Chiller plants are commonly used to provide cooling water to data centers. While chiller plants are designed for the maximum projected heat load, majority of data centers produce a fraction of the design load. Chiller plants that operate at partial load may not perform efficiently. Evaluating the performance of a working chiller plant is challenging because of the limited data available at the site. Moreover, it is not possible to know how the system will perform after making an adjustment to improve the efficiency. In this paper, we illustrate the use of a scientific approach based on the Flow Network Modeling (FNM) technique for improving the operating efficiency of a real-life data center in Rochester, NY

A computer model of the Chilled Water System (CWS) was created using the FNM technique. The flow and thermal characteristics of the components of the CWS were taken from hydraulic handbooks and manufacturer data. The computer model reported the water flow rate, pressure, and temperature throughout the entire CWS. These values were validated by comparing them with field measurements at limited locations.

The computer model showed that the CWS was not working efficiently for several reasons including recirculation of a large portion of chilled water inside the chiller plant, flow imbalance in parts of the system, and low chilled water temperature. Various alterations to the CWS were tested using the computer model. Modifications such as reducing the speed of the pumps and increasing the chiller supply temperature were proposed. These modifications were implemented and the energy consumption of the CWS was monitored over a few months. The proposed modifications resulted in energy reduction of 33% to 50%, which reduced the annual cooling cost by approximately $60,000.