Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Transformer-based Capacity Prediction for Lithium-ion Batteries with Data Augmentation

Gift Modekwe, Saif Al-Wahaibi, Qiugang Lu

TL;DR

This work addresses the challenge of accurately predicting lithium-ion battery capacity (a key facet of state-of-health) by capturing long-term temporal dependencies in multi-channel data (current, voltage, temperature) while mitigating data scarcity through Gaussian-noise data augmentation. It proposes a hybrid auto-encoder–transformer framework that compresses per-cycle battery profiles into latent representations and uses a moving window to forecast the next-cycle capacity as a scalar. Validations on NASA Group 1 and UofM aging datasets show that a transformer-based predictor with data augmentation outperforms CNN and LSTM baselines and benefits further from augmentation, highlighting improved robustness and prediction accuracy for battery management systems. The approach offers a practical pathway to more reliable SOH estimation in real-world BMS deployments, with potential extensions to other temperatures and augmentation schemes.

Abstract

Lithium-ion batteries are pivotal to technological advancements in transportation, electronics, and clean energy storage. The optimal operation and safety of these batteries require proper and reliable estimation of battery capacities to monitor the state of health. Current methods for estimating the capacities fail to adequately account for long-term temporal dependencies of key variables (e.g., voltage, current, and temperature) associated with battery aging and degradation. In this study, we explore the usage of transformer networks to enhance the estimation of battery capacity. We develop a transformer-based battery capacity prediction model that accounts for both long-term and short-term patterns in battery data. Further, to tackle the data scarcity issue, data augmentation is used to increase the data size, which helps to improve the performance of the model. Our proposed method is validated with benchmark datasets. Simulation results show the effectiveness of data augmentation and the transformer network in improving the accuracy and robustness of battery capacity prediction.

Transformer-based Capacity Prediction for Lithium-ion Batteries with Data Augmentation

TL;DR

This work addresses the challenge of accurately predicting lithium-ion battery capacity (a key facet of state-of-health) by capturing long-term temporal dependencies in multi-channel data (current, voltage, temperature) while mitigating data scarcity through Gaussian-noise data augmentation. It proposes a hybrid auto-encoder–transformer framework that compresses per-cycle battery profiles into latent representations and uses a moving window to forecast the next-cycle capacity as a scalar. Validations on NASA Group 1 and UofM aging datasets show that a transformer-based predictor with data augmentation outperforms CNN and LSTM baselines and benefits further from augmentation, highlighting improved robustness and prediction accuracy for battery management systems. The approach offers a practical pathway to more reliable SOH estimation in real-world BMS deployments, with potential extensions to other temperatures and augmentation schemes.

Abstract

Lithium-ion batteries are pivotal to technological advancements in transportation, electronics, and clean energy storage. The optimal operation and safety of these batteries require proper and reliable estimation of battery capacities to monitor the state of health. Current methods for estimating the capacities fail to adequately account for long-term temporal dependencies of key variables (e.g., voltage, current, and temperature) associated with battery aging and degradation. In this study, we explore the usage of transformer networks to enhance the estimation of battery capacity. We develop a transformer-based battery capacity prediction model that accounts for both long-term and short-term patterns in battery data. Further, to tackle the data scarcity issue, data augmentation is used to increase the data size, which helps to improve the performance of the model. Our proposed method is validated with benchmark datasets. Simulation results show the effectiveness of data augmentation and the transformer network in improving the accuracy and robustness of battery capacity prediction.
Paper Structure (13 sections, 9 equations, 8 figures, 2 tables)

This paper contains 13 sections, 9 equations, 8 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. Transformer Network
  4. Data Augmentation
  5. Battery Dataset
  6. Transformer-based capacity prediction with multi-channel Profiles
  7. Multi-channel Profiles for Capacity Prediction
  8. Architecture of the Proposed Transformer Network
  9. Case Studies and Discussions
  10. Setup of Case Studies
  11. Experimental Result and Analysis:
  12. Conclusion
  13. Acknowledgment

Figures (8)

  • Figure 1: Capacity degradation profile for the batteries used in this work.
  • Figure 2: The proposed auto-encoder and transformer architecture for battery capacity prediction.
  • Figure 3: Voltage, current and temperature profile for battery B0005.
  • Figure 4: Logic flow of the case studies.
  • Figure 5: Predicted capacity vs. ground-truth values for NASA Group 1 dataset.
  • ...and 3 more figures