Cutting-Edge Manufacturing Made More Efficient with AI Smart Meters

The Challenge

Katerra is using modern technology to optimize every aspect of building design and construction. The company’s Tracy, CA factory contains some of the most advanced manufacturing equipment in the industry; at full capacity it will be able to produce the equivalent of 12,500 multifamily units a year. The factory also recently installed a 1.117 megawatt (MW) solar array to offset a majority of its energy use.

To further measure and reduce the energy impact and carbon footprint across all their construction and manufacturing processes, Katerra turned to Verdigris.

The Approach

Verdigris installed six AI energy meters and 216 smart sensors on six electrical panels and 72 circuit breakers at Katerra’s Tracy factory. Total installation time was 8.7 hours, averaging under 1.5 hours/panel. For the same number of panels and data points, a typical submetering installation time could be up to five times higher.

Data was live on the Verdigris dashboard within 24 hours after installation, including highly granular energy data down to each phase of each circuit.

One of Verdigris’ most advanced AI features is end-use classification, enabled by very high-frequency energy data sampling (8,000 times per second). At Katerra’s factory, Verdigris identified several key end-use categories: lighting, HVAC, compressor, dust collector, large automation, plug loads, and other process equipment by analyzing the unique waveform for each load type.

To develop the AI, 21 days of energy data plus external inputs like weather data were fed into Verdigris’ algorithm. Fourteen days acted as the training set; the remaining seven days as the validation set. Of the validation set, Verdigris AI automatically “blind labeled” all the load types. 98.9% of energy was labeled accurately.

The Results

Energy Usage Anomalies
By looking at the past three months of data for every circuit in a building, Verdigris constructs an expected region of energy usage with 95% confidence. Customers are alerted if energy usage falls outside of this expected region.

This feature gave Katerra an immediate payback when Verdigris detected abnormally high demand at night. The anomaly was traced back to dust collectors running overnight. The cost for this was estimated to be a couple hundred dollars per incident. Left undetected, the energy costs for this equipment alone would be significantly higher. With the Verdigris intelligent alert, the facility team was able to take quick action to correct wasted energy usage.



Phase imbalance
For motor loads, phase imbalance can cause reduced output horsepower, overheating, and even shutdown of the equipment. Being able to proactively identify this problem is critical to avoiding equipment downtime, financial losses and energy waste. Verdigris detected phase imbalance at several breakers for Katerra’s Tracy factory, which the team then fixed before problems arose.

Compressor short-cycling
Short cycling occurs when a mechanical system that operates in cycles turns on and off more frequently than expected. Short cycling can occur in air compressors as well as other equipment such as chillers, AC units, fans, and pumps. This rapid state-change adds unnecessary stress on the equipment, reducing its lifespan and efficiency. Short-cycling is difficult to detect among millions of pieces of equipment running different processes at a big manufacturing factory, but Verdigris’ highly granular energy data made it possible. The AI identified a potential short-cycling compressor, which, as confirmed by the facility team, uncovered a number of minor leakages.

What’s Next?

For Katerra, these energy efficiency improvements identified by Verdigris are just the beginning of a more sustainable manufacturing operation.