To round out the controversial critique of solar power started here, I reproduce below the most detailed critical analysis I’ve read on solar thermal electricity. It’s written by a University of NSW academic, Ted Trainer. I’d strongly encourage you to read his full 44-page set of arguments on the inability of renewable energy to sustain an energy intensive society (an earlier version of which was reviewed last year on BNC — this provides an updated summary of his 2007 peer-reviewed book on this topic, published by Springer).
For context Ted and I agree on many things, but not all. We both acknowledge the seriousness of the climate crisis and the magnitude of the sustainability problems caused by humanity’s overexploitation of natural environments. We differ in that Trainer sees only one viable solution — a rapid, planet-wide ‘power down’ of civilisation to some ’simpler way’ (read here for his well-mapped-out thesis).
In the past year Tom Blees and I have exchanged a number of emails with Ted, in which we’ve talked about Generation IV nuclear power (IFR in particular), proposing it as a potential ‘uranium-thorium bullet’ to solve the energy crisis — and, as a result, providing the clean energy to solve a whole host of global problems such as climate change, water supply, agricultural sustainability and repairing the damage we’ve inflicted on natural systems. To his credit, Ted has looked at my arguments seriously, and has come up with a range of questions on Gen IV nuclear on which he requires ‘clear and convincing information’. As such, in a future post on BNC, I intend to address his nuclear critiques. But that’ s for another day.
The examination of solar thermal electricity given below is quite detailed, yet necessarily incomplete. Reliable data are simply lacking on many critical points. At the foot of this post, I list some key knowledge gaps on which Ted seeks further data. Perhaps you can help. For now, read on!
SOLAR THERMAL ELECTRICITY (for the reference list, see here)
The major drawback for renewable energy is the inability to store electricity from intermittent sources. Solar thermal technologies are especially valuable because they can store heat and use it to generate electricity when it is needed. Some believe this capacity will be the key to enable renewable energy sources to meet all electricity needs (e.g., Trieb, undated, Czisch, 2004), for example plugging gaps left by PV and wind.
Solar thermal systems are best suited to the hottest regions and it is not clear how far into the mid latitudes they can be effective, apart from via very long transmission lines. They seem to be especially doubtful in winter, even in the best locations. (For a more detailed discussion of solar thermal’s limits and potential see Trainer, 2008.). Trough systems will be considered first, then dishes.
The winter electrical output for the US SEGS VI trough system is reported at about 20% of summer output. (NREL, personal communication.) Modelling for Central Australia, possibly the best solar thermal location in the world, by Odeh, Behnia and Morrison (2003) produces a figure closer to 12%.
The SEGS VI plant with its north-south troughs was not designed to maximise winter performance. Arranging the troughs on an east-west axis, as distinct from the usual north-south axis, would raise the winter/summer ratio for energy entering a trough. (“Polar axis” alignment of troughs enables maximum energy yield, but is not feasible for large scale power generation.) However even in good solar thermal regions the performance of east-west troughs in winter (and summer) is relatively low, compared with the summer and annual average performance of north-south troughs. This is evident in Figure 1 from Odeh, Behmia and Morrison. Summer thermal energy (not electrical output) entering a NS trough at Alice Springs would be 780 MJ/m/month, (in this document “m” represents square metre) whereas in winter from and EW trough it would be 430 MJ/m/month, or 4 kWh/m/day.
The radiation data given by RREDC (undated), Meteonorm and ASRDHB, 2005, point to the same general conclusion. These sources indicate that Alice Springs is a better location than Egypt, receiving possibly 50% more solar energy per metre in winter. It also seems to be a little better than the SW US. Thus if solar thermal technologies are problematic in winter at Alice Springs they are not likely to be viable in the US or for Europe.
A critical problem for solar thermal systems is what proportion of collected heat is above the threshold level required for generation of sufficient steam pressure. In regions where radiation is low to moderate, considerable heat energy could be collected without enabling generation of a significant amount of electricity. For SEGS VI radiation appears to have to reach 700 W/m (DNI or direct normal irradiation, not global radiation) before generation becomes moderate, and at 500 W/m it is only about 33% of maximum. (NREL, undated, Jones, et al., 2003, Figs. 5 and 14.) Thus if east-west trough collects 4 kWh/m/d, as Fig. 1 from Odeh, Behnia and Morrison indicates, not all of it will be at a sufficient intensity to generate electricity.
