On the basis of sample entropy

this paper proposed a method for state stability analysis of gassteam combined cycle units.In this method

the sample entropy's sensitivity to fluctuation amplitude and change trend of signals was used to obtain the steady factor of the target sequences

thus to carry out selection and classification of the massive data.The sample entropies of several typical simulation signals were calculated respectively

and the relationship between sample entropy and signal-to-noise ratio(SNR)was discussed

which verified the sample entropy's good recognition ability to unsteady signals.Moreover

this method was applied in the extraction of steady data of part parameters in a gas-steam combined cycle

and the selected samples were used for neural network modeling of the gas-steam combined cycle unit

which improved the accuracy of the network model and decreased the misdiagnosis rate of test.