Disruption in Electricity Markets: AI, Renewables and Demand Growth
Energy markets are undergoing major changes. Demand for electricity is increasing and at the same time supply from intermittent renewables is gaining market share. Artificial intelligence (AI) is another source of disruption for electricity markets.
To help us understand the changing market dynamics from an electricity trader perspective, our guest Cory Paddock, President and Co-Founder of GBE Energy joins the podcast.
Here are some of the questions that Jackie and Peter ask Cory: How is electricity traded? As renewables gain a larger percentage of total supply, how will electricity markets be impacted? Are market reforms, such as capacity payments, needed to ensure power reliability as the share of renewables grows? What is a “duck curve”? How could AI change electricity markets? Is AI already being used? Could AI replace human electricity traders?
Content referenced in this episode:
- Small Vermont utility quietly builds a fleet of 4,000 Tesla Powerwalls (Electrek, Aug 22, 2022)
Please review our disclaimer at: https://www.arcenergyinstitute.com/disclaimer/
Check us out on social media:
Twitter: @arcenergyinst
LinkedIn: @ARC Energy Research Institute
Subscribe to ARC Energy Ideas Podcast
Apple Podcasts
Google podcasts
Amazon music
Spotify
Episode 204 transcript
Disclosure:
The information and opinions presented in this ARC Energy Ideas podcast are provided for informational purposes only and are subject to the disclaimer link in the show notes.
Announcer:
This is the ARC Energy Ideas podcast with Peter Tertzakian and Jackie Forest. Exploring trends that influence the energy business.
Jackie Forrest:
Welcome to the ARC Energy Ideas podcast. I’m Jackie Forrest.
Peter Tertzakian:
And I’m Peter Tertzakian. Welcome back. Well, it was a beautiful weekend here in Calgary.
Jackie Forrest:
Like summer heat.
Peter Tertzakian:
It was. Over the weekend, there was a different kind of heat in different kind of rooms, with the OPEC meetings over on the other side of the pond where they’re talking about production cuts. And they did agree to cut, or at least Saudi Arabia did.
Jackie Forrest:
Yeah, that’s right. So Saudi Arabia announced a million barrel-a-day cut. This is potentially doubling the cut for the entire group relative to October 2022. What’s different about this one is Saudi Arabia is making all of the cut. The rest of the 23 different countries involved are making no cuts, although they are committing to keeping the cuts they already had. So this is a little bit different of a strategy. Saudi Arabia wanted all the others to also take some pain-
Peter Tertzakian:
Proportionally.
Jackie Forrest:
And it wasn’t always the case because of a lot of reasons in terms of where the quotas were and things like that. But this time. Saudi’s saying, “Hey, we’re willing to take the hit here and we want to stabilize oil price,” which had fallen below $70 last week. It’s now at $72 as of this morning as we’re recording.
Peter Tertzakian:
Right. All of this lowering of output is a consequence of weakness in the broader oil market as a consequence or an expectation of weakness in largely the Chinese market, right?
Jackie Forrest:
Well, it’s interesting. It’s kind of a weird market. If you look at any analyst, it shows a very tight market in the second half of the year, maybe draws of 2 million barrels a day coming out of inventory because we don’t have enough crude oil to meet demand. So that would imply an $80 + type market, especially when you consider where inventories are, which one kind of at average levels, but we’re not there. We had $68 oil last week and that’s because people are concerned about a recession. And those outlooks for the big draws and the tight market require about 2 million barrels a day of oil demand growth this year. And people are concerned, whether it be China or just broadly the interest rate hikes that are slowing down the economy that we’re not going to see that.
So it’s a strange market that way. And Saudi’s saying, “Hey, I believe in the underinvestment. I think the market’s going to tighten at some point. I’m willing to take a hit here in the meantime just to get prices high,” because I think they know that their supply will be needed at some point.
Peter Tertzakian:
Right. That is the overarching narrative is that demand is going to level out here depending upon global economic activity, and that underinvestment and supply will be insufficient to even meet that demand. That’s the expectation as we roll into the second half, which is only three and a half weeks away. So is summer driving season. We’re revving up for that if you pardon the pun. So, it’s interesting that they felt they had to cut now. I guess it’s really to shore up the price above $70 bucks.
