(293c) Offset-Free Model Predictive Control of a Heat Pump

Building comfort control is a complex task, consisting of not only conditioning but ventilation and transportation of a medium to/from regions requiring comfort maintenance. Such a task, is completed by various mechanical and electrical equipment usually interacting in a hierarchical manner, all in an attempt, which in certain cases, comes with an unknown degree of certainty, to simultaneously meet comfort and safety objectives. Improving such a process can be too vast when considering all interacting components at once. As a result, this work focuses on removing the uncertainty in the operation and control of the lowest level component within the hierarchical operation, which we consider as the heating, ventilating and air-conditioning (HVAC) unit.

One common HVAC unit commonly used to condition (by supplying both heating and cooling) building air is that of a heat pump. A heat pump is identical component-wise to that of a vapor compression cycle (VCC), with only one additional component being a reversing valve which allows for the refrigerant flow within the cycle to be reversed. Because of this additional valve, the heat pump can provide conditioning at any time in the year due to it's bimodal dynamics (i.e. reversing dynamics can be considered a discrete effect). As a result of the heat pump being operational in both summer and winter months, it, and it's respective control strategy, must be robust to fluctuations common in both seasonal periods, while also ensuring it's able to handle the usual fluctuation in demand common to the building occupants. The former requirements are taken care of during the system/component design and outside the scope of this work as our focus is on designing a control strategy which can account for such fluctuations while doing so in the most energy efficient manner. Our previous work has considered the problem of improving the local control approach in one mode of the system (i.e. cooling), however, by considering the bimodal nature of the heat pump, an opportunity arises to present the potential energy savings and tracking improvements associated with both it's modes of operation (i.e. cooling in summer months; heating in winter months).

Current practice fails to take into account interactions within a system and also handles constraints in a way that is separate and often damaging to control performance. Energy use is also not usually featured as an objective in the control performance and is handled instead by setting set-points in an ad-hoc way based on experience. Model predictive control (MPC) is a control methodology that has reached a certain level of maturity in several industries but has not yet been applied to buildings at any substantial scale. MPC is designed to handle many of the shortfalls of existing control mentioned earlier including interaction between variables, constraints, and the incorporation of higher level objectives such as energy use. The work reported here involves applying MPC methods to a heat pump system used as a roof-top conditioning unit. An MPC strategy is developed and tested for manipulating the electronic expansion valve and also the speed of a variable speed compressor in both winter and summer conditions. During both seasonal periods, the control objective is to regulate the air temperature being blown over the evaporator subject to constraints on the superheat. An optimization objective is formulated to minimize the energy used by the compressor while meeting control objectives and constraints. A linear model is used for the air and superheat temperature predictions with different models being active depending on the mode of operation which the heat pump is operating at. The linear MPC framework incorporates an offset-free mechanism in the form of an augmented model and state estimator in order for the controller to achieve offset-free tracking. This offset-free mechanism is able to account for variations in the air and load conditions outside the heat pump. A detailed simulation model was used to evaluate and compare the performance of the MPC strategy with that of a traditional control strategy. Results will be presented showing both improved temperature regulation and energy savings when regulating the heat pump using the MPC strategy.