Avenal in ascendance: Taking a closer look at the world’s largest silicon thin-film PV power plant

Aug. 18, 2011 - Tom Cheyney - pv-tech.org

One of the largest photovoltaic power plants in the world started officially sending electricity to the grid earlier this month—and hardly anyone seemed to notice. The commissioning of the 45MW (AC) Avenal Solar Generating Facility in rural Kings County, CA, was drowned out among the buzz of First Solar’s gigawatt-scale PR burst around the Agua Caliente, Topaz, Desert Sunlight, and Copper Mountain Solar II installations and SunPower’s imminent kickoff of construction on the California Valley Solar Ranch site. After all, what’s a mere 45MW compared to the nearly 1.8GW represented by those megaprojects?

In Avenal’s case, not only has it joined the cast of the current top 10 largest operational PV plants, the site stands as what may be the biggest silicon thin-film-based solar generating station on the planet. A closer look at Avenal also provides an opportunity to examine two of the solar industry’s trickier metrics: AC:DC conversion ratios and kilowatt-hour output.

The co-owners of Avenal, Eurus Energy America and NRG Solar, didn’t offer a wealth of details in their announcements, but a bit of Googling revealed additional info both on their websites as well as on the site of general contractor Quanta Services.

Avenal is actually three projects in one, all located west of the Kettleman Hills just off Highway 33 in central California--Sun City (20MW), Sand Drag (19MW), and Avenal Park (6MW)—each of which has signed 20-year power purchase agreements with Pacific Gas & Electric. According to Quanta’s project portfolio, the site features Japanese-made Sharp thin-film modules (about 450,000 of the tandem-junction units) in fixed-tilt arrays, Emerson inverters, a 34.5kV substation (designed and built by Quanta subsidiary and project lead, Ryan Co.), and a generating capacity ample to provide “enough energy to power 36,000 homes.”

NRG’s project factsheet adds some other details, such as the combined acreage of the sites (500), while the Eurus site lists the overall project cost as $220 million. A quick math turn shows that maps out to about 11.1 acres and $4.9 million per megawatt AC installed (total balance of systems, etc.).

One other info-bit on the Quanta site triggered my geekish curiosity: The converted DC equivalent of the plant’s AC nameplate rating is listed as 57.7MW. A back-of-the-envelope (BOTE) calculation reveals the DC:AC ratio for Avenal to be about 1.28:1.0, a much higher number than I had seen previously.

My own standard BOTE DC:AC ratio is 1.2:1.0 (or flipped over, 0.8:1.0, AC:DC), a number used publicly by the likes of First Solar, PG&E, and others. Other ratios in use are even tighter.

SolarPlaza’s Tom van der Linden told me that the group uses a 0.87 AC to 1.0 DC metric in its just-published top 10 solar PV power plants report.

Southern California Edison calculates the ratio at roughly 0.9MW AC of power produced for every 1.0MW DC installed on its owned and operated systems when they’re operating at peak generating capacity, according to utility spokesman Gil Alexander.

Since the Avenal plant came online Aug. 5, it’s not surprising it didn’t make the deadline cutoff to be included in SolarPlaza’s brand-new top 10. If it had been listed, the 45MW (AC) site would crack the chart in 6th place, if the 57.7MW (DC) equivalent were used, or it would land in the 10th spot at about 51.5MW (DC), if the Dutch publisher/consultancy’s metric were used. In either case, it means that of the current 10 largest PV generating systems, four employ thin-film technology.

But the top 10 ranking would also be shaken up if First Solar’s 1.2:1.0 ratio were used. The undisputed number one--the 80MW (AC) Sarnia project in southern Ontario, Canada—receives a 92MW (DC) nameplate rating by SolarPlaza, but would jump to 96MW if First’s own ratio were applied. The current holder of the sixth slot—Copper Mountain Solar I (Boulder City)—has its 48MW (AC) converted to 55MW (DC), but would jump to 57.6MW if the CdTe leader’s metric were the norm.

The horse race between Avenal and Copper Mountain gets more complicated when comparing the DC equivalents offered by the project owners/EPC firms. Although the Nevada site is obviously 3MW larger than the California farm in terms of their respective AC ratings, the ratio conversion apparently used by Quanta pushes Avenal past Copper Mountain on the DC side.

I reached out to someone with intimate knowledge of utility-scale PV projects and EPC, who shared some insights with me on the condition that he remain anonymous. Noting that the issue of DC:AC ratios is “a very good (and rarely asked) question,” he explained that “a novice will design so that the DC ratio peaks at the point of maximum sunlight (roughly noon time). Depending primarily on the modules, this will yield a ratio around 1.2.”

“However, if the goal is to maximize energy production (and there is sufficient land), you can design with a higher ratio, thus generating more energy on the shoulders of the parabola but clipping at the peak of production,” he said. “For modules that perform well at lower light conditions (e.g., CdTe thin-film modules), this can be a big and very important difference. Add in the complexities of trackers and other details, and there is further optimization to be had. Again, the details are highly proprietary and few have optimized this.”

As a result of the range of optimization options, in some cases, he told me, the DC:AC ratio can even be quite a bit higher than Avenal’s 1.28:1.0.

But the conversion ratio of the installed modules is just the tip of the iceberg when it comes to how a utility-scale solar power plant is designed and how much power it actually is capable of generating. Is the owner’s goal to “maximize energy, maximize revenue, or maximize power and therefore peak production?” the source asked. Is the project size under any kind of space or land constraint? In the case of the megafarms starting to sprout up in the southwest US and other regions, it rarely is.

