In Defense of Space Solar Power

Al Globus, January 2009

Space solar power (SSP) refers to huge satellites (PowerSats) in Earth orbit that gather sunlight, which is converted to electromagnetic waves beamed to Earth, where it is converted to electricity. This could supply very large quantities of environmentally friendly baseline electrical energy to Earth for the next few billion years. We know we can do it, most satellites are powered by solar energy today and wireless power transfer has been demonstrated with very high efficiency, but we don't know if SSP can be economically feasible anytime soon. For example, in a recent paper Space Solar Power: An Idea Whose Time Will Never Come? Steve Fetter claims that SSP cannot possibly out-perform ground solar economically because working in space is so expensive; and for that reason SSP should receive no federal research and development (R&D) funds. At the most basic level, Fetter is misled by high space system costs driven by one-off hand-crafted systems, today's norm. Electrical energy demand when the sun isn't shining is extremely large. To satisfy this market would require hundreds or even thousands of PowerSats, each consisting of very large numbers of identical components and launched by nearly identical vehicles enabling economies of scale -- doing the same thing over and over -- which in other industries have brought prices down dramatically. Contrary to Fetter, this paper presents evidence and arguments, based on economies of scale, suggesting that a reasonable R&D program has a decent chance of bringing SSP production within reach of the private sector. If successful, such a program would have enormous positive impact on energy independence, wealth generation, greenhouse gas reduction, space development and the international balance of power.

In his paper, Fetter constructs a simple mathematical model of SSP economics and, on the basis of this model, claims that SSP must meet six conditions to compete with ground solar:

  1. "solar supplies ~100% of total electricity demand;
  2. the cost of space-based solar arrays is reduced to $1,000/kW and earth-based arrays do not cost less than space-based arrays;
  3. SSP transmission is no less efficient and no more expensive than storage or intercontinental transmission of electricity generated by earth-based systems;
  4. SSP operation and maintenance is no more expensive than operations and maintenance of earth-based systems;
  5. total on-orbit system mass is less than 5 kg/kw; and
  6. launch cost (currently about $10,000/kg to low-earth orbit) is less than $350/kg."
First, Fetter's launch cost numbers are a bit off. One can order a Falcon 9 launch of 12,500 kg for $36.75 million ($2,940/kg) today and Russian vehicles have reportedly launched for as little as a few thousand dollars per kg. Second, Fetter makes no attempt to validate his model. It is very easy to construct models that seem reasonable but do not reflect important elements of reality. Thus, it is important to check models against existing data to see if they are reasonable. Fetter doesn't do this, although, to be fair, validating this kind of model is very difficult. Given the lack of validation, however, it is important to look for clues that the model doesn't fit reality. We will examine one such clue.

This paper will focus on meeting five of the six conditions, and explaining why condition 1 makes no sense, a sign of possible problems with the model. There is evidence that conditions 3, 5 and 6 can probably be met given a robust federal R&D program. Conditions 2 and 4 depend on the design of PowerSats and we present arguments that they might be met as well.

When considering these arguments remember that the case for SSP need not be perfect. To rebut Fetter's claim that economic SSP is all but impossible, an R&D program to develop SSP need only have a decent chance of success. It seems reasonable to suggest that an SSP R&D effort on the order of the thus far unsuccessful fusion energy effort is warranted. We have spent over $20 billion on fusion energy research in the last 50 years, including $300-900 million per year for the last 30 years. Depending on one's opinion, this may or may not have been a good investment. However, it indicates how much effort developing a major new energy source is worth -- what the customer is willing to pay -- even with substantial risk of failure. SSP, if successful, is a major, positive game changer for energy, global warming, space development and the global balance of power and, unlike fusion, requires no breakthroughs in physics and the space development benefits would be incalculable. For comparison, NASA's budget is roughly $17 billion annually. Great benefits warrant great effort, so a 50 year $1-2 billion/year SSP R&D program seems appropriate. About $30 billion is for launcher development and the rest for energy transmission research, system design, component development, in-space transportation and assembly and maintenance research. There is no claim that this program is optimal in any sense, only that it may be sufficient to meet Fetter's conditions. In particular, it may be longer and larger than really necessary.

We now turn our attention to Fetter's six conditions.

  1. solar supplies ~100% of total electricity demand. This condition makes no sense. Does it mean U.S. demand? Global demand? The demand of Lichtenstein? If it means U.S. demand, what if demand doubles? Halves? Obviously these things matter since a successful SSP must sell enough electricity to cover development costs. The author emailed Fetter for clarification, and he replied, "The calculations in the paper do not refer to any total electricity demand," which makes one wonder what condition 1 does mean. In another email on the same subject Fetter says, "If solar supplies substantially more than 10-20% of total demand (~100%), then most ground-based PV electricity would have to be stored or transmitted very long distances, incurring costs and losses similar to those that are inevitable with all SSP." Leaving aside the fact that, say, 50% is a lot more than 20% and much less that 100%, this condition appears to be a proxy for ground solar storage and transmission. However, obviously, to supply electricity when the sun isn't shining ground solar requires storage and/or transmission regardless of market share. Furthermore, economic SSP development requires sufficiently low operating costs and a very large market to amortize those costs, not that solar capture all of the market. This condition appears to be an over-interpretation of Fetter's model. All models are approximations of reality tuned to answer certain questions. If the wrong sort of question is asked, you get nonsense.

