Olea Edge Analytics helps conserve a precious resource with Coral

In the United States alone, household and industrial uses consume gallons of fresh water every day. Six billion gallons of that gets lost to aging infrastructure through broken or faulty meters, leaky pipes, and water main breaks. At the same time, thanks to growing populations and more frequent droughts in already dry locales, global water supplies are under greater stress than ever.

Consumer education efforts, household conservation measures such as the use of rainwater tanks, and even municipal recycling of gray water (leftover from drains) for watering plants and lawns all have roles to play in stretching water supplies. But understanding commercial water use offers greater opportunities for conservation.

One company based in Austin, Texas is helping water utilities find solutions using Coral’s AI-at-the-edge platform.

The problem with meters

“Some of the larger companies that utilities serve are very, very conscious of trying to conserve water,” Frank Kaplan, Olea Edge Analytics vice president of sales, explains. However,“If they don’t accurately know what they’re using, it’s hard to conserve.”

In your home, the only way to tell how much water you are using is with a water meter. Commercial buildings operate the same way but use large high-volume water meters that track millions of gallons of water, compared to the 400 gallons per day that might be used by a family of four. The stakes are much higher for water conservation in the commercial realm; high-volume meters take a pounding and often fail with no indication or warning, costing utilities millions of dollars and squandering a precious resource.

Olea Edge Analytics helps solve the problem with monitoring devices that put eyes, ears, and touch sensing on the meters, providing independent checks on their accuracy. To work effectively, the devices need to interpret data using onboard processing rather than relying on the cloud. Edge processing is the only solution for Olea’s use case because of the massive amounts of data required to solve the problem. We use ultra low power chipsets and rely on a combination of cameras, optical and magnet sensors working in conjunction with each other. The Coral Accelerator radically alters the time it takes to get to a result. This changes the power management profile radically and allows for sampling more, better quality of data. Olea turned to machine learning algorithms running on Coral’s USB Accelerator to dramatically enhance overall system performance.

Tracking water use with AI

In Olea’s meter health solution, a camera watches the dial on the meter, a magnetic sensor tracks the turbine rotation in the meter, and an acoustic sensor listens to water flow in the meter and pipe.

Water accuracy is calculated using compared data from the three sensors that feed an AI model. The model’s results are then compared to what the meter billing records to provide a clear financial understanding of the failure’s impact on revenue. Benchmark data, event data, and alerts are transmitted to the cloud via cellular communications.

It’s all powered by a single solar panel outside the pit, putting energy for powering the system at a premium. “Especially in a city environment,” Olea chief technology officer Stefan Grefen explains, “you may have limited recharge capability due to lack of direct sunlight.

That means the system has to be as energy-efficient as possible, ruling out sending raw data to the cloud for processing the discrepancies. “We can’t transmit the data in real-time,” Grefen says. That’s because radios for sending and receiving data draw more power than an energy-sipping processor and the data transport costs would be excessive. And that makes the Edge TPU on the Coral Accelerator essential for the system to work. Low energy edge processing is essential.

Realizing the benefits of conservation

The Meter Health Analytics Solution helps Olea save its customers water and, ultimately, bill for every drop of water delivered. In an early pilot program, Olea helped one utility in a major U.S. metro area identify and recover $1 million in revenue due to meter inaccuracy in just a few months.

The City of Atlanta got started with a limited-scale pilot project in mid-2018 and realized significant benefits in just three months. “The City of Atlanta DWM conducted a 20-meter trial with Olea Edge Analytics,” explains former deputy commissioner of Atlanta’s Department of Watershed Management Quinn Jackson-Elliott. “We found over a million dollars in recoverable revenue

Conservation measures that result from increased visibility into water use on the part of commercial customers happens on an individual basis and aren’t explicitly tracked by the utility. Even so, it’s certain that higher water bills due to more accurate measurement encourage better management of the precious resource.

Recovering revenue also provides funds for utilities to modernize their facilities. In the case of the Atlanta DWM, for example, the revenue generated in the Olea trial allowed the utility to expand the monitoring solution to many more meters. “This funded Phase One implementation of 700 meters,” Jackson-Elliott says

“We can find enough money quickly for the utility to pay for our solution and repurpose dollars to pay for upgrades,” Olea CEO Dave Mackie says. Ultimately, that will pay dividends in years to come as municipalities face aging infrastructure, such as leaking pipes that further depress dwindling reserves.

Continuous monitoring of meters also saves on operational costs. “They tend to replace the meters every ten years,” Kaplan says of Olea’s utility customers. “If we’re on the meter, we can tell them, 'The meter’s working just fine, no need to replace it.’” Not having to swap out good meters on a preemptive replacement schedule frees up yet more funds for conservation projects and upgrades.

From prototype to production

Olea has only just begun to realize the full potential of the Coral platform. At the time of this writing, the company was still in the proof-of-concept stage for the Coral-enabled version of its monitoring devices.

The prototypes rely on the Coral Dev Board for development. “We are collecting data,” said Grefen. “For character recognition, there’s a gazillion learning sets you can find on the internet.” That’s good for training the AI to read meters. But that enables just one of the monitor’s sensors, the camera, to work.

Acoustic monitoring, for the system’s microphone, is a tougher challenge. “If you look for [data sets for] noise on a pipe and how that relates to flow, there’s none.” Olea engineers have to collect their own data so they can train the model how to measure water flow based on sound. Once they get this piece of the system nailed down, though, and run the resulting software on the Coral USB Accelerator, it will provide a valuable new capability to Olea customers, giving them an additional check on the accuracy of their meters.

Grefen said he and his team are looking forward to integrating the Coral Mini PCIe Accelerator into production units. Coral radically reduces time to resolution on these systems as well as reducing the energy to process.

If you're interested in how Coral can help you build your next product please contact us.