Optimal partitioning of a boiler-turbine unit for Fuzzy model predictive control

In order to effectively control the multi-input multi-output (MIMO) characteristics of a boiler-turbine unit Fuzzy model predictive control (FMPC) has been previously applied successfully. An important step in the FMPC design is the partitioning (also called the design of the antecedent rules). In this work it is shown that by exploiting the plant characteristics and by using suitable metrics of the nonlinearity a representative one-dimensional partition space can be defined. Furthermore, it is demonstrated that a simple partition into two linear models suffices to obtain a stable and performing control system. In order to achieve this simple FMPC a nonlinear stationary feed-forward decoupling is implemented. Simulation results demonstrate stability and performance of the resulting control system even for large reference steps. The methodology developed in this work is quite general and can be easily applied to the FMPC design for other nonlinear MIMO plants.