Energy Capital Podcast Using AI to Strengthen the Grid with Mary Cleary
What if utilities could see infrastructure failures before they happen? What if grid operators, regulators, and energy companies could harness the power of artificial intelligence to prevent outages?
Heat waves, droughts, wildfires, hurricanes, and winter cold snaps are getting more extreme. But as their intensity ramps up, so does computing power, and particularly the capabilities of artificial intelligence.
In this episode of the Energy Capital Podcast, I spoke with Mary Cleary, VP, Marketing, Communications & Public Policy at Neara, a company pioneering AI-driven predictive modeling for utilities and grid planners and operators. Their technology is already helping major power providers anticipate and mitigate damage before storms, wildfires, and grid failures occur.
We discussed how Neara’s platform builds hyper-detailed digital models of the grid, down to individual poles, wires, and even the exact species of trees near power lines. Using LIDAR scans and AI classification, utilities can create real-time risk maps that help prioritize maintenance, allowing them to proactively prevent outages before they happen rather than reacting after the fact.
This technology is already delivering results. In Houston, Neara’s partnership with CenterPoint Energy has cut processes that once took a year and a half down to just a few hours. Rather than relying on slow, manual inspections, utilities can now simulate storm impacts and predict which infrastructure is most vulnerable before a storm makes landfall.
AI is also changing the way utilities manage wildfire risk. By analyzing environmental conditions like wind speeds, temperatures, vegetation density, and the age and vulnerability of equipment, Neara’s models help utilities pinpoint the highest-risk areas, allowing for targeted prevention measures instead of costly, broad-stroke fixes.
Beyond disaster response, we also explored the role of AI in grid modernization and demand forecasting. With Texas’ rapid energy demand growth, driven in part by AI data centers and industrial expansion, utilities need smarter ways to anticipate and manage electricity needs. Predictive modeling using artificial intelligence is giving them that capability.
One of the most fascinating parts of our conversation was how AI can help utilities make more cost-effective investments. Instead of replacing entire sections of aging infrastructure, Neara’s software allows for surgical upgrades—determining exactly which poles need reinforcement, which lines need reconductoring, and where distributed energy resources could provide the most resilience.
AI is often discussed as a massive new energy consumer, but as Mary pointed out, there’s far less conversation about how AI can be a critical tool for making the grid more efficient and reliable.
This episode provides an exciting look at the future of grid resilience, extreme weather preparedness, and how AI is changing the way we think about energy infrastructure.
As always, please like, share, and leave a five-star review wherever you listen to podcasts. Thank you for being part of the conversation!
Timestamps
00:00 – Introduction: AI, predictive modeling & the future of the grid01:58 – What is Neara? AI-driven modeling for utilities & extreme weather response04:33 – Case study: CenterPoint partnership (Houston’s 27,000 miles of lines)06:30 – How AI interprets LiDAR & optimizes storm recovery09:30 – Reliability vs. Resiliency: Measuring grid performance, customer expectations12:33 – AI disaster modeling: Hurricanes, floods & wildfire case studies15:45 – LiDAR & Dynamic Line Rating (DLR): Unlocking hidden grid capacity
18:00 - Could AI help relieve the transmission constraints that contributed to energy emergency on September 6, 2023?20:1 6 – Winter storms & gas infrastructure: Predicting failures before they happen23:00 – Optimizing energy resource placement: AI’s role in siting generation resources24:22 – AI’s next areas of development and limitations & regulatory roadblocks28:00 – Flaws in reliability metrics (SAIDI & SAIFI) & need for predictive benchmarks33:34 – Balancing cost, reliability & AI-driven efficiency36:01 – Effective policy constructs for reliability and resiliency38:24 – AI, Senate Bill 6 & large load growth: The power of micro-solutions43:21 – House Bill 2555 & Texas grid investments: Balancing cost, data & outcomes46:45 – Regulatory innovation & final thoughts on AI-driven transformation48:18 – Conclusion & call to action
Shownotes
Further Reading from the Texas Energy & Power Newsletter
* You Get What You Pay For: Let’s Pay Utilities for Performance
* Large Loads at the Lege: Grid Roundup #40
* Texas’ Energy Future: A Conversation with Jimmy Glotfelty
AI-Powered Grid Resilience & Predictive Modeling
* Neara’s Website – Explore Neara’s AI-driven platform for grid infrastructure modeling and predictive analytics.
* How Neara Uses AI for Utility Infrastructure – Detailed breakdown of Neara’s AI capabilities, including digital twins and predictive risk assessment.
* Hurricane Preparedness Utilizing a 3D Digital Twin
* LiDAR Utilization for Utilities
* CenterPoint Energy & Neara Collaboration – Announcement of their partnership to enhance grid resilience in Greater Houston following Hurricane Beryl.
* AI for Grid Optimization – The International Energy Agency's insights on AI’s growing role in modern energy systems.
* Why AI and Energy Are the New Power Couple
* Electricity Grids and Secure Energy Transitions
AI & Wildfire Risk Mitigation
* Neara’s Wildfire Risk Modeling – How AI helps utilities predict and mitigate wildfire risks.
* Neara’s Collaboration with Southern California Edison (SCE)
* CenterPoint Energy & Technosylva Collaboration – Using predictive analytics for wildfire and extreme weather preparation.
* How AI is Being Used to Fight Wildfires – A look at AI’s role in early detection and prevention.
* AI-Powered Camera Networks for California Wildfires
* Google’s Satellite Surveillance initiative (FireSat)
* USC Scientists Use AI to Predict Wildfire’s Next Move
AI and Demand Forecasting for Energy Grids
* Artificial Intelligence for Energy – U.S. Department of Energy’s take on how AI is shaping power grids.
* How AI Can Help Clean Energy Meet Growing Electricity Demand
* DOE Announces New Actions to Enhance America’s Global Leadership in Artificial Intelligence
Transcript
Mary Cleary (00:01.982)
And what we're seeing so far already is that CenterPoint has been able to compress processes that once took them a year and a half down to just a few hours.
