Most existing fault diagnosis approaches based on Petri net are applied in transmission network, and usually utilize the action information of relay protection and breaker to form the diagnosis model which is not suitable for radial distribution network. In this paper, a novel Petri net model based on fault information of FTUs is proposed. The rapidity and fault tolerance of distribution network fault diagnosis are both guaranteed through comprehensive utilization of the uploaded over current signal and the amplitude measurement information from FTUs. According to the topology characteristics of distribution network, a hierarchical diagnosis approach is employed to identify the fault feeder, fault unit and fault section successively. The simulative ability of the proposed diagnosis model is enhanced by utilizing the probability information and Petri net with inhibitor arcs. The incidence matrix and the state equations are used to describe the model. Simulation results verified the effectiveness and feasibility of the proposed approach. The high commonality of the method makes it possible to be applied in distribution systems with renewable energy sources such as solar photovoltaic generations and wind generations.