#2 What Actually Makes a Solar Project Bankable? | Logan Boutilier

- Hello everyone, welcome to Beyond the Layout. I'm here with my friend and old boss Logan Bulltier and he is currently VP of engineering and Aspen Power and previously worked at DNV and the renowned independent engineering firm. And so we're here to talk about banking solar projects and making projects that are simultaneously bankable but also will live to the lifetime of the asset class. So Logan, thanks for hopping on. How's your day going? - But it is going great, Rockhead. Thanks for having me. I appreciate that we get to do this 'cause we talked about it in person alone. - Yeah, yeah. So you taught me everything that I knew about banking projects before I was under you, I was working at Safari Energy and we were running Helioscopes and once in a while we'd run a PVCIS but there were no actual standard procedures and ways of making the assumptions necessary to then go and finance that project and to make sure that that model is really accurate and used by our asset management team. So could you speak like starting from soup to nuts about how to bank projects? Like what does it mean to, what does bankability mean? Let's start with that. - That's a good question. It's a loaded question. - Honestly. - Bankability to me means that a project has the confidence in within the banking community to invest not only construction debt but also tax equity money into the project knowing that returns are going to be provided to the banks and all the stakeholders involved. That's kind of a very complex answer to say does the project receive banking money? And the banks are very conservative in their investing and they're going to ask a variety of questions to any level of depth that they feel necessary and I feel that it's the independent engineers' duty and the developers and the owners' duty to provide some sort of pushback on the banks in certain commercial environments. Ultimately, if the bank is willing to fund projects on favorable terms, that's interest rates, pay back periods, screening projects through their underwriting process, that to me is ultimately the definition of bankability. If a bank is willing to fund whatever debt is needed for the project to move forward, that's what I call bankability. - So realistically bankability and quotation marks is kind of like a spectrum where the most favorable terms is the most bankable of a project and the least favorable terms. Maybe that project is still receiving some of that money or that equity, but you're getting it at such unfavorable terms at that project now maybe not, does not pencil. Is that what I'm hearing from that? - Yeah, that's accurate. - Perfect. So how do we make it so that we have the most favorable bankable projects and not the least favorable bankable projects? - Are you doing what your, are your projects performing to what you say that they're going to perform to? So are you selecting equipment that is going to be reliable, not going to have tracker uptime reliability issues, no snail trails, although I think they were pretty much passed snail trails in the module sector, but I've been around the industry long enough to know when I go out, that used to be a problem in the industry. Snail trails lead to module failure that you don't really see unless you're doing a thermo flyover, like a thermographer, aerial thermographer flyover. You have somebody inspecting the project on a, it consisted enough basis for these things to come up. So equipment selection is very important. You need reliable products that are shown to be tested to known standards for IEC, for UL, and that it's, that all that equipment is installed in a consistent manner and a reliable manner to where really it can be, touch it in, forget it. - So what was I- - So let's set it in, forget it. - Yeah. - So let's start with the equipment and then we'll go to the energy modeling aspect of this. How do we as an industry, like this industry is very nascent, right, we've only been around for maybe a couple of decades, three decades. How do we, and when we're financing these projects and trying to say these projects are gonna last 25, 30, 35 years, how do we actually get to the point where this equipment vendor or the equipment that you're using is vetted to the point where you're confident that that is going to be around in 30 years, because industry itself hasn't really been this developed for that long. It's such a good question. If I can provide a little bit of background story, a little bit of background to what we're talking about in terms of useful life. When I started at DNV, this was back in 2014, I started with DNV, all financial models at the time were looking at a 25 year useful life for the project. So what does that mean? The financial model is taking into account monthly expected revenues for the projects that are based off of all of the capex going into the project, the op-x once the project is operational, and the revenue is expected for the project based off with the predicted energy models, that are whether adjusted over time, and those ended at 25 years. So we knew there was a 25 year window that the project's expected to generate revenue. As that number started to increase from 30 to 35 to 40 plus years, you're now going out into the uncertainty of predictability of these projects to be producing the same amount of energy or the same amount of energy with increased downtime, increased degradation over the period of 40 plus years. There was no data in the industry for large scale solar projects to be operating for 40 plus years. But the banks and the developers wanted to say, it's similar to like a car loan. You could take out a 33 year car loan and have a very high monthly payment or you could go out for a 72 month car loan and have to pay it back over a much smaller monthly payment over a longer period of time. - And the banks would that make more money because there's higher interest payments. So that's a win-win win for everybody if the promise of that bears fruit. - Correct. Yeah, so what we started to come up with in specifically, my experience has been in distributed generation solar. So instead of deploying a single 300 megawatt project, we were representing companies that were building 30 five megawatt projects that were scattered around the US, that were scattered around even microclimates, say Pennsylvania, New Jersey, Massachusetts, Maryland at the time, there's some confidence that if the problems or the weather in one location is down in one year, that the weather will not be down across all projects, across all regions. - Across all years. - Across all years, exactly. So there's a little bit more confidence in a geographic distribution. When you look at a five year operating period of 30 assets versus a single asset in five years at one location. The ability to predict on an annual basis when it comes to energy modeling, if we jump to energy modeling from equipment, when I started in the industry, we were using NSRDB2 and TMI2 and TMI3 data sets that were ground-based stations that were well vetted, but had some historically dated data. All the data was from the mid 1900s to 1991 and the TMI2 data sets and then the TMI3 data sets were 91 to 2009-ish for establishing what the P50 or long-term average expected weather would be at a given location. At the specific location of the station. So this is, I feel like I joined in. I'm fortunate that I joined into the energy modeling guru community which I didn't understand until I joined DNB. I thought I was very good at running PV systems and understanding energy modeling until I joined DNB and started to hear from the experts and I learned that I was actually an amateur. So my time at Barrego, I was trained by the historic B.E.W. crew that were the first, they underwrote as the independent engineer, they underwrote the Sun Edison deals. They started to become a very prominent independent engineer in the industry, set a lot of the PV systems standards. Some of the experts at B.E.W. at the time consulted with PV systems to upgrade the software and get it to be the software that it is today. I joke to with you all the time in training PV system, like don't go into PV system for because there's ghosts and the buttons are terrible and you have to do post processing and it was just the growth of that software has been really to benefit the user and benefit the community as a whole over time. So as the satellite data provider started to come in, so this is solar anywhere data, Vaisala, DNB has a sole cast, - And there's our solar GIS as well. - Solar GIS there, a bunch of different vendors out there that were pushing, this is a much better, if you go down to your 10 kilometer grid, your five, your three or even your one kilometer resolution, you're gonna get much more accurate, predictable long-term energy models from those weather resources, then a station that's 50 miles away. However, what, as an independent engineer, our job was to be skeptical of all new information coming in. The incentive is for the satellite providers to sell more of their product and it took the IE community a number of years, it was actually just two years ago that DNB dropped all of the NSRBB stations and now really only consults with satellite providers because it's taken a decade for the satellite providers to understand the regression analysis that they do when they run the, It's the NREL satellite that reads the expected GHI at a given location that they tracked uncertainties properly through all of the calculations. Early on, we discovered at DNB, a man named Jeff Newmiller, who speaks calculus when the rest of us speak arithmetic. He was describing this to me in a bunch of the younger engineers at DNB at the time, is that the satellite providers were not tracking the uncertainty of the data that the regression analysis used to kind of fill in the holes. And so you would have a recording of 6% uncertainty in the data, but you got to really add on 4% to that because they did not include the uncertainty in the pieces of data that they used to fill in the holes or the gaps. Also, things like reflectance from large bodies of water or snow cover. Now, all of that has been updated with the help of the independent engineer community to provide more confidence in the satellite data so that when you read a tile, you now, they've overlaid like the thermal images to be able to filter out periods of snow, or filter out areas of snow, so that you're not getting the GHI fluctuations in the middle of winter that would be affected by reflectance from snow or large bodies of water. - Yep. - So inputs to your model are very important. One PV system report, and I would do this with dozens of different developers during my time at DNB. They would come and say, we run PV system, we use the closest NREL station, what else can go wrong? Like, why is this not accurate? Well, if there's a giant mountain range, Massachusetts is a great example. Lots of microclimates. Lots of unreliable NREL stations, but some very, very reliable NREL stations. And you've heard me say this, anything to do with, any project that's outside about a 20 mile radius of Boston moving west. Wister TNY2 is almost always going to be the selected data source when we were using ground-based stations. And it was weird when I joined DNB, I heard the experts say this, and I was skeptical, and I'm like, how do they do that right off the top of their head? So then I would go through, and I do this analysis, and it was painstakingly awful to go through this analysis in Excel, where we have to pull down all the NREL data manually from NREL's website, plug it into an Excel tool in different tabs, and then we'd have to, actually, we'd have to pull the data down, plug it into PV system, output a specific monthly table of GHI, DNI, wind speed and temperature. Plug it, it pasted into Excel, populate these charts, filter out the outliers, physically by looking at charts. Now this is all done through DNB's solar resource compass, available online, and most developers are starting to use that, or some other quicker means to decide on a long-term weather solution. - Well, they're doing that with multiple different, to the best of my knowledge, and CREP, if I'm wrong, the IEZ are recommending to compare all those weather resources together for the same parcel, for the same exact location, to find the one that's at least the median or the meter, to be able to select that with higher salons, and just saying, here, I'm just using this insert, the name of the weather resource, and I'm only using this, I'd rather have you pick and choose which resource you'd like for