Observer-Based Model Predictive Control of Rotary Kilns - Application to Limestone Calcination

This paper presents an observer-based model predictive control approach of rotary kilns. A one-dimensional, object-oriented modeling approach of rotary kilns is used as the basis for the control scheme. After linearization of the nonlinear modelica model at an operating point, the generated LTI model is reduced by applying a balanced truncation method. The reduced LTI system, i.e. the result of an equivalence transformation of the original LTI system, is then used for the observer design. The process observer is enhanced by disturbance models to cope with input and output disturbances. Using the estimates of all immeasurable process and disturbance variables, a linear model predictive controller is applied. With this design, disturbances, stemming in particular from the input, are dealt with. Also, the economically needed weighting of the actuated inputs is dealt with implicitly by the MPC algorithm. The main goals of this approach are stable operation, robustness against model uncertainty and reduced energy consumption.