IN-24-C045 - Novel Chiller Modeling Strategies for Low-Cost Measurement and Verification of Retrofit Measures

Commercial buildings consume a significant portion of a country’s energy resources. Without considering electrical losses, about 35% of this energy is used for providing space cooling, heating, and ventilation. A significant portion of this energy is attributed to operation of central chiller plants, with chillers alone consuming over 60% of plant electrical energy. Enhancing the chiller efficiency can play a pivotal role in improving the overall energy performance of the buildings sector. Chiller efficiency can be improved by either upgrading chiller plant equipment or enhancing chiller control systems. The critical process of measuring and verifying savings from chiller retrofits is often constrained by budget considerations of energy efficiency projects. This study proposes and investigates modeling schemes that aim to address the metering cost constraints of the process by utilizing readily available measured parameters such as outdoor air temperature and humidity to predict the year-round energy performance of chillers. Existing chiller models utilize chilled water and condenser water temperatures that are rarely trended on chiller plants. The modeling schemes are validated using chiller performance data for a chiller serving an office building located in a hot and humid climate. Two modeling strategies, a multi change-point regression approach and a multiple regression strategy, were developed and assessed. The multiple regression and seven-parameter change-point models showed average statistical performance, (R2, CV) of (0.933, 10.8%) and (0.935, 10.9%) respectively. In comparison, the Gordon-Ng model showed average performance of (0.959, 8.2%) with the same data. Despite a slightly inferior performance of the novel modeling strategies to the Gordon-Ng model, they have an advantage of using readily available outdoor parameters as regressor variables. With a lighter training effort, the seven-parameter change-point model showed the best overall performance for measurement and verification applications. Further ongoing investigations involve assessing the models’ performance across different climate zones and refining the chiller evaporator load modeling process to enhance the multiple regression strategy.