On The Abnormal Parameter Searching And Fault Warning Of Power Plant Equipment

With the comparison of the measured value and the threshold of the parameters, traditional on-line monitoring systems have realized the alarm function. However there are still some deficiencies in applying this single warning mechanism, which works after the fault appears. It can't detect the early fault signs of systems timely and tracked the abnormal trends during monitoring process, usually leaving the system in downtime and resulting in heavy losses. So an improved warning mechanism is in urgent need.Aiming at the imperfection of the existing warning mode, this thesis proposed a new system of on-line condition monitoring and early fault warning for a complex system. The system could make full use of real-time/historical database to timely indicate the dynamic changes of parameters and the health of system status.Firstly, an algorithm of searching abnormal parameters was put forward, which divided time subsequence into each subschema based on the important points, then extracted each subschema's eigenvalue respectively according to the types of parameter monitoring signal and mapped the eigenvalue into high dimension space. The system would search out anomaly parameters sequence which was obviously different from the other data. Taking into account the advantages of Grey Model and Time Series Prediction Model, the thesis proposed a new combined prediction method of time siries, using the Gray Model to predict the tendency, and applying the Time Series Prediction Model to predict the residuals. With this method, the trends of abnormal parameters were advanced found, which laid the solid foundation for the analysis of system abnormality, and the Prediction accuracy was improved.Secondly, the hierarchy of the complex system was devided effectively. The method of determining the dependence between faults and parameters was proposed, and the correlation between parameters was also studied. Moreover, the calculation of the possibility of potenitial fault occurrence was given. All studies above could provide the knowledge-based information for the analysis of system state according to the abnormal parameters.Finally, the idea of on-line condition monitoring and early fault warning for a complex system was applied to turbine. Results indicate that this system is effective and superious. It can dig out the evidence of units'deterioration from immense amounts of monitoring data and predict the potential fault. According to the abnormal system status, measures can be taken timely and effectively, leaving the potential failure nipped in the bud. In other words, forced shutdown of units can be avoided and the damage losses caused by failure can be ninimized.In short, the early fault warning mechanism can provide a solid guarantee for systems'secure and long-term operation, provide reliable decisions for intelligent maintenance, and pave the road for the company's management goal of "zero fault".