High-pressure heater is an important component in turbine regenerative system. Long-term operation at high temperatures and pressures can easily cause the pipe leakage

which affects the safe running of units when the leakage turns serious. In this paper

a horizontal high-pressure heater of a large thermal power unit was taken as the research object

a fault simulation model of high heater leakage was built up on the basis of a discrete numerical thermal calculation method. The data set related to the faults was obtained. Moreover

a fault diagnosis model was constructed through the improved BP neural network. Furthermore

the model was used to solve the leakage fault of No.3 high-pressure heater in a 600 MW steam turbine system. Compared with other diagnostic models

the results show that this improved BP neural network model has high accuracy and reliability.