Anomaly Detection Method for Lithium-Ion Battery Cells Based on Time Series Decomposition and Improved Manhattan...

Area of Science:

• Electrical Engineering
• Materials Science
• Data Science

Background:

• Individual lithium-ion battery cell failures can lead to electric vehicle (EV) system malfunction and safety hazards.
• Early and precise detection of anomalous battery cells is essential for preventing accidents and mitigating economic losses.

Purpose of the Study:

• To propose a robust battery cell anomaly detection method for EVs using real-world operating data.
• To accurately identify failing battery cells and determine the precise moment of malfunction.

Main Methods:

• Applied time series decomposition to EV battery pack voltage data to extract individual cell voltage trends.
• Utilized an improved Manhattan distance algorithm to compare adjacent cell trend components for anomaly identification.
• Calculated Manhattan distances within the data sequence at specific sampling moments to detect the exact time of cell malfunction.

Main Results:

• The proposed method successfully identified abnormal battery cells within the pack.
• The technique accurately diagnosed the specific time of abnormality in early-stage battery cell failures.
• Experimental verification confirmed the method's effectiveness and robustness on actual EV operating data.

Conclusions:

• The developed method provides accurate and early detection of abnormal lithium-ion battery cells in EVs.
• This approach enhances EV safety and reduces potential property damage by identifying critical failure points.
• The combination of time series decomposition and Manhattan distance offers a reliable solution for battery management systems.