In view of the demand for intelligent operation and maintenance of distribution networks and the trend of digitalization of the smart grid, a comprehensive evaluation and intelligent early warning method for cable passages based on improved adaptive network-based fuzzy inference system (ANFIS) in the edge computing scenario of the distribution internet of things is proposed. First, leveraging the data collection and processing computing advantages of edge IoT terminals, a comprehensive monitoring system is designed based on the device perception, IoT network, data platform, and application display layers. Then, to minimize communication and computation delays, a hierarchical edge computing model is constructed, and corresponding task unloading and scheduling schemes are proposed from three aspects: real-time value density, execution urgency, and importance of quantitative analysis to improve resource utilization and task execution efficiency. Finally, the adaptive fuzzy inference system is improved by phase space reconstruction, and used for comprehensive evaluation of cable passages operational status after iterative training. The cable passages are calibrated to different risk levels through hierarchical clustering, and the feasibility of the model is verified through engineering experiments in a case analysis.