Atoms for California

Apr 1, 2015

A Cascade of Errors in the CEC’s Estimates of the Cost of New Nuclear, part #2: Understanding CoG, and a Preliminary Correction

If this is the first post you’ve seen in my series on the California’s Energy Commission’s Cost of Generation model (CoG), you may want to start at the beginning. Here are all the posts for easy navigation:


The goal of my analysis is to deconstruct the CEC’s estimate item-by-item, checking their sources and math, making corrections where needed, and observing the result of those corrections. The final effect will be compiled into a “waterfall chart” showing the cumulative contribution of each error or faulty assumption to the total result. A good example of a waterfall chart is from Georgia Power’s reports to the Georgia Public Service Commission on the construction progress at Vogtle 3 & 4. The chart summarizes the increases customer benefits and cost savings accrued in the time since those two AP 1000’s were originally approved by the regulator:

image

The idea of a waterfall chart is that it shows you how different numbers add up together (or subtract, as will be primarily the case in my posts) to generate a final result. As another example, let’s consider how the each of the different components of the cost of an AP1000 per the CEC add together to generate a total result.

For the sake of simplicity, I will limit my posts in this series solely to a discussion of the results under financing assumptions associated with investor-owned utilities (IOUs). IOUs serve roughly 75% of electricity demand in California. I could look at the analysis under publicly-owned utility (POU) financing assumptions, but that would make the pro-nuclear argument too easy, as POUs have access to substantially cheaper financing than IOUs. Since nuclear power is a notoriously capital-intensive method of electricity generation, cheaper financing gives it a greater boost than to its competitors. There is a third financing option in CoG, that of the merchant generator (non-utility). While many merchants generators purchased existing nuclear capacity during deregulation, there is no historical instance of a merchant generator constructing a new nuclear power plant in the United States. That would likely require a power purchase agreement and serious, non-discriminatory climate policy (climate policy that doesn’t discriminate among technologies for GHG abatement). So, let’s just focus on IOUs.

Here is what CoG says an IOU-owned AP1000 would cost if construction had begun in 2009 and it was put into service in 2018. I’ve taken the liberty of adjusting the numbers for inflation to 2014 dollars. You can click here for a larger image. Unfortunately, tumblr seems to compress all my images, so I’ve uploaded them to imgur for closer inspection.

image

$256 per MWh?! How is that possible?! For comparison, the 2009 version of CoG estimates that a combined-cycle gas turbine (CCGT) natural gas plant with IOU ownