The fast and accurate diagnosis of fault components is one important precondition to realize the on-line regulation and operation of power grids. For the sake of its understandably modeling and precise logic, the Petri net fault diagnosis method are widely studied. However, most current methods only consider the digital information rather than the analogue information, despite that the analogue information is more accurate, fault tolerant and so on. Thus, a fault diagnosis method of power grid based on Petri net is proposed by considering both the analogue and digital information. Firstly, the fault current of the components is extracted by Hilbert-Huang transform, and the fault probability of the components is obtained through fuzzy logic analysis and fuzzy logic Petri net reasoning. Secondly, based on the digital information, the Petri net model of the components is established, and the constraint of fault sequence information is probed, and the fault probability is obtained by fuzzy reasoning calculation. Finally, D-S evidence theory is utilized to explore the fault probability level to obtain the final failure probability of components. Several fault cases of the New England 10 machine 39 bus system are tested. The results show that the method can quickly diagnose fault components, improve the limitations of the previous single fault source and single diagnosis method, and hence effectively improve the accuracy and fault tolerance of fault diagnosis.