Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Episode 25 – Part 1 – How AI is changing how buildings are surveyed with Nic Cory, Absolar
45 Plays1 year ago
In this week's episode, we speak with Nic Cory from Absolar.
Across the three parts of this episode, we are discussing how AI is changing how properties can be surveyed and the role of solar and heat pumps in the future of the built environment.
Nic is an entrepreneurial Finance Director with a diverse background in Big 4 audit, deals, private equity and property fund management. His journey has led him to Absolar, where they are using technology and data to drive Net Zero opportunities and decarbonise property portfolios whilst ensuring financial returns for clients.
Absolar exists to help people and businesses reduce their energy costs whilst adopting renewable energy sources. Using unique AI-based remote sensing and GIS technology, Absolar can carry out remote solar surveys for any building, portfolio, and city, wherever you are in the UK.
In Part 1 of this episode, we discuss:
🌱 Using AI to accelerate the change to renewable energy across property
🗺️ How AI helps surveyors assess large numbers of buildings more efficiently
🏫 The importance of understanding a roof's makeup, including its orientation, pitch, and solar potential, to determine whether solar PV is a viable option.
🤖 The limitations of AI in real estate
👷♂️ The potential replacement of surveyors with AI
🦾 How AI will become more accessible to different parts of surveying
Recommended
Transcript
Introduction and Guest Background
00:00:00
Speaker
Welcome to Survey Booker Sessions. Tune in to hear from people working in a range of industries and roles to provide you ideas that you can take away and use in your own business. I'm your host, Matt Nally, the founder and director of Survey Booker, which is the leading CRM and survey management system for surveyors. On this week's episode, we have Nick, who's the director at Absolar, so thanks for coming on. Thanks very much, Matt. Thanks for having me.
00:00:22
Speaker
For those that don't know you, do you want to give us a bit of background as to I suppose who you are and what you do at Absolla? Yeah, of course. From a personal point of view, I'm an accountant by training, but gave that up a long time
Absolar's Mission and Technology
00:00:33
Speaker
ago. Did my degree in real estate at College of Estate Management a number of years ago now. Three years ago, joined up with a few colleagues from University of Southampton, from engineering and computer science, really with a mission set out at Absolla to help
00:00:50
Speaker
accelerate the change to renewable energy across property.
00:00:55
Speaker
Interesting and I suppose there's three topics we're going to cover today around this. Later will be the sort of future of the built environment and we'll look at SOLAR specifically on different buildings but one of the first things I thought was really interesting with what you're doing is the use of AI and I suppose changing how you survey buildings in the first place. How do you use AI to cover that process versus going out on site?
00:01:22
Speaker
Yeah, indeed. And the artificial intelligence and the various aspects are incredibly useful for how you repeat something at scale. Any surveyors will know, you're looking at one building, the results are normally pretty obvious. If suddenly you've got a thousand buildings to look at, you need a bit of help to start assessing that.
00:01:40
Speaker
So we use, a lot of our base data is LIDAR, if you've come across that. If not, in our case, it's lasers bounced off a plane, helps us understand the built environment underneath in sort of minute detail, allows us to build a 3D model of what's going on on the ground.
00:01:58
Speaker
We then have various bits of AI that can detect from those models. What's a roof line? What's a dormer window? What's a skylight? So you can start to understand the makeup of a roof, which is obviously important when you're looking at solar PV. So you can understand what are the roof facets, which way is it facing? What's the pitch of that roof? Start to overlay a bit of satellite imagery and you can essentially plot each
00:02:24
Speaker
solar panel, how much can you fit on a roof across an entire city at
AI in Building Assessments
00:02:31
Speaker
one go? We actually measure into five kilometre blocks.
00:02:35
Speaker
We then use a lot of radiation modeling to say how much power can a rooftop generate at every half hour across any given period of time. So you can begin to understand how much electricity can any rooftop generates within a particular environment. And then the rest is sort of pretty simple, pretty complex, but in theory, pretty simple maths and business cases to then say, and should you actually go ahead and do that.
00:03:03
Speaker
So how does it change the number of properties you might actually visit and do a traditional roof survey on, I suppose, in terms of understanding whether solar can be fitted? Yeah, so on PV in particular allows you very quickly to say, is there enough power that's going to be generated on this rooftop compared to the electricity consumption inside to even warrant investigating whether PV is suitable.
