SOH Estimation for Li-ion Batteries Based on Features of IC Curves and Multi-output Gaussian Process Regression Method

Abstract

The state-of-health (SOH) estimation of li-ion battery is an important basis for the safe operation of battery system. There is no mature solution for the determination of characteristic parameters and methods for SOH estimation, and it is a great challenge to obtain high-precision estimates with as little data as possible. Considering the actual operating condition of the power battery, we extract multiple features reflecting battery aging state from incremental capacity (IC) curves. A multi-output Gaussian process regression model is built by taking the capacity and features as model outputs. The results show that the proposed method can utilize the underlying correlation between battery capacity and features. Different covariance kernels are used to exploit the data and improve estimation accuracy further. In view of the lack of historical data, the relationship between the percentage of training data and the accuracy of estimation is studied. The analysis illustrates that the multi-output Gaussian process has great advantages for the regression problem under the condition of small samples, which is conducive to the practical application of the proposed method.