Abstract:
Liquid chillers are widely used in commercial and industrial applications to provide chilled water for airconditioning purposes and are the single biggest consumer of energy in commercial buildings. However a review of the literature revealed little experimental or theoretical research dedicated to the in-situ performance of these machines. The experimental studies were primarily concerned with energy consumption and neglected the physics involved in chiller operation. Existing modelling studies were predominantly empirical in nature. Screw-type chillers were notable due to their absence in the literature. The objective of this research was therefore to these points by performing a detailed experimental and modelling study of an in-situ, vapour-compression liquid screw chiller. Data was collected from a Trane RTHB255S screw chiller with a rated cooling capacity of 650kW over a period of two (summer) months. Part-load conditions predominated while the operation of the chiller was never truly steady state. The performance varied significantly over its operating range with cycle efficiencies being severely compromised at low loads. In particular the performance of the heat exchangers and the efficiency of the compressor varied significantly with cooling load. Steady state and transient models were both developed based on a first-principles approach using equations of heat, mass and momentum conservation. The distinctive and unique facets of the steady state model are numerous. They include the use of complex elemental (discretised) heat exchanger models, the use of an economiser component model and the implementation of refrigerant inventory routines. The dynamic model was believed to be the first developed for and applied to a screw-type chiller and is unique in that the heat exchangers are modelled physically and not empirically, as is the case with existing formulations. The models were both calibrated and validated with independent data sets. It was found that the steady state model predicted the performance of the chiller to an accuracy of ±5% while the dynamic model predicted steady-state data to within ±5% and transient processes to within ±10%. The models were structured in such a way that they can be reapplied and extended as research in the field progresses.