What? The sun isn't always shining?!

In this episode, Princeton professor and energy modeler Jesse Jenkins tackles the question of how we can build a decarbonized energy system that relies on inherently variable wind and solar power.

(PDF transcript)

(Active transcript)

Text transcript:

David Roberts

If you’ve spent much time discussing clean energy on the internet, you’ve probably come across a disturbing piece of information: the sun, it seems, is not always shining. What’s worse, the wind is not always blowing!

It’s crazy, I know.

Unlike coal or natural gas or nuclear — “dispatchable” power plants that we can turn on or off at will, when we need them — we do not control solar power and wind power. They come and go with the weather and the rotation of heavenly bodies. They are, to use the term of art, “variable.”

Many people, bringing to bear varying levels of good faith, conclude from this fact that we shouldn’t or can’t shift to an electricity system that is based around wind and solar, at least not without occasionally shivering in the cold.

Is that true? Do we know how to balance out the variability of wind and solar enough that we can fully decarbonize the grid with them? This is probably the number one question I hear about renewable energy, the number one reservation people have about it, so I decided it’s time to tackle it head on.

To help, I called on longtime Friend of Volts, Princeton professor and energy modeler extraordinaire Jesse Jenkins. We walked through the basic shape of the problem, the different time scales on which variability operates, and the solutions that we either have or anticipate having to deal with it.

This one is long and occasionally gets a bit complex, but if you’ve ever wondered how we’re going to build an energy system around wind and solar, this is the pod you’ve been waiting for.

All right, I am here with Princeton professor and longtime friend of Volts, Jesse Jenkins. Jesse, welcome back to Volts. Thanks for coming back.

Jesse Jenkins

Hey, Dave. It's always good to chat with you.

David Roberts

Jesse, the reason we're doing this is that in the course of my research, I have come across some extremely disturbing information which I felt like I needed to share with you and the world as soon as possible. Apparently, the sun is not always shining and the wind is not always blowing.

Jesse Jenkins

Wait, what?

David Roberts

I know this changes everything, so we're going to have to talk through this.

Jesse Jenkins

Oh, man.

David Roberts

But seriously, I don't mean to make too much light of this. This is a subject about which people say lots of dumb things, but it is at its heart, I think, a perfectly valid question, a perfectly valid area of concern. In fact, it is the central area of concern about renewable energy. It is the central question to answer, which is the term that used to be used is intermittent. Renewable energy, wind and solar are intermittent, I think. Now the preferred term of art is variable, but I think probably the most accurate terminology for our purposes is non-dispatchable.

It just means we don't control it; we don't turn it on and off. It comes and goes with the weather.

Jesse Jenkins

Yeah, I prefer just weather dependent. Right. I think it makes more intuitive sense to people. Like you said, it's solar and wind power, so it depends on the weather. That is not shocking, but also defining of what the resource is.

David Roberts

Also dependent on the turning of the planets and the solar system.

Jesse Jenkins

That's true.

David Roberts

But anyway, people know what we mean. We don't control them. A lot of people, I think, especially people who are coming, who haven't given a lot of thought to the clean energy transition and are starting to grapple with it for the first time, I think intuitively run up against this question early on in their thinking, which is "how do we deal with this?" So, I want to take those questions as good faith questions and talk through answers to them to the extent we have answers, to the extent we do know how to deal with it, to the extent we do have the tools to deal with it, and the extent to which it remains to some extent unsolved.

I want to start with a couple of really big picture questions before we hone in on the details. I think the first big one to ask is just what greenies, what climate people seem to be recommending, and what we seem to be doing, at least in the early stages, is shifting from an electricity system based on dispatchable power plants that we can turn on and off at will to a system that is fundamentally based on non-dispatchable weather-dependent power plants that we can't turn on and off, which, as we're going to talk through, raises a whole host of issues and problems to solve.

So I think the first thing to address is just why do that at all? Why take on that trouble? Why not just shift from dirty dispatchable energy to clean dispatchable energy like nuclear, hydro and geothermal? Why take on the burden of dealing with variable energy at all?

Jesse Jenkins

Yeah, it's a great question, and the reality is that now the reason is that wind and solar are really, really cheap. That wasn't the case a decade or two ago. And we sort of set off on this path supporting wind and solar and other clean energy technologies in different countries, not sort of knowing exactly where the cost declines would travel. And what we've seen is that basically, in the west at least, the cost of building new large scale nuclear power plants, which is our sort of most mature and previously scaled carbon free generation technology, they've only gone up over the last couple decades, and we can talk more about why or that's probably a better subject for another podcast.

Just observe that that is a factual statement. In other parts of the world, like China and South Korea and UAE, where they're used to building large scale civil works projects, they've been able to build nuclear plants at a reasonable expense. And it's a big part of the mix for those countries. But what we've seen is a tremendous decline in the cost of wind and solar power. And what we chalk that up to is what's known as experience curves. I know there's a great Volts cast in the archives on this topic. You can go back to listeners.

But the idea is that as we scale up and deploy basically any technology, but in particular wind and solar at scale, we drive a whole set of processes kind of centered around innovation and competition that lower the cost of the technology. And that's done through economies of scale, either in manufacturing or production of the technology itself. It's done through incremental innovations and sort of improvements that just get more efficient and better at producing and building these things over time. It's done by learning, by doing sort of tacit knowledge. The skilled workforce develops and the engineers of the processes develop over time.

And all of that drives the cost of technologies down as we build more of them. That's true for ships and flat panel TVs and aircraft and also for wind and solar. And particularly true for wind and solar because they are very small modular technologies that are repeatedly built both in manufacturing and in installation characteristics that make them well suited to not just learning, but rapid learning or rapid experience curves. The sum product of that is that wind and solar are the cheapest way to get electricity, period. Not just clean electricity, but electricity in most of the world today.

