By Jason Deign in Barcelona

What happened at Shams 1 would nowadays probably be classified as a schoolboy error. Last year, the CSP plant’s owner, Masdar, confessed that it had overestimated how much sunlight would reach the parabolic trough field. It had not factored in the amount of dust in the air.

“The Shams original estimate was based on satellite data and was off by more than 10%,” says Marcel Suri, managing director of GeoModel Solar, a solar resource consultancy.

Although Masdar claimed it could correct the problem through the addition of extra mirrors within the margin for error in its financial plans, the setback was a timely reminder for CSP developers of the need to take solar radiation levels very seriously indeed.

Solar irradiation, commonly measured in terms of direct normal irradiance (DNI), is essentially a measure of the amount of fuel available to power a CSP plant, Suri says. “If you have 30% more fuel it makes a big difference.”

And substantial differences are not hard to find. The minimum threshold DNI for solar thermal development is generally taken to be 1,800 kWh/m2/year. In Spain, the cradle of modern CSP, DNI levels range from about 2,000 to 2,100 kWh/m2/year.

But head down to South Africa’s Northern Cape province and the DNI easily tops 2,900 kWh/m2/year, a 40% increase. “This is a very important consideration when you are talking about economics,” says Suri.

GTM Research senior analyst Brett Prior illustrates this point with reference to DNIs across California, using figures from the US National Renewable Energy Laboratory’s Solar Radiation Data Manual for Flat-Plate and Concentrating Collectors, one of several online data sets.  

A big difference

Even within a single state, he notes, there can be big differences in DNI. Daggett, renowned for its sunshine, gets 7.5 kWh/m2/day (equivalent to more than 2,700 kWh/m2/year), while Arcata, on the other side of San Francisco, gets just 3.5 kWh/m2/day (less than 1,300 kWh/m2/year).

What this means, Prior says, is that if in Daggett a CSP project can deliver power at, say, 10 cents per kilowatt-hour, the exact same facility in Arcata would only produce half the amount of power, “so your cost per kilowatt-hour would exactly double. It's a very big difference.”

Another way of looking at the issue is that for lower DNIs “you need a bigger collector field,” says Jenny Chase of Bloomberg New Energy Finance, although clearly this will have an impact on plant costs.

This is why CSP is feasible in places like China. On one hand, says Chase, “The DNI in parts of China is good; the Tibet plateau is quite decent,” and on the other, “everything is cheaper in China.”

Because of DNI, states Prior: “You have to choose your location carefully. I think it's one of the reasons people are very excited about going to a place like the Atacama Desert.

“The Atacama Desert is even better than Daggett, so you potentially have even lower costs than putting the plant in the Mojave Desert. DNI has a massive influence on the economics. It's the reason that you generally only see CSP in very sunny places.”

However, the DNI levels published online by organisations such as the National Renewable Energy Laboratory (which also offers a system advisor model with performance parameters for CSP plants) are only part of the story, as Masdar found out with Shams 1.

Satellite database

“If you take an irradiation number from a satellite database which is not corrected you could be off by 10% or 12% in some regions,” Marcel Suri cautions.

The problem is that satellite data does not take into account particulates such as dust, water vapour or smoke in the air, which can cut the amount of sunlight reaching the earth by vital percentage points.

The level of airborne particles varies significantly by region, country and even locality. In countries such as Spain or South Africa, for example, it can add a margin of error of 5% to 7% to standard satellite data.

But in the Middle East, where dust storms are more frequent, or India, where fires are commonplace, the margin of error can be twice this level.

The only way to get round this problem, Suri says, is to carry out on-the-ground readings using high-quality measurement devices once a site has been selected. This could take some time; Suri advises 12 to 18 months in order to have a full data set.

However, he says, this can help reduce the margin of error in DNI readings to about 3% or 4%. And the good news is that, since the Shams debacle, developers are increasingly aware of the need to carry out the exercise.

“That was three or four years ago, when the whole industry was learning on the fly,” he says. “Now DNI knowledge is a lot more elaborate so it is less likely this sort of thing would happen.”

To respond to this article, please write to:

Jason Deign: jdeign@csptoday.com

Or write to the editor:

Rikki Stancich: rstancich@gmail.com

Image credit: Courtesy of Africa