The National Grid is testing computer-manned drones that can save millions in maintenance work.
PHOTOGRAPH: CHRIS RATCLIFFE/BLOOMBERG/GETTY IMAGES
IN MARCH, A troop of engineers gathered in an unkept green field in rural Nottinghamshire, England. They were there to test a drone piloting software that they hoped could one day be in charge of maintaining the high-voltage pylons that transmit electricity across the country. Assuming the software was working, a drone was about to inspect a pylon from a few meters away, maneuvered not by a nearby pilot but a computer in a control station hundreds of meters away.
Seconds later, the dance began. Whizzing around, the drone took 65 photos that documented the condition of the pylon’s steel arms, fittings, and conductors. After only six minutes, the drone returned to the ground to a round of applause. By the time it had landed, it had already sent the photos to be analyzed for corrosion by an AI-powered system.
“What we’re doing is sending a super high-level instruction to the drone, like ‘Go to that pylon,’ and the drone is using its own intelligence to understand where the pylon is, where the parts of the pylon are that need to be imaged, and then it organizes its own route to the data capture itself,” says Sees.ai founder John McKenna, whose company was behind the drone test.
Until now, data about the condition of electricity pylons has almost exclusively been captured manually by using ropes to climb pylons, which is dangerous, or by helicopters, which is expensive and polluting. (Helicopters also deliver poor data because they can only gather it from afar.) Manually-flown drones, on the other hand, can't be rolled out on a large scale because they're extremely slow and require a pilot and an observer to follow them.
As such, the companies responsible for these pylons have had to settle for scheduled maintenance, which is not only inefficient but unsafe. Faults in the UK power transmission network are expensive, shutting down entire regions, but in drier regions they can cause wildfires. Unlock unmanned drone flight and you can, in theory, eradicate this problem.
Other countries have been working on similar efforts: Last year, the Florida Power and Light company used automated drones manufactured by Israeli company Percepto to detect problems in the power grid after hurricanes. In Norway, utility company Agder Energi Nett announced in April 2021 that it will rely exclusively on automated drones, mostly flown by KVS Technologies, to monitor its power grid. The system the company uses is tailored to speed and scalability in that it flies a minimum of 15 meters over the top of the grid for a “broad inspection,” says the company’s COO, Jimmy Bostrøm, rather than inspecting each pylon individually. A key part of the inspection is identifying vegetation that may have fallen on the grid during strong winds and storms. Three of Sweden’s core electricity distributors have also recently signed contracts with Airpelago, another company that flies automated drones, and have committed to exclusive use of automated drones for inspection over the next two years. “There are real signs that operators are steadily moving away from helicopters,” Max Hjalmarsson, the company’s cofounder and CEO, says.
Back in England, the control station powering the drone was only a walk away, but it could have been anywhere in the world, explains McKenna, and the pilot would only need internet connectivity to issue high-level instructions and override the system if anything goes wrong. Instead of humans and helicopters, McKenna’s vision is to have armies of drones inspecting and maintaining the electricity transmission grid using preprogrammed templates. This is possible because of commonality between towers. By taking photos in a consistent, perfectly repeatable process, the company’s system can digitally reconstruct each pylon, capturing data optimal for automated processing. Sees.ai sends the data it captures to a company called Keen AI, who will use it to digitally reconstruct each pylon, identifying precisely where corrosion is developing and, possibly in the future, where it’s likely to develop.
And instead of one pilot observing a single drone, each pilot could observe several, operating like air traffic control at an airport. Because the drone understands how to position itself, it can execute the mission autonomously even if communication fails.
Sees.ai designed a drone software that works in a similar way as autonomous cars. Using information gathered from six on-board sensors—two LIDAR, three fish-eye cameras, and an IMU (Inertial Measurement Unit)—it creates its own 3D world that it then presents on a computer screen, along with a live videostream from the cameras. Instead of relying on potentially inaccurate or outdated historical data from asset design files, Google Maps, or satellite imagery, the software captures its own from scratch, and will evolve in real time throughout the drone’s mission.
McKenna says this test flight in Nottinghamshire was a step towards developing a command and control system that’s going to allow for autonomous aerial vehicles to be approved on a large scale. The trials so far include the remote inspection of Sellafield’s nuclear site, the rail infrastructure governed by Network Rail, and Vodafone’s telecommunications network. Alongside the Lancashire Fire & Rescue Service, Sees.ai has been exploring whether the system could be used to transport medical supplies, and eventually persons, to and from incidents.
