Research on Nonlinear Modeling Method of Gas Turbine based on Hammerstein-Wiener
The modeling of a gas turbine serves as the foundation for its control,performance optimization and state monitoring.This paper proposed a nonlinear modeling approach for gas turbines based on Ham-merstein-Wiener methodology.Firstly,the dynamic linear model of the gas turbine was established using the transfer function.Then,the piecewise linear function and the wavelet transform network were em-ployed as the functions of the nonlinear input and output links,respectively,to depict the gas turbine's nonlinear characteristics.And the loss function was constructed.The information vector of the model was decoupled from the parameter vector,and the parameter vector was estimated using the Gauss-Newton nonlinear least squares method based on actual operational data of the gas turbine,thereby obtaining a nonlinear model for the gas turbine,which was compared with ARX,BP and CNN models.The test data was utilized to validate the nonlinear model's extensibility.The results demonstrate that the established nonlinear model exhibits enhanced accuracy.The maximum root-mean-square error(RMSE)is 0.010 9,the maximum mean absolute percentage error(MAPE)is 3.305 4%,and the minimum FIT is 0.998 7,indicating the model's capability to accurately fit actual gas turbine monitoring data.
gas turbinenonlinear modelingHammerstein-Wiener modelGauss-Newton method