A methodology for mapping the performance of variable-speed residential cooling equipment using load-based testing. (December 2021)

• Record Type: Journal Article
• Authors: Cheng, Li; Braun, James E.; Horton, W. Travis
• Is Part Of: International journal of refrigeration. Volume 132(2021)
• Page Start: 133
• Page End: 144
• Publication Date: 2021-12
• Subjects: Load-based testing -- Variable-speed residential cooling equipment -- Performance mapping -- Optimal training dataset; Essais basés sur la charge -- Équipements de refroidissement résidentiels à vitesse variable -- Cartographie des performances -- Ensemble de données d'apprentissage optimal; Refrigeration and refrigerating machinery -- Periodicals; 621.56
• DOI: 10.1016/j.ijrefrig.2021.09.001
• Languages: English
• ISSNs: 0140-7007

Abstract

Abstract: A load-based testing methodology has recently been developed for laboratory testing of residential cooling equipment with its integrated controls that employs a model to emulate building dynamics and their interaction with equipment controls. This approach is part of a new performance rating standard that utilizes a bin method to estimate a seasonal energy efficiency (CSA EXP-07, 2019; Cheng et al., 2021). However, a "holy grail" for future equipment performance rating is to be able to map equipment performance using load-based testing results and then implement the map as a "model" in building energy simulations to generate seasonal performance ratings that are specific to various building and climate types. With this goal in mind, this paper presents a performance mapping methodology that incorporates a "gray-box" model structure that uses inputs that are consistent with building energy simulation programs (sensible cooling load and equipment inlet conditions) and outputs key performance metrics (total sensible and latent cooling rates, power consumption, and COP). A strategy and a test matrix for training the model with a relatively small number of testing points were established by developing a successive optimization approach to identify the best 12 test conditions to apply for training from a larger set of available data. In order to develop, demonstrate, and validate the modeling and training approach, load-based laboratory testing was set up within psychrometric chambers and testing was performed to generate 39 quasi-steady-state data points over a range of loads and boundary conditions with a test unit operating with its normal integrated controls. The best model trained using the optimal training subset of 12 data points was able to represent the equipment performance across the operating envelope within ± 10 % .