ASRDHB data show that for Alice Springs in winter the intensity of DNI per square metre entering an east-west trough averages only 408 W/m, over a 12 hour period. It is over 700 W/m for about 7 hours. Fig 3 from Odeh, Behnia and Morrison shows that at Alice Springs 26% of DNI received over a year is under 500 W/m and 18% under 350 W/m. These figures suggest that the 4 kWh/m/d entering the trough is a misleading indication of electricity likely to be generated and that the actual output could be less than half this amount.
More direct evidence comes from the SEGS VI record. Hayden (2004, p. 190 .) reports that the 2.3 million square metres of collectors average 77 MW over a year, which corresponds to a continuous flow of 33 W/m. The above evidence is that winter performance is about 30% of the average performance, which is a solar to electricity efficiency of 10.7%. This suggests that the winter figure would be a 24 hour average c.13 W/m.
From this very low gross output a number of factors must be deducted, the main two being the energy required to build and run the plant. The latter energy losses, mostly for pumping fluid through the absorber, are given by Sargent and Lundy (2003, Section 4 – 3) at 17% p.a., although they estimate that in future the figure will be under 10%.
The embodied energy cost, i.e., the amount of energy needed to build the power plant, is reported by Dey and Lenzen at c. 4% of gross output for a plant of normal size in normal conditions. However a plant capable of delivering 1000 MW in winter would have to be 2.5 times as large as one capable of this output as an annual average, so its embodied energy cost would be that much higher.
(It would then generate much more than 1000 MW in summer and the ratio of embodied cost to total output would remain c. 4%, but a problem would then be that summer output would be far in excess of demand. On problems in storing such a surplus as hydrogen see below.)
The embodied energy cost analysis of solar thermal systems must also take into account the energy cost of building and maintaining the long distance transmission lines, e.g., from North Africa to North West Europe and of transforming from DC to AC power. The lines might add one-third to power plant dollar cost. (Czisch, 2004.)
The loss of energy from solar thermal storage is low but has been estimated by Sargent and Lundy as .9%.
Finally, the loss of energy in the very long distance transmission has to be taken into account, e.g., from Egypt to NW Europe. This could be 15% of gross output.
Some of these numbers are uncertain but when combined they indicate that the total energy loss might be 35% of the meagre gross output, meaning that a net delivered amount well under 10 W/m might reach users in winter. If so plant capable of delivering 1000 MW in winter would need 100+ million square metres of collection area. At the estimated SEGS cost of $800/m (i.e., per square metre, not per kW)) the plant would cost $80 billion. (Trainer 2008
More confident data on trough performance in winter would be desirable here, but troughs would not seem to be viable. (Heydon’s account comes to a similar conclusion.) Note that the vision of abundant winter supply to Europe via troughs would involve harvesting radiation at about 55 degrees from the vertical.
Sargent and Lundy (2003) put the capital cost of solar thermal plant at $(US)4,589/kW ($(A6,556) for the “near term future” (including heat storage, which reduces required generator capacity and cost, by enabling the generation rate to be levelled out.) NREL say the 2003 equivalent price of the SEGS plant is $(US)7,700. These figures are to be compared with $(A)3,700 million for coal plant plus fuel (early 2000s price) over plant lifetime. These figures are for peak outputs and the average output from a coal plant is c. .8 of peak whereas for a solar thermal plant it is around .25 of peak capacity (in the best locations). Thus capital cost per gross kW delivered on average (as distinct from peak) from solar thermal plant would be over 7.5 times as great as for coal including fuel. (See Trainer, 2007, Chapter 3.) Transmission lines from the Sahara to Europe under the Mediterranean Sea would probably add more than 33% of generating plant capital costs. (Czisch, 2001, 2004) indicating a multiple of 10. Note that these figures are not for plant large enough to deliver well in winter and for SEGS VI this factor might multiply by a further 2.5. Note also that dish costs are at presently much higher than trough costs.
Again future materials, energy and construction costs are likely to be far higher than at present so these figures are not very meaningful guides to future viability.
