MPC Design for an Auditorium Building Using Data Driven Modeling

This paper describes the design of a model predictive controller (MPC) for preheating/-cooling of an auditorium building with a central air handling unit (AHU). The data-based approach uses system identification to estimate a low order nonlinear MPC model. The data used for identification is generated by simulation of a high order process model that was developed in Modelica using the open source Buildings library. This model is further used for validation of the predictive controller within Matlab/Simulink using Functional Mock-up Interface (FMI). Advantages of MPC compared to a conventional control strategy are simulatively proven using co-simulation, where an overall cost reduction of approx. 25% was achieved.