonlinear. In order to model this nonlinear relationship for SOH prediction, ANN as a powerful nonlinear fitting tool or method is explored in this paper in order to characterize this relationship. The designed ANN can update its parameters based on the feedback data from the operation of the system. This feature makes the proposed method be able to adapt to the changes in the operation conditions and aging conditions of the battery, which enables better SOH prediction accuracy compared with the static SOH model methods when the operation conditions or battery conditions are different from the ones that the static SOH models are derived from. The proposed SOH estimation method also allows for fast prediction compared with the conventional capacity fading methods. This is mainly because the parameter used for SOH prediction, i.e. battery impedance phase, can be obtained within a short time during the online operation of the system. A preliminary experimental prototype is built in the laboratory to verify the proposed method.