The Curtailment Paradox

The Curtailment Paradox

Show Video

All right, it's a few minutes past the hour,  so let's go ahead and get started here. Thanks,   everyone, for joining, and good morning from  Colorado. I don't know what time zone everyone is   in, but glad you all could join. And we're gonna  be talking about the solar curtailment paradox.   My name's Bethany Frew, and I'm an engineer  at the National Renewable Energy Laboratory.  

I've done work with capacity expansion modeling,  production cost modeling, with a more recent in   trying to couple those together to be able to  explore wholesale electricity market design   questions. That's kinda high-level areas of  interest, but today's talk about curtailment   is really looking at flexibility in the system and  how, as we evolve to futures with high levels of   PV and batteries, what sorts of interactions we  see between different components of that system   specifically related to curtailment.  Make sure I get the slides here –   there we go. So this talk is drawing from a  paper that was recently published in Joule,   "The curtailment paradox and the transition to  high solar power systems." The link is on the   screen. An amazing team of co-authors; many of  them are on this call as well, and may help with   questions at the end: Brian Sergi, Paul Denholm,  Wesley Cole, Nathaniel Gates, Daniel Levie,   and Robert Margolis. And as always, grateful for  the funding from the US Department of Energy,  

in this case the Solar Energies Technologies  Office, who provided funding for this work. So what is a paradox? Some of you may know,  but this actually came up as a question   in our review process. The definition that I like  and I think is most applicable to this particular   paper and talk is "a seemingly absurd or  self-contradictory statement or proposition   that, when investigated or explained, may prove to  be well founded or true." So basically something   that at first doesn't really seem to make sense,  but as you dig into it, you can start to explain   what's going on underneath. And so in this paper  we found two pieces of what we're calling the   solar curtailment paradox. The first is that  thermal generator parameters that relate to  

flexibility, especially when we restrict minimum  generation levels and ramp rates, impact variable   renewable energy or VRE curtailment more in  mid-PV contribution levels than at lower or   at higher contribution levels. In our particular  system, that mid-level was about 25 to 40 percent,   and then the lower levels were about 20 percent,  and higher, about 45. But those exact numbers may   not translate to every system. The point is that  there was sort of this wider band in the middle,  

and I'll dig into this later in the talk. The  second piece of this paradox is that allowing VRE   and storage to provide operating reserves results  in reduced operating cost and curtailment from a   system-wide perspective, but the price suppression  effect from these resources participating   reduces the incentives for PV to provide  operating reserves with curtailed energy.   So in other words, this price suppression  effect really kind of degrades any sort of   monetary compensation that these resources might  get for the benefits that they're providing from   a system-wide operating cost and curtailment  perspective. So I'll dig into each of these in   a few minutes here, but I wanna give a little  bit more context and background first before we   do that deeper dive. So how did we get to this  point with these two curtailment paradox pieces?   We wanted to systematically explore the impact of  flexibility options on curtailment levels as PV   contributions grow. For example, how operational  flexibility of thermal generators impacts   curtailment; the role of battery storage; the role  of transmission constraints, et cetera. And so we  

assess the potential value also – and this is kind  of that second curtailment paradox piece – the   potential value of allowing VRE and storage to  provide operating reserves with that curtailed   energy. So basically let's understand how, from  a very systematic perspective, various factors   impact curtailment levels as we have more PV and  batteries on the system; and then let's explore   at least one way that we could try to use that  curtailed energy for an additional value stream.   Existing work in this space has looked at some of  the factors that we explored, but not all of them   in this sort of systematic comparison, and  typically most of those studies were at much lower   PV contribution levels, generally around 5 to 20  percent. There was one study, notably, that had  

about 50 percent solar that explored the impact  of minimum generation levels and start-up cost,   but that was a more narrow kind of focus study. So  why the focus on curtailment? Curtailment happens   whenever generation exceeds demand and/or system  conditions impose certain operational constraints   that prevent available energy from being utilized.  There's a lot of different factors that can play   into this, but some of the sort of leading reasons  are transmission congestion; minimum generation   levels of thermal generators or hydropower, or  back-feeding in the distribution system. And   there's lots of literature out there if you're  curious to learn more about kind of generally   speaking curtailment. But we see curtailment  really being a new normal in grid operations,   and this is something that we've actually  published on in the past, which is shown in the   paper on this slide, where economic curtailment  – and when I mean economic curtailment, I mean   it's part of the least cost operations of the  system, where curtailment is actually contributing   flexibility to the system. It's really a tool that  the operator has at its disposal. And so this is  

really becoming a new normal in grid operations,  again by helping to provide flexibility to ensure   reliability. And in this previous work that  I mentioned, we found many hours of very high   levels of instantaneous curtailment, above 40  percent in these high PV penetration cases. So why the focus on PV? Many of the studies, both  in the US and globally that are looking at sort of   projecting or forecasting what deployment of  various systems will be in the future, especially   out to about 2050, are projecting for PV to be the  largest renewable technology deployed out to that   horizon. So basically we're anticipating a very  large growth in PV deployment. And then also,   because of the coincident nature of solar,  where you have sort of that restricted sunrise   and sunset element, there's a more rapid increase  in curtailment as contribution levels increase for   PV. At least PV alone; if you combine it with  wind, it kinda dampens that effect, but wind   doesn't see quite as sharp of a rise. So PV is  a little bit more sensitive to curtailment, and   we're expecting to see a lot of it in the system.  So that's why we're looking at PV in curtailment,  

