The real-time and accuracy of power system fault diagnosis are facing challenges, especially in the case of multiple equipment failures and complex environments. This paper aims to improve the intelligent fault diagnosis capability of digital substations through multimodal data fusion technology of Internet of Things and artificial intelligence. First, multi-source data of the substation is collected, including sensor information, equipment operating status and environmental parameters. Secondly, the Internet of Things technology is used to realize real-time data transmission and upload the data to the cloud for centralized storage. Then, the CNN model is used for feature extraction, and data from different modes are fused to build a fault diagnosis model of multi-modal data fusion. Finally, the model is iteratively optimized through hybrid training and verification methods. The model achieves a fault identification accuracy of 96.7% and shortens the response time to less than 1.1 s. The multimodal data fusion method based on the Internet of Things and artificial intelligence significantly improves the fault diagnosis efficiency and accuracy of digital substations, providing strong support for the intelligent development of power systems.