Self-optimizing control with active set changes

► A control structure is “self-optimizing” if it is simple, model-free, and close to optimal for any disturbance. Here, we consider the case of changes in the active set. ► The work is based on the “nullspace method”, a method for selecting controlled variables and parametric programming. ► It is shown that the value a controlled variable may be used to decide when to switch regions. ► Demonstrated on a model of an ammonia production plant.