Jackie Forrest:
I think so. And they kept saying that we value market stability, which means probably prices being in a range that works for them.
Peter Tertzakian:
Well, we’re going to be talking about this more because as we roll into the second half, that’s when all the expectations of a tightening market and higher prices so we’ll follow it and see.
Jackie Forrest:
Definitely. But today we want to talk more about electricity and all the changes that are coming to the electricity markets, whether it be from growing demand, we know the EVs are coming and are pushed to electrify everything. We know wind and solar generation is growing fast, and although it’s still a small amount of generation in many places today, it’s expected to grow and it brings with it a lot of volatility in terms of how it comes on and how unpredictable it is. And then the third thing that’s coming is artificial intelligence. And no matter where you look, there’s a discussion of how AI’s going to change industries, whether it be writing scripts for movies or writing songs, but we haven’t talked on the podcast about how it’s going to affect energy and it will have an impact, we think, on electrical markets.
Peter Tertzakian:
Yeah, for sure. I mean, AI, we’re going to talk about how it is going to be used to predict supply and demand, but importantly, AI has to be powered by electricity. And we know that data centers are expanding, and we’re going to talk about that, which means more electricity demand against the backdrop of changing supply structure, the shift from base load to more intermittent renewable. Anyway, I’m already talking too much, and we have had electricity experts on our program podcast before, but we have never had somebody who is in the trenches of actually trading the electricity. So on that note, welcome Cory Paddock, who’s President and Co-founder of GBE Energy, an electricity and energy trading house here in Calgary. Welcome, Cory.
Cory Paddock:
Thank you very much.
Jackie Forrest:
Okay, well, maybe first, Cory, just tell us a little bit about yourself and GBE Energy.
Cory Paddock:
Great. Yeah. So Cory Paddock was born and raised Calgarian. My dad is fully born here too. Very much a part of all the institutions that exist here in the city. Went to the University of Calgary, have a couple of degrees from U of C, and played volleyball at U of C after I graduated, I was interested in finding a career to match the competitive aspects of sport sports analytical skills. And trading is a very interesting place to express that. Every day you’re either winning or losing, you know whether you’re getting better or worse, and you get immediate feedback-
Peter Tertzakian:
In your wallet.
Cory Paddock:
In your wallet, right? The numbers do not lie. So I learned the business here in Calgary actually at TransAlta, and this is going back into the late 2000s, and I guess was lucky enough to work my way up from the back office there onto the trade floor. And actually, TransAlta has produced several high-class traders in the power markets over the years, many of whom have gone on to found their firms as well. So, quite a significant institution when you think about power markets here in Calgary.
Peter Tertzakian:
Yeah. When I tell people about electricity trading and what goes on behind the scenes when they flip on the light switch, it’s a little bit hard for the average citizen to think about electricity trading because it’s a commodity that is somewhat nebulous. I can touch a barrel of oil if I wanted to as a trading, trader-visualize a barrel of oil, but electricity is nebulous. Can you give us a sense of exactly what’s being traded?
Cory Paddock:
Sure, yeah. So the way to think about this is to start with how you arrive at an electricity price. So you’ve got renewables, obviously, wind and solar, and then on the kind of dispatchable side or fuel side, you’ve got nuclear, you’ve got coal, you’ve got gas, and all these different energy sources have different characteristics about them. So meaning how long they need to run in order to be economic the ramp rates, sand, and how quickly they can ramp up and down. So that’s a major consideration just on the supply side. And then on the demand side, the demand is very variable. So you’ve got a certain amount of baseload that you can quantify and think about temporally over the span of years. So there’s obviously like day of the week, there’s hour of the day, and then there’s seasonality to a lot of these things as well.
Peter Tertzakian:
So, baseload, just to be clear for our audience is just sort of that background base level of electricity you need to keep society running, including factories that run 24/7 are collective refrigerators and furnaces, well, no, not furnaces, typically gas-
Cory Paddock:
All kinds of loads roll up into something predictable.
Peter Tertzakian:
Predictable, right? And then there’s the morning rush and the evening rush.
