Predicting Wildfire Damage with AI – Kaitlyn Albertoli, Buzz Solutions

This is texturong TV. Welcome to techstrong TV. I'm sustainability contributor Bonnie Schneider while climate change has led to an increase in Wildfire size and frequency.

How can these devastating events that often destroy integral power lines be better detected through improved technology. Well Buzz Solutions a Stanford california-based startups AI solution to help utilities detect and repair power line in Grid faults quickly before wildfires can start maybe the solution their platform utilizes Ai and machine vision technology to analyze millions of images of power lines and Towers from drones helicopters and aircraft this system can help detect dangerous faults and flaws as well as overgrown vegetation. And of course in the grid the infrastructure this will help you utilities identify and repair problem areas in advance of a disaster joining me now is Kaitlyn Albertoli Lee the CEO of Buzz Solutions.

Caitlin thanks so much for joining us. Thanks so much for having me today excited to be here to share more great. Well, I'd love to dive right into the conversation and learn more about how you're using Ai and Predictive Analytics and this important Mission really to get people more aware of the wildfires before they start and and take those precautions to prevent the damage.

Yes. Absolutely. So at Buzz Solutions just a bit about our story.

We launched in 2017, which at that time was before some of these major wildfires that have brought California even had even occurred but our solution we take all of the data that's collected from drones helicopters fixed when aircrafts that are part of these mandated inspections of power lines and we provide a much faster way to analyze that data so that utility alignment field technicians and workers can have the answers they need much faster so that they can conduct the important maintenance before a wildfire starts or before a power outage can occur. That is my background is as a meteorologist and I just think that is such a brilliant solution to get ahead of it because these fires are becoming more frequent due to climate change. They're affecting more areas and spreading faster than ever you of course, you know that you're in California.

Did you have a personal connection to this being there and experiencing this rapid growth that we've seen of wildfires? Absolutely. I mean, I grew up in Southern California and as I was growing up in Southern California, I had several friends throughout my childhood who work evacuated due to wildfires.

It was I'm incredibly devastating to see the loss that wildfires have continued to cause throughout the state. But then even my during my time at Stanford, we were rocked with several unfortunate wildfires as well and they were sparked by a variety of different issues including some of them being from you know, grid sparked wildfires and it's it was devastating. I mean one instance particular the skies love entirely orange, and she couldn't even be outside or see outside.

How bad the smoke was with the visibility? So as we see the climatic effects continue to have an even greater impact on the weather. We're having in California.

We you know Wildfire season isn't just in those late summer and early fall months like it once used to be we're seeing wildfires that are happening, you know throughout the year. We were even impacted with one here in Southern California last May and you would never have thought it was so close to the ocean in an area that you know does get some rain but still it's it can happen at any time and so we just have to be more Vigilant especially with our critical infrastructure, which is aging it's now more important than ever to take serious serious notice of some of those issues. That's a great point.

But I'm really wondering how you made the connection to focus on AI and drone technology for power utility companies. It's this is an extremely Innovative concept. So what led you to it?

Absolutely. So at that time I will be launched us. We actually interviewed 35 different major power utilities across North America and they all had the exact same story that the grid was aging.

They were mandated to do many of these more frequent inspections. And as a part of those inspections, they were collecting massive amounts of data. So they were going from collecting hundreds of images.

They used to have lime and walking underneath the lines looking up with binoculars or they were collecting imagery that was from high zoomed out helicopters, but now with the technology such as drones and high Zoom high resolution cameras on helicopters, they are collecting tens of thousands and millions of images on an annual basis. But as they're collecting all of this data the means of analyzing that data was still entirely manual. They had people looking through every single image seeing any sort of anomaly that was happening in the image.

And that was causing a massive time lag. So with some of the utilities we were speaking with it was a four six eight-month time lag from the time an image was collected before they had the images fully analyzed in a prioritized path to maintenance. Of course in that time.

The data becomes outdated. It's no longer actionable and they have to send out a field crew to go reinspect that that particular poll or Tower to see how severe that issue is, of course lending to a even greater risk of an outage some sort of failure or you know, an extreme cases a wildfire. And so we said, you know, what if we can plug in and really just focus on that key issue to help them streamline that process of analyzing the data and then give them a prioritized path to maintenance.

We identify hotspot areas which pose the most risk on the grid and we can also direct the right teams the right place at the right time to reduce reduce risk save time and also save massive costs as a part of the Action process. Can you walk us through the process of how it works? Our audience are a lot of people that work in the tech industry and they might be interested in how your the AI and the Predictive Analytics actually worked with predicting when you're inspecting these power lines, maybe just get into the technical process of it.

Absolutely. So we and build to these proprietary machine learning computer vision algorithms, which allow us to analyze all of the visual data. So as I was mentioning, we're primarily working with visual imagery.

