Control of gas turbine power plants by means of the weighted one-step-ahead adaptive technique

Abstract

The one-step-ahead controllers represent a branch of minimum prediction error adaptive controllers. They combine the parameter estimation of the controlled system model with the control scheme; therefore, they are especially suitable for non-linear and time-varying systems. Since the estimated parameters are updated at each time step (by using the sampled data), these methods can be adopted for non-linear applications. Consequently, the one-step-ahead controllers do not require knowledge of the dynamic characteristics of the controlled system (e.g. state-space systems or transfer functions). Sometimes, in the gas turbine field, the control system could produce an excessive control effort, due to sudden variations of the electric load. In order to reduce this control action, the weighted one-step-ahead adaptive (WOSAA) control algorithm considers a penalty associated with the control effort using an appropriate cost function. In this way, the control variable does not assume values that are too large, even when the gas turbine undergoes sudden changes in the external load. As a consequence, the robustness and stability features of the WOSAA control system are increased. The results show that the WOSAA control technique, applied to both the double-shaft aero-derivative gas turbine and the single-shaft heavy-duty gas turbine, effectively counteracts the load reduction with limited overshoot in the controlled variables with reduced control effort.