Recently, in discussions of solar technology, I've been hearing more and more talk of scalability.
Being a skeptical engineer, I chalked most of it up to marketing hype. It did get me thinking however, so recently I dug into it a bit to see if anything meaningful could be pulled out. My goal is not a policy plan or pitch for solar, but the following: 1) to see what metrics can be used in
engineering design and
technology development to quantify scalability, and 2) to imagine extremely large scale solar and see what innovations this might inspire.
Ambitious GoalsThe first questions that frame a discussion of scalability are “how big?” and “how fast?” Every policy wonk seems to have a different version of what is best.
Al Gore's goal is 100% carbon free electricity in 10 years. NREL’s
Ken Zweibel and Cal Tech’s
Nathan Lewis have both discussed various aspects of the “Terawatt Challenge” and have established a rough worldwide goal of 20 TW by 2050.
This isn’t a political blog, so I don’t want to argue about policy; this is a discussion of solar scalability. To bracket the problem, let’s say, in round numbers, that the target is to provide 1/3 of the current worldwide total energy consumption (electricity, transportation, and industrial/residential heat) using solar power by approximately 2025. As of 2008, the world was
using energy at a rate of 15 TW with the U.S. accounting for about 22% of that (
3.3 TW). To meet our goal, we need around 5 TW of total installed capacity by 2025. Of course, since solar only produces power during the day, it suffers from a pretty lousy capacity factor, so we’ll have to install 20-35 TW of peak capacity depending on location and module characteristics. To keep numbers round and to put a stake in the ground, I’m going to use a target of 25 TW, which is about 3500 times the current worldwide installed capacity.
Constraints on Ultimate Size?Before addressing scaling rate concerns, what about ultimate size? Solar energy has the
largest fundamental resource base of all types of energy, at 120,000 TW total and 600 TW practically available. In the following sections, we’ll touch upon a few more potential technological constraints related to total size.
Scaling RatesNow that we have a target and a timeline, we can establish the required “scaling rate.” Once this is in place, we can start to think about questions like “What constrains scaling rate?” and “What are the requirements that a technology must have to scale at this rate?”
Digging into this subject, one of the things that surprised me the most is how feasible this goal actually is. Everyone who has worked in the solar industry is familiar with solar’s historical growth curve – an
exponential increase since 1975 averaging somewhere around 30% per year, which is pretty astounding on its own. What I had never seen until I made the graph myself is how the rate of increase correlates with annual production capacity. Intuitively, one would expect that as the industry grew, it would become harder and harder to maintain exponential growth. This is certainly true in the limit, or the entire universe would eventually be converted into solar panels. However, looking at the data over the past 35 years (especially the past 10 years), we see the
opposite trend, i.e. as the industry has grown, the
rate of increase has increased as well! This trend is due to both the rapidly dropping cost of solar power and more widespread public acceptance.
Based on this, what growth can we expect over the next 15 years? More importantly, what growth levels do we need in order to achieve our target of 25 TW installed capacity by (or soon after) 2025. Bearing in mind the old Yogi Berra quote, “It's tough to make predictions, especially about the future,” I invented 3 scenarios for discussion. The “aggressive” scenario is simply a best fit to the growth of the past 10 years. While interesting, looking at physical and societal constraints, this scenario is neither possible nor desirable. The “flat” scenario is continued growth at a steady rate equal to the average of the past 10 years. The “front-loaded” growth is something I made up, based on the growth level we’re at now and the desire to have a sustainable industry in the future, while conforming to the bounds of reality.
Photovoltaic industry historical growth and projections
The aggressive scenario reaches 25 TW by 2019, flat just after 2026, and front-loaded by 2025. However, for the front-loaded case the size of the industry in 2023 is about a third that of the flat case. The front-loaded scenario gets us to the goal in the most sustainable and realistic way, as we will see when we discuss scaling constraints. Furthermore, the size of the industry (annual production) is much smaller in the front-loaded case (3.6 TW) than the aggressive (14 TW) or flat (9.6 TW) cases. This is essential for the sustainability of the industry, as we don’t want to just stop making solar panels when we reach our goal. In the front-loaded scenario, the industry can simply stop growing and produce at a level that eventually provides for our entire energy demand, along with replacing panels that have started to degrade after 30+ years.
Again, these aren’t necessarily “likely” scenarios. They are scenarios that both fit with what might physically be possible and achieve the stated scaling goals.
Grid ParityBefore we jump into thinking about different constraints, a note about grid parity (the point at which solar power matches the cost of the current electric grid). Ever since I started working in solar, the incessant drumbeat has been “grid parity.” Somehow, I expected more of a D-Day type celebration when I read that
First Solar had achieved this elusive goal (Not that grid parity is the sole possession of First Solar – I have seen convincing cases that string ribbon Si, high efficiency tracked Si, thin-film Si, CIS/CIGS, and CPV technology are there or will be there soon). So, congratulations everyone, we did it! However the reality is that grid parity is a sliding scale, depending on a huge number of factors, but that discussion is too big to get into in this post. In reality, we may have only achieved 20% grid parity, so the quest towards lower cents/kWh continues!