A solar thermal plant near Sydney, NSW, some 34 degrees south, has been constructed to pre-heat water for a coal-fired power station. ( Mills, Le Lievre and Morrison, 2000.) This is sometimes taken to show that solar thermal systems are viable in the mid latitudes. However this system delivers heat at about half the temperature required in coal-fired power stations, and therefore does not have to concentrate solar radiation intensely. The absorber is about 1 metre wide and therefore reflectors can be wide with little curvature. Thus the capital cost is quite low. These features indicate that this plant is not a good guide to the effectiveness or cost of solar thermal plant at this latitude that would generate electricity without augmenting fossil fuel power generation. In a world that did not exceed safe greenhouse limits there could be few if any fossil fuel plants. Also the performance of the system falls markedly in winter as the above discussion would lead one to expect.
Dishes collect more energy in winter because they can be pointed directly at the sun, but there are two significant drawbacks. Their dollar costs are reported as being 2 – 4.5 times those of troughs (Sandia, undated), although costs will surely fall considerably with further development and mass production.
The data I have been able to access indicates somewhat surprisingly low but useful winter output from dish–Stirling devices. Some US dishes seem to have an average 24 hour flow equivalent of around 20 – 30 W/m (Davenport, 2008.) An output plot for the Mod dish-Stirling device shows that the January average flow (averaged over 24 hours) was c. 18 W/m, and for December, 22 W/m. Commonly published power curves show that at 700W/m output falls to around half peak output.
However this is not very relevant to our problem. If solar thermal systems are to provide electricity 24 hours a day, and also to solve the general intermittency problem set by other renewables, then heat must be stored. This means that the efficiencies will be much lower than those represented in the literature on dishes, which almost entirely deals with the direct generation of electricity via dish-Stirling systems. Dishes are not well suited to large scale heat collection.
Trough systems transfer heat long distances to the power block mostly through the absorber pipes, which are heated as they collect radiation (nevertheless 4% is lost, according to Sargent and Lundy, 2003). With a dish system this would not be so and either very long pipe distances would have to be insulated and would still lose much heat energy or many small generators would have to be located close to groups of dishes. (The equivalent of a 1000 MW plant in winter would involve tens of thousands of big dishes; below.) For these reasons the European and US dish developers I have contacted regard the use of dishes to collect heat as not being viable. (Personal communications.)
Kenaff’s pioneering work at White Cliffs, Australia on dish-steam generation (without storage) achieved 9.1% annual solar to electricity efficiency. The ANU Big Dish has a 13.9% efficiency, which it is expected can rise to 19% in future. (Note that this is a measure at a point in time under ideal conditions, not a measure of recorded annual average performance; which would be considerably lower, e.g., because of dust build up, warm up delay after cloud, etc. and winter performance.) I do not have figures on the winter performance of either system. If we assume 5 kWh/m/d radiation, the White Cliffs 15% loss of heat between collector and engine room, the 19% heat to electricity generation efficiency Lovegrove expects, and an 8% energy cost for pumping (the trough figure), then output might correspond to a gross 31 W/m 24 hour average flow. However this assumes 1000/m radiation and in winter radiation barely rises above 700 W/m, which for dish-Stirling generators cuts output in half. Again Kenaff’s evidence is that steam generation is significantly affected by lower DNI. (See Trainer 2008 for more detail.) Very important is the fact that the White Cliffs system involved only 14 rather small dishes and thus a very short distance for heat to be moved to the engine room, and the Big Dish is a single unit close to its steam generator. For the equivalent of a 1000 MW plant very long distances would be involved (or many small power blocks.). Thus it is not possible with this information to estimate a winter average net output after heat storage, but it would seem likely to be well below 30 W/m.
The heat storage strategy using dishes which looks most viable is that being developed by the ANU group, involving the use of ammonia dissociation as a means of heat storage (Lovegrove et al, 2004). Perhaps its main advantage is that no insulation is required, so heat would not be lost from the storage pipes or tanks. (The energy is stored chemically, in the splitting of ammonia into hydrogen and nitrogen.) This system is being built into a commercial plant at Wyhalla, South Australia. It is estimated that half the energy entering the dish might be available for generating after storage by this means. (Wizard Power personal communications.)