but who cares? Who do we expect to try to  take these results and do something with them?   Our overarching question was really  kind of broadly, from a very stylized   manner, to understand or explore what challenges  or interactions do grid stakeholders need to   confront as PV contributions grow. So we  see two main groups of these stakeholders   that we think would benefit from these results.  First is system planners and operators where   thermal generator operations  are particularly important   at these mid-PV penetration levels – that's the 25  to 40 percent that I mentioned earlier, but again,   may vary by system. And this suggests that  there might need to be a phased approach   as our power systems continue to transform and  presumably move towards these higher VRE and PV   contribution levels. So there might need to  be sort of different regimes for both planning  

and operations. And the second group are market  designers, who might need to potentially explore   revising operating reserve eligibility rules and  compensation structures as PV contributions grow,   and where there might need to be kind of resets  or tweaks to be able to better align the benefits   and value that resources are providing the system  and the compensation that they're receiving. But   I should say that's an even broader conversation  in the market design world, and worthy of its own   presentation. So our study system that we looked  at in this particular analysis was based on the   Los Angeles Department of Water and Power, or  LADWP. You might've heard about the LA100 study  

that recently came out from NREL. We were using  basically one of the data sets that that team used   with some slight modifications, and the link for  the LA100 study is here. If you haven't seen it,   you should go check it out. Some really  great and really innovative work.   So first we used a capacity expansion model  to develop six PV and storage build-out cases   where we assumed consistent load, both on the  profile and the magnitude across those six cases.  

We really wanted to have six different PV and  storage contribution levels with all else kind   of being uniform for those load levels, at least.  And so within this capacity expansion modeling   framework, it captures the declining capacity  credit of variable renewable resources and storage   resources, and so it inherently is endogenizing  within its solution the need to over-build from a   name-plate capacity perspective in order to have  sufficient firm capacity for resource adequacy   purposes. That was a question that I've gotten in  the past, so just wanted to clarify that point.   And also, because it's sort of this co-optimized  framework, kind of a central planner perspective,   it's inherently guaranteeing cost recovery  for these six kind of base cases only. We then took those six different build-out  cases and applied them to a production   cost model. PLEXOS was our tool in this  particular case. It's a commercial-grade   software that some of you might be familiar  with. And we did that to then explore various  

sensitivities on top of those six base cases,  which allowed us to explore those different   operational factors and how they impact  curtailment. So we looked at curtailment of   course, generation costs, and price outcomes for  each of those sensitivity cases as well as the   base case. But because we've effectively modified  the system from an operational standpoint – again,   just those operational parameters; we didn't  change any of the portfolio or the system   build-out itself – we can't guarantee necessarily  long-run cost recovery, because we're not   accounting for that feedback effect on how those  operational parameters would impact prices and,   in turn, investment decisions, and even from  a central planner perspective. So just wanted   to highlight some of the kinda caveats and  nuances of the modeling tools that we used here.  

And then this picture on the right is showing  kind of the broad footprint that was used for the   capacity expansion modeling, which is the various  different-colored circles. We used NREL's RPM,   the regional planning model tool, for that  to get those six different build-out cases.   And so it looks at, in this case, the full western  interconnection with sort of this more resolved   kind of granularity from a spatial and  operational perspective for just that LADWP   footprint. But by having sort of a less resolved  representation of the rest of the interconnect,   it's able to account for the boundary effects.  And so within that LADWP footprint, it's able to   do mixed integer programming as well as – so that  means unit commitment is represented, and that's   done at an individual unit level, and individual  transmission lines, as well. So slightly different   from the ReEDS model, if any of you are familiar  with that, so a little bit more granular, and it's   meant to kinda zoom in, as the name implies, to  a specific region. Okay, so these six build-out  

cases that I mentioned, and this'll be consistent  through the rest of this talk, we've named them   by their annual PV contribution level for that  base case. Now, of course when we've adjusted   the operational parameters, the exact penetration  levels or contribution levels change, and I'll get   to that later. But you'll keep seeing these same  six numbers throughout the talk, and that's just   referring to these different build-out levels. Son  installed capacity on the left, total generation   on the right. The key point here is that you see  solar contribution levels increase significantly  

across these different cases, as well as battery  contributions. And so the PV _____ note includes   rooftop PV, utility-scale PV – and that's  kinda this lighter color yellow – and then   also we had hybrid PV-plus-battery resources,  and we've separated in many cases at least the   battery from the PV contribution of that overall  hybrid system, just to kinda give a better idea   of the breakdown. And so that's kinda the two  different slightly gold colors in these figures.   All right, so if we zoom into just, again,  these six base cases without any of the other   sensitivities, this table is just giving  a little bit more information about what   the contribution levels from the various resources  are, kind of what's happening. I'm not gonna talk  

through every number here. You can go look  at the paper if you really wanna study this   a little bit more. But the point is that we have  a wide range of renewable contribution levels:   if we just look at PV; if we look at PV plus wind;  and if we look at sort of a broader renewable   energy level, all the way up to a 44 percent PV  contribution level with 96 percent instantaneous   penetration cases, and then 75 percent if we  account for wind. Storage ranges its contribution   as a percent of peak load, so on a capacity basis  all the way up to about 37 percent. And then  

you'll see curtailment, the average annual ranges  for total VRE about 8.2 percent in the maximum   case and almost 10 percent if we look at just the  PV resource alone. And then the max instantaneous,   we're hitting 57 percent for wind and  solar combined, and 65 percent for PV.