Solar is the cheapest in most of the world. And if it's not, it's probably wind, with a few exceptions. And so that's the main reason to rely on it. It's a cheap source of both clean and abundant electricity.

David Roberts

Right. So this would be something that the market would be pulling people to do anyway. These set of problems around variability would be problems that we would be trying to solve regardless because people want cheap energy and here's cheap energy. And so, people are going to figure out how to maximize their use of the cheapest energy available. In large part, this is being driven by forces that are not directly related to climate change. Obviously, we need to do it as fast as possible because of climate change, but this is not, as I think many people naively think when they first encounter it, something that we're taking on just because of climate change.

It is because this prize is out there, this super, super cheap electricity. And if you can run your widget on super cheap electricity, you're going to want to figure out how you can do that.

Jesse Jenkins

Yeah, I mean that's true. Now, I think it's important to remember that we supported wind and solar when they were expensive alternative energy technologies. That's what we called them back in the mid-2000s, right? And we supported them for a variety of reasons. Right. Because of climate change. Yes, but actually originally because of energy security concerns that were sparked by the Arab oil embargoes in the 1970s. And that's what drove the early era of solar and wind development in the US, Denmark, Japan, and other countries that wanted to get off of imported fossil energy.

We still burned a lot of oil in power generation at that time. And so, finding alternative sources of electricity was important for energy security. That has come back to the fore of our attention, of course, with the invasion of Ukraine and Russia's unprovoked invasion there, which sparked Europe to really dramatically reorient on a rapid pace away from imports of fossil fuels, gas, coal, and oil from Russia, which they were very dependent on. And it shot the cost of natural gas and gasoline here in the US. Where we, yes, produce more fossil fuels than we consume. So we're sort of physically energy secure, but we're still connected to these global energy markets.

And so when a dictator on the other side of the world decides to invade his neighbor for no reason, that drives up the cost of gas at the pump and the cost of natural gas for our heating here in the US, like, overnight. So there's a bunch of important reasons to pursue these fuels. They also, of course, have no air pollution. Right?

David Roberts

Air pollution. Yes. Let's throw air pollution in there because the science on air pollution, as you and I know, just gets worse and worse and worse. The evident damage of it gets worse and worse.

Jesse Jenkins

Yeah. And not just mortalities but also it makes us dumber. There's a lot of clear indication that particulate pollution, actually it affects our cognition, it affects our hearts, it affects our lungs, it impedes development of young children. I mean, it's just nasty stuff. And so if we can produce energy that's made from a domestic resource, like the abundant wind and solar that we have across the United States and other countries, that we can do so affordably and that we can do so without any air pollution, those are all really good reasons to rely on wind and solar and to want to tackle the associated challenge of dealing with their variability and weather dependence.

David Roberts

One other general level question. This is something else I think people kind of come to intuitively and there are not great straightforward answers out there, which is and this is a variant of the first question, but I think importantly different. A lot of people want to say there are times when the sun will not be shining and the wind will not be blowing, i.e., there will be times when renewable energy output is at zero and demand will still be there. So you'll have to have backup resources capable of satisfying all that demand. But if you have to have 100% backup, why not just make the backup the main thing?

Again, why go to the trouble, you know what I mean? This idea that because they are variable, they require basically 100% redundancy with non-variable resources strikes a lot of people as sort of crazy. Like, why don't we just build the non-variable resources and skip the first step? So what do you say to the 100% backup hang up?

Jesse Jenkins

Yeah, so the reality is you don't need 100% backup. You do need a sufficient amount of what I call firm capacity available. It's capacity that you can use whenever you need it for as long as you need it. Which makes it a really important complement to weather-dependent resources like wind and solar, as well as to we'll get to their role later, but to energy limited or time-limited resources like batteries or demand flexibility, which are key parts of the puzzle as well. And so you need a certain amount of firm capacity. It's a pretty significant amount.

But the reality is it's not 100% backup. You don't need one for one, because there actually are really no times when there is no wind or solar across a large area, unless you're talking about maybe an island grid that really has no geographic diversity. But yes, there's nighttime and there's winter, but generally there's some wind somewhere, right, at all times. And so you don't need 100% backup. So that's the first thing. And the second is that, again, for all the reasons I just went through, we want to rely as much as we can, again, not all the time, but as much as we can on wind and solar.

Because the fuel is free, the cost of installing them is incredibly cheap. And when you have wind and solar, you displace other dirtier fuel-consuming resources like natural gas or coal, and that saves money and it saves lives and it improves energy security. So all of that is the sort of main value add of wind and solar. I call them fuel-saving variable renewables, because when you've got them, you don't need to consume other fuels. And it turns out that if that's the kind of grid you're building, then there are pretty cheap sources of standby capacity that don't cost very much upfront and are perfectly fine to pair with also very cheap wind and solar to play that backup role. I mean, the one example is combustion turbines.

David Roberts

We should say that it's also unlikely that a trough in renewable supply is going to overlap with a peak in energy demand. Those peaks in energy demand tend to be during the daytime.

Jesse Jenkins

That's true today, although I would worry more about that in the future as we electrify heating when the demand is likely to peak in the winter overnight. And so it may be more likely that we do line up one of those periods, what the Germans call Dunkelflaute — You have no solar output or very little solar output even during the daytime because it's winter, it's very cloudy, and then you have a prolonged period of a big, high blocking high that sits across a wide region, a weather front that limits the wind output that can occur both in the winter and the summer.

And so, it is a challenge and it's something we have to plan our renewables-based grids to be resilient to. But again, that's why we don't depend entirely on wind and solar. We need a portfolio or a team. The way I describe it is there's a couple of metaphors. One is you need a balanced diet in your day-to-day life. And the fact that starches or wheat is cheap, right, as a cheap way to get calories, means that the bulk of your food pyramid or whatever is going to be from those sort of cheap sources of calories, rice, starch, all the sort of staple crops.