The designers cannot predict performance confidently at this stage (personal communications), and understandably will not make the technical information they do have available. It would seem from above that if Big Dish efficiency can be raised to 19% then after storage at 70% efficiency, then can overall efficiency of 1.3.% could be achieved. Gross winter output in Central Australia then might correspond to the region of 28 W/m continuous 24 hour flow. (At present Big Dish efficiency the figure would be 19.8 W/m/.)
Several factors would reduce this gross figure, including the effect of warm up delays after the passage of cloud, operating energy costs, energy embodied in the construction of the dish and the long trans mission lines (“emergy”), energy losses in those lines, and especially the embodied energy cost of the ammonia processing plant (including the reactors in the dishes which dissociate the ammonia, and the one in the power block that recombines it.). The emergy implications of the ammonia processing plant are difficult to assess, and could be problematic. The attempt sketched in Trainer 2008 suggests supply from a 1000 MW plant, taking the most favourable of the estimates for storage volume received (17 litres per kg of ammonia, and 4 MJ/kg), might require some 2,800 km of one metre diameter gas pipe, the intended containment vessel.
Also a concern is the fact that big dishes (Whyalla will use 500 square metre dishes) involve disproportionately higher materials and energy costs for structures, foundations and drive equipment, in view of the higher wind stresses they will have to cope with. An uncertain estimate based on the materials in the ANU Big Dish indicates an embodied energy cost three times that of troughs, i.e., in the region of 13% of plant lifetime output. (Trainer 2008.) However the developers believe other advantages of big dishes outweigh these factors, although I do not know whether this is only a dollar cost calculation or one focused on embodied energy costs. There is a high probability that in future the cost of materials and construction will be far higher than they are now.
These figures suggest that a solar thermal ammonia storage system will be capable of low but significant/useful net output in winter. From above, net energy delivered to distant users would seem likely to correspond to around 20+ W/m of collection area (not plant area including space between dishes). However it is not clear that the very large numbers that would be required could be afforded. On the above estimates a plant capable of delivering 1000 MW in winter might have to include up to 50,000 big dishes each of 400 square metres. These would probably have to be spread over an 11+km x 11+ km area. Just to connect the dishes to the power block would probably require some 2,800 km of pipe, although it would not have to be insulated. This seems to be about the length needed for storage above, but easily overlooked is the need for the same length of pipe to carry the recombined ammonia back to the dishes from the power block. Having many small plants rather than one big 1000 MW plant would not alter the overall ratio of pipe length per kW output. The total 5,600 km of steel pipe capable of taking 15mPa pressure might weigh 280,000 tonnes and have an embodied energy cost of 11.2 PJ. This is around 45% of the annual output of a 1000 MW power station (assuming .8 capacity), so the embodied energy cost of the pipe alone might add 2.2% of lifetime output to the total embodied energy cost figure.
Also of concern is the fact that if net delivered output corresponded to a say 25 W/m flow in winter, then it would take 46 metres of dish collection area to sustain one person at the Australian average electricity consumption rate, meaning that the ANU Big Dish would provide for only 8+ people.
Direct hydrogen production.
It is possible to produce hydrogen by splitting water at high temperature, around 800 degrees, and a practical application of solar thermal to this strategy is being discussed. (Taylor, Davenport and T-Raissi, 2008 ) A theoretical 40% solar to hydrogen efficiency is thought to be achievable. If this becomes viable it would probably be the best option, although it would involve the usual problems in large scale handling of hydrogen. These include pipe embrittlement, leaks, and the very low energy density meaning either very large storage volumes and/or high compression. Especially problematic are the energy losses in long distance transport. Bossell estimates that to pipe hydrogen from North Africa to Western Europe could require more than half the energy despatched from Africa. Ideal solar thermal sites are a long way from demand.
A system designed to deliver 1000 MW after storage would need a 1000 MW hydrogen-fuelled power station in addition to the dish system which generated the 1000 MW supply of hydrogen to run it, indicating high capital and embodied costs. The efficiencies of the various steps (e.g., .4 for hydrogen production, .8 for handling/transport, .4 for fuel cell generation) suggest an overall gross solar to wheels/use efficiency of 13%, from which the embodied and operating costs of materials-expensive hydrogen handling plant would have to be deducted. It is therefore not clear that this path would be more viable than the others considered above.
