Application of a Multivariable Self-Tuning Controller to a Power Plant Boiler

A multivariable self-tuning controller is applied to a power plant boiler for control of three outputs using three input variables. It is assumed that at a constant load, the relation between the selected inputs and the outputs can be closely approximated by a linear autoregressive (AR) model. The order of the autoregressive model and the time delay in the system are identified off-line. The results are used to design the multivariable self-tuning controller. Under this adaptive controller, some of the controlled outputs show relatively large deviations from their set points during transient conditions. These deviations can be greatly reduced if the parameter estimator of the self-tuning controller closely tracks the relatively fast changing parameters of the system, or if a gain scheduling is used with the self-tuning controller. Also, in the case of self-tuning control/gain scheduling, increasing the frequency of the control calculations does not result in further improvement.