Nonlinear identification of a gas turbine system in transient operation mode using neural network

In this paper ANN (Artificial Neural Network) identification techniques are developed to estimate a General Electric frame 9, 116MW combined cycle, single shaft heavy duty gas turbine dynamic behaviors during loading process based on available operational data in Montazer Ghaem power plant in Karaj. Related Input and output data are chosen based on thermodynamics and first order linear models. Electrical power and exhaust gas temperature are chosen as system main outputs which can be expressed by fuel flow, shaft speed and compressor inlet guide vanes considering the ambient temperature effects. The operating condition of the gas turbine during identification procedure is considered from full speed no load to full load. Comprehensive results perform that this model outputs is closer to the experimental data than conventional NARX models and can predict system behaviors perfectly.