LiBMap: A Large-Scale Simulation-Derived Design–Performance Dataset for Lithium-Ion Batteries

Lithium-ion batteries (LIBs) play an important role in the development of new energy technologies, yet their design still relies heavily on costly trial-and-error experiments and simulations. Artificial intelligence has shown considerable potential for solving complex design problems, but its application to LIB design is constrained by the scarcity of design–performance data. To address this limitation, we construct a large-scale design–performance dataset for LIB design—the first of its kind, to the best of our knowledge. The dataset integrates cell format, geometric dimensions, material systems, kinetic parameters, and porous-electrode design parameters into a unified design space. Using material balance calculations and a coupled electrochemical–thermal model, we generate approximately 770,000 battery designs and corresponding performance data through large-scale parallel computing. Each sample contains over one hundred design parameters and 29 performance metrics covering typical operating scenarios in energy storage, electric vehicles, and consumer electronics, together with complete electrical and thermal process data for each metric. The dataset provides complete design information, comprehensive performance characterization, and fine-grained process data. Furthermore, we develop a retrieval–generation framework that uses knearest neighbor search to retrieve local data and fits the retrieved data using radial basis functions. The framework requires no large-scale neural network training and has low computational cost. For both LIB performance prediction and inverse design, a single task can be completed in tens of milliseconds; relative to simulation results, the mean absolute percentage error (MAPE) for both tasks is below 5%. We believe this work will provide a data foundation for LIB design and further advance LIB design capabilities.