Norway – could smart drones monitor the power grid?

Long range drones equipped with artificial intelligence could provide real time monitoring of power grids.
Published: Wed 23 Aug 2017

Norway’s power grid has a length corresponding to approximately ten times the circumference of the earth or about 400,000km.

Much of this is in remote, hard to access areas with the country dominated by high mountains, deep fjords and hundreds of islands and a variable and harsh climate.

To monitor the power grid, the state owned transmission system operator (TSO) Statnett uses aerial photography from helicopters. These are costly to fly and the process is time consuming, requiring post flight manual analysis of the photographs.

But a plan being pursued by Norwegian University of Science and Technology (NTNU) students could see these being replaced by drones equipped with artificial intelligence based monitoring systems, which can collect and analyse data without human intervention.

The use of drones for power system monitoring is not new – drones have been gaining growing interest in the energy sector in the past two decades and especially in the last few years, with their shrinking size and wider choice and availability. But this initiative adds some new layers of innovation to the technology.  

Long range drones

The plan, which is now under detailed development, was the winner in a student competition run by Statnett to come up with new ways to use ICT to bring improvements to the planning and operations of the power grid.

Stated the competition announcement: “Statnett faces a large portfolio of investments, reinvestments, operations and maintenance over the next ten years. By using ICT to a greater extent, we believe we can make this safer and more effective than today.”

Two of the two students are Jørgen Veiby and Per Magnus Veierland, who have also turned entrepreneurs to advance the technologies through their respective startup companies, Sevendof and Veierland Intelligence.

The plan essentially involves two aspects, Veierland told Engerati in an interview.

One, which is the business of Sevendof, is the development of a long range industrial drone. The concept, named StormPetrel, combines a vertical take-off and landing architecture with a custom hybrid gas-lithium polymer battery engine.

“The lift gained from having wings in forward-flight mode increases range efficiency, while the hybrid engine allows for carrying more energy,” Veierland says.

He adds that estimations show that a pure electric version of the drone would have a range of around 130km, while the hybrid engine increases the range up to 500km.

Over such a range, navigation of the drone would be autonomous, based on specified geographical coordinates with inertial measurement and global positioning systems.

Sevendof has been awarded two years of funding from the Research Council of Norway to develop the StormPetrel to the point where it is able to gather inspection quality images in the field.

Currently, a full-scale version has been constructed but flight development and testing is under way with a 50% scale model, Veierland notes.

Grid artificial intelligence

The second aspect is developing the artificial intelligence system that would be carried on the drones, which is Veierland’s focus.

He explains that he is developing the basis for this application utilising historical inspection data from the Norwegian grid company TrønderEnergi Nett, and using machine learning to detect faults in the images.

“The idea is to capture the same type of images from drones that are captured from the helicopters and to analyse them automatically in real-time with an onboard computer,” he says.

“This allows for notifications to be sent back to the user immediately and also allows for interaction with the drone autopilot to ensure that sufficient data is gathered.”

In addition, images would be uploaded to the cloud both during and after the flight for further analysis and for further improving the machine learning models.

Initially the focus is on visible spectrum imaging but the aim over time is to investigate how other types of sensors, such as infrared cameras or LIDARs, may help in fault discovery, Veierland adds.

Currently, data is available covering approximately 60 fault categories although just a few of these are responsible for the majority of fault cases. These include rotten wooden traverses, missing or damaged top hats and woodpecker holes.

Drone monitoring

Compared to using helicopters, drones are obviously able to significantly lower the costs and risks involved in gathering power grid data.

As a “fun exercise,” Veierland says he and Veiby have calculated that to cover Norway, eight drones appropriately stationed throughout the country could “comfortably cover the mainland,” assuming that each drone returns to its home station.

They also believe that rather than a utility owning their own drones, an ‘as a service’ model would be more cost effective. The drones could then perform work for several companies, which would increase their utilisation and thus lower the cost of the service.

“In this way, the drones could become a key data source for power grid companies. They could use their existing systems to schedule a drone flight using from the nearest and most readily available drone. And with tight integration with their existing systems the data can be presented in the same way as with current manual analysis.”

Related Webinar