Updated 5/1/2011

Concentrated Solar Power (CSP)

In 2000, the National Research Council (NRC) recommended a detailed review of CSP and its potential be carried out by an independent engineering firm, as they felt the DOE’s earlier conclusion that CSP did not merit further support was unjustified. In early 2003, Sargent & Lundy, submitted the 340-page report of their findings, assessment, recommendations, and projections to the DOE-NREL. Fortunately, they concluded that CSP had enormous potential and deserved further support. The technical material collected in this report is of enormous value.

However, their cost projections beyond 2003 had limited value. Unfortunately, they are still generally accepted by many program managers and CSP advocates, even though today they are low by factors of 2 to 4.

Our cost data reported below are mostly based on discussions with corporate personnel in early 2008 and are largely reflective of 2007 prices. An updated study similar to the 2002 Sargent & Lundy Report is desperately needed. There have reportedly been some CSP deals at about $2.5 per peak watt (W_PE), but small early deals are often done at a big loss. The biggest recent announcement that included price information was the Lockheed-Starwood deal for a 290 MWPE (peak electrical output) trough plant for $1.5B. That comes to $5.2/W_PE. Below is our best effort to summarize and update the big picture.

The Mean Price Referent (MPR) for electrical power in California today is about $0.13/kWhr, but electrical power is more expensive during the hot afternoons when solar is producing most of its output. Hence, CSP plants typically sell their power at over 1.5 times the MPR, or about $0.2/kWhr because of favorable Time-of-Day (TOD) pricing. With the 30% investment tax credit (ITC) that has been available and good TOD pricing, CSP briefly became a break-even proposition in some places. This would suggest the cost of energy from CSP without the ITC is about $0.26/kWhr, or $72/GJ in choice sites. (Off-peak wind energy in ideal locations with a 25-year project lifetime is less-than one-fourth as expensive.)

Solar, wind, nuclear, and geothermal energy costs depend as much on the interest rate, tax credits, project lifetime, and inflation rate as they do on the initial plant cost per peak watt, capacity factor, and O&M (operating and maintenance) costs. Interest rates have dropped and project lifetimes are increasing. Calculating the effects of these rather unpredictable factors is one of the difficult parts of determining the cost of renewable energy, but these factors are all similar where fuel costs are negligible and O&M costs are small. In these cases, useful comparisons between various options can be made simply on the basis of F/C_PE, where F is the capacity factor (the ratio of mean electrical output power to peak output, averaged over the year) and C_PE is the initial cost per W_PE.

It is useful to present some data here on costs and performance, but it is essential to keep in mind that these data still have limited validity with respect to estimating future prices. The cost of CSP is strongly dependent on the price of high-quality mirrors, extruded aluminum, and the vacuum-insulated receiver tubes.

There are sound reasons to expect reductions in mirror costs with higher volume production, but this will be offset in part by the increasing costs of aluminum and steel. The cost of the receiver tubes should also be able to be reduced considerably with very high volume production.

Nevada Solar One, the most advanced trough-plant (commissioned in mid-2007), has peak electrical output capacity of 70 MW_PE. It cost about $3.5/W_PE and has peak collection efficiency h_C (the product of receiver efficiency and concentrator efficiency at design operating temperature) of about 66%. Its peak thermal conversion efficiency h_T (the ratio of thermal input power to electrical output power at the grid voltage) is about 35%. Its capacity factor F is about 22%.

With no thermal storage and for the typical solar insolation patterns where CSP is being installed, the mean solar efficiency (MSE), the ratio of total electrical energy output to solar energy striking the mirrors over the course of the year, is given approximately by the following expression:

MSE = 2.3 F h_Ch_T .

(There are limitations to the above because it is based on how mean performance generally relates to the only three numbers usually publicly reported – two peak efficiencies and the capacity factor. For more detailed calculations, see page 253 in the Sargent-Lundy Report.)

For the Nevada Solar One case, MSE is about 11.7%. However, it is worth noting that to achieve even this level of performance, the land area required is about four times the collector area, primarily to minimize collector shading and secondarily for service access and fluid lines.

