Saturday, November 3, 2007

Scaling solar power, midway to success

As of summer, 2007, the solar electricity plant with largest annual output was near Serpa, Portugal. In 2009 a larger plant was scheduled to begin operating near Brandis, Germany. Both use photovoltaic modules. One would like to use these plants as practical measures for the progress of solar power, but information about them has been slippery, coated in marketing muck that unskilled journalists fail to penetrate.

Following are approximate land area, capital cost and operating estimates released by plant developers:

Power PlantLandCost, US$PowerEnergy
locationsq mibillionGW, peakTWh/yr
Serpa, Portugal0.
Brandis, Germany0.

Serpa, Portugal, was developed by SunPower of San Jose, CA, United States, using single-crystal modules made by themselves and other manufacturers. One of SunPower's latest panels, the SPR-95, was rated at 95 W, peak, and had an area of 5.89 sq ft. If these panels covered 150 acres at full density, they would provide rated peak power of 105 MW. The Serpa installation is quite sparse, providing about 10 percent of full density.

SunPower was currently developing a similar plant rated 15 MW, peak, in the United States at a southwest desert location near Las Vegas, NV. The plant area is 140 acres. If SunPower SPR-95 panels covered this area at full density, they would provide rated peak power of 98 MW. The Las Vegas installation is denser than Serpa, providing about 15 percent of full density.

Brandis, Germany, was being developed by Juwi of Mainz, Germany, using thin film modules from First Solar, sold commercially by Blitzstrom. One of Blitzstrom's latest panels, the CTS-272, was rated at 72.5 W, peak, and had an area of 7.75 sq ft. If those panels covered 297 acres at full density, they would provide rated peak power of 121 MW. The Brandis installation is somewhat sparse, providing about 33 percent of full density.

The Brandis plant uses fixed mounts, while the Serpa plant uses horizontal, single-axis moving mounts. The Las Vegas plant uses inclined, single-axis moving mounts for which SunPower has not yet released performance information. Data comparing the systems are not consistent. Curtright and Apt of Carnegie Mellon Electricity Industry Center [3] report results from fixed-mount and horizontal, single-axis moving-mount installations in United States southwest desert locations, latitude about 33 degrees, both achieving annual average capacity factors slightly over 19 percent. Marion, et al., writing for the U.S. National Renewable Energy Laboratory [4], report Arizona results from 12 inclined, single-axis moving and 2 fixed-mount installations achieving 21.4 and 17.4 percent annual average capacity factors. Results are affected by plant layouts, with sparser layouts providing less self-shading and achieving higher capacity factors.

By comparison, Serpa plant developers were estimating annual capacity factor 22 percent for an upland location about 50 miles from the ocean at about 38 degrees latitude. Brandis plant developers were estimating annual capacity factor 11 percent for an inland, low-altitude location at about 51 degrees latitude. NASA records for average insolation at Serpa, Brandis and Scottsdale, AZ, are 4.4, 2.8 and 5.1 KW/sqm-day [5]. Scaling from Curtright and Apt measurements, the Brandis annual capacity factor estimate appeared realistic, while the Serpa estimate appeared optimistic by about 30 percent, possibly justified by its sparse layout and its moving mount design.

The Brandis installation provides the most favorable benchmark, about US$4.3 million and about 7.4 acres per peak MW. For a benchmark plant with output of one TWh per year, using 19 percent as an achievable capacity factor for southwest desert locations in the United States leads to estimates of US$2.6 billion in capital costs and 7 square miles of land.

The benchmark solar plant output of one TWh per year is 25 times the output predicted from the Brandis plant but only about 1/5 the output from a modern, high-efficiency coal-fired plant complying with current U.S. new source emission standards [6]. The announced capital cost of that coal-fired plant was US$1.3 billion [7].

Estimates of how much electricity might be supplied by solar power vary. Lack of practical resources to store very large amounts of electrical energy over several days may limit solar power to no more than 20 percent of total consumption. Total U.S. electricity consumption for 2004 was 3,548 TWh [8]. Supplying 20 percent of that would take 710 benchmark solar plants. Benchmark estimates predict total cost at US$1.8 trillion and total land area at 5,000 square miles, less than 5 percent of the area of Arizona.

