High Accuracy and Applicability Battery Aging Models for Electric Vehicle Applications

Abstract

Batteries have gained significant attention due to their numerous advantages in applications such as electric vehicles. One of the factors limiting industry adoption is the aging of batteries. The characteristics of battery aging vary depending on many factors such as battery type, electrochemical reactions and operating conditions. Here we document the comparison of semi-empirical aging models (SEM), highlighting limitations and challenges. In addition, four SEMs are proposed. The usability and compatibility of these models are evaluated using experimental data from various sources including the Horizon 2020 Helios Project. The optimized parameters of each model are documented via linear regression and genetic algorithms. The results show that the genetic algorithm approach provides higher accuracy in comparison to the linear regression. The documented SEMs reveal better prediction performance than the literature of calendar obsolescence with SEM-3 and 7 performing particularly well in predicting capacity loss for the Helios dataset with low errors, i.e. 0.43 and 0.79 RMSE, respectively. The range of RMSE values for model predictions across all the datasets ranges from 0.196 to 3.903. This study aims to document the accuracy of SEMs both from the literature and proposed in the paper relative to battery ageing data from distinct sources.