A humble power pole in Auckland’s Maraetai was one of the first to be examined by a new AI that’s learning how to spot wear and tear on our network, helping us prioritise what to fix first.
This is the beginning of a revolution in how we look after the network. How did we get here? Several years ago we started informal engagements around a shared vision for the future of electricity networks, then in 2021 made a formal collaboration announcement with Tapestry, a project from X, The Moonshot Factory, Alphabet’s innovation lab, to implement cutting-edge AI technology to optimize asset inspections. The first tool to be deployed from this partnership, is an innovative platform that collects aerial images from drones and helicopters and uses advanced AI and machine learning technology to assess their condition. The goal is to improve the precision of network maintenance, helping keep the power on for customers.
Rather than settling for the traditional approach of sending people out on foot to check every last power pole and wire, we’ve deployed GridAware to harness the power of aerial technology to get the job done more efficiently and accurately, and we’re taking it a step further by using AI to speed things up even more!
First, the aerial survey technology – from helicopters and drones – captures precision images of our equipment, as well as other data like thermal scans. With lots of high-quality data to hand, our network inspectors review this back at their desks and mark up images when they spot something that needs fixing. Two things happen next: we take this information and model which poles or wires need fixing first, and we use it to train an AI to spot similar things automatically as more images come in from elsewhere on the network.
Every 2.5 years we’ll record and analyse more than 1 million images to complete 170,000 inspections of Auckland’s entire overhead network, including high resolution images, thermal scans, and LiDAR.
What are the benefits?
The benefits are clear when compared to the traditional approach, because it’s simply much faster to capture images by air than on foot. But we’re aiming further than that. Our goal is to fully automate this process. As our network inspectors record their assessments on the images, machine learning algorithms are teaching the tool to automatically assess condition, prioritise maintenance and schedule jobs. This will improve the efficiency and accuracy of our programmes to look after the network by leveraging the full potential of AI.
Another key benefit is that following a damaging weather event like Cyclone Gabrielle, aerial imagery can be collected from areas where access is difficult for line patrols, and assessed quickly to speed up repairs.