But of course, you also can't subsist entirely on those staple crops. You need a balanced diet of different things, playing different roles and combining with each other in a way that gives you a balanced diet. So the same thing's going on in the grid. We have imperfect substitutes here for each other. They all produce electricity just like all foods produce calories, but they have other characteristics as well. And just like starches and staple crops are the staple of our diet but not the exclusive makeup of our diet, wind and solar can be the staple of our energy diet as well, but have to be complemented by other things.

And so we just need to be clear about that. No one's saying only use wind and solar power all the time for everything. We're saying these are cheap, clean, energy secure ways to produce electricity that are scalable across most of the world. And so they're going to play a really central role, a star role in our overall energy mix.

David Roberts

People might be aware there is some controversy about — there are people out there who want 100% renewable systems versus people who want some nuclear or natural gas with CCS involved. But the argument there is not whether you need balancing resources to balance renewables, right? Even the people who want 100% renewables acknowledge you need storage and hydro and et cetera, et cetera, et cetera. They acknowledge you need resources to balance variable renewables. It's just an argument over which resources. Right? And we'll get to that later.

Jesse Jenkins

That's right. And I should just say before we dive into solving the renewables challenges, it is worth noting that it's a big diverse world out there and we have countries that are situated in vastly different ways in terms of their geography, their population density, their available renewable resource potential. And so there are going to be parts of the world that can't rely on wind and solar as the dominant source of their energy mix. Places like North Korea or Japan or the UK.

David Roberts

I was just in Iceland, which gets 100% carbon-free electricity with zero wind and solar. It's hydro and geothermal.

Jesse Jenkins

Yeah. So we should acknowledge that up front. And I'm not saying like this is the solution for the world, it just is for a big chunk of the world. And even in places like Japan or the UK, which are pretty dense and limited land area, they can rely on renewables for a good chunk of their energy needs. And they're trying to do so because of the energy security, affordability and climate clean air benefits that they offer. So it's a piece of the mix. Whether it's the dominant majority or not, it depends on the local circumstances. Some parts of the world are probably going to need nuclear power or geothermal or other more energy-dense resources to complement or even fully supplant wind and solar because of local resource constraints.

But that is going to probably be the exception, not the rule.

David Roberts

Okay, so I want to hone in a little bit on the timescales here. I'm going to run through and we'll sort of proceed in the discussion from the first of these to the last. So I'm going to run through real quickly the different timescales of variability because renewables are variable, but they're variable on several different timescales which pose distinct problems. So let me just run through this real quick. So at the shortest level, you have variable in terms of seconds or minutes. So you can think of something like clouds drifting in front of the sun that causes a slight dip.

There are those constant slight dips in the wind and the sun. And so you need something that is balancing in terms of near instantaneous short-term balancing. Then second, you have what you call minutes to hours. So you think of like ramping. So for instance, the sun goes down at the end of the day. You go from 100% solar resource to 0% solar resource relatively quickly over the course of an hour or two. That's a different kind of intermittency. And then you go up to hours and days. Here you get to what are called diurnal cycles. Overnight, for instance, the sun goes down at night and occasionally the sun and the wind will flag for a couple of days and then come back.

So there's the hours to days cycle. Then you get up to weeks. You can have weeks of unusually high demand or unusually low renewables. And then beyond that, you have what's called seasonal variability. So there can be entire seasons or years where solar insulation is unusually low or wind is unusually low. So at each of those timescales, you have a distinct problem to be solved in the electricity grid. And we have I think it's fair to — I mean, tell me whether you think this is fair or too crude.

I think that is roughly also the order of easiest to solve to most difficult basically. But we can get into that. But let's start at the normal second to second, cycle to cycle variability of wind and sun. What's our solution there?

Jesse Jenkins

Well, here's where it's important to remind folks that the electricity system is a pretty unique supply chain, in that supply and demand have to be balanced every millisecond instantaneously, basically, in this market. So if you're consuming electricity somewhere out there, someone has to be producing it at the exact moment that you're consuming it. That's true for every location across the entire grid all the time, which is different than, say, like Amazon's supply chain, where there's a package in a warehouse somewhere, it may or may not go out to get to you when you ask for it. It'll take anywhere from a day to five days right, to get to you.

David Roberts

It's the ultimate just in time delivery.

Jesse Jenkins

It's like Amazon Prime on steroids. So, yeah, it has to be balanced everywhere. Supply has to equal the demand in real time. And there's actually really significant physical implications if that doesn't happen, because you have a whole bunch of generators and induction motors that are actually synchronized with the alternating current frequency of the grid. That means that in the US, every 60th of a second, the grid's frequency is reversing back and forth. And the motors that are spinning to generate that electricity and the motors that are induced to spin by that electricity to do useful things like run industrial processes and other things are all synchronized with that frequency.

So they're spinning at 60 Hz as well.

David Roberts

What a wild thing it is that it works.

Jesse Jenkins

Oh, yeah. I start my classes like this just to remind ourselves that this is like the craziest continent scale Rube Goldberg machine that we've built with incredible physical tolerances. And it just works. Yeah. So that's important to remind ourselves because it's not like there isn't variability in that system already, right? Demand goes up and down. You can flip on a light switch or plug in your EV, or flip on an electric kettle, whenever you want. Right. You don't have to ask the grid in advance. You just do it. And that's true across millions of consumers all over the continent.

And power plants, transmission lines, they fail sometimes. Substations go down, transmission lines fail, generators break. And so not only do you have small changes in demand from your light switch, but you can have big changes in that supply and demand balance that happen pretty much instantaneously.

David Roberts

And this is a good distinction to mark here, which is the distinction between predictable variability and unpredictable variability, which are very different.