The intermittency problem.
The heat storage capacity of solar thermal systems overcomes some of the intermittency problems that trouble wind and PV systems, such as the occurrence of night time. The standard provision will be 12 hour storage enabling continuous 24 hour electricity delivery. However examination of climate data reveals that even at the best sites sequences of 4 or more days without sunshine are not unusual. The best US sites often have 2 runs of 4 consecutive days of cloud in a winter month. (Davenport, 2008)
If 1000 MW(e) output was to be provided for four cloudy day from stored heat, some 290,000MWh of heat would have to be stored. Storage cost has been estimated at $(A)10/kWh(th) meaning that the required storage plant would cost more than $8 billion, or around twice the cost of a coal-fired plant plus fuel. However this refers to trough technology and it is likely that for the ammonia process costs would be higher.
Again we would be faced with the prospect of very high capital costs for a large amount of plant that would not be used most of the time, and would still be insufficient occasionally. There would also be the question of whether there would be enough solar radiation in winter to meet daily demand and also recharge a large storage sufficiently to cope with the next run off 4 cloudy days.
The climate evidence given in Trainer 2008 seems to show that solar thermal systems even at the best locations would suffer a significant intermittency problem, despite their capacity to store energy.
Another problem is that if solar thermal plants are to help buffer the intermittency of inputs from other renewable sources then a major cost saving often claimed for solar thermal systems would not be available. The ability to store heat from peak mid day collection and generate with it at a much lower constant rate, perhaps .2 of peak capacity, means that much smaller and cheaper generators can be used, perhaps one fifth of the capacity that would be needed to use heat energy at the mid day rate of collection. The power block can make up around half of a solar thermal system’s cost so the saving in capital costs, energy costs and operations and management is considerable. However if the solar thermal component of a renewable supply system must at times plug gaps left by variable wind and sun, then there will be times when it must meet almost all demand and so its individual stations must often be cable of generating at much greater than average rate.
There would also be a problem regarding the need for solar thermal plant to rapidly ramp up to high levels of output, in order to meet most of the demand when sun and wind energies fall suddenly. Thermal generators can’t be brought up to full output quickly.
It is sometimes assumed that solar thermal systems will enable the gaps left b y other renewable sources to be smoothed out, by use of solar thermal heat storage capacity. For this to be plausible storage capacity would have to be extremely large. The following figures seem to show that this proposal is not viable.
If the solar thermal system was to average .3 of total electricity supply yet had the storage capacity to meet total electricity demand through 4 calm and cloudy days, then its storage capacity would have to be 24 times as great as for 12 hour storage. In addition its collecting capacity would have to be considerably greater then that required for average output, in order to accumulate such a reserve.
This evidence seems to mean that there is no chance that the capacity of solar thermal systems to store energy could overcome the problem of gaps left by combining the output from the other renewable energy sources, as some have hoped.
Solar thermal conclusions?
The climate data seems to show that despite their storage capacity solar thermal systems would suffer a significant intermittency problem and in winter would either need storage capacity for four or more cloudy day sequences once or twice each winter month, or would need back up from some other sources. This means they could not be expected to buffer the intermittency of other components in a fully renewable system.
It seems that troughs suffer a big drop in output in winter, that dish-steam systems cannot operate well enough on stored heat and that hydrogen generating systems are too handicapped by the usual difficulties associated with hydrogen. The prospects for satisfactory winter supply of electricity from solar thermal systems therefore seem to depend on whether or not the dish-ammonia system will be viable on a large scale, and capable of overcoming intermittency problems. The unsatisfactory information available suggests that they will be significant contributors but are not likely to make possible reliable winter electricty supply at a tolerable cost, that they will suffer a significant intermittency problem, and that they cannot be a solution to the integration problems left by other renewables.
