Advanced battery diagnostics for electric vehicles using CAN based BMS data with EKF and data driven predictive...

Area of Science:

• Electric Vehicle Battery Technology
• Data-Driven Diagnostics
• Lithium-ion Battery Management Systems

Background:

• Accurate battery State of Health (SoH) and State of Charge (SoC) are crucial for electric vehicle (EV) safety, performance, and longevity.
• Conventional monitoring techniques like Coulomb Counting can be susceptible to sensor noise and model inaccuracies.

Purpose of the Study:

• To develop and validate a novel hybrid diagnostic framework for improved EV battery monitoring.
• To enhance the accuracy and reliability of battery State of Health (SoH) and State of Charge (SoC) estimations.

Main Methods:

• Implemented a hybrid framework combining statistical analysis, machine learning, and model-based estimation.
• Utilized Extended Kalman Filter (EKF) to mitigate sensor noise and model errors.
• Employed Random Forest regression for SoH assessment, k-means clustering and Dynamic Time Warping (DTW) for cell behavior analysis, and Principal Component Analysis (PCA) for voltage imbalance identification.

Main Results:

• The EKF demonstrated superior performance over conventional Coulomb Counting techniques.
• The Random Forest model achieved higher accuracy for SoH assessment compared to linear regression.
• Identified outlier cells and analyzed voltage imbalance trends effectively.

Conclusions:

• The proposed hybrid framework offers a practical and expandable solution for reliable and interpretable EV battery diagnostics.
• This approach significantly improves battery monitoring capabilities, contributing to enhanced EV safety and life-cycle management.