that specific project, to be as representative for that specific microclimate, or that location as possible, and then use that instead of just picking one and sticking with one. - Right, 'cause the danger of picking one specific source for your data is, all data sources have inherent bias. In the calculation and how they interpret the results, there's inherent bias in selecting a sole source satellite provider. So the IE community typically has taken it upon themselves to look at the available resources on a month-to-month basis, and decide which resource best represents that specific location. Sometimes it's solar anywhere, sometimes it's by-sol, sometimes it'll be solar GIS, that is the closest to the median, or DMV has a new term, it's the MLE, which is I think the maximum likelihood expected, I'm getting the words wrong, but it's way too many acronyms in our history. - It is, I mean, even NSRDB in NREL is without defining those. - Yeah, it's a natural renewable energy laboratory, which has since been renamed under the Trump administration, too. I don't even know what, because it doesn't exist. (laughing) - Or NSRDB, the National Solar Resource Database. - Yep, and sometimes even the local weather, like on the ground, NSRDB stations will be even more accurate, especially for snow losses, than the satellite interpolation. So it's really up to the developer to figure out what that, what the ideal combination or single source of truth is for that specific site, when they're going through it. - Yeah, and I feel fortunate that I grew up and grew up in, was raised, grew up in San Diego, and now live in San Diego, so I can understand microclimates. And when I was looking at, when I was helping provide independent energy reviews and providing energy models for projects in San Diego, we have, I think, probably eight NSRDB stations in San Diego, two TMI, two stations, and five TMI, three stations. Some of them are right on the coast, some of them are five miles inland. Big difference. We're sitting in Rancho Fernando right now, and it's beautiful blue skies. If you drive five miles west of here, it could be 10 degrees cooler with mist and overcast all day. Meanwhile, we're sitting here at beautiful blue skies. That comes out in the data. And so you can't just take a project, say a project is gonna be built at my house. If I wanted to fund a project that's being built on my house, it would be a little bit misleading if I chose a station that had more reliable data right on the coast. Or vice versa. If I go five miles further inland, the temperatures are gonna be much higher, but they don't get as much cloud cover as we do here closer to the coast. So understanding microclimates is also important. And I know we're getting very detailed into. - Well, let's resource selection. - Well, let's take a step back here. The biggest, you said it could be misleading, right? Let's go into why developers would want to be misleading in the first place about, and even the resource providers themselves. I know historically for these satellite providers, they started a little bit high. And now, as you said, have gone closer to the mean for what these specific locations should be able to do, why would we want to say bolster our numbers? What are the downstream effects of that on not only that individual project, but then also the industry as a whole? So like, let's say what not to do, right? If we bolster our numbers, why would we want to, what's the incentive there? What happens when we do do that? And what happens when industry does that? So depending on where the developers involved in the process, they are incentivized to increase energy yields. They want higher overall performance with less uncertainty. Some developers, and the incentive is to find the resource that is the highest justifiable GHI input into PV system, or whatever energy modeling software is being used, because it allows them to borrow more money on the project. So if the long-term 25 to 40 year returns are going with a 1900, a 1900 GHI value versus a 1700 GHI value, they're going to be able to borrow more money at that rate, at the 1900 versus the 1700. And that all plays directly into the yield numbers that come out of the project, which is the energy generation and then the money that gets brought back into the project, because you've spent in the meter more. - Well, when you're talking from the person who's at, or the company or entity, that's actually owning the project, right? Upstream of that, our developers are pre-developing that project, right? And there, I'm speaking more so about the upstream developers. - Okay, okay. - What's been unique at Aspen is that we're kind of, we play both roles. We're developing projects to crank out as much IR as we possibly can. But then on the flip side, we're also responsible for actually sleeping in the bed that we make. - Exactly. So some developers don't actually are IPP's independent power producers that own that project for the lifetime of the asset. And for those developers, right? They're almost always incentivized to have a higher number attached to what the energy can produce from that. - Correct. - So that when they sell to an IPP, they get more money right off of that. They're out of the equation. Now the IPP is stuck footing the bill, proverbially speaking for a site that they expected to produce say 110, and now it's actually only producing 100. - Correct, right? - Yeah. And one of the things that I've done at Aspen with the help of the entire team that was there when you were there and a few other folks is, we have a very detailed energy model that's going to be a procedure document that gets updated occasionally, somewhat depending on which independent engineer we're working with, which banks we're working with. We're all learning, we're all refining things to try to be as precise as we can be when we can be precise. But something you taught me is, we still got to make a decision and take an action and actually move this thing forward and we can't spend 20 hours going through all the fine-toothed comb details of the project if we're just looking to purchase it from a developer. We have this fully detailed document. I think it's 18 to 20 pages showing exactly what we're doing. of how to run energy models giving background to our engineers of why we do things the way that we do. I took that document, removed all the background, shortened it down to about a four-page document, and now we hand that to all of our code developers. So the developers that are incentivized to crank those yields as high as possible, to get us into a letter of intent into exclusivity and reviewing the projects, they get paid their final payment, their developer fee at construction and TP. So once the construction starts, Aspen is now taking full responsibility for the energy yields for the project. I found it very challenging to continually talk with these code developers about the same themes, about resource selection, how you run LID, sub-hourly modeling, there's so many detailed nuances in running an energy model that developers aren't really paying attention to unless somebody tells them to. At DNB, I would find myself continually trying to inform these developers that were selling projects to our customers at the time of the same themes that I was seeing at Aspen, where is LID coming from? What does a qualified LID test report include in terms of the data points, and duration, and light soaking, to be able to plug that into an energy model? So we took it upon ourselves to shorten that energy modeling procedure and share it with our code developers and say, we want to align on yield as early as possible so that we're not fighting come closing time about a 5% difference in our yield numbers. Let's do it very early. This is how we will run it. You run it the same way that we do, and then we can compare. That has been the single most liberating and single most valuable effort in reducing conflict with these code developers that we lift the skirt and show them how we do things at Aspen. It's largely similar and based on my experience in what I learned at DNV, how to justify this long term with these code developers. That look, we know this hurts your returns for the project, your developer free fee that we pay you as the project matures, but it does no good for the deal to be selling us, to be selling us projects that are not funded properly. If we are left holding the bag that's going to be underperforming for the entire life of the project. >> So we dove pretty deep into the solar resource and that being a very big pillar of an input to these models, right? And you talked about this procedure that you've made at Aspen and I was a part of that and running the PV system back then. What, like you said, most of that comes from your experience at DNV and as the independent engineers, really the buck starts to stop with them, right? And so could we go down through what does it look like to have good assumptions? What do you think those good assumptions are? And then when you deviate from those assumptions, what in which cases can you do that? So we really did that, I think quite well for solar resource, but what about, as you said, LID losses, module quality factor losses, all the losses that go snow and snowing losses, all the losses that go in, what I like to say for an energy, for a quote unquote bankable energy model is there's three different pillars. The pillar of the inputs to the model, those are all of your losses that you put on the PV system or any other software that you use, as well as the weather resource. There's a layout itself with the near shading and then there's a model that you use, right? And so we've all taken for granted, the PV system is the model engine, right? You do the near shading in PV system or somewhere else and you can import that to PV system. And we're really gonna focus on the assumptions that are going into this project, right? And so can you go more into those assumptions, those inputs that were bringing into these models, because I think that's where the core of the energy bankability discussion happened. - Yes, so if I could take one specific example of that is PV system comes with a built in database of pan and on D files. So pan files represent essentially the code behind solar modules and how they interact in the software. PV system models a solar module in the piece of software based off of the environmental conditions, how many modules per string you have, how many strings per inverter. The pan file definition process, as you know, it's very nuanced. The installation that you include, it sucks that we do it. However, day one at DnB, when I joined DnB, one of my mentors in the industry, Colleen O'Brien pulled me in with this other engineer that was starting at the same time and she said this database is not as good as we'd like it to be. So we take it upon ourselves, every energy model that DnB makes, they make their own pan and on D files, based off of data sheet values, based off of test reports that are available for specific modules. Now, you've been around long enough in the industry, know that module vendors will produce a data sheet on day one to represent what they expect, the project modules to produce. But then you get the actual data sheet from the vendor when you purchase the modules that have slightly different characteristics. And they could also have a report from a test at a facility where the modules they intended to be made, but then you bought modules from a different facility and that requires a different report. Well, so at DnB and Aspen largely follows the same procedures that I followed at DnB, at Aspen we make all of our own pan and on D files, based off of if we have just a data sheet, that's sufficient for us to move forward with an energy model. But we make generalized assumptions about LID, low light losses. What else is there? IAM losses, there are a number of different test reports that will consider. There are also some things that we will not consider. So if we see a test report come back for IAM losses, that shows that at a 90 degree incident angle, you have a 102% current value coming out. Something wrong. You're saying that the sunlight hitting the module and you're getting a boost to the expected energy, which is a little bit odd. So the independent engineer community brought a lot of skepticism into the use of low light test data or IAM test reports when we started to see these anomalies come through. And there's no real way, nobody was really tracking the uncertainty in the measurement of those data points. So the independent engineers that I've worked with have taken things like