Elliott, Matthew Stuart (2013-07). Distributed Control of HVAC&R Networks. Doctoral Dissertation. Thesis uri icon


  • Heating, ventilation, air conditioning, and refrigeration (HVAC&R) systems are a major component of worldwide energy consumption, and frequently consist of complex networks of interconnected components. The ubiquitous nature of these systems suggests that improvements in their energy efficiency characteristics can have significant impact on global energy consumption. The complexity of the systems, however, means that decentralized control schemes will not always suffice to balance competing goals of energy efficiency and occupant comfort and safety. This dissertation proposes control solutions for three facets of this problem. The first is a cascaded control architecture for actuators, such as electronic expansion valves, that provides excellent disturbance rejection and setpoint tracking characteristics, as well as partial nonlinearity compensation without a compensation model. The second solution is a hierarchical control architecture for multiple-evaporator vapor compression systems that uses model predictive control (MPC) at both the supervisory and component levels. The controllers leverage the characteristics of MPC to balance energy efficiency with occupant comfort. Since the local controllers are decentralized, the architecture retains a degree of modularity--changing one component does not require changing all controllers. The final contribution is a new distributed optimization algorithm that is rooted in distributed MPC and is especially motivated by HVAC&R systems. This algorithm allows local level optimizers to iterate to a centralized solution. The optimizers have no knowledge of any plant other than the plant they are associated with, and only need to communicate with their immediate neighbors. The efficacy of the algorithm is displayed with two sets of examples. One example is simulation based, wherein a building is modeled in the EnergyPlus software suite. The other is an experimental example. In this example, the algorithm is applied to a multiple evaporator vapor compression system. In both cases the design method is discussed, and the ability of the algorithm to reduce energy consumption when properly applied is demonstrated.

publication date

  • August 2013