With the rapid development of Internet of Things, various automated monitoring systems have been widely used. As the main technical means of the monitoring system, the node operating state of the wireless sensor network has a direct impact on the reliability of the system. Therefore, in order to diagnose the faults of wireless sensor network nodes accurately and quickly, a fault diagnosis method of wireless sensor network based on improved frog leaping algorithm was proposed. First, the wireless sensor network node structure and fault classification system were analyzed. Then, the traditional BP wavelet neural network method was adopted, and the frog leaping algorithm was used to optimize it, so as to overcome the problem of local search into the problem. Simulation results show that the proposed method has higher reliability and accuracy than those of the BP wavelet neural network method.