B (3:59)

Like you said, this is a general problem with power systems around the world. About 130something countries now have targets for renewable power deployment. Now all those countries are at very low levels of penetration right now. So I mean, even in the US which is a very robust renewable power market, the penetration of wind and solar power is just over 4%, which is very, very low. At those levels of penetration, these issues are negligible. But when we start getting into 20, 30, 40, 50, 60% penetration, which is what a lot of countries are imagining now, the intermittency starts to really add up and start to impose serious costs. You mentioned coal. One of the reasons I focused on South Africa is that this is a country like India and to some extent like China and other developing countries that is very dependent on coal. Over 90% dependent on coal for electricity right now. Problem is coal does not play particularly well with intermittent renewable energy sources. Gas plays much better. It's able to fluctuate over time to help smooth out the wind and solar.

A (5:05)

You can ramp it up and ramp it down quickly.

B (5:08)

Exactly. But coal is a different beast. It's thermal in nature. It's not as responsive. So that adds another layer of complexity and difficulty to the planning problem.

A (5:18)

And what happens then? If you are like South Africa and you have a very ambitious renewable target, you'll probably tell us, I think it's what, 10 or 20% within 10 or 20 years, something like that. What happens if you build up lots of wind and solar and you don't take this into account?

B (5:38)

So the way I explain it is that this is really a two pronged problem. It's a problem of cost and it's a problem of reliability. And the two are related. So as we build out wind and solar power, the most obvious problem is cost. These technologies are more expensive, at least in the short term. They increase the price of electricity. That's problematic in South Africa. South Africa has historically had some of the lowest electricity prices in the world. And only recently they've realized those prices are completely unsustainable. And now electricity prices are going up 20% per year, not because of renewables, but because they need to have money to invest in the power system. So if renewables are going to add to that problem of rising costs, that's a political issue. And then there's the reliability problem. South Africa has had years now, about five years now, of really serious reliability and blackout problems. Again, if renewable power is not going to, is going to make those problems worse, it's going to make it more difficult for them to reduce their emissions. So South Africa is a very interesting case study in terms of addressing both the reliability and cost issues associated with renewables.

A (6:49)

Just to draw you out a little bit on cost, I'm imagining that part of the cost issue is if you're going to install large amounts of wind and solar and then they generate the energy in a period when you don't need it, then the cost per usable kilowatt hour is going to be even higher because you're going to have some generation when perhaps you don't have sufficient demand. Am I right? Is that part of the problem?

B (7:14)

That's right. So the environmental community has not liked to admit this, but the bottom line is that anytime you take what we call a dispatchable source of electricity, like coal or gas, and you replace it with an intermittent source like wind or solar, you by definition increase the cost of running the power system reliably. There's no way around that. Now we can do things to try and reduce these integration costs, to try and make the technologies play nice, but at the end of the day, it does oppose an additional cost. And the goal of a good system planner is to try and minimize that cost.

A (7:49)

What about storage? Can't you just put the power in batteries and pull it off when you need it? Or pump water up to a reservoir and let it sit there and then run back down when you need the.

B (7:59)

Power, you can use pump storage. And that is probably the most realistic and cost effective large scale storage technology Right now. Batteries are just. Unless we had large scale electric vehicles, battery technology is really too expensive to use for this Type of thing. What makes again, South Africa an interesting case is that although they have some pumped storage research, some pumped storage resources that are quite good, they're limited in hydroelectric. And elsewhere in the world where we see renewables being integrated easily, it's usually because there's a large amount of hydroelectric power which is cheap, clean and can be ramped up and down quickly. So again, South Africa has these unique challenges that make it a great case study.

A (8:43)

Presumably with storage there are some efficiency losses as well. You don't get 100% of the power back out. You use energy in order to store.

B (8:50)

Right. And the bottom line is that if it comes down to using storage to try and address intermittency or using more natural gas, especially where gas prices are today, it's just much more cost effective to use natural gas.

A (9:03)

Okay, so we're going to take a break. When we come back, I want to ask you to walk us through how you approached identifying the locations for wind and solar that could help to address the intermittency problem. We will be unpacking spatio temporal modeling. Stay tuned. We'll be back in a bit. Welcome back to the Global Prosperity wonkast. I'm Lawrence MacDonald. My guest today is Kevin Ummel. He's a visiting senior associate here at the center for Global Development and we're talking about a new paper in which he has used spatial temporal planning, spatial temporal modeling to try and figure out where South Africa should put its wind and solar in order for it to be not only maximally efficient, but in fact even to make it possible to have large scale wind and solar without having disruptions from the fact that the sun doesn't shine at night and the wind may not blow when you need it. The problem of intermittency and ways that that could be addressed. Kevin, given the problem, you approached it in a way that was massively data intensive. Explain to me the approach that you took.

