Abstract: With the constant consumption of traditional fossil fuels, people have gradually realized the importance of protecting the environment. Therefore, in recent years, new clean energy such as wind power has gradually attracted attention from all walks of life, and new energy vehicles have also gradually replaced the traditional car rapid development. Among them, lithium-ion battery, as the energy storage equipment of clean energy and the power source of new energy vehicles, has attracted attention from all walks of life. However, lithium-ion batteries are prone to thermal runaway failure during use, and their safety cannot be guaranteed. In order to ensure the safe operation of energy storage equipment and new energy vehicles, aiming at the phenomenon that the internal short circuit fault of lithium ion battery causes runaway heating, the fault diagnosis methods of whale optimization algorithm optimization variational mode decomposition (WOA-VMD) and particle swarm optimization support vector machine (PSO-SVM) are proposed to diagnose the internal short circuit fault voltage signal of lithium battery. Firstly, the short circuit fault voltage signal in the lithium battery is decomposed by VMD to obtain a series of natural mode components. There are two parameters in the VMD algorithm that have a great impact on the decomposition results, namely, the number of decomposition layers K and the penalty factor α， In order to achieve the best decomposition effect, WOA is used to find the VMD decomposition level K and penalty factor α And get the optimal decomposition level K and penalty factor α The parameter combination of is: K= 10, α= 1997, the optimal parameter combination was introduced into VMD decomposition to decompose the internal short circuit fault signal of lithium ion battery to obtain 10 modal components; Then, for the 10 decomposed modal components, the sample entropy of each modal component is calculated and used as the eigenvector; Finally, the eigenvectors are input into the SVM model respectively, and then the PSO optimized SVM model is used for fault diagnosis and the diagnosis results are output. The final results show that the diagnostic accuracy of the direct SVM model is stable at 66.667%, and the diagnostic accuracy of the PSO optimized SVM model is stable at 100%. Compared with the direct SVM model, the internal short circuit fault of lithium ion battery in the SVM model after feature selection by particle swarm optimization algorithm is effectively identified, which proves the effectiveness of the PSO-SVM diagnostic model.