Here are the contestants in this important match. In one corner we have Duke Energy, in the old days “Duke Power,” headquartered in Charlotte, North Carolina. Duke is a $57 billion company, serving 7.4 million customers in six Midwestern and southeastern states. In the other corner is Georgia Power, headquartered in Atlanta, GA and serving 2.5 million customers. Because it is a subsidiary of the Southern Company, there is no comparable market value for it as an independent entity. The Southern Company’s market value as of 1 September was $48.5 billion. But Georgia Power is Southern’s largest sub, so it should be a fair fight.
At the end of August, Duke announced that it would abandon the construction of a planned two unit, 2200 MW nuclear plant called Levy, to be located in Florida, and instead build 700 MW of utility scale solar plants and 50 MW of battery storage in its place. This announcement followed by about a week an earlier one abandoning a similar plant called “Lee” proposed for North Carolina. Both plants were to use Westinghouse reactors. This would have been a bad week for Westinghouse’s shareholders if the company weren’t already bankrupt.
On the 31st of August, Georgia Power announced to the Georgia PSC that it would like to continue the construction of two nuclear units, Vogtle 3 and 4, despite the fact that the price of these units has more than doubled in cost and construction is two or more years behind schedule. And there are no guarantees of ultimate cost or final completion date. And despite the fact that a very similar project in South Carolina had been abandoned by its sponsoring utilities about two weeks earlier.
And so, the scene is set.
To make a determination in this contest of corporate intelligence, we have chosen to look at cost of construction, total price of power to be delivered from the plants, status of construction contractor, fuel supply, environmental risk, and risk of over-building, as well as other factors. Let’s start with the cost of construction.
The cost of utility scale PV plants has fallen dramatically over the last several years. One analytical publication had data for H1 of 2016, indicating that utility scale solar plants cost from 1250 to 1350 per KW at that time. More recent reports tout the arrival of one dollar per watt ($1000 per KW) plants. This would translate into five to six cents per KWh electricity, although the sun regime at the plant site can affect this number up or down. It seems likely that if Duke aggressively pursues a competitive third-party approach to their solar construction program, it should do even better than these numbers. And they can use multiple suppliers to compete against each other for the key component, the solar panels. A January 2015 Wikipedia article listed nine US and 48 non-US providers of this technology, and there are no doubt now many more, especially in China.
The Vogtle cost is unclear. Most estimates have it at $25 billion, although some argue it could go as high as $29 billion. The truth is that no one knows. Georgia Power did originally try to cap the costs by signing a fixed price contract with Westinghouse, the reactor provider and constructor, and a subsidiary of Toshiba, but that has worked out badly, for all parties.
Westinghouse has declared bankruptcy because of the cost and schedule overruns of this and the sister project in South Carolina. Speculation is rampant that Toshiba will do the same. Bechtel has been appointed to take over the construction of Vogtle, but Westinghouse is still the supplier of the key component—the reactor. This particular design, called “AP1000,” is still being designed as it is being built. But no one else can step in at this point and supply a different design, the plant construction is too far along. Besides, there are only ten suppliers of nuclear plants in the world, each with its own proprietary design. Trying to fix the cost of a complex and large piece of engineered metal before it is fully designed is not the way to do a capital project if you have any interest in cost control. Not good for the customer, not good for the supplier.
Cost to customers is perhaps the single most important variable in this analysis. Since the capital cost is unknown for Vogtle, we have to use estimates. Assuming that 25 billion is the final number, and using even very simple assumptions (30-year debt at 4.5%, mortgage amortization, a 50/50 debt to equity financial structure, 10% return on equity, and 85% capacity factor) then the cost to customers of the capital alone comes out at 12+ cents per kwh. And nuclear plants have non-trivial operating costs which are generally acknowledged to add four to six cents per kwh to the cost of making electricity. So the Vogtle electricity leaves the bus bar at north of sixteen cents per KWh.