Now we return to the discussion of the path solar growth must follow.
Scaling Rate Constraint 1: System cost and bankabilityEven at grid parity, there is a finite amount of investment and money flowing through the economy, which in some sense corresponds to the total amount of effort we as a society can collectively direct towards everything. One measure of this is the worldwide GDP. The main requirement here, simply put, is that the amount of money required for the up-front cost of all solar power systems should not make up a significant portion of the worldwide GDP (
$69.7 trillion in 2008); if it does then this is an unreasonable amount of resources to expect.
If we look at the front-loaded growth scenario, $3-6/W system costs are sustainable for the next few years after which point costs will have to start dropping closer to $3/W (approximately current best practice for large thin film installations) However, by 2017 or so, system costs will need to start dropping closer to $1/W or growth will suffer. This cost reduction roadmap (shown below) falls pretty well in line with solar’s historical cost learning curve.
Projected cost learning curve for solar
I should also note that these $/Wp numbers are for one set of assumptions (18.4% capacity factor, 4.5% discount rate, and 25 year life). As has been shown, actual $/kWh electricity prices can vary quite a bit for systems of similar $/Wp ratings. The point here is to bracket the problem: are we talking about 0.1%, 1% or 10% of the GDP?
Based on the $/W roadmap I mention above, solar energy power systems will consume between 0.1%-5% of the worldwide GDP in the front-loaded growth scenario. Looking at the graph below, we can see why the aggressive and flat growth scenarios are undesirable – both the level and the rate of increase of total GDP seems effectively impossible near the tail ends of their growth curves.
Investment requirements for different solar growth projections
The question remains, is the front-loaded growth case reasonable? Technically, yes; therefore it is mainly a social priorities question – do we want to do it? The graph below shows how it compares with some other industries in 2008. I use U.S. GDP numbers, as these are easier to come by, but the picture is similar on a worldwide scale. On one hand the 2025 GDP requirement is about 2.5 times as large as the current utility contribution to GDP. On the other hand, it’s interesting to note that it is approximately the same percentage magnitude as the decline in U.S. manufacturing GDP over the past 10 years. My take-away is that given positive political and popular sentiment it is well within the bounds of reality, but it is likely still one of the most limiting scalability factors. Grid parity is not enough.
Comparison of projected solar industry size with other US industries
Furthermore, in terms of securing financing, the concept of “bankability,” (in short, the certainty of making money) was discussed quite a bit at Solar Power 2009, as it is becoming increasingly important as the size of systems increase. For solar, this comes down to reliability (both demonstrated and perceived), as well as company balance sheets (a metric of the likelihood that module warranties will be honored). All emerging technologies will have to rigorously address reliability in one way or another to prove to the banking industry that their technology is the Will Smith of Solar. For example, CPV systems must contend with thermal and tracking issues, which is leading to the development of reliability standards; CIGS tends to be more sensitive to moisture ingress and corrosion, leading to innovative module packaging; CdTe must rigorously address life cycle management due to cadmium toxicity concerns.
Scaling Rate Constraint 2: Energy Pay Back TimeAlthough there was much discussion of high energy pay back time (EPBT) in the 70s and 80s, it is mostly a non-issue today as
nearly all PV technologies have a EPBT less than 2 years (i.e. the power produced during the first 2 years of a panels 25+ year operating life will equal the amount of power required to produce it), with some technologies paying back in less than a year. However, when discussing scaling quickly to large portions of the energy supply, even 2 years can look like a long time. Consider if the industry doubled in size every year and the system EPBT was one year; in this case 100% of the energy of the panels produced would go into producing the power required for next year’s increase, thus no net power. I categorize this as power yield, or percent of produced power that is not consumed making and installing increasingly more power.
Based on the front-loaded growth curve, 1 year EPBT gives energy yields of 50-60% in the near term high growth phase, easing off to 70-90% in the longer term. I see this as a minimum metric for scalability. Ideal EPBT would be closer to 0.5 years, which keeps the energy yield in the 80-90% range even during the high growth years.
Scaling Rate Constraint 3: Capital expenditure per wattThis constraint is related to system cost, but also how fast a factory can be scaled up. For capital expenditures (CapEx) we can look at specific historical spending. Looking at
government census data, average CapEx per year in the U.S. for all sectors is $1.2 trillion/year, including $200 billion/year for the “manufacturing” sector and $100 billion/year for the “utilities” sector. With this in mind, CapEx costs <$1/W are almost certainly required, with CapEx costs less than $0.50/W much preferable.
