Researchers at the University of Pittsburgh, University of Massachusetts Amherst, and Microsoft Research India have developed a system — WattScale — that leverages AI to pick out the least energy-efficient buildings from a city- or region-level population. In a preprint study, they used it to show that half of the buildings in a 10,000-building data set were inefficient, in large part due to poor construction.
Buildings — including offices, homes, and stores — use 40% of U.S.’ energy and 70% of its electricity, according to the Alliance to Save Energy. They also emit over a third of the nation’s greenhouse gases, which is more than any other sector of the economy. Solving for the disparity requires identifying buildings that are the least efficient and thus have the greatest need for improvements, but approaches that rely on the age of a building or its total energy bill don’t work well; greater energy usage doesn’t necessarily point to inefficiencies.
WattScale aims to address this with (1) a Bayesian modeling technique that captures variable distributions governing the energy usage of a building and (2) a fault analysis algorithm that makes use of these distributions to report probable causes of inefficiency. The open source tool offers two modes — individual and region-based — that flag inefficient buildings either by comparing their distributions with similar homes in a city or with distributions learned for the entire population in a region with comparable weather conditions.
In experiments, the researchers applied WattScale to data sets from three cities — Austin, an unnamed small city in New England, and Boulder, Colorado — and tapped the Building Performance Database, the largest publicly available data set of energy-related commercial and residential information, for per-region building distributions. The Austin and Boulder data sets contained a breakdown at an appliance level, while the New England corpora tracked energy usage from electricity and gas meters and real estate information including buildings’ sizes, the number of rooms, bedrooms, and property type.
The coauthors report that WattScale identified faults in nearly 95% of cases, finding inefficiencies in more than half of the 10,107 residential buildings from the data sets. It pinpointed building envelope — the physical barrier between the conditioned and unconditioned environment of a building — as a top cause of inefficiency, accounting for around 41% of the homes. Heating and cooling system faults were the next major contributors, affecting 23.73%, and 0.51% of the buildings spotlighted by WattScale, respectively.
In the New England city, which has more winter days than the other two regions, WattScale found that 18.06% of the homes had issues with either high heating or low cooling setpoint temperature and that the majority were built before 1945 and co-located. It also found that, excepting HVAC appliance-related faults, mixed-use developments had the highest share of inefficient buildings followed by multi-family and single-family property types.
The researchers envision utility companies and policymakers using WattScale to identify inefficient buildings within a cohort and assessing the impact of various subsidies on energy consumption. Even homeowners can benefit from the tool, the researchers assert, because of its ability to compare a home’s efficiency to any region when combined with geolocation data.
“Since WattScale uses coarse-grained daily and annual energy consumption to create distribution for a building and region, respectively, we see enormous potential in applying our data-driven approach for various energy-efficiency related analytics,” the researchers wrote. “We intend to deliver individual inefficiency report generated from WattScale to the different homeowners. These nudges can be used to motivate and incentivize homeowners towards energy efficiency measures.”
In the future, the team plans to augment WattScale with satellite data and building occupancy patterns. In addition, they hope to investigate how the tool might be used to track energy savings throughout the day and seasons and quantify the effectiveness of retrofits in homes.
The paper detailing WattScale — which builds upon the researchers’ earlier inefficiency-detecting system, WattHome — follows a proposed AI model that uses smartphone location data to predict power grid usage. In a preprint study published in June, Microsoft and the University of Washington researchers detailed a system that uses smartphone location data to forecast electrical load. They claimed their architecture, which takes into account data from geographical regions both within the U.S. and Europe, can outperform conventional forecasting methods by more than three times.