Intelligent Positioning And Diagnosis For The Leakage Faults In The Water Steam Circulation System Of Boiler Turbine Unit

Leakage is a common fault of the power plant steam-water circulation system which seriously affects plant operation safety and economy. The entire steam circulation system including condensate, feed water, regenerative cycle, boiler water circulation, superheated and reheated steam and other subsystems. The steam-water working medium in each subsystem is interconnected and tightly coupled. When the leakage occurs within a subsystem, it not only impacts the parameters itself, but also influences the relevant subsystem. At present, most researches are for a single subsystem, thus misdiagnosis is not easily avoided when fault occurred in other subsystem leading to parameters abnormal changing in this subsystem. In addition, a power unit may operate at various load conditions, while current fault diagnosis researches are most for a single operating condition(such as the rated load point),coupled with the complex structure of the thermodynamic system and faults diversity,the application of fault diagnosis system in actual power plant is greatly limited.For this reason, leakage faults occurred in the unit water-steam circulation system,are better to be overall considered by fully taking account of the coupling characteristics among various subsystems. Therefore, this paper puts forward a fault diagnosis method for power plant steam circulation system based on two-stage neural networks. Among them, the first-level neural network model is employed to locate the subsystem of the fault occurred, and the second-level BP neural network fault diagnosis models are then used to identify the specific fault types. Symptom optimization technique is also employed to improve the fault diagnostic effect for faults of varying severity under different load conditions.This paper takes a 600 MW supercritical unit as the object investigated and fault diagnosis study is carried out with its full-scope simulator. By simulating of typical leakage faults in the water-steam circulation system, a series of fault samples are extracted and symptom fuzzy calculating method is adopted to construct the knowledge base for fault diagnosis neural network models development. According to simulation operating data at different load conditions, interpolation method is used to predict the expected normal values of the characteristic parameters. The fault positioning and diagnosis program is developed with Matlab software combined with symptom zoom technology. By communicating with the simulator, detailed fault diagnosis simulation tests are carried out, which shows that the suggested method can achieve good diagnosis results for the power plant steam circulation system at different load conditions and varying degree of faults.