Energy market design in the US with Gary Cate (Director of Energy Market Analytics @ Fluence)

- The market shouldn't pick winners and losers. It shouldn't pick a singular technology. I don't think we should come in and say, we need x amount of this generation or type, coal or gas, or even storage, or wind, or solar. We should say, we need these type of attributes. We need something that has ramping capability. We need something that has frequency response capability. We need voltage response, grid forming, those type of things.

- How do current market designs support or not support battery energy storage.

- There's all kinds of things that can happen on the grid from the time that you cleared the day-ahead market to the time that you're actually delivering that power. If you just clear those ancillary services in the day-ahead market, you've positioned those ancillary services on assets that you may not want ancillary services on.

SPEAKER: Hello, and welcome back to Transmission. The US has multiple system operators, serving different areas of the country, each encompassing a unique mix of generation types. These differences and the variation in demand across each area means the market design of each operator differs too.

In this episode, Gary Cate, Director of Energy Market Analytics at Fluence, joins Q to discuss market design in the US. As always, if you are enjoying the podcast, please hit Subscribe, so you never miss an episode, and give us a rating wherever you listen. Let's jump in.

- Hello, Gary. Welcome to the podcast.

- Hey, how are you.

- Yeah, very well, Thank you. Thanks for joining us today.

We're going to talk about Fluence, where you work, and a little bit about the work you do around US markets. But before we get started, do you want to just explain your background. I know you've been in the energy industry for a long time.

- Yeah, about 17, 18 years, I don't like to think about how long it is, but electrical engineering background, started at Southwest Power Pool, essentially straight out of college, was working on SCADA AMS systems initially. And then I found out pretty quickly I don't really dig the hat, so moved pretty quickly over to markets.

We were implementing an energy imbalance market at Southwest Power Pool that went live in 2007. So that was my first taste of energy markets. And then, essentially, pretty quickly, after we started doing the energy imbalance market, we started moving into a full day to real-time TCR, FTR market that went live in 2014.

So it was the tech lead for a lot of that and then move more into a management role, doing market design, basically setting market policy for Southwest Power Pool, worked with a lot of the other RTOs ISOs. And then the last thing before I moved to Fluence that we were working on was all the markets in the west.

So we had put in an energy imbalance market in the front range utility area in Colorado and was helping lay some of the groundwork for the markets in the west that SPP is working on now markets plus and RTOS.

- All right. And it would be rude not to talk about SPP for a minute. So for the audience, what is SPP, and where is it?

- Southwest Power Pool is an RTO. It's right in the middle of the United States. It's to the left of MISO, to the right of CAISO.

Basically, you've got the Colorado Mountains, the Rockies, cuts right there, where your DC ties are. It goes all the way down into Texas, all the way up into Canada, Oklahoma, Missouri, Nebraska.

Peak load is around 55 gigawatts, heavily wind. It's usually going back and forth between them, CAISO and ERCOT about who's setting the most renewable grid record that day. It all depends on how the wind's blowing or the sun shining.

- And then, of course, you've been in the battery world for a few years now. So I'm sure that most people know what Fluence is, but it'd be good to take it from the top. What's Fluence? How long have you been at Fluence, and what does the business do?

- Yeah, I've been at Fluence around three years now, heavily focused on energy storage from a asset operation perspective. But Fluence, as a whole global market leader in energy storage, both from a standalone storage perspective, but also a services and an optimization software perspective. So we're in over 47 markets, globally, like I said, standalone storage.

We have an Aspera, which is an asset performance management software that's doing a lot of AI around predictive maintenance. And then we have Mosaic, which is our bidding software, which we're using in CAISO, ERCOT and the NEM market in Australia. And NEM is optimizing both renewables and storage. Right now, our focus in CAISO and ERCOT is primarily just standalone or co-located and hybrid. We just released a hybrid model for CAISO energy storage.

- And so a lot of people know Fluence from, well, what were the original cube batteries, but Fluence is a OEM, original equipment manufacturer for battery energy storage systems, but also has some software businesses. You have the Mosaic, which gets you into the markets and your asset management software. For the listeners, what proportion of Fluence is--

is it mostly building batteries and being an OEM, or how big are the software parts of the business?

- Initially, it was certainly more focused on OEM storage and then it's grown significantly, moving towards both software for the battery management side, software for services as well, and then software for energy bidding. So those are growing, and the software side is growing a lot. Because, I think, what we're seeing a lot is people want turnkey battery solutions.

They want to--

for them, it'd be great if they could go to one vendor, and you get your battery. You get your EMS., you get your OS. You get your performance management software.

