Power System Fault Diagnosis Based On High-level Petri Nets

Accurate diagnosis of power system failure has important significance for accelerating accident treatment and restoring safe-economical operation of power system. Massive alarm information sent by local signal generator assembles at dispatch center when complex power grid failure occurs. How to help operators identify fault elements accurately in a short period is a critical and urgent research problem. Loss and distortion of the alarm information also further increases the difficulty to find out fault elements. Among many methods proposed by the research scholars, a power system fault diagnosis method based on the Petri nets has attracted more and more attention for its simple reasoning and graphical description of the topology structure and the dynamic characteristics of the power grid.In order to narrow the searching range of power system failure and avoid fault judgement of all components successively, an identification method of fault area in power system based on colored self-modifying Petri nets is proposed. Through this method, the fault diagnosis problem is limited to one or more passive network. First, a colored self-modifying Petri nets model used to detect the fault area is built based on Petri nets. This model can interpret the power network topology change, a discrete event dynamic process, caused by the breakers trip. All components are assigned with color attributes and the weights of some directed arcs are variable in the model. Then, searching of the fault area is realized by Java in the Net Beans, the developing environment of the software.In order to reduce error caused by artificial subjective factors in power system fault diagnosis process and improve fault tolerance and accuracy of fuzzy Petri net models, a power system fault diagnosis method based on improved dynamic adaptive fuzzy Petri nets and back propagation algorithm is proposed. First, the general Petri net models are built by introducing complementary arcs tuple in dynamic adaptive fuzzy Petri nets, which is able to dynamically adapt to updated fuzzy knowledge in expert systems. Second, the back propagation algorithm of neural network is used to train model parameters. Finally, the adaptability and the fault tolerance of the algorithm are analyzed.It is indicated that the proposed method can make full use of the parallel processing capabilities of Petri nets by the simulation results on 8-bus testing system and Siping actual power system in Jilin province. Simple and clear in derivation, satisfying diagnosis results against single fault, multiple fault and the complex fault with malfunctions of several protections and breakers can be obtained by the proposed algorithm in this paper, also with good fault tolerance.