the IAM test reports and said we just won't rely on that. However, we've found that this is reliable. These data points are reliable, which is, and even during my time at DNV, we switched from the ASHRAE model in PVSIS to the Frenel model PVSIS, depending on an anti-reflective coding on the module or not. So these are the details that are needed to produce bankable energy models that the independent engineer community is now confident in proceeding and presenting to the banks. It's very nuanced. It's very nuanced. And all this gets baked into the pan files. So if you read a PVSIS report, you don't really know what exactly is included in that pan file. So when I was at DNV, we'd produce the PVSIS reports, but we'd ask for whoever our customer was that was producing their own PVSIS reports if they wanted us to review their PVSIS reports. We would ask for the whole project file. So that we could open the pan files and the O&D files and see how intricate they were, or how intricately and detailed they were kind of pulling the levers and changing things behind the scenes. As you know, we did, during your time at Aspen, we accepted some LID test reports for a particular module, but we found that the developer had modified the pan file in other ways that resulted in an artificially higher module performance in PVSIS. And to kind of identify where the rubber hits the road here, when we went to fund the project, we had not performed that review of the pan file. We kind of just accepted the pan file with these LID test reports. And the results of that verified that in the pan file but didn't verify every little other detail. Aspen took a 1.5% hit on those two projects because when the independent engineer looked at it, they identified that, oh, by the way, your low light data was changed and your IAM was changed, you're not following your own procedures. And so there was some crawling back and saying, okay, we need to update our procedures now to verify every little detail, which is why we make our own pan files in the first place so that we can control all of that. And I want the ability, this is what I do. This is why I produce these energy and modeling procedures that are externally facing to our code developers is my story to them now is, if I accept, if we are meeting halfway on LID test results, I then have to take that half percent boost and I have to justify it to the banks. And if I can't do that with the data that I have, then we can ask when takes the hit but you've been given your developer fee. So my goal and my job is really to communicate that friction back upstream to the earlier stage developers. And it's tough because they want to crank more megawatt hours out of every project and I'm saying they can't get that when it comes to justifying it to the banks. And this is a perfect example of an industry where everybody is incentivized to show nicer numbers up front. But somebody eventually is going to be holding this asset and it's going to have to re-up your, like the re-update. baseline what those expected yields are to match real actual yields of the system that is being produced. Right. And so the devil, especially in pan-on-d files are in the details with here. The devil's are in the details with pan-on-d files. And you said to create those, it's really finding the data sheet and then any external validate studies for that specific pan-on-d file for that specific module or inverter. Right. And then embedding that in the pan-on-d file, that's the gold standard, which is what any i.e. would do and any good developer that is attempting their best to represent the most fair p50 case for these modules and inverters for that system. Yeah. So moving on from modules and inverters, pan-on-d files to the rest of the assumed losses, just take like a bird's eye view, what are the biggest losses that matter that most developers get wrong, especially let's harken back to your experience at DNV where you were viewing most of all birds or ports, you're seeing the inconsistencies between what they were assuming and what an i.e. would assume. Well, the largest is snow and soiling. I think that's the largest contributor. I think the industry's kind of come upon, this is a phrase that I joke that I'm going to trademark someday, but snowling losses. Snow and soiling together. I'm just trying to bring back the snails. Oh, yeah. Keep the snails trails away, but I'll talk to the snowling as long as I can. But a good example is early on when Aspen was just eight people, when I came on to join Aspen, the founders at Aspen said, Logan was such a pain in our butt being an i.e. at DNV. We wanted to come in house and be a pain in the butt to all of our customers and people were buying projects from. So bring Logan in, he's going to run what he does on the diligence level and review projects and he's going to run energy models. We're doing a project in Copenhagen, New York, where there were two different snow stations, no national oceanographic administration, forgetting the other A there, but no, no, no, A, another acronym. I think it's atmospheric and atmospheric administration, ocean atmosphere. It's snow loss models that the towns and the, the town's and model for snow losses and the Kimber Mitchell model for, uh, soiling, dust, soiling in particular are very sensitive to, uh, the inputs as we are discussing now how much rainfall, how much snow, how long is the snow lasting on the ground, the angle, all of that. Exactly. Yeah. And the buildup of the soil, soiling and what is what a rainfall event constitutes as a cleaning event versus a delayed soiling event versus less than an inches, basically no change to what, what the existing soiling conditions are on the module. But the sensitivity of the snow model to elevation or, let me, let me track that back. So it's really the sensitivity of snowfall to elevation is huge. So we're looking at this project in Copenhagen, New York, where we were building a project, our project was being built up on a bluff, but the closest know a station with fully reliable data. That's, I think they have three different classifications of fully reliable, kind of questionable and some are just not reliable because of data, data that's missing. They've had the backfill it with nearby stations. They've had to interpolate to get the monthly data inputs into, uh, into the snow loss model. And what we found was Copenhagen, New York, the town was at a much lower elevation, probably five to 800 feet or five to 100 meters lower than the project site. And because of that, we went from our very first models, use the Copenhagen know a station and resulted, I think in an annual eight percent snow and soiling loss. Yep. Remember, this is New York, so there's a lot of snow regardless of if you're a high elevation versus low elevation. Correct. So this was like 12 to 14% in December, January, February months of snow losses. We were using the closest station. We were following the, you're doing everything right. We're doing everything right. And when we took it to DNV to look at it, they had experts look at the surrounding know a stations and said, well, although, although that station is technically the highest quality data that you could get, the elevation difference. It's substantial. It causes a substantial difference in the winter time. And in New York, they get enough rainfall to wear. Soiling losses are going to be no more than 2% per month. I think we modeled that 1% per month. Yes. There's enough rainfall to wash them off. No seven northeases. Yeah, 1%. Correct. For the timber. Yeah. So snow losses are the largest driver of, of annual influence on, on energy model in this particular location. So then DNV at the time was our independent engineer. They said you should use this station. So we went from 12 to 14% monthly losses in winter to 35 to 40% monthly losses in winter. A substantial rate. Up to like a 13 to 15% annual loss in energy because of the difference. So we increased 5 to 7% on our, or we decreased our yield, 5 to 7%. Simply because of a 5 to 800 meter elevation difference in two stations in the inputs. And what's, what I think this gets to is we can, as a developer, try to do everything right. And then there's this one little detail that we forgot, big detail, which is elevation, right? And we're talking these numbers, which are, while we're taking this much more, soiling and this much more off of the production. But that is actually more accurate and should be done because when you are, then owning that asset, it's not like in the Catskill Mountains of New York. There's going to be less snow. There will be snow. There actually will be more snow. So it is, it behooves us to do this as early in that process as possible. Yep. Yeah, absolutely. And some of the creative things that I, being in the industry for almost 20 years now, some of the creative ideas that I love hearing because I tend to be more industrious and not as creative. I'm learning to become more creative, but aren't we all trying to? You know, when I was at DNB, I had the pleasure of working with 10 towns and who is the town end of the towns and it's no lost model. It'd be great to have a talk with him, Rocco, to have him share the story of how he actually developed the snow loss model. I hope you can connect this so that we can have that conversation. Yes, I'd love to, because that would be a who have a conversation together about this. But I was talking with him about the potential for snow clearing. So a way to reduce those 30% January, February, March and monthly losses to be able to go out and clear all the snow off the modules and get back to zero or 1% on a daily basis. Obviously the snow will build up. And this was largely before trackers were largely implemented. The developer, the developer IPP that we were working with every two years would come to us with a plan here. We plan on doing this for snow clearing. It's not just a matter of removing the snow from the modules. You have to move the snow from the, you have to put that snow somewhere. And who's going to go be removing the snow from the modules? Are you going to hire highly skilled laborers that are going to understand the nuances of if you nick that wire, if you hit that piece of equipment, if you're, how do you remove this snow from the modules while still protecting the structural integrity and not scratching the module surface? And so we consistently just brought up, we again, as an independent engineer, we were very skeptical and we tried to find the holes in this created plan. And this developer every two years would bring us a plan and we would shut it down. They'd bring us a plan and they would, and we would shut it down. When I say shut it down, I just said it really means we can't justify adjusting the snow loss model based off of the based off of your plan to keep the snow off the module. You need an ample amount of evidence or at least a logical consistency to the plan to be able to modify gray standard. Right. And the whole thing of this conversation is bankability and independent engineer, they would bring us this plan. We would then have to justify to the banks why we think that this plan would result in a consistent enough reduction in the snow losses to justify those increased yields. And so we would also review the O and M plan and the qualifications of the laborers going out and clearing the snow. And what do you do in a P P 50 or a P 90 event or even a 10% likelihood events of snow that you just can't put anywhere? Yeah, well, if you're in the Northeast, anybody, this is 2026. So we just had a gigantic snowfall and that snow was about a foot, could have been a foot and a half and other two foot in other areas. And that didn't clear for six weeks, right? And that's just where I was. If you're on top of or on a hill and you have a lot of snow and you move that somewhere, that snow physically has a volume to it. It's not like you can just melt it and it goes away. We're talking in the winter where you're not melting the snow. You're putting it somewhere else. You can only do that so many times before there's nowhere to put the snow where you need to get dump trucks and then is the price of the dump trucks and all the labor worth the swiss of getting extra production out of the system. Absolutely. And again, being bored and raised in San Diego, when I learned in big snow events in Boston that they have to put snow on the dump trucks and into parks. It's wild. It just blew my mind. I'm like, hey, you got to actually find a place to send the snow and ice. It's not all white and fluffy. It is. and icy and brown and it's ugly and it's not as simple as San Diego where oh you