You get your market bidding software. So we're really seeing this ecosystem where people want one singular manufacturer company to go to, which is really neat from a Fluence perspective.

Because that's been really cool from a software perspective, just all of the data that we have. So being an OEM, building the actual batteries, and then having all that software where we can pull data from degradation curves, degradation of form bidding. There's just a lot of things you can do when you have all this data in a single ecosystem. So it's been very neat.

- Today is all about market design. And the idea of market design is quite an unusual thing. I mean, most markets just exist, right? Someone builds something, and someone else--

someone makes something, and someone else buys it. But in the electricity markets, there's a lot of thought that goes into designing markets to get certain outcomes or behaviors or investments. So when we're talking about market design, could you just talk about the different structures that they have in the US, maybe zoom in on a couple of them and key differences in design of markets in electricity markets.

- Yeah, I mean, it's interesting because they're all fairly similar from a structure perspective, as far as, generally, you're going to have a day-ahead market, a real-time market. Some people have a market in between. They're like a 15-minute market, like CAISO has, but they all go about clearing that market and the products that they're asking for--

It's not just an energy market. It's an energy and ciliary services market, and each RTO ISO has implemented those in a different way, and rightfully so, because each market has intricate differences, either from a fuel mix or a load growth or a load shape perspective. So a long time ago, there was a lot of calls, I think--

and you hear it come up from time to time of the central market design of all the markets in the US should operate very similarly.

And they do, structurally. But again, what happens in CAISO is different than what happens in ERCOT. The shape of the load is different. The mix of renewables is different. So they all do need somewhat of a different design. And I think that's worked fairly well, I think, for us in the US.

- Gary, if you could, in a nutshell, let's talk about ERCOT. So what's the market design of ERCOT? And what behaviors or what outcomes does Texas want in ERCOT?

- Well, I think that fluctuates, depending on the legislatures that are involved. But generally, it's more of a real-time market. It's just an energy market. There's no capacity market there, like you see in some of the Eastern Interconnect RTOs. There's no RA concept, resource adequacy concept, like they have in CAISO, where that's a significant portion of the battery revenue.

So essentially, the majority of your revenue in ERCOT is coming from day-ahead/real-time market. And generally, most people heavily participate in the real-time market. So it's very much real-time focused. I think if you look at any time there's an event in ERCOT, with the price spikes, a lot of people see that as an issue.

A lot of people see that as a feature of the ERCOT market, because that's what's creating asset revenue.

Certainly, there's less price administration. I think you see in ERCOT than you see in other markets, the curve for energy prices is allowed to go much higher than a lot of the other markets. And you see that in operations, but it's generally a well-functioning market.

- And how do--

if we just think about battery energy storage now, how do current market designs support or not support battery energy storage?

- That's a good question. So I think, generally, all of the markets are moving in a direction to support energy storage. Obviously, you've got the two main markets with the highest penetration of storage, and that's CAISO and ERCOT. And batteries are performing very well in CAISO and ERCOT, heavily reliant both in CAISO and ERCOT on battery storage, either from an ancillary service perspective or peak energy, ramping at the peak.

I think, CAISO takes a more heavy handed approach in how they do battery management. So they have state of charge management where they'll step outside of the economic algorithm. If you have ancillary services, and they may move you in a direction that's counter to the market price to make sure that you can provide your ancillary services.

They do a lot more checks on ensuring you have enough state of charge to meet your ancillary service position.

If you look at ERCOT, it's a little more nascent in the design, right? They're moving in that direction. And I think if you look at a lot of the NPRRs that are out there, ERCOT is going to move.

And I think everyone's going to move to a design that's a little similar to CAISO. They're going to put their own flavor on it, but CAISO has learned there has to be some checks and balances from energy storage perspective. And I think that comes a lot with how the market is. It's a new market. Energy storage is new. Asset operators are new to this.

And so state of charge management and how to bid in the market, everyone's figuring that out. And sometimes, that can lead to not great outcomes from a reliability perspective. And that's one of the things that CAISO has done is, they're putting in checks and balances to make sure that the reliability of the grid is maintained.

- So we're talking specifically here about reporting of state of charge to the TSO and also ensuring you have enough energy in your asset to deliver on your obligations. And then, of course, there's a question of penalties and how that is managed. So how do they do it in CAISO then?

- Yeah, so from a CAISO perspective--

and it's changed over--

and this is one of the things I want to touch on, too, from not necessarily a negative, but just something that happens when you're making rules for energy storage. But the way CAISO handles it is, they're constantly checking your state of charge from a ancillary.