And so we trained our algorithms across a variety of different types of data collection with the variety of various different backgrounds geographies of the images and also various angles as well of how the data was collected and we also had to spend a tremendous amount of time collecting a lot of faulty data so data that had the the issues existing within the image and some of these component failures don't happen with high frequency. So it was pretty difficult for to actually collect enough data to train a successful algorithm. So then we train our algorithms across a variety of different anomalies our computer vision algorithms right now are detecting 28 different types of as we call them failure modes or different types of anomalies and that list continues to grow.

We're with our AI we're detecting Various types of components. So we're doing things like insulator damages conductor damages any sort of structural damages like, you know rusting or pull rot some of the major components that you see fail along the grid infrastructure and that's where the artificial intelligence has become quite valuable because at scale the AI continues to learn and improve and become even more accurate and so it's been it's been a great Learning System that you know continues to bring even greater value at scale. We mentioned a little bit about the process and and developing have what were some of the challenges that you faced in developing this technology.

The data is incredibly proprietary in the space, of course, you know, we're dealing with critical infrastructure. So this data has to be handled with great care and being able to get access to this data has been a very challenging process particularly in the early days. And so we were fortunate to work with several entities.

We worked with several utilities and then we also worked with research organizations like every which is the Electric Power Research. Institute to gain access to these proprietary data sets. So that was definitely one of the biggest challenges the second one I would say is as I was mentioning before at some of these failure modes don't occur with high frequency for different utilities, but they can be you know, very severe issues that utility must detect such as you know arrested sea hook or any sort of smaller component like a cotter pin which is missing or loose.

You know, those are very fine minute types of anomalies. But in order to train a successful algorithm, we have to have the that faulty data to be able to successfully train it and so it took us a good deal of time to be able to collect that very data. So I would say that those are are some of the Two of the biggest challenges, but the third is you know data is not standardized in the way.

It's collected in this space as well. And so you have data that's being collected from various types of drones helicopters camera sensors angles backgrounds, and so being able to train an algorithm successfully performed across all of these various scenarios was really important to the industry. And so that took us a great deal of time but it took us about a year and a half to overcome some of those challenges for us to successfully, you know, build up technology and what has been the feedback that you've gotten from these these power companies maybe even some of the people that are working there that can can actually use the information that you're providing.

What kind of feedback have you ever said that we're able to provide them huge Time Savings and make their jobs much easier and more efficient, you know, one of the values of bringing drones into the equation is it helps keep those those lime and field technicians on the ground in a much safer. It's much safer for them to actually perform that inspection work and so being able to Analyze that data seamlessly so that they can identify critical issues before before they become a sparking issue or before they're at their most severe point also reduces the risk in their jobs as well. So the fact that we're able to provide them huge efficiency and Time Savings, we're able to provide the utility on the business side cost savings as well as a part of their inspections.

And then lastly the big risk reduction has been one that's been a value for the the boots on the ground teams as well as the business units and it's been a win for for those across the board. That's great. And do you find that the interest in this concept that you're doing is is people want to know well, wait a minute we can do this with power utility companies that are already gathering this data.

There can be other applications. So I'm just wondering if you're what you're looking towards what the future where you can even apply this specific proprietary technology or develop something similar for another industry. Yes.

It's a great question. You know, our vision is to as we say Safeguard the world's energy infrastructure. Our goal is to work far beyond just the the power infrastructure.

Although as we've you know seen that's been the the biggest burning need is the utility space first off but we see our technology applicable in many other parts of infrastructure inspections. So we're actively working on some of those now and we're excited to be bringing those to the market here and be able to share a bit more about it here pretty soon. That's great.

And a lot of entrepreneurs are looking into some of these hot areas like Ai and drone technology to develop some big idea and you you did it and you created a successful concept. Do you have any advice for anyone that's watching that's thinking about well, I'd like to do something like that as well. Now what we found is that we listened closely to the customers needs and really what they were saying their biggest pain points were and I think that's been so valuable for us as we continue to grow and we you know, innovate and continue to to progress our company our vision and as we're we're growing our offerings in the space keeping closely the mind on what the customer needs at.

The Forefront has been something that's that's worked well for us and I always say, you know, you you want to tackle tackle a new issue or tackle a space get out there and start talking to people because they're the people that are boots on the ground doing this job day in and day out. They're going to be the ones that can provide some really good feedback on what their biggest challenges are and what they To see which would improve their day-to-day operations and you know, once you hear that story time and time again for us, it became really clear that we needed to do something to really, you know, we had a solution that we could bring to this market. And so it's if you find that other needs somewhere else.

You know, I take a chance of it and really see what can come of it because it can be really exciting Journey. Absolutely a Caitlin Albert Tony CEO of Buzz Solutions. Thank you so much for joining us on Tech strung TV with your really incredible story and how you're really on the front lines fighting these problems that we're seeing with wildfires as a result of climate change and coming really up with really innovative solutions.

Thank you so much for joining us so much for having me. Really appreciate it. All right.

Well stay with us. We have a lot more on Tech strong TV coming up.