Modeling Heavy-Duty Power Plant Gas Turbine Using NARX Identification Techniques

In this paper time-domain NARX (Nonlinear Auto Regressive with eXogenous inputs) identification techniques are developed to estimate a General Electric gas turbine dynamic behavior which is mounted in Montazer Ghaem power plant. Related I/O signals are chosen based on thermodynamic and first order linear models by dividing the system into three MISO subsystems. Compressor discharge pressure, exhaust gas temperature and electrical power are chosen as system main outputs which can expressed by fuel flow, shaft speed and guide vanes position considering the ambient temperature effects. During identification procedure various nonlinearities are exploited in order to estimate these three outputs. According to residual analysis validation tests, the NARX model with the wavelet network and polynomial regressors is found applicable to the case study. Also comprehensive results perform that this model outputs is more close to the experimental data than conventional ARX models and can predict system behaviors perfectly.

کلیدواژه ها:

Gas Turbine ، System Identification ، NARX ، ARX ، Power Plant