A lower-cost CSP plant design by Ausra was expected to achieve h_C = 62%, h_T = 19%, and F = 17%. This translates into a mean solar efficiency MSE of 4.6%. Company promoters indicated the price would eventually be $2.5/W_PE, but available data on the 5 MW demo plant in Barkersville and the 177 MW plant that had been planned for San Luis Obispo CA (in 2010) suggest the cost would be at least twice that amount. (Ausra tried to re-invent their goals several times in 2008-2009, but eventually failed. Their wreckage will apparently be bought by Avera. We pointed out the fallacies of Ausra’s approach to CSP in mid 2007.)

A high-end CSP plant (such as the PowerTower by Solar Reserve, a union of Rocketdyne and US Renewables Group) will probably cost over $6/W_PE (still without storage) and perhaps have h_C = 77%, h_T = 38%, and F = 27%. In this case MSE would be about 18%. (These are recent, orally reported data from company personnel. The numbers from ~2000 in the Sargent-Lundy Report are much worse. They calculate MSE=13.7%)

The biggest difference between these plants is the peak thermal conversion efficiency, which is mostly dependent on the peak operating temperature and how much is invested in the power block. The high-end plant operates at up to 840 K, the mid-range plant up to 660 K, and the low-end plant up to 520 K. The second biggest difference is the capacity factor, and this depends mostly on the quality of the off-peak tracking, which is not as effective in the linear Fresnel array used in the low-cost plant. Of the above three CSP examples, the high-cost design seems to be the least cost-effective (based on F/C_PE), and the other two look similar. O&M costs (where weekly washing of the collectors may be the biggest component) would be higher with the low-cost design, as it requires four times as much collector area as the high-cost design.

We have several pending patents that will help bring the cost of CSP down in the future – particularly in areas where geothermal is also available to enable a novel Geo-CSP hybrid. However, CSP (and even our Geo-CSP hybrid) will not compete with off-peak wind for synthesizing fuels from CO₂.

In 2003, Sargent & Lundy projected a 100 MW_E Power Tower in 2010 with 13 hours of thermal storage would have capacity factor of 73%, mean thermal collection efficiency of 46%, peak thermal conversion efficiency of 42%, mean solar efficiency of 16.1%, and would cost $4.6/W_PE. Not surprisingly, all their cost projections were quite low, especially the cost of thermal storage. As a result, Power Towers are usually built with only 3 to 8 hours of storage, capacity factor is much lower, and the plants being built are smaller (usually in the 15 to 50 MW_E range). They also expected Power Towers to be more cost effective today than trough plants, but that does not yet appear to be the case.

Fuels by High-Temperature CSP. There has been steady effort by several groups for the past decade to show that it may be useful for CSP to go well beyond 840 K – the highest temperature utilized thus far in a large-scale CSP demonstration. The motivation is that it may be possible to drive some endothermic chemical reactions that could permit the synthesis of renewable liquid fuels at least partly from water and waste CO₂. One of the more studied routes has been “dry reforming” (also called “CO₂ reforming of methane”) – producing syngas (a mixture of CO and H₂) from methane and CO₂ by the following extremely endothermic reaction:

CO₂ + CH₄ 2CO + 2H₂ ΔH_600K = 257 kJ/mol

The syngas produced can then be converted to fuels and chemicals by Fischer Tropsch synthesis. Half of the carbon in the above syngas is from waste CO₂, and half is from fossil methane. Unfortunately, the thermodynamics for the above requires the reaction to be over 900 K to achieve usable CO₂ conversion and CSP at high temperatures is expensive.

The best documented large CSP case at moderately high temperatures (840 K) is the PowerTower (as noted earlier), and it achieved a peak collection efficiency of 77% at 840 K. The mean annual thermal collection efficiency (product of field and receiver efficiencies) in Solar Tres (Spain, 14 MW_PE, 840 K, 2004) is 41%. The cost of the thermal energy (prior to conversion, and backing out the cost of the conversion equipment) was over $70/MWhr in Solar Two (twice the cost of wind electrical energy in prime sites). The 4th-power temperature dependence in the Stefan-Boltzmann law suggests the concentrator challenges would increase at least quadratically with temperature, and other factors (especially emissivity) contribute further rapid increases in energy cost with temperature. This is borne out by the very limited cost and efficiency data that are available for CSP between 660 K and 1000 K.

