Subspace identification for disturbance rejection control design in gas turbines

This paper discusses approximate system identification of a gas turbine operating with closed loop feedback control of the fuel and airflow. The objective is to obtain multi-input multi-output linear dynamic models to capture the system dynamics over a wide operating range of the turbine. In the proposed approximate modeling, the dynamics of a gas turbine at multiple operating conditions is described by a single multivariable model with additional scaling to account for gain variations. A covariance-based subspace system identification algorithm is applied to obtain dynamic linear models of a generalized plant model of the combined controller and gas turbine from data generated from a first principles simulation. The procedure is demonstrated to sufficiently replicate the operation of closed loop system to enable a single generalized linear model to replicate the nonlinear turbine dynamics and enable subsequent robust control design.