IoT and AI team up for energy management and fault prediction.
Usually, NASA is the developer of new technology; this time it’s a beta tester.
NASA’s Ames Research Center contains one of the most energy-efficient federal buildings, a LEED Platinum facility called Sustainability Base. It was designed to have a minimal environmental footprint, including ample daylighting throughout the building, passive cooling, water conservation technology, renewable energy generation, and an intelligent building automation system (BAS).
Although the BAS can monitor and control central devices like lights, pumps, and HVAC units, it can’t tell if those systems are working at peak efficiency, nor can it detect electrical anomalies that could indicate impending unit failure. Verdigris Technologies was developing an IoT energy monitoring system that uses clamp-on current sensors to detect individual devices turning on and off, and a sophisticated signal processing algorithm that identifies the device and analyzes its electrical “signature” to detect potential faults. Verdigris had several prototypes that were tested in-house but wanted to see how they performed in a more rigorous setting with a wider variety of equipment. Through a nonreimbursable Space Act Agreement, Verdigris installed its equipment at no cost to NASA, allowing NASA to optimize Sustainability Base while Verdigris put its product line through the paces.
Smart Energy Analysis
The Verdigris system is installed at the service panel, with a current sensor clamped onto each circuit:
So how does the system know which devices are on or off at any given time? It turns out that most electronic units have an electrical signature that causes tiny fluctuations in the power line. Sampling the power line thousands of times each second, the current sensor uses an artificial intelligence algorithm to determine the device’s “fingerprint,” enabling it to identify which unit just powered up or down. (That’s some pretty cool signal processing, if you ask me.)
Upon installation, the system has to be taught which device corresponds to each signal; from there, it can track exactly which units are turning on and off at any given moment. That information is transmitted wirelessly to the BAS for energy analysis, giving the facility manager a clearer picture of what’s using electricity, and when.
Fault Prediction
When a device like a motor turns on, it causes a fluctuation in the power line. Once the Verdigris system recognizes that signature, it’s able to detect slight variations which could indicate impending failure, as shown here:
Here on the home planet, this kind of fault prediction can reduce downtime in manufacturing facilities. In off-world applications, the technology can warn mission controllers when a probe is about to fail. “When you have advance notice, you may be able to isolate a device, so it doesn’t bring down other devices. You may be able to make remote changes that prevent the failure,” says Rosalind Grymes, Deputy Director of NASA Ames Partnerships Directorate. And if failure is imminent and not preventable, Dr. Grymes says, “You can now optimize the use of that device from now until the predicted failure, so you can get the best science out of the remainder of its life.”
Here’s Matt Bereman, head of business development at Verdigris, describing what Verdigris does:
Although Verdigris is currently aimed at the commercial/industrial sector, demand response will eventually affect all customers, so a variation of this technology could soon be in the residential market as well.
Images courtesy of NASA. Video courtesy of Verdigris.
_______________________________________________________________