So the sensitivity scenarios that I've mentioned  a few times now – again, we layered those on top   of the base case, and so I know there's a lot of  words on this slide with this table. I'm gonna   kinda talk through the categories individually.  Not gonna cover every single word here, but again,   you can go to the paper if you wanna revisit  this. So the base case is this first line,  

so that's referring to those six build-out levels.  And so this table is basically saying for each of   these six build-out levels, we're going to apply  an additional 13 sensitivity cases which in total   results in 84 total scenarios that we ran and  included in this report. So the first group here   are what we're kind of lumping together as thermal  plant flexibility. So this is where we have   more and less flexible from a minimum generation  perspective, from a minimum up and down time,   and from a ramp rate perspective. So this  is really trying to capture the ramp rate,  

the duration, and the magnitude of  these thermal flexibility parameters,   to understand how they impact curtailment  and specifically how much VRE variability   and uncertainty the system can absorb. And so the  duck curve I know is overused, but a lot of the   challenges that folks point out from  that figure are embedded within these   sensitivity cases. So the belly of the duck,  for example, is the minimum generation point.   The next category here is really looking at  those operating reserve eligibility rules,   so that was kind of that second paradox element  that we talked about earlier. So here, this is   where we're allowing VRE and storage to provide  operating reserves in different combinations   of that eligibility, and basically to see if by  allowing those resources to participate, if they   can use otherwise curtailed energy in the upward  direction, but also if they can curtail even   further in the downward direction to contribute  to those reserves, 'cause we have both directions.  

And then other constraints, sort of a  miscellaneous bag here, but looking at   how the rest of system operations – for example  unit commitment decisions resulting from each   of these – can impact curtailment levels.  And so we're looking at a greater resolution   with the five-minute; we're looking at sort of  the pass down from the day ahead to real time,   with forecasting errors being seen when  the real-time phase of the model runs. A   no-storage case is really a counter-factual,  and then copperplate to look at the impact of   relaxed transmission constraints. So just a little  bit more information on the operating reserve  

scenarios before I dig into results. The eligible  generators may provide operating reserves up to   their operating limits, so just wanna clarify this  in case there's any confusion. So that means that   all of the ramping limits, the maximum resource  availability – for example, you can't provide more   reserves than the PV resource that you have  available – and also online status constraints   are all applied. So PLEXOS was able to consider  all of these different factors to be sure that   a resource actually would have had room to be able  to provide reserves. And within PLEXOS we only   modeled the sort of holding of the capacity;  we don't actually model kind of the releasing   or the deployment of that capacity. So  that's a slight nuance to clarify as well.  

But the point is that even with holding that  capacity, its ensuring that all of these   operational factors would be maintained if that  capacity was released into energy. So there're six   products. Not gonna go through all the details  here, but just wanna highlight a little bit.   We have regulation up and down; spinning;  non-spinning; and flexibility up and down.   Different response speeds for each of these  resources, different scarcity pricing assumptions   with sort of a hierarchy being maintained with  those assumptions of the prices. Somewhat of an   arbitrary choice here for how we just changed  it by one dollar, but the point was do maintain   that normal hierarchy that we see  in the actual system operations. So we're gonna dig into the two different paradox  components now. The first one was really that that  

thermal flexibility only matters in the middle.  Kind of a colloquial spin on the wording there.   This figure, which is a key figure, shows  VRE curtailment. So this is wind and solar,   on the vertical axis; PV penetration on the  horizontal axis; and the text boxes above the   lines show the storage contribution levels, which  if you remember were shown in that kind of summary   table of those six different build-out cases  that I showed earlier. And then the different   lines here are the different sensitivity cases,  with all the cases except for the ones that refer   to operating reserves. We're gonna save that  and focus on those when I discuss the second   curtailment paradox item. So what you see here is  that the no storage case, which is kind of that   goldish-tan color on the far left – we see massive  increase in curtailment as PV penetrations grow.  

Again, this really serves as a counter-factual  case. This would never really happen in reality,   because pretty much all the PV being built these  days has battery coupled with it, or at least   paired with it as a co-located resource. Or  hybrid, I guess there's a combination of both.   But the point is that this was just showing  the importance of having storage to minimize   curtailment in the system. Kind of a bookend  result here. And so the key point is that we   have this transition zone in the middle, or if you  wanna call it a Goldilocks zone, but it's at these   mid-PV- contribution levels where those thermal  generator parameters impact curtailment the most,   much more than at the lower or higher levels. And  this is where the Goldilocks term kinda comes in:  

it's because in those mid-levels you  have enough PV to kinda see curtailment,   and you also still have enough thermal generators  on the system for the flexibility parameters to   impact curtailment. So as we start to get the  higher PV contribution levels, for example,   there's just not as much thermal online for those  parameters to matter. So kind of that sweet spot,   just right – the Goldilocks zone. And so we  see significantly less curtailment impacts   from that miscellaneous bag, that last column  in the scenario table that I showed earlier.   So the forecasting errors, the resolution,  and the transmission – we saw significantly   less curtailment impacts from those parameters,  at least in this particular test system.

