The Fault Diagnosis System For Thytistors In TCR Of Static Var Compensator

In this paper, the development of fault diagnosis and fault-tolerance control is introduced , and the application of various methods in recent years for fault diagnosis of power electronic circuit, especially those based on neural network, are retrospected at first. Then the shortcomings of BP algorithm are analyzed after introducing the three common structures of neural network and its learning methods, consequently the combination of genetic algorithm and BP algorithm is adopted to determine the structure parameters of neural network in order to overcome the shortcomings of BP algorithm, and the simulation in MATLAB proved its superiority. Thirdly, the circuit for receiving signals and man-machine conversation based on DSP TMS320F240 are designed and the corresponding programs in assemble language are devised, including diagnosis part, keyboard and display service, communication with monitor, neural network calculation, external interrupt service routine, timer interrupt service routine etc. This diagnosis system, which perform the fault diagnosis for each thyristor of TCR of large SVC using neural network, can give on-line real-time detection to each thyristor and its trigger circuit as well as its circuit for receiving signals, can calculate the total number of the fault of each phase, and can also display the state of each thyristor, if it has any problem, the system would record the fault, including the location of the component, the fault type which is numbered, and the time which appeared, then send these information to the monitor. At the end of this paper, the assemble program for neural network and its calculation process and error are presented, and the diagnostic result for corresponding fault is offered in photograph.