didn't have rain for a few months you can just wash the panels off and get the dust off you actually have to physically move this no exactly so exactly so following down we're we're talking the key factors kind of the higher impact was in a PV system model if you follow down the the waterfall diagram in a in any energy model output PV system waterfall diagram at the end so no installing is a big one especially in if you take a step up actually in that waterfall diagram you have your your near shading losses yep that's just making sure you get the instructions the tree heights the distance from the trees your inter row distances all that has to be right correct for that to manifest yeah and I know all of these conversations have an impact on how you actually test the system when you turn it on so getting the near shading model correct and accurate as accurate as accurate as you can with accurate tree heights tree line setbacks sometimes don't let you clear any trees so you have to model 60 foot trees within 30 feet of the array that really they killed production but they may not kill the deal and that's something that my mentor at at Aspen Power is he's really taught me Scott Delaney has taught me that just because the yields get reduced or you have severe losses associated with trees it doesn't mean the deals going away well you're so because you're putting panels that close to the trees you're actually getting an uptick and gross power of the system you're just getting a downtick and normalized power of the system right so if I say it in a different way you're making a choice the project is making a choice to increase capacity into shaded areas mm-hmm at the expense of yield but you're getting a benefit of increase megawatt hours throughout the year yep because you are investing more by putting modules in poor producing locations so that you can actually increase the overall energy output of the project yields go down but megawatt hours go up mm-hmm because your system capacity increases also just because yield goes down by 5% doesn't mean the project returns go down by 5% you may it and this is something I've had to learn in being an IPP and working with the developer so closely is engineers aren't the only decision makers when they come to moving a project as much as we wish we were what we want to it's it's it's follow the money if we if we take a 5% hit on yield but our IRR is only only reduced by a half percent yep then well because you're already the labor is there you already have the boots on the ground you're hiring the EPC it might not be that much cost to add an extra few hundred panels and a racking there to get an increase in gross megawatt hours because that's only a little bit extra on the on the estimating side or the actual spend at the system side even and then you get that gain for the lifetime of the asset where you're getting more megawatts for the lifetime yeah yeah totally agree so again that step up in the the P.B.S.S. lost diagram to your near field shading model something that is that that I'm particularly sensitive to given that I've always worked in distributed generation um meaning smaller parcels more wetlands less civil work we're not we're not going down with masquerading sites to try to maximize or minimize terrain losses which we'll get to we're trying to squeeze all the modules that we can into tight areas in the to connect on the distribution grid large scale 100 megawatt plus solar plants don't have the same sensitivity to shading as the smaller scale ones do because if you think about it you take a thousand acres and you install solar solar solar modules all over that thousand acres it's really just the edges of the arrays that are affected by tall trees on the stones however if you have the same the same trees that are as close to the array for a five megawatt facility you have a much higher impact your window of clear sky no shading gets minimized when when the trees are a hundred yards apart from each other you only have a hundred yard corridor to install all of your modules in um so modeling for the the trees nearby objects uh rotor row shading and modeling terrain that's one of the more complex things that we have to try to model is the terrain so um you know there are a few different software tools out there to model terrain losses uh it's it's becoming I think there's some tools in the work in the works to take out to to be able to map modules to terrain accurately and then do a 3d model representation that's you and I've talked about this the best way right now that most developers are using or that are using PV case to model with the terrain right and then they're taking that and they're importing that into PV systems to do the actual energy estimates right but in PV systems correct me from wrong but I believe they don't actually simulate it in full 3D they actually take an estimate of what those terrain losses are going to be and then flatten that array and simulate it in 2d correct could you go into that please so that like to me when I learned that I was very confused because we spent all this time trying to model the terrain just to stick a 1.9 percent extra loss right so we're understanding of this and and you can you probably know more about this than I do from your your recent research but the way you described it is is accurate to my understanding is you take PV case provides a 3d mapping of the terrain at PV case doesn't it amazing job where you have the option of you can either take this terrain that they pull in from US GIS data or from Google data and they they map this 3d space this this terrain net for lack of a better term a mesh that's the term a 3d mesh and in PV case you have the option of choosing to map to that mesh so fixed tilt tables can map to that mesh for fixed tilt systems or you can establish a maximum height of your modules off off of the ground and PV case will adjust the tables at a it they will adjust the elevation of the tables to meet your minimum and maximum height requirements off of that that mesh for trackers they will try to match the tracker talk to slopes to that mesh again based on the constraints that you're putting in again and as I mean as I start to build this in my head you can see all the little tiny in just so many have that have a dramatic difference in a dramatic effect on the overall yields and the broder rotating is a big problem if you don't map it correctly yep very easy if there's no terrain that there's no hills but if you have a hilly site right it could be you actually get more production say you're on a south facing slope of this hill and you can actually squeeze them closer together I have less road or row spacing but because there are different elevations say for a fixed tilt system you might actually have more yield than if it was flat but it had now say for a north facing slope you might have significantly less yield correct those flat right yeah so so PV case outputs a file a PVC file then we that is imported into PV cyst and what PV cyst does is it kind of pulls out that shade model and runs an hourly simulation and calculates the percentage base it's so when you pull in you can define in the PV case you can define in the PV cyst the stringing ends with with the split cell modules we had to make a change about 10 years ago to where one module is really two modules yes because of the half cut model correct split yeah yeah so you have two circuits in each solar module now seven three with the diodes you have six yeah well two circuits and the three essentially six circuits inside of each module whereas before you'd only have three with the diodes or three in the the full cut cell modules but what PV cyst is doing is it's estimating just the shading percentage and then it it basically maps that shading percentage at every hour of the year and then does a 2d representation of that in hitting the energy model or the available resource hitting the solar cells what's in the works and where where the industry is going next is the ability to do a 3d model of that without the the simplify without the simplified 2d version it's going to a 3d version yeah I'm pretty sure plant predict and solar farmer the only two models in the market that do it actual full 3d simulation yeah no so wonder if PV cyst is actually working on it by all I imagine they are I hope that they are I hope they are too yeah so that's kind of where we're at with terrain modeling and ensuring that you're getting your rotor row losses accurate now a caveat to rotor row losses as the solar industry has grown and we're going to go back to snow losses here rotor a losses typically hit you the most in the middle of winter now if you're at a 35 to 40% monthly loss due to snow it's not going to impact you because the system I shouldn't say the rotor row losses are going to be minimized because you're going to have snow cover on the modules whereas if so a developer can look at that and say I'll take that rotor row hit because we're going to have the so the snow losses at the same time and the lowest the your your your GHI is much lower anyway we want to match production in the summer. So we may take a road-to-row hit because everything's covered in snow. Plus, you have a lot more ambient light, diffuse light in the winter, typically because you have gray skies. There aren't a hardshade lines in a fully cloudy day where all of the light is diffused. You don't have any direct beam hitting the module. So a developer really has to play with the ground coverage ratio to figure out what's best for them and best for project research. That's what's beautiful about solar development is that there are so many little things you can tweak to try to fit the project to what you want it to be, that there is no optimal project. For any one person what is optimal may be different for the next person or a developer. So we went through snow and solar losses, very big obviously modules and inverters, terrain losses, what would be the next and obviously all near shading, which includes trees, obstructions on roofs, nearby roofs, right? What would be the next biggest one would you say after those four? So those are the big ones. Those are the big ones. You get those right, you're like how much percent of the way there? You're 98 percent of the way. All right. So now we get into the fun parts of module and string mismatch and module quality. When Logan says fun, we mean like fun for nerds and engineers and not approximately fun for other people that want to to bottle this. Yeah, so because it's quite technical, ridiculously technical. And when we say technical, it requires quite a bit of time to investigate and back up and provide the justification for the values that go into these, the module quality factor and mismatch and how it's run. And LID as well. And LID. And so about. Well, some are released in a module. Depending on how it's run in which software you're using. So as some backgrounds, DNB would run independent energy models, meaning from ground up, DNB would build the energy model based off of the plans that are available for the project. At the time, we would also review a developer's PD system to see the difference. Well, we would not, we would produce our own as the bankable model that the IE standing by as their own independent model. But as, you know, if if one of our customers at DNB and we do this at Aspen as well, is we say, look, we have 20 projects coming through. We want you to do an independent model for five. And then we want you and then we want you to review what we do for 15. So there's that, that like would set off alarms alarms in my head if I didn't know when I was at DNB, if I didn't know what their procedures were, the biggest question I always got, or that I always had back to my, the developer customers that we were working with at DNB. And that I give now upstream to the developers that we purchased projects from now that I'm at Aspen, tell me what goes into your modular quality factor and tell me what goes into your mismatch losses. Those two things can be interchanged and different people, different companies use those two, those two terms or inputs slightly different. Well, they're kind of a catch off for all of your ambiguous losses. Correct. Yeah. So this is where you capture things like nameplate losses, where you deal with the tolerance, the zero to plus five watt, nvening class, where you deal with the uncertainty in the measurements of the power output of the modules, where you deal with sub hourly losses that come in because of the limitations of PVs as to being hourly only, although I think that they're already making a change to be able to do