No really, space based solar power is not a useful idea, literature review edition

This blog is part of my series on countering common misconceptions in space journalism. It’s an extension of my post dismantling the case for space-based solar power.

As a writer in the public sphere, I am fortunate to have almost always received positive feedback for my writing. That said, I do listen to criticism and my post on space-based solar power was criticized for failing to address the substantial published literature on this topic.

One interlocutor recommended reviewing the Japanese literature on the subject, and provided some links to (English, fortunately) articles on aspects of the problem. While I am unable to read Japanese, I am extremely doubtful that writing the physics in another language will change the result.

In my previous post I consciously chose to ignore most of the hype and, instead, drilled into the core physics of the problem. On numerous occasions in my career I’ve been called upon to quickly find an answer to some poorly understood problem, such as magnetic levitation or numerical convergence of gravitational wave simulations. There is some skill to finding the biggest, most communicable issue and understanding how it changes everything else. In the case of space-based solar power, it just doesn’t matter what the rectenna is made from, or what orbit the power station is in, if there is another effect with a thousand times the impact on the bottom line.

As a result, none of the papers I’ve reviewed for this post have changed my analysis or my conclusions. I know incrementally more about high frequency power converters, but at the end of the day, power is cheap and rapidly getting cheaper on Earth, while it remains expensive in space. I want huge space stations as much as anyone, but I’m unwilling to lie myself into thinking that space power is the way to get them.

In this blog I’m going to review three papers on space based solar power. I had initially planned to do five, but the remaining two were not focused on economics and covered ground already dealt with in the first three.

The first is Space Solar Power Programs and Microwave Wireless Power Transmission Technology, by James O. McSpadden and John C. Mankins, appearing in IEEE Microwave Magazine in December 2002, and garnering 414 citations since then. It begins with a handy review of a 1970s-era study suggesting a R&D budget of $300b (in 2002 USD), before focusing on wireless power transfer (WPT). Although at this stage only small-scale prototypes of high power microwave transmitters had been built, the article’s working assumptions are for 70-85% transmitter efficiency and 90% receiver efficiency. Including beam control losses and antenna side lobes reduces net WPT efficiency to the order of 45%, which is consistent with my earlier assumptions. While essential for any space-based solar power system, the precise cost and efficiency of the wireless power transfer doesn’t make a huge difference to the bottom line, compared to launch costs.

The second article is A Fresh Look at Space Solar Power: New Architectures, Concepts and Technologies by John C Mankins, published at the 38th International Astronautical Federation in 1997, and garnering 159 citations since. Interestingly, the second sentence reads “During the first decades of the new century, global demand for electrical power is projected to grow dramatically – perhaps doubling from 12 terawatts to more than 24 terawatts. Achieving this power growth while managing environmental impacts effectively is a cruical [sic] international challenge.” In 2017, world electricity consumption averaged 2.7 TW, with a peak capacity of about 6.5 TW. Growth rates hover between 2% and 3% a year, meaning that consumption will hit 12 TW in 2080 and 24 TW in 2110. Additionally, more than 2/3 of current production growth is in renewables, whose installation costs are falling at up to 10%/year. In short, the hypermalthusian assumptions that underpin space-based solar power advocacy are not economically sound, nor are they reflected in the real world. Later, the source of these projections is revealed to be the US DoE’s Energy Information Agency (EIA), about whose predictions’ accuracy nothing further needs to be noted.

The article continues by summarizing and critiquing the 1970s DOE study, that called for thousands of person-years of space construction and, quite obviously, failed to deliver a competitive product. The most salient change mentioned in 1997 for what had changed since was a recognition by NASA that prohibitive costs to orbit were strangling commercial space opportunities. For all this recognition, the first commercial cargo transport contracts weren’t awarded until 2008, and the shuttle wasn’t retired until 2011.

According to this paper, “transportation system concepts that could deliver SSP elements to LEO for costs on the order of $100-$200 per pound … appear feasible.” As a reality check, SpaceX’s ridiculously cheap, industry-destroying, partly reusable Falcon Heavy can deliver cargo to LEO for as low as $1000 per pound. A factor of 10 here is not a small problem. It has the potential to elevate logistics costs from, say, 20% of the total cost to more than 70% – the single most dominant cost factor. Advocates for SBSP maintain that SpaceX’s Starship could deliver cargo at a cost as low as $100 per pound, and that’s great. But in the meantime, ground-based solar power has also gotten more than ten times cheaper.

