Abstract: Aiming at the problems that distribution network live working robots cannot predict, prevent faults, and the maintenance response is delayed after faults occur, an intelligent fault diagnosis method based on multi-source heterogeneous data fusion and a three-level diagnosis model is proposed. Through clock synchronization of each component of the robot, a spatiotemporal synchronization acquisition mechanism is constructed to obtain multi-source data. A dynamically adaptive weighted fusion algorithm combining Kalman filtering and fuzzy logic is adopted to eliminate data heterogeneity and spatiotemporal deviations. By establishing the three-level diagnosis model, fault prediction and maintenance scheme generation are realized. Experimental results show that fault prediction reduces the probability of faults by 90%, automatic recovery shortens the fault recovery time to 9s, and the generation of maintenance schemes reduces the manual fault recovery time to 1min, pushing maintenance plans for potential hazard components avoids the occurrence of faults, which effectively improves the operational reliability and maintainability of the robot.

Key words: distribution network live working robot, intelligent fault diagnosis, multi-source heterogeneous data fusion, three-level diagnosis model, clock synchronization