sub hourly modeling in PVs as well. I'm pretty sure solar, farmer and plant predict also do sub hourly losses. Like say, when we say sub hourly, an 870 is one hour for every hour of the year, 8,760 hours per year, then you also have 15 minute models. You can even do five minute models. Some provider are even giving one minute models, though I am dubious of the certainty of a P50 case with a one minute model. So that's what we're talking about when we talk about sub hourly losses is capturing the losses in between that single hour that we're running that simulation. Right. So in that module quality factor, the way that we model things is you're accounting for the nameplate bidding of the modules. Now it's all zero to plus five watts, but when I started, it was some of it was plus or minus 3%. So it's negative five to positive five. I haven't seen any of that. That's wild. It's it. I mean, you remember looking at that pan files that I would bring into Aspen of older modules. You say, I remember you're bringing this up like this. This one's way different than anything I've seen. I said, yeah, because the data sheet says minus five to plus five watts. That was wild. Yes. And I'm glad that the industry has at least standardized on a zero to plus five watt bidding class. So the way that the way that the independent engineer community looks at that is, and we can get into, again, bankable, how do we bank our modules as well with equipment, but the way that the independent engineer community looks at that is typically utility scale projects are having a 100 plus megawatt project is hiring a company like CEA InterTech now to go and perform a factory site visit of their actual modules being produced over in Asia. It's something very verifying that the modules themselves are being produced to the standard by which the module the manufacturer is saying. Correct. And you can afford that when you are making the utility scale project to your purchasing of a gigawatt of modules over five years, you have the purchasing power to hire somebody to go and inspect the actual modules that are going to your project. But our DG developers typically don't have that purchasing power. Yeah. So if we could take a slight tangent to years to talk about module purchases in the DG community, the way I always prided myself on trying to be as commercially as commercially minded as possible in closing DG deals when I was at DNB, understanding things like frost heave ramifications for projects that are already in the ground. To tell a developer to go or a contractor to go rip a project out because they didn't account for an additional two feet of depth needed to resist frost heave forces is not a commercially viable opinion to go say go rip it all out. You've already built this project now rebuild it. Right. So while the industry started to learn about frost heave and how to properly account for it, how to mitigate it with like frost sleeves or backfilling with non-frocessed up toable materials or foam, we still add projects that were in the ground that needed to have funding close. So when I was at DNB, we developed a monitoring and a monitoring and mitigation plan that we would put into our IE reports that included surveying the project at its baseline, which is its installed condition that mechanical completion or substantial completion, and including in the O&M agreement to update that on an annual basis. I won't get into too much detail, but that seems that works to get it through financing. And we had developers, we had vendors that swore that there was no frost heave on any of their projects, hundreds of megawatts installed in the Northeast. And on the flip side, DNB was also hired by a dozen different IPPs to investigate frost heave from those same developers and those same vendors. So we knew it was a risk, we knew it was a problem. And what is frost heave? You went right into frost heave that's to find it just for a moment. So when there are frost susceptible soils, which include fatty clays, similar to expansive soils that get saturated with the spring melt and freeze the cycles that happen in anywhere in the Northeast where it snows, you have this estimated frost depth and you have what's called an ad freeze bond strength. So as the soil around the meta, and we all know this, people who park their cars outside, the ground will be perfectly dry, but your car is wet in the morning. Because water tends to condense onto a metal surface, not onto a stone surface. So all the moisture that's in the soil condenses onto the metal piles that are in the ground. And then that water forms ice lenses that causes the water to expand. And then that tops those metal things out like a pimple. So if you have frost susceptible soils, there's pressure that builds up inside of all the voids that are filled with water that now you have these ice these ice lenses that are expanding. And there's only one place to go. And that's up. Yep. So we saw occurrences when I was at DnV, we saw occurrences of 18 to 24 inches of frost tracking occurring on projects that I'm not joking. We saw pictures at mechanical completion as we were in the in-appen engine here. At mechanical completion, the table perfectly flat. We go out two years later and we see we see foundations jacked up through the back of modules and it just looks like an undulating terrain. It blew my mind the difference between the same picture taken at mechanical completion versus two years of terrible linters. And I say that because it's all building into the bankability conversation. Well, yeah, this is an equipment selection part of the bankability conversation. And the specific design. So you have to take the equipment that's all listed for its purpose, assemblies, individual components. But then you have to have to put that somewhere and understand what the environmental loads are going to be on that particular system through the 25 to 40 year useful life of the project. And that's really going back to our first topic is you could have great manufacturers, great equipment, but then you have something like Ross you've come in. And you did not put something to a mitigation strategy in place. And even three years into that, two years into that project, it looks nothing like what you expected it to, right? Right. So to tie it back, I think I started to talk about my goal in providing good customer service was to find commercially reasonable solutions, rather than just putting up a brick wall as an independent engineer and saying, no, you must do it this way or else. If only all developers had someone as reasonable as that, as their independent engineer representative of choice. Well, I mean, the proofs in the putting my customers kept coming back and wanting me to do the work. So I think that that is really the only way that I could judge. And I did not get access to any of those projects operating data after the fact. In terms of how the project was performing relative to my energy models. And they weren't just mine. They were DNBs. They were the people that were reviewing and approving my energy models at DNV. And that's something that we do at Aspen Power. We have a peer review process. We have an approval process as well that only highly qualified experienced energy modelers perform. So anytime an energy model goes to our project finance team, it's already been peer reviewed and very scrutinized down to all the details that we're talking to. And for a time, I was providing all the approvals. And you know this 50% of the projects that would come from me to me for approval, I would reject and say do this. And this mimics exactly what DNV would right? Because you have that the IE would have somebody that makes the model that you validate the about trust, but verify you're verifying it. And then you then make that iteratively better. Right. And you can point out me because most of those rejections were my models at the time. So it's a humbling process. I guess that's a very humbling process going from Barrego at the time learning how to do PV cyst through BWs training. And then I joined DNV and I thought it was an expert now. I was a general. Exactly. So kind of drawing back to this module quality factor in this match. At DNV when I would review a developer's energy model, just their PV cyst report. I would ask for the PV cyst file so I could verify the contents of the Pananone D files. But the largest question I would ask was always tell me what goes into your module quality factor what goes into the mismatch. Because you have things if you have undulating terrain that is modeled as a 2D flat surface. You're going to need a digital awsome match. Yeah. If you have a module pointing at five, just a five degree difference in in it's asmet third, it's incident angle. There's some inherent mismatch there. And PV cyst has the ability to do mismatch among electrically connected components. But PV cyst does not have the ability to calculate mismatch. Say if you have a string of 24 modules across three tables of eight each. And the middle one is pointed straight up. And the two on the side are pointed slightly east and slightly west. You have a mismatch across how much what the incident angle is for the incoming irradiance. So PV cyst does not have the ability to model that. And that's something that's really cool about some more farmers that they have every module they take like 36 points. And then they actually string it all together in their 3D scene. And then they do all those calculations in real 3D, which is why it takes forever for that simulation to run. But it's really cool that you can actually model that now. Yeah. So this bankable energy theme that we're talking about, that's kind of where the industry is going is being able to calculate and quantify what that mismatch loss is without taking an overly punitive approach. Well, and to be clear, what is currently considered bankable is a PV cyst approach, which is not even doing that additional 3D validation. Right. Right. And we're all trying to become more accurate because I think you know, it benefits the industry to have systems producing what they're expected to produce. Yeah. It benefits, no, it actually hurts the industry when you have overly aggressive developers, overly aggressive vendors like module suppliers that or resources, whether resources, solid resources. Correct. Yeah. The resource providers, it really hurts the industry if we have projects that are not performing as they should. There are companies that are that have operating fleets that are performing very well, but across the industry, we have a shortfall where the systems are not producing what the developers and the IPPs have said they were when the project originally went to to financing now. And the worst part is that the incentives typically are all there for people to make this as optimistic as possible. Right. Instead of taking a probability 50 case or saying it's a probability 50 case, but really it's a P 25 P 10 case where it's very optimistic. And the incentives are there for that part of the reason for this conversation is so that if people wanted to design a truly P 50 case, hopefully they can listen to this and have better insight into how to do that than before. Yeah. Right. And so we're talking about putting all of the extraneous losses that may not be captured in a perfect third pillar of the model itself into that module quality loss factor or into the mismatched losses. And then having that baked into that energy model as that right down. Right. So you can add it there and you're heavily scrutinizing that as naive and ask putting your IE hat on at aspect. Right. And then after that, right, because that is very we could talk about that for hours. Right. What goes into that? Those specific just details. I think can you give a generally held wisdom approach like say, if somebody was going at the system and they wanted to do this reasonably well with minimal time, what would you ask them to do? That's a loaded question because I'm an engineer. I want all the details. All right. All that will say we'll plead the fifth here and not give a very goes. I mean, it was deep as you have time for that you go to PV Watts and run a very simple calculation. They installed DC capacity, AC capacity and generalized DC and AC losses basically go to solar resource compass. We get the sub hourly losses. Right. They could go to solar resource compass and get the if they pay for the well, to get the sub hourly losses, you have to pay for solar resource compass. But solar resource compass will also give you a yield number with all your monthly output for the inputs and the outputs needed for a financial model. It