Doug Lewin (00:13.134)
How long do you think it takes a utility to figure out which power lines will fall in the next big storm? Days or weeks? You think maybe months? In many cases, it's actually years. And with extreme weather happening more frequently and customers expecting power back in hours, not days, the old ways just aren't cutting it anymore. But what if AI could predict grid failures before they happen?
What if utilities had a digital model of every single power line, pole and tree so they could prepare for hurricanes, floods, wildfires, winter storms, whatever nature may throw at us before disaster strikes? This actually does exist. And that's actually what we're talking about in today's episode of the Energy Capital Podcast. I'm your host, Doug Loon, and my guest today is Mary Cleary, vice president at Nira, an AI-driven company revolutionizing grid reliability and resiliency. Today we'll break down AI's role in wildfire prevention, flood modeling, hurricanes and winter storms, why outdated reliability metrics are holding utilities and the industry back, how AI is changing the game for power markets and regulation as well. I'm excited about this one. It isn't just about technology. It's also about the future of the grid. Stick around because by the end, you'll know exactly why utilities that don't adapt will be left behind. As always, please like, rate, and review this podcast. Give us five stars wherever you listen, and please forward the episode on to friends and colleagues. It really does help people find us. Word of mouth is the most powerful, and we really appreciate your support of this podcast. Let's dive in. Hope you enjoy the show.
Mary Cleary, welcome to the Energy Capital Podcast.
Mary Cleary (02:00.056)
Thank you so much, Doug. Really nice to see you. Great to meet you in person in Austin last week and really excited to be here.
Doug Lewin (02:05.71)
Yeah, it was great to see you in Austin. You guys pulled together a bunch of utilities to talk about what they're doing on resiliency and what you're doing on resiliency. This company, Nira, is really, really fascinating to me. Let's just start with a little explanation for the listeners as to what it is y'all do.
Mary Cleary (02:23.95)
Sure, yeah. So, Neera is a predictive modeling software platform that helps utilities take a more proactive approach to reliability and resiliency across the entire T &E system. So essentially what we do is we build a digital model of the whole network, including all the little details. you know, think things like the widths of the individual cables, the tree canopies, the fluctuations in ground slope, all those little nooks and crannies.
Mary Cleary (02:49.888)
And utilities then use that model to simulate scenarios they need to prepare for. So everything from ice storms to hurricanes, cetera. And that model then becomes a flexible environment for utilities to essentially test drive different types of remediation solutions. So once you have, for example, simulated what a category three could look like in your network, you might know which polls are going to fail.
Mary Cleary (03:15.574)
And what utilities might have once done is a simple rip and replace, right? Which as you can imagine can get quite expensive. But in our platform, what you're able to do is see that perhaps this, you know, force rank your risks and then match up solution by risk. So perhaps only a small percentage of those polls actually need to be replaced. And maybe of the, you the ones that don't, you know, maybe it's
Mary Cleary (03:41.198)
maybe you just need to add a stay and that's sufficient in the category three. And for the uninitiated, the stay is the thing that looks like hypotenus.
Doug Lewin (03:48.142)
Yeah. So we're going to get more into this as to different kinds of extreme weather events and what this works for. And I encourage folks to, we'll put a link in the show notes to a little video so people can actually see this. But so that people could get just kind of an understanding in their heads of what this is. This is, like you said, down to the individual wire, down to the actual, tell me if I'm wrong on this, but the species of tree and exactly how big it is. I mean, you're using LIDAR to get...
Doug Lewin (04:18.218)
extremely granular and take every single line in a utility service territory and the actual conditions on the ground so that then they can kind of rank order where they're going to go. Is that right? Yeah. So for Centerpoint, you guys announced a partnership with them. Centerpoint and Nira announced a partnership last fall.
Mary Cleary (04:32.225)
Exactly.
Doug Lewin (04:42.83)
And I was watching a news story about what y'all do with them. And I believe they said, I may be getting this wrong, but I think this is right. 27,000 miles worth of lines in their service territory. And Centerpoint actually has a fairly compact utility service territory. You start thinking about in Texas, Encore that has the entire Dallas, Fort Worth area, huge swaths of West Texas.
Doug Lewin (05:07.48)
To actually send people out and do visual inspections, which is what the utility industry has done forever, takes a long, long time and you just can't possibly get everywhere. So this is a use where AI can actually look at that whole system and figure out what the needs are. Can you talk a little bit about, and I know you're very early days, but there has been some press on this and some things you guys have said about it. Give a little glimpse for folks into what's going on in Houston.
Mary Cleary (05:36.622)
Yeah, absolutely. So while I can't comment on the specific things that Centerpoint is doing necessarily, what I can say is we're extremely excited about what we've seen so far. So, you know, in the broader scope of network governance processes, there are all these things that are very manual, as you alluded to. And what we're seeing so far already is that Centerpoint has been able to compress processes that once took them a year and a half down to just a few hours.
Doug Lewin (06:06.68)
So things that took a year and a half, basically getting people out to a site, recording what went on there, probably sending somebody back, having some kind of committee meeting. You're down to a couple hours because you are able to take this lidar and look at it. And then can you talk a little bit about, and this doesn't have to be specific to CenterPorter Houston, you're working globally, but how does AI figure into this? Because you guys are also an AI and machine learning company, correct?
Mary Cleary (06:32.876)
Yeah, yeah, great question. So the primary role of AI in our software is the classification of the lidar. So have you ever seen a raw lidar scan before? Have you ever seen a raw lidar scan?
Doug Lewin (06:42.647)
say it again.
Doug Lewin (06:45.646)
I'm not sure I have.
