비지도 학습 기반 전기차 배터리 전기적 이상 탐지 관한 연구: 충전 방식, 노화에 따른 비교 분석

다국어 초록 (Multilingual Abstract)

The trend toward carbon neutrality and the rapid adoption of electric vehicles have significantly increased the demand for lithium-ion batteries. However, lithium-ion batteries pose potential safety risks such as thermal runaway, highlighting the need for early detection and mitigation technologies. This study proposes an autoencoder-based unsupervised learning model for the detection of electrical anomalies in lithium-ion batteries, such as voltage drops and overheating. The experimental results indicate that the proposed model exhibits high sensitivity in identifying individual anomaly scenarios, with particularly enhanced performance under fast-charging conditions. In addition, the model exhibited significantly enhanced sensitivity in complex anomaly scenarios. Furthermore, it maintained consistent detection performance after the battery had undergone aging through repeated charge–discharge cycles. These findings suggest that the proposed approach can serve as a valuable contribution to the study of lithium-ion battery anomaly detection and the application of unsupervised learning in electrical fault diagnostics.