This paper “Better sustainability assessment of green buildings with high-frequency data” is part of a bigger project funded by the U.S. National Science Foundation Faculty Early Career Development (CAREER) Program to improve energy efficiency evaluation via big data analytics. This NSF project started in 2017.  The overarching goals of the bigger NSF project include providing more precise estimates of energy savings from building energy efficiency as well as the associated environmental, social, and economic benefits, using the increasingly penetrated smart meter data and advanced statistical and econometric methods.

In the summer of 2013, the principal investigator (Yueming Qiu)’s was working with a U.S. electric utility company to evaluate the impact of one their dynamic pricing programs on consumer energy consumption patterns (See Qiu, Kirkeide, and Wang, 2018). Knowing the hourly energy consumption is critical for utilities to operate such dynamic pricing programs which charge different electricity prices at different hours of day. At that time, the utility company was just starting to deploy large scale Advanced Metering Infrastructure (AMI) or what we normally call smart meters in its service territory. When Qiu was first working on the dynamic pricing project, the utility company needed to send out technicians to the field to manually install and turn on a specific type of meters that enabled the utility to know the electricity consumption of a household at any given hour of the day. Such manual process made the evaluation studies highly limited in terms of the scale of the study sample as well as the types of feasible study designs.  After the utility started to deploy AMI, these smart meter data fundamentally revolutionized load research methods and enabled timely and more precise program evaluations.

(Source of photo: https://blogs.sap.com/2017/08/02/meter-data-management-with-sap-your-way/)

It was at that time Qiu started to think about what other research benefits smart meter data can bring to environmental researchers.  In 2014, the Clean Power Plan was first proposed under the Obama administration. At that time there was lots of discussion about how the mix of fuel for electricity generation might change towards cleaner fuels. With rapidly evolving electricity grid in the U.S., it is important to distinguish between marginal environmental pollutant emission factors versus average emission factors, because at different hours of day the marginal power plant switched on is different. Meanwhile, any demand side interventions such as distributed solar energy, energy efficiency and demand response programs can have different environmental impact due to the varying marginal emissions factor by hour of day. However, no prior research was able to provide precise environmental sustainability impact of these demand side interventions because smart meter high frequency electricity demand data was not largely available before the 2010s.

With her experiences working with the utility’s smart meter data, Qiu started to connect the demand side intervention evaluation at high frequency level with the supply side marginal emission factors as well as other supply side economic factors such as capacity value and wholesale electricity prices. For example, any energy savings that happen when coal as the marginal fuel will save more on environmental pollutants than savings when natural gas was the marginal fuel. Qiu and Kahn then started their research discussion about whether green buildings can deliver the promised environmental gain. Combining these discussions she had with Kahn and researchers at the utility company, Qiu started to write the NSF proposal and eventually got funded and started working on these papers. This paper is a great example of how new research ideas can originate from the collaboration between academic researchers and practitioners in industry.

(Photo credit: Xu Tan)

Before the widespread deployment of smart meters, electric utilities collected data at the customer account/year/month level.  With twelve monthly observations per account, researchers faced a challenge of teasing out "cause and effect" at high frequency intervals.  For example, suppose that a researcher seeks to measure how heat waves affect household electricity consumption.  The researcher observes that a household consumes an aggregate total of 900 kWH of power in July 2017.  During this month, the daily temperature was 70 degrees for the first two weeks and the daily temperature was 90 degrees for the second two weeks.  In this case, the average temperature was 80 degrees that month.  A researcher will use such aggregate data to measure the impact of summer heat on electricity consumption but the actual household was never exposed to 80 degree temperature that month. This example highlights how temporal aggregation of the climate data impinges on the ability of the researcher to study the key question. As the volatility of dynamic prices increase, the importance of evaluating energy efficiency programs using high frequency electricity data will only increase.  

Contributors to this blog post:  Yueming Qiu and Matthew Kahn

Read the full paper, " Better sustainability assessment of green buildings with high-frequency data," in Nature Sustainability at https://doi.org/10.1038/s41893-018-0169-y 

Reference cited:

Qiu, Y., Kirkeide, L., & Wang, Y. D. (2018). Effects of Voluntary Time-of-Use Pricing on Summer Electricity Usage of Business Customers. Environmental and Resource Economics, 69(2), 417-440.