Driving behavior's hidden impact on EV battery safety: Paving the way for smarter fault detection in electric vehicles

Research Background

As electric vehicles (EVs) surge in popularity to combat climate change and reduce fossil fuel dependence, ensuring the safety and reliability of their battery systems has become a critical challenge. Battery packs in EVs consist of hundreds or thousands of cells connected in series-parallel, but inherent inconsistencies in cell voltages can lead to faults like over-voltage, under-voltage, or even thermal runaway. Traditional fault detection methods often rely on fixed thresholds for voltage consistency, which can result in false alarms or missed detections due to varying driving conditions. This study, conducted by researchers at Wuhan University of Technology, investigates the microscopic associations between real-world driving behaviors and battery cell voltage consistency (VCC), using high-frequency data from naturalistic driving experiments. By revealing these links, the research lays the groundwork for adaptive, behavior-aware fault detection algorithms that could enhance EV safety and longevity.

Results and Benefits

The study divided EV driving processes into four micro-segments (A, B, C, D) based on accelerator and brake pedal actions, allowing for precise analysis of driving behavior parameters (DBPs) and their effects on the voltage variation coefficient between cells (VVCC). Key findings include:

- Strong correlations identified via Pearson correlation coefficients (PCCs): For segment A (acceleration phases), the average accelerator pedal stroke showed the highest PCC of 0.724 with VVCC. For segments B (cruising), C (deceleration), and D (coasting), average speed emerged as the top correlate with PCCs of 0.789, 0.554, and 0.553, respectively.

- High-accuracy predictive models: Four random forest (RF) regression models achieved goodness-of-fit (R²) values exceeding 0.919, demonstrating reliable prediction of VVCC based on DBPs.

- Nonlinear impact patterns: Accumulated local effects (ALE) plots revealed positive-nonlinear relationships overall, with approximate linearity in most intervals. The maximum VVCC growth rates were 48.09% for average accelerator pedal stroke in segment A, and 55.70%, 29.01%, and 23.68% for average speed in segments B, C, and D, respectively.

These outcomes highlight how aggressive driving—such as rapid acceleration or high speeds—exacerbates voltage inconsistencies, potentially accelerating battery degradation. Socially, this could lead to safer EVs, reducing accident risks from battery failures and extending vehicle lifespans, which supports broader adoption of sustainable transportation. By integrating driving behavior into diagnostics, the approach could minimize false alarms in battery management systems, saving costs for manufacturers and consumers while promoting environmental benefits through prolonged battery efficiency.

Future Application Prospects

The research's insights open doors to practical applications in EV battery management, such as developing adaptive-threshold algorithms that adjust fault detection in real-time based on driving states. For instance, integrating these models into onboard systems could enable predictive maintenance, alerting drivers to potential inconsistencies before they escalate. Further research could expand to include stationary vehicle states or diverse VCC indicators like entropy or curve distance, allowing for multi-metric prediction models. Comparative studies of advanced neural networks (e.g., LSTM or TCN) versus RF could identify optimal frameworks for voltage prognosis. In practical terms, this could refine energy management in fleet operations, optimize charging strategies, and inform policy for safer EV infrastructure, ultimately accelerating the transition to net-zero emissions by making EVs more reliable and user-friendly.

Conclusion

This pioneering work innovatively bridges driving behavior with battery voltage dynamics, offering a data-driven foundation for intelligent, adaptive fault detection in EVs. By quantifying correlations and nonlinear effects with real-world precision, it addresses longstanding gaps in fixed-threshold methods, promising enhanced safety, efficiency, and sustainability in the electric mobility era.

Reference

Author:  Shaopeng Li ^{a b c}, Hui Zhang ^{a b}, Naikan Ding ^{a b}, Matteo Acquarone ^c, Federico Miretti ^c, Daniela Anna Misul ^c

Title of original paper: Associations of battery cell voltage consistency with driving behavior of real-world electric vehicles

Article link: https://www.sciencedirect.com/science/article/pii/S2773153724000884

Journal: Green Energy and Intelligent Transportation

DOI: 10.1016/j.geits.2024.100236

Affiliations:

^a Intelligent Transportation Systems Research Center, Wuhan University of Technology, Wuhan 430063, China

^b Engineering Research Center of Transportation Information and Safety, Ministry of Education, Wuhan University of Technology, Wuhan 430063, China

^c Department of Energy (DENERG) and Center for Automotive Research and Sustainable Mobility (CARS@Polito), Politecnico di Torino, 10138 Torino, Italy