Fault diagnosis method for water vapor circulation system of large steam turbine

The invention discloses a fault diagnosis method for a water vapor circulation system of a large steam turbine. The method comprises the following steps: S1, segmenting working conditions of the water vapor circulation system; s2, constructing a normal data set and a fault data set of the water vapor circulation system; s3, training the recurrent neural network of the water vapor circulation system, and establishing an RNN recurrent neural network model for different working conditions according to the loop series relation of the water vapor circulation system; s4, performing recurrent neural network prediction on the water vapor circulation system to obtain a prediction result of the whole system; s5, carrying out fault output, carrying out similarity calculation on a predicted value and a normal data set, and if a similarity result is greater than a threshold value, considering that the system is normal; the prediction value and the fault data set are subjected to similarity calculation, if the system is no