Experience with NARX model identification of an industrial power plant gas turbine

This paper reports the experience with the estimation of a nonlinear autoregressive with exogenous inputs (NARX) model for the power plant PGT10B1 general electric-Nuovo Pignone gas turbine nonlinear model. Two operative conditions of the turbine are considered: the droop mode and isochronous mode. This two conditions have forced the identification scheme to operate, respectively, in closed and open loop. In isochronous mode, the turbine supplies power directly to a specific load at the factory and the frequency need to be controlled acting on the FSR (fuel stroke reference), representing the opening percentage of the fuel valve. In the droop mode, the frequency is imposed by the network but the direct power need to be regulated. In order to study the stability of the power generated and the electrical network, an efficient, reliable and low complexity prototype model of the turbine need to be developed.