Jesse Jenkins

Exactly. So there's a certain amount of this variability and uncertainty that already happens in the grid. There's sort of demand changes that are both predicted and also errors in those predictions. So we have demand forecast errors every day that are off by several percent right, from what we thought the demand was going to be. And we have what grid operators call contingencies sort of the unplanned forced failures of certain grid equipment that we have to be ready for at all times. Because if you lose a 1000-megawatt substation with a big factory on it, like an aluminum smelter, or you lose a generator, a big coal plant or a gas plant complex, or a nuclear power station, instantaneously, you have to rebalance that because if supply and demand get out of balance, the inertia of the grid physically responds.

So it's a little bit like if you remember playing on a merry-go-round at a playground where you could sort of have a couple of friends on it and you're spinning it around and then somebody jumps off and all of a sudden the rotational inertia is the same, but the weight is different. And the merry-go-round speeds up really fast, right, because there's less mass to move around or somebody jumps on and it slows down, right? And that's the same thing that's happening on the grid in aggregate is if you add load, is what the electricity system calls demand because it acts like a physical load on the force that the generators have to induce to create the electricity. It slows those generators down just a little bit and they have to work a little bit harder or you have to add more supply.

One of your friends has to come run up and help you push the merry-go-round as more people get on. And the same thing, the opposite happens. If supply exceeds demand, it gets easier to push, just like when somebody jumps off the merry-go-round. And so the generators all speed up and so do all the motors that are connected to the grid. And what's challenging here, again, is that the tolerances there are incredibly narrow. So just a 1% deviation in that speed of the grid is enough to trigger devices to disconnect to avoid damaging themselves because they spin up.

If you have a generator that's designed to go a certain speed and it starts to go faster than that, it can start to throw turbine blades out at very high velocities and self-destruct a whole building, right? A whole very expensive generator that you don't want to blow up. And same thing with industrial equipment, right? If they start moving too fast or even too slow, they can cause damage. So we have these protection devices that will trip offline devices as the frequency gets out of this very narrow range. And that can also cause a cascading failure because if you lose one generator because the frequency is too high or too low, then you'll start to lose the next generator and then the frequency will drop even more and then you'll lose the next generator and it'll go even lower.

And so you get these cascading failures. And the grid operator's job, really the number one job is to avoid that outcome at all costs, right? To make sure that this crazy Rube Goldberg machine is resilient to those kinds of scary unplanned contingencies. So we always have enough backup generation, enough flexibility what we call operating reserves or contingency reserves or spinning reserves — lots of different names for these products — of basically backup generators that are there able to increase their output if they're already producing or decrease it very rapidly or that are offline but can start up quickly to step into the gap when something occurs.

And that's how we keep this crazy system running right now. So there's sort of a physical response of inertia as you turn on or off devices and then we have all of these sort of cascading markets of different paces of response time that we have backup capacity waiting for from seconds to minutes to half hour, hour long kind of startup times. And that system of redundancies is how we keep the grid running today. And it will be the same set of solutions and some new ones that will come in to help augment what we do today.

That'll help us deal with the variability that we now are adding from wind and solar to a system that is already variable and has dealt with variability since its very beginning.

David Roberts

Right. It's fair to say, though, that we have a lot more.

Jesse Jenkins

Yes, we will have more.

David Roberts

The second to second variability, for instance, is going to be a lot more from a wind and solar based system.

Jesse Jenkins

Yeah, it adds a new source of forecast error. Right. Because your wind and solar is now also variable with the weather and we get better at forecasting that the closer to real time we get. But there are still errors in those forecasts. And certain power plants, coal plants, nuclear plants, others are slow to react to changes. And so we actually commit them to operate a day ahead of time. Usually we give a day ahead schedule for the next 24 hours and that predicts the sort of average demand over the course of each hour that we're trying to meet.

And so generators get turned on and are ready to meet that demand based on the forecast. And if the forecast is wrong a day ahead, then we need to deploy those flexibility resources at different timescales to cover the surpluses or deficits that we have in the system. And again, that's already how it works. We're just going to do more of that. And in some ways, we're going to be reducing the conventional sources of that flexibility. Because right now we get most of that flexibility from hydro and fossil power plants that are committed and operating on the grid but are held back from operating at their maximum or minimum levels to have some flexibility to ramp up or down quickly.

So the less of that we have because we're shutting down those plants to make room for wind and solar when they're producing, the more alternatives we need or we end up actually having to curtail wind and solar output in order to keep a minimum amount of those fossil generators online to maintain reliability. So that's the first option is you just curtail the renewables. But of course that's wasting free energy. And so we'd like to have other ways to take advantage of the wind and solar and still manage that short term variability. And that's where things like batteries and synchronous condensers and capacitor banks and other devices that we can add to the grid to augment their flexibility on those short timescales come in.

David Roberts

Is it safe to say that with those options, especially batteries, do you worry as we get closer to net zero, closer to a 100% carbon-free system, do you worry about this second to second variability or do you think basically with batteries, we basically have that problem solved and can handle that?

Jesse Jenkins

It's very low on my worry list. And that's not because it's not something that somebody has to worry about. It's just that I think there are very good control engineers and power engineers and grid operators out there solving these problems already. And we've known about these problems for decades. And so there are a lot of solutions already out there. And so pretty much everything is figured out in this space, I would say, with the exception perhaps of the physical inertia that really immediate microsecond, microsecond response that we get from the physical spinning mass of all of these interconnected generators.

Beyond that, the next line of defense is what we call frequency regulation or frequency reserves. Those are the ones that sort of move up and down on a second by second time scale to track a control signal. They say go up a little bit, go down a little bit to sort of keep things balanced out. And a few years back, maybe about a decade now, some of the grid operators in the US opened those markets up to batteries and particularly lithium-ion batteries —

David Roberts

Grid services.

Jesse Jenkins

Yep. And it turns out that lithium-ion batteries are incredibly good at this job because you don't need very much battery capacity, right?

