International Journal of Control, Vol.74, No.11, 1118-1130, 2001
Multivariable model predictive control of a thermal power plant with built-in classical regulation
Multivariable control systems are preferable to classical regulations based on frequency decoupling in plants where the dynamics of the controlled variables are strongly interacting. A multivariable solution generally lead to better performance, since the plant is considered as a whole, allowing the optimization of the control action. Nevertheless, such solution is scarcely adopted in practice. In fact, classical systems are preferred since they allow the operator to use his experience when necessary, acting manually on single parts of the plant. The design of a multivariable control system that corrects the action of classical regulation would allow both optimal performance and maximum flexibility. In this paper, a novel solution based on a multivariable control added to the traditional structure is proposed for a conventional power plant. Simulations show that the improvement in performance is significant, especially in the critical situation of a sudden change in the load request.