Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Degradation Self-Supervised Learning for Lithium-ion Battery Health Diagnostics

J. C. Chen

TL;DR

This work tackles the challenge of diagnosing lithium-ion battery health under dynamic operating conditions with limited labeled data. It introduces degradation self-supervised learning that leverages unlabeled voltage–current time-series, enhanced by empirical wavelet transform preprocessing and a transformer-based health predictor, to yield a health indicator that reflects irreversible aging. The approach achieves a strong association with actual capacity degradation on the Stanford aging dataset (average $R$ around 0.95 for eight cells) and demonstrates effective transfer to a downstream capacity-estimation task with limited labels. The method offers a cost-efficient, scalable path for battery health monitoring in electric vehicles and other energy-storage systems, while acknowledging the need for real-world, end-of-life data for further validation.

Abstract

Health evaluation for lithium-ion batteries (LIBs) typically relies on constant charging/discharging protocols, often neglecting scenarios involving dynamic current profiles prevalent in electric vehicles. Conventional health indicators for LIBs also depend on the uniformity of measured data, restricting their adaptability to non-uniform conditions. In this study, a novel training strategy for estimating LIB health based on the paradigm of self-supervised learning is proposed. A multiresolution analysis technique, empirical wavelet transform, is utilized to decompose non-stationary voltage signals in the frequency domain. This allows the removal of ineffective components for the health evaluation model. The transformer neural network serves as the model backbone, and a loss function is designed to describe the capacity degradation behavior with the assumption that the degradation in LIBs across most operating conditions is inevitable and irreversible. The results show that the model can learn the aging characteristics by analyzing sequences of voltage and current profiles obtained at various time intervals from the same LIB cell. The proposed method is successfully applied to the Stanford University LIB aging dataset, derived from electric vehicle real driving profiles. Notably, this approach achieves an average correlation coefficient of 0.9 between the evaluated health index and the degradation of actual capacity, demonstrating its efficacy in capturing LIB health degradation. This research highlights the feasibility of training deep neural networks using unlabeled LIB data, offering cost-efficient means and unleashing the potential of the measured information.

Degradation Self-Supervised Learning for Lithium-ion Battery Health Diagnostics

TL;DR

This work tackles the challenge of diagnosing lithium-ion battery health under dynamic operating conditions with limited labeled data. It introduces degradation self-supervised learning that leverages unlabeled voltage–current time-series, enhanced by empirical wavelet transform preprocessing and a transformer-based health predictor, to yield a health indicator that reflects irreversible aging. The approach achieves a strong association with actual capacity degradation on the Stanford aging dataset (average around 0.95 for eight cells) and demonstrates effective transfer to a downstream capacity-estimation task with limited labels. The method offers a cost-efficient, scalable path for battery health monitoring in electric vehicles and other energy-storage systems, while acknowledging the need for real-world, end-of-life data for further validation.

Abstract

Health evaluation for lithium-ion batteries (LIBs) typically relies on constant charging/discharging protocols, often neglecting scenarios involving dynamic current profiles prevalent in electric vehicles. Conventional health indicators for LIBs also depend on the uniformity of measured data, restricting their adaptability to non-uniform conditions. In this study, a novel training strategy for estimating LIB health based on the paradigm of self-supervised learning is proposed. A multiresolution analysis technique, empirical wavelet transform, is utilized to decompose non-stationary voltage signals in the frequency domain. This allows the removal of ineffective components for the health evaluation model. The transformer neural network serves as the model backbone, and a loss function is designed to describe the capacity degradation behavior with the assumption that the degradation in LIBs across most operating conditions is inevitable and irreversible. The results show that the model can learn the aging characteristics by analyzing sequences of voltage and current profiles obtained at various time intervals from the same LIB cell. The proposed method is successfully applied to the Stanford University LIB aging dataset, derived from electric vehicle real driving profiles. Notably, this approach achieves an average correlation coefficient of 0.9 between the evaluated health index and the degradation of actual capacity, demonstrating its efficacy in capturing LIB health degradation. This research highlights the feasibility of training deep neural networks using unlabeled LIB data, offering cost-efficient means and unleashing the potential of the measured information.

Paper Structure

This paper contains 14 sections, 3 equations, 6 figures, 1 table.

Table of Contents

  1. Introduction
  2. Methods
  3. Data description and preprocessing
  4. Self-supervised degradation learning framework
  5. Embedding module.
  6. Transformer encoder.
  7. Loss function.
  8. Training procedure
  9. Results
  10. Health status estimation
  11. Representation visualization
  12. Ablation experiment
  13. Knowledge transfer to downstream task
  14. Conclusion & limitation

Figures (6)

  • Figure 1: Preprocessing workflow of proposed method.
  • Figure 2: Overview of the proposed self-supervised degradation learning framework
  • Figure 3: (a) Discharge capacity and the corresponding health indicator $h$ estimated by the proposed model for the 8 cells. (b) The distribution of surface temperature during the diagnostics tests. (c) The regen resistance calculated according to the HPPC at third diagnostic test.
  • Figure 4: The averaged estimated $h$ of each cycle and the corresponding capacity of (a) V4 and (b) W10 cells.
  • Figure 5: Visualization of the representations extracted by embedding module using (a) PCA and (b) Isomap algorithms
  • ...and 1 more figures