You don't need a bunch of energy in the tank to be able to do this because it's generally about neutral, right? You're sort of going up and down and up and down and up and down around a middle point. And so you can maybe only have 15 or 30 minutes of full power discharge capability. And that's still enough to provide frequency regulation because you're really only charging discharging on few second to minute long timescales and they're incredibly fast. So the power electronics responds really quickly to the control signal and it can flip from full on to full charge very quickly, much faster than a physical generator could do, even a hydro generator, which traditionally were the fastest response.

And so when PJM and other grid operators opened up these markets in around maybe 2009, I think, to storage, we saw the first commercial scale deployment of grid connected batteries and they basically ate the entire market because they're just the best way to do this.

David Roberts

Yeah, there's not much left of that market in places where it opens up. It's pretty easy to cover those needs.

Jesse Jenkins

Yeah, you only need probably a few thousand megawatts of frequency regulation nationally. So that's like a few nuclear power plants worth of capacity nationally. And we have built that and a few hundred megawatts usually per grid region. And so the batteries just came in and you built a few grid batteries and they have taken on that role very capably. And the market is sort of full.

David Roberts

Is that what they call synthetic inertia?

Jesse Jenkins

No. So that's the next challenge. So yeah, this is frequency regulation, which is on the sort of second by second timescale. That initial response of the physical inertia is like milliseconds that just happens instantaneously because again, all of the devices are interconnected and physically synchronized. And so when the demand goes up a little bit, all the generators kind of lean into it a little bit and produce a little bit more and vice versa when the demand drops. And so that is where we currently depend on the physical inertia of generators who are connected to the grid and producing power.

And we get that for free. It's not something we pay those generators for, it's something that they just physically have provided for free. And it's been ample and well in excess of the amount that we generally need with rare exceptions like islands or micro grids where it's much more challenging to keep enough inertia. And so we haven't been paying generators for that. It's just sort of a bonus that we get from having these synchronized generators online and grid connected inverter based resources, which includes wind, solar and batteries and fuel cells and any other kind of direct current device like that electric chemical device.

They're not synchronized spinning masses of copper and steel like generators are and so they don't provide that physical inertia. And so synthetic inertia is basically a computerized control strategy to make those inverter based resources act like a physical inertia device would and to sort of automatically compensate based on local measured characteristics. This is too fast to send a control signal out even from a centralized dispatch. It has to be locally metering what's going on, and directly responding to the local conditions without knowing what's going on in the rest of the grid. And so you are basically designing control strategies to use the power electronics in an inverter to change the reactive power production or consumption of the battery or the solar panel or the wind farm, which can, if you do it right, can tune it well, can simulate and replace the physical inertia that you get from the system.

This is something that, again, people have been working on for decades in the lab. We've done lots of experiments. It's one of those ones that grid operators are very reluctant to deploy at scale and rely on in a field experiment because if it goes wrong, the grid goes down potentially. And so it's one of those ones like, it probably would work if we were willing to just jump off the cliff and try it. But for obvious reasons, this is an incredibly conservative industry. And so there's been various small scale deployments to try to see how it works.

But nowhere in the world that I'm aware of is relying substantially on synthetic inertia today. Again, with the exception, maybe, of small micro grids. I should say that there's a dumber, simpler and slightly more costly solution that we can fall back on, even if that doesn't work, which are called synchronized condensers, which is basically a generator without the turbine, without the spinning prime mover that are just spinning hunks of copper wires in magnets that are on the grid and are synchronized. They consume a little bit of electricity to spin around and stay synchronized. So they do use up some variable, some of the energy production, and they do cost money because they're basically half of a generator, the magnet part without the turbine.

But these have been around for a long time, and they're used in certain locations to buffer short term variability from, say, starting up a steel mill, electrical steel mill or aluminum smelter. That is this big new demand that comes on very quickly. They've been used in that context and to support the voltage at certain little pockets in the grid where it's been hard to do so. And I recently read a thesis a dissertation from University of Melbourne PhD student who modeled this as an option without any synthetic inertia in the grid, but a minimum physical inertia requirement and found that it would add to a fully decarbonized system about 1% or 2% to the cost of that system, if we only relied on synchronized condensers to do the job.

So, again, these are mature technologies we know how to build. At worst, they add a couple of percent to the system. At best, they're free, because all of these inverter connected devices that we're adding can perform the same role as physical inertia via synthetic controls. And again, that's more perspective at this point, but I think it's an imminently solvable challenge.

David Roberts

Okay, so this is the super short-term variability. Let's call it a solved problem, at least as these things go. So let's move up a little bit. Then you get big ramps, ramps in the morning when the sun comes up and goes down. Occasionally wind will die down quickly. What do you do about these sort of minutes to hours midday variability?

Jesse Jenkins

So I'll say what we do now and then what we could do, which would be better. Right now, again, we rely on fast-acting thermal or fossil power plants to play that role.

David Roberts

Mostly natural gas, right?

Jesse Jenkins

Mostly natural gas. Sometimes diesel, internal combustion engine reciprocating engine generators. So what we do is we commit a bunch of generators that are ready to act when the sun is about to set, and they are operating at their minimum stable output level, which is not zero. So generally, they don't get to just sort of sit there and park at zero. They have to be on at somewhere between 40 and 60% of their output, usually, or 30 and 60% of their maximum output is as low as they can go. So during the middle of the day, when the solar is at its maximum, many of these are shut off, but then they start to get recommitted in the afternoon hours, right before this evening ramp and run at sort of crouch there at their minimum output level and then ramp up really quickly as the sun sets to compensate.

And so gas turbines are really good at this. They're really fast to respond. I mean, they're, what, run jet engines, right? I mean, jet engines are basically gas turbines. And we derived our gas turbine generators from jet engines. So the fastest ones are as fast as a jet fighter. They're literally the same engine. We have one here at Princeton in our central plan as an aero derivative gas turbine. It's the same. It's used in like an F-16 fighter. So they're really fast to respond because they can handle a dog fight. But then you also have bigger what are known as frame combustion turbines and combined cycle power plants that usually also use these frame turbines connected to a second steam generator, so they use the hot gas from the combustion turbine as the steam generation source for a steam turbine as well.

