Concentrated solar power in the USA: a performance review
A review of concentrated solar power (CSP) plants operating in the US reveals that they are costly, heavily-subsidized, generally performing below expectations and no more efficient than utility-scale PV plants. The need to jump-start them in the morning can also require the burning of substantial quantities of natural gas. And although CSP’s sole advantage over PV is that it can store energy for re-use only one of the plants considered has built-in storage capacity. As discussed in the earlier concentrated solar power in Spain post , however, it is unlikely that enough storage could be installed at a CSP plant to provide more than short-term load-following capability when the sun is not shining. (Inset: Ivanpah Unit 2 tower catches fire, May 2016).
This review originated from a comment posted by correspondent “Thinks Too Much” (T2M) on the Blowout Week 172 thread which bewailed the lack of publicity being given to the poor performance of the Crescent Dunes CSP plant. After further exchanges T2M sent me a copy of a spreadsheet he had painstakingly constructed from the EIA’s Electricity Browser monthly data, which, supplemented by Wikipedia data on the Genesis plant I have used to develop the data presented here. So a thank you and a hat tip to T2M.
The locations of the six CSP plants reviewed (Mojave, Solana, Genesis and the three units at Ivanpah – Crescent Dunes is discussed later) are shown in Figure 1. Installed capacities are Mojave 250 MWe, Solana 250 MWe, Genesis 250MWe and Ivanpah 392 MWe (126 + 126 + 133). Nameplate capacities (MWp) are about 10% higher. Only the Solana plant has storage capability (reported to be 1.68GWh), but no details are available on its performance. Mojave, Genesis and Solana are “parabolic trough” plants and Ivanpah and Crescent Dunes “solar tower” plants. Additional details on CSP plant design are given in the “concentrated solar power in Spain” post post linked to in the introduction.
Figure 1: Plant location map
All plants except Crescent Dunes have monthly production data for the two-year period from January 2015 through December 2016. Over this two-year period Solana generated 1,363GWh, Ivanpah 1,355GWh, Mojave 1,128GWh and Genesis 1,246GWh, for a total of 5,092GWh. This represents less than 1% of the electricity consumed in Arizona, Nevada and Southern California in 2015 and 2016.
A plot of monthly generation from the plants is not very instructive, so we begin instead with a plot of capacity factors. Figure 2 shows average capacity factors by month. Solana leads with an average of 31.1%, followed by Genesis with 28.4%, Mojave with 25.7% and Ivanpah with 19.7%. The weighted average capacity factor for all four plants is 25.4% (calculated using MWe; calculated using MWp it’s around 23%). This, however, is not significantly higher than the capacity factors achieved at conventional utility-scale PV plants in the Southwest US. According to the EIA data I reviewed in solar capacity factors in the US, which yielded values of 28.7% in California, 27.0% in Arizona and 26.7% in Nevada, it is in fact lower:
Figure 2: Monthly capacity factors
A question that arises here is why CSP plants located in the same desert environment don’t give more consistent results. The most likely reason is malfunctions in plant operation, with the Ivanpah plant the most seriously affected. Figure 3 plots capacity factors for the three Ivanpah units:
Figure 3: Monthly capacity factors for Ivanpah Units 1, 2 and 3
The three Ivanpah units cover an area of only about 10sq km, so there is no meteorological reason why any one unit should outperform any other – so long as the units are working properly. But they generate comparable amounts of electricity only about half the time. There are large discrepancies in early 2015 and also between April and June 2016 (partially but not entirely explained by the Unit 2 tower catching fire in May, a result of misaligned mirrors). Other features of Figure 3 are also not credible, such as higher solar generation in February 2016 than in May 2015. Seasonal variations are less pronounced and more erratic than one would expect from a properly-functioning solar plant, and the capacity factors (21.1% for Unit 1, 17.9% for Unit 2 and 20.2% for Unit 3) also seem implausibly low. The implication is that the Ivanpah plant is not working as planned, with problems both in the solar side of the operation and probably also in the power generation circuit, which is a complicated system that uses heat exchangers to produce the steam that drives the generators from the molten salt.
Problems with Ivanpah operations other than the May 2016 fire have also been reported by the media:
Wired Magazine: Ivanpah initially struggled to fulfill its electricity contract, and it would have had to shut down if the California Public Utilities Commission didn’t throw it a bone this past March, approving without discussion agreements that would give the owners of the plant, NRG Energy, BrightSource Energy and Alphabet’s Google, up to a year to work out the problems.
Wikipedia: In November 2014, Associated Press reported that the plant was producing only “about half of its expected annual output”. The California Energy Commission issued a statement blaming this on “clouds, jet contrails and weather”. Performance improved considerably in 2015 — to about 650 GWh, but ownership partner NRG Energy said in its November quarterly report that Ivanpah would likely not meet its contractual obligations to provide power to PG&E during the year, raising the risk of default on its Power Purchase Agreement.
Greentechmedia: The (Ivanpah) plant….. kicked off commercial operation at the tail end of December 2013, and for the eight-month period from January through August, its three units generated 254,263 megawatt-hours of electricity, according to U.S. Energy Information Administration data. That’s roughly one-quarter of the annual 1 million-plus megawatt-hours that had been anticipated. Output did pick up in the typically sunny months of May, June, July and August, as one might expect, with 189,156 MWh generated in that four-month period. But even that higher production rate would translate to annual electricity output of less than 600,000 MWh, at least 40 percent below target.
