Thermal System Performance Research On Substitution Of Mining Data For Failure Value Of Measuring Point

In large power station, the thermodynamic system is so complex that the parameters which need to be monitored is very a lot. However, these monitoring points work in the high pressure and high temperature and vibration environment, sensing device are easily damaged. So the accuracy of measurement is seriously affected. This undoubtedly bring great difficulties to the analysis work of performance trend of unit heat consumption.This paper introduces the basic principle of forecast model on the basis of the similarity modeling, set up its super ball dynamic model using normal historical operating data of a 600 MW unit regenerative system in PI database calculated the predicted value; similarity curve performance calculation model and the measuring similarity curve points which depend on the distance measure function and test points of similarity calculation formula. Depending on the method of distance measurement function, the degree of similarity of each equipment and monitoring points are computed out at normal operation time, then to get the minimum value as the similar reference value, and thus the reference value as a screening failure determination standard data. The data are regarded as failure when the real-time data received in equipment state model which is similar to the value is lower than the reference value, and this time model of the measuring point data mining prediction value as there placement of the failure data, into the performance calculation model to calculate, the normal operation to ensure the performance calculation model, and the system test.Research shows that: the performance analysis by the way of the method of ball model is highly fault-tolerant using. This characteristic make the correct result can be given out, even when the thermal parameters which are inputted are movement and damage caused by the measuring system instead of the thermal parameters. Thereby, this mothod make it possible to do the work of performance analysis and fault diagnosis correctly.