In 2015, PG&E customers received about 97% of Ivanpah’s contracted electrons, which is a massive improvement over its first year.

But this raises a question:
Ivanpah PG&E PPA solar required versus generated in year 1 and year 2
Source: SEC/EIA data Unit 1, Unit 3

What exactly were the engineering challenges at Ivanpah – and why did they take a year to solve? 

To find out, I spoke with engineering experts at NRG, which is the operating partner as one of the three Solar Partners which developed the project, along with BrightSource and Google.

“We encountered the kinds of engineering problems that can really only be seen and solved in a first full-scale deployment,” NRG spokesman David Knox told me.

“And in that first year, an inordinate number of partly cloudy days impacted not only the energy output, but also the plant’s ability to commission and actually fine tune all of its control systems.”

At 377 MW net, Ivanpah is the first-ever utility-scale direct steam solar tower: any similarly novel technology relying on solar would also need a succession of sunny days to diagnose and try out solutions to engineering problems. 

“But now that we have learned the lessons that we have learned, and the industry has learned the lessons that we have learned; others don’t have to,” he added.

“But the first large-scale installation needed to go through those steps.”

Correcting a too-easily-tripped steam sensor

Adjusting the settings for the sensor on a steam drum to prevent tripping too easily was one of the biggest contributions to the increase in generation in the second year, according to Mitchell Samuelian, NRG’s vice president of operation for utility-scale renewable generation.

“If even a wispy little cloud came over in the morning, the plant would trip. We would actually have four or five drum level trips on start-up every morning,” he told me.

Samuelian has an engineering background and worked in traditional thermal and hydropower electricity before coming to NRG, which itself is well-versed in operating traditional thermal plants that use a power block in the same way as a CSP plant does, but they are fueled by gas or coal, not solar.

In both the newer “tower” types of CSP, whether they operate on direct steam like Ivanpah, or on molten salt with energy storage, like Crescent Dunes, the transfer medium is heated by the moving path of sunlight continually reflected off mirrors (heliostats) onto a receiver in a central tower.

The steam drum has to be downsized in a solar tower.

At Ivanpah, sunlight concentrated onto the tower receiver heats water to steam in the steam drum.

At 500 feet up in the tower, the drum had to be smaller than a typical steam drum in a conventional power plant. A steam drum has water in the bottom, and heating the water creates steam in the top. In the Ivanpah plant, that gets piped down the tower to operate the steam turbine below.

“In a gas or coal plant the steam drum is much bigger, so they don’t have the same issues with the level going up and down,” he told me. “Small changes in level don’t cause problems. So it was just the fact that it was up in the tower and it’s hard to put a big huge steam drum up there.”


A sensor reset helped reduce morning startup time down to 25 minutes.

Initially the water level sensors in the steam drum were set like those for a fossil fuel plant, which trips off if the water level raises too much, indicating inadequate steam production to run the turbine.

But in a solar configuration, every passing cloud was tripping it off, turning off the plant unnecessarily, and especially during morning startup.

“So we went through with the engineering and redesign so we could go to a higher level and a lower level when we are operating,” said Samuelian. “The operating range was plus or minus I think three inches. And now I think we’re in the range of plus or minus 11 or 12 inches.”

“And now we get a trip on start-up only every couple of weeks”

Simply resetting the steam drum water level sensors to be a little less sensitive to water levels (we’re talking a few inches here) while not endangering turbine operation was a big part of how the morning startup time was cut from four hours to under 25 minutes:

“In the first couple of months, it was taking us right around three to four hours to startup, and now on a normal sunny day from the time that the sun comes up over the horizon to the time that we actually synchronize the unit is in the 25 minute range,” he said.

Many other small engineering fixes included ongoing improvements in integrating the control system that operates the movement of the mirrors in the solar field with the one operating the power plant itself.

Ivanpah was the world’s first attempt at utility-scale direct steam solar tower CSP. Abengoa built the first direct steam tower CSP in Spain at just 11 MW in 2011, which is pilot sized. Abengoa’s direct steam Khi Solar One came online in South Africa in 2016, but at only 50 MW, compared to Ivanpah’s three towers totaling 377 MW two years ago.

New technology takes time to refine. I asked NRG whether traditional power plants had similar start-up troubles:

“Oh yeah,” said Samuelian. “You see that in all new technology. In fact if you look at early on; I remember in the first number of years they will tell you that the forced outage rate was in the 30 or 40 percent range for that new technology, when they were first using gas turbines to drive generators. and nowadays those things are at between 97 percent to 98 percent availability.”

“It is just like any other technology. It just takes a while to get all the bugs out of it.”

For engineers behind the scenes, improving technical problems is just routine.

But CSP is not like any other technology. Our previous first-of-its-kind energy technologies didn’t start up in the glare of a hostile spotlight from today’s highly politicized media:  

Nobody cared that engineers took years to fix the start-up problems of coal or gas turbines. Solar PV could take care of any start-up buggy-ness in the privacy of space.

The Wall Street Journal, now owned by Rupert Murdoch, is widely quoted with its factually wrong statements about Ivanpah’s generation requirement for PG&E, making it appear that the first direct steam solar tower has failed spectacularly to meet the target.

When I asked the journalist responsible why she ignored the facts in the SEC filing, she said because BrightSource wouldn’t also “go on the record.”

When PPAs are confidential, the parties are liable if they reveal PPAs that are not in the public domain: However, another journalist outed the SEC filing last year so it is in the public domain.

SEC filings are reliable sources; in covering most businesses, the WSJ cites them, because investors need facts.

The SEC filing states the mature year contract quantity of generation required after a four year ramp up is 640,000 MWh for PG&E’s two units.

“The “contract quantity” for each year is expected to be 304,000 MWH for Solar Partners II (Unit 1) and 335,600 MWH for Solar Partners VIII (Unit 3) throughout the delivery term, and the seller must deliver a guaranteed amount of energy in two-year measuring periods.”

Now here is the math for the first two-year measuring period: 2014-15:

“The production guarantee generally is 140% of the contract quantity during the first measuring period after the commercial operation date.”

So to get PG&E’s two-year requirement, multiply 640,000 by 140% = 896,000 MWh.
EIA shows two-year generation from PG&E’s  Unit 1 and Unit 3 as = 723,153 MWh.

Over both years that’s 81% of the contracted quantity.
Ivanpah PG&E PPA solar required versus generated in year 1 and year 2

But most of the shortfall was in the first year: The reason that PG&E petitioned the California Public Utility Commission to allow it to keep the contract was the improvement after these engineering challenges were resolved.

In 2015, PG&E customers received about 97% of Ivanpah’s contracted electrons.