In Situ Estimation of Li-Ion Battery State of Health Using On-Board Electrical Measurements for Electromobility Applications

Jorge E. García-Bustos, Benjamín Brito Schiele, Leonardo Baldo, Bruno Masserano, Francisco Jaramillo, Diego Troncoso-Kurtovic, Marcos E. Orchard, A. Pérez, Jorge F. SilvaBatteries 2025 · DOI

TL;DR

This paper introduces a novel, real-time method for estimating the State of Health (SoH) of lithium-ion batteries in electric vehicles using only on-board voltage and current data. By combining Maximum Likelihood Estimation (MLE) with Bayesian Filtering (specifically, a Particle Filter), the approach accurately tracks battery degradation without requiring battery removal or specialized diagnostic tests, achieving mean absolute errors below 1%.

What Problem Does This Paper Address?

Electric Vehicles (EVs) rely heavily on lithium-ion batteries, but their performance degrades over time, leading to reduced range, potential safety concerns, and decreased reliability. Accurately monitoring this "State of Health" (SoH) is crucial for effective battery management, ensuring user confidence, and optimizing EV lifespan. However, existing SoH estimation methods often require battery removal, lengthy diagnostic procedures (like full charge/discharge cycles or Electrochemical Impedance Spectroscopy), or access to data not typically available from an EV's Battery Management System (BMS). This makes real-time, on-the-go SoH estimation a significant challenge. The authors aim to bridge this gap by developing a non-invasive, efficient, and accurate method that works solely with readily available electrical measurements.

Methods

The core of the proposed methodology is a hybrid approach combining Maximum Likelihood Estimation (MLE) and a Particle Filter (PF), termed MLE-PF.

  1. Feature Extraction Module (FEM): This module identifies specific voltage and current measurements during typical EV driving (discharge cycles). The key is to find moments when the current is low enough that the measured voltage closely approximates the true Open-Circuit Voltage (OCV). This is achieved by applying thresholds for current magnitude and duration. This module also compensates for internal resistance effects.
  2. Maximum Likelihood Estimation (MLE) Module: Using the extracted OCV approximations and corresponding charge extracted, the MLE module estimates the battery's current capacity. This is framed as an optimization problem where the goal is to find the capacity that maximizes the likelihood of observing the measured data, given a model of the OCV-SoC relationship.
  3. Particle Filter (PF) Module: The capacity estimates from the MLE module are then fed into a PF. The PF acts as a Bayesian filter, refining these estimates by incorporating a degradation model and accounting for uncertainties and noise in the measurements. This helps to smooth out estimates, correct for potential outliers or inaccuracies from the MLE, and provide a more robust and reliable SoH prediction. The PF also allows for the simultaneous estimation of State of Charge (SoC).

This integrated approach allows for real-time estimation directly from on-board data, adapting to varying driving conditions without needing predefined profiles.

Datasets

The methodology was validated using the publicly available lithium-ion battery aging dataset from Stanford University, released by the MIT-Toyota Research Center. This dataset contains detailed cycling data (voltage, current, charge, discharge capacities) for ten commercial INR21700-M50T lithium-ion cells cycled over 23 months under realistic EV use scenarios. The authors selected cells W4, W8, and W10 for their higher degradation, and cell W9 was used exclusively for parameter tuning.

Key Insights and Ideas

  • The Power of "Almost" OCV: The critical insight is that even without full battery rest periods, carefully selected low-current discharge segments can provide voltage readings that are close enough to the true OCV to enable accurate capacity estimation.
  • MLE for Initial Guess, PF for Refinement: MLE provides a good initial estimate of battery capacity from the extracted features, but it can be sensitive to noise and short discharge cycles. The PF then acts as a crucial layer of robustness, smoothing these estimates and handling uncertainties.
  • Scalability is Key: The framework's modular design and reliance on a Thévenin-equivalent circuit model make it adaptable to different battery chemistries, modules, and configurations with minimal offline parameter tuning.
  • Real-World Adaptability: The method doesn't require specific driving profiles or controlled lab conditions, making it highly suitable for practical EV applications where driving patterns vary daily.

Key Results

  • The proposed MLE-PF method demonstrated accurate tracking of battery degradation across multiple cells.
  • The methodology achieved a Mean Absolute Error (MAE) below 1% for SoH estimation. Specifically, for cell W4, the interpolated MAE was 0.235% and the MAE at diagnostic tests was 0.349%.
  • The sensitivity analysis showed that the framework is robust to different combinations of low-current detection parameters (current threshold and duration), with the best performance achieved when prioritizing the quality of detected pseudo-OCV points over their quantity.
  • The 5-95% credibility interval provided by the PF remained narrow, indicating low estimation uncertainty and consistent tracking of the ground truth.

Limitations and Future Work

  • Internal Resistance Model Simplification: The paper acknowledges that the polynomial model used for internal resistance is a simplification and doesn't account for factors like temperature or hysteresis. While errors are mitigated by the PF's noise modeling and the low currents used, this remains an area for potential improvement.
  • Data Quality Dependence: While the PF enhances robustness, the initial quality of the extracted pseudo-OCV points is still important. The sensitivity analysis highlights the trade-off between the number of detected points and their quality.
  • Future Work: The authors suggest exploring the integration of additional data sources, such as temperature and external load variations, to further enhance accuracy and applicability. They also aim to refine model parameters and investigate the impact of temperature variations on ESR model parameters.

Summary updated

Abstract

The well-balanced combination of high energy density and competitive cycle performance has established lithium-ion batteries as the technology of choice for Electric Vehicles (EVs) energy storage. Nevertheless, battery degradation continues to pose challenges to EV range, safety, and long-term reliability, making accurate estimation of their State of Health (SoH) crucial for efficient battery management, safety, and improved longevity. This paper addresses a compelling research question surrounding the possibility of developing a real-time, non-invasive, and efficient methodology for estimating lithium-ion battery SoH without battery removal, relying solely on voltage and current data. Our approach integrates the fitting abilities of Maximum Likelihood Estimation (MLE) with the dynamic uncertainty propagation of Bayesian Filtering to provide accurate and robust online SoH estimation. By reconstructing the open-circuit voltage curve from real-time data, the MLE estimates battery capacity during discharge cycles, while Bayesian Filtering refines these estimates, accounting for uncertainties and variations. The methodology is validated using an available dataset from Stanford University, demonstrating its effectiveness in tracking battery degradation under driving profiles. The results indicate that the approach can reliably estimate battery SoH with mean absolute errors below 1%, confirming its suitability for scalable EV applications.
BibTeX
@inproceedings{garciabustos2025situ,
  author = {Jorge E. García-Bustos and Benjamín Brito Schiele and Leonardo Baldo and Bruno Masserano and Francisco Jaramillo and Diego Troncoso-Kurtovic and Marcos E. Orchard and A. Pérez and Jorge F. Silva},
  title = {{In Situ Estimation of Li-Ion Battery State of Health Using On-Board Electrical Measurements for Electromobility Applications}},
  booktitle = {Batteries},
  year = {2025},
  doi = {10.3390/batteries11120451},
}
Figure