Realistic Fault Detection of Li-ion Batteries from 347 Electric Vehicles

Over 690,000 charging snippets from 347 electric vehicles provide data for evaluating Li-ion battery safety. The dataset was released by author Jingzhao Zhang as part of a paper on a dynamical deep learning approach for anomaly detection. It addresses the lack of real-world testing frameworks with large-scale data for complex battery failure mechanisms.

Use Cases

• Training anomaly detection models based on real-world battery charging data.
• Validating fault prediction algorithms for Li-ion battery safety.
• Benchmarking dynamical deep learning approaches for time-series data from electric vehicles.
• Studying complex failure mechanisms of batteries in realistic operational settings.

Strengths

• Contains over 690,000 data points, providing a substantial sample size.
• Originates from 347 real electric vehicles, suggesting real-world operational data.
• Explicitly designed for realistic battery fault detection, addressing a noted research gap.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and last update date are unknown, which may limit suitability assessment.
• Data may reflect temporal or source bias inherent to the specific collection from paperswithcode.

Provenance

Source

Jingzhao Zhang

Collection Method

Likely collected from operational electric vehicles, though the specific method is not detailed.