AI Creates Lithium Battery Database: 3,000+ Papers Mined

An AI agent has created a comprehensive database for lithium metal batteries. This agent, called LLMB, mined over 3,000 scientific papers. It extracted data from 3,606 papers, building a database of 8,074 battery cells. This database includes details on components and how long the batteries last. LLMB achieved high accuracy in its data extraction, with a 96.4% F1 score for cell name extraction and 99.3% for data merging. The agent also features machine learning models that can predict battery performance. For NCM-based batteries, it predicts initial capacity and 50th-cycle capacity with R2 scores of 0.75 and 0.69, respectively. It can also identify how solvent polarity affects battery performance. The research even validated new solvent systems, finding that low-polarity solvents can improve battery performance. This technology could significantly speed up the development of better batteries.