    For the purpose of this paper we will grant ground solar market dominance when and where the sun is shining brightly, thus we compete only at night and when little sunlight reaches the ground. For this part of the market, ground solar must pay the storage and transmission costs. In practice, of course, SSP will produce power when the sun is shining and it will make more sense to sell the power at a loss than throw it away. In any case, SSP economics depend on economies of scale so the market must be large to justify SSP development. However, total world energy use is about 15tw per year [Wikipedia] of which about 2tw is electricity, a substantial fraction of which is needed when the sun isn't shining brightly. Furthermore, demand is rising as billions of people currently have little electricity and wealthier societies are expected to move to plug-in hybrid cars which will tend to charge at night. The size of the potential market does not seem to be a problem.

    If the market is large, can SSP meet the demand? As the total solar energy available in space is vast, SSP can be scaled to deliver whatever level desired simply by building more, larger PowerSats and receiving antennas. To take a somewhat over-the-top example, at 10% end to end efficiency SSP could supply the entire 15tw global demand with 115 billion square meters of collecting area, or roughly 370 PowerSats each with a radius of 10 km. Obviously a big job, but if one can be built there's plenty of room for more and economies of scale are relevant.

    The fact that condition 1 makes no sense is a red flag that the model has problems. Perhaps the model needs a more sophisticated means to distinguish between demand when the sun is shining and when it is not. Perhaps there needs to be explicit terms to represent competition from other forms of energy. Even though condition 1 makes no sense, that does not mean the other conditions are not important, and we will now examine them.

  2. space-based solar arrays < $1000/kw. This number is provided without references. A casual google search indicates that nuclear power plants cost up to several thousand dollars per kw ($6,267/kw for a 2008 Florida plant [Barnhard, personal communication]) and coal plants are only a few hundred per kw. On the other hand, both of these, unlike solar, require fuel, coal has a large environmental cost, and nuclear increases risk of mass casualty terrorist attacks. While the nuclear costs suggest that reaching $1000/kw may not be necessary, it would certainly be worthwhile.

    Current space solar costs are reputed to be $750,000/kw, a factor of 750 too high. Is there reason to believe that economies of scale can cost reductions on this order? Consider the computer keyboards used by air traffic controllers. These are custom, not used anywhere else, and are produced in low volume. They cost approximately $2,500. In 2009 one could buy a standard, mass-produced computer keyboard at Office Depot for as little as $13. The price difference is roughly a factor of 192, not too far off of 750. If there are major optimizations in approach, perhaps we could get the factor of 750. There may be. On solar energy systems for existing satellites the entire collecting area is expensive semi-conductor cells (and market size is very small). One could use large, inexpensive mirrors and relatively small higher cost silicon solar cell areas. This is, in fact, how the lowest-cost terrestrial solar systems work, they use inexpensive mirrors and smaller, relatively expensive sunlight-to-electricity conversion systems. Also, current demand for new space power is perhaps a few hundred kw a year. As we have seen, potential demand for SSP power is multiple orders of magnitude greater enabling economies of scale.

    The cheapest mirrors might be built in space. Such mirrors need be little more than a thin reflecting film, perhaps only a few atoms thick. Mirrors built in space wouldn't have to be strong enough to survive launch and would not need to be folded or rolled to fit in a launch vehicle fairing. One only need launch the feedstock for a thin-film mirror production machine that takes advantage of solar heating, orbital vacuum and zero-g. While expensive, developing such a machine may be worthwhile since an enormous mirror area is needed. This is an example of economies of scale. To deliver current space solar power to existing satellites is far too small a market to justify the cost of developing a space mirror production machine. However, SSP will require acres and acres of mirrors, so the development costs can be spread out over a large number of systems. An in-space mirror production facility may be a key portion of the proposed R&D program and a potentially valuable task for the International Space Station (ISS). At the moment, there is no way to know the ultimate cost of such mirrors.

    As there is some reason to believe that large optimizations are possible, this condition may be within reach of the proposed R&D program.

  3. SSP transmission is as good as ground storage and intercontinental transmission. SSP transmission would be electromagnetic at some yet-to-be-determined frequency. Most designs call for PowerSats in geosynchronous orbit (GEO), although some have proposed sun-synchronous orbits (SSO) with relay satellites in low earth orbit (LEO). Ground solar can transmit using wires, but not between continents, so ground solar requires beamed power to meet this condition as well. The simplest architecture for ground solar is to beam power between continents using GEO repeaters, but this is twice as difficult as GEO SSP since power must be beamed up and then down for ground solar, but only down for SSP. In the relay satellite case, to first order both systems require roughly the same repeater satellite constellation (or network of high-altitude long-duration aircraft). Thus, unless an inexpensive means to transmit electrical power between continents by wire is developed, meeting this condition does not appear to be a major problem.