B (10:40)

So let's be clear. South Africa, the government has committed to having wind and Solar provide about 20% of generating capacity by 2030. We call that 20% penetration. And there are plans being developed by Escom, the state owned utility, to push that up to 40% by 2040. Those are quite impressive penetration numbers.

A (11:03)

It sounds heroic. All countries should be so ambitious. That sounds wonderful.

B (11:08)

It's great. I mean, compared to what some other developing countries are doing, the large ones like India and China, it's relatively modest. But for South Africa, it is very aggressive and very laudable. The issue is that we have lots of experience planning Power systems that have no renewables in them. And I like to use the analogy of a Rubik's Cube, the cube that you twist and turn different ways and try to get each side to be a uniform color. I'm sure everyone's familiar with it.

A (11:37)

I never could do it, but I once saw an 8 year old do it in about 30 seconds. Blew my mind.

B (11:42)

So that makes my point for me, which is that solving a traditional Rubik's Cube is a bit like planning a conventional power system, one that depends on fossil fuels. Now imagine I took that same Rubik's cube and maybe 2 or 3% of the squares on the surface. I modified them so that instead of being a constant color, they actually changed color every few seconds. And now I asked you to twist and turn that Rubik's cube so that you had solid colors on every side for some maximum percentage of the time. That's like planning a power system that has a low level of wind and solar power penetration. It's difficult, but it's not that hard to do. But now imagine I took the same Rubik's cube and Instead of just 2 or 3% of those squares alternating colors every few seconds, imagine 40 or 50% were doing that. It's a psychedelic Rubik's trying to arrange that in a way that maximizes the probability of having solid colors at any one time on every side. That's like planning a high penetration wind and solar power system. Very, very complex.

A (12:46)

It's devilishly hard. I shudder to even think of it, given my bad experience with the basic Rubik's Cube.

B (12:53)

Yeah, so that's why we have to bring to bear so much data and new analytical tools, is because we're dealing with a problem with which we simply haven't dealt with before. So in this study, what I did was use a large amount of high resolution data from NASA, from European satellites, from numerical weather models, to simulate how photovoltaic concentrating solar power and wind turbine technologies would perform hour by hour over a 10 year period across South Africa. And we can take that massive amount of data and put it into a model that says, okay, how should we arrange these technologies across space to provide electricity reliably at the lowest cost?

A (13:41)

So you have data that's showing where and when the sun shines and for how many hours, taking into account both weathering patterns and seasonal fluctuation and where and when the wind blows. Is that basically those are the two parameters that you're looking at?

B (14:00)

There's actually other parameters that are secondary, but Those are the main ones. So for example, ambient air temperature impacts the efficiency of photovoltaics and CSP as well. So one of the reasons that this study I think is quite unique is that I've taken an approach that is really quite detailed with respect to how these technologies are going to perform.

A (14:23)

And if I remember correctly, you've also taken into account some sort of non power related things such as biodiversity. You wouldn't want to put a wind or solar farm in a place that's got high biodiversity. Also where the people are, where the demand is and where the existing grid is. So I think of it, and correct me if I'm wrong, as sort of a series of filters or layers where the best places to put the wind and solar are going to emerge. Only after you have run these many, many filters to identify them. Is that right?

B (15:01)

Right. So there's an extensive, what we call terrain screening layer involved here. It basically screens out places like national parks or game reserves, places where the soil is insufficient. There are lots of things that prevent us from putting technologies in certain places.

A (15:19)

And out of all of that, do you have an answer that the ESKOM in South Africa, if they chose, could look at it and say, great, we don't have to do all this hard work. Kevin Hummel has done it for us. We know where to put the solar and we know where to put the wind.

B (15:36)

Well, how ESKOM should respond is maybe a different issue. But I'll tell you what the results are first. So if we take what I call a traditional planning approach to this problem, which is the approach that South Africa is using right now, we can simulate what that scenario would look like in terms of cost and emissions. And we can calculate something called the abatement cost, which the economists in your audience will be familiar with. And that's just a measure of how cost effectively is this policy option, how cost effective is this policy option at reducing greenhouse gas emissions? A low abatement cost is better. If we look at the traditional planning approach, we find that South Africa could hit 40% penetration by 2040 with an abatement cost of about 26.$5, which is consistent with estimates elsewhere in the literature.

A (16:24)

I want you to say that once more slowly, because I think it's important. You said they could reach 40% penetration by 2040 and what would the abatement cost be?

B (16:36)

$26.5 per ton of CO2 eliminated.