For extremely endothermic reactions near atmospheric pressure, the CSP receiver needs to be ~200 K above the minimum reactor temperature. Even with expensive Pt/ZrO₂ catalysts, CH₄ conversion is still typically only ~25% (half of thermodynamic limits) at 900 K, and only ~70% at 1100 K. Separation of CH₄ from syngas is an order of magnitude more costly and energy intensive than the other separations in syngas processes, so most dry-reforming demonstrations (such as the SPARG process, Sterling Chemical Inc.) have operated ~1200 K, both to limit CH₄ slip and to prevent catalyst coking (above 980 K, the reverse Boudouard reaction converts the C to CO).

It has recently been shown that sub-micron-sized carbon particles will also “catalyze” dry reforming; but they must be continuously supplied, as they are also consumed in other reactions. Moreover, achieving practical conversion rates requires reactor temperatures above 1350 K.

The H₂/CO ratio from dry reforming is half that needed in the FT process, so additional reactions (steam methane reforming and water gas shift) are needed, which further reduce the fraction of renewable C that ends up in the fuels.

It is really not difficult to achieve CSP temperatures above 1300 K when neither efficiency nor cost of energy is an issue – children all over the world have been doing it for centuries with simple optics. Many research groups and small companies have reported only stagnation temperature at essentially 0% efficiency. Some experiments have reported excellent efficiency above 1200 K, but the apparatus is prohibitively expensive and of short useful lifetime.

The challenges are more than just getting solar energy above 1100-1400 K at an affordable price. For example, one of the more extensively researched processes begins with endothermic reduction of Fe₃O₄ to FeO at over 1400 K followed by exothermic splitting of water (by reacting it with the FeO at ~1000 K) to yield H₂ and Fe₃O₄. After nearly a decade of well-funded efforts toward thermo-chemical production of hydrogen from solar heat via this approach, some of the more recent experiments (reported July, 2009, ASME Sustainable Energy Conference, by a group from the German Aerospace Center) apparently achieved a mean hydrogen production rate of about 2E-8 kg/s from about 4 kW input for several hours. This is equivalent to about 0.1% conversion efficiency, and it was about one-tenth of what their simulations were predicting. It appears their more recent simulations for a larger system with a 90 kW solar flux are predicting H2 production of about 1.2E-6 kg/s, which would correspond to an efficiency of ~0.19%.

One of the highest experimental efficiencies reported thus far in a solar thermo-chemical path (CO production from C+CO₂ in the reverse Boudouard reaction) briefly achieved about 12% efficiency (at about 1400 K, and again, one could expect another 50% loss in a practical concentrator); but the efficiency of the reactor dropped in half in about an hour. Few (if any) experiments have exceeded more than a few percent system efficiency after a few days of operation, and none of the designs appears to be scalable.

A paper published in early 2008 by perhaps the most famous research group promoting thermo-chemical routes to hydrogen production projected over 60% solar-irradiation-to-chemical efficiency should be possible at the 30 MW level. Some simple calculations are useful in shedding light on efficiencies typically reported by this group. For example, one of their experiments reported achieving up to 30% efficiency in a process that produced 95% zinc (from ZnO+C) at a peak rate of 50 kg/hr from a receiver solar flux of ~300 kW into the receiver. However, the peak theoretical limit to the hydrogen that could be produced from the above zinc would be ~1.5 kg/hr, which represents a chemical power (HHV) of 60 kW. The reaction also consumed carbon at the peak rate of 9.2 kg/hr (which represents an input chemical power of 85 kW) and produced CO at the peak rate of 21 kg/hr (which represents an output chemical power of 60 kW). From this perspective, their peak theoretical chemical conversion limit would be 35 kW – or about 12% efficiency before considering some of the larger process losses (or the losses in the solar field).

The total zinc production during the day was about 2.4 times the maximum produced in one hour, and they reported a peak solar field efficiency of 60%. Since the total solar insolation over the day is typically equivalent to about 6 hours at the peak rate, it looks like their mean theoretical conversion efficiency from solar field to syngas might approach 2.9%. But this does not include numerous parasitic losses: (1) preparing and loading the pulverized reactants (ZnO and beach-wood charcoal); (2) production of zinc particles from the zinc vapor; (3) separation of the nitrogen sweep gas from the CO; (4) cleaning the reactor and quartz window for the next batch; and (5) production of hydrogen from zinc particles and water.