  4. operation and maintenance no worse than earth-based systems. This condition may well be the most difficult to meet and is by far the most difficult to demonstrate without actually building and operating SSP. Current ground-based costs can be ascertained but ground-based costs can be driven down by R&D and experience gained. Space systems are at a much earlier point in the learning curve and thus have more potential for improvement. In the end it will be a contest between teleoperated robots working in the mostly predictable space environment vs. dealing with wind, dust, corrosion, rain, vandalism and theft on the ground. Space operations costs are very high today, but, again, this is mostly operating one-off custom spacecraft not doing the same things over and over and over as would be the case for SSP. Telerobotics is in its infancy, so R&D effort should provide significant improvements.

    There is some reason to believe that this condition might be within reach of the proposed R&D program.

  5. total on-orbit system mass is less than 5 kg/kw. Fetter claims current cost is about 10 kg/kw, so only a factor of two improvement is needed. Economies of scale can easily produce this kind of benefit. Furthermore, most of the collecting area for SSP can be mirrors, which are very light weight. Thus, the key is to reduce the mass of the relatively small power conversion and transmission components. In the as-yet-unpublished paper An Isoinertial Solar Dynamic Sunsat Phil Chapman argues that space solar dynamic power generation technology developed at NASA Glenn Research Center can form the basis of a 5.2 kg/kw system. This technology is relatively mature and was seriously considered for use on the ISS. As the Glenn system is of value for other projects, development costs could be shared. There is a big disadvantage for this approach, solar dynamic has moving parts and high temperature fluids that will inevitably drive up maintenance costs to the detriment of condition 4. It would be desirable to develop solar cell systems to replace the generators.

    Update: In "Towards and Early Profitable PowerSat," presented at the Space Studies Institute Space Manufacturing 14 conference in 2010 I note that the Japanese Ikaros satellite, a solar sail, is producing power at about 0.8 kg/kw for the power production system. This extremely low value is accomplished by using a heliogyro thin film solar sail design, which has no structure, and thin film solar cells on part of the surface. Given reasonable improvements in thin film solar cells this figure could be 0.16 kg/kw for power production, not including other parts of the satellite.

    This condition appears to be well within our reach.

  6. launch cost < $350/kg. Although purchasing one 12,500 kg per launch Falcon 9 costs $36.75 million, SpaceX representative Lauren Dreyer reports that packages of 1,000 launches can be purchased for $10 million apiece, which works out to $800/kg. At 5 kg/kw, 1,000 launches could provide roughly 2.5 gw -- a tiny fraction of current global energy demand. This cost is only a little over a factor of two more than the condition requires.

    A less developed, but perhaps better, option is the Sea Dragon design from the 1960s. This 150m tall, 23m diameter ocean launch and recovery design uses simple pressure fed engines: LOX-RP1 for the first stage and LOX-H for the second. LOX and H is manufactured on-site by a nuclear aircraft carrier until SSP is available. The design calls for 8mm steel tankage so the Sea Dragon could be built in shipyards, not expensive aerospace facilities. The payload target was 500 tons -- enough for a 100 mw PowerSat from a single launch at 5 kg/kw. Development costs were expected to be $27 billion in 2007 dollars. Estimated launch costs, including amortized development costs, for 240 flights in 10 years was $242/kg in 2007 dollars according to design documents available at This would meet the condition.

    Clearly, the target launch cost has not been met to date. Equally clearly, both of these vehicles, and quite likely others, are in the ball park. This condition appears to be well within reach for the proposed R&D program.

Fetter further argues that the 'probability' of meeting all these conditions is so small that SSP cannot be economical. However, we are not throwing dice nor are we passive observers. Whether SSP becomes economically viable may well depend on what we do -- and that is under our control.

We see that there is evidence to suggest that three of Fetter's six conditions (3,5,6) may be met by a reasonably sized R&D program and one condition (1) makes no sense. Weaker arguments suggest that the other two conditions (2,4) could be met. The core reality behind Fetter's assertions is that space systems and operations are far more expensive than those on Earth; but current space system prices are based on one-off hand-crafted and therefore necessarily expensive systems. Electrical energy demand is so large that to satisfy even a fraction would require hundreds of PowerSats and allows economies of scale. This unquestionable fact gives hope, although nothing resembling certainty, that a vigorous R&D program can develop SSP technology to the point that profitable SSP businesses can be established.

SSP's greatest weakness is that profitability depends on supplying very large quantities of power unless a small, niche market willing to pay high prices can be found. One such market might be space-to-space power, i.e., PowerSats supplying power not to the ground but to other spacecraft or perhaps the lunar surface. A research effort in this direction may be worthwhile regardless of the fate of space-to-ground power.

The bottom line: in contradiction to Fetter, there is reason to believe that the U.S. federal government should institute a vigorous space solar power research program without delay. While there is a definite risk of failure, potential benefits are major: vast quantities of energy, substantial reductions in green-house gas emissions, and great wealth and power for those who succeed.