Ogre AI: Predicting the unpredictable – AI forecasting for renewable energy systems

00:00:02: Empowering Tomorrow, the podcast from The Bond X.

00:00:09: You have a grid in the middle that is the same as twenty years ago so... ...the only thing that lets this physical infrastructure

00:00:16: to cope?

00:00:17: It's the intelligence!

00:00:18: The energy transition will not fail because we don't build enough wind and solar.

00:00:23: it'll fail.. ..because we can't predict well enough When you've got bad market conditions.... ...you require more intelligence & accuracy.

00:00:30: Welcome to Embowering tomorrow The podcast from Verbund X. Together with top experts, we explore the future of energy in Europe.

00:00:40: We talk about innovations technologies and bold initiatives that are already shaping a sustainable energy future today.

00:00:49: How can you reach a climate-neutral Europe?

00:00:51: how Can be close to innovation gap?

00:00:54: These are questions will tackle this podcast.

00:00:58: My name is Franz Zeuchbauer.

00:01:00: I am the Managing Director of VerbundX, Innovation & Adventure Unit at Verbund Austria's leading energy company.

00:01:07: Let's get started!

00:01:10: Welcome to Embarking Tomorrow, the podcast where we explore ideas that acknowledges and people shaping the future of energy.

00:01:18: I'm Franz Zeigbar, Managing Director for Verbund X, the Innovation and Venture Capital Unit at Verbund – Austria's Leading Energy Company.

00:01:26: In this third season our founder editions will put in a spotlight on entrepreneurs from the VerbundX venture portfolio.

00:01:34: These are the builders working at the forefront of energy transition.

00:01:37: Developing solutions that will define how we produce, manage and consume energy in years to come.

00:01:43: Today's guest is Mathe Stradan co-founder & CEO of OgréAi a fast growing startup on cutting edge of energy forecasting.

00:01:50: Welcome Mathe!

00:01:51: Hello Franz thank you for your invitation.

00:01:53: OgrëAi develops advanced AI powered models which help energy companies better predict both production demand a capability that is becoming increasingly important in the energy transition.

00:02:04: In this episode, we will dive into Mattis' personal journey to energy innovation – the founding story behind OKAI and real-world challenges of building a startup rapidly evolving an often unpredictable energy landscape….

00:02:18: …and talk also about collaboration between OKAI & Verbund.

00:02:22: So, Matti's really great for having VSS today!

00:02:25: But before talking about OKAI can you take us back and share how your own journey led you to the energy innovation.

00:02:31: Sure, thank-you Franz!

00:02:33: So what I can say is that i had no plans at the beginning of going into an energy business.

00:02:38: probably like every other person in a startup industry... ...I have a background in Business and Finance as a degree.. ..and my career started off in two thousand six as consultant.

00:02:51: My luck was that in two hundred nine ten I got a first-hand witness into the regulatory changes in Romania with respects to the balancing market.

00:03:03: And hence, I got my first hand witness basically what it means for forecasts and what it mean to nominate values of energy markets.

00:03:12: This is how I basically got insights on how much money was being lost.

00:03:18: Being already passionate about technology having spent years studying things such as programming networking as many other unrelated things to energy.

00:03:29: I basically put my thirst for technical knowledge and background, then i put it into the test in a real-world problem.

00:03:38: Can you take us back maybe to a specific moment where the idea of OKI was first emerged?

00:03:43: Sure!

00:03:44: It was more than just one process at that time – threefold.

00:03:48: First and foremost what I mentioned is the insights from the balancing market were extremely important and I witnessed some operators losing up to forty percent of their revenues.

00:03:58: I remember due to forecasting inaccuracies, secondly i had some insights that were brought about by other competitor solutions which I have the opportunity to work with an installed certain DSOs.

00:04:11: And thirdly as any other startup it's a moment when you are able put together great team.

00:04:18: So when these three things I think happened, this is where the idea came about and business actually became a proper business or proper business idea to catch on.

00:04:29: OK, I was founded in twenty-twenty one And when did you found your co-founder so maybe the team?

00:04:35: Now it's the moment of time for you to find how many months before

00:04:40: Sure.

