Presentation Details
  1. Home
  2. Presentation List
  3. Presentation Details
Robust and Reliable Early-Stage Lifetime Prediction of Batteries via Transferable One-Shot Learning
ID：88 View Protection：ATTENDEE Updated Time：2025-11-10 12:17:55 Hits：231 Oral Presentation

Start Time：2025-11-23 08:30（Asia/Shanghai）

Duration：20min

Session：S3 Parallel Session 3 » S3-2Parallel Session 3-23 AM

Presentation File

Tips: Only the registered participant can access the file. Please sign in first.

Abstract
Accurate early-stage lifetime prediction of lithium-ion batteries is essential for their optimised utilisation and accelerated product development. However, obtaining full-cycle degradation data under normal usage conditions is both costly and time-consuming. Accelerated ageing tests offer a practical alternative, but a key challenge remains: how to generalise models trained on such data to normally aged batteries—an out-of-distribution (OOD) problem common in real-world applications. To address this, we propose a transferable one-shot diagnostic framework that requires only limited lifetime labels from short-life cells (averaging ~484 cycles) subjected to accelerated ageing, along with a single long-life cell (>1200 cycles) under normal conditions, to facilitate accurate predictions across the long-life population. Validation results show that the proposed model generalises effectively to long-life cells exceeding 800 cycles (with an average lifespan of 1147 cycles and a maximum of 2227 cycles), achieving an average prediction error below 15% across 10 independent runs of model training and testing. Compared to conventional baselines, the framework yields a substantial accuracy improvement of up to 68%, demonstrating its efficacy for robust and reliable early-stage lifetime prediction under OOD conditions.
 
Keywords
Lithium-ion batteries, life prediction, early life, one-shot learning, accelerated ageing test, variational autoencoder
Speaker
Ruohan Guo
Postdoc The Hong Kong Polytechnic University

Submission Author
Ruohan Guo The Hong Kong Polytechnic University
Jinpeng Tian The Hong Kong Polytechnic University
Dandan Peng The Hong Kong Polytechnic University
Chi-yung Chung The Hong Kong Polytechnic University
Submit Comment
Verify Code Change Another
All Comments
Important Date

  • Conference Date

    Nov 21

    2025

    to

    Nov 23

    2025

  • Oct 20 2025

    Draft paper submission deadline

  • Dec 08 2025

    Registration deadline

Sponsored By
IEEE Instrumentation and Measurement Society
South China University of Technology
Organized By
South China University of Technology
Contact Information
Login to view full contact details
Previous Conferences