Cory Paddock:
Even a lot of that is very predictable, depending on the time of year, the time of day, when the sun’s rising and setting, et cetera. And then where things get more volatile is just the weather. So at a high level in the winter, how much electricity people are using for heating, they’re tending to stay inside more often. And then you get into summer and then you’re talking about air conditioning loads.
Peter Tertzakian:
So in conceptualizing the units of electricity that flow off the wires, like if we consider the wire the assembly line, the conveyor belt coming into the house, that every hour we have a kilowatt or kilowatts of electricity flowing in, and that you can source it from different sources and pair it to the demand and that’s what you’re trading.
Cory Paddock:
Yep, and I think it’s probably worth discussing too, just how different markets can be throughout North America. So this is going back to the early 2000s, late ’90s. There was a big effort to deregulate electricity markets and several jurisdictions did that, including Alberta. So Alberta was an early adopter when it comes to that.
Peter Tertzakian:
So it’s just like a market that you would go to trade anything from vegetables to whatever, right-
Cory Paddock:
Yeah. So-
Peter Tertzakian:
… in this case, it’s electricity.
Cory Paddock:
Yeah, exactly. We didn’t discuss this part about electricity in that. Of course, it’s talked about quite often that you can’t store it. So by that, this is a very, very dynamic process. And in all the deregulated markets of North America, you’ve got a price every 5 minutes. And then it can even get more granular than that when you’re talking about most of the grids in the United States that are deregulated have nodal prices or locational marginal prices where you’re trying to calculate or not trying, you do calculate the price at several nodes within the system. And when I say several, I’m not talking about one or two. I’m talking about several thousand.
Peter Tertzakian:
Right. In a lot of places that are regulated also the regulator price comes into the home, the homes of our listeners, and they pay a constant price. But behind the scenes, there is all this kilowatt-hour trading going on.
Cory Paddock:
Yep.
Jackie Forrest:
And it’s also probably worth talking about the major areas of North America’s power grid.
Cory Paddock:
Sure. Yeah. And this is kind of an interesting history in that if you were starting from scratch right now you’d map everything out quite differently. But like everything, these things started as simple systems and then work their way up to complex systems. So all the grids in North America do operate at 60 Hz. So that’s the frequency of the grid and the frequency that all of our major machines work at, all of the devices that we have in our homes. However, they’re not all synchronized. So there’s a western grid, there’s an eastern grid, there’s a Quebec grid, and there’s a Texas grid, meaning all these four grids are operating at 60 Hz, but they’re not synchronized with each other. So Alberta resides in the western grid, and Saskatchewan resides in the eastern grid. So you’ve got a situation here where Saskatoon is electrically connected to Miami and Fort McMurray or Edmonton is electrically connected to San Diego, but Calgary and Regina are not electrically connected.
Peter Tertzakian:
So, this also creates some of these cross-provincial issues.
Cory Paddock:
Yes.
Peter Tertzakian:
When we try to think about synchronization and the broader discussion which we won’t get into of achieving net zero, et cetera, by electrifying the country because the grid has evolved over North America in such a way that it’s challenging to politically synchronize, let alone physically synchronize.
Jackie Forrest:
Well, you can connect them though with different types of connections where they’re not connected in a way that-
Cory Paddock:
Yeah, so you can connect them with DC ties, direct current ties, which is essential, you can conceptually think about it like you can connect a power plant from Saskatchewan sending power into Alberta, but those flows can only go in one direction at one given time.
Peter Tertzakian:
Well, we don’t need to get deep into the physics other than to say that this conversation illustrates the complexity of trying to electrify everything in the pursuit of net zero. But let’s talk about now what also happens with this complexity when you start to radically change the composition of the supply, which is happening with the introduction of renewables, which as you said earlier, there are all sorts of different types of supply of electrical energy. Some are more intermittent than others, et cetera, et cetera. So now you have to start thinking about how each supplier comes into the market.
Cory Paddock:
Yep. We could talk about PGM, which is the biggest deregulated market in the world.
