Fault Diagnosis Method Research And Application Of Gas-steam Combined Cvcle Generating Unit

With the rapid development of our gas-steam combined cycle power generation technology, a large number of combined cycle power plants have been constructed around China’s coastal areas, due to the situation of complex structure of these units,high operating parameters and frequent variable load, all kinds of faults are prone to happen, so it is essential to carry out fault diagnosis study for safety and reliable operation of combined cycle power plants.Contents of this research can be divided into two modules, the first one is power plant sensors fault diagnosis module, the measurement data of sensors is used to compute the performance of the units and give some corresponding decision, the accuracy and reliability of the sensors and must be guaranteed, but some sensors work in the harsh environment of high temperature, high pressure, high corrosion and often alternating load, various types of failures will occur. In this paper,after the mechanism group analysis,a mathematical model was built with historical data between diagnosed sensor and relative sensors to achieve real-time monitoring of diagnosed sensor, NN-PLS method proposed here can assess the average contribution rate of each independent variable on the dependent variable in the model and filter out the main modeling parameters,these parameter are trained to get a good the spline transform PLS model which has an excellent nonlinear fitting ability and relatively few parameters,so the model can be used for online prediction and fault diagnosis. As a validation,the sensor of active power of gas turbine was chosen to check the model accuracy because the measurement data of this sensor is relatively accurate,the result show that the NNJPLS model has a good prediction accuracy at different loads, and in a reasonable set of thresholds and window length, sliding window threshold method is able to effectively detect the occurrence position of various typical sensor faults and give a right fault category.Another module is fault diagnosis module of the gas turbine flow passage,after the analysis of typical fault types and the corresponding change of the characteristic parameters,the similar flow parameter was chosen as a basis of the fault diagnosis. Three groups of known-fault actual data of the power plant were selected as the validation sample,the diagnostic results match the known fault information perfectly,so the diagnostic method is proved to be effective.