In Figure 4, there is a concrete business model with numbers. It projects a public/private partnership to develop 24 satellites producing a total of 4GW over a lifetime of 40 years. It assumes $400/kg launch costs, a total installation cost of $40B ($10/W), and total revenue of $270B. The power cost (less investment) is projected to be 4c/kWh and the retail price 21c/kWh.

Screenshot from 2019-09-19 23:27:01

These costs are very optimistic (based in part on severely faulty demand assumptions), but I’m happy to grant them and instead compare them to the development of a modern solar plant, such as Pavagada Solar Park. Total rated output is 2GW, at a total cost of about $2b. Even taking into account capacity factor, this is 40% cheaper than the space solar station. Total development time is about three years, instead of 20. Retail electricity prices started at about 5c/kWh but have since fallen to 1.7c/kWh due to economies of scale. It is one of about ten GW-scale solar farms currently under construction. In Los Angeles, solar+storage is retailing at 3.5c/kWh – obviating concerns about capacity factor. To summarize, despite the significant optimism of the space based solar power business model, actually existing ground based solar power is 10x cheaper and 10x faster to install, 10x cheaper to purchase power from, and growing at about 30% a year. It requires almost no front-loading of development costs, it is getting much cheaper every year, and it also doesn’t require unbelievably cheap launch.

The third paper of note is SPS-ALPHA: The First Practical Solar Power Satellite via Arbitrarily Large Phased Array, also by John C. Mankins who by this time had set up his own consultancy: “Artemis Innovation Management Solutions LLC”. It was published as part of a NIAC study in 2012 and has 92 citations. What we lack here in author diversity we will make up in a time-series study of assumptions.

To begin with the executive summary, “it appears that a full-scale SPS-ALPHA, when incorporating selected advances in key component technologies should be capable of delivering power at a levelized cost of electricity (LCOE) of approximately 9c/kWh.” In other words, if we add enough advanced widgets, we can get the cost down to the standard US 2012 baseload electricity cost. In 2019, it turns out that technology improvements lowered the cost of ground-based solar+storage to 3.5c/kWh, so SPS-ALPHA will need yet more shiny widgets to compete in the future.

It can look suspicious when a high level cost estimate, employing numbers with uncertainties as great as 1000%, returns a final number that is exactly what it needs to be to compete with the going electricity rate in whatever year that report is published. And yet, in all three studies reviewed here, that is exactly what happens. All three reports are happy to predict wildly optimistic demand growth, and none even mention the steadily falling costs of competing supply.

The remainder of the report goes into vast detail regarding the conceptual design of self-assembling hexagonal space LEGOs, all of which can be regarded as cheapish mass-produced advanced satellites with solar panels and stuff on them. The actual details and costs of these parts doesn’t matter a whole lot unless their delivery cost to orbit drops below where even Starship might end up.

In section 4.2 we find that costing was done using a spreadsheet (NASA-grade, of course) derived from the 1997 Fresh Look study addressed above. We find that cost effectiveness depends on satellite systems cost falling below $500/kg, which occurs after the 260th unit is produced, assuming that the learning curve results in production costs halving as manufacturing quantities double, nine times in a row. In comparison, solar panel manufacturing costs drop about 5% per technology iteration, every 6 months or so. Solar panels are much simpler than satellites, which are more like cars. Henry Ford drove incredible reductions in car price, but at the end of the day there is a certain cost floor determined by the commodity cost of materials and specialty suppliers, which is why our cars are not disposable. So I think this assumption is rather dubious, but in the spirit of generosity I’ll grant it and make no further mentions!

There is a long discussion of various energy markets, as well as the identification of certain niche markets willing to pay 10-100x the going market rate for energy, such as military bases and small islands. Unfortunately, the net contribution to the market of such niche uses is so small it can’t support a SBSP industry, or even much of a terrestrial small/cheap power industry. There is also an extensive discussion of more speculative uses for SBSP, such as powering electrically propelled space probes and crewed Mars and Moon bases.

In Section 7 we get a detailed description of the optimal design cases identified by their unified systems design spreadsheet, including (in one case) a total platform hardware mass of 11,795,271kg, with no error bars. This solar power station is to deliver 500MW of power. Even accepting wildly generous launch cost estimates of $500/kg, the total launch cost of this satellite is about $6B, with manufacturing cost about the same. For the same money, any number of contractors could deploy 12GW of solar power. That’s 25x as much power, delivered in years rather than decades.

