Recent rapid growth of social media and cloud storage has resulted in significant construction and development of data centers in the United States. With the introduction of the 5G network and increased interest in cryptocurrencies, the need for data storage will continue to increase, and the data center market will continue to expand. Sources of air emissions at data centers typically consist of diesel-fired emergency generator engines [combustion emissions, primarily nitrogen oxides (NOX), carbon monoxide (CO), and particulate matter (PM)], diesel belly tanks [volatile organic compound (VOC) emissions], and cooling towers [PM emissions]. Historically, intermittent sources such as diesel-fired emergency generator engines have been exempt from demonstrating compliance with short-term air quality standards. The US EPA has also previously issued guidance and rationale for effectively exempting intermittent sources from worst-case modeling assumptions. However, with the growing prevalence of data centers consisting of a large number of emergency generators, some states are requiring 1-hour modeling for nitrogen dioxide (NO2) using maximum hourly emission rates and worst-case meteorological conditions for these sources. Given the short amount of time that these generators operate on an annual basis, resulting in a low probability that the emergency generators will operate during the worst-case meteorological hour in any given year, this conservative approach results in unrealistically high concentrations. One option for these facilities may be to employ a Monte Carlo statistical postprocessing step to randomly sample and summarize thousands of 1-hour NO2 modeled impacts, resulting in modeled concentrations that more accurately reflect the intermittent nature of the diesel-fired emergency generators. This presentation will outline the current regulatory environment as it relates to short-term air dispersion modeling at hyperscale data centers. In addition, this presentation will discuss the benefits and challenges of the Monte Carlo approach for modeling 1-hour NO2 at hyperscale data centers.
.jpg)