Peter Tertzakian:
That’s Pennsylvania, Jersey-
Cory Paddock:
Maryland. So this is an area that runs from New Jersey down through into Virginia, includes Ohio, and then also includes metro Chicago. So lots of different politics, lots of different fuel sources. And this is an area that has all the fuel sources that we think about in modern times. So there’s a lot of nuclear power. There’s still a lot of coal power. There’s gas generation. It’s sitting on the Marcellus. There’s a lot of wind generation and then we’re adding a lot of solar as well. So even since my career started, I’ve seen the grid evolve quite a bit where we’ve seen a lot of coal and gas switching, which changed the market quite a bit in that the entire transmission system was set up to deliver coal energy eastward to where demand is. And then all of a sudden gas generation shows up and starts to beat coal, and then all the flows and all the lines start to move in different ways that were originally designed. So you started to see different congestion in the grid that didn’t exist previously.
Peter Tertzakian:
So this is that kind of broad market and deep market, I’ll call it, because of the population base there, is sort of an in indicator of what potentially can happen to other markets such as their own. And I know, just as a side note, I find it intriguing that GBE Energy, which you co-founded, trades actively in those markets, so when somebody flips a switch on in Maryland, basically you have bought and sold it to them.
Cory Paddock:
Yep, this is true. Yep.
Jackie Forrest:
Deregulated markets can become efficient-
Peter Tertzakian:
Out of Calgary.
Jackie Forrest:
Yeah.
Peter Tertzakian:
Well, that’s amazing.
Jackie Forrest:
Well, let’s talk about how much is the percentage of renewables generation in that market. Like in Alberta, I think we’re at 12% of the generation. Texas, I think, is already near 30%. And have you seen any issues come from the greater amount of solar and wind coming on that is a little bit less predictable than those base load generations?
Cory Paddock:
I don’t think necessarily has issues. I mean it’s very different. And the operators, they adapt to this new situation, but these things also happen kind of gradually, and then suddenly you look back and it’s like, “Oh wow, actually a pretty significant amount of the electricity generated last year.” So I think when it comes to renewables there’s one thing to consider as well in that renewables are reliable to be what they are currently without modern computing. We don’t necessarily make those two associations immediately. And that has to do with a couple of things. One, we’re getting quite good at figuring out where to source these things based on historical weather data and historical cloud cover data, et cetera-
Peter Tertzakian:
By source, you mean to place them?
Cory Paddock:
Place them. Exactly.
Peter Tertzakian:
Yeah, yeah.
Cory Paddock:
Where do we have high solar capacity capability? Where do we have high wind capacity capabilities? And then weather forecasting has improved significantly. So we’re still pretty lousy at three to four weeks out. We know roughly what’s going to happen seasonally, but I would say seven days out we’re getting pretty good. So operators are already looking a week out and making their forecasts and making dispatch decisions around how the composition is going to work in that trial-timer
Peter Tertzakian:
So as a trader, you’re looking for potential volatility in those predictions between buyers and sellers, is that fair to say?
Cory Paddock:
Yep.
Jackie Forrest:
Now, some people have said that as we get more and more renewables on, even if you can predict it better, it’s going to create issues because some of those baseload generators are going to maybe not make as much money, not be needed as much of the day as they were before, or maybe these peak generators, although they might get high prices when they’re needed, they’re used less and less. So people have said, we’re going to have to move these free markets to some sort of hybrid. For example, Texas is now moving towards capacity payments, I understand, to create more incentives for those baseloads. So they get paid just for being there, even if they’re not generating. Do you think we need changes in the market design as we see more and more renewables to ensure that reliability is there, that and those baseload generators that aren’t used as much are still available?
Cory Paddock:
I don’t think we know exactly how we should be structuring these markets if you’re thinking about long time horizons. We’re still kind of trying to figure things out as we go along here. So I think a lot of the debate around capacity markets has to do with acute events that have never happened before. And the problem with acute events that have never happened before is, well, how do we forecast when those events are going to happen again?
Peter Tertzakian:
Like the weather, basically-
Cory Paddock:
Like weather-
Peter Tertzakian:
… like an ice storm in Texas.
Cory Paddock:
… so I think roughly speaking, we know generally when wind and solar are going to be around at a seasonal level, but after that, it becomes a little bit more difficult to forecast out where this is going to kind of land. So I think there are a few jurisdictions that are worth watching for the next several years. One is California, because there’s high penetration of solar. We’re seeing duck curves start to show up. This is becoming more of a common term now in the energy landscape.