These estimates were performed in 2012, only 7 years ago. And yet, in the short time since, we have seen the numbers go from “don’t look too closely” to “not even on the same planet”.

At this point I’ve pulled out the relevant data from a series of papers by an authority on the topic, spanning nearly 20 years. Over and over again we see the same key weaknesses:
– Faulty projections of unmet electricity demand.
– Consistently optimistic estimates of construction and launch cost, at least an order of magnitude beyond current best industry standards.
– A disproportionately large focus on the relatively irrelevant cruft such as scifi applications and the minutiae of RF transmitter design.
– A failure to understand the forces driving innovation and cost improvement in the renewable electricity industry.

Space-based solar power is still not a thing.

Do you know of another article that you think changes the calculus here? Send it my way. If it adds to the conversation, I’ll happily add it to this post.

16 thoughts on “No really, space based solar power is not a useful idea, literature review edition

  1. Casey, you wrote “In 2017, world electricity consumption averaged 2.7 TW, with a peak capacity of about 6.5 TW.”

    What is your source for this. I thought electricity usage was much higher than that. Also why is it in TW’s instead of TWh’s; I am not sure I understand the difference.


  2. I wonder if anyone looked into the transmission cost of territorial solar vs space solar. As the recent California blackout shows, it doesn’t matter how efficient your powerplant is, if your transmission line is crap you still have a power problem. Could space solar have an advantage here? Can you put microwave receiver anywhere, in forest or on top of cities even?

    Liked by 1 person

  3. One article that should be considered for this article is your own “The SpaceX Starship is a very big deal,” which says that SpaceX could soon be launching large amounts of cargo to LEO for $35/kg. That would have to be adjusted to GEO cost, but that’d still be well under the “wildly generous launch cost estimates of $500/kg” that you mention here.

    It’s also worth mentioning that many “solar plus storage” installations don’t actually have enough storage to get through the night. A realistic comparison should look at the actual total cost to get 24/7 energy from ground-based solar.


    1. Some cost price distinction here but I don’t think it changes the outcome much. The EROI for Starlink is at least 50x better than SBSP, and probably infinitely better.


      1. But that doesn’t really matter since Starlink will not absorb much of SpaceX’s launch capacity with Starship. The competition is ground-based solar with sufficient storage to get through the night, and sufficient overcapacity to get through a few rainy weeks in winter. If SPS launched at $35/kg can do better than that, it’s competitive.

        It might not be competitive *right now* because instead of doing the hard stuff we’re backing solar with cheap natural gas at zero carbon price. But we’re going to have to stop doing that.


  4. While I agree that terrestrially-launched SBSP seems like a fool’s errand right now, I’m not quite ready to give up on it yet, for two disparate reasons:
    1) All of the studies you’ve cited make the assumption that all materials are launched from the surface of the Earth to the target orbit (usually GEO). In that case, it’s fairly easy to compute the LCOE just from the launch costs, and rapidly despair.
    But you can get an awful lot of the mass for an SPS from materials available on the Moon or near-earth asteroids.
    This is pretty much science fiction right now, but it won’t be in the 15-20 year timeframe. If Starship launches for $10M/launch, then specific cost to the lunar surface comes in below $1000/kg. Assuming that somebody’s willing to put in the R&D investment to build largely automated lunar mining and manufacturing, then the bulk of the mass for an SPS comes from the Moon, using a mass driver, and the total system cost stops being a function of specific cost to orbit and becomes one that looks a lot more like a traditional power plant.
    2) Assuming that we need to de-carbonize 100% over the next 30 years or so, there’s some limit to what percentage of available nameplate capacity can be renewables, simply because bad weather sometimes happens. I’m prepared to believe that that percentage might be as high as 75%, but beyond that, there seems to be 4 broad classes of solutions:
    a) Massively overbuild renewable capacity. If your renewable nameplate capacity is 2x-4x your peak demand, then you’re immune to all but the 4- and 5-sigma weather events, which gets you about the same grid reliability as we have today. But the business case for this is terrible, because the average cost of electricity is so low that they can’t amortize their investments.
    b) Build out a lot of backing storage. If you’re at 100% renewables, you need something like 10-15 Wh of storage per W of renewable nameplate. At current prices, that comes out to an LCOE that’s well over $100/MWh, and possibly over $200.
    c) Build massive numbers of gas peakers. Their emission footprint is tiny, because they hardly ever run, but because they hardly ever run, they’re going to charge thousands of dollars per MWh for their energy. When you average that across the energy cost of the renewables, you’re going to wind up with an average electricity cost that’s pretty expensive.
    d) Make about 30% of the production portfolio come from baseload sources. As a practical matter, the only decarbonized baseload sources today are geothermal and nuclear. (And biomass, I guess, but that’s basically a scam…) We don’t know how to do geothermal at scale, and everybody hates nuclear.
    But SBSP would be baseload power. In GEO, you’re in sunlight well over 90% of the time. So if you can get SBSP to just a bit more than nuclear in LCOE (i.e., about $100/MWh), then it’s a winner.
    Which brings us back to lunar or asteroidal resources. I completely agree that without them, SBSP isn’t gonna cut it. But if the technology exists to make a substantial number of the components enumerated in the 2012 Mankin paper on the Moon, and toss them into GEO from the lunar surface, then SBSP stands a chance.
    That’s a lot of “ifs”, and I remain skeptical that the investment case will close. But it’s not hopeless.
    One other thing: Once you’ve made the investment, SBSP scales really, really well. You may start out with a biz case that assumes only a few percent of global capacity, but the marginal cost of producing more and more SPSes is very low. You could easily wind up with the majority of global capacity being delivered from space within a century.

