CATL Intelligent Cell Design Platform Combines Physics-Based Models and Machine Learning

CATL's intelligent design platform for lithium-ion cells is trained on more than 100,000 battery design cases, 600TB of test data and aftermarket data from a broad range of NEVs, reaches prediction accuracy of up to 95 percent, and generates design recommendations in seconds and virtual cell designs in minutes. The World Economic Forum honored the project, "Augmented Intelligence Leading Next-Generation Lithium-ion Battery Design," with its 2026 MINDS Award; only 15 organizations worldwide were selected this year, judged across strategy, talent, data, technology and governance. Ni Jun, Chief Manufacturing Officer and Co-President of Engineering Manufacturing System of CATL, said: "In complex industrial scenarios like battery production, relying solely on 'Artificial Intelligence' can't solve core problems—I advocate more for 'Augmented Intelligence'. By integrating our intelligent design system with engineers' in-depth understanding of material properties, process principles and system engineering, we can strengthen the human–AI partnership and therefore ensure consistent, reliable product quality." Traditional lithium-ion cell design has long depended on individual engineers' experience and extensive manual testing, leading to prolonged evaluation cycles, high prototyping costs and repeated trial-and-error iterations. CATL's Intelligent Cell Design project shifts the development model from reverse design and experimental trial-and-error to forward design and prediction before production, drawing on the company's strengths across materials science, cell design, manufacturing processes and equipment. The platform combines physics-based electrochemical models with machine learning. It runs on an on-premises private cloud and draws on more than 50 million data records. With physics-informed machine learning and agentic AI, it works like a "digital engineer," automatically generating, evaluating and refining design options in a closed loop of data, physics and computation. It evaluates designs against electrochemical constraints, applies AI-driven optimisation, and can dynamically adjust design priorities to accommodate highly customised performance requirements. CATL said it will keep advancing AI from "enabling" to "creating," and from solving known problems to discovering new frontiers.