Peter Tertzakian:
So duck curve is the curve of the surge of electricity demand in the morning. Then it kind of levels out and then surges when people come home in the evening and turn on their stove to cook and stuff like that.
Cory Paddock:
Yeah. So what we’re seeing is completely different demand curves than we’ve previously seen for the same weather and same time of year.
Peter Tertzakian:
Why is that?
Cory Paddock:
What you’re seeing is… Let’s just use the example of solar. So solar at its limit is peaking in the middle of the day at the same time that air conditioning used to peak, but all of a sudden air conditioning and solar are this amazing offset. So it turns out to be no load. And back to what your original point was, you’re seeing a little peak in the morning followed by a trough in the middle of the day, followed by another peak in the evening when solar’s coming off. But air conditioning is still hanging around because it’s still pretty warm and your buildings are warm. So the previous curve was kind of a gradual ramping up of fossil generation, let’s call it, and nuclear et cetera, to meet that load. And then it was quite a predictable thing.
And now what you’re seeing is this trough in the middle of the day that is also associated with really, really low prices and then a peak in the evening with quite high prices. And just to give you an idea of the scale, you could see a situation where in a matter of two or three hours you need to ramp up 20,000 megawatts or 30,000 megawatts kind of at the limit. It might even be bigger than that as-
Jackie Forrest:
Just in California, just in one state?
Cory Paddock:
Just in California, just in one state.
Peter Tertzakian:
So it’s the swing consumption profile that is causing issues and whether or not the supply can respond to that. And is it expected that the mismatch or potential mismatch between what we can supply to meet these swings in demand is that… Some people say, “Oh my god, this is going to be a huge problem.” Others say, “No, we can smooth it out.” So where are you? What are you seeing in the trading pit?
Cory Paddock:
It very much depends on the jurisdiction, I think. So in the example of California, conceptually we’ve been talking about renewables and storage together for a while, and it’s one thing to conceptually talk about it. It’s another thing to see some really low prices and see a price signal for storage to get a return. So now if you’re seeing all these days and years of very, very low prices in the middle of the day and an arbitrage, well, storage looks a lot more attractive.
Jackie Forrest:
Right. If you’re getting a very low price, you want to sell into that higher price-
Cory Paddock:
The cure for low prices is low prices.
Peter Tertzakian:
So let’s pick up on that because this notion is so important for price signals. Again, I sort of equate it to a market or an auction or whatever. You go and there’s a seller that says, “I’m not going to sell to you. You’re not offering me enough.” But as soon as the price goes up, the seller then at some point says, “Okay, fine. I’ll sell to you.”
Jackie Forrest:
Only if they were able to store those electrons though.
Peter Tertzakian:
In this case. So on the storage note, I mean, I want to go back to your comment. A couple of minutes ago you said, “We don’t know whether or not the supply side because it’s just evolving. Can you meet this?” One of the big wild cards, of course, is storage and home storage and electric vehicles as potential storage such as the Ford, what is it?
Jackie Forrest:
F 150-
Peter Tertzakian:
The Lightning.
Jackie Forrest:
… and other cars. I think California’s mandating that all cars need to be bidirectional by a certain time.
Peter Tertzakian:
So basically your car becomes a battery and we’ll see how that goes. I mean, how are you thinking about all that stuff as you-
Cory Paddock:
It’s becoming real. Yeah, it’s becoming real. And I think again, it’s important to think about what jurisdiction you’re in and where this can apply. In Alberta in the dead of winter, that is going to be very, very challenging and not going to work. But if you’re in the sunbelt and you’ve got quite reliable solar output, the economics, I think, are going to look more and more attractive over time to shave off these peaks from your home usage depending on the extent time of uof se pricing makes its way to the retail customer.
Jackie Forrest:
I think we should switch to a talking-
Peter Tertzakian:
AI.
Jackie Forrest:
AI, yeah. So one thing we haven’t talked about is one way to deal with all of these changes, whether it be the growing demand from EVs or the more volatile type of supply that’s a bit harder to predict when it’s coming on, is to use AI. Peter and I are already using AI in our research for energy, and it’s already making a difference.
Peter Tertzakian:
How do you know this is me here? You don’t.
Jackie Forrest:
We trained AI to sound like Peter and it’s amazing. Now that’s why he’s able to do 26 hours a day-
Peter Tertzakian:
Exactly.