    Liked by 3 people

      1. Isn’t a lunar space elevator feasible with the bulk of the load-bearing material being basalt fiber? The cost of stuff in space depends less on what it is and more on where it is than we’re used to, and the moon seems like a much better location to source our oxidizer, silicon, and aluminum from than the ground, if mass drivers or a lunar space elevator, or both, turn out to be toward the optimistic end of the plausible range.

        Liked by 1 person

  5. If I were D. D. Harriman, it would be for deep space, and for handling asteroid material in cis-lunar space. I think we should be not bi-planetary but multi-world, artificial world’s with rotational pseudo gravity and lower delta-v than Mars. But since Elon Musk is D. D. Harriman and I’m not, it probably would be primarily for Mars.

    People can only go to Mars during the normal launch windows, but if I understand the conclusions correctly (I definitely don’t understand the actual orbital mechanics), low-cost bulk supplies can be sent slowly on low delta-v trajectories at a much wider selection of departure times. Being able to rendezvous with a cheaper tank of oxygen at Mars should cut the overall cost of Mars.

    It might also be for LEO, for de-orbiting Starships and filling them for Mars, GEO, or deep space. I think a lunar space elevator can be long enough to let you just drop payloads into a transfer orbit to LEO, but I don’t know how much delta-v it is from the transfer orbit to LEO. Cheap oxygen at elevator-LEO transfer might not be a bargain if you need too much kerosene/H2/whatever to go with it, brought into an orbit there’s no other reason to have it in.

    De-orbiting Starships after Starlink launches is what could be big, since it would be serving an actual high-volume commercial market.


  6. I was assuming that the reducing component of rocket fuel has to be hydrogen and/or carbon, lifted from the ground. But in solid boosters it’s aluminum. If we can make aluminum/oxygen rockets work with no C or H, they can be made entirely of stuff that’s abundant on ordinary rock as found on the moon. It’s a really bad way to make a rocket if you’re in a place that’s full of air and water. But if the boosters for leaving near-Earth space could be made without lifting anything from Earth at all, and so can the ones for returning to the ground from LEO, it would bring asteroid platinum that much closer to being economically viable.

    Rocket propellant normally has to be hot gas. (Or it can be stuff thrown from a mass driver, with pathetically low specific impulse and worse maximum thrust.) By including excess oxygen, we might be able to have the exhaust consist of a mixture of hot O2 gas and alumina dust. An alumina rocket would have crummy specific impulse compared to what you have if you don’t deliberately get the stoichiometry wrong, but if it works at all it probably would do better than lifting from the ground.