00:04:40: In twenty-twenty one, I was actually working alone on the first solution which is for PV forecasting and only after i built my first solution that I met with my partners in late twenty-one.

00:04:50: And then in twenty-two it's when we took our first pre-seed round.

00:04:56: this started to build on top of the MVP that initially worked by myself.

00:05:01: so Six months we were, you know discussing talking planning going to competitions and benchmarking processes.

00:05:08: And in late twenty-twenty one We started up the company.

00:05:10: so mid twenty-Twenty One uh...we Were working together and probably at The start of Twenty twenty one I was Working alone

00:05:18: okay?

00:05:18: So really one man startup.

00:05:25: Okay, that's also quite convincing.

00:05:27: That you put all your effort to build it on the scratch for me own and then after a first proof point find out about others?

00:05:34: Yes!

00:05:34: It was like every other innovation thing for me.

00:05:37: There is passion in something I am very curious about... ...very driven to solve as a problem.. ..it was real need.

00:05:43: so i'm excited to built this energy brought by people around me And to build such great team.

00:05:50: It took quite a lot of great energy, let's see.

00:05:55: Okay you just pointed out some use of the specific problem in the energy system.

00:05:59: maybe you can explain it little bit more for our listeners.

00:06:02: which problems are we solving with OGAI?

00:06:04: Basically the problem that we're solving at Energy AI is need to these operators.

00:06:10: so they have legislative needs on day-to-day basis to forecast the energy quantities that they will produce.

00:06:17: The energy quantities are their consumer base, we'll consume you know losses on the grid

00:06:23: etc.,

00:06:23: so all of this is a mandatory activity mission critical and we're helping these operators to forecast better in order to reduce there costs through balancing activities.

00:06:33: I mentioned also for trading activities which should become more efficient by knowing better the quantity's available to acquire.

00:06:43: So, this is in the simple way how what we are doing and because it's regulatory driven or mandatory of course there a lot space on market for work.

00:06:56: Okay!

00:06:56: And MITE is so critical mission today that you have really accurate forecasting.

00:07:02: We're going into stage having a lot renewables with a lot of immobility in sector but its' really so critical from your point

00:07:10: Sure.

00:07:11: Because if we look at what has happened, let's say over the last ten years before you had a stable dispatchable base which was replaced by millions of distributed weather-dependent assets for example on demand side.

00:07:28: it used to have predictable demands.

00:07:30: now there are EVs that didn't exist five year ago.

00:07:36: You have the presumers, which are generating a lot more energy.

00:07:40: We had heat pumps that were not so common back then swinging with temperature and you also have of course... ...the elephant in the room is now the data center load….

00:07:49: …which has grown faster than anybody else planned for.

00:07:52: So what happens at both ends of this system?

00:07:56: They're becoming stochastic at the same time And we've got a grid in the middle.

00:08:00: That's exactly like it was twenty years ago.

00:08:02: The only thing that lets this physical infrastructure to cope is the intelligence behind it to basically forecast and manage these DRs, things connected with infrastructure.

00:08:14: So in essence without good quality forecasts I think you have massive imbalance.

00:08:20: costs curtail more renewables frequency is more unstable so far.

00:08:26: The energy transition will not fail because we don't build enough wind and solar.

00:08:30: it'll fail as well to integrate it fast enough and make the most of it.

00:08:37: Yeah, many things.

00:08:37: that's a key problem and massive problems when you're solving this OGI.

00:08:41: if we look back at your founder story Mathe song You found out as one man shows on OGI in twenty-twenty-one What were the toughest challenges you met?

00:08:51: And do have to overcome these years.

00:08:53: So first and foremost, as you said is the team.

00:08:56: so building the team and energy and AI Is pretty rare intersection with human resources, people that both understand the sector and people who can build serious machine learning technologies.

00:09:09: Both of these people are hard to find while finding them in the same room is pretty easy but harder.

00:09:15: so this was first and foremost hardest thing.

00:09:18: Secondly I think building a product outperforms current market.

00:09:22: This one's very difficult because in let's say our services, the customer does not really care about their story.

00:09:29: They interfaces pretty presentations.