This technology is pushing the limits of what drones can do in British airspace. While the uses of drones are multifarious, especially when it comes to transportation and delivery, the rules that govern their operation have made it difficult to roll them out at scale. In the US, for example, the Federal Aviation Administration (FAA) prohibits companies from flying drones beyond the visual line of sight (BVLOS). Though it has approved 230 waivers, most of them have been for academic or research purposes. The waivers that have been granted for commercial purposes have been limited on time, airspace, and often both. (In March, a report issued by the FAA recommended an overhaul of these existing regulations to enable the commercial drone industry to scale.)
“It’s like this in almost all countries,” says David Wickström, CTO of Skyqraft, a Swedish company that uses AI to analyze data acquired by drones. Some drone operators, including Zipline, a US startup, have resorted to developing its systems in Africa.
In the UK, the Civil Aviation Authority (CAA) also requires the pilot to be within the visual line of sight (VLOS) of the drone. But in 2021, the CAA granted Sees.ai explicit authority to begin operating BVLOS flights in nonsegregated airspace, up to a height of 150 feet. There are only 10 or so companies in the world that have permission at this level, McKenna says. The list also includes American Robotics, the Massachusetts-based company that in January became the first company authorized by the FAA to operate automated drones without anyone on-site to monitor them. Its system relies on an acoustic Detect-and-Avoid (DAA) technology that ensures that its drones maintain a safe distance from other aircraft.
“We’re moving into a future where these drones will fly themselves all over the countryside,” McKenna says. “But the long-term future of this software is that it will fly people around.”
With the UK’s National Grid, which operates the country’s energy supply, the relationship has been more concrete, after the organization committed funds to accelerate development of Sees.ai’s technology. The partnership’s first goal is to prove that the system can be used to better maintain the grid's 21,900 steel pylons.
The network needs constant tune-ups to stay reliable, and regular inspections are important. The National Grid boasts 99.99 percent reliability: something it wants to improve on by locating critical issues long before outages occur. In the UK’s wet climate there’s a high risk of corrosion, which is difficult to stop once it has started. Pylons need to be replaced when the rust has affected their structural integrity, so early detection saves costs in the long run.
The National Grid spends around £16 million each year painting its pylons, and it has anticipated a cost of £35 million over the next five years to replace corroded steel. Factoring in the high costs of R&D, Sees.ai’s drone system isn’t necessarily cheaper than other methods of inspection, but the National Grid anticipates that it’ll enable more frequent and timely data capture that in turn will save costs through more targeted asset replacement. If the trials are successful, the National Grid anticipates savings in excess of £1 million for UK consumers by 2031.
But until cost-effective drones are deployed at a large scale, the only option is to use helicopters. A helicopter can inspect 16 pylons every hour at a cost of £2,000 per hour, but flying a VLOS drone is not much better because it’s laborious and slow with the pilot below. On a good day, VLOS drone teams can inspect no more than 10 pylons. “It’s the human element of it that causes the problems,” Mark Simmons, National Grid’s condition monitoring manager, says.
Sees.ai is not alone in tackling this problem, but the systems that many other companies rely on use GPS and compass for positioning. The problem is that these technologies are vulnerable to failure, especially when close to steel or strong electromagnetic fields, which occur around high-voltage power lines. Relying on preexisting data can also be precarious because the world is constantly changing.
According to David Benowitz, head of research at the research platform Drone Analyst, GPS technology is also not always accurate, especially when being used to measure altitudes or in rural areas with poor satellite coverage. Because there’s always going to be that “bubble of doubt,” he says, there’s a higher risk of collisions in busy airspaces. With more vulnerability comes more risk.
The only way to roll out these technologies, then, is to limit risk in other ways, such as by flying simpler flights farther away from potential collisions. But with each limitation imposed, “the applicability and scalability of the solution reduces,” Benowitz says. If we are to replace manned helicopters, we need to develop a solution that “doesn’t have these limitations,” that can safely carry out overviews and detailed inspections of assets over the majority of the grid, not just remote sections.
For this to happen, there needs to be more reliable and robust technologies: Each operating system needs to have multiple layers of safety. “In order for us to fly close enough to the pylons to acquire the best data, we need more intelligence than GPS,” says Hjamlmarsson. But there also needs to be change among the regulatory bodies like the FAA and the CAA to create the space for these more advanced systems to be developed and properly tested so that they can ever be proven to be safe. “It’s the chicken or egg scenario,” Benowitz says. “These systems are not bleeding edge, so there’s no problem rolling them out at scale and at cost, but the regulations need to get up to date.”
Updated 6/7/2022 09:30 ET: This article has been updated to clarify that Sees.AI does not process the data from the drones and that it is sent to a third company, Keen AI.