NARX Identification Development for a Heavy-Duty Gas Turbine

In this paper, time domain MISO NARX identification techniques are developed to estimate a General Electric gas turbine dynamic behavior. I/O data set, consists of HRSG loading sequence, and are chosen based on thermodynamics and simplified models. During identification procedure, various nonlinearities are exploited in order to make the model valid in different operating points. According to residual analysis tests results, the NARX model with the wavelet network nonlinearity and polynomial regressors is found more applicable than low order ARX models. Finally, the NARX model is subjected to the load demand variations individually. Model responses are more close to the experimental data than conventional low order models in the wide range of turbine operating points.

کلیدواژه ها:

Gas Turbine ، Identification ، NARX ، Modeling ، Nonlinear system