Automating power grid inspection with AI and drones to prevent wildfires

Machine learning and computer vision methods are being deployed by industry to help manage critical infrastructure. AI-enabled monitoring and maintenance systems rely on robust AI methods that can be deployed in real-world environments.

Sparked when a powerline came into contact with a tree, the second-largest wildfire in California's history scorched almost one million acres and destroyed a small town in 2021.  As extreme weather events become more frequent in nearly 75% of the world, including Europe,  the regular inspection of power lines will become increasingly important to prevent devastating wildfires.

Currently, some five million kilometres of electricity gridlines across the European Union are inspected by manned helicopters, creating pollution and risks for humans, as well as a hefty bill for power companies of up to €150/km. To enable more frequent, safe, and cost-effective inspections, FuVeX aims to replace helicopters with specially-adapted long-range drones. The start-up’s automated solution harnesses AI to enable drones to collect the data energy companies need to ensure gridlines are in good order.

Inspector gadget
“The main issue inspecting power lines is capturing high-resolution data, as the camera has to point to the power line towers” explains Carlos Matilla, FuVeX CEO.

With ELISE funding, the company has developed a prototype drone with cameras and a computer on-board that can automatically point a gyro-stabilised camera at a power line to capture vital visual data needed for inspections, such as clear images of insulators and joints.

The system can do this by combining computer vision and machine learning. It uses neural networks to first automatically detect and locate power line towers in real-time using a low-resolution camera connected to the computer. Using this information, an algorithm calculates the position of the tower, which enables the gimbal with the high-resolution, colour camera on it to point in the right direction and take photographs.

FuVeX first validated the performance of its machine learning in a lab by simulating a drone flight. The team projected power line images in front of its drone to analyse its reaction, and to ensure the system could point its main camera at lines. More tests followed including flights beside real powerlines and now the company is honing its technology with a powerful partner.

A powerful alliance
“The main challenge to the application of these technologies is the vast amounts of data needed to train the AI models,” Matilla says. Luckily, FuVeX has access to Naturgy’s data and power lines. It is working with Spain’s third largest energy company to test its technology and is the first company to receive authorisation to perform beyond visual line of sight (long–range) flights over power lines in the country.

FuVeX inspected 1,000km of power lines with Naturgy last year and aims to survey the rest this year, totalling 12,000 kilometres. Matilla says Naturgy is already spending a third less on surveying its power lines thanks to FuVeX’s innovation. “In the next two years, we are working towards achieving savings of 50 to 75%. And that’s just the beginning” he adds. The start-up believes that when fully automated, its system could save companies 94% compared with using manned helicopters.

While the company is focused on working with its main customer, the team knows that building a larger data set will be challenging because there are various types of powerlines operated by different companies. Training the system to operate alongside these different types of infrastructure is an important part of ensuring its effectiveness in deployment. However, it is confident it will be able to collect the data needed and hone its drone further while testing its reliability. “Robustness is a huge challenge we’re facing right now,” says Matilla. “The technology has to be completely reliable.” This requires testing the system on different types of infrastructure, as well as embedding safety mechanisms to signal when the system is at risk of failure. “Modern society relies on being able to flip a switch to power its homes and businesses. But this simple action requires millions of kilometres of powerlines to function well, so FuVeX’s system has to guarantee it can always capture accurate data to ensure power lines are well maintained,” Matilla explains.

In the future, FuVeX aims to expand within Europe. The company could grow fast, with utility companies eager to transition to autonomous power line inspections, which could become the norm in as little as five years, according to Matilla.

A hot topic
Climate change could also accelerate adoption, as fears of larger and even more frequent wildfires abound. “Having a reliable power network that doesn't generate wildfires is extremely important,” he says.

Matilla believes regular inspections are a key prevention strategy. He estimates that power grids inspected by manned helicopters currently only take place every one to three years. Consequently, utility companies are spending time and money mending damage that could have been prevented and there is a risk of fires starting. “Our drone technology could keep flying in continuously over power lines ensuring their stability for the same or much less cost,” he says. The company is even working on a new type of aircraft that could make surveying power lines as cheap as €10/km."

The business case to make inspections completely autonomous is there,” says Matilla. There is also a moral case if autonomous monitoring of power lines could save habitats and lives that could otherwise be lost to wildfires. “Crewed helicopters are not enough to inspect powerlines to make sure they do not cause fires. We want to change the paradigm of inspections so that power lines, which are critical for society’s transition to green electricity, are reliable and safe." This makes autonomous monitoring one hot topic.