A Novel Approach for the Open-circuit Voltage Estimation of Lithium-ion Batteries by epsilon SVR

Abstract

Open-circuit voltage (OCV) estimation is an essential challenge for battery management systems (BMSs). Accurate OCV estimation can benefit the state estimation of lithium-ion batteries (LiBs), including state of charge (SoC) and state of health (SoH) estimations. When obtaining a SoC-OCV curve, a battery test method based on low current discharge cannot overcome the effect of battery polarization, and the discharge time usually takes several hours. Hence, an ε-support vector regression (ε-SVR) model for battery OCV estimation was proposed in this research. In accordance with the voltage relaxation behaviour of LiBs, the sample data were collected by hybrid pulse power current (HPPC) experiments on LiBs under different ageing degrees. The features were selected from the data samples using grey association analysis (GRA), and the hyperparameters of the ε-SVR model were obtained by K-folding cross-validation (CV). To validate this approach, the proposed model was trained and tested over the dataset acquired from the LiBs with varying degrees of ageing. Based on the experimental results, the model only needs some short-term battery characteristic data to achieve high-precision OCV estimation.