There are a number of companies that have this argument built into their value proposition. For example,
Skyline Solar plans to use a combination of
already existing auto manufacturing capital and highly leveraged (via concentration) crystalline solar manufacturing capital. Another company which explicitly has this in mind is
Nanosolar with their largely non-vacuum printing-based deposition.
Ultimate Size Constraint 1: Material AvailabilityRecently, some folks at Berkeley did
a great study of resource bases and material availability for a whole host of potential photovoltaic materials. This study and others like it have shown that there are some device designs limited by material availability to less than 1 TW/yr (CdTe limited by Te, CIGS limited by In and to a lesser extent Se). However, there are many other device structures that show effectively unlimited material availability.
Ultimate Size Constraint 2: Energy StorageThe other potential limitation of scale is energy storage. Energy storage is a huge topic which deserves its own discussion, so I will only touch on some of the strategies being pursued which would enable 25 TW peak electricity to be supplied to the grid. While
some studies in Germany have found that up to 30% of peak electrical demand can be supplied via solar energy without significant modification of existing infrastructure, much deployment beyond that will depend on a number of energy storage technologies. Some envision the
widespread use of compressed air energy storage.
CSP with massive amounts of
thermal energy storage appears to be another potential solution. Yet another viable strategy is vehicle-to-grid, or
V2G. The idea is to use the distributed storage capacity that will come with increased deployment of
electric cars and
plug-in-hybrid vehicles, especially with newer
advanced battery technologies. Finally, the holy grail of energy storage technology is
direct solar fuel. The basic idea is: instead of converting solar energy into electrical energy, convert it into chemical energy in liquid or gaseous form, enabling temporal and spatial separation of generation and use.
Ultimate Size Constraint 3: LaborWhile less of an inherent constraint than the others listed, this is certainly something to consider. At the cycle time session at Solar Power 2009, Laks Sampath mentioned that “best practice” large scale ground mount installations took about 4 h/kW to install. If we look at the growth projections, this number will have to come down by an order of magnitude by the end of the next decade, otherwise somewhere around 7 million people (worldwide) will have to be employed simply installing panels. Although this number is not completely absurd (in the U.S.,
the entire construction industry employs around 11 million people), it seems likely and desirable that this number can be drastically reduced. This could be accomplished in ways similar to the innovations shown by Applied Materials with their
gen 8.5 module, SunPower with their
T20 tracker, or SunTech with their
Reliathon module. As a reference number, the
Apollo program at its peak employed 400,000 people.
Ultimate Size Constraint 4: LandMost studies on this issue have concluded that it is not a constraint. To make this easier to think about, we’ll only consider the United States’ portion of energy use. For the United States’ energy needs, depending on efficiency, array packing density, insolation, and other factors, this number is somewhere around 1-2% of the land area of the U.S.. It is similar in magnitude to the total paved area of all roads and parking lots in the U.S. It is also the same order of magnitude as the
total area reserved for U.S. military bases and bombing ranges.
So, on one hand, using this amount of land to provide for a significant portion of our energy needs seems like a relative non-issue. On the other hand, it is huge! On average, just for the U.S. portion of energy demand, we will need to install 10 square miles (~6,700 acres)
every single day for the next 15 years (assumes 10% AC module efficiency and 40% packing density). Thinking about installing PV over an area of that size in a day is admittedly quite daunting. Not insurmountable, but daunting.
To look at it on another scale, I spread the 5 TW of PV to be installed in the U.S. into 50 evenly sized installations across the nation (square installations 34 miles on a side). I put a few more in the sunnier states, but the distribution is mostly random, trying to avoid national forests/parks. Somehow, visually, this seems more doable, as I can imagine 50 separate teams working for 15 years rather than 1 team of a few hundred thousand working in the same place. Or, in the (admittedly silly) extreme case, over the next 15 years, each of the 304 million people in the U.S. would have to go outside and install one 220 W panel every other month. People between 16-65 would have to pick up the slack for those who couldn’t install solar panels, but it is still about the same magnitude of effort as a regular household chore. In one sense, the beauty of solar (at least PV) is that it works on anything from kW scale to GW scale.
Land area of solar needed to provide 5 TW peak power to the US
Parting thoughtsThere is no doubt that solar power at this scale will be mind-bogglingly huge compared to its current scale. However, it is entirely within the realm of physical and societal constraints. Cutting through the hype, I’ve found there are a few fundamental metrics of scalability to keep in mind when assessing technological innovations and strategic directions.
To quote a solar pioneer (MIT Professor
Ely Sachs), "The science is understood, the material abundant, the product works, all that is left is to build the biggest manufacturing industry in the history of human kind. Time is a-wasting."
