An estimation model for state of health of lithium-ion batteries using energy-based features

Abstract

Lithium-ion batteries are pervasive in the renewable-energy based market. A key but challenging issue is accurate state of health (SOH) estimation in battery health monitoring (BHM). The complete discharging curve of battery is rarely available in real world. The incomplete discharging operation affects the subsequent constant current (CC) charging process, which extremely limits many conventional aging features extracted from the complete cycle process. Therefore, under incomplete discharging, the energy-based features are extracted to realize accurate and reliable SOH estimation. The purpose is achieved by an improved Gaussian progress regression (GPR) model. First, the features extracted from direct measurement curves are considered as the inputs of degradation model. A multidimensional linear mean function and a novel covariance function are proposed to adapt the fluctuations. So as to realize accurate batteries SOH estimation. Additionally, several batteries from NASA dataset are applied for the verification of the proposed model from different initial health states. Results demonstrate that this model outperforms the counterparts with a mean RMSE of 0.97% in the testing set.