  7. A problem with microwave transmission is the large area required on Earth for the receiving antennas. And a problem with optical transmission is that it’s largely blocked by clouds.
    Could terahertz transmission be used instead. It’s rather midway between microwave and optical.
    Another possibility: metamaterials. Some recent research suggests you can go beyond the diffraction limit in optics using metamaterials for lenses or mirrors. This means a lens or mirror can be smaller than the classical diffraction limit suggests and get the same resolution.
    The same is true in the microwave range and should also be true in the terahertz range.
    Bob Clark

    Liked by 2 people

  8. A superheavy will loft 100 tons to orbit.

    Let us say one such launch may loft 1000x 100kg 20m^2 panels each with attitude control and a microwave transmitter. Let each cost $20k, in volume, so $20m, + $5m launch cost. (BFR might launch cost will drop to $2m, eventually.)

    20 launches, 20k satellites, $500M. Not peanuts, but in easy range of commercial capital.

    The satellites orbit at maybe 16k km radius, distributed 10k km wide along a 100k km circumference. Precise orbits are not needed, because the aircraft will broadcast a synchronizing signal to all the satellites in line of sight for them to phase-lock their transmitters to.

    An Airbus A330-200 normally takes off with ~160,000 lbs of fuel: 80t, 23k gallons, say $60k worth of Jet-A. (Rated max is 111 tons fuel, 50 tons cargo.) Rated engines generate 72 tons thrust, which seems to take about 85 MW to produce. We will assume this is more than needed, for reasons to be seen below, so we plan to supply only 60 MW, sustained.

    A 20 m^2 perovskite or GaAs panel might extract 20 x 1.36kW x 30%, or 8 kW, and radiate 7 kW of precisely targeted phased-array microwaves constructively interfering with thousands of others, to deliver 6 kW each to an aircraft-mounted rectenna. So, 10k satellites in line of sight can power a fully loaded A330-200 through takeoff, climb and cruise. After it reaches cruise altitude, it needs much less.

    We can do 2 flights a night. This displaces fuel cost of 2 flights x 350 nights x $60k = $42m / year.

    But it doesn’t stop there. Fueled flights are limited to ~7 hrs because they need to land and refuel. Longer hauls are worth more, and incur less airport costs. We can fly as long as we like; usefully, up to ~13k miles. But leave that aside.

    Also! Speed is normally traded off against fuel cost. But with zero marginal fuel cost, we can offer shorter transit time, which is, again, worth more. But leave that aside, too.

    But look! Our cargo capacity has been limited to 50 tons because we need to loft 80 tons of fuel with it. Absent fuel, we can take (say) another 50 tons of paying cargo; and carry 10t of batteries, for safety. Each flight generates *twice* the gross income. How much is that?

    Each fueled 50t flight had to generate net $60k just to pay for the fuel. Guessing a 20% gross margin, we collect an extra $50m in gross income, for the extra freight, on top of the $42m fuel savings, for $92m/year.

    But there is more! We can fly the same aircraft in the opposite hemisphere, when it is night there, and generate a *second* $92m/y.

    And, we can operate passenger flights in the daytime. Maybe those generate less? Say $60m/y each. So, we get $304m ARR against a one-time $500m investment (plus whatever it costs to retrofit 2 or more aircraft).

    I need to justify saying we only need to supply 60 MW instead of the rated 85 MW. Takeoff load is 100t cargo + 10t batteries = 110t, instead of the current 50t cargo + 80t fuel = 130t. 85 x 110/130 is 72 MW. We are carrying 10t of batteries for safety, and they can provide a boost for takeoff. So, we can draw 12 MW from the batteries for a few minutes on takeoff, and top them back up at altitude.

    Alternatively, the satellites can radiate more than the rated 8 kW, for a few minutes, using *their* batteries. Or, some of each.

    A 20 m^2 solar panel satellite for $20k might be over-optimistic, but the reliability demand on these is much, much lower than usual for a satellite. A $40k satellite cost would be tolerable.

    Bigger satellites would power more than the two simultaneous aircraft cited, which will come seem like a good idea. Maybe instead of 20,000x 20 m^2 satellites per launch, 2000x 200 m^2 sats would be better, or even 200 2000 m^2 sats. After a time we probably want enough to keep 10,000 aircraft aloft, not just the 2, which is 100,000 launches.

    At some point, we might start collecting carbon taxes for fueled aircraft. Carbon injected into the stratosphere is much worse than terrestrial release.

    Probably, after the 1st 1000 launches various things will be different. Hydrogen-fueled aircraft might become competitive. Maybe people will ride rockets more and planes less. Maybe an orbiting counter-rotating wheel that dips into the stratosphere will make rockets mostly unnecessary. Maybe civilization will collapse.

    (And who knows, maybe the horse will learn to sing.)

    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s