For the solar plants, Duke should bid out the cost number. Recent pricing has fallen dramatically. Published numbers for utility scale PV bids won in 2016 started at seven cents/Kwh for a contract with 8 Minute Energy and the Los Angeles Department of Water and Power, and by the end of the year had fallen to 2.4 cents for a bid in Dubai, won by a joint venture of Marubeni, the Japanese trading company, and Jinko Solar, one of the larger “A” quality Chinese panel manufacturers. The sun in Florida may not be the same as Dubai, but even adjusting for that, Duke should be able to secure bids in the three to five cents per kwh range. Or, in a couple of years, less.
Vogtle’s construction contractor, as mentioned above, is now Bechtel, an improvement on Westinghouse, but Westinghouse is still a sub-contractor. No one else can supply the reactor vessel. And since this design, the AP1000, is a first of the kind, the performance warranties are important. What happens if something breaks and has to be replaced at the vendors’ cost? What happens if the reactor vessel and associated hardware don’t make the full output? We note again, Westinghouse is bankrupt. How good are those warranties? The original construction contract has been set aside, so they’re gone. Are there new ones? How much would a reasonable person pay for this? How about nothing, which is what they’re worth.
We don’t know who will build Duke’s solar plants, but we do know that solar plant construction using photovoltaic panels is simplicity itself. Ram some posts in the ground, put up the supporting racks, lift the panels out of carboard cartons and screw them into the structures. Connecting the panels is not unlike plugging in a toaster, no electricians required. The warranties are on the panels and the inverters, a relatively simple piece of electrical equipment made by numerous suppliers. Many, many reputable constructors can build these plants, and many already have.
On to fuel supply. Solar panels run on, yes, the sun, so the only issue is how much sun will there be. There is 20 years of NASA data on this important variable. There will be some intermittency, but it is easily manageable, as many battery installations are now proving. These have been established to deal specifically with the cloud caused intermittency, and are both effective and declining rapidly in price. A case in point: in May Tucson Electric announced that it had signed a solar plus storage contract for “significantly less than 4.5 cents per Kwh.”
Uranium supply is substantially more complicated. We won’t go through all the steps necessary to “enrich” the uranium, boosting the U-235 concentration in the fuel rods to 30%, the minimum amount needed to support a fissile reaction, not the naturally occurring 0.72%. The enrichment process is costly, and adds operating cost to a nuclear plant of 3.9 cents per Kwh—just for the fuel. The sun is, of course, free.
Some justify nuclear generation as a “clean” option, and it is true that it has no CO2 emissions. Neither does a solar plant. Nuclear plants use steam and steam turbines to produce electricity, and thus need highly purified water. This is a cost, and the purification process results in waste that much be handled. So that part is surely not clean. Solar plants have no inherent water use. Finally, the mastodon in the control room is the spent fuel, which has highly radioactive materials in it—strontium, cesium and plutonium—and for which there is not yet a reasonable solution. And this despite government work on a spent fuel repository that started in the Carter administration, and is hardly any further along than when it was begun, 39 years ago.
Ok, last point. Utilities in the US have found in the last ten years that demand for electricity has not grown. In fact, it has fallen slightly. Total generation from utility scale facilities in 2007 was 4,156,745 thousand megawatt hours; in 2016 that number was 4,079,079. Energy efficiency (think LED bulbs, LEED buildings, and better heat pumps) and the recession are both causes, but what has always been difficult to predict is now even harder. Hence a cautious utility will not make commitments that require one and a half or two percent growth per year to justify. The nice thing about solar plants is that they are essentially modular, there is little economy of scale in building them; you can build ten megawatts or a hundred megawatts for about the same cost per MW. The minimum nuke plants size seems to be upwards of 2000 MW.
No power technology is perfect. Solar plants don’t work well at night, or when clouds pass over the sun. But they are so much simpler and cheaper and more reliable, and they have so little environmental impact and construction risk, that the judgment of what to build should be a very simple intelligence test for a utility. It is almost impossible to construct a scenario of the future where expensive nuclear plants making 16 cent electricity will prove to be a better investment than photovoltaic solar plants making 5 cent electricity. Duke wins the intelligence contest hands down, and Georgia Power will have to go back to utility school for remedial math. Maybe one of their instructors can assign this article, or any of the hundreds of others which make the same points, as homework.