Application Research Of Model-Based Diagnosis In Fault Diagnosis For Traction Power Supply System

The safety and reliability of traction power supply system are important prerequisites for the safety of trains. Once the traction power supply system fault occurs, not only the operation condition of the train in the fault area will be affected, but also the operation condition of other trains in non-fault area will be affected as a chain reaction, which may result in significant economic losses. In order to find the fault causes quickly and correctly, and take measures to restore power as quickly as possible, a lot of artificial intelligence systems have been developed in last decades, in which expert systems are the most successful. But the expert systems are based on experts’experiences which need a long time to obtain. The method of model-based diagnosis can diagnose the system according to its structure and function, which avoid collecting experts’experiences. Therefore the application and research of using model-based diagnosis in traction power supply system’s fault diagnosis are done in this paper, and the key problems are analyzed and discussed.Firstly, in view of the shortcomings of existing intelligent search algorithms in computing the minimal hitting sets, a hybrid evolutionary algorithm (BPSO-GA) based on Binary Particle Swarm Optimization (BPSO) and Genetic Algorithm (GA) are proposed. Furthermore, the fitness function, the minimal hitting set assurance strategy and the termination condition of BPSO-GA are discussed and analyzed. The results of programming and testing show that the new algorithm is better than the standard particle swarm optimization in search efficiency. Compared with the standard genetic algorithm, it has a faster convergence speed. Most of all, the new algorithm is not sensitive to the problem scales and it can find out almost 100% minimal hitting sets.Secondly, the method of model-based diagnosis is applied in the diagnosis of traction power supply systems and a scheme of using the model-based diagnosis in traction power supply system’s fault diagnosis is proposed. As an example, the Yongle Section in Beijing-Tianjin Railway Line is used to describe the realization methods of the key steps in the diagnosis process, such as the abstract representation of system components, the modeling of system, the recognition of the minimal conflict sets, the acquisition of the fault diagnosis candidates and the fault matching. Through programing and solving the fault data obtained from the Simulink, the results show that the model-based method can diagnose system faults quickly and accurately. Meanwhile a strategy is proposed to avoid the influence caused by the measurement device failure. And the experimental results show that the diagnosis system using the proposed strategy can also find the system fault when some measurement devices break down.Thirdly, considering the characteristics of traction transformer faults, the faults are divided into sudden electrical circuit faults and latent faults. For the sudden electrical circuit fault, the consistent reasoning and adductive reasoning are mixed to diagnose faults, and the experimental results show that the mixed reasoning is feasible. For the latent fault, a fusion method using the combination of consistent reasoning and expert systems is proposed and its realization process is analyzed.In the end, the analytical model is introduced into the fault diagnosis of traction power supply system. The diagnostic problem based on analytical model is transformed into an optimization problem which can explain the warning information by analyzing the action of protective relays and circuit breakers. The result of experiment shows that the method based on analytical model, which fully uses the topology of power system, the data of relay protection and the actual alarm information, can determine the actual action of protections and circuit breakers and also can judge the leaking information and misinformation.