Integrated leakage fault positioning and diagnosis for water-steam circulation system of a supercritical power unit at different load conditions

A novel approach based on two-stage neural networks is proposed for integrated leakage fault positioning and diagnosis in the water-steam circulation system of a supercritical power unit. By constructing leakage-related ratio-parameters as the model inputs, a primary neural network model is employed to locate the subsystem fault occurred. Then, the secondary BP neural network model is applied to identify the specific fault types in each subsystem. Symptom optimization technique is also employed to improve the diagnostic effects for faults of varying severity degrees under different loading conditions. By taking the water-steam circulation system of a 600MW SC power unit as an object studied, detailed fault diagnosis tests are carried out on the full-scope simulator of the given unit. The results show that the proposed method can achieve good diagnostic results for varying-degree leakage faults in the water-steam circulation system under different loading conditions.