In order to monitor the condition of the heat recovery steam generator(HRSG) and to ensure the healthy operation of the HRSG

the three-pressure main steam temperature and pressure prediction model was established by using the data from the healthy operation of HRSG and combining the three methods of principle component analysis(PCA)

sparrow search algorithm(SSA) and long short-term memory(LSTM). PCA was used to reduce the input parameters of the model from 22 to 9 dimensions

and taking the reheat steam temperature prediction model as an example

it was concluded that the model with PCA dimensionality reduction reduced the hyperparameter optimization time by 11.3% compared with the model without PCA dimensionality reduction.Compared with the model without SSA

the value of coefficients of determination of these models is significantly improved

mean absolute error and root mean square error are significantly reduced

and the alarm threshold of the main steam temperature HRSG is determined according to the distribution of absolute error. Therefore

the condition monitoring model of HRSG based on PCA-SSA-LSTM has short training time and high prediction accuracy

and the model provides theoretical basis and technical support for fault monitoring and diagnosis of HRSG in gas turbine combined cycle power plants.