A (16:41)

Okay, I'm going to make it easy for us. I'm going to round to the 26, so it's $26 a ton.

B (16:49)

Now, if we took the same amount of wind and solar technology, but now said, okay, let's allocate it optimally across space to minimize the abatement cost. If we do that, we find that we can reduce the abatement cost by almost 15%, not by changing the amount of wind power, the amount of solar power, just by changing where it's located spatially. And that's in part by putting the facilities in more efficient locales, but it's also by dispersing them spatially to reduce the cost imposed.

A (17:20)

So you get a 15% savings by taking into account the placement, to put it very simply, to make maximum use of, of the wind when it blows and the sun when it shines.

B (17:36)

Right. And getting these resources to play nice with each other hour by hour. So if we disperse our wind turbines so that production drops off in the late afternoon, and then disperse our solar facilities so that their production ramps up in the late afternoon, now we can reduce the intermittency cost. Now we did another scenario run where we allowed the model to decide how much of each technology to deploy. And this result was very surprising because what it found is that if we allow the model to decide how much wind, how much solar, and where to put it spatially, we can reduce abatement costs by 40% compared to the traditional way of planning.

A (18:18)

Wow. I was happy with the 15%. That sounded like a big savings to me.

B (18:23)

Yeah, 15% would be fantastic. This 40% number. What's particularly intriguing about it from a research standpoint is that in that scenario, the way it achieves that savings is by going gangbusters on photovoltaics. So the conventional wisdom is that it's better to sort of use an all hands on deck approach. Let's disperse our investment across many different technologies. That may make political sense in terms of getting buy in from everybody when you're putting together a strategy. But these results suggest that in the long run it may be optimal. I'm not saying always, but it may be optimal to focus on a particular technology or two that seem to do a really good job of driving down the abatement cost.

A (19:07)

I feel like I'm on a late night infomercial where we say, but wait, there's more. The 40% is very exciting, but I'm puzzled. I'm not sure that all the listeners will follow this, but some of them will. My understanding had been that when you look at utility scale power, it's concentrated solar power that is having mirrors that heat A salt medium to generate steam. So it looks more like a conventional power plant, but the power is from the sun. That that was going to turn out to be more cost effective than the photovoltaics, which are these, you know, these solar cells that everybody's familiar with that you can put on your own roof. Can you address a little bit? That seems like a surprising thing to me that you would say the photovoltaics is, is the way to bring down the cost so sharply.

B (19:52)

Well, I'm not suggesting that this is a band aid for all problems. In this study, I used the cost assumptions that the government has used in their own planning documents. And it just so happens that in their planning documents they assume that the cost of photovoltaics will decline quite precipitously in the future. So it could be that that assumed cost decline over time is driving this PV dependent result. When we allow the model, sort of when we, the model complete freedom to decide what to do on the question of PV versus concentrating solar power. In this case, again with the cost assumptions that I'm using, the model decided that it was actually better off to just use a lot of photovoltaics and then introduce more natural gas to provide electricity during those evening periods when the demand is high. So one of the reasons people have thought CSP would make more sense is that it has an ability to produce electricity in the evening period when the.

A (20:53)

Sun is down because of the heat contained in this liquid salt.

B (20:59)

Exactly. It has this ability to sort of extend its production period into the evening peak periods. However, what these results suggest is that given the cost, the technologies, which I want to stress have considerable uncertainty. It's better off, or at least more cost effective just to use gas in those peak periods and let PV do what it does best, which is produce a lot of electricity during the day.

A (21:21)

It's a fascinating finding and obviously very important for South Africa. I want to zoom out if you will pan out and think for a minute about the implications for others. Obviously, as we've seen in exploring the result between the PV and the concentrating solar power, the results you get are going to depend a bit on your assumptions and those will change over time as technology changes. I think the broader contribution of your paper is that you have been very detailed and careful in showing how you did the work. And other countries, which, as you point out, have similarly ambitious renewable targets, could then apply this method. And you know, they would come up with different results depending on their assumptions, depending on their terrain, but they would be able to do the spatial Temporal modeling. If they can find somebody as smart as you and then presumably get pretty substantial savings, is that what you're hoping will happen with this paper?

B (22:24)

That was one of the original motivations. So I mean, take the US as an example. We have probably the world's preeminent renewable energy research institute just down the street from me in Golden National Renewable Energy Lab. They recently published a renewable futures study which did this amazing modeling study to show how the United states could get 80% of its electricity from renewables by mid century, taking into account all the intermittency transmission issues and all the rest. Fabulous study. Problem is it required over 100 co authors in 35 organizations and the data it's relying on goes back five years. We're talking millions of dollars to get this result. I'd like to see a capability for countries everywhere to do low cost, not as high quality, but do low cost initial modeling of this kind to give them an idea which technologies seem to make the most sense, which areas of our country are likely to be most suitable for developing these technologies. I think that there should be a market for more low cost initial modeling to help countries down this path.