David Roberts

That's combined cycle.

Jesse Jenkins

That's why we call them combined cycle, because they combine a Brayton and a Rankin cycle, a gas and steam turbine.

David Roberts

So we don't want to do this. We can't do that in a fully decarbonized grid. I mean, you can, I think, keep some fossil plants online and use them very, very rarely. But I don't think you could do that on a day-to-day in a decarbonized system.

Jesse Jenkins

Yeah, I mean, theoretically you could do this with a hydrogen turbine or something like that, but you would probably consume way more hydrogen than you want because hydrogen is a very expensive fuel to produce. And so, yeah, you don't want to keep doing this on a day to day basis. But I want to add that, again, we do this now. And this is how we keep the grid running even when California gets nearly 100% of its electricity during the middle of the day from solar and wind. The downside is that because you have to have those generators running at their minimum level before they can ramp up, because it takes between 30 minutes and several hours to turn on once you call on them.

And like a combined cycle power plant takes the longest of the gas generators. The air derivative turbines can maybe turn on in 30 minutes, but generally they have to be sort of on and parked and ready for that ramp. And you need some of them just sitting there, even for the unforecasted variability. Right. We know that the afternoon ramp is happening, but —

David Roberts

They're displacing wind and solar while they're sitting there —

Jesse Jenkins

Exactly. And so that limits the ability of wind and solar to displace their fuel consumption because they're on, not because they're the cheapest generator to meet demand at that hour, but because we know we need their flexibility for the ramping periods or the contingencies that we're waiting for. So it would be great if we had a really fast way to flexibly produce or consume energy to match the variability of wind and solar. And fortunately, there are lots of good ways to do that too, batteries being the first and most significant new source of that kind of hourly flexibility. But also the demand side can be called upon much more as well.

David Roberts

We should note that batteries in this capacity are way faster than the turbines.

Jesse Jenkins

Yeah, once again, they can do that frequency regulation on a second by second basis. So they surely can deal with the sunset. And yeah, they don't need to be committed. They can sit there, they're on the grid all the time. They can go from fully discharging to fully charging or back. So they actually have twice the ramping capability. Right. Because if you have 100 megawatts of battery, it can switch from being 100-megawatt consumer to 100-megawatt producer, giving you 200 megawatts of ramping in that battery and they do it in a second, right. From one to the other. And so they're really good at this.

And we're already seeing them deployed at gigawatt scale in a lot of markets in the world, particularly those with high solar penetrations, because this daily cycle is so predictable. You get really cheap power during the middle of the day and really expensive power in the evening ramp and so they can make money arbitraging that spread, buy low and sell high.

David Roberts

Yes. And you also mentioned demand shifting, which is just trying to move large sources of load under that curve when solar is producing all this energy away from the times of sharp ramps.

Jesse Jenkins

Exactly. You mentioned at the beginning that we're shifting from a system of dispatchable generation to one of variable or nondispatchable generation. Well, we're also hopefully shifting from a system of nondispatchable demand, constant demand that doesn't know what the price of electricity is and just keeps consuming no matter what to dispatchable or flexible demand. Because power electronics are cheap, computing power is cheap, controls technologies are very sophisticated and it would not be very hard to wire up a whole bunch of HVAC controls and hot water heaters and EV chargers to be much more flexible on both minutes to hours to even daily timescales.

There's a lot of flexibility in an EV, right? I mean, I have a 300 miles range EV. That's enough for five to seven days of driving in my typical driving pattern. So not only can I shift which hours during the night or daytime, if it's plugged in at home during the day, that I consume energy, but I can even choose which days to consume, right? I can shift from Monday to Wednesday or Wednesday to Saturday, right? And that's probably the most flexible of these loads. But think about a hot water heater. That's just a big thermal battery.

It's a big insulated tank of water and when you charge it and heat it up or not, it is quite flexible. You can do it right as you're drawing down the hot water, or you can preheat it and get it above the desired temperature. And there are even more sophisticated ways to do that. Parts of the world that traditionally relied heavily on hydro or nuclear power, where you had the problem of too much generation overnight: What those parts of the world did way back in the 50s through to today is they have ceramic brick heaters that heat up a big ceramic brick when the power is cheap, and then let that brick reradiate heat into your house during the daytime right when —

David Roberts

Thermal storage!

Jesse Jenkins

It's cheap —

David Roberts

We love thermal storage.

Jesse Jenkins

And again, this is not Sci-fi. This is like they do it in Quebec and the UK and they've done it since the 60s. So we could be building big thermal batteries in everyone's home whenever we put in a new HVAC system, right? It could just be part of the HVAC design, is that you have a big hot water tank or a big hot brick tank.

David Roberts

We also get to what I think is one of the most fascinating questions and I think unexplored as yet questions in this whole area, which is when you are talking about big industrial loads, how much of that load is shiftable, how much of big industrial load could be shifted? I don't think there's been a ton of exploration of that to date. And I think we're going to be finding out soon what the answer to that question is.

Jesse Jenkins

Yeah, and actually, I have a paper that we just resubmitted this week. After revisions on this, we can provide a link to the working paper in the show notes on what we call demand sinks. So these are consumers that are extremely flexible in when they decide to consume and will basically match their consumption to the availability of low-cost power. And since wind and solar are the cheapest way to make low-cost power, that will mean they can sync up their output to wind and solar. And so we actually offer four different categories of demand in that paper to try to help kind of talk through the options here.

