Interpretable Deep Learning Using Temporal Transformers for Battery Degradation Prediction
Accurate modelling of lithium-ion battery degradation is a complex problem, dependent on multiple internal mechanisms that can be affected by a multitude of external conditions. In this study, a transformer-based approach, capable of leveraging historical conditions and known-future inputs is introd...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-06-01
|
| Series: | Batteries |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2313-0105/11/7/241 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Accurate modelling of lithium-ion battery degradation is a complex problem, dependent on multiple internal mechanisms that can be affected by a multitude of external conditions. In this study, a transformer-based approach, capable of leveraging historical conditions and known-future inputs is introduced. The model can make predictions from as few as 100 input cycles, and compared to other state-of-the-art techniques, our approach shows an increase in accuracy. The model utilises specialised components within its architecture to provide interpretable results, introducing the possibility of understanding path-dependency in Li-Ion battery degradation. The ability to incorporate static metadata opens the door for a foundational deep learning model for battery degradation forecasting. |
|---|---|
| ISSN: | 2313-0105 |