Yes, that’s right! No hydrogen was produced, and the amount of CO that was captured was not reported. So it seems they simply consumed reactants and produced nothing but a few hundred dollars worth of dirty zinc particles (from what was probably a $10M experiment).

The hydrolysis of water using zinc particles does not appear to be practical, as the particles are immediately coated with a thin film of ZnO and C from the exothermic reverse of the reaction that forms them. To limit oxidation, the Zn-CO+N₂ gas from the reactor is immediately quenched, which also means the enthalpy of these products cannot be well utilized. The hydrolysis reaction proceeds very slowly below 650 K, and at higher temperatures the oxidation reaction proceeds at a similar rate unless the CO has been replaced with an inert gas. Separating micron-sized particles from a gas carrier is complex and expensive. Thus, we see no hope for this route to hydrogen production ever becoming practical.

Another paper was published by the Steinfeld group in Science in February 2011 that again generated a lot of attention with yet another method of attempting to produce syngas using concentrated solar energy. This time, they claim to have achieved 0.4% efficiency using a ceria catalyst at up to 1800 K, but again numerous factors would lead to actual efficiency being extremely negative in any real application – at any scale.

A number of even higher temperature endothermic reactions have also been proposed, but they make even less sense from a practical perspective. Amazingly, one company has reportedly received hundreds of millions of dollars from (naive) investors for thermal dissociation of water into hydrogen and oxygen at over 1800 K. Even more amazingly, research funding agencies continue to support high-temperature CSP ideas that have little or no chance of competitively producing renewable fuels (such as using dish concentrators to drive stirling engines, or other similarly cost-ineffective proposals). After more than a decade of research by several groups, the first demonstration of significant syngas or hydrogen production by CSP-driven reactions, irrespective of cost, still appears to be many years away.

After more than a decade of research by many groups, the first demonstration of substantial syngas or hydrogen production by CSP-driven reactions, irrespective of cost, still appears to be many years away.

The dish-concentrators driving Stirling Engines are another example of HT-CSP ideas that never compete in the real world. In early 2009 SES was expecting to have 300 MW of their systems installed by mid 2010, but the actual amount was 1.5 MW. There is still no publicly available cost information, and no data showing such systems will hold up over years. In late 2010, these developments ground to a halt, after showing runaway losses. Again, investors were left empty handed.

The argument always made for solar compared to wind is that good quality solar energy is much more widely distributed than good quality wind. However, high-temperature CSP is unlikely to ever be practical in plants below 100 MW_T, and such plants require a large area of cheap land. Dual-use of the land is virtually impossible with CSP (unlike with wind energy), and cheap land is usually dry and dusty (or at least it will be after the mirrors are put up, as they prevent growth of ground cover). High precision mirrors are required for HT-CSP, and they must be kept extremely clean. Even in desert areas, dew falls at night, and the dust sticks to the mirrors. The Nevada Solar One (200 MW_T, operating below 660 K) washes their mirrors (with water) weekly on average. To operate efficiently at 1200 K requires over 3000 suns and solar tracking errors under 0.05^o in both directions. Daily mirror washing may be required for CSP above 1200 K.

However, there are more fundamental reasons for doubting that CSP above 1200 K will ever prove practical, with the most significant being the receiver thermal flux problem. Current towers (840 K) can operate with surface fluxes in the range of 0.6 to 1.2 MW/m² because low-melting salts are available that have sufficiently low reactivity (with the other materials present) up to at least 900 K. If the receiver thermal flux does not increase as the 4th power of temperature, receiver efficiency will drop. Heat transfer liquids that will work stably and conveniently with affordable receiver, ducting, pump, and sealing materials above 1300 K, with receiver heat fluxes above 2 MW/m², can probably not be developed. Hence, one plan has been to dispense with the liquid intermediary. However, the highest gas-phase heat fluxes that are practical (200 W/m²K) are lower than what is needed (for high receiver efficiency) by two orders of magnitude. This hasn’t been a problem in the experiments thus far because they haven’t worried about cost, lifetime, and scale-up. Obviously, these issues can’t be ignored in commercial power or fuel production.

