Since the beginnings in the chemical and process
industry, model based predictive control strategies have become
widely accepted. Often mentioned success factors for MPC
are the use of optimization based on a plant model, the
consideration of constraints, and an intuitive tuning. Indeed,
if a nominal plant and overall objective are known, the tuning
can become straightforward. However, as soon as disturbances
have to be taken into account, the tuning effort increases and
becomes less intuitive. Against this background, a novel strategy
to address the issues with unknown disturbances is proposed.
The idea is to separate the nominal tuning process and extend
the control by an outer loop, which ensures offset-free control.
The inner, nominal loop decouples the system and essentially
leads to a first order response. This inner loop addresses the
performance targets in the nominal case, and the outer loop
provides offset-free control in case of unknown disturbances.
The outer loop consists of feedback controllers adapting the
reference, which due to the decoupling can be tuned by known
guidelines. The proposed strategy is presented and evaluated
using a simulated case study.