And this is all related to ancillary services, for the most part, right? Because you're procuring those ancillary services. You need those ancillary services to ensure grid reliability. So they want to make sure that those ancillary services that they're clearing are actually deliverable.

- And these are the frequency response reserve products, that kind of stuff?

- Exactly. Regulation up, regulation down, frequency response. Spin, not so much. But generally, your regulation up, regulation down, and fast frequency.

Fast frequency for ERCOT, regulation up and down for CAISO. But the big change in CAISO, recently, was--

and this is something I see eventually coming to ERCOT as well was when you're clearing in the day ahead market.

So if you look at CAISO and ERCOT, the overwhelming majority of your ancillary services are cleared in the day-ahead. market. ERCOT doesn't have real time co-optimization. CAISO has 15-minute real-time co-optimization, but it's still not full real-time co-optimization.

- And what does that mean, real time. Just so we make sure we are clear on this.

- When you're clearing ancillary services--

like let's take the day-ahead market, for example. When you're clearing ancillary services, there's this constant co-optimization between energy and ancillary services. So you're making sure that you're taking a position in energy that can still support your ancillary service position.

And then from a pricing perspective, it's ensuring that no resource is worse off for clearing ancillary services, right? So you don't want to go into a market and clear someone for ancillary services. And they lose money versus what they could have made on energy. So there's this loss of opportunity that goes into the ancillary service price setting.

So it's not just your bid in price, it's your bid in price plus the loss of opportunity from maybe not doing energy. Because if you take, for example, energy price spikes to $1,000, and you've bid $9 to do regulation up, well, I don't want to clear regulation up. That's a much worse outcome for me. So you have to have this loss of opportunity that says, OK, they're going to make $9 for regulation up.

They could make $1,000 for energy, and that pushes that ancillary service regulation price upwards. Now what happens in a full day-ahead real-time co-optimization is all that we just talked about in the day-ahead market, that is pushed down into real-time. So when you're doing this 5-minute dispatch, some markets--

like, SPP, for example, MISO, PJM, they're taking and reclearing those ancillary services.

So it's not just what you clear in the day-ahead market. It's recleared when you go into the real-time market. And what that helps is, it's a much more efficient way to clear ancillary services. Because when I run the day-ahead market, obviously, it's a day ahead, right? And we know there's a massive amount of changes that happen from the day-ahead time frame to real time.

You may have a generator trip. You may have a weather event roll through. There's all kinds of things that can happen on the grid, from the time that you cleared the day ahead market to the time that you're actually delivering that power. If you just clear those ancillary services in the day-ahead market, you've positioned those ancillary services on assets that you may not want ancillary services on.

You may have congestion, or maybe you really need this asset for energy, and you need another asset for regulation. And so the way the operators in non-5-minute real-time co-optimization markets handle that is it's typically throughout of market actions. You, as an operator, will physically say. I don't want this asset to provide regulation.

So you kick them out of the regulation market, and you clear that regulation on someone else. And so that's where I see ERCOT moving is. They're going to have--

I mean, actually, it's on their roadmap. They're going to move to real-time co-optimization, which, I think, will create a more efficient market from an ancillary service perspective.

- One of the interesting and fun things about electricity markets is that they are evolving all the time. So you can't design a market and set and forget it, and especially because there's so much renewables being built. And the way that these electricity markets are operating has to change with the generation mix. So as there's a lot more renewables coming onto the system, what are the key considerations for market designers?

- Yeah, that's a great question. I think the biggest thing that's made sense to me, over time, certainly moving to the other side, moving from an RTO ISO to an asset optimizer, it has to--

A, you have to have a focus that's not just--

I mean, you want to talk about energy markets, right? And that's what your question is. But it goes deeper than just pure energy market, right?

It goes all the way back to the planning process of how are we deciding what generation gets built? How are we deciding what transmission gets built? How do we have this cohesive plan? Because I think you see a lot today, where there's definitely disconnects in those plans, where, OK, we're getting a lot of renewable generation built, but we're not getting a lot of transmission built.

Or we're getting some transmission built, but we're not seeing the buildout of a certain type of asset. And so I think, A, you have to take this holistic, cohesive look at both generation planning, transmission planning, and energy market design. The other thing there is, I don't think--

and this has shifted back and forth over time, but the market shouldn't pick winners and losers, right?

It shouldn't pick a singular technology. I don't think we should come in and say, we need x amount of this generation or type, coal or gas, or even storage or wind or solar. We should say, we need these type of attributes, right? So it's more of an attribute-based market. We need something that has ramping capability.

