Multi model predictive control of a power plant heat exchanger network based on gap metric

This paper is focused on centralized model predictive control of both superheated and reheated steam temperatures of a supercritical power plant with once through boiler. Steam superheating and reheating is performed by a network of several heat exchangers, spray attemperators and diverting valves. This network is strongly nonlinear and it cannot be controlled by a single linear controller. At the same time, tight temperature control is essential for plant efficiency and prevention of plant damage due to thermal stresses. This paper proposes a multi-model predictive control scheme based on linearized first principles models of the heat exchanger network. The selection of individual models is based on gap metric nonlinearity measure. The performance of this control system is evaluated in the wide range load following operation (from 50 to 100% of the nominal load). It is shown that there is a significant control performance improvement over classical control system based on gain scheduled PID/PI controllers.