Tomorrow’s hydropower will start by fixing today’s dams

The way the tests are performed has also changed slightly.

Historically, it was the responsibility of the men who walked the line to check the condition of the electrical infrastructure. When lucky and on the way out, line workers use bucket trucks. However, when electrical constructions are not accessible to a mechanical elevator in the backyard, on a hillside, or otherwise, line workers must still start armed. In remote areas, helicopters are equipped with optical zoom cameras that allow you to inspect electrical lines remotely. These long-term inspections may cover more ground but cannot be replaced by a closer look.

In recent years, power plants have been using drones to capture information about their power lines and infrastructure. In addition to magnifying lenses, some are adding temperature sensors and LEDs to the drones.

Temperature sensors absorb excess heat from electrical components such as insulators, controllers and transformers. If neglected, these electrical components could spark or, worse, explode. Lidar can help with plant management, by scanning the area around the line, and by collecting information that the software will later use to create a 3D model of the environment. The model allows power system administrators to determine the exact distance of plants from power lines. This is important because tree branches can cause short circuits when they are too close to power lines, or they can catch flashes from other electrical components.

AI-based algorithms can process tens of thousands of aerial images in days and identify locations where plants violate electrical currents.Buzz Solutions

It is good news to bring any technology that allows for more frequent and better testing. And that means using state-of-the-art monitoring tools, major utilities are now drawing more than a million of their grid infrastructure and surrounding areas each year.

AI is not just good for analyzing images. It can predict the future by looking at the patterns in the data over time.

Now for the bad news. When all of this visual information is returned to utility databases, field technicians, engineers, and line technicians spend six to eight months analyzing the months. This prevents them from doing maintenance work in the field. And it’s just too long, and the data is outdated.

Now is the time to enter AI. He has begun to do so. AI and machine learning have begun to be used to identify faults and faults in power lines.

A number of energy utilities, including XL Energy and Florida Power and Lighting, are testing AI to identify problems with electrical components on high- and low-voltage power lines. These power supplies are expanding their drone testing programs (optical, thermal, and lead) to increase the amount of data collected, with the expectation that AI will make this information useful immediately.

My company, Bath Solutions, is one of the companies providing such AI equipment to the energy industry today. But we want to do more than just identify problems – we want to anticipate them. Imagine what it would have been like if a power company had known where a firefighting team was going to go in order to take precautionary measures before a fire broke out.

It’s time to ask if an AI United States Forest Service could be a modern version of the old Smoke Bear window
before They happen.

    Landscape of water, trees and hills.  At the front are electrical equipment and power lines.  The device on the left is marked Green  u201cPorcelain Insulators Good  u201d and  u201cNo Nest  u201d.  The center is surrounded by red 201cPorcelain Insulators Broken  u201d.Damage to electrical equipment due to overheating, damage or other problems can cause a fire.Buzz Solutions

We started building our system with data collected by government agencies, such as the Electricity Research Institute (EPRI) for non-profit organizations, power providers and air traffic controllers, helicopters and drones. Taken together, this data collection includes images of thousands of electrical components in power lines, insulators, controllers, connectors, hardware, poles, and towers. Includes a collection of images of damaged components such as broken insulators, damaged connectors, damaged controls, rusty hardware structures and cracked poles.

We worked with EPRI and power tools to create instructions and taxonomy to label the image data. For example, what exactly does a broken insulator or broken connector look like? What does a good insulator look like?

Then we had to combine different information, images from the air and the ground, using different types of camera sensors that work at different angles and resolutions and that work in different lighting conditions. We have increased the contrast and brightness of some images. Also, instead of considering the whole picture, we had to adjust our algorithms to focus on the needs of each image, such as an insulator. For most of these adjustments, we used machine learning algorithms that work on the artificial nerve network.

Today, our AI algorithms detect damage or errors related to insulators, connectors, dampers, poles, cross-arms and other structures and physically highlight the problem areas for maintenance. For example, we can identify what we call flashing insulators — the damage caused by excessive electrical current. It can also look at breaker conductors (also caused by overheated lines), damaged joints, damaged wooden poles and crosses and many other issues.

Unplug the green wires, surrounded by green and labeled  u201cConductor Good  u201d.  A piece of silver hanging from it holds two conical pieces on either side, burnt and surrounded by yellow, with the symbol Damp Damaged  u201d.Developing algorithms for analyzing power system equipment is necessary to determine exactly what the damaged components look like from different directions in different lighting conditions. Here, the software indicates problems with devices used to reduce wind vibration.Buzz Solutions

But one of the most important issues, especially in California, is when our AI plants become very close to high-voltage power lines and when they will grow, especially when paired with the wrong components, a dangerous combination in a fire country.

Today our system can go through tens of thousands of images and see issues in hours and days, for months to analyze manually. This is a great help for utilities trying to protect their power infrastructure.

But AI is not just good for analyzing images. It can predict the future by looking at the patterns in the data over time. AI has already done this to cite a few examples to predict weather conditions, corporate growth and disease risk.

We believe that AI will be able to use similar forecasting tools for power utilities, anticipation of disasters and areas where these disruptions could lead to wildfires. To do this, we are developing a system in collaboration with industry and utility partners.

Combining historical data with power line inspections, we are feeding the machine to the relevant region with historical weather conditions. We ask our machine learning systems to look for designs related to damaged or damaged parts, healthy bodies and overgrown plants and weather conditions around the lines and our strategies for predicting the future health of the power. Linear or electrical components and plant growth around them.

Buzz Solutions’PowerAI software analyzes power infrastructure images to identify current problems and predict the future.

Now, our algorithms can predict after six months, for example, that five insulators can be damaged in a particular area and then there is a high risk of overheating in the area of ​​the line, which can lead to fire.

We now use this error detection system with several major utilities in pilot programs: one in New York, one in New England and one in Canada. Since launching our pilots in December 2019, we have analyzed nearly 3,500 electric towers. We found that 5,500 of the 19,000 healthy electrical components could cause a faulty power outage. (We do not have information on repairs or replacements.)

Where are we going from here? To avoid these pilots and to deploy predictive AI, we need a wealth of information gathered over time and by different geographies. This requires working with multiple power plants, collaborating with control, maintenance and plant management teams. Major energy supplies in the United States have budgets and resources to maximize data on drone and aviation testing programs. But as drone prices continue to plummet, small businesses are gaining more and more information. Making devices like ours widely used requires collaboration between large and small devices as well as drone and sensor technology providers.

Fast forward to October 2025. It is not hard to imagine that the western US will experience another hot, dry, and extremely dangerous season, when a small flash of lightning could lead to a major catastrophe. People living in a fire country should be careful to avoid activities that could cause a fire. But nowadays, the threat from their power grid is minimal, as utilities, transformers, and other electrical components that were damaged during the arrival of utility workers months ago have been repaired and replaced by cutting down trees, and even cutting down trees. Access to power lines. Some asked the staff why all the activity. “Oh, our AI systems suggest that this transformer could explode next to this tree, in the fall, and we don’t want that to happen.”

Of course not!


Leave a Comment