Mary Cleary (06:47.374)
It kind of looks like, you can Google image it, but depending on the quality of the LIDAR, it kind of looks like something between a Jackson Pollock painting or TV static, like kind of poltergeist-y. So it's not super intelligible. And the traditional processes for classifying LIDAR, which is what makes it actually usable, is manual, right? So someone actually sits and looks at the images and says, ah, this is a tree. Wait a minute, that tree looks about the same height as the distribution pole. It might be the same thing.
Mary Cleary (07:15.47)
It's very error prone, it's manual, and it takes a heck of a lot of time. So we're taking in LIDAR across an entire network. And what our AI does extremely well is classify that LIDAR automatically. So tree distribution poll and not just what's what, but like looking at the, you know, the sink depth of that tree and the fact that there might be multiple tree canopies layered over a line and be able to tell if there's a line under those tree canopies.
Mary Cleary (07:43.394)
So the primary role of the AI is in making sense of the actual network data.
Doug Lewin (07:50.422)
And then rank ordering, right? So it's really like a little bit of like a triage, right? Because you just, after Hurricane Barrel, right? There were so many problems and leaders all over the state, citizens all over their service territory saying, fix this, fix this, fix this. Well, you can't go to every single line and trim every single tree in the space of a few months. You have to know where the, what is the tree that is most likely to fall on a line? Can you talk about how you guys figure that out?
Mary Cleary (08:19.63)
Yeah, yeah, 100%. So I think you just raised a really, really important point, which is that when you're doing these processes manually, you're sending teams out into the field, you're only covering, you know, an inspector and a team can cover maybe like 80 to 100 polls in a visit, depending on the complexity and such. But, and then they're answering 100 questions per poll in some cases, right? So like, you see in that given field visit, you might come across a vegetation risk that is, you know,
Mary Cleary (08:49.25)
within, it's under two feet from a line, right? But then one circuit over or even 100, maybe it's one mile down the road, there's something that's six inch clearance risk. And you don't see that otherwise. the way you, it's all relative, right? Like it's whatever the biggest risk is that day is highly unlikely to be the thing that actually requires resources and attention.
Doug Lewin (09:13.582)
It's so fascinating. mean, it's just, it really is. You know, it's so interesting in the energy world, everybody these days is talking about AI and everybody's talking about AI in terms of the power demand that it has, right? my God, AI is going to crash the grid because it needs so much power. But I feel like there isn't nearly enough discussion. I brought this up on a couple of podcasts recently. I did on the one I recorded with a former commissioner, Jimmy Glodfelt as well.
Doug Lewin (09:40.782)
There's not nearly enough discussion on the use cases for AI to actually make the grid more reliable and more resilient. You and I have talked before, I know you have some thoughts about this. Could you talk a little bit about reliability and resiliency and even how you define those terms? I you are a company that is specifically devoted to those and I feel like sometimes we don't spend enough time defining terms. And when one person says reliability,
Doug Lewin (10:08.386)
They may be talking to somebody who thinks about it entirely differently. What do these terms mean to you and how are they the same? are they different, reliability and resiliency?
Mary Cleary (10:16.79)
Yeah, sure. So, you know, obviously, neither of them are new terms, right? But they're both increasingly becoming moving targets. And something that's been really encouraging to see, I think, is both utilities and policymakers actually think about how both of them need to be approached with a different or increased level of rigor, right? So on the resiliency side of things, the asset hardening capital allocation framework that maybe worked beautifully five to 10 years ago probably needs a rethink.
Mary Cleary (10:45.41)
And on the reliability side of things, you can no longer just look at safety and call it a day. Safety is, it doesn't actually include in many cases the severe weather impacts, right? So you're essentially excluding all the stuff that makes the biggest impact on the system and only looking at, excuse me, duration and frequency in quote, unquote, normal course operations. And so...
Mary Cleary (11:10.008)
You know, the other dynamic at play is loan growth, right? So objectively slash obviously that creates more demand that the system needs to service. But a vastly underappreciated dynamic there is what does that do to customer expectations? So customer expectations are, they're going up. They're not getting more forgiving. So we ran one of our recent consumer polls where a third of Americans
Mary Cleary (11:37.366)
said that they expect power to be restored after severe weather within one to two hours, and another third said they expected the same within three to five hours.
Doug Lewin (11:46.518)
Is that even possible? that a completely unrealistic expectation, or do think it's actually possible?
Mary Cleary (11:51.33)
I think it's, I mean, it's not gonna happen overnight, right? But I do think it's not impossible over the right time scale.
Doug Lewin (11:59.522)
It also depends what kind of extreme weather we're talking about, right? If it's a category five and you're on the quote unquote dirty side of it, right? The east side of a category five in the northern hemisphere. Are you in Australia?
Mary Cleary (12:13.676)
No, I'm in the Northeast.
Doug Lewin (12:15.534)
You're in the Northeast. Okay, okay. It's an Australian company, which is why asked that. But if you're the Southern Hemisphere, I guess it would work the other way. But anyway, the Northern Hemisphere, you're on the Eastern side and you get a category five, like one to two hours is probably completely unrealistic. But I think there are a lot of events where that could be possible. So are you guys actually like... So a hurricane is, for instance, forming in the Gulf of Mexico and you're a Texas coastal...
Doug Lewin (12:44.75)
city or utility that covers a Texas coast, if they were using you guys, could they actually put that storm with its features into AI and then into your AI engine and then actually see, if it goes this way at this level, this is the path it's going to take so they can pre-position crews and things like that? Is that a use?
Mary Cleary (13:09.006)
Exactly, exactly. So what you're describing is a use case we're really, really familiar with. we've got, so actually, did you, I don't know if you had a chance to speak with South Australia Power Networks at the event last week.
Doug Lewin (13:20.718)
heard his presentation. Yeah, was fascinating. So that was a flood situation, Yeah. Yeah. Yeah. Tell us about that.
