Anomaly Detection in Battery Charging Systems: A Deep Sequence Model Approach

Abstract

While the popularity of electric vehicles brings great convenience to our lives, battery charging also leads to an increase in accidents, resulting in personal injuries and economic losses. The methods currently embedded in charging hardware mainly focus on the short-term state of the battery and fail to leverage historical information effectively. The development of the Industrial Internet of Things (IIoT) enables data collection from sensors on industrial devices, which can be analyzed using deep learning methods to support sophisticated analysis. This paper proposes an intelligent and secure battery charging system in the IIoT that establishes an interaction between battery charging devices and cloud-based algorithms. A novel anomaly detection method is introduced to deal with anomalous charging sequences by making good use of historical data. We evaluate our system using real-life data from 4,940 batteries in electric vehicles, and our experiments achieve satisfactory results in detecting anomalies in battery charging.