So thermal generator flexibility. I wanna  kind of dig into those six different scenarios   that are really looking at, well, the thermal  generator flexibility, so kind of ignoring that   miscellaneous set that I just described at the end  of the previous slide. So this slide is showing   the change in generation from the base case.  So positive values here mean greater generation   relative to the base case; negative means less  generation relative to the base case. And I'm   showing the six different thermal flexibility  cases that have greater and less flexibility   with sort of the green labels indicating  greater, and the red showing less flexibility,   and then we have kind of an aggregated view of  the different generator technology types here.   So key points are that in scenarios with  greater flexibility, so the green labels,   we generally see more generation from solar, gas  CCs, and coal, with less generation from gas CT   generators. In scenarios with less thermal  generator flexibility, there's typically less  

coal and solar generation, with slightly nuanced  trade-offs among the natural-gas-fired generation.   And what's really going on here, kinda the crux of  all this, is that there's a trade-off between cost   and flexibility of these different generators.  So for example, coal tends to be relatively cheap   but inflexible. Gas CTs, on the other hand, tend  to be relatively expensive but very flexible,   and gas CC is in the middle. And so as you  restrict or relax different flexibility   constraints for those thermal generators,  it allows the system to sort of shift to   whatever's the most economic, or what's  available to meet the system needs.   And so I'm gonna show an example in a  minute here, but just wanted to note,   too, that the minimum generation levels and  the ramp rates, or at least the less-flexible   case of the ramp rate, the 10 percent case,  yielded the largest difference in generation.  

So I'm gonna, for this example, to kind of talk  about that trade-off of cost to flexibility. I'm   gonna look at the zero min gen case, kinda that  far left bar in this plot that I just highlighted   in the red box. So in that particular case, at the  33 percent PV contribution level, this figure is   showing a dispatch for four representative or  kind of just example days, with the base case   on the top row, and then that zero min gen case  on the bottom row. So this is, again, a greater   flexibility 'cause you're allowing those thermal  generators to go all the way down to the minimum,   and they don't have to stay at this sort of higher  level when they're remaining online. The textboxes  

in sort of the upper right of the upper right of  these two different rows are showing the total   generation cost and the curtailment for just  those four days. So what we see here, again   this is that trade-off in cost and flexibility,  is that the additional flexibility afforded to   the coal generators – because remember, they tend  to be the most inflexible – so now they're pretty   flexible, but by not having a min gen level, the  system can utilize more of the zero margin of cost   solar and this relatively low-cost but now  sufficiently flexible coal, as well as the gas CC,   which also has a fair amount of flexibility and  is kind of the next tranche up in terms of cost.   And so relative to the base case for these four  days, we see a greater utilization of solar, we   see quite significantly less curtailment, and also  some operating cost savings, about 16.5 percent.   So now I wanna move on to the second solar  curtailment paradox piece, which – again kind   of a colloquial phrasing here – is that curtailed  PV eats its own lunch. At least with respect to   operating reserves in this particular case, or  this application. So what we see in this first   figure, which is the same layout as that other  curtailment figure that I showed before – so   VRE curtailment on the vertical, PV penetration or  contribution on the horizontal – now we're looking   at just the cases that refer to operating  reserves. So we see the base case still  

with this thick black line, but we see the three  different operating reserve sensitivity cases with   the various orange lines. So the first thing  I wanna point out here is that at high VRE   contribution levels, by not allowing VRE and  storage to provide reserves – so that's the   no VRE or storage reserve scenario, which is the  thin, dotted line, the upper one in this figure –   we see significant increases in curtailment. And  so the base case is the one where we allow both   VRE and storage to provide reserves. So just to  kinda make this super clear, I'm looking at this   upper right point. And so this is the highest PV  contribution level, where we see about 53 percent   more curtailment relative to the base case when we  restrict VRE and storage from providing reserves,   so a massive increase. So let's look at  what's happening at that particular point.

So this is the base scenario on the left column  – the no VRE or storage reserves, which was that   one that was 53 percent more curtailment, on  the right-hand side here, and this is at the   44 percent PV contribution level, so that exact  same point that I just highlighted in the previous   slide. These figures are showing generation or  committed capacity for the three different main   kinda thermal units that impact the system the  most: the gas CCs, the gas CTs, and in this case   biopower. At least for this example,  these three are the most important.   The darker-colored lines in these figures  show the available committed capacity,   and the lighter-colored lines show the generation  that's actually resulting from those generators.   The text boxes that are kind of in the upper left  of each of these panels is the generation. In this  

case, it's for the full year, so the 8760 total  annual generation for each of these individual   generators. So what we see is that when VRE and  storage can provide reserves, which is again the   base scenario, thermal capacity that's otherwise  committed to meet operating reserve requirements   is no longer needed. And that's pretty clear  in these figures, right? If you compare kind of   the left case to the right, you see large portion  of the year with this base case where we don't   even need to commit these resources, and so  therefore we also don't generate from them,   because we're able to use that PV  generation to help provide reserves.   So this reduces generation for these  thermal generators by about half,   and you can see that from the numbers in these  figures. And it also, as I just mentioned,  

enables greater utilization of lower-cost VRE  and storage resources not only to provide those   reserves, but also for energy, thereby  reducing curtailment. And that's why we   see that 53 percent difference. So we thought  this was a pretty interesting result, but first   it seemed a little larger than we were expecting.  But again, as we dive into this, it makes sense.   Oh, and then it also reduces system-wide operating  costs by about half as well, in this case. So some operating reserve price trends, just wanna  touch on this quickly. Despite the curtailment  