References:

The end of solar Stirling engines:
http://www.greentechmedia.com/articles/read/stirling-tessera-projects-frozen-amid-turmoil/

A Pessimistic outlook for CSP from an optimist
http://www.greentechmedia.com/articles/read/is-CSP-doomed/

First Power Towers to be built soon in the U.S.
http://www.energy.ca.gov/sitingcases/ivanpah/index.html

Solar Millennium has final approval for a 1000 MW plant, but needs investors
http://www.greentechmedia.com/articles/read/largest-solar-plant-in-the-world-receives-final-regulatory-approval/

French nuclear company Areva will buy Ausra:
http://www.renewableenergyworld.com/rea/news/article/2010/02/areva-to-acquire-csp-company-ausra

Stirling CSP progress
http://www.renewableenergyworld.com/rea/news/article/2010/06/maricopa-solar-project-is-big-test-for-stirling-technology?cmpid=WNL-Wednesday-June23-2010

Perhaps six 250 MW CSP trough plants will be built in the US Southwest (Solar Trust of America) over the next 4-7 years, with the first possibly to begin producing in 2013:
http://www.blm.gov/pgdata/etc/medialib/blm/nv/field_offices/las_vegas_field_office/energy/proposed_solar_millenium.Par.48331.File.dat/Copy%20of%20AFR%20FINAL%20POD%20120408.pdf

http://www.solartrustofamerica.com/news.cfm?newsID=19

http://www.basinandrangewatch.org/AV-SolarMill-scoping-Aug2009.html

http://renewableenergydev.com/red/solar-energy-mojave-solar-park-csp/

The Lockheed 290MW, $1.5B CSP-trough announcement:
http://www.greentechmedia.com/articles/read/lockheed-martin-starwood-to-build-290mw-solar-thermal-plant-in-arizona/

http://www.nrel.gov/csp/troughnet/

http://en.wikipedia.org/wiki/Solar_power_tower

http://www.solarpaces.org/Library/docs/SargentLundyReport.pdf

http://www.solarpaces.org/Library/csp_docs.htm

http://www.solarpaces.org/CSP_Technology/docs/solar_tower.pdf

http://www.solarpaces.org/Library/CSP_Documents/2007_concertrating_solar_power_en.pdf

E Prabhu, “Solar Trough Organic Rankine Electricity System (STORES)”, NREL/SR-550--39433, 2006.
http://www.nrel.gov/docs/fy06osti/39433.pdf

R DiPippo, “Geothermal Power Plants: Principles, Applications and Case Studies”, Elsevier Ltd, Oxford U.K., 2005.

A Borsukiewicz-Gozdur, W Nowak “Comparative analysis of natural and synthetic refrigerants in application to low temperature Clausius-Rankine cycle”, Energy, 32, 344-352, 2007.

H Price and V Hassani, “Modular Trough Power Plant Cycle and Systems Analysis” NREL/TP-550-31240, 2002.

High-temperature reforming:

M Neises, F Goehring, M Roeb, C Sattler, R Pitz-Paal, “Simulation of a Solar Receiver-Reactor for Hydrogen Production”, ES2009-90273, ASME Joint Conferences, San Francisco, 2009.
T Kodama, SI Inuta, T Hatamachi, N Gokon, SG Yamashita, T Seo, “Molten-salt Tubular Absorber/Reformer (MoSTAR Project..”, ES2009-90230, ASME Joint Conferences, San Francisco, 2009.

HH Klein, J Karni, R Rubin, “Dry Methane Reforming Without a Metal Catalyst in a Directly Irradiated Solar Particle Reactor”, J Solar Energy Engr., 131, 021001-1-14, 2009.

C Singer, R Buck, R Pitz-Paal, H Muller-Steinhagen, “Assessment of Solar Power Tower Driven Ultra Supercritical Steam Cycles Applying Tubular Central Receivers with Varied Heat Transfer Media”, ES2009-90476, ASME Joint Conferences, San Francisco, 2009.

C Wieckert, E Guillot, M Epstein, G Olalde et al, “A 300 kW Solar Chemical Pilot Plant for the Carbothermal Production of Zinc”, ASME: J Solar Energy Engr. 129, pp 190-196, 2007

http://www.ub.utwente.nl/webdocs/ct/1/t0000001.pdf

G Olah and A Molar, Hydrocarbon Chemistry, 2nd ed., Wiley, 2003.

M Aresta, ed., Carbon Dioxide Recovery and Utilization, Kluwer Publishers, NL, 2003.