Pretrained Battery Transformer (PBT): A battery life prediction foundation model

2025-12-18

AI Summary

Early prediction of lithium-ion battery cycle life is hindered by data scarcity and heterogeneous aging conditions. To address this, we propose the first foundation model for battery lifetime prediction—introducing a novel foundation-model paradigm for battery analytics. Our method employs a hybrid expert (MoE) Transformer architecture explicitly infused with electrochemical domain knowledge, enabling cross-dataset, cross-operating-condition, and cross-chemistry generalization. Leveraging joint representation learning across 13 public LIB datasets, domain-knowledge-embedded pretraining, and few-shot fine-tuning, the model achieves a 19.8% reduction in average prediction error on the largest publicly available battery database. It attains state-of-the-art performance across all 15 heterogeneous benchmark datasets, demonstrating significantly superior generalizability compared to existing approaches.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Cognitive Modeling & Cognitive Systems: Other Foundations of Cognitive Modeling & Systems
• Intelligent Robots: Other Foundations and Applications

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Vertical and domain-specific search

Abstract

Early prediction of battery cycle life is essential for accelerating battery research, manufacturing, and deployment. Although machine learning methods have shown encouraging results, progress is hindered by data scarcity and heterogeneity arising from diverse aging conditions. In other fields, foundation models (FMs) trained on diverse datasets have achieved broad generalization through transfer learning, but no FMs have been reported for battery cycle life prediction yet. Here we present the Pretrained Battery Transformer (PBT), the first FM for battery life prediction, developed through domain-knowledge-encoded mixture-of-expert layers. Validated on the largest public battery life database, PBT learns transferable representations from 13 lithium-ion battery (LIB) datasets, outperforming existing models by an average of 19.8%. With transfer learning, PBT achieves state-of-the-art performance across 15 diverse datasets encompassing various operating conditions, formation protocols, and chemistries of LIBs. This work establishes a foundation model pathway for battery lifetime prediction, paving the way toward universal battery lifetime prediction systems.

Problem

• Develops a foundation model for battery life prediction
• Addresses data scarcity and heterogeneity in battery aging
• Enables transfer learning across diverse battery datasets

Innovation

• Pretrained Battery Transformer foundation model for life prediction
• Domain-knowledge-encoded mixture-of-expert layers for transferable representations
• Transfer learning achieves state-of-the-art across diverse battery datasets