Split-Model MPC Architecture for Complex Systems

The use of complex algorithms to model systems is rapidly becoming a standard within the context of Industry 4.0 (I4.0). In particular, the practical inclusion of machine learning methods to create predictive models is an idea that is taking hold. Model predictive control (MPC) algorithms are uniquely suited to practically implement these modelling techniques on real systems at a controller level. One important challenge in achieving this is the large processing times often required to solve complex models as well as making time varying adjustments to models. In this research paper, a MPC adaptation designed to leverage complex modelling methods while mitigating large control loop processing time is presented and tested in simulation. The MPC adaptation methodology is tested using dynamic matrix control and shows potential for convenient use in systems with complicated dynamics.