Process-optimizing multivariable control of a boiler system

This paper presents a method to apply multivariable controllers as optional process optimizing extensions to existing conventional control systems. The basic idea of the concept is to consider a conventionally controlled power plant boiler-85MW coal-fired drum boiler-as the process to be optimized. The control errors from the conventional controllers are used as inputs to a LQ controller. The outputs from the LQ-controller are added to the outputs from the conventional PI-type controllers. The control performance has been evaluated by computer simulation, using a detailed non-linear boiler model. The results show improved control in all cases compared to a conventionally controlled system. Furthermore the method offers an attractive and safe way to introduce multivariable controllers in power plant control.