00:03:31
Speaker
So it allows you, within Absolar, typically we can look at 2000 buildings a day and quickly whittle down which ones are those that are actually interesting to people and which ones should they park and either discount or put at the back of the queue. Particularly with emerging technologies, it's key to get those quick wins first to get the adoption rates up and get everyone happy that technology does work.
00:03:59
Speaker
And we're increasingly employing that in air source heat pumps, identifying where is their space to install that with all the noise requirements, for example. You can't have them looking at a neighbor's window. So it allows you just to identify where they can go, or perhaps more importantly, where they can't go, where there are restrictions. That's interesting. How does that compare? So if you're doing 2000 a day, how does that compare to, I suppose, old process where it was literally just going out?
00:04:28
Speaker
Yeah, we still do manual surveys. We still sort of, for those that are suitable, we still go out, we do three a day. So you can see the difference that you can get to. And so recently, I think we did a study of Basingstoke, the entire sort of borough of Basingstoke. And within a week, that's it, you've understood where in Basingstoke is good and bad, as opposed to what would have been a two-year study for a surveyor looking at buildings.
00:04:55
Speaker
There's a massive difference.
Human Role in AI-Driven Surveys
00:04:57
Speaker
Is there an element then of, I suppose the question that we ask in every industry around AI is, could it replace surveyors long term or is it you always going to need that? Yeah, pretty confident that it won't. What it helps you do is do that identify stage pretty quickly. So it does
00:05:17
Speaker
it helps you, it gives you a bit of shortcuts, you still need to go out and actually look at that building if you are going to be doing a design or a specification. And it's the grand old problem of real estate, you might have, you know, nothing's homogenous, you might have two buildings that are identical next to each other, actually the tenancies are different, the occupants is different, there's a surrounding environment that's slightly different, therefore the requirements are totally different.
00:05:41
Speaker
So you still need that personal touch, that there's no point training AI to act over something that's not entirely homogenous data set for the whole process. So it helps you at the beginning of the stage, definitely needs to hand over to the surveyors when you start getting into action.
00:06:02
Speaker
Yeah, I suppose it'll be interesting with the development of smart buildings, and I know this is a long way off, but obviously there are some newer buildings that are much smarter now, you know, how many people are on each floor, how many people have scanned in, what their usage of different things is. I suppose that could start to feed in the future, but you're right, I think it's always going to need someone to actually have feet on the ground, see what's going on.
00:06:25
Speaker
Yeah, it does. And also, even our data, we gather data every two years. It's out of date. Buildings change. You've got to pitch up to actually look. Before you send your containers of panels to go on the roof, you've got to pitch up and do that manual survey.
00:06:42
Speaker
Interesting. On that note about containers of panels, I was walking through a field the other day that had what looked like 30 lorry loads of of of solar panels just in crates in a field. I've been there a long time so I was intrigued as to what they were for. I suppose what are the limitations of AI then? Is it the data sets in the first place?
00:07:03
Speaker
Yeah, you've got to train it. It's taken us two and a half years to get to the point where we can realistically point to the roof and get results back that we're comfortable with. Even then, it's a nine out of 10 success rate. If you go and say that's 90 percent accurate, great, it sounds wonderful, but actually looking at 10 buildings, that means one of your results is just slightly weird and it only takes one. If you're looking at a portfolio and you look at a couple of results that look weird,
00:07:32
Speaker
as surveyors, we tend to say, well, I don't trust the whole thing. So it does need, the quality of data is reasonably good. AI is particularly bad. It doesn't perform as you expect all the time. Therefore, you still need that manual review process that actually corrects some of its errors. Interesting. Is that something you, I suppose, you get better at over time identifying
00:07:59
Speaker
Yes, or the odd anomalies, I suppose, that are in the data or in the results area. Yeah, exactly. And we run two different types of calculation within Absolla when we're looking at a property, and that allows us to pick out, well, there's two disagreements here within a different AI model, so it allows us to spot those errors. But it needs a human to sit and resolve the argument that the two machines are having with each other. Interesting, okay. And with, I suppose, just continuing on the AI side of things,
00:08:29
Speaker
What do you think the changes will be as we go on?
