The use of Model Predictive Control (MPC) is commonplace in many industrial applications. The anticipative nature of MPC and the inclusion of physical constraints into the control framework presents many advantages over classical control strategies. Despite these advantages, obtaining an accurate open-loop model of the underlying process is often a difficult and time consuming process. In this paper, a methodology is introduced to identify linear open-loop models of gas turbine engines from closed-loop data. The closed-loop data can be obtained by any sufficiently informative experiment from a plant in operation or simulation. We present simulation results here. These open-loop models are then used in the design of model predictive controllers at a number of operating points of the turbine. The predictive controllers we designed include physical constraints on the fuel and air flow into the turbine. The performance of these predictive controllers is compared in simulation against existing classical control techniques in a number of typical operating scenarios including off loads, on loads and set point changes.

This content is only available via PDF.
You do not currently have access to this content.