00:09:31: this is a numbers game.

00:09:32: so we sell numbers and they benchmark with our quality versus theirs.

00:09:38: So if you're not measureably better then conversation pretty much ends.

00:09:43: This was very difficult getting past these thresholds which are enough performance to solve solution.

00:09:51: And I think thirdly was to build the right references.

00:09:53: As you know, very well in energy, I think references are their real currency.

00:09:58: so one serious customer and production really unlocks them.

00:10:00: next ten conversations.

00:10:02: So this is also a Very valid point for work with Verboond then accelerator that i'm sure we'll talk about later.

00:10:09: That has helped us alot with the references

00:10:15: You mentioned.

00:10:15: the first point was finding the right people, where you have found their rights.

00:10:19: At universities in other countries and your neighbourhoods?

00:10:24: I'm very fortunate to know my co-founder partners that we started this business up with one of them being a professor at Oxford University and now later at UCLA in the States.

00:10:36: We had some contacts, let's say with very remarkable resources that could help us with research side which sits on base of technology basically quality service we deliver.

00:10:46: so I think this was a important thing also an edge within our company because we have talent access through my partner Mihai, who is leading the research within your organization and of course has supported us in identifying... ...the right people for their job.

00:11:06: In terms of machine learning and AI research.

00:11:09: Okay!

00:11:10: If you look back at the years of OKI's on the market there have been several shocks also in the market.

00:11:16: The latest one song called Crisis in the Middle East.

00:11:19: How these crisis and shocks have maybe shaped your strategy?

00:11:24: or your products?

00:11:25: Yes, sure.

00:11:25: I need some careful framing here because of course there's nothing to celebrate in this and what is going on.

00:11:32: They have definitely reshaped the European energy markets in lasting ways.

00:11:38: first of all i would mention that regulatory got a lot harder with price cap.

00:11:44: you have windfall rules emergency interventions.

00:11:47: You know we have complex sector which has gotten more complex.

00:11:52: We were counting with our team, I think one or two years ago and there was a two hundred and sixty something legislative changes only in Romania.

00:12:00: With respect to the energy markets it's very difficult to operate.

00:12:04: so this has been let us say stress The retail competition of course has intensified And all retailers had problems with prices spiking contracts moving and margins being a lot compressed compared to the past, and I believe this in turn of course has helped us a bit because every forecast matters more now when you have these very small margins.

00:12:31: So let's say the underlying logic is pretty simple in my opinion.

00:12:34: so we need bad market conditions that require intelligence and accuracy and efficiency.

00:12:40: That itself increases demand for our AI technologies and large margins tolerate inefficiency while smaller ones do not.

00:12:50: In terms of new products, I think this was also something that has changed over the past years.

00:12:55: For example for us we have built a new solution to help retailers manage the presumers portfolio and reduce the imbalances related to that.

00:13:03: The legislations have pushed for the pursuers increase in euro.

00:13:08: so This is another good part And if i had to look five years from now there's going be huge opportunity for European AI energy startups, because the energy sovereignty is becoming a big topic.

00:13:23: Data center security and data security's become a big topics.

00:13:27: so I think it's good environment for us although of course on basis something not so good.

00:13:31: So this what i'm thinking.

00:13:34: Okay, okay.

00:13:34: Now many things.

00:13:35: so these are your perspectives for scaling a startup like ODI based on the frameworks and the market conditions but other also some issues For example in the regulatory framework or in the market that holds you back bringing innovation to the market Which has to change in the car?

00:13:50: In the next years that really start up.

00:13:52: Like you we can scale and make a difference of the future.

00:13:56: Sure I think There are few things especially if we look at their regulatory side.

00:14:02: I think there could be a more active push for AI into utilities and with pressure such as regulators that are asking utilities how they're using AI to reduce the waste.

00:14:14: You can have regulatory sandboxes test AI in real use cases, you know?

00:14:21: With good security but with real data.

00:14:23: so we really testing out and just have lab tests data privacy frameworks.

00:14:28: So this is a problem that we are very afraid of because we see an increase in the security of data, but does not really go hand-in-hand with AI and building AI around it.

