Dynamic Modeling and Controls of Variable Refrigerant Flow Systems

Building heating ventilation and air conditioning (HVAC) systems have significant impact on the energy consumption of residential and commercial buildings. The Variable Refrigerant Flow (VRF) systems, by distributing refrigerant instead of air flow, have emerged as an appealing class of HVAC system that features quieter operation, higher flexibility, and lower cost of installation and maintenance. However, such systems also present higher challenges for controls that can realize its achievable performance. This dissertation research proposes to investigate the dynamic simulation modeling and modelfree control strategies for energy efficient operation of VRF systems with single or multiple outdoor units (ODUs) under fluctuating and uncertain load and ambient conditions. Modelica based dynamic simulation models are developed for VRF systems of different configurations, which involve modeling of indoor units (IDUs), ODUs, and various control valves for thermal regulation. Motivated by a preliminary study on applying multi-variable extremum seeking control (ESC) for an air-source heat pump (ASHP) system, a model-free self-optimizing control strategy is investigated for efficient operation of a multi-functional VRF system under both heat pump modes and heat recovery modes. With the feedback of the total power only, the multi-variable ESC takes different combinations of manipulated inputs, including IDU superheat setpoints, compressor pressure setpoints, ODU fan speed, and ODU superheat setpoint. Input selection is carried out based on their respective impact on the total power. To realize automatic and smooth switching between different operation modes of multi-functional VRF system, the mode switching strategy is proposed. Whether to turn on or off an IDU is determined by the zone temperature and a preset hysteresis band about the temperature setpoint. Based on the thermodynamic analysis, a decision variable for determining the mode of ODU heat exchanger (HX) is proposed as the ODU-HX air-side temperature differential normalized by the dimensionless outdoor unit fan speed. For the smooth switching between two compressor pressure controllers, two bumpless transfer methods are applied. The simulation results validate the effectiveness of the proposed strategy and performance of bumpless transfer strategies. For the multi-ODU VRF systems, an integrated efficiency operation strategy is proposed to optimize the energy efficiency, which consists of three respects: i) for a given operating condition, a multi-variable ESC strategy is used to optimize the settings of manipulated inputs of operating ODUs, by use of a load-sharing valve array and feedback of the normalized total power; ii) a model-free ODU compressor staging strategy with ESC integrated control logic; and iii) a modelfree ODU heat exchanger mode switching strategy with ESC integrated control logic. For staging on additional ODU compressor, the saturated operation of compressor speed (i.e. saturated to the higher limit) is used as the indicating variable. With ESC-alike real-time optimization strategy in operation, the least efficient compressor would be driven, which will then be staged off. As for automatic mode switch of ODU heat exchangers during heat recovery operation, the saturation of IDU EEV opening is utilized as the indicating variable, and the ODU heat exchanger (or fan-coil unit) with least efficiency under ESC operation will switch its operational mode. Similar to many other HVAC systems, the ESC operation of VRF systems is subject to state and/or input constraints. In this dissertation research, the general problem of constrained ESC is studied. The penalty-functions based framework of constrained ESC is studied. The dither-demodulation design is modified for penalty-function based ESC with assumption of Wiener-Hammerstein system composition. An online penalty-weight adaptation scheme is proposed with online Hessian estimation, and its convergence analysis is conducted in the context of numerical optimization ESC (NOESC).