Jackie Forrest:
… not like the rest of us. But ChatGPT is good. I have to say though, I don’t think it’s going to replace me. So I still have some job security. I’ve been testing it.
Peter Tertzakian:
No, not right now.
Jackie Forrest:
It’s not always right.
Peter Tertzakian:
You’ve heard of the term hallucinating where it makes up stuff and confidently reports it as an answer to your question. I find the questions that I ask the algorithm hallucinates a lot, in other words, dishes me a lot of nonsense that you better check, otherwise, it’s wrong.
Jackie Forrest:
Yeah, but I mean it’s right sometimes too.
Peter Tertzakian:
Yeah, that’s right.
Jackie Forrest:
I’ve seen some efficiencies. I think it gets 70% of there, maybe half of the way there.
Peter Tertzakian:
Well, it’s amazing the tool and other things, but check your work when you use it.
Jackie Forrest:
Well, so I use ChatGPT to help me with this podcast, and I said to it, “Make a list of ways AI could change electricity markets for our discussion.” And I have to say it did a very good job.
Peter Tertzakian:
Stuff like that is great.
Jackie Forrest:
So let’s go through the list, Cory.
Cory Paddock:
Okay.
Jackie Forrest:
It talked about energy optimization and efficiency. So it said that AI could be used to optimize energy generation, transmission and consumption. It didn’t give me any examples. That’s one of my value add here. I was thinking, well, one thing is shifting demand that we talk about that a lot, especially for things like EVs that could have a choice to shift the demand. And another thing I thought about is they talk about dynamic rating than transmission lines, that transmission lines could carry much more electricity than they do today, but we’re being very conservative in terms of how we’re rating them. And so those are two areas that could see big changes.
Peter Tertzakian:
Is that true?
Cory Paddock:
It is true. One way to think about this is we’ve got a lot of steel in the ground currently that is not fully optimized and building more steel in the ground is getting more and more expensive and-
Peter Tertzakian:
Or permitting is difficult.
Cory Paddock:
Exactly. So I think just the economics of optimizing the existing system will be preferable.
Peter Tertzakian:
So in the pipeline world, it’s like figuring out ways to optimize flow to get an incremental 100,000 barrels a day through the pipe.
Cory Paddock:
Yep.
Jackie Forrest:
Yeah. Hey, the Enbridge mainline, almost doubled their capacity or the amount they were flowing through these incremental improvements. And I think there’s a lot of, with AI, there may be a lot of opportunity with our power grid, and that’s the thing people maybe aren’t thinking about.
Peter Tertzakian:
To put more-
Jackie Forrest:
Trends through the same infrastructure.
Cory Paddock:
So we’ve got winter line ratings and summer line ratings already, but those measurements can’t happen in real-time yet. So this is one use case there, yep.
Jackie Forrest:
I will put this article in the show notes. I found this interesting article in Vermont, and it’s a bit about this shifting of demand, and we talked a little bit about consumers getting economic value for doing this. But this company is helping to pay for people’s power walls as long as they can control them. And the article says that there’s like one week they saved a whole bunch of money just because they could use the power wall of the people and that storage allowed them to avoid a bunch of bottlenecks in their system. So I think this is all just starting, but I think AI can enable a whole bunch of decisions to be made over a whole bunch of power walls that one human couldn’t have done.
Peter Tertzakian:
So what you’re saying is I have a Tesla power wall in my house and that the utility that supplies my electricity can tap into that, unbeknownst to me, and pull on my battery to power a neighbor who’s pulling hard on their electric car.
Jackie Forrest:
Exactly. And maybe the neighborhood-
Peter Tertzakian:
Charge-
Jackie Forrest:
… grid, can’t-
Peter Tertzakian:
Right.
Jackie Forrest:
… be at that level. So they start to use the power walls from the other houses to meet that-
Peter Tertzakian:
Basically, as an aggregate all the power walls are in one giant battery and kind of managed.
Cory Paddock:
But what groups are doing fundamentally is doing the math on, “Oh, should we do these power walls or should we build new transmission?” And they’re saying, “Oh no, this is cheaper and it gives us more visibility and we can manage our grid better.”