Mary Cleary (13:27.798)
Yeah, so what you're describing is a very, very similar scenario where we had a flood model kind of pre-set up that the model actually existed and this flood was coming in and they were able to layer on in real time those conditions. So as if that storm in the Gulf is brewing. So you essentially bring that situation to life in real time in the model and you see exactly where the water levels are going to get dangerously high, where you need to...
Mary Cleary (13:55.352)
de-energize or it's safe to re-energize and to your point where you can send people where it's safe.
Doug Lewin (14:01.23)
amazing. So we've talked about a couple of use cases, floods, hurricanes. Wildfires is another one I think is probably an obvious use case for this too. Can you guys talk about how you're trying to mitigate wildfire damage?
Mary Cleary (14:17.806)
Yeah, for sure. know, wildfire is such a multifaceted challenge, right? It's not, it's not a, you know, it often gets pinned on veg and veg is certainly a part of it, right? But that's really just one part of it. There's asset failures, there's design flaws, there's a lot that can and does go wrong. So in terms of how we mitigate vegetation, it's typically vegetation, sorry, wildfire risk. It's that typically a combination of a utility using our model.
Mary Cleary (14:46.092)
to combat their risk on all relevant fronts.
Doug Lewin (14:50.03)
So with Wildfire, you're trying to... I'm trying to think of what you would input into that model when you're dealing with the hurricane. I think everybody can kind of picture it. You've seen the images of the eye and the clouds and the rain and all that around it. But with Wildfire, it's a little tougher. You're trying to deal with what are the wind speeds? What is the humidity? Are these all inputs that you can enter into the platform?
Mary Cleary (15:18.112)
Yeah, so some that come to mind off the bat are wind speeds for sure, but also conductor temperature, right? So when the conductors get hotter, you know, on hot days when wildfires more likely, they are more likely to sag and come into contact with things like vegetation or a structure, potentially worse and ignite. So we're doing all that very granular clearance modeling to figure out where that's actually a risk.
Doug Lewin (15:45.614)
So do you guys actually, I'm just gonna ask the stupid questions, because sometimes the stupid questions are the ones everybody's wondering and just is afraid to ask, but I won't be afraid. The stupid question here is, have you guys, with LIDAR, you just have the whole globe or do you have to go, like somebody hires you and then you do their section of the world, or is this model just has everything in it already?
Mary Cleary (16:08.526)
Yeah, no, so we don't actually capture LIDAR utilities. Sometimes they have it already. Sometimes they don't. Sometimes they'll capture LIDAR in advance of working with us because the model requires it.
Doug Lewin (16:20.206)
Describe exactly what LIDAR is. I think I understand what radar is, but explain to me exactly what LIDAR is.
Mary Cleary (16:26.286)
Yeah, sure. So starting with the acronym, the fun part, LIDAR stands for light detection and ranging. And essentially what it is, is imagine a bunch of lasers shooting off of, sometimes it can be, sometimes it's a helicopter, sometimes it's a fixed wing aircraft or drone even in some cases, or a handheld device. But picture laser shooting off of some device that hit whatever is in front of them, right? So that could be the ground, it could be a wire, it could be a pole. And they bounce back.
Mary Cleary (16:54.664)
onto the device that is doing the capture and takes measurements of that distance. And so because it's measuring distance, it's fundamentally really, really good at solving for 3D questions versus imagery, which definitely has its place and role in all of this, but is more 2D.
Doug Lewin (17:18.743)
Got it. When you were talking about line sag, and you said it can get very granular, that led me to think, I'm wondering if this actually has, and maybe you're already doing this, maybe it's just something that's on the roadmap or something you wouldn't do at all, but the dynamic line rating gets talked about a lot in Texas. I've written about it before. Folks that are regular readers of the newsletter will remember probably.
Doug Lewin (17:46.04)
depending on how long they've been regular readers, but it was one of the most read pieces I ever wrote was on September 7th, the day after the only energy emergency Texas has had since winter storm Uri. And on that day, there was a decision made by ERCOT to actually reduce the amount of power flowing over a line. And the thought was, well, we could
Doug Lewin (18:10.914)
we could overload that line. It was really hot. This was September 6th. It was still extremely hot. And they were worried that they were going to overload that line. But my understanding is that their sort of dynamic line rating for those just listening, I'm doing air quotes right now, is not super dynamic. They take a couple of different buckets. And like if it's over this temperature, we turn it down this much. But if you actually had measurements and AI modeling, you might find you could flow more power over those lines. You might.
Doug Lewin (18:40.404)
I you could flow less power over those lines, but you would have... Is that a use case for either what you do or other AI engines?
Mary Cleary (18:49.184)
Yes, absolutely. And we do do that. We do do that. yeah, I mean, what you're describing is, I think a common, one of the reasons this hasn't been done in the past, right, a common challenge with this is like, it's totally impractical, right, to send someone out in the field and manually measure the distance between the graph clearance of every wire, particularly at distribution, which is, I think, totally, totally underappreciated component to overall network capacity.
Mary Cleary (19:16.096)
And so when we talk about modeling those granular clearances, that's exactly what it is. So sometimes our customers will find that there are bottlenecks, e.g. clearance requirements where they can't run the line any harder without reaching clearance. But what those bottlenecks can then become is potentially a punch list of things that they can do, design modifications to their network, where they can then actually run more power.
Doug Lewin (19:40.834)
reconducting things like that.
Mary Cleary (19:42.68)
Exactly,
Mary Cleary (19:43.121)
exactly. Or maybe, maybe reconducting, but also, or maybe, you know, what if we made this poll a little bit taller? Would that work?
Doug Lewin (19:51.17)
Very interesting. And do you actually need physical sensors on the line for that or can AI sort of handle that knowing what the line is and what the temperature and humidity is?
Mary Cleary (20:02.86)
Yeah, so we don't rely on physical sensors for that.
Doug Lewin (20:06.476)
It can entirely be done with AI. It's so fascinating how quickly the world is changing. It's such a challenge just to even try to keep up. It's amazing. Yeah. OK, I want to ask about a couple other use cases. Winter storms, that's something people in Texas think about sometimes. Yeah, is there a use case there?