and system cost benefits when we allow VRE and  storage to provide reserves, and the base case,   this doesn't necessarily translate into increased  revenues. So this is where there's kind of that   potential misalignment in compensation  that gets into some of the market design   items. But higher curtailment levels associated  with the increasing contribution of PV resources,   as well as the interaction with storage – because  we're assuming zero margin of cost for these   resources, which is maybe another point to come  back to, at least for storage – leads to lower   prices for reserves, especially during the times  of PV curtailment. So basically this is where   it's eating its own lunch. Because  they're low-cost resources,   they effectively suppress prices and thereby  prevent themselves from actually getting   significant revenues by participating  in these operating and reserve markets.  

Operating reserve shortage conditions, and  this is just kind of a secondary point here,   generally only affect operating prices during  non-curtailment periods in the transition zone.   So, one, not really super prominent to begin with,  and two, not very relevant for PV, but just kinda   wanted to note that this was happening. It was  just sort of a less-important dynamic that was   happening. And this is the last set of figures I'm  going to show. I wanna talk through a series of  

price results to kinda help show what's happening  with this price suppression effect that's kind of   a high-level finding. What I'm gonna do is show  kinda different panels here. Within each of those,   it's the mean reserve price, and in this case  looking at spinning reserve as the example.   And so just kinda taking the average for three  different kind of sets of hours across the year. So all hours of the year are kind of this dark  plum/purple color; hours with no PV curtailment   is this green or teal; and the only hours where PV  curtailment is happening is the yellow color. And  

then the six different PV contribution levels or  build-out cases are shown for the different sets   of bars. So what we see in this first case are no  VRE or storage reserves. That's the one where we   saw the most curtailment, the one that was  53 percent more than the base and that high   PV contribution level. So what we see is that  all three price metrics generally increase as PV   contribution levels increase, so kind of an upward  trend across all of these bars which is driven   by the thermal generator commitments that we were  just discussing. So you have higher-cost resources  

that are being committed to meet reserves, and so  they're the ones setting the prices for reserves.  There's a much smaller secondary effect from the  presence of operating reserve shortage or scarcity   events during periods without PV curtailment.  Those are the teal bars, so I kinda mentioned   this already. It has the largest impact in that  transition zone, but even with the sort of price   impact, it has a very small impact on the overall  system cost – less than 1 percent. But just kinda   wanted to note that, not a super-strong trend  happening. So now, if we move to the no storage   reserves case, so now we allow VRE to provide  reserves but not storage, we see reserve prices   generally decreasing, because now curtailed VRE,  which is zero marginal cost, can provide operating   reserve capacity at the zero cost instead of  the higher cost of those thermal generators.  

And this is particularly prominent during  periods of PV curtailment, which you can   see by the yellow bars, and operating reserve  shortage events play a much smaller role here.   If we move to the bottom left here, where  we now allow storage to provide reserves   but not VRE resources, we see a much greater  decrease in operating reserve prices which   is driven by the participation of storage. Again,  we're assuming a zero cost here, and it has ample   capacity to provide reserves, especially during  periods of curtailment, which are the yellow bars.   Operating reserve shortage events play a very  small role here. And then when we allow storage   to provide reserves, we see a pretty significant  reduction in total curtailment from a megawatt   energy basis but also the number of hours  where curtailment is happening, implying   that there are fewer periods when PV could even  use curtailed energy to provide reserves if it   were permitted to do so. And in fact, that kind  of thinking leads us into our final point here,  

which is now going to our base case, where we have  both VRE and storage allowed to provide reserves.   We see the operating reserve prices don't really  change much from this last case where only storage   could provide reserves. Prices remain near  zero, especially during times of curtailment.   And so storage has really already taken most  of the benefit of providing reserves, and so   there's really little left for PV to benefit from  that, kinda to that last slide and the last point,   because there's just already fewer periods of  curtailment in the first place. And so this   highlights the importance of really understanding  the role and interaction between PV and storage,   but also for storage providing reserves. And the  devil's in the details with a lot of these rules   related to markets, and so there's maybe some of  those details we need to explore to understand the   implications or impacts that they could have  on other resources in the rest of the system. So just wanna touch on a few wholesale market  design considerations that kind of were relevant   based on what we were seeing from this report.  Again, this is more of a discussion and not  

necessarily things that we are demonstrating  in this study. Because the first caveat is   that we're not actually modeling a formal  wholesale market. As many of you probably know,   LADWP is a vertically-integrated utility. It's  not even really part of a formal market, although   it certainly interacts with the surrounding _____.  So really we're using sort of this production-cost  

model, cost minimization approach to simulate at  least from a high level how a potential market   dynamic could play out, and how prices from  those results would give us some sort of   insight into how market design-related elements  would be affected in these high PV contribution   levels in the various sensitivities that we  ran. So again, not a formal wholesale market,   but the way that we modeled this gives us some  idea of how prices might be affected to comment   on where future work and future discussions should  go with this topic. I didn't explain that well,   but hopefully that made sense. So first,  market structures and policies that allow   all VRE resources to provide operating reserves  yield lower-cost operations in curtailment. So   this is really that second paradox item. But this  sort of open participation is actually not common   practice in competitive wholesale markets  in the US, ERCOT being excluded from that,   where they do allow basically any resource  to participate as long as they can qualify.  