Peter Tertzakian:
To the extent that people want somebody else to control the electricity in their house.
Jackie Forrest:
But if you got it for free… I talked to someone from Vermont, and he has two of them, and he’s like, “It’s great,” because now his solar is there and he’s backed up all the time. And so for him, it was like a no-brainer. They have control of using it when they want, but he gets a lot of benefits too.
Peter Tertzakian:
It’s a no-brainer until such time somebody turns into some conspiracy theory. So communication of what is being done is paramount, in my opinion.
Jackie Forrest:
Well, let’s go to the next idea that ChatGPT had, which was-
Peter Tertzakian:
Okay, let’s hear it.
Jackie Forrest:
… renewable energy integration. So basically saying that AI can do a better job of analyzing weather and predicting when we’re going to get renewable generation. So Cory, do you think that, is happening already? Is there room to improve that still? What are the benefits?
Cory Paddock:
That has been happening for a decade.
Jackie Forrest:
With AI or just other systems?
Peter Tertzakian:
Well, I mean, weather prediction is AI in many ways.
Cory Paddock:
A lot of the work that’s been happening in weather forecasting for years has filtered into the energy systems in ways that we don’t necessarily talk about. Just based on what I was talking about before.
Jackie Forrest:
So this morning you’re trading electricity in the Midwest. Are you getting a prediction for Chicago’s sun and how much solar generation is going to come because of that?
Cory Paddock:
Of course.
Jackie Forrest:
Are people doing that?
Cory Paddock:
Of course.
Peter Tertzakian:
Yeah.
Jackie Forrest:
How accurate is it?
Cory Paddock:
Pretty accurate. Yeah. But I mean, this is still on a day-to-day basis, there are still plenty of anomalies. So if you’re thinking about the capacity factor of a solar farm over a year, we can get pretty good accurate readings. But then once you get into the kind of granular hour-by-hour, there were obviously differences, but we’re getting better at predicting that for sure.
Peter Tertzakian:
Sure. And what about managing or, I don’t know what the word is, reconciling the gut feeling of a trader versus the gut feeling of an algorithm in this?
Cory Paddock:
This is a good question. Yeah. I guess going back into any kind of trading day, you’re always fundamentally looking at all the fundamental data that can arrive at a price. And sometimes supply wins out, sometimes demand wins out, and you just get this gut feeling around a particular day and which is the factor that is going to move the needle in one way or another.
Peter Tertzakian:
And which, in that regard, put your money on the table.
Cory Paddock:
Yep.
Jackie Forrest:
Well, this gets me to your job, Cory, because ChatGPT also wants to take your job. It’s telling me that it’s going to be useful for energy trading and market optimization to support energy market participants in making informed decisions. Well, I mean eventually maybe it could just take over their jobs though, right? So give me some examples of how it might help you with trading, and do you think that there’s a future where the algorithms are better than the people?
Cory Paddock:
I have no comment.
Jackie Forrest:
You’re conflicted in this answer.
Cory Paddock:
I was going to… No. We’ve been thinking about this for quite a while. It’s a pretty complex thing to wrap your head around. I think just in general, AI/ML suffers from this bandwidth issue where practitioners and the folks that have all this tacit knowledge don’t necessarily have the technical knowledge and back and forth, so it’s actually about increasing pipelines between those two pathways to get stuff to function better.
Peter Tertzakian:
Yeah, I think this is the thing, that there are differences of… There are going to be big differences in AI capability in these different regions and sub-regions in the markets, and each region and the traders in each region are going to be using different data sources, which in doing so, will create anomalies and the arbitrages that traders work off of. Is that fair to say?
Cory Paddock:
I think it’s fair to say, yeah. But it’ll help calculate much bigger data sets than previously possible and compare data sets that weren’t necessarily being looked at together at the same time before.
Jackie Forrest:
Increase the efficiency then. Maybe you need fewer traders and they’re more informed than they used to be, potentially. Right? Or fewer people in the back room.
Cory Paddock:
Yeah. The arch of history for a lot of-
Peter Tertzakian:
Yeah, I mean, I’ve studied and worked with predicting data using algorithms for most of my career, and the adage of course was garbage in, garbage out. So to what extent is the data going into these models cleaned up sufficiently in your mind these days relative to where we were, say 10 or 15 years ago? Has there been improvement in such that there’s less garbage in?
