Cloud Sourcing Electricity Usage | Michigan Technological University News

What do energy usage in buildings and traffic congestion have in common? Crowdsourcing.

Sometimes an outside perspective is all it takes to tackle a problem in an innovative
way. And inferring electricity usage in a building without using a meter could be
as simple as correlating average occupancy at a given time.

Measuring electricity consumption can be time-consuming or require installing expensive
equipment that requires regular updating. In comparison, transportation engineers
only occasionally use sensors to measure exact numbers of vehicles on a given stretch
of road, preferring to infer route usage instead.

About the Researcher 


Chee-Wooi Ten, associate professor of electrical and computer engineering, and Kuilin Zhang, associate professor of transportation systems and computer science, both at Michigan Technological University, put their heads together to create an
interdisciplinary approach to model electricity usage using inference and correlated
occupancy information.

Additionally, using the researchers’ methods in the context of COVID-19, the quasi-online
correlation between occupancy within a specific location and electricity consumption
at home may infer a shifting load of stay-at-home individuals. In the bulk power system,
research suggests there is a significant reduction of electricity loads in the communal
spaces — office buildings and entertainment and shopping districts — in a region.

More specifically, these studies can be used to determine if societal conformity with
national recommendations to stay home to flatten the curve is occurring. As most people
own a smartphone today, aggregated spatial information connects to individuals. At
this critical time, the reliability of power delivery to individuals is vital to those
who stay and work from home. This information could be vitally important to infer
and improve the quality of life at home.

Ten and Zhang published “Establishment of Enhanced Load Modeling by Correlating with Occupancy Information” in the journal IEEE Transactions on Smart Grid together with coauthors at Michigan Tech, the Global Energy Interconnection Research
Institute North America and the School of Technology and Engineering at the University
of Washington-Tacoma.

The paper proposes a statistical approach — a regression model that correlates occupancy
within physical proximity and associated loads to generate a time-dependent model
— to establish the correlations between estimated occupancy of buildings based on
simple sensors we all carry in our pockets — cell phones.

“If there are no people involved, there is most likely no electrical load,” Ten said.
“Streetlights have a constant value. Traffic lights, too. But in shopping malls, factories
and houses, you characterize the consumption behavior based on when people are there.
Based on the number of occupants in a building, we can infer electricity consumption
and build a profile of that, so we don’t necessarily put a meter in to measure the

Grants and Funding 

National Science Foundation projects 1541000 and 1538105, State Grid Corporation Technology
project 5455HJ180018.

Ten noted that he was able to model electric usage for the Electrical Energy Resources
Center (EERC), a multistory academic building on the Michigan Tech campus, based on
class enrollments and class times. Another way to infer occupancy in buildings can
be based on cell phone locations and from devices on the Internet of Things (IoT),
much the same as how Google or the app Waze acquires real time traffic data from cell
phones to infer congestion on roadways.

About the Researcher 


“Smart and connected devices, such smartphones and connected vehicles, have been widely
used as crowdsourced sensors to collect individual trajectory data to understand human
activity and travel behavior at each location and road along the trajectory,” Zhang

The length of time a person (estimated from their smartphone or vehicle data) spends
at a given location provides occupancy data that can be used to understand load patterns
on the power grid.

Ten notes that by using statistical correlations, utility companies could stand to
save on installing meters, a significant upfront investment. For occasional meter
readings in person, companies could use temporary smart meters to check against the

“Because of cell phones, which can be cloud sourced, the way traffic congestion is
tracked has changed,” Ten said. “How you see a problem is how you understand the problem.
We are coming at this problem from an interdisciplinary angle in a way that could
be disruptive, not incremental.”

Future research includes using block-by-block occupant data (rather than by individual
household) to estimate between different power distributions, how much power will
flow through and how many occupants based on time of day.

Michigan Technological University is a public research university, home to more than
7,000 students from 54 countries. Founded in 1885, the University offers more than
120 undergraduate and graduate degree programs in science and technology, engineering,
forestry, business and economics, health professions, humanities, mathematics, and
social sciences. Our campus in Michigan’s Upper Peninsula overlooks the Keweenaw Waterway
and is just a few miles from Lake Superior.