Second, operating reserve markets may not provide  a significant revenue source to compensate for   reduced revenues resulting from greater levels of  curtailment and declining energy prices. This is   also really the second curtailment paradox  piece. So really it's just saying that there   might be a misalignment between the value  that resources are providing the system and   the compensation that they're receiving, and  so there might need to be some adjustments   to really compensate resources for the full value  they're providing. And this extends, honestly,   beyond even just the PV and battery discussion.  This is sort of a broader grid system services  

and getting the prices right within markets,  and that's, I think, a topic that there's a   lot of discussion and work being done on right  now. But the third point is that markets may   need to continually update pricing structures to  better reflect value of reliability, flexibility,   externalities, and the true lost opportunity  cost. So again, this kind of is just playing on   that same "getting the prices right" theme, where  there's a lot of different rules and factors that   can impact the price outcomes, and we wanna make  sure that those price outcomes are appropriately   signaling for the full set of grid services,  and there's not any unintended consequences.   So some of the factors that can play into  this are the rules around shortage pricing,   and so ORDCs – operating reserve demand curves –  are a popular idea right now being implemented in   multiple systems to try to better signal for that  scarcity or shortage. Also bidding practices of   different resources to make sure that they're  bidding in to account for their full cost in   the full set of considerations. So for example  we assumed zero marginal cost for storage, but  

really there's some sort of degradation cost since  those batteries are being used – kinda this wear   and tear idea, which extends to other resources  as well. There can be participation rules that   can affect bidding practices, and there could  be other policy factors, like RPS – renewable   portfolio standard – that may impact how certain  resources need to be used, or how they did.   There's also the avoided emissions pricing, so  a lot of conversations right now about carbon   pricing and carbon taxes. So that plays into how  the prices turn out in these different markets.   And also uplift payments, which are  out of market, but they're used to   really make sure that generators that are  needed for reliability get compensated   fairly. And so there's a whole conversation about  that, and transparency is a big issue there.   But the point is that there's a lot of details  that play into how prices turn out, and trying   to get those prices right is a big item to still  be figured out within wholesale market design.

And the last point here is that markets may need  to consider factors that impact how the rest of   the system is operated, so min gen levels being  an example of how different factors can impact   operations. So for example, different emerging  technologies like hybrids and distributed energy   resources, and also resources that are effectively  self-scheduled – for example if they are through   some sort of bilateral contract or PTA where they  may have flexibility that they can provide the   system, but the operator may not see that or have  access to it. And so there's lots of conversations   about how to integrate resources to tap into  the flexibility, or even just understand how   the lack of that flexibility can impact  operations in the rest of the system.   Okay, so just to wrap up here, a few more  kind of bullet lists. Sorry for all the words,  

but I just wanna flag what we see as being  some potential future work. Of course,   this may not be exhaustive, but work that  could look at differences in dimension, so   different systems beyond just this LA-like  model that we used, different build-out levels,   different range of storage and hybrid contribution  levels, different durations for storage. So   lots of different cases where you look at a  different build-out of the system and typology.  

There could be cases that look at flexibility,  both things that can impact flexibility but   also supply flexibility needs to the system, as  well as trade-offs between different options.   So for example, and I think this may be the best  example, is doing a full cost-benefit analysis   of the thermal generator flexibility upgrades. So  we basically just assumed, in our kind of bookend,   stylized scenarios, that you would be able  to have a zero min gen, for example. Is that   possible? If it is, what's the cost? We didn't  account for the feedback effect of that cost   in the capacity expansion model, where it  was making decisions to build and maintain   certain generators in the system. So something  where we try to co-optimize at a unit level   to account for the full cost and benefit trade-off  of these different flexibility options would   be a really cool project to do in the future.  Another one is looking at a larger role of demand   response, or more broadly just price response of  demand, particularly seeing how it could be an   alternative to thermal generator upgrades. For  example, could demand response result in the  

same outcome as the zero min gen scenario? Also  looking at consideration of other end uses for   curtailed energy, so this gets into the ESI,  energy systems integration, realm, where we're   looking at kinda sector coupling and combining  with other end uses beyond just the power sector.   We could also look at more robust  evaluation of the role of uncertainty,   so this kind of gets into forecasting errors, and  even using stochastic forecast, and the treatment   for look-ahead horizons and resolutions, and  how that sort of information flow, if you will,   information flexibility in a sense, can impact  operations as well. Also technology and market   representation, so this kind of is a little bit  of a repeat for some of the other points, but   just looking at other types of resources and  different ways that those resources can be   operated, including the rules surrounding them.  Particularly in this case, the storage and hybrid   representation would be of prime importance. Also  improvements to the operating reserve treatment,   making sure that we're capturing the true cost  of these resources and kind of the interaction   with policies that may disincentivize,  for example, down-direction of reserves,   where there's some sort of RPS requirement.  And finally, linking to investments – again,   this also sort of touches on previous points as  well, but kinda the feedback even more broadly   from that cost-benefit co-optimization would  be just more generally a feedback effect of   operational sensitivities on investment decisions  over multiple years of a system evolution. So  