Even Solana, the best-performing of the plants, had its problems:
Phoenix New Times: (Solana) was knocked out by a microburst for a few days in late July and won’t generate power normally for months, a new report reveals. The severe problem comes on top of generally poor performance from the $2 billion project over the past two years. As New Times reported in November 2014, in its first year the plant produced only about two-thirds of the power that its former owner, Spain’s Abengoa Solar, said it would. The company and Arizona Public Service, which is contracted to buy the electricity the plant generates, said at the time that performance would improve. Publicly available production figures reviewed by New Times this week showed that Solana did generate more electricity in its second year but is still well below its advertised potential. The plant also did worse in the second quarter of 2016 than it did in the same period in 2015, the numbers show. And considering the new report on the July microburst, the plant’s third-quarter results for this year — which haven’t been released yet — are likely to be abysmal.
(Note that Figure 2 confirms a large drop in capacity factor between July and August 2016.)
No specific malfunctions have been reported at Mojave or Genesis, but the fact that the capacity factors at these plants are lower than at Solana suggest that they also had their share of them.
Next on the agenda comes natural gas. The Genesis, Solana and Ivanpah plants (but not Mojave) need to burn it to get the plant warmed up in the morning. Again this is a particular problem at Ivanpah:
Wikipedia: The plant requires burning natural gas each morning to get the plant started. The Wall Street Journal reported: “Instead of ramping up the plant each day before sunrise by burning one hour’s worth of natural gas to generate steam, Ivanpah needs more than four times that much.” On August 27, 2014, the State of California approved Ivanpah to increase its annual natural gas consumption from 328 million cubic feet of natural gas, as previously approved, to 525 million cubic feet. In 2014, the plant burned 867,740 million BTU of natural gas emitting 46,084 metric tons of carbon dioxide, which is nearly twice the pollution threshold at which power plants and factories in California are required to participate in the state’s cap and trade program to reduce carbon emissions.
How much natural gas is actually consumed in the warm-up process? According to T2M’s spreadsheet Ivanpah consumed 1.29 trillion btu in 2016. If this much natural gas had been consumed in a typical CCGT plant (heat factor 7,650 btu/kWh according to EIA) it would have generated 169GW, almost a quarter of the 703GWh of solar electricity Ivanpah generated in that year.
And as shown in Figure 4 there is a fairly strong correlation between the amount of gas Ivanpah burns and the amount of solar generated (R^2 = 0.51). Clearly the more gas the plant burns in the morning the more solar energy it generates later in the day. (Although it’s only fair to note, as the WSJ reports, that Ivanpah is a particularly bad example. As far as I have been able to determine Genesis and Solana burn significantly less gas.)
Figure 4: Natural gas consumption vs. solar generation in 2016, Ivanpah Units 1, 2 and 3, monthly data
Last but one on the agenda is the question of project costs. Based on data from a number of sources, not all of which are necessarily reliable, I have put together the following table. It contains Crescent Dunes for completeness:
The five listed plants, which between them generate less than 1% of the electricity consumed in Arizona, Nevada and Southern California, cost over $8 billion to construct, and over 70% of this cost was covered by federal loan guarantees. All of the projects were also eligible for a 30% federal tax credit. With these generous subsidies and a bit of creative wheeling and dealing it might well have been possible for the developers to complete construction without forking out any of their own money at all.
Of particular interest is the ~$6,430/kW installed cost, which is in the same range as the 3.2GW Hinkley Point C nuclear plant. According to NREL’s 2015 cost estimates it also exceeds the cost of installing the same amount of utility-scale PV capacity by a factor of over three.
Another consideration is electricity sales price. Electricity from the plants is sold to various Southwest US utilities at cents/kWh rates and reliable data are again hard to come by, but the following numbers indicate a range of between 12 and 20c/kWh, or $120-200/MWh:
- CSP unspecified: 15-20c/kWh
- Ivanpah: 12c/kWh
- Ivanpah: 20c/kWh
- Crescent Dunes: 13.5c/kWh
- Southwest US: 13-17c/kWh
- Solana: 14c/kWh
These rates equal or exceed the all-sector rates that local utilities charge in Arizona and Nevada, which according to EIA data are presently 9.62c/kWh and 8.02c/kWh respectively. After addition of transmission charges, administrative charges, taxes etc. Arizona and Nevada utilities will therefore lose money on each kWh of CSP energy they buy. With the all-sector rate at 15.02 c/kWh California utilities will probably lose money too, although not so much. And ultimately the consumer will finish up paying.
Last on the agenda is Crescent Dunes, the project that gave birth to this post. As shown in Figure 5, Crescent Dunes started operations in October 2015 and took its time ramping up, but by the late summer of 2016 it was achieving respectable capacity factors of between 30 and 40%. But then in early October a leak developed in the molten salt circuit and the plant was shut down, and it has stayed shut down in the five months since (probably now for six months. On March 2 of this year it was expected that it would be “another few weeks” before operations recommenced. But as of the time of writing there are no reports of the plant restarting, so presumably it’s still down):
Figure 5: Monthly capacity factors since startup, Crescent Dunes
Now there’s nothing unusual about a power plant shutting down, but it’s not often that a “low-tech maintenance issue” shuts one down for six months:
PV Times, March 2, 2017: “We expect to be back online in a few weeks,” CEO Kevin B. Smith said. A hot salt tank issue “took a while to get it fixed, but it’s a pretty low-tech issue,” Smith said …… I understand you guys have got to figure out what’s going on, but you just seem so infatuated with this hot salt tank issue. It’s a maintenance issue ….”
One has to wonder how long a real breakdown might shut the plant down for.