Abstract: With the development of new energy power generation, electric vehicles, etc., the requirements for energy storage are constantly increasing. Lithium-ion batteries are widely used in various energy storage systems due to their advantages of environmental friendliness, high energy density, and long lifespan. Providing reasonable thermal fault diagnosis for lithium-ion batteries can avoid thermal runaway and ensure safe and reliable operation of the batteries. This study proposes lithium-ion battery intelligent perception (LBIP) to build a thermal fault diagnosis model for lithium-ion batteries. LBIP includes the backbone for feature extraction, region proposal network (RPN) for proposals generation, and fine-grained localization. The Ansys Fluent software is selected for finite element simulation of lithium-ion batteries. The model processes the thermal imaging image of the battery surface, identifies the problematic battery, and localizes the problematic battery. Results shows that the recognition accuracy of the faulty battery can reach 95%.