AI's Broader Applications in Energy Systems
00:08:32
Speaker
Because I suppose we've made quite, you've made quite good advancements already with being able to scan on scale or analyze on scale. What other applications do you think there might be going forward, whether they're in this space or elsewhere?
00:08:46
Speaker
It still feels like quite an unknown area. It certainly is. It's certainly going to be opened up to more and more people. You know, at the moment it's still quite a novel thing and surveyors get access to our type of stuff, but the end users don't sort of get the instant access. And that will come with time you've seen with open AI, kind of as soon as general public start to get access with AI.
00:09:09
Speaker
Then come the mass adopters, but then also come the debates over whether it's a good thing or not. We're seeing a lot more rolling into actually monitoring our systems performing as expected. So am I getting value for money from what I've put in? And actually that's where
00:09:29
Speaker
AI does have a good role to play, because on day one, you can feed in all the parameters, that's not going to change. And then you can get quick alerts as to when something's going wrong within a system, as opposed to what has happened the last 10 years. And so the PV of the system goes wrong, people turn it off, and they never turn it back on again, which is, which is obviously frustrating. Yeah, definitely. And so with the, I suppose case study wise, because I think, I think it was the, in Southampton, you've done a
00:09:57
Speaker
a site already, haven't you, from start to finish? Yeah, we've done loads of sites from start to finish, yeah. How accurate have you found, I suppose, the analysis process through to when you're
00:10:12
Speaker
looking at how things are performing after installation and so on. Has it matched quite well to that initial present? It does. So it matches the radiation predictions. We also feed in what I used. It's an acronym of NREL. I can't remember the full name of it. So we feed in sort of industry standard calculations as well into that radiation data. So it gives us a bit of robustness.
00:10:37
Speaker
There's one site we are monitoring in particular because we've put PV panels on every side of the roof, including the north side, which is bordering on wrong, but actually for that site was the right solution because of a low pitch and high energy consumption.
00:10:52
Speaker
and it's within two to three kilowatts a month accurate, which is incredible accuracy, so we're quite proud of that one, keeping to track that. It only goes wrong when we have a serious weather event when actually there's a heat wave or we've got really bad weather for a while, in which case we kind of have to allow for a bit of difference there.
00:11:13
Speaker
That's an interesting point, though, actually. Is that getting more tricky than with weather patterns seemingly being a bit all over the place at the moment? And I can talk to that site in particular. It averages out to still being in line with expectations. So we have the extreme weather events that you can argue whether are one off or not. But actually, it's still averaging out, as you would expect. So don't see a material impact from that.
00:11:42
Speaker
Interesting.
Business Cases and Property Types
00:11:43
Speaker
Again, I think my final question on this then, on this particular part of it is how do you decide, is it through the analysis process or later on between potentially using solar or the heat pump side or insulation, where does that fit into the process?
00:12:00
Speaker
And our analysis will produce essentially summary business cases for interventions. So it becomes quite obvious early on if you're looking for financial payback, which ones are those that are going to be of interest and how they interplay with each other. There are still
00:12:19
Speaker
restrictions outside of a business case as to whether you want to go ahead with something. Solar PV can be done when you've got occupants in the building isn't going to impact on a tenancy. Obviously going in and re-insulating or changing out a heating system means landlords are not going to be making those interventions. We present the information, so here's the impact it'll have, but it has to be down to that end user as to how that fits into the realities of the building.
00:12:45
Speaker
How much does this suit commercial versus residential? Is it beneficial to both in the same way or does it suit sort of maybe a commercial self industrial area better?
00:12:56
Speaker
It's highly suited to commercial because each commercial property is so different. The profile of an occupant within a commercial building as well as the rooftop is so different, it does meet that kind of bespoke assessment. Residential, you can actually start to apply some rules to various types of property and types of occupants. And so an AI assessment is useful for a portfolio and then you can start to extrapolate out
00:13:24
Speaker
So you get less use on a per property basis on residential, but you get very good if you're trying to assess a city or what kind of social housing portfolio works very well. For commercial, it works all the way down to each individual.