The dumping and additivity problems (for technosolar in general)
Discussions of renewable energy contributions to total demand also typically give little or no attention to the problem of energy dumping set by the variability of renewables. This increases the need for energy conversion and lowers average capacity factors. Consider a system in which over time wind and PV each supply one-third of demand (i.e., .33 x total demand). The world average wind farm capacity is .23, which means that at times output from a farm will approach 4.3 times average output. For PV systems in good locations annual average capacity is probably about .18, meaning that on a sunny day a system will be producing about 5.6 times average output. Now consider a system in which wind and PV components each contribute on average .33 of total demand. On a sunny day which is also quite windy these two components might be producing about 3.3 x (total demand). So even if the non-renewable components in the system can be shut down, 2.3 times as much energy as is needed would have to be dumped, or stored inefficiently as hydrogen. This would have a dramatic effect on the system’s average capacity measured in terms of energy actually used. (Storing as hydrogen and regenerating via fuel cells might yield no net electrical energy when system operating and embodied energy costs are deducted from the perhaps 25% of energy available after conversion, storage and retrieval.)
Stern’s Fig. 9.4 shows that this dumping problem has not been taken into account. Total demand is divided into components reflecting averaged or annual contributions with no consideration of what would happen when some or all are performing at their peak capacity rather than at average capacity.
The variability between summer and winter would more or less double the magnitude of this problem for solar sources, given that in good solar regions winter insolation is about half the summer value. (The multiple is greater in the lower latitudes.) Thus a PV system designed to meet 30% of demand in winter might meet 60% of it in summer. The effect would be offset to some extent by the fact that winds tend to be higher in winter. However with daily variability the effects compound rather than compensate; i.e., at night when there is no solar input winds tend to be lower.
Renewables are not additive
Renewable energy sources are usually thought of as additive, that is, as if building X GW of wind capacity and X GW of PV capacity would give us 2X GW of generating capacity. However on calm nights these two sources would give us no generating capacity at all. Thus they are best thought of as sources which at times can be alternated with or substituted for coal fired power, but not as sources which can always be added to each other. (Stern’s Fig. 9.4 reveals that the various components are being thought of as additive.) This means that we might have three or more very expensive systems each capable of more or less meeting demand while the others sit idle, and in addition we must retain a coal or nuclear system capable of meeting most or all demand when most or all the renewables are down.
Questions that arise: (from a recent email to me by Ted)
1. How viable is storing heat from dishes? My European contact says not worth it. There would be long distance piping from tens of thousands of big dishes for a 1000MW power station, and thus large heat loss. With troughs heat moves through heated absorbers most of the way.
2. Where can we get performance data, for the different ST types, for every day of the year, along with solar radiation data for the sites in question?
3. What is the embodied energy cost of the ST types; Lenzen and Dey say 4% for troughs, but my attempts to check this are unsatisfactory, and it seems that for dishes the cost would be much higher (Derived from big dish materials inventory.)
4. Analysis is required of climate data for Eastern Sahara, SW US, and Central Australia, especially actual daily pattern of radiation (ideally pattern of hourly radiation every day.)
Another issue that require resolution concerns the angle between sun and reflector (the ‘cos‘ problem). With dishes this is no problem ever because you can point them at the sun anytime, so I have tried (in the above) to deal with dishes mainly, because if that’s problematic then towers and troughs are out…because they do have an intractable cos problem, especially severe in winter even in the best regions. Troughs it seems just can’t do it, as performance goes down to 20%- of summer output in winter in the US good sites. Towers (CR) are of course good for storage, but I’m assuming their cos problem is serious. It would be great to get some actual data on their year round performance. I have found it fiendishly difficult to get such data out of anyone; they seem not to want to make it public, and this makes evaluation of claims very difficult.
Another problem is that if heat loss is 1% over a 12 hr period, which is what Sargent and Lundy seem to say, then over a 4 day period this is 8%, so would have to be added to other factors detracting from overall ER and dollar cost situation. BOM is where to go for climate data (in Australia) and I have put some time into this. I could get more detailed information than I have but what I have indicates that there is a problem with sequences of cloudy days in winter, even in Central Australia. US seems much worse. Can’t get much confident data on North Africa, which is what ST electricity for Europe would depend on. Note a problem is time taken for systems to come back up to output after “transients”; i.e., passage of cloud. Can be significant I believe, and thus average radiation per day could be misleading; if this is made up of a lot of passing cloud and sun you might not total much time on full output.
If anyone can help Ted source relevant information on the above (or if you have insider information, email him), I’m sure he’d greatly appreciate it.