even just looking at market design impacts, for  example, where there may not be a technology cost,   but there's a feedback effect through the price  signals. And also how market design changes could   support revenue sufficiency for resources  that are needed for reliability purposes. Okay, so I am done now. The reference is  here if you wanna come back to those later,   but thank you for your time. I know that was kinda  maybe drinking from a firehose, but I will open   this up to questions. And oh, I forgot to mention  this at the beginning, sorry, but hopefully,  

it looks like folks figured it out. There's a  Q&A box, so please put your questions there,   and I'm gonna go through those  individually starting now,   and may call on some of the co-authors who are on  the phone to jump in if needed. So I'll just go   in the order here. So, Joe Seale: "You mentioned  cost recovery is only guaranteed in the base case.   Can you comment on how PV revenue compares  to LCOE with a large amount of curtailment   and lowered wholesale prices at the times of  solar generation?" Ooh, that's a great question.   I don't know what the revenue versus  LCOE breakdown is, but yeah, obviously   with more curtailment, you're effectively  increasing your levelized cost of energy from   those resources because you're getting less  utilization from them. We commented on this   in the earlier paper – not this Joule paper, but  the other one that I mentioned, from 2019 – but   I don't remember the exact numbers off the  top of my head. But I know it's in the text.  

We didn't crunch these numbers for this study,  so I can't give values there. But the point   was just that there's sort of that feedback  effect being accounted for in the base case,   feedback with kinda prices within that capacity  expansion model, but not in the sensitivity cases.   So if that didn't answer your question,  please feel free to add another follow-on.   And if any of the co-authors have  additional thoughts there, please jump in.   Okay. "On slide 25, why do you say that storage  has zero cost?" Let me go back to slide 25.   Except I don't see – oh,  there it is. I'm blocking my  

– _____. Oh, I'm not saying that it actually has  zero cost; I'm saying that's what we assumed,   which really, y'know – Degradation cost, it's  not actually zero, but it was a simplification   that we made. But the point here is just  that because we made that assumption,   that's driving storage to really capture the bulk  majority of the benefit from providing reserves.   So future work definitely should explore  what that actually cost for storage could be,   and I know folks are working on that. It's kind  of a can of worms, 'cause it gets hard to – How do  

you parameterize that cost, like what is the cost  curve? How do you track the usage of the battery   to then align with that cost curve, assuming  you're using a curve or function – whatever   the math is behind it. But it's a tricky problem,  and I know folks are looking into that. Okay,   next question. "Did you include hydrogen as one  of the sources – ? " I'm sorry, my screen just   changed. "Did you include hydrogen as one of the  storage options, in particular for longer-duration   energy storage requirements at higher VRE?" No,  we did not. We only included battery storage,   as well as pumped hydro. And Brian, correct me  if I'm wrong, we had a compressed air, I think,   CAES. I can go back to that. We actually  did have a small amount of demand response.  

I guess I'm wrong about the case. Actually,  let me go back to the – do-do-do –   here we go. Oh, we do, okay, and we just didn't  plot it in the other one 'cause it's not getting   used. Okay, so there's a little bit of CAES, a  tiny bit of demand response that you can see in   the capacity. I guess the CAES and the demand  response show up on the capacity but not on   the total generation, so they're not actually  being used. So pumped storage and battery are  

really the two storage options that we used. But  agree that in the future, looking at hydrogen,   especially for long-duration storage, would  be super interesting and, I think, valuable.   Sorry, I'm getting my cursor tied  up here. Okay, let me get back here.   So Keith also asked, "On slide 30, one of the  other end uses for curtailed energy is the whole   H2 at scale wheel, such as the – yeah, exactly  – steel, ammonia, yeah, transportation,   et cetera. Yeah, totally agree, there's a ton  of end uses where if you produce that hydrogen,   even if it doesn't go back to the  grid per se for balancing purposes,   there's a whole bunch of industrial  and commercial applications that   can use it and be part of this kinda broader  energy systems integration/ sector coupling. Yeah,  

totally agree, and yeah, this would be  an interesting area for future work. So the next question. "Did you consider a scenario  with no coal generation, or where only the gas   thermal plants could be run more flexibly?" Not  explicitly in the sense that we did a no coal, but   if you look at the – Actually, this is convenient  'cause it's shown right now. If you look at this  

slide that's shown right now, you'll notice  that at the two higher PV contribution levels,   the 38 percent and 44 percent, there is no coal.  Or, I guess, look at the installed capacity on   the left-hand side. So the coal is all retired  by that point, and this is part of the LA100   trying to get to 100 percent system. They had  certain retirement decisions within the model.  

So, yeah, so there was no coal in those cases.  And this has actually come up before – one of   the interesting things to do would maybe be to  go back and look at an hour-by-hour basis from   some of these runs, and break down by technology  type sort of what's running versus what's not,   to see like capacity factor values of  the generators, to see maybe if we could   kind of back out whether coal or gas is  contributing more or less to the curtailment.   That's not something we've had time to do, but  there could be some just even post processing   with these that could get at kinda the  relative role of gas versus coal. Okay.  

