State of Health Estimation for Lithium-Ion Batteries Using Random Charging Data

Abstract

Precise state of health (SOH) estimation is pivotal for reliable operations of lithium-ion batteries in electric vehicles. However, the collected charging data is usually incomplete, which makes it difficult to generate health features and brings great challenges to SOH estimation. To conquer this defect, Gaussian process regression (GPR) is developed using random partial charge segment to estimate battery SOH. Firstly, a voltage fitting process is proposed to reconstruct the constant charging voltage trajectories from random partial charging data. Then, the charging time is inferred to characterize battery deterioration. Correlation analysis is conducted and high correlation between SOH and health feature is verified. Following this endeavor, GPR is presented to effectively predict SOH with the input of charging duration. The proposed method can extend the random partial charging segment to complete charging data, thereby relieving the pain there is little chance that the drivers charge the battery from a predefined voltage data. Train and validation are executed on four battery cells, highlighting that the developed approach can maintain the SOH within 2% error.