depends on which stage you're at because you asked for a general overview kind of when you're getting the first touch on a project. And you and I would go back and forth when we would now I'm always first touch and you're always last touch. So correct. Yeah. And a big bounce. And my response was always where we can be precise we should be. If we have the information available to do it precisely now, let's do it now. Yep. It'll save us effort in the future. It'll help close that gap that's typically a five to eight percent difference in energy yield between what the developers initial model was and what the IPP's final number was or what the final bankable i.e. model is it's typically a five to eight percent difference is what we've seen. And to put it in perspective, the final bankable, like the bankable number is the i.e. best representation of what the size actually can produce. Correct. So without any further assessment, we could try to say that as close to a perfect p50 as possible. Correct. That's the aim at least. Yes. And what I think the developers and the IPPs are trying to avoid is what they call it haircut. iIs are always accused. Oh, the iIs just take a haircut. No. A haircut means I'm at 100 and the iIs saying 95, which is a 5 percent hair. Yes. So the mv just takes a haircut across the developers production number. No, it's it is a it's a scientific and an engineering process to go through. And it may be 5 percent difference. It doesn't mean the iIs taking a haircut or it doesn't mean that the banks are taking a haircut. It just means that there's a difference between what the developer how the developer modeled it and how the iE modeled it. And it's not iIs are not out there just saying oh, take 1 percent hit no matter what. No, they're doing their models and then saying hey, based off of our models, our models are different than your models by 1, 5, how many percent? Correct. And then they're saying, well, we believe this and the banks believe them. So the accessible capital is based off of the iIs number and not the developer's number. Now the developer is well in the right to keep trying to use that number for their asset management team and to see if that actually hits and if the developers right, great. They have more profit from the project, right? But you still don't have access to that accessible capital, which is really the whole purpose. Well, one of the two purposes of bankability, right? The first is to get the accessible capital, but then the second is to have your asset management team judged and projections in accord with what is actually real and reasonable for those projects. Correct. Those are the two real, crap if i'm wrong, but that's what i'd say there's like a, i hate to say there's a fourth leg in this stool because i don't like four leg stools. It's unstable in in mesh environments, 3d mesh environments. You're engineering brain. But it also and i alluded to this when we talked about the near shading scene, the importance of getting that near shading scene accurate in your simulation is that when you go to run a capacity test, what is front shade beam losses? Yeah, so your upshade beam losses are whenever that is what is it? When the factor is less than one. One means no shading. There is no shading on the entire system. There's no shading from the sun, say you have a tree next to your system, and the sun hits that tree, and then it only hits 50% of the panels, Fshabie will be 0.5 or 50. If it's 30% of the modules, it would be 0.7. Yes, correct. So the capacity test procedures that are accepted by the banks and included in EPC contracts as a performance test, you have to filter out those periods where Fshabie is less than 0.99 really. Yes, there's 1% buffer in there. And so this is where timelines and EPC contract payments and everything comes into this bankable model of looking at a project that's bankable, are you going to be able to close this project in March? Yep. Will. Fshabie, during winter, could be for these systems in the Northeast. Never. Never won. And we did this. I was working on a project in Massachusetts. This was, I hate to say it, but 12 years ago, 12 years ago, it doesn't feel like I've been in the industry for that long, but I looked at this site. So before PVSIST had the ability to do PVSIST 4, I've already said there were ghosts. One of the more frustrating aspects of it was their near field shading model interface was atrocious. It was no. I said, people think it's bad now. They made it. They made a fast improvement. And I love it. And I use PVSIST every day at work. And I'm so thankful for all user interface improvements that they've had over the past decade. But they started out where you would define a table. And then you would have to, you could not grab the table and move it with your mouse to a particular location. You had to enter in coordinates of a table for each table that you made. And then you had to do the same thing for trees. So at DNV, we would do a stopgap measure in Google SketchUp, where we build everything in Google SketchUp, we would then go to the Equinoxes and the Solstice, the summer, the winter solstice, and one of the Equinoxes. And we would click hour by hour throughout the day and count how many module strings were affected by this shading. We'd plug those numbers into three different columns for the two solstices and the Equinox. Then we would extrapolate an annual shade loss, a percentage shade loss on an annual basis. So this is not accurate. This is on an annual basis. We would say 8% shade losses. And then we'd go into PVSIST and manually adjust the fins on the horizon losses to match that 8% shade loss. Oh my god. So totally inaccurate, but it's somewhat worked. And that was considered bankable at the time because there was a dearth of better ways to do it. And as industry progressed, the goalpost has reasonably moved for what is concerned bankability with better and more realistic estimates and better models. Right. Yes. To tie that back to the performance model or the capacity test, when we went to test this same system, when the EPC went to test the same system, there was never a day where the F-shade bean factor was one. Because of the, like we had south side shading, we had north side shading, we had east and west side shading, it just had to do with the particular nuances again. In the DG market, we're trying to smash as many modules as we can into super tight spaces, which is when you have wetlands where you can't clear trees, you have property lines that are very close and tight. The developers and the IPPs are making a choice to increase capacity in favor of megawatt hours, but the yield drops. So the efficiency of the system reduces. And this is an example. So what is amazing about PPSIST is that once you have that 3D model built, you can then simulate every hour of the day and see the gray hardshade lines and you can see the yellow affected strings inside of PPSIST. So again, to be commercially reasonable in trying to close projects and help make them bankable, I asked for drone footage at 9am, noon, and 3pm. To see the shade. On a particular date. And what I did was I said, okay, let's take these real world drone photos on a clear sky day and compare them to how the PVSIST model shows the shading should be. I wrote that up in the report, put the pictures in the report in the bank, accepted it. We're having difficulty passing the capacity test. Or the performance test because I think we ran a 24 hour performance test for this project. Much more qualitative, much like a PR test. Yeah, not a five day or a real question. No regression analysis. It's very qualitative. And my language in the report was, although we don't have a qualified performance test passing result, what we have confirmed is that the shade model characteristics on this date are consistent with what the inputs to the model are. Therefore, we're confident that long-term, the energy model is accurate to the installed conditions. They closed. They're making the project. Yeah, they're making the project. Yeah, so they're commercially reasonable ways of moving beyond just this brick wall closed-in environment that IE sometimes get criticized for. But there has to be a thought out. There's the most large-seger methodology. You have to be a scientist. You have to put on your science hat and actually do that. Correct. So to kind of close out this big contributors to the PV system, loss tree, where there are differences. I mean, AC and DCO maclosses are calc based off of wire size and distances. Yeah, equipment selection and locations. Those are as trivial as it gets even though they're difficult to produce. Correct. Time intensive, not difficult to produce. Transformer losses both resistive and constant. And load or no load. The load or no load. And then what's challenging sometimes in that model of reviewing a developer's PV system model is where does availability come in? Where does degradation come in? Availability and degradation. Big ones. Yeah, so recently when I first started at Aspen, we took a 99% availability for fixed tilt systems and a 98.5% availability for tracker systems on annual energy. Period. Period. That's three to four days of downtime every year is expected to be built in for the model. The IE's have now done much more research to make it a more robust model to where it accounts for teething issues. So you take an additional one to one and a half percent hit in the first two years, I believe it is. But then you're rewarded at year three of operation with only a half a percent hitter. It's like, yeah, so it's a 99.5% availability for fixed tilt, 99% for trackers. But then as the inverter warranty start to expire and most the G projects are built with string inverter or central and small central inverters these days. Whatever, everybody calls string inverters. Well, we're not connecting string inverters to a string, to an SMA or a chin or a some chin models. Anyways, it gets complicated. But what colloquially is held is string and verger, small central and verger. Correct. So as the warranties expire, those inverters start to fail based off of the bathtub curve, essentially for any electronic device that's out operating in the wild. Especially being constantly exposed to sod and heat and winter and snow and all the elements. Yeah. So from years eight to 15 as those inverters expire, they will be replaced. The expectation is that all of the inverters will be replaced by year 15. The shorter age ones will be replaced starting in year eight. So you take an availability hit in years eight through 15 to replace those inverters. But then you get that back. You get it back from years 15 to 26. And then from year 26 onward happens again. Yeah. But then as we discuss uncertainties, this this 25 to 40 year life, like the uncertainty of well, how many inverters are going to start to fail at year 26. You now have the issue of you know, mismatched module connectors or even matched module connectors not being mated properly that build into the unseen degradation of systems that asset managers kind of throw out their hands and they're we know that's why the degradation is a consistent loss. And even though most module degradations follow a logarithmic curve to us, a quasi asymptote around 80% or 75%. We assume a consistent loss throughout all that to your.64. A little bit. Yeah. Based off of D&B studies. Yeah. And so you have that baked into the model. That is something that is typically outside the 8760 because you're looking at that from year two onward, right? Not your one. And you're one, you have almost no unavailability, especially for the capacity test. And then you add that unavailability on for the actual numbers. And so that's part of the bank ability of the lifetime of the energy simulation. Not just the first years 8760 p50. Dumbers. So we talked a lot about a little bit about equipment selection, a lot about energy model. Yeah. What I, there are a couple other aspects I just want to touch on. Please. That's why we're here. Equipment selection as it specifically relates to trackers. Logan has many opinions on trackers. Yeah. I'm somewhat proud that the ABL that I manage at Aspen is very short when it comes to trackers. I've seen too many trackers. not operating as expected and not in a reliable way. - Yeah, so I remember there's many, I grew up in O&M and many a time there would be trap roles rolled to fix trackers. You'd have one not working, you'd have many not working. There's a litany of problems. I remember one week I would just go out and I would actually grease up the track there gears and I was listening to a book, I was listening to the