Mary Cleary (20:30.05)
Yeah, absolutely. So ice loading is something that comes up quite a bit. And it really comes down to the same variables, essentially, as the pull load, the tension on the line, and temperature. So freezing temperatures, what that looks like, what precipitation looks like, and wind.
Doug Lewin (20:48.942)
Yeah, so that would deal with what is obviously a very real problem. It's not talked about as much, but February, 2023. Obviously, February, 2021 was Winter Storm Uri. February, 2023 sometimes referred to as Winter Storm Mara or Mara. I've never actually heard which way it's supposed to be pronounced. But in Austin and other utilities around central Texas, they had a ton of ice really started to cause tree branches to fall. So again, that's sort of a...
Doug Lewin (21:17.678)
an obvious use case for this. I'm also just wondering about there's a winter storm coming, you're looking at temperatures in the Permian of a certain level. Can you get into modeling what might happen to gas infrastructure? Could you get into modeling what demand actually might look like, which is an area ERCOT has really struggled with? And I've been critical of them for that, but frankly, that's really hard to do. You have to really understand the building stock.
Doug Lewin (21:47.182)
Can you start to get into how much power will actually be needed and where are the areas where it's most likely that you would see problems with power infrastructure, gas infrastructure, et cetera?
Mary Cleary (22:02.092)
Yeah, absolutely. you know, one of the things that's great about our technology is it can be used to, it's very good at modeling physical assets. And so whatever the physical asset makeup is, it's very adept at identifying risks therein. And in terms of like system-wide modeling, one of the things that's also great about our technology is the ability to answer questions that go upstream and downstream across the, across the energy supply chain, right? So like coming back to your point on
Mary Cleary (22:32.194)
reliability earlier, I think something that is changing in how we talk about reliability is not just obviously looking beyond safety and including things like severe weather, but looking all the way downstream to things like generation mix and understanding how does that affect customer outage minutes from a resource availability perspective.
Doug Lewin (22:51.906)
Tell me more about that. Yeah, what is, I mean, I know what that means, but like, how does that factor into an AI engine that is mapping? I don't, I understand what you mean, I don't understand the connection.
Mary Cleary (23:03.244)
Yeah, sure. So what I'm saying is, you know, obviously the energy ecosystem is very complex, right? Lots of moving parts. And that's one of the reasons why it's so hard to make better, faster decisions. in helping utilities forecast resiliency, we're looking at all those moving parts and how they turn together. And it's almost like the AI plays this role of a programmatic dial that you can turn up or down to look at reliability and resilience risk.
Mary Cleary (23:33.046)
evaluate objectively how much of that risk is there and how much and what does it cost to buy it down.
Doug Lewin (23:39.416)
That's so interesting. So you could then conceivably, when you're talking about resource mix, you could look at, this would be an area where on a certain, say, a side of a transmission line, maybe there's a constraint. This would be a great place for battery storage to go. Or you have so much battery storage in this one particular place, but there's really no gas resources here. This would be a place to look to put a gas peaker.
Doug Lewin (24:07.086)
I'd assume there'd be a big relevance there for distributed energy resources too. We can't put any really large infrastructure in this particular place, but maybe some customer-sided stuff would be a particularly good solution. exactly. How far along... I think part of what I'm trying to understand, and I'm sure everybody listening is trying to understand, is how far along is AI on this at this point? Because I think Neera, it's safe to say, is...
Doug Lewin (24:36.814)
state of the art, you guys have been at this for what, seven, eight years, you're through a series C, you have a bunch of different deployments, but it still feels like this is early days. Like there's just like the advancements are probably going to like... So what are the next areas where you're excited, where you're like, I can't wait to see what AI will do with fill in the blank?
Mary Cleary (25:00.974)
So that's excellent question.
Mary Cleary (25:05.474)
You know, I think that there's a lot to be excited about, but I think there's also a lot to be cautious of, right? AI is, I'll use the hackney phrase because it's true, but it's not the silver bullet. And there's a change management journey that goes along with adopting AI, right? it's, can't, like, in other words, I guess an example of what I'm saying is you can't suddenly...
Mary Cleary (25:31.054)
cover much more ground in your network inspections, right? Or you can't suddenly discover 100 risks where you were only able to previously see one to 10 at a time and then not be able to do anything about it. So there need to be surrounding processes or infrastructure, not in the physical sense of the horse, but that actually support these processes. So that's one thing I would say about it. But I think the bigger opportunity is something I'm extremely excited about.
Mary Cleary (26:00.622)
is that AI can actually help between both utilities and policymakers help encourage and really drive this idea of a proactive approach to resiliency and reliability, which I don't think is controversial. don't think anyone thinks that's a ridiculous statement. But unfortunately, the reality says otherwise, where a lot of the processes and regulation that are in place today actually actively discourage taking a proactive approach.
Mary Cleary (26:29.41)
And think Eric, Brian.
Doug Lewin (26:31.402)
Yeah, what's an example of that? That makes sense.
Mary Cleary (26:34.318)
typical of that
Mary Cleary (26:34.918)
is California, there's something they not so affectionately call the 45-day rule, and I'm sure there are examples of this in other states as well. But it's essentially that if you discover a risk, you have 45 days to actually remediate it, which is fine. It sounds sensible, right? No one wants risks lingering out there unattended. Except that is essentially, how is that different from playing Whack-A-Mole? You find a risk, you fix it. You find a risk, you fix it.
Doug Lewin (26:54.614)
Festering, yeah.
Mary Cleary (27:02.456)
What if that risk that you're suddenly pouring all of your resources into at that moment isn't what you actually need to be paying attention to?
Doug Lewin (27:09.11)
It may not be the biggest risk at all. There could be a bigger risk somewhere else, but you found it 20 days into that 45-day period, and so you let it go.