So if we're thinking about the demand side, there's firm demand. That's the normal stuff that we're used to having where it wants to have three or four nines of reliability, we usually say, which is like 99.99 or 99.999% reliability basically all the time. And that's most of our current demands. Residential, commercial, lighting and cooking and refrigeration, industrial, most industrial loads, hospitals and other critical loads, and most heating demand today. And that's the demand we expect to serve. And if you don't, it's a big problem, right, that's when you're having rolling blackouts. Then we have interruptible demand or curtailable demands.

This is the sort of category of demand response that we have. So these aren't necessarily shifting their production, they're just stopping consumption when the price of electricity is really high or when they're being paid a lot of money to do so. And that's where things like aluminum smelters or industrial demand response contracts that they have with a whole bunch of industrial refrigeration warehouses or consumers with backup generators who can turn on and get off the grid when the price of power is higher than the cost of running their generator. That's a lot of the demand response we have today.

That's also where hydrogen boilers, other things could potentially play a role. So there's some new ones coming in that category too. Those ones, you don't want to call on those very often, but they can help you avoid building a bunch of generation that just sits there for that like half a percent of the hours of the year when you really need some backup because they can consume less during those periods for a few hours at a time. Then you have what we call shiftable demands. These are the ones we were just talking about where you can move around when you consume within a kind of hourly or even daily scale.

Flexible EVs, heating demand, data centers potentially can do this — something Google has explored, moving around in both space and time where they do the compute loads.

David Roberts

Yeah, Google is doing a ton of work trying to figure out how much of their compute load is shiftable.

Jesse Jenkins

Yep. Yeah, one of my former MIT classmates who I happened to see last weekend at a wedding was working on this with Google. Yeah, fun stuff to optimize, right? Great control problem to play with. Then things like agricultural pumping is another one that's often done already. Like California irrigation districts will shift when they pump their water into the canals and the reservoirs and things like that. So that's another tried and true demand. And so those demands, they meet their needs, right? It's just a question of when they do it. So it's different than curtailable or interruptible demands. And then this last category of demand sinks are the really price sensitive consumers who really can choose when to consume.

And this is where it's an interesting question which categories will emerge here? What we found in our paper is that in order to do this, you kind of need a weird combination of things. You need something that's highly automated because you can't have a lot of labor sitting around idle when you stop consuming. Right. Because that's usually too much of a cost. It needs to be very energy intensive, meaning a big chunk, if not the most of your cost of production is the cost of energy inputs. And it needs to produce something of value that isn't so valuable that you never want to turn off.

This is the current problem with crypto mining with bitcoin, is that the bitcoin prices are so high — or they have been, I don't know, they're all over the place now, so maybe they're lower now — that you want to consume even if the electricity is several hundred dollars per megawatt hour, $100 per megawatt hour, it means you basically consume 98, 99% of the time anyway. So that makes you more like an interruptible demand, not a flexible consumer. But if the price of the product is lower, where your willingness to pay is only ten or 20 or $30 a megawatt hour, then you want to concentrate to when the load is — or the power is cheap.

And then finally it has to be not very capital intensive because if you're going to idle your production and shift your consumption around to low price hours, you're going to have a low utilization rate for that capital, all that equipment. And so it can't be too expensive or you'll need to run it all the time. And that's where kind of direct air capture fails the test right now because it meets the other requirements, highly automated, totally energy intensive, but it's too capital intensive to run at anything less than maybe 95% of the time. So there are a few here that I think may work.

And one is, I think we share is one of our favorite technologies out there, which is resistance heating with thermal storage. Right. So Rondo or Antora, who you've interviewed on here I'm on the advisory board of Rondo Energy, I should disclose and big fan of what they're up to. But here you basically take in renewable electricity whenever it's available and you use a big thermal battery like the hot water tank or the ceramic bricks that we're talking about in the home to decouple —

David Roberts

Box of rocks.

Jesse Jenkins

Yeah, or a box of rocks or even just rocks in the ground covered up with dirt to decouple the constant heat demand of an industrial process from the variable input of the wind and solar. And that's a great option. Another option is to just install resistance heaters alongside gas boilers. So don't replace the gas boiler fully or at all, but run it in a hybrid mode, where when energy prices are cheap electricity prices are cheap, you switch off of gas to electricity, and when electricity prices are higher than the gas cost, you go back to gas.

And that makes it look like a very flexible demand sink. That was a technology we put in the model for the Net Zero America study and we saw like terawatts of that load in the final Net Zero system. Right.

David Roberts

So the system wants —

Jesse Jenkins

Wants that cheap renewable electricity if it can use it. Right. So if you can find a way to meet your constant energy demand for industrial heat while tapping into this cheapest source of energy, period, whenever it's available, that's a really valuable thing to do.

David Roberts

Yeah, I think that one's going to be huge in like a decade.

Jesse Jenkins

Yeah, I think so. I think the industrial heating is the biggest one that people largely are sleeping on. Although not you and I of course. And then the other one that's getting most of the attention right now I should say is hydrogen production from electrolysis.

David Roberts

Right? Yeah.

Jesse Jenkins

Where again today electrolyzers are pretty expensive so you probably want to run them at least 70% of the time. But that's still very flexible. I mean 30% of the hours is a lot of hours you can shut down. And as the cost of electrolyzers fall, which we expect they will just like solar and batteries did, probably by 50% over the next six years or so, then you can afford to run them at 30 or 50% utilization rate and then they're a really good flexible consumer. Now I want to add that both of these, any of these demand sinks, what we found in our paper, they don't really help the broader grid operate.

What they do is allow you to just tap into that weather-dependent but very low-cost clean electricity and make greater economic use of it and displace fossil energy consumption elsewhere in the energy system. But it's sort of additive to all the demands that we already are going to have in the grid and the flexibility that we need to handle those demands. So we added a whole bunch of demand sinks in our modeling and we found is that it didn't really reduce the amount of firm generating capacity or battery capacity that you wanted on the system, but it also didn't increase it much. It just sits there and soaks up that good cheap renewable energy when it's there —

David Roberts

But allows you to use more wind and solar.