00:14:39: so if you protect Data and don't have access to it You're going to be capable to develop something very worthy.

00:14:45: I think The market has helped us recently And should help us Because We had the fifteen minute change Intervals as you very well know.

00:14:55: So this means more reporting, more data and of course it means more business for us.

00:15:01: There's more transparency now into the reporting.

00:15:04: This should also be helpful to our businesses.

00:15:08: In terms of the data Let see what problems with the data.

00:15:11: I think The biggest problem is standardization Of data.

00:15:16: I think it's very, very important to have machine-readable access to grid and market data.

00:15:24: So the unitarity of data from DSO to DSO across Europe to be able to use this fast and efficiently would be very helpful so that data management does not take ten years and you know its built somehow smart form the beginning.

00:15:39: You have a big project in mind with clear strategy but do not mess things up.

00:15:46: What's holding innovation back?

00:15:47: I think there is a lot of funding gap here.

00:15:50: So even with startups, you know very well as an investor that we are always pressured to make profits and to make revenue in.

00:15:56: this does not always align very well with research.

00:15:59: so this has to be some sort of balance between these two.

00:16:02: the sales cycles This Is Always Something To Worry For Startups.

00:16:07: When You Have One Year Long Sales Cycles It's very difficult to plan, very well sometimes.

00:16:13: And of course the risk aversion of utilities is the last thing that may pop into my mind because naturally and I think this is a fair point they are conservative with start-ups as they want to work in serious companies.

00:16:25: so it has something expected but must be managed somehow.

00:16:28: So

00:16:29: if you look back we started not investing.

00:16:32: We started incorporating via the ex-excelerator.

00:16:36: What was your experience working together with Verbund?

00:16:39: And where this collaboration also made an impact to the further development of OK.

00:16:45: Sure, so first of all I want to say that and honestly the Verbund accelerator was very different from other accelerators we participated in... ...and i think we've been through four or five if am not mistaken.

00:16:57: And in my opinion, and very honestly it was not just a social media event as we have witnessed at many other numerous companies.

00:17:04: This is about the real problem solving thing.

00:17:08: It's really relationship building exercise with real opportunities In the most programs that we have seen elsewhere, you get a lot of mentorship and demo days but... ...you don't really get the opportunity to work with companies as we've seen here.

00:17:22: So very shortly I would say that Verbund gave us a real problem on a real system….

00:17:26: …with real consequences in clear path from the beginning which is also something important because other companies are saying this isn't their target!

00:17:39: For us, with Verbund of course you may remember we did the demand forecasting for EVs.

00:17:44: It was very straightforward.

00:17:46: We implemented it and I say...I think we implemented it very swiftly with you And it helped a lot.

00:17:53: So not only because of knowledge but see You know..we developed our product more so that product maturity was improved.

00:17:59: The validation tier one energy utility That says this product works in production on their infrastructure Opened doors that we could not have opened otherwise.

00:18:10: So I think the knowledge, let's say the opportunity is really important.

00:18:17: Okay later on you did then as long as for both ventures all decision to invest in you and your company?

00:18:23: And how?

00:18:24: from a perspective they dance on the collaboration change or maybe also some of impact because it's another thing.

00:18:30: if you've referenced customer or utility CVC on your cap table was there something you recognized afterwards?

00:18:38: Yes, of course.

00:18:39: The great part and the smart thing was that you've made investment only after technical demo is done and solution has been implemented.

00:18:46: so... That sequence I think matters a lot.

00:18:49: and Verbund saw technology work in their own operations before they invested.

00:18:54: So when round actually happened we weren't selling you pitch as remember had something already implemented.

00:19:01: i think capital was very meaningful but capital was I don't think the essential thing here.

00:19:07: So, I think that credibility and depth of relationship in long-term horizon is what we value most out this endeavor.

00:19:19: We know each other from two perspectives being really a client having customer-client relationships on the other side.

00:19:26: so it can be an investor's role to make these roles work well for you and for OKI, maybe you can share some experiences in this respect?

00:19:35: Sure.

00:19:36: So I think the first thing that really made it work is people.

00:19:39: so i think uh... The team we were exposed to France on the VerboNext team understand both sides of both let's say the technical side also the capital side.