Mary Cleary (27:17.526)
Exactly. And even worse, what if there are a hundred or a thousand versions of that same risk and in aggregate, there's a more cost-effective way to solve that risk.
Mary Cleary (27:29.73)
instead of just doing it one off.
Doug Lewin (27:32.27)
It brings me, you when you're a hammer, everything's a nail. But this brings me back to it. I wrote a piece on this like not quite a year ago. was right after Hurricane Barrel about you get what you pay for. And starting to think about performance-based models where we're really paying utilities a lot more if they do their jobs exceptionally and less if they're not. I won't ask you to weigh in on that, but I do think there is some merit in that because then you can get out of that sort of...
Doug Lewin (27:57.454)
hamster wheel of the way things have been done and it just has to be done this way as opposed to putting the focus on the outcome you actually want to drive towards. do, want to ask you about, you've mentioned a couple times, Sadie and Safie, and I think probably most of our listeners know what that is, but I actually want to take a minute and just kind of double click on that. So these are, I'll give my explanation, then you correct me and say it better than me.
Doug Lewin (28:25.624)
And then I want to get into a little bit like what might be a better metric. So, salient safety, whatever its system average, but the D and the F are the important... Go ahead. What is it?
Mary Cleary (28:37.556)
It's System Average Interruption Duration Index.
Doug Lewin (28:40.872)
So
Doug Lewin (28:41.122)
duration and frequency, right? So how often are outages happening and how long do they last? But as you mentioned, a lot of times the extreme weather events are removed from that metric because it's sort of viewed as like force majeure. It's like, well, this is what are you going to do? Sometimes the power just goes out. But as you said earlier, customers are not cool with that. Like that's not okay. expect...
Doug Lewin (29:07.362)
to have the power back on. Have you seen any... I was on a panel at a conference a couple of months ago with a guy from Con Ed, the utility in New York, and he said they are actively trying to come up with different metrics because the old ones are so flawed. You work on this stuff around the world. Have you seen better ways to actually measure the outcomes we're after?
Mary Cleary (29:28.782)
don't think there's something perfect just yet. think the perfect metric will be something that is forward-looking, that's more of a leading indicator instead of a lagging indicator. And I think using predictive modeling technology puts utilities in a place where they can actually do that reliably and then communicate that to policymakers so that there's a sheer understanding.
Doug Lewin (29:31.298)
That's for sure.
Doug Lewin (29:53.39)
So basically, that's really interesting. So you could basically run a predictive model and say, hey, if a category three hurricane hit Corpus Christi, if a wildfire with wind speeds of 55 miles per hour hit Amarillo, God forbid any of these things happen. But you could, if a flood of Hurricane Harvey in 2017, we got a repeat of that and...
Doug Lewin (30:20.238)
2028, you could plug those into the model and you would say, based on existing systems and processes and infrastructure, we would expect power to be out this long. And then basically say, here's the triage list, utilities fix as much as you can, and the better you do against that benchmark, there could actually be a performance incentive potentially. Could be an increase in rate of return, could just be a cash payment, could be any number of different things.
Doug Lewin (30:48.226)
which I think the vast majority of customers would support because then you would actually have lower outage times, which is what everybody's after. Is that what you mean or do you mean something different?
Mary Cleary (30:56.588)
I think that's, I don't think that angle is out of the realm of possibility.
Doug Lewin (31:01.91)
Is there a different way that you're thinking of that though? how, what, because that's the way I understood it when you said it. Am I understanding it wrong?
Mary Cleary (31:09.804)
Yeah, no, maybe another a little bit more color I would add to it is I guess it avoids this really unsavory situation, right? Where the pendulum's going back and forth between reliability and resiliency and affordability. And you got something bad happens, no one likes the outages and the impacts of them, but then the shop wears off and people go, my gosh, all this stuff is really expensive. Do we really need all this? Is this necessary?
Mary Cleary (31:36.064)
And then you end up in a situation where you're deliberating until the cows come home about your torturing cost line items and then nothing happens. And then something bad happens again and people wonder why nothing's changed. So that's really the situation I'm referring to, which is like, let's stop having this high temperature back and forth between utilities and policymakers where policymakers sees a really big price tag and utilities see the
Mary Cleary (32:04.674)
since the thing that needs to get done. And they're not speaking the same language.
Doug Lewin (32:11.564)
I think some of the reason why that happens in my view is there's this sort of information asymmetry where the utilities know their system better than anybody else and the regulator doesn't have nearly as much information. But I'm not sure how we get over that problem. Have you guys ever worked for utility commissions or grid operators or is it like utility side?
Mary Cleary (32:39.884)
No, we're defying the utility sign, but we are seeing our technology being used as the vehicle to objectively communicate what's needed. And so when policymakers bristle at, you know, eight, nine, 10 figure investment, utility can then use the technology to say, okay, you don't like that. Well, here's what happens if we don't do that. You're probably not going to like that situation either, right? But then it becomes this way to evaluate how much risk you can buy down and at what cost by saying,
Mary Cleary (33:09.742)
Okay, maybe we haircut that investment by 30 % or 15 % or whatever. Can we look at the outcome, the proposed forecast outcome on the other side of that? And is that enough to justify the investment? Do we feel comfortable with that? And that kind of thing, that kind of dynamic really takes that temperature down when utilities and policymakers are trying to make those really difficult decisions together.
Doug Lewin (33:30.446)
Yeah. mean, you say eight, nine, 10 figures. And in Texas, for the resiliency plans the utilities are doing, I think we're bumping up against 11 figures at this point, because, yeah, Houston's is 5.7. Center points is 5.75 over three years proposed, not approved. Encore was a little over 3 billion over three years. I think, though, there is... I don't think anybody... I'm certainly not arguing, and I don't think anybody could...
