The foundations of Data Predictive Control
Decisions on how to best optimize large scale energy systems and industrial operations are becoming ever so complex and conflicting, that model-based predictive control (MPC) algorithms must play a key role. However, a key factor in prohibiting the widespread adoption of MPC for such systems, is the cost, time, and effort associated with learning first-principles based dynamical models of the underlying physical system. Data predictive control (DPC) is an alternative approach for implementing finite receding horizon control for large scale systems using control-oriented data-driven models. The data predictive controller enables is most useful for systems where there is a significant cost to learning white/grey box models of the systems dynamics. Applications include, Demand Response for buildings (energy systems) and inferential sensing and composition control (industrial control).