account of money, Chris, at the time. I was just listening to that for two or three days straight as I had this grease gun looming up, the tracker bushings, right? And so there's a lot that has to do with trackers. - Yes, so hiring an independent engineer and reviewing a bankability study is very important when it comes to operating these assets longterm. Now, some IPPs are incentivized, much like developers incentivized to maximize profits. We're all incentivized to maximize IRR. - Mm-hmm. - I mean, it's a business. If we weren't doing that, the whole industry would not work. - Absolutely. - You need to make profit from these systems to actually build these systems or else no one will be building them. - Correct. And the challenge I have sometimes is making the balance between gold plating a project and installing a project to a sufficient enough quality to support all of the assumptions in the financial models, the O&M costs the downtime. All of that. And it's not unique to me or to Aspen or any particular company, it's industry-wide is how do we make that balance? You could gold plate something and install it for $6 a lot, which used to be the install cost back in like 15 years ago. - Yep. - Back when Silicon was a lot of-- - But that's not reproponatable. And you're not going to build a lot of those projects, right? - Correct. So how do you do things like wire management? Now we're talking about the actual implementation and construction of the project. So having high construction standards and inspections by qualified third parties really helps in supporting an argument with the banks that the project was built in a bankable manner. - And it also helps make sure that when you hand this off to asset management, that you're not giving them a crap sandwich that you're giving them something that is going to be hopefully as easy as possible to maintain for the lifetime of that asset. - Correct. - And so that's where the sleeves for the frosts even come into play, right? That's where having a tracker brand that is cost effective, but then also isn't going to break down and will be in business. I know there are many tracker brands that were in business 10 years ago that are not now and then when there's a warranty, they're not even a warranty, but a part that breaks, how do you even fix it when the brand's not in? - Yeah, so it exists. - Aspen owns a couple of projects, actually, owns about a dozen projects with tracker vendors that are no longer in business. We better go out on the third party market to try to find replacement parts and where we can't do that. We actually have some projects that are at risk of not meeting their minimum production guarantees. So we're actually looking at modifying the system to be very expensive fixed tilt systems just to keep them operational. - Yep. Instead of having a tracker be flat at stowed at zero, right, getting this on, it might be even better to change this to a fixed tilt system if the track lines aren't working. - Yeah, and stowing at zero is a very sensitive subject when you're in the same place. - Yeah, yeah, yeah, because especially with the wind issues with that. - Yeah, so and just to provide background for people that are listening is solar racking systems had gone from robust unistrat, square stock steel or even one frame on rooftops. So these are thick steel materials. You know how they behave, you know what they're modular are for all of how they're gonna behave under stressful conditions. And as the industry has progressed, we've had much more creative solutions come on board that resulted in lighter materials, more engineered solutions, much more elegant on paper solutions. But as we started to lighten all of the steel components or aluminum components that support the modules, you're changing the way that these structures behave dynamically. So when I started in the industry, almost all solar structures were considered to be rigid structures. You did not have to account for dynamic excitation from oscillating vortices or trailing rows being affected by the vortices that come off of the upwind side of the wind modeling. - What Logan is saying in engineering terms is when wind passes over structures, it creates waves and that waves can incite especially tracker structures which are movable to bend and contort in ways that are not desired. - Correct. And again, when I started the industry, in addition to considering them to be rigid structures, the only real ASCE, American Society of Civil Engineers, which is a code book that prescribes environmental loads onto building structures. The only way to really analyze PV systems was with a static wind load of a single row, as we discovered, which it worked at the time because we had rigid structures that were built with rigid materials that would somewhat overestimate. So you had factors of safety at every calculation along the way, somewhat overestimate the wind loads on a structure. As the structural components started to get lighter, we started to see that these structures are not rigid. They actually move in the wind and they are affected almost to a, so if you have the static wind load, which is, you have a design wind load with certain characteristics of wind hitting a tilted surface like a wing, you're gonna have uplift, you're gonna have overturning moment that all has been calculated by a licensed engineer. What we found though is if you put another row in front of that. - That causes the disturbance. - You not only have the static wind load, but now you have these oscillating vortices that cause what are called dynamic loads that can be upwards of 20% of the static load. So we used to design it to 100% of the static load. When we discovered, we actually need to do 120% of the static load based off of wind tunnel testing. And then we kind of locked that in for fixed tilt. So I was really fortunate to be invited by a plan reviewer at the Division of the State Architect in California to join a small group of industry experts to help write the new wind load code and seismic code. It was the C-OXO structural engineers association of California put together this panel of folks that got together and said, we ASCE7 can't accurately interpret the wind loads on PV systems that have like the sawtooth type structure for a rooftop and fixed tilt systems. We need to come up with a prescriptive method to describe the wind loads on these systems so that the vendors don't have to have a wind tunnel report to support the wind loads. Number one, and then number two, the seismic loads of ballasted systems. These are all bankability concerns for projects that ballasted systems move in the events of an earthquake. They scrape that the roofing material, you have to have certain offsets from existing other structures on the roof so that if this ballasted system is sliding around on the roof, it's not gonna hop and then hit another. - Something else. - Another piece of equipment. I was really fortunate to be invited to do that. That's how I met, to partake in that panel. That's how I met the folks that I eventually hired me at D&B was through that panel discussion in helping write the code. So I was really fortunate to be invited to go do that. And then to, we discussed fixed open now trackers as they have gotten lighter and they've gotten much more elegant in their design. The modules now are a much lighter component and they're easier to rotate. So now you get these. - They're like big fins on airplane that are able to be susceptible. - It's scary as hell, right? If you're sitting at the window seat on the wing and you see the wings flat out there, that's what they're supposed to do. And it's designed to do that, but it's scary 'cause you think that they should be rigid, but they were rigid, it wouldn't be able to take off. - No, there would be too heavy. But that's where, you know, as we, as the industry progressed and the IE started to investigate failures of these trackers, the wind tunnel studies, like, or the wind tunnel labs like RWDI and University of Ontario and CPP, they started to develop testing standards for trackers that allowed us to investigate things like torsional divergence, torsional galloping, oscillating portacies. There were four or five different failure mechanisms that they were seeing out in the field that they then developed a test standard for. And it's important that if you're considering a tracker for your project, that you are reviewing a bankability report from a qualified engineer based on a wind tunnel laboratory test result from a qualified wind tunnel lab. - And this is not just for single access trackers, even though it's very pertinent for single access. It's also for fixed tilt on ground mounts and on rooftops, it's just have good wind tunnel testing. And if you're fortunate enough to be able to have a vendor that does that prioritize that at-alt-falx, correct, would be your recommendation. - That's my recommendation. - Yeah. - And then so that happens for a racquet, that's a perfect example of something moving beyond just the simple PV system and energy model, but how to make sure that what you've built actually stands the test of time and by proxy is bankable and you can get financing for it. - Correct. And just to bolt on this last piece, And I'm certainly not an expert in this view. is do the questions, part of the bankability review, or a project level review by the banks, is a review of the operations of maintenance agreement. - Yeah, I'll have somebody on that stress that management that can talk to all about that. - Yeah, so to put it in a very high level nutshell and summarize it, the IE is going to apply on whether the budget and scope in the O&M agreements are reasonable to support the predictions in the energy model and the financial model. I think they most own emigrements assume one truck roll for preventative maintenance, a specific set of things, a task that need to be done during that truck roll. And then they authorize a corrective maintenance budget in the O&M agreements, but it only is used in the events that corrective maintenance is needed. So there's preventative maintenance that's done every year. And then corrective maintenance is, if something is found to be out of whack, the corrective maintenance is a budget to go fix it. You're assuming you don't have to fix anything in the preventative maintenance budget, but the corrective maintenance budget is there to fix things. - And if there's not enough budget there that the IE assumes that would be necessary for a project with this specific build material, right? So the RAC and components of modules, the infirters and the specific build configuration, then they will opine and say, hey, maybe you actually need more budget for this. Or there is additional quote unquote haircut to the system. So we talked about the projects where we received that we're hand filed and cost us a percent and a half at deal close time with the banks. Those same projects were using an unproven unproven unbankable tracker model that we really, these projects were being built. So we had to use that. - We had to use that vendor. We were not gonna change the blockchain. - Yeah, so at Aspen, we were looking for the narrative to tell about the project of how we're accounting for a non-bankable product. I shouldn't say non-bankable, a product that I had considered to be unproven in the industry. - One that is not a surefire, $6 per watt product. - Right, so well, yeah, correct. I'm not sure about the $6 per watt. - I mentioned so far as like, it is the gold plate. - The gold plate got you. - It's not a gold plate product. Largely because they didn't have a full bankability report from an IE, it the scope that they did have an IE look at wasn't robust enough to talk about torsional divergence. Some of the minor tracker failures that we've seen across the industry. So we took a creative approach and said, well, let's take, if this tracker is not, the performance of this tracker cannot justify our availability assumptions number one, let's just take another availability hit in our energy model. So we took an additional one and a half percent availability hit. The number still works for the project. Okay, well, let's triple, like which parts of the tracker are we ready to boast? Let's triple the amount of spare parts, keep them at the site. So we're gonna have to pay for that out of CapEx. And then we are likely gonna have more truck rolls. So let's increase our opEx budget by 20%. That all made the project work financially and it was bankable. So the project became bankable because we used money, we used capital in other areas. We reduced the energy model, which is the future