Ryan Pollin for Zondits, November 23, 2015. Image credit: TORLEY
The traditional method for identifying energy efficiency opportunities in a building is to hire a professional to walk through the space, look at every boiler, motor, and fan in the building, and determine just how well they are working together. In recent years, however, several groups have been working on a different approach, an approach that can provide energy study type information before anyone sets foot in the building.
It’s called remote auditing – and it’s just what it sounds like. Recent developments in machine learning– algorithms intended to recognize patterns and learn from them and make predictions–have allowed for greatly improved understanding of buildings based solely on remotely available data on a building. These algorithms are applied to interval billing data to disaggregate the loads, making inferences about where the power in that building might be going based just on the one data series. Add other information that can be gathered from satellite imagery and public data records (like building size and shape, age, maybe even rooftop equipment, window area, and shell construction), and suddenly the program starts to “understand” quite a bit about a building before engaging an engineer.
If you are skeptical thus far, that is all right. There are limitations to just how far these algorithms can go. Can a piece of software know as much about a building’s operation as a trained building auditor? Can you determine that spaces are being overcooled, that steam traps have failed, or even that a building hasn’t upgraded from their standard efficiency T-8s without having a look and listen? The answer to those questions, at least in some cases, is yes.
To be clear, at the moment the concepts described here work for buildings with highly regular usage patterns and less-complex systems. The further a building deviates from regular patterns, the less likely it is that algorithms will be able to determine what is happening. However, despite difficulties analyzing complex facilities, the technology is finding a place. Take FirstFuel, a company that won Project of the Year with the Department of Defense’s (DoD’s) Environmental Security Technology Certification Program in 2014. Tasked with auditing 25% of their facilities, the DoD was facing hundreds of surveys. When compared to ASHRAE Level 2 on-site assessments, the interval analysis methods by FirstFuel’s Remote Building Analytics platform is reported to have found 16%–37% more energy conservation measures.
While your friendly neighborhood energy auditor will not be entirely replaced by software, projects like the DoD show that the remote audit has a place. The tools we use to audit buildings have changed greatly over the years. Modeling has become easier, and all manner of metering and measuring have become mainstream. While each tool has promised to replace the trained engineer, the fact is what they really do is to enhance the engineer’s productivity.
Looking beyond just one building or even a campus of buildings: How do we rapidly improve the efficiency of our whole building stock? Relatively few buildings are surveyed by engineers every year, because the industry just doesn’t have the capacity to look at every single structure in the country. However, if auditing efforts were prioritized to buildings that fail a series of first-pass, software-based tests, then these analytic tools could augment the energy engineer’s productivity. By using automated screening across a whole population of buildings, an inexpensive first-pass audit could quickly reveal those buildings in the most need of help. By using broad-based analytical tools up front, energy engineers can be more selective about the buildings they need to visit, preselecting those with a very high likelihood of opportunities.