We need something that has frequency response capability. We need voltage response, grid forming, those type of things. So I think breaking it down into the individual components of what you need to run the grid--

I mean, obviously, you need energy, and you look at ELCC and a lot of things they're doing from resource adequacy perspective of ensuring there's just enough energy on the grid.

But then, I think, when we do planning and people are building out assets, and we build these energy markets, we need to bring our focus back into individual capabilities of the asset. I know when I was at Southwest Power Pool, for a long time, we thought it doesn't really matter, right? So you're just, we need something to respond, right?

And we can group those in, just these big homogeneous products where we need a response upward. So we just get a upward product, or we need a response downward, so you have a downward product. But when you look at that, over time, that doesn't really create the efficient outcomes that you want. You need this more specificity so that people will design to those specific needs that you want.

So if you need something to respond in 4 seconds, people can bring something that can respond in 4 seconds to the market. If you need a 4-hour response, people will bring four hours to the market. You need eight, people will bring eight. So I think that's, to me, the big focus we need to have from a how we build the grid of the future.

- And what about policy? What are the big policy hot potatoes that are currently being discussed that are especially important for owners and operators of battery energy storage?

- I think a lot of the biggest ones are on accreditation of how do we treat these batteries. Do we treat them like it's a conventional asset? Do we treat them like it's a special asset? In some cases, you want to treat them like it's a special asset. In some cases, you don't, right? So I think that is one thing. I think, to look into how much storage do we need?

How long of the duration storage do we need from a policy perspective? So a lot of people say, 4-hour storage is enough. We don't necessarily need long duration energy storage. But then, you look at the pricing models, and you look at the advancements that we're having, we can already see 5- to 6- to 7-hour batteries becoming profitable in CAISO, and that's going to make sense long term.

So I think figuring out that mix and those market products that you can mix a 4-hour set of assets with a 6-hour set of assets and a 8-hour set of assets, I think that's the big policy shift that I think you're going to see over the next 5 to 10 years is, we have to expand the markets backwards, in my opinion.

So a day-ahead market, when you're heavily, solar, wind, and battery storage, I don't think a day-ahead market is necessarily enough time to get all of that prepped and prepared and ensure that you have the right mix for that day. So looking back in these longer term capacity products, especially with battery storage, right?

Because battery storage is reliant upon the grid to charge, right? So there is--

it's not like--

I mean, everyone listening to this podcast knows this, but batteries charge from the grid, they're not taking fuel, like gas and coal, right? So that planning of where you're going to set your generation from a day-to-day basis, you've got to start thinking about moving that back.

There may not be enough load. You're essentially trying to flatten out the peaks. But how you do that over a one or a two-day window is going to be interesting. And I think you're going to see markets build longer-term products.

Most of our products, you look at--

ERCOT has ECRS. That's two hours, right?

Something's going to be moving back into the 4-hour time frame, the 8-hour time frame. So I think you'll see this shift in energy markets to trying to figure out how do we build these longer-term capacity products in markets, but dealing with the fact that everyone is structured in this day-ahead market, real-time market construct.

- So if I've got this right, your take is--

or one of your takes is that you think the duration of these products will get longer, but there'll be more and more procured closer to real time.

- No, I definitely think the duration of the products will get longer. You'll see some shifting in real time, but I think because the duration of the products needs to get longer, you may have to expand past a day-ahead market, where you're running some other process prior to the day ahead market that's an economic clearing.

So you've got--

you're setting one set of batteries of, hey, you're my longer-term capacity battery. You're my short-term frequency response type battery. And so you're going to see different products, different frames.

And I think you're going to see a shift from a lot of these products. We just talked about it earlier of loss of opportunity.

And that's how a lot of these products, some of them are bid in.

Most of your ramping products are just loss of opportunity. So what was the difference between what you could have done in energy and what you're doing in this ramp product? I think you're going to see a shift to all of these products are going to start being biddable.

So you need to shift away from loss of opportunity. Because loss of opportunity doesn't actually represent the value that those products provide.

And so is it really fair that all I get is just the loss of opportunity, and I'm providing you an essential reliability service? So I think you'll see a shift to products are going to have to be biddable so that we can have an efficient market outcome for everyone that's involved.

- All right. And now onto the finishing questions. So the first one is, this is your opportunity to plug something. Is there anything that you or Fluence wants to get out there to the world of our listeners, which are generally battery people?

- Yeah, I think it's not necessarily--

I mean, certainly something I'm working on here, we're working on at Fluence, but it's really just been interesting to me to, even in this 2 and 1/2, three years that I've been here, see this shift. We're certainly seeing a shift away from pure AI to this concept we're calling and others are calling human-in-the-loop trading, right?

