In order to improve the accuracy of power grid multiple fault diagnosis, a power grid fault diagnosis algorithm based on multi-layer data fusion is proposed. Through the analysis of the data characteristics of the whole process of fault, the fault data and the power grid structure diagram are used to model the primary equipment, and the diagnosis model of circuit breaker, line, bus and main transformer is formed. The data from different sources of information (SCADA, prudential system, recording system) is treated uniformly to form the whole process of fault intelligent diagnosis data source, and on the basis of in-depth data fusion, the quality of information is improved. This paper makes full use of the circuit breaker status and electrical information after fault to accurately diagnoses the fault and determine the fault area, and then adopts the recovery path algorithm based on fault rules to give the optimal path of the rapid restoration of power supply. The operation results show that the proposed method has strong practicability. This work is supported by Central China Division of State Grid Corporation of China (No. SGHZ0000DKJS1700140).