Jesse Jenkins

Much more. Yeah, much more.

David Roberts

Okay, so we've covered seconds and we've covered minutes and hours and it sounds like on the minutes to hours thing, combining batteries and then all these demand, as you say, these various sources of shiftable demand. Do you think that the sort of ramping problem is solvable to solved? Let's say it's also low on your list of worries.

Jesse Jenkins

I think we have solved that problem in the sense that we know the technologies; they are not Sci-Fi, they can be deployed at scale now. They are not deployed at scale yet at the scale we would need. So in the next ten years or five years we are still going to have to rely on those gas turbines and things like that to do a good chunk of this. But over time, as we build more batteries, as we wire up more flexible loads and give them the incentive to participate in this demand shifting, as we get more interruptible consumers signed up, we'll be able to do more of this without relying on so much gas backup capacity.

And that's a good thing from a decarbonization perspective.

David Roberts

So then we get up to hours to days variability in terms of diurnal cycles, the sun going down every night in the sort of daily storage needs. What are our options there?

Jesse Jenkins

Yeah. So here we probably again, we rely right now on fossil generators, right. Ramping up and down. We can rely to some degree on lithium-ion batteries. They are most economic to operate for just the highest price periods for that sort of peak in the evening ramp, or maybe twice if there's a double peaking system in the morning too. And it's not a function of like physically you could slow the discharge rate and run a lithium-ion battery for 24 hours of discharge. You just have to discharge at a much slower overall rate than you could. And so, economically, batteries and given their cost today, are really best suited to somewhere between like two and six hours of duration.

David Roberts

Yeah, although that number has been edging upward, I feel like, for as long as I've been paying attention.

Jesse Jenkins

Right. Because the reason I'm emphasizing that technically they can do longer is that what limits that is not the technology per se, it's the economics of the battery.

David Roberts

I mean, you could theoretically just stack batteries to the heavens and solve all of this if you had infinite money.

Jesse Jenkins

If you've got ten four hour batteries, you've got a 40 hours battery. Or if you have one four hour battery that you discharge at one 10th of its rated capacity, you have a 40 hours battery. Right. So that's not rocket science. The problem is you need to make enough money every time you charge and discharge to cover your overall fixed cost of a battery. Right. So batteries make money off of kind of capacity contracts and flexibility services. So they're sort of paying for their standby ability, but also from buying low and selling high. Right. This sort of buy sell spread.

David Roberts

Right.

Jesse Jenkins

And the problem with any arbitrage play is that the more you buy low, the higher the low price gets. And the more you sell high, the lower the high price gets. And you're not the only one playing this game.

David Roberts

Everybody else is arbitraging too.

Jesse Jenkins

Exactly. And so we've seen this happen is that basically the price spreads start to collapse as you build more of these flexible demands and more of these batteries all kind of playing on the same price signals. And so that creates a race between the declining cost of these technologies and their declining value as you do more and more with them. And as long as the costs keep falling, or we develop cheaper lower cost per kilowatt hour of storage capacity batteries or storage technologies, then batteries can stretch to play a longer and longer duration role. And so, just to put some numbers on that, right now lithium-ion battery systems are probably $250 to $350 per kilowatt hour of capacity installed. If they fell to 100 ish dollars —

David Roberts

Isn't that DOE's stretch goal?

Jesse Jenkins

Yeah. So the pack costs are already falling below $100. So the actual battery pack itself, but you have to install it and give it cooling and power control electronics and wire it up to the grid and everything. And so the labor and the balance of system costs, just like for solar modules, which are only like a third of the cost of a solar system now at scale, the pack cost is a piece of it. And so we would need to get the pack cost down a lot more. If you want to hit the $100 per kilowatt hour total system cost level, you'd still need to probably a stretch for lithium-ion to hit that target.

But maybe lithium-ion phosphate batteries are looking like a better option to do that. Sodium sulfur batteries or sodium-ion batteries, sorry, are being introduced now as a cheap, low-range option for EVs. Well, EV batteries need to charge and discharge very quickly and they need to have a high enough energy density to give you a lot of range in a small package for not very much weight. None of those apply to a grid battery.

David Roberts

Yeah, right.

Jesse Jenkins

A grid battery can charge and discharge over hours, not minutes. And the energy density doesn't matter. The gravimetric density, the weight part, doesn't matter at all. The volumetric density only really has a small impact on the amount of space you need to put them, which can impact the installation costs and cost of the land.

David Roberts

So here you get into this weird because we're going to discuss later long-duration energy storage, where we're talking about days and weeks. So here you're getting into this weird sort of liminal space between —

Jesse Jenkins

Yeah, diurnal storage.

David Roberts

lithium-ion batteries, which gets you up to whatever, six, maybe eight hours.

Jesse Jenkins

Yeah, maybe 16. Right. This sort of diurnal scale is what it seems like is necessary to manage most of the day-to-day variability of demand and solar and wind, which they have pretty pronounced daily cycles because the sun goes up and down every day.

David Roberts

So that's where you get into flow batteries and things like this, which are sort of like longer than short term, but shorter than long term.

Jesse Jenkins

Yeah. So, iron flow —

David Roberts

I don't know how much of that space is going to be left. My sort of instinct is that that space is going to get eaten from below by lithium-ion and eaten from above by long duration and there's not going to be much of it left. But, what do you think?

Jesse Jenkins

Yeah, I mean, it depends. It's sort of a race to market, I think, and which technologies kind of get to scale and get on that cost curve first because there's a lot of path dependency here, right. If you can get to market and scale up and start driving down your costs before another startup can, you may be able to edge them out of the market. And this is something that any of these startup battery companies need to keep in mind. Because lithium-ion and sodium-ion and all of the automotive battery technologies are coming like a freight t