So we talked about earlier how people want a turnkey single asset solution, but they're wanting that same type of thing from a trading perspective and a bidding perspective of--

I think, initially, everyone saw, OK, batteries are complicated. It's a very complex algorithm, both from a forecasting perspective to a bidding perspective when your fuel is a part of the market cost, right?

So we went very quickly and said, OK, we're going to create AI algorithms to do this because a human can't do these things. And then it went very hard on the AI side, but then you start to see, wait, OK, humans still know a lot about energy trading, and they know a lot about pattern recognition. Now, granted, computers typically find patterns better than a human.

But from an energy market perspective, I think we've started to see, humans are pretty good at this as well, right? They've been doing it for years. And so we're really starting to see this drive and push to where it's human-assisted AI trading. So very much reviewed, reviewing the outcomes, feeding human input into the algorithm, and then also self-learning algorithms. So we're working on AI self-learning of that outcome wasn't what I wanted.

Let's feed in the fact that we had a bad outcome and feed that back into the algorithm, so it knows, hey, this was a bad decision. And so that's happening both from an AI perspective, but then also from a human-in-the-loop perspective of we're telling it. That was a bad decision. Here's what you need to do to move forward.

So it's really been a shift that--

and we've seen a lot of promising data from more human interaction with the algorithms.

- OK. And now on to my favorite question, which is what's your contrarian view? What's the thing that you believe, Gary, that not necessarily everybody else believes?

- I think long-term power price forecasting is an exercise in futility.

Granted, I know we have to do it, but if you look at how quickly the grid is changing now, I just don't think any of the price forecasts have the accuracy that we are typically used to. It's a grand thing to work on. And obviously, it's important.

But if you take ERCOT, for example--

and this is using some of your data actually from a Modo index, Right? So battery storage revenue in 2022 was one 140, 145k megawatt.

2023, it was like 196.

And this year, it's shaping up. It's going to average around maybe 90 to 100. And that's a huge shift from--

it's basically been this upward curve.

And now it's taken a huge dip. And that's because ERCOT put on, I think, 11 gigawatts of generation in six months, right? From the start of this year till June, they had added like 11 gigs of generation. You can't forecast that with any accuracy. And the thing that I see is very difficult. And this goes all the way back into the first of our conversation of, OK, the grid is changing rapidly.

We're trying to change policy rapidly.

And that, to me, is the hardest piece, maybe, to get in a price forecast is policy. If you look at the ERCOT change that's coming--

I forget the exact NPRR number, but earlier release of ECRS capacity, right? That is going to really dampen those ERCOT price spikes.

And so that, alone, will create a pretty big shift in how the ERCOT, certainly from a peak perspective, prices are going to look. And so that's just really hard to get into a policy model for forecasting is difficult. And so I think when you look at how we're moving forward, we're going to have to find new ways to come up with these price forecasts.

You can run a million Monte Carlo simulations, but that only gets you a little bit of the way there because there's still actually a singular outcome that's going to happen. And so I know we have to have long-term price forecasts, but I don't think right now the long-term price forecasting is providing the value that 5 or 10 years ago, it was providing. Because the grid's just changing too quickly.

- I would love to spend some time talking you through how we do our modeling, because I think I can convince you otherwise. But there are some things that--

so the details, you're right. So the details of policy changes like that are nigh on impossible to predict. Agreed. But there are certain themes which, I think, will always be true, which is that market designers will aim to remove efficiencies in the market wherever they are that cause unusual price behavior.

And we've seen that across Europe, and now the US markets we're looking at. Whilst the details of the policy changes, are very difficult to predict, you can, with a fairly high degree of accuracy, figure out that where there is--

where the market is inefficient and causing huge prices, that either the regulator or the market designer or RTO, whoever in that particular jurisdiction is responsible for it, will take measures to reduce the impact of that.

But yeah, I agree with you, very, very difficult thing to do.

But, Gary, I would implore you to come and spend some time with us. We'd love to talk to you how we do modeling. I think we can change your mind.

- Yeah no, that'd be awesome. Yeah.

- All right, Gary, I want to say a massive thank you for coming on the podcast. It's been a delight to have you on, and thanks for talking to us about market design.

- You're welcome. I really enjoyed it.

SPEAKER: Thank you for listening to Transmission, a Modo Energy podcast. Transmission delivers conversations from industry leaders and experts, exploring energy markets and the operations and technologies related to grid scale, battery energy storage. Check out our other episodes by searching Transmission wherever you get your podcasts. Check out the Energy Academy, our video crash course on how markets in Great Britain and ERCOT work.