Doug Lewin (33:57.194)
there doesn't need to be investment made in the distribution grid. That seems obvious and axiomatic. The question is how much and how do you balance that affordability? And I think so much of it is about getting better data out there and letting everybody see it. Like you said, here's the predictive modeling, here's what happens if this happens. Here are the targeted things that can be done and rank order and what they cost so that policymakers and regulators
Doug Lewin (34:25.014)
and utilities together, stakeholders, et cetera, consumer advocates can kind of make decisions together about, you know, this is you do Can you, you can definitely respond to that. I'm also just interested in like, how does all of this play into the affordability side of things? Does this give us, does it do what I just described? Are there other ways I'm not thinking of where we're using AI for utility modeling and that sort of thing could actually significantly save money, reduce the cost of investment?
Mary Cleary (34:54.252)
Yeah, so I think on the affordability piece, the primary role that AI can contribute to that is by helping size with the remediation solution is, right? So we're getting away from this, these pulls are weak in a category three, they're not going to make it. Let's just replace, let's panic and replace them all. Instead, let's add accessories like stays were appropriate and maybe even the ones that you need to replace. Maybe some of them can be one for one, wood to wood.
Mary Cleary (35:21.506)
And maybe there's only a handful of select hyper-targeted scenarios where you need something stronger like a steel or composite.
Doug Lewin (35:29.664)
And in the model, we'll get that granular to say in this spot because of maybe the geography, this one sits a little higher, it's a little more exposed to a coastal wind or whatever. This is the one that needs to be replaced as opposed to 50 of them. It'll get that granular. That's really interesting. All right. Good stuff. let's, course, it's the Energy Capital podcast. Let's talk about...
Doug Lewin (35:58.67)
policies. And again, and it actually may be before we jump to policy or maybe just as part of the same answer, I am curious just kind of where in the world you guys are working right now. We've talked about Centerpoint, you've mentioned Australia. Where else are you guys doing work around the world?
Mary Cleary (36:15.82)
Yeah, so we're working with utilities in pretty much every major region of the US right now. We're also in Europe. We have several customers across Europe, Asia, UK, South America increasingly. So I think we've got the continents covered except the Arctic.
Doug Lewin (36:33.826)
Yeah, no Antarctica yet, but know, they're in dream one day. So you see policy constructs all over the place. And this is one of the things I just think is not done nearly enough. know, Nehruq and other organizations, International Energy Agency, those kinds of organizations do a good job trying to like cross-pollinate, help people see the way, you know, folks do it differently around the world. But you guys get to see that.
Doug Lewin (36:59.136)
Are there any particular policy constructs that you think are particularly effective around reliability and resiliency?
Mary Cleary (37:05.836)
I don't think there's something yet that I would point to and say that's the thing that everybody should be doing that. I think, to come back to this idea of taking a more proactive approach, I think there are baby steps being taken around the world that kind of smell like that, but I don't think we're there yet.
Doug Lewin (37:25.826)
Yeah, I think that's right. I think this is, I always like to say, think, you know, there's a lot of innovation happening, obviously, on the tech side. There's a lot of innovation happening in the markets. Policy innovation is important too. You have to be able to like think differently because the regulatory constructs pretty much everybody's operating in were developed and evolved over a hundred years ago. And they've evolved some, of course, since, but like not a ton.
Doug Lewin (37:55.566)
It's an area that for the technology to actually have as much impact as it could, you're going to need the regulatory side to catch up, I think.
Mary Cleary (38:05.888)
Agree? Agree.
Doug Lewin (38:07.566)
But is there anything I should have asked you that I didn't? Anything else you want to cover?
Mary Cleary (38:14.414)
Let's see.
Doug Lewin (38:16.782)
There's so much, right? I mean, it's such a fascinating space, but.
Mary Cleary (38:21.1)
Absolutely. So Doug, think maybe there's one last thought I'll leave you with, which is that when we talk about the grid, we're often talking very big terms, right? We talk about massive amounts of money, we talk about sweeping projects and massive time scales and things taking a very long time. But something I am confident we're going to see more of and very excited about is this idea that we could see grid transformation happen on a more micro scale.
Mary Cleary (38:50.306)
with more micro outcomes. this idea that the energy transition and the idea of resiliency doesn't need to be this massive nail biter, right? Like everyone is on the edge of their seats every time there's a bad weather report. There's the Senate Bill 6 action earlier this week, so everyone's wondering what's gonna happen with large loads and such. And so I think technology like ours can play a really critical role
Mary Cleary (39:20.062)
and allowing utilities to take better, faster, more thoughtful, smaller steps, but in faster succession. So it keeps everyone on the same page and working towards the outcomes that we all want in a way that's much more affordable for consumers on the other end of the energy bill.
Doug Lewin (39:37.144)
So I was going to end, but now I want to ask about that. So we're going to go to another couple of minutes if that's okay. So Senate Bill 6, obviously I wrote about it a little bit and we'll put it in the show notes where I wrote about it and I'm sure I'll be writing about it a lot more in the coming months. And we're recording on February 14th, so the bill was just filed a few days ago. By the time you're listening to this, there could have been hearings or what have you.
Doug Lewin (40:01.368)
But it's a bill that really does kind of change some of the ways that large loads interconnect to the grid, change potentially some of the requirements for the generation they need to bring, and then change how they pay for transmission. I think, and Mary, I'll just say this the way I perceive it, and then you can change it, say it however you want. I think, I'm not sure if this is what you're getting at, but I think there's a little bit of sort of a freak out happening right now, frankly, where a lot of folks are like, my God, they're seeing the numbers of we might need 50, 60, 70 more gigawatts in the space of five, six, seven years, and that looks really scary. And I get it. We have added 13 gigawatts last year, an average of 11 gigawatts the last four years. Is what you're saying that these... And obviously, this is...