So a question from Yang. "If I understand  correctly, this is based on a regional grid. Would   you mind making some comments about whether it  is technically feasible to scale this study up at   a national level?" Sorry, my thing keeps moving.  "…at a national level, or even at a global scale?   Are conclusions here applicable to other regions  and/or grid conditions?" Yeah, great question,   and sorry if I didn't make this clear. Yes.  These results we can only guarantee for this   particular system, and that was sorta why we had  flagged in future work looking at other systems   and other configurations. It would be, I mean,  technically feasible, 'cause there's other studies   that have looked at the full US in a production  cost-modeling perspective. It gets challenging to  

run those. They take days with parallelization  techniques on our fancy HBC high-performance   computer. So it would be possible, but for the  budget and time frame that we had in this study,   doing those scale of runs was not possible. But  certainly it could be done. I think it could be  

interesting, especially where you start looking  particularly in the Eastern Interconnection, where   you have larger loads, and you have a more kind  of interconnected and meshed grid. Seeing how the   imports and exports from different regions  – basically that transmission question,   how that can play into this. Because yeah, we only  looked at sort of the small footprint in the LA.   So even though the capacity expansion model  looked at the full region, the full Western   Interconnection, we only looked at the LADWP  footprint for the production cost modeling runs,   effectively treating it as an island. Which  may seem like a strong assumption, but it's   actually consistent with how LADWP does their  planning studies – basically kinda worst case,   "Let's assume it's an islanded system." So it just  made it easier to kind of maintain tractability  

and dig into what was going on. But yeah, future  work should definitely explore other systems.   And we try to caveat in the report that some  of the exact numbers and the quantitative   results may not extend exactly. But  yeah, future work should explore that. Okay, I'm not seeing any other questions on  here, but let me just scroll through again.  

I guess if no other questions pop in  in the next minute or so, we can just   end early and give folks a few minutes  before what is likely their next meeting.   And of course, we're always happy  to answer questions by email, if you   wanna follow up later, and I can put  my last slide with my email on here   in case you wanna note that. Here we go.  Oh, I think I saw a question come in.   "Did you look at individual impact on thermal  plant profitability?" No, we did not. We could, or   at least a net revenue perspective. This is what's  tricky, and at least in the past work I've done   with production cost modeling runs, net revenues  is really the best we can do. Net revenues are  

the total revenues from kind of the energy and  ancillary services, their operating reserves in   this case, minus the operating cost – so basically  the production cost within the PLEXOS runs – and   so that difference is called the net revenue.  And that's actually used by the independent   market monitors when they're assessing the  various ISO and _____ regions for market design   considerations, and market power, and that sort  of thing. That revenue basically indicates what's   left to cover the fixed costs, and so often  that's compared to like a net CONE value,   for example, just to determine whether there's  sufficient net revenues to invest in new   resources. So profitability gets a little tricky  to report, but we could if we wanted to, and if we  

had time and resources, could go back and  calculate the net revenues for those generators. Great, so I guess maybe I'll just  give it a few more seconds if there's   any more lingering questions. Oh, now I'm  seeing – Oh, there're several questions.   Okay, also a follow-on. "Do you think that current  market design broadly speaking incentivizes  

updates of existing combined cycle plants at lower  min load?" Oh, geez. I mean, there's a lot here.   I mean, this kinda gets into the flexibility  market design topics, and all the talk of adding   flex products, flex ramping _____, for example. I  mean, I think the idea is for those to signal for   more flexible capabilities of resources. I can't  comment on whether they're sufficient or not. I   guess it just depends on the system. But I mean,  I think that's the direction some of the market   designers are doing with, and I mean and certainly  some regions have already implemented it.   Whether it's getting the right press or not is  another question. But yeah, this is part of, also,  

that full kind of co-optimization with the cost  benefit that we'd wanna account for the cost to   do those upgrades versus the revenues. "Does the  model take into account scarcity pricing?" Yes,   it does. So there was one slide, I kinda  went over it quickly, but it showed the   different scarcity pricing assumptions that  we used for the operating reserves, which   tend to be the more important set versus energy,  for example, because you're gonna violate your   reserve requirements, which leads to scarcity,  before you ever drop load. So those scarcity   pricing values are what you're gonna actually  see. Oh, I think I have it much further back.  

Oh, here it is. Yep, so here's the  shortage or scarcity pricing values.   Again, they're somewhat arbitrarily chosen just  to have a hierarchy, where our regulation reserves   tend to be at the top of the "pyramid" for our  operating reserve hierarchy. So yeah, and we   did see scarcity pricing, or shortage pricing,  events that were driven by having a shortage in   the reserve requirement – or meeting the reserve  requirement. "Has the transmission constraint  

have been considered in case of excess solar  _____ generation?" Not sure I understand – So I'm not sure I understand the question,  but I think what this is asking is about   the transmission treatment. And in every  scenario except this bottom and this figure,   the one that says "copperplate," that's  assuming transmission are not enforced,   so a copperplate system, where there's no flow  limits or anything. Aside from that case, every   other scenario that we ran enforced transmission  limits within the footprint that we modeled.   Okay. Unless I've missed any, I think we've  gotten through all the questions here.   Give it a few more seconds here, and then if not,  we'll wrap up. We're almost out of time anyhow.  

Bah, okay.   Oh, I see. Got it. Okay, thanks, Brian. Thank  you for that. Cool. Well, I'm gonna call it.   Like I said, we're almost out of time. But thank  you all for joining. This was actually the first  

time I've done one of these NREL webinars, so I'm  excited to be able to share this work with you.   Hopefully, if you're interested, you'll  be able to go and look at the full report.  

2021-07-08 02:44

Show Video

Other news