Estimation of Battery Capacity Fade using Real-World Vehicle Data for Diagnosis of Abnormal Capacity Loss

Accurate estimation of battery capacity and diagnosis of its degradation state are essential for safe battery management. This paper presents an advanced method for accurate capacity estimation and abnormal capacity degradation diagnosis of electric vehicle battery systems. Base on the real-world electric vehicles (EVs) data, the reference capacity of the battery system can be calculated by integrates incremental Capacity (IC) curves and Coulomb counting method. Main factors, such as mileage, temperature, charging current, and depth of discharge, affecting the battery performance and life were discussed. And then, a fusion model developed by combining the XGBoost and LightGBM algorithms is used to estimate capacity. The results show that the proposed model outperforms the single model with a mean absolute percentage error (MAPE) of 2.45%, and has a better ability to follow the abnormal capacity degradation, which can evaluate the battery capacity and ensure safety.

I. Introduction

With the increasing global energy demand and escalating environmental concerns, electric vehicles (EVs) have gained significant attention as a viable alternative to traditional fossil fuel vehicles [1]. Lithium-ion batteries, known for their high energy density and long cycle life, are widely employed as the power source for electric vehicles [2]. However, battery degradation over time remains a significant challenge that impacts the performance and lifespan of these batteries [3]. Therefore, accurately estimating battery capacity degradation is crucial to ensure the reliability and safety of Evs’ battery systems.