Control and Optimization of Vapor Compression Cycles Using Recursive Least Squares Estimation

Vapor compression cycles are the primary method by which refrigeration and air-conditioning systems operate, and thus constitute a significant portion of commercial and residential building energy consumption.

This thesis presents a data-driven approach to find the optimal operating conditions of a multi-evaporator system in order to minimize the energy consumption while meeting operational requirements such as constant cooling or constant evaporator outlet temperature. The experimental system used for controller evaluation is a custom built small-scale water chiller with three evaporators; each evaporator services a separate body of water, referred to as a cooling zone. The three evaporators are connected to a single condenser and variable speed compressor, and feature variable water flow and electronic expansion valves. The control problem lies in development of a control architecture that will minimize the energy consumed by the system without prior information about the system in the form of performance maps, or complex mathematical models.

The control architecture explored in this thesis relies on the data collected by sensors alone to formulate a function for the power consumption of the system in terms of controlled variables, namely, condenser and evaporator pressures, using recursive least squares estimation. This cost function is then minimized to attain optimal set points for the pressures which are fed to local controllers.