Cory Paddock:
Yeah. Yeah, I think so. But at the same time, I think you brought up a good point in that AI does a pretty good job of explaining what happened already, but does not do a great job of explaining what’s going to happen.
Peter Tertzakian:
Right. Well, getting back to my comment about how AI hallucinates on qualitative stuff, makes stuff up. Well, it could make stuff up easily and numerically, which can have major consequences for your wallet as a trader. Right?
Cory Paddock:
Yeah. It would do poorly in tail events. Put it this way.
Peter Tertzakian:
Tail, as in rare sorts of ice storms and things like that.
Cory Paddock:
Yeah, exactly.
Jackie Forrest:
All right. Well, let’s talk a little bit about how demand, everyone knows demand’s going up, and it depends on if you think we’re going to electrify heating and how many EVs are going to come on, but what about computing power? There’s already a whole bunch of server farms being built, and could there even be more because of the use of AI? Could AI actually… You see that with Bitcoin in Texas, how it added a material amount of demand? Could that be another issue here be that demand and supply get further apart because of this growing use of AI?
Cory Paddock:
Yeah, this is an interesting one. I don’t think it’s very well intuited just how much electricity is being used for data centers and data transmission. I’ve seen figures that are anywhere from two and a half percent to 4% of total electricity demand worldwide, which is a pretty big number. It’s like the load of 50 Calgary’s around the clock, which is powering the data complex which didn’t exist 15 years ago.
Yeah, it’s still growing. To the extent it’s growing, I don’t know, but it’s going higher.
Peter Tertzakian:
Well, if you look at the management discussions around the companies that build server farm facilities, because a lot of this stuff is contracted out. I mean, they are growing like compounding. I don’t know, it’s like 20% or something. It’s just ridiculous how fast they’re growing.
Cory Paddock:
They’re growing fast.
Peter Tertzakian:
And so this tells me that the demand is going up. I mean, the chip makers are going through the roof in terms of their stock prices and the expectation that this is going to create even more data computing power. Just makes sense. And I mean, I think you did some numbers in terms of how much electricity is being… Well, you just said, 2 to 3%, but it’s five or six times the amount that is currently being used to charge up electric vehicles. That’s another way of thinking about it.
Jackie Forrest:
Hey, I asked ChatGPT if I should be concerned by the amount of electricity needed because-
Peter Tertzakian:
Oh, did you? Okay. Well-
Jackie Forrest:
… and actually, it said that they’re continually improving and getting more efficient and that over time you’re going to be able to do more with less electricity just like everything else, right.
Peter Tertzakian:
Interestingly, it would bias towards itself like that.
Jackie Forrest:
It didn’t seem a problem.
Peter Tertzakian:
But this is to me, like a lot more base load demand because these server farms are just whizzing away answering people’s questions and controlling things.
Jackie Forrest:
Well, there’s probably a pattern to it though like during the workday it might be higher, in the evening, it’s less, you’d think.
Peter Tertzakian:
But there are time zone shifts around the world, like-
Jackie Forrest:
Oh yeah, globally. That’s true. The server farm could be anywhere in the world. I did one example. If you just Google search something, but then you ask the same thing to the Chat, it takes quite a bit longer to come back, which just makes me think-
Peter Tertzakian:
On Chat?
Jackie Forrest:
… there’s a way more computational… Yeah, on the Chat… More computational energy is being used to come up with the same answer.
Peter Tertzakian:
Wouldn’t doubt it. Yeah, it makes sense. Okay-
Jackie Forrest:
I think we should wrap up.
Peter Tertzakian:
Well, this has been an amazing conversation. We can talk a lot more because it intersects with commerce and philosophy, and I love talking about all that stuff, and I know you do too, Cory. Thanks so much for the time for coming in. We should let you get back to your trading desk and make your bets. Thanks so much for coming in.
Cory Paddock:
Oh, thank you. It’s a pleasure.
Jackie Forrest:
Thank you. It was a great discussion. And thank you to our listeners. If you enjoyed this podcast, please rate us on the app that you listen to and tell someone else about us.
Announcer:
For more ideas and insights, visit www.arcenergyinstitute.com.