revenue predictions for the project. And a caveat about tracker availability, it's not just downtime up time. It's also, it's a tracker. It's the tracker pointing where it's supposed to be every row, every hour and every day. If half the rows are pointed east and the sun's in the west, that's unavailability. Yeah. So we justified to the banks that we are reasonably accounting for this risk and mitigating it through increased on, increased on availability or decreased availability, increased spare parts and increased truck rolls. And the bank said, that's acceptable. We will fund this. I think that's an excellent point to bring up, which is you don't need this gold plated product correct to reach bankability, right? Every project, especially in the GG space, has its story to it. And if you properly account for the risks associated with that project in that story and build a plan to mitigate those risks, you can take something that would ordinarily be a crap sandwich and make it into a very, very bankable product because you've mitigated those risks effectively. Correct. And you can do that on any case by case basis for any project, whether it's a gold labeled product or the newest vendor on the market, right? And so anything could be bankable if you've packaged it and accounted for the risk to it properly. I mean, to circle back to what we talked about at the top of the PV6 lost diagram, you could fund snow clearing for every project you have and make it work. But is that a new asset management cost? Well, that's exactly, that's what a competent IE will ask that question and make sure it's funded in the financial model and included in the UNESCO. Yeah. A good note on trackers in snowy regions because as the solar market has grown, we've grown into places like New York and Wisconsin and Minnesota where there are frosty brisk, there's snow loss risk because it's so variable, it's important to review the warranty terms of trackers because some tracker vendors require the owner to keep a maximum snow cover at the site. And again, it's impossible. It's almost impossible. And Copenhagen, New York, it's almost impossible to keep an 18 inch maximum snow cover when you get three feet and 24 hours there. You're not gonna do that. So it's important to review the warranty provisions and what is going to violate the warranty. And if you think about it from a mechanical perspective, it makes sense that the tracker motors are designed as efficiently as they possibly can to rotate under environmental loads. If you're throwing snow on the ground that's going to resist that motor movement, the tracker vendors like we can't control that. We're going to protect ourselves and they're in their right to do that. But it also, you have to have a competent asset management team to say, where do we stow these trackers to where we're not gonna get wind damage from the storm, where it's going to not allow snow build up. Stowing flat again sensitive subject. That's your considering. Stowing flat increases the fortices of the wind. And so it's like, you don't want to stow flat. You actually want to stow off flat. Right, that's where wind issues. Torsional divergence, much like the Tacoma Nero's bridge. Normal wind gusts, but you have aeroelastic flutter. That was the fourth method. The aeroelastic flutter. I don't, I always forget that. It's rather nuanced, but it's a fun one. So, yeah. And so we have, basically, I think we've laid out from soup to nuts, right? We have the energy model, we have the equipment, we have the plan, we have the way we're going to install it, we have mitigating risks for that specific location, right? We have the asset management plan. That is all wrapped up into one gigantic hundred page bov of a document that you're going to have validated by the I in the panageneers and then also have the bank's review and eventually approve. Is there anything there that we missed that you think would be beneficial for developers to know to read you a little pine on that you have experience in? That would, that would help. I think the last piece that I would add in there is just an add on to what we discussed. So, large scale utility module purchasers, developers and builders. They hire companies like CEA, InterTech. There's another company based out of Arizona that will go visit these factories in Malaysia and Indonesia, China sometimes, India, wherever they're producing modules. Again, my expertise and experience has been in the DG field where we're buying 10 to 30 megawatts at a time, not 300 megawatts. So, we don't really have the justification to spend the, to expel the costs of hiring an independent party to go inspect our particular modules. Usually we're buying modules on the second hand market. So, from a distributor. And this really helped in us closing deals with banks, banks that had, and part of what's happened in the finance community is you have these experts in house at the banks that are engineers that are used to doing 300 megawatts deals now coming down and doing smaller and smaller portfolio deals. Again, but expecting the same dupelands, - Dilligence. - Dilligence, not dilligence, as a utility scale project, which DG, we just simply don't have the capital for. Right, we don't, and don't do. And so, what really I found beneficial in describing the way that DG, IPPs, and developers, procure modules is, you know, an AES building a 300 megawatt facility will say, "Ginco, here's my construction schedule. I will need 30 megawatts of projects or 30 megawatts of modules per month for 18 months." Oh, guess what? We're three months in the construction now we're delayed. What does Jinco do with those 30 megawatts of modules? Do they shut down their factory? No, they keep it churning, they keep it going. And there are second, there are like second tier, I shouldn't say tier, secondary market suppliers that come in and say, "Ginco, I will buy those 30 megawatts of modules and I'll sell them on the secondary market." That's where we really, Jeopardy, [BLANK_AUDIO] We capitalize on that access. - That access. - As capitalize on that access. And when I get asked the question, well, when are your modules being inspected? I say, never, we don't pay CEA to do it. However, CEA was there a month before, and they'll be there a month after. And with Chinko's permission and AES's permission, we've been very thankful that we can rely on those two CEA reports as saying, although your particular module is weren't inspected. - The ones before and after. - And after, and the ones after were all inspected, and they were consistent and found to be green flags, some amber, but no red flags. And so we've been thankful that the industry has opened up enough to where we can share these quality inspection reports as part of our bankability for the equipment that we're producing. And that worked. And I'm not, it provided a narrative as to how we made the decisions that we did when it comes to module for carements. And I think most IPPs in the DG space operate under the same guidelines. We're not gonna pay a third party inspector to inspect our particular modules and find 12 of them to ship to a PVL or a Kewa, I think is what they're called now, to do all the robust testing for specific LID and low light and all that information. - But we'll leverage the existing tests from the utilities that are doing that to make the case for your specific product. - Correct. - Yeah, the product that you're using, perfect. So that covers bankability, at least as it stands right now, thanks for going over all that. I'm really curious, based off of your experience, you've been in the industry for 20 years, where do you see this going, right? We just took a snapshot and actually like a journey to the past to see where we, how do we get here? What is industry best practices right now? Where would you like to see the industry going in the future? What would you think better practices in the future would entail? This is a ability to opine without expertise. This is a prediction into our crystal ball. So that is not meant to lock you into this box. It's just meant to try to say, "Hey, what are we thinking in terms of innovation?" - Wait, are you trying to get me out of an industrious box here? 'Cause attempting to be creative. - Attempting to. - With more sites. - With nice language attempting to, what would you say? - Cajol into insight. - Correct, yeah. - So I think as you see the industry changing, and the solar coaster is a reality. We all don't really know where the industry is gonna go, what the federal regulations and state regulations are gonna have for the industry as a whole. I've tried to stay away from really getting into that, 'cause as an engineer, that's very industrious and I just wanna build good projects. I've had the luxury of being able to kind of put my blinders on and say, "Just give me a project to design and build and operate correctly." I think what we'll see is an increased reliance on extrapolating or extracting capital from operating fleets. So Aspen is very incentivized to keep our operational fleet as close to 100% availability as possible, that's not possible, but as close to it as possible. In retrofitting our existing projects that have issues and continuing to build good quality projects up front. - And potentially repowering projects and already have an interconnection project, interconnection agreement that are nearing end of life to then have even more energy being produced from them. - Yep. And you and I've talked about innovations and not only the testing and commissioning activities that happen between mechanical and substantial completion, integrating and replacing full IV curve tracing every year with aerial thermography that has targeted IV curve tracing. I know some asset managers are already deploying that and they're finding some success in doing so. Again, you talked about repowering. So you have old products out in operating projects, that no longer have warranty coverage, that no longer have replacement parts. Part of my duties and my team's duties at Aspen is to support our asset management team in redesigning these projects and finding, hey, these inverters are failing. That vendors out of business they provide no support, help us find another inverter that we can plug in and give us a new set of drawings to retrofit. You have an engineer come in and provide the electrical engineering. Sometimes you work with the utility to make sure that they're okay with that inverter, yeah, and whatever settings are. - Our compliance with other connection agreements. And I also see a future that's longer term of taking existing interconnection agreements and supplementing them with non-solar. So battery storage combined, LNG, liquid natural gas, compression expansion cycles, even micro-nuclear, supplementing the solar during the day 'cause it's the cheapest value, the cheapest energy source when the sun is shining in a predictable manner, but what do you do at night? You can still use that interconnection agreement to inject energy into the grid. - Yeah, because it's already assuming that energy coming. Oh, it's brilliant. That's brilliant. All right. I love the ideas there. - Now we gotta put it in the action. That's me being creative. - I love both. I'm gonna take a step. - This is why engineers don't like being creative 'cause now we're like, all those are so many problems to solve. - How do we actually do that? - LNG didn't do that. Look, it's been five years and that didn't get implemented. But it's not about that. It's about hopefully somebody can do that. - Yes, right? - All right. Well, thank you. - Is there anything else you'd like to say before we close out? - I think this-- - This has been pleasurable. Longer than I had expected. It's been great. We got into a lot of details. - Yeah, thanks for providing us my details in your experience. I think one of the goals here is to have it so that people who have actually done the work, right? Who actually know this intimately can help other people who haven't had that experience to get to glean that wisdom. - Yes. - And I think you did a really good job today. So thank you for everything. - Yeah, we're here to make sure that the solar industry is representing the best quality that we can have going forward. It benefits everyone to have these conversations. And I don't think there was any secret sauce shared here. It's just very, we know what bankable looks like and there are efficient ways to make commercial decisions and work with the project stakeholders to find a project that's viable in a manner that's not going to kill projects. - Sweet. - Well said. - Yeah, I think we'll sum up there. Thanks.