Doug Lewin (40:55.246)
but there's a whole lot of companies working in this space that AI actually, while it is itself a big load, actually, as we were talking about earlier, help solve some of the problems too. Almost itself be energy aware of when it's using energy, when it's using the energy that's on site. Maybe even helping to deploy distributed energy resources, help folks that have batteries, either in their garage on the wall or just in the car, like actually deploying assets and resources in a smarter way that makes us more reliable. And I didn't give my own definition earlier, Mary, but my own definition is reliability is do you have enough supply to meet demand? Resiliency is can you deal with extreme weather in a way where you stay online as long as possible and then you're back up that restoration is a very short period of time and distributed energy resources help with both of those things, right? Because they're adding supply and they're giving you resiliency because it's close to where you're at. So maybe I'm completely twisting around what you meant when you brought up Senate Bill 6, but that's where my head goes. What did you mean when you said that?
Mary Cleary (42:06.092)
Yeah, so for me, it was just yet another example of all the anxiety about energy. But I think you make an interesting point about how much is AI contributing to the anxiety and how much anxiety can it potentially detract from the equation. I don't know exactly what the answer is. I can't speak to a self-aware, energy-consuming AI anytime soon, although that'll be very exciting to see. But I do think there is massive opportunity for it to be a net anxiety relief.
Doug Lewin (42:39.692)
Yeah. And when you talk about the micro solutions, can you talk more about what you mean? Is that just like what you were saying earlier, like getting down to the individual poll? Or does that also deal with like micro generation? Does that come into the picture too?
Mary Cleary (42:52.046)
It's more thinking at the asset level, asset resiliency.
Doug Lewin (42:56.814)
Yeah, yeah, yeah. No, it makes a lot of sense. And I think it's really important for all these different things, whether it's distribution infrastructure, generating resources. We have to think of the micro and the macro together and how they fit together, because the really small things can actually add up to a whole lot. All right, cool. I'm going to ask the same question again. Is there anything else you'd like to say, because then I just added a whole lot to what you said at the end? Anything else you want to say before we end, Mary?
Mary Cleary (43:20.866)
No, I just had a question for you and it just went right under my head. No, you know what? It was actually about House Bill 2555 been out in the wild for some time now, but how is your level of optimism changed, if at all, since then?
Doug Lewin (43:40.366)
Well, honestly, this conversation actually has made me more optimistic about 2555. What I worry about 2555 is it's a little bit of a blank check. Now, I don't want to be Dougie down or just be negative here because I do think that that overall is a good policy. I think that you need utilities to spend money on the distribution grid and on resiliency, and it's a regulatory framework that gives them some certainty emboldens them a little bit to actually invest in the distribution grid, which as we all saw after Hurricane Barrel and other events, that wasn't the only one. Again, the Austin ice storm in February, 2023, there's all kinds of examples of where the distribution grid needs investment. My worry about it is there isn't enough data that folks can really look at and together and agree on. And I think that what y'all are doing,
Doug Lewin (44:38.078)
And on this podcast, I'll interview a lot of different companies going forward that do this, because I think the competition in this space is pretty interesting. And I think that there's going to be a lot of competition in this space. I think you guys are an early mover, and I've seen examples of your technology. I'm pretty excited about it. But to me, that is one of the most exciting things that the regulator, the utility, the stakeholders can all kind of see. No, here's the predictive modeling. Here's what happens if a cat three, cat four, cat five hits this particular area with the infrastructure you have in place. It ain't pretty and here's the rank order of what you could do to make the most impact at the lowest cost to reduce the damage and the potential pain and suffering that would happen after that. So I have no doubt that we need to spend on resiliency and I understand why lawmakers wanted to pass 2555. The information asymmetry and there's sort of the general lack of data that's been out there is one of my biggest concerns about that balance between reliability, resiliency, and affordability. I think what you guys are bringing to the table is potentially transformative. So I'm pretty excited about the technology. What do you think of 2555?
Mary Cleary (45:52.846)
You know, I am really excited, obviously, based off our work with utilities in Texas, what I'm seeing so far there. But I think it sets Texas very well apart from other states in this country, as well as other folks globally in a lot of ways. Because a lot of utilities, I won't name names, but they're still thinking about things through very narrow lenses, right? They've got their pet problem and Texas is a bit unique, right, because it experiences a whole laundry list of things, whereas other areas of the world maybe experience two or three of them. But I think...
Doug Lewin (46:30.811)
We've got it all here in Texas. You name your climate risk. It's right here.
Mary Cleary (46:36.226)
But just zooming out and thinking about everything through that resiliency lens, think that's a really good example that others would move themselves to follow.
Doug Lewin (46:45.548)
Yeah, yeah. I it makes a lot of sense. Yeah, I just think the layer that hasn't been there yet is the regulators really insisting on seeing the data and not saying that they haven't done it yet. It's just like we haven't had these tools. Right? So again and the data wasn't necessarily available, you know? Exactly. I get my pen and paper and I scribble something down for you and it's like, as a regulator, are you going to put much stock in that or do you want to audit something and not take my word for it?
Doug Lewin (47:19.886)
Yeah, again, regulatory innovation, right? think regulators really insisting on we need the best possible data. They have a huge job. It's damn near an impossible job. I don't envy them for having that responsibility, but they've got to make sure they're looking out after ratepayers and making sure that, we always say keep the lights on. No, it's like making sure the heat and the air conditioning are on. That's so people can find a flashlight. It's the heat and the air conditioning. Keep that running. Mostly air conditioning in Texas, though. Anyway. Mary, I really appreciate you taking the time. I'm excited about Nira. I am excited about everything I've heard so far. I'm looking forward to following the growth and evolution of this company and all of these kinds of use cases to hopefully improve outcomes for customers. And yeah, thanks for taking the time.
Mary Cleary (48:15.438)
Pleasure speaking with you, Doug. Thanks so much for having me. Take care.
Doug Lewin (48:18.04)
Thanks, Mary. Thank you for listening to the Energy Capital Podcast. I hope you enjoyed the episode. If you did, please like, rate and review wherever you listen to your podcasts. Until next time, have a great day.
This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit www.douglewin.com/subscribe