Large Language Model-Based Intelligent Pre-Design of Lithium-Ion Battery Accelerated Life Tests

Purpose: This study proposes an intelligent pre-design method for lithium-ion battery accelerated life tests (ALT) based on large language models (LLMs) operating in a local environment, in order to address risks related to design errors and data security.<br/>Methods: A local retrieval-augmented generation (RAG) system, utilizing an open-source LLM and a vector database, derives intelligent pre-design results using literature-based thresholds. The effectiveness of the design is validated by reviewing analytical results from actual datasets to ensure compliance with design requirements.<br/>Results: Optimal thresholds for LiCoO2-graphite cells were derived to generate pre-design reports. Design validity was verified through comparisons with actual datasets and supporting literature. High consistency and accuracy were demonstrated across 10 repeated experiments, confirming the practical stability of the proposed approach.<br/>Conclusion: The proposed method serves as a practical tool for improving design precision and analytical objectivity. It enhances the efficiency of battery reliability test design by safeguarding technological assets and automating the review of relevant industrial standards.