
Home About Projects NEXTCAR ARPA-E SHARKS BMS Media People Publications Contact	Integrated Data-Driven and Physics-Guided Frameworks for Battery Module Health and Degradation Monitoring Critical Need: In practical applications, lithium-ion battery cells are connected in parallel or series to form modules that meet the power and energy demands of electrified vehicles. A major challenge in monitoring the health and degradation of battery modules composed of parallel-connected cells is the presence of intrinsic cell-to-cell variations and the lack of cell-level measurements. Differences in internal resistance, capacity, and thermal behavior among cells can lead to uneven current distribution and nonuniform degradation, which are difficult to detect and characterize using only module-level measurements. This nonuniformity not only accelerates aging but also complicates the interpretation of measurement signals, particularly under dynamic loading or general charging profiles. Conventional degradation monitoring approaches that treat the module as a homogeneous system often fail to capture these subtle but critical internal dynamics, resulting in inaccurate state-of-health (SOH) estimations and delayed fault detection. Consequently, there is a critical need to develop accurate, robust, and computationally efficient health and degradation monitoring frameworks tailored for battery modules with parallel-connected cells, enabling reliable implementation in onboard battery management systems. To overcome these limitations, there is a growing need for integrated frameworks that combine data-driven methods with physics-based modeling. Data-driven methods excel at extracting patterns from large, noisy datasets but often lack interpretability and generalizability beyond the training domain. Conversely, physics-based methods provide physical insight and can generalize across operating conditions but typically involve complicated modeling, require extensive parameter calibration, and are sensitive to modeling inaccuracies. An integrated approach can harness the strengths of both paradigms by leveraging the physical knowledge of the battery system to constrain and inform data-driven learning. Developing such hybrid frameworks, especially those optimized for real-time onboard implementation, is essential for ensuring the safety, reliability, and longevity of battery modules in electric vehicles and energy storage applications. Project Innovation + Advantages: This project, supported by the UM-Ford Alliance Program, aims at developing an integrated framework that combines data-driven techniques with physics-based analysis methods for health and degradation monitoring of battery modules. Compared to the existing methods, the innovations of the integrated framework under development are as follows: It achieves reliable module-level state-of-health (SOH) estimation with high accuracy, confidence, and computational efficiency in the presence of cell-to-cell variations, using only onboard module-level measurements. It generalizes the degradation monitoring framework from constant-current charging profiles to general charging profiles whose current is not necessarily constant, such as fast charging and constant-power charging profiles. It establishes new methodologies for estimating cell-to-cell variations solely from aggregated module-level data, using only module-level measurements. Sponsor: Ford Motor Company Publications: Qinan Zhou, Dyche Anderson, Jing Sun, "State of health estimation for battery modules with parallel-connected cells under cell-to-cell variations", eTransportation, 22, 2024. (PDF) Qinan Zhou, Erik Hellström, Dyche Anderson, Jing Sun, "Sensitivity Analysis of Support Vector Regression-Based Incremental Capacity Analysis for Battery State of Health Estimations", 2023 IEEE Conference on Control Technology and Applications (CCTA), 2023.
	Copyright © 2025 Real-time Adaptive Control Engineering Lab at the University of Michigan.

Integrated Data-Driven and Physics-Guided Frameworks for Battery Module Health and Degradation Monitoring

Critical Need:

In practical applications, lithium-ion battery cells are connected in parallel or series to form modules that meet the power and energy demands of electrified vehicles. A major challenge in monitoring the health and degradation of battery modules composed of parallel-connected cells is the presence of intrinsic cell-to-cell variations and the lack of cell-level measurements. Differences in internal resistance, capacity, and thermal behavior among cells can lead to uneven current distribution and nonuniform degradation, which are difficult to detect and characterize using only module-level measurements. This nonuniformity not only accelerates aging but also complicates the interpretation of measurement signals, particularly under dynamic loading or general charging profiles. Conventional degradation monitoring approaches that treat the module as a homogeneous system often fail to capture these subtle but critical internal dynamics, resulting in inaccurate state-of-health (SOH) estimations and delayed fault detection. Consequently, there is a critical need to develop accurate, robust, and computationally efficient health and degradation monitoring frameworks tailored for battery modules with parallel-connected cells, enabling reliable implementation in onboard battery management systems.

To overcome these limitations, there is a growing need for integrated frameworks that combine data-driven methods with physics-based modeling. Data-driven methods excel at extracting patterns from large, noisy datasets but often lack interpretability and generalizability beyond the training domain. Conversely, physics-based methods provide physical insight and can generalize across operating conditions but typically involve complicated modeling, require extensive parameter calibration, and are sensitive to modeling inaccuracies. An integrated approach can harness the strengths of both paradigms by leveraging the physical knowledge of the battery system to constrain and inform data-driven learning. Developing such hybrid frameworks, especially those optimized for real-time onboard implementation, is essential for ensuring the safety, reliability, and longevity of battery modules in electric vehicles and energy storage applications.

Project Innovation + Advantages:

This project, supported by the UM-Ford Alliance Program, aims at developing an integrated framework that combines data-driven techniques with physics-based analysis methods for health and degradation monitoring of battery modules.

Compared to the existing methods, the innovations of the integrated framework under development are as follows:

It achieves reliable module-level state-of-health (SOH) estimation with high accuracy, confidence, and computational efficiency in the presence of cell-to-cell variations, using only onboard module-level measurements.
It generalizes the degradation monitoring framework from constant-current charging profiles to general charging profiles whose current is not necessarily constant, such as fast charging and constant-power charging profiles.
It establishes new methodologies for estimating cell-to-cell variations solely from aggregated module-level data, using only module-level measurements.

Sponsor:

Ford Motor Company

Publications:

Qinan Zhou, Dyche Anderson, Jing Sun, "State of health estimation for battery modules with parallel-connected cells under cell-to-cell variations", eTransportation, 22, 2024. (PDF)
Qinan Zhou, Erik Hellström, Dyche Anderson, Jing Sun, "Sensitivity Analysis of Support Vector Regression-Based Incremental Capacity Analysis for Battery State of Health Estimations", 2023 IEEE Conference on Control Technology and Applications (CCTA), 2023.