A (23:31)

Well, I certainly hope that you're right. It's very exciting work and as somebody who couldn't begin to do this kind of work myself, just to have the opportunity to understand it is very exciting. Do you have a parting thought for those people who may not be particularly concerned about South Africa, but they're trying to get their head around how they should think about doing this kind of modeling in relationship to renewable targets?

B (24:00)

I mentioned that the results that I've gotten here are very much dependent on the cost of technologies over time. This is the great uncertainty. So I'd like to see investors like the World bank, for example, government, energy ministries, approach this power system planning problem the way a really good financially savvy person approaches retirement planning. In other words, we know the future is uncertain, we don't know what's going to happen with cost of technologies, but we should try to develop deployment strategies and transmission, build out plans that leave open as many low cost possibilities as we can so that as we move forward and learn more, we have the ability to go down those paths that make the most sense. In other words, if we make bad choices now about where we put our facilities or where we build out transmission, we may effectively close off opportunities 20 or 30 years down the road that are going to get us those biggest savings. So I'd like to, I call this developing deployment plans that are robust to future uncertainty. I think that should really be the big picture goal.

A (25:10)

So you haven't used the word diversification, but having recently reviewed my own retirement planning, I know that's what my advisor told me. You know, you wouldn't want to put all your retirement money in stocks or indeed all of it in bonds. And by the same notion, you wouldn't want to rush and put all of your renewables into one technology or in fact, to make bets on several technologies, but perhaps in the wrong place, right?

B (25:37)

I mean, the one big difference between portfolio theory for retirement and renewable systems planning is obviously that the latter is concrete and steel. So if we make a decision now about where to put a transmission line, for example, that infrastructure is going to be there for 60 years and that's going to dictate where we can build out our facilities in the future. So what I'm suggesting is that we need a more holistic and robust approach to planning that takes into account the fact that we may need to change where and how we deploy technologies in the future. So we should try to build things out in a way that leave open as many options as possible.

A (26:18)

Kevin, that's terrific. Thanks so much for joining me on the show. I hope we'll have the pleasure of welcoming you back on a visit to CGD before too long. Of course, next time you come, we won't be in the current office. We're moving next year week to our new offices. And I look forward to welcoming you.

B (26:32)

Sounds great, Lawrence. Thank you.

A (26:35)

This has been the Global Prosperity Wonkast from the center for Global Development. My guest today is Kevin Ummel, visiting senior associate here at cgd. And we've been discussing a recently released paper, planning for large scale wind and solar power in South Africa. Kevin has been unpacking for us what spatiotemporal modeling means and why we should do it. You can find the Wonk Cast online and on itunes. Just search for Wonk Cast or CGD if you have itunes. I hope you'll subscribe to hear a new interview every week. Until next time, I'm Lawrence MacDonald. Thanks for listening.

B (27:44)

The people who want to do power planning don't have much incentive to do it differently because they don't have the tools. They may have the tools, but I don't think they have the political incentives. I mean, in South Africa, I've been struck by the fact that the classic case is the wind industry comes to Eskom and says, hey, we want to build these projects. We've got investors. But you Guys aren't giving us access to the transmission grid. And Eskom comes back and says, well, tell us where you're going to put your facilities over the next 20 years and we'll build a transmission system. An industry says we can't look that far in the future. Right. So I think you're right. The actors in the system themselves sort of have the incentives to do things very, very cautiously and no risk. So I think there's some market for some third party. I mean the World bank would be the obvious person to play this role, but they don't seem interested in doing it, which is to provide a third party vision of where it should go that both parties can sort of just sign onto with some modification. It could be. I mean, I actually have serious. I think South Africa itself may be too far down the road. In other words, they're moving and to the extent they're going to change, I don't know. I think an avenue that gets me more excited is how can we develop a larger southern Africa grid that effectively shares all the renewable resources, hydro, wind, solar, that are fabulous across the region, but shares them with a built out transmission system.

A (29:25)

The tie in here then, I think is this Power Africa Obama initiative. Right. Are they involved in that?

B (29:33)

I don't know if they are, but even that initiative, I'm not sure it's looking at this holistic approach. Yeah, I imagine they'll do things that are more conventional because they're easier. But I still think there's a real argument because the cost of this research is nothing. So there's an argument for doing it if all it does is get people thinking about, oh, you know what, there might be big benefits from thinking a little more, more holistic and integrated. Sam.