Other practical considerations will intrude. The U.S. electrical grid currently uses up to 750 KV lines to transmit power up to about 400 miles. Line voltage would need to increase by a factor of 5 or more to supply the continental United States from southwest desert locations. Without proven technology to do that, solar plants would have to be built in less favorable locations. A comparison of projected capacity factors for Serpa and Brandis shows that less favorable locations can double the cost and the land area required.

The slim and brief experience operating large solar electricity plants provides inadequate background to estimate long-term maintenance costs. Tucson Electric Power published a summary of experience with a fixed-mount installation growing over 5 years to 5 MW, peak [9], showing a 19.2 percent average capacity factor and 0.3 percent down time for 2004 but giving no year-by-year performance indicators. The system, located in United States southwest desert, supplies auxiliary power to one of the coal-fired plants that then produced 99.95 percent of the utility's electricity.

Outputs from solar modules are known to degrade over time, but there are no data measuring trends for modern solar modules over several decades of operation. Maintenance of exposed, moving collector arrays also has no long-term track record. Desert locations are most favorable for solar intensity, but they provide little rainfall to wash away dust, and they carry risks of sandstorms that abrade exposed surfaces. It may be possible to provide replaceable surfaces and moving parts, but the potential cost and effectiveness of such approaches are as yet unknown.

Capital costs dominate all types of new electricity generation in the United States [10] except combined-cycle natural gas. Fuel and other operating costs for new coal-fired plants are estimated to be less than half the carrying costs of capital. Benchmark solar plants based on the Brandis design would carry capital costs 10 to 20 times those of new coal-fired plants with the same total output, depending on solar plant locations, with zero fuel cost providing only a small offset. Unsubsidized solar electricity prices from plants in favorable locations would be more than 5 times the economic prices for coal-fired electricity from new plants.

Capital costs for photovoltaic systems have declined substantially over the past 30 years, but prospects for future reductions in solar plant costs are uncertain. In the twenty years up to 1995, inflation-adjusted prices of unmounted solar modules dropped about 11 percent per year [11]. Installation costs now comprise more than half the total costs of large systems [12]. The inflation-adjusted trend between mid-1998 and mid-2005 shows the installed cost for large systems falling about 5.5 percent per year. If that trend were to continue, unsubsidized solar electricity prices from plants in favorable locations would become competitive with current, coal-fired electricity from new plants in about 30 more years.

[1] SunPower, Serpa solar power plant, 2007, at

[2] Juwi, Brandis solar power plant, 2007, at

[3] Aimee E. Curtright and Jay Apt, The character of power output from utility-scale photovoltaic systems, Carnegie Mellon Electricity Industry Center Working Paper CEIC-07-05, 2007, at

[4] B. Marion, J. Adelstein, K. Boyle, H. Hayden, B. Hammond, T. Fletcher, B. Canada, D. Narang, D. Shugar, H. Wenger, A. Kimber, L. Mitchell, G. Rich and T. Townsend, Performance parameters for grid-connected PV systems, U.S. National Renewable Energy Laboratory Report NREL/CP-520-37358, 2005, at

[5] U.S. National Aeronautics and Space Administration, Surface meteorology and solar energy database, at

[6] Arkansas Department of Environmental Quality, Draft operating air permit 2123-AOP-R0, July, 2007, at

[7] American Electric Power, SWEPCO announces Hempstead County as site for new baseload generation power plant, August, 2006, at

[8] U.S. Energy Information Administration, U.S. electricity consumption by sector, 2004, at

[9] Larry Moore, Hal Post, Tom Hansen and Terry Mysak, Photovoltaic power plant experience at Tucson Electric Power, Tucson Electric, 2005, at

[10] U.S. Energy Information Administration, Annual energy outlook 2007 with projections to 2030, at See Figure 59.

[11] U.S. Department of Energy, Technology Characterizations, Overview of PV Technologies, 1997, at See Figure 1.

[12] Ryan Wiser, Mark Bolinger, Peter Cappers and Robert Margolis, An empirical investigation of photovoltaic cost trends in California, Lawrence Berkeley National Laboratory Report LBNL-59282, 2006, at See Figure ES-2.