00:19:53: they understood very well utility operations start-up dynamics.

00:19:57: This is rarer than it should be.

00:20:00: And I think secondly, what has worked very well was the sequencing that we have mentioned.

00:20:05: So venture client first and investor second.

00:20:07: so this order i think is the reason why it works really well?

00:20:12: On my advice to founders ,I would say when a corporate invites you to an accelerator To treat at the operational collaboration as a prize not the investment.

00:20:20: This will be my advice.

00:20:22: In the investment if comes It's consequence of value created Not really goal.

00:20:27: Thanks for this advice

00:20:31: Mati.

00:20:32: and maybe one further question.

00:20:34: What's your long-term perspective on OKI?

00:20:37: If you look back, may be fifteen or twenty years.

00:20:39: what do we like to have built out of business company?

00:20:43: Sure it is very hard to predict as you know regulatory changes all the time.

00:20:48: but in essence I believe that due forecasting always being a critical layer.

00:20:53: And if you have to look towards the future at fully decentralized electrified grid, You will have millions of generation points.

00:21:01: You should have hundreds of millions of flexible loads and probably a physical grid not yet designed for any of this.

00:21:09: So in my opinion we won't be able to operate all these on schedules.

00:21:15: We'll need to operate it with a continuous probabilistic forecasting system which works every node.

00:21:22: So, if we have to analyze the last decade in terms of forecasting I think it was all about renewables forecasting.

00:21:29: If you look towards future its going probably be more focused on consumption with a lot of sophistication for EV charging behavior as said previously heat pumps cycles industrial flexibility and data center load which i mentioned.

00:21:45: so If we get it right, I think forecasting becomes the energy system intelligence.

00:21:49: And if you get it wrong... ...the transition slows down and not from physics but a lot of uncertainty.

00:21:55: Madam to close our conversation maybe i question if we beam ourselves back into twenty-twenty one The moment song where decided to start with all that knowledge and experience You made in last couple years What things would make differently?

00:22:13: the things you did and what would do something exactly same as before?

00:22:17: Add some learnings to share.

00:22:23: I think two thing that i can probably think about when we are talking differently, so first of all i will probably hire senior talent faster Especially technical leadership, so this may be one of the things that I would and also related to your previous question.

00:22:39: And secondly i will say you know that energy sector does not really forgive mistakes in the sense maybe we were a bit very too cautious In pushing out solution for market.

00:22:50: then maybe could have had few months extra time if started selling earlier but We don't know this yet.

00:22:56: So maybe we would have failed otherwise, and those wouldn't have not opened the same.

00:23:00: so This is what I would've probably done differently in terms of you know The same.

00:23:05: And um i Would say that uh You know?

00:23:08: We resisted a huge Temptation to expand into other things such as optimization Such As trading bots etc.

00:23:15: and we resisted That.

00:23:17: so i think staying focused To be the best at forecasting is something that I think was a good decision.

00:23:25: And we want to become the best forecasting layer in the European energy system, and I think the discipline to stay narrow was important... ...and it will be the most innovative thing that we may do.

00:23:38: so We want to stay at forecasting.

00:23:40: you want to becomes what we're doing and go-to place for this around the world eventually.

00:23:46: That wraps up our conversation quite well, Mathe.

00:23:49: Many thanks and many things not only for this interview but also for all of our collaboration.

00:23:56: I think it's quite unique that we had this cooperation first with the Federal Accelerator then we invested together And that really tackles a problem which is mission critical to make better forecasting using AI in the right way.

00:24:13: So it's really a pleasure, so many thanks Mathe.

00:24:15: And we are looking forward to an exciting journey together and supporting you further with OKI.

00:24:21: Many thanks!

00:24:22: Likewise thank you so much Frans and

00:24:26: thank you to everyone.

00:24:30: have a great day Thanks a lot !

00:24:41: Your support helps us to bring more fascinating guests and keep the conversation going around.

00:24:47: innovation, climatic adventure capital.

00:24:50: And remember – The future of energy depends on things we do today.

00:24:54: Let's

00:24:54: empower tomorrow together!