Euan Mearn’s recent Red Eléctrica de España (REE) post drew my attention to the fact that REE has now begun to show grid data for solar PV and concentrated solar power (CSP) generation separately instead of lumping them together. In this post I use the REE data to review the performance of Spain’s CSP plants and to check among other things whether the claim that they are capable of providing baseload generation, as this 2011 Forbes article claimed, holds up in the light of operating experience. Spain is a good case study because the lion’s share of world CSP capacity (2.3 of 3.4GW in 2013) is installed there.
Gemasolar 20MW CSP plant, Spain. “The molten salt storage tank permits independent electrical generation for up to 15 hours without any solar feed”. Image credit Torresol.
Types of CSP plant
First a brief review of the different types of CSP plant. The Gemasolar plant shown above is a solar tower CSP plant, where dual-axis tracking mirrors reflect solar energy onto a central tower that contains a working fluid, usually molten salt. The salt is heated to temperatures of 300˚C or more and passed through heat exchangers that generate steam to drive turbines. Solar towers are not the most common type of CSP plant but they are easily the most photogenic.
The most common type of CSP plant is the parabolic trough. Here a linear single-axis tracking mirror array concentrates solar energy onto a tube containing a working fluid – again usually molten salt – located along the focal point of the mirrors, and the molten salt is again used to generate steam to drive turbines. The Andasol in plant in Spain is an example:
Detail of single-axis tracking array, Andasol parabolic trough plant
Other types of CSP plant include dish stirling, which uses stand-alone parabolic reflectors to concentrate light onto a receiver and delivers the heated working fluid to a Stirling engine and fresnel reflectors, made of many thin, flat mirror strips that concentrate sunlight onto tubes through which working fluid is pumped. More details are available here.
These different CSP technologies have their advantages and disadvantages but they all have a common feature that distinguishes them from solar PV plants. They convert solar energy into heat, and heat can be stored for re-use. In short, they have a built-in storage battery that allows them to generate electricity when the sun doesn’t shine.
They also have another less-well-publicized common feature. They need an auxiliary source of heat to jump-start them in the morning, and the heat is commonly provided by natural gas. Spain in fact allows CSP plants to generate up to 15% of their electricity from natural gas. A list of Spanish CSP plants is here.
REE June grid data
Now to the REE grid data. REE only started segregating PV and CSP generation in April of this year so a year-long evaluation isn’t possible. Here I review the data for just two months – June 2015, which I have assumed was a typical summer month in Spain, and November up to the time of writing (November 26) which is as close to a winter month as I can get.
Figure 1 is a stacked bar chart showing PV and CSP generation for June, a month during which the sun shone brightly except for a cloudy period before mid-month. CSP generation tracks PV generation but has a somewhat different shape and tends to lag PV by about an hour, presumably because of thermal inertia. The PV capacity factor over this period based on 4.7GW of installed capacity was 25% and the CSP capacity factor based on 2.3GW of installed capacity was 45%. More about capacity factors later.
Figure 1: PV and CSP generation for June 2015, Spain
The CSP plot shows nighttime generation of power from stored heat. However, so does the PV plot – over 600MW of it on June 29 and 30. There is obviously a problem here but I have assumed that it affects only REE’s PV data and that the CSP data are correct.
Figure 2 shows CSP generation segregated into daytime generation and “nighttime generation from storage”. Two points should be noted:
- 1. The plot sums the generation from over 40 CSP plants with built-in storage capacities ranging from zero or near-zero up to 15 hours (although the storage capacity in MWh is never specified).
- I define “nighttime generation from storage” as the CSP power generated between the time the sun went down, as defined by the cessation of PV output, and minimum daily CSP generation, which usually occurs shortly after sunrise.
Calculated using this definition only 19% of Spain’s June CSP generation came from storage and the amount of energy stored was never sufficient to deliver stable power through the night. Power delivery always began to fall off before it picked up again in the morning:
Figure 2: CSP generation for June 2015 segregated into “daytime” and “nighttime from storage”
In addition, no attempt was made to match CSP generation to demand (Figure 3). Daytime generation tops out around 2.2GW, most probably because the 2.3GW of CSP turbines were “maxed-out” at this level, but the working fluid continued to accumulate heat that was released through the turbines later. The blue areas can in fact be visualized as having been scraped off the top of the orange areas:
Figure 3: CSP generation for June 18th through 22, 2015 segregated into “daytime” and “nighttime from storage” and compared with demand. Note that the left and right Y scales are not the same.
REE November grid data
Figure 4 plots PV versus CSP generation for November. The PV capacity factor has fallen to 14% from 25% in June but the CSP capacity factor has fallen to 14% from 45%. Clearly mucho sol is needed for CSP plants to work efficiently. (Note that PV plants generate electricity in the absence of direct sunlight while CSP plants do not):
Figure 4: PV and CSP generation for November 1 through November 26, 2015
Figure 5 plots daytime CSP generation versus nighttime generation from storage for November. The percentage of total generation from storage has decreased to 10% from 19% in June, which when combined with the decrease in capacity factor reveals that Spain’s CSP plants delivered six times less stored power in November than they did in June. On the first three days of the month they delivered no stored power at all:
Figure 5: CSP generation for November 1 through 26, 2015 segregated into “daytime” and “nighttime from storage”
And again no effort was made to match generation to demand, as shown in Figure 6. (The double peak in CSP generation is particularly intriguing. I have no backup evidence but speculate that the first peak may be when they turned the gas off):
Figure 6: CSP generation for November 6 through 9, 2015 segregated into “daytime” and “nighttime from storage” and compared with demand. Note that the left and right Y scales are not the same.
As things stand Spain’s CSP plants are nowhere close to providing baseload generation. But then, they have no incentive to provide it. As I understand it they get paid the same kWh rate regardless of when the power is delivered, which gives them an incentive to produce power as quickly as possible after the sun gets up to minimize working fluid heat loss. If they were paid a large enough premium for nighttime delivery the generation plots would be flatter, but how much flatter is impossible to say.
As to whether the plots could ever be smoothed to the point where Spain’s CSP plants provide baseload generation at roughly the same level year-round, however, the answer is an unqualified “no”. Even if the plants had been able to smooth out diurnal variations so that the same generation level was maintained through the day we would still be left with large day-to-day variations in June and even larger seasonal variations between June and November, as shown in Figure 7:
Figure 7: Average daily CSP generation, June and November, 2015
Smoothing out the June variations so that power could have been delivered at the monthly average of 1,030MW at all times during the month would have required a large amount of storage, and delivering power from June through November at the average June-to-November level of 670MW a prohibitive amount. Figure 8 summarizes the power storage and release requirements that would have been needed to maintain a constant 670MW of baseload generation during June and November. Approximately 260 GWh of storage would have been needed to cover the shortfalls in November alone. Installing this much storage to support less than a gigawatt of baseload generation is clearly not a viable option:
Figure 8: Power to and from storage needed to maintain constant 670MW baseload generation, June and November, 2015
A final question is whether the higher capacity factors of CSP plants means that they are more efficient than PV plants, as is sometimes claimed:
Torresol says that the plant will provide electricity for about 20 hours each day on average, with numerous days in the summer seeing 24-hours of supply. How does that compare with a similar-sized PV plant? The 21.2 MW Photovoltaic Solarpark Calaveron in Spain generates about 40 GWh a year. This smaller 19.9 MW power tower plant will generate about 110 GWh per year.
CSP plants have higher capacity factors than PV plants simply because capacity factors are calculated based on the capacity of the turbines, not the capacity of the mirror array. When the capacity factor is calculated based on the capacity of the mirror array they are actually less efficient than PV plants. Andasol 1 is an example (data from NREL):
Turbine capacity: 50 MW
Generation at 100% CF: 438,000 MWh/year
Actual generation: 158,000 MWh/year
Capacity factor: 36.1%
Mirror area: 510,120 square meters
Solar resource: 2,136 kWh/sq m/year
Mirror array capacity: 124.4 MW
Generation at 100% CF: 1,090,000 MWh/year
Actual generation: 158,000 MWh/year
Capacity factor: 14.5%
If Andasol’s dual-axis-tracking mirrors were dual-axis-tracking PV arrays we would expect a capacity factor of over 20%. By this yardstick the Andasol plant is roughly a third less efficient than a PV plant of the same size.
ADDENDUM: December 2, 2015
Willem Post suggested in comments that it would be helpful to include the graph below in this post. It plot Spain’s CSP generation and demand for November 2015 on the same scale and illustrates how little CSP presently contributes to Spain’s generation mix: