The Research Of Parameter Identification & Fault Diagnosis Of Turbine Regulating System Based On BP Neural Network

The essential task of hydro turbine regulating is to adjust the output of the active power continuously according to the load fluctuations, and keep the rotate speed (frequency) in the set range. The research on fault diagnosis system of Hydro Turbine Governing System (HTGS) is aimed to find a way to diagnose, prevent or eradicate various abnormal conditions or fault conditions, and to give some guide to the operation of hydro turbine group. This will improve the reliability, security and effectivity of turbine group, and thus minimize the loss caused by faults.According to the characteristics of control system and identification mechanism, the dynamic response of a system can fully reflect the information of the system. For the same system, the changing of model parameter indicates that some character of the system has changed, and if a fault has occurred in the system, the model parameter must not remain in the normal span. Mathematical model is the recurrence of a real system in some accuracy range. So it is feasible to diagnose faults by identifying whether the parameter of the mathematical model is normal. Different types of faults have different changes in the parameter, which can help to classify, locate the faults and analyze their situation and causes. Based on the above, this paper brings forward the method of Fault Diagnosis based on System Parameters (FDSP) to study the fault diagnosis of HTGS.Basing on the modeling and emulation of the BP Neural Network identification theory, this paper has identified the parameters of different parts of the HTGS. This provides a foundation to monitor the condition of HTGS. Basing on this foundation, the condition evaluation of HTGS, can be done, which provides a platform for the fault diagnosis of HTGS, based on abnormal parameter.For the complex features of HTGS, and its huge physical structure, the amount of parameters which need to be identified is enormous, and their features are far different. In order to do the research, it's necessary to subdivide these problems. This research is a process from simple to complex, from part to the whole, and from linear to nonlinear. Analysis of the system itself is first based on the fact that the system identification is correct and then this research method is an analogy in the research of the abnormal situation in the system.