This is “very complicated” business, he stressed. “There are many factors that, in the end, must be run through the model. Consider: geographic location, all angles of the modules, module type, ambient temperature, weather patterns, system design parameters, parasitic loads, module type, inverter, etc.”

Which brings us back to a key piece of information for any solar power plant, especially larger ones: the annual kilowatt/megawatt/gigawatt-hour output or energy yield. Missing from any of the respective companies’ documents or websites is the project’s expected MWhr/GWhr metric. Both Eurus’s Mark Anderson and NRG’s Lori Neuman told me that it’s their company’s policy not to release that information.

As it turns out, this information is a matter of public record and exists both in the California Public Utilities Commission’s final advice letter for the project and in the California Energy Commission’s investor-owned utilities contract database.

Both documents show Avenal with a rated capacity of 48MW, or 3MW more than what was actually built. I’ve been unable to find out what happened to those other megawatts, but the companies involved have consistently been reporting the size as 45MW since the project was first announced and began construction last year.

The data I sought, listed under the categories of “minimum expected deliveries (GWh/yr)” and “maximum expected deliveries (GWh/yr)”on the CEC spreadsheet—there seems to be no difference between those two, btw—show the silicon thin-film-powered system anticipated to provide 77GWhr/yr. After recalculating that number based on the actual 45MW size of the plant, the figure shrinks to about 72.2GWhr/yr.

A BOTE formula for estimating kilowatt-hours per year goes as follows: take the installed capacity of the system, say 10MW, and multiply that by the average number of sunny hours per day, say 6. Then take that product and multiply it by the estimated number of sunny days per year, say 300 for well-irradiated parts of California, and you come up with a very rough approximation of the total annual megawatt-hour output of that system; in this case, about 18GWhr/yr.

If we apply that same formula to Avenal, we would overshoot the stated gigawatt-hour total, hitting 81GW. But if you ratchet back the numbers, reducing the sunny days per year to about 267, the new total is right around 72GWhr/yr.

Of course, this BOTE calculation flattens out all the optimizable variables cited by the expert source earlier and doesn’t address the whole AC:DC controversy again. (But then again, I’m an AC kinda guy.)

Earlier I mentioned that the Avenal project principals have claimed the power plant would provide sufficient energy yield to power 36,000 homes. Anyone who’s perused reports about utility-scale solar is familiar with this statistic, but not everyone may realize that the “house count” is nonstandardized and can vary widely, depending on several factors such as load factor, consumer’s consumption pattern, energy efficiency, location and climate, level of insolation, type of PV system, etc.

But one key metric, one that is often misstated or at least left out by project developers/owners, is whether the number of households cited is the total that will be served by a system during peak-demand generating hours or the customers who will have their electricity needs met on an average annualized basis.

SCE’s Gil Alexander helped put this into perspective. He explained that the average Edison residential customer in the company’s 15 Central and Southern California county service area uses about 600KWhr of electricity per month, or 7.2MWh/yr. (PG&E’s average customer uses 6.6MWh/yr, sources say.)

The utility consistently—and conservatively—states that for every megawatt of solar installed (and again, this is for plants owned and operated by Edison), an average of 650 residential customers are served when the system is generating at peak capacity. That same 1MW facility produces close to 1500MWhr/yr, or enough to serve 207 average residential customers, according to the spokesman.

“It’s the old ‘capacity factor’ issue,” he noted. “Maximum generating capacity under perfect conditions versus energy produced over time.” SCE uses a capacity factor of 16-20% in its own model, he added.

Alexander then shared an related anecdote. “A couple of weeks ago I hosted a journalist at our [5MW AC] Porterville solar station. As it happened, we arrived at the site at a point when the inverters said we were producing 98% of the nameplate capacity. This was because the sun was at its peak overhead and there was a slight cloud cover meaning there was a ton of sunlight but the panels weren't as hot as they can get in clear sunlight, [since] very hot panels don't produce power as efficiently.

“So, at that moment, we were producing enough power to meet all the needs of 3,250 average Edison residential customers. However, to state the obvious, that does not mean the site can meet all the needs of 3,250 customers when conditions are not peak.”

Unfortunately, not all project announcements include verbiage that clarifies whether the “number of households served” figure is based on peak-time system performance or annualized energy production. First Solar, for example, has used phrases like “at full capacity” and “during peak demand hours of the day.”

In the case of Avenal, this was unstated though it’s pretty clear from some quick calculations that the “36,000 homes” stat—about 800 customers per megawatt—represents the system’s peak-demand output capability. Given PG&E’s residential kilowatt-hour average, the annualized number of households comes in at 10,930 residences (243 per megawatt); like the similar SCE figures, the yearly amount rounds out to less than a third of the peak number.

Despite the slightly loose use of terminology and numbers-gamesmanship by Team Avenal, they still have bragging rights for awhile, given their overlooked status as the biggest tandem-junction (or any other kind of amorphous silicon) thin-film PV power plant in the world.

But their roost-ruling days may be numbered. Construction on a 73MW project, also using Sharp TFPV panels, is reportedly under way in the Lop Buri province of Thailand, north of Bangkok. Although there have been no recent announcements by NED, Mitsubishi, Sharp, or the other parties involved, the original timeline called for the power plant to be operational by the end of 2011.

If true, the Thai installation would overwhelm Avenal in the installed capacity category and vault into the top tier worldwide, regardless of whether that 73MW is—and I can’t confirm it either way though I suspect it's direct current—an AC or DC nameplate rating.