Centralized & Decentralized Temperature Generalized Predictive Control of a Passive-HVAC Process

In this work, an application of conventional generalized predictive control (GPC) to a new HVAC (Heating, Ventilation, Air Conditioning) design is presented. As the process is composed of three nonlinear subsystems, two approaches for temperature control are compared; the centralized one, which solves the global control problem as a full MISO (Multi-Input Single-Output) problem, and the decentralized approach, which decomposes the global control problem into manageable subproblems. In this studied case, the system nonlinearities are seen as additive parametric uncertainties affecting the output, which appears as load disturbances to be rejected by the GPC regulator. For this purpose, both a nominal MISO discrete-time linear model of the global system and two nominal SISO discrete-time linear submodels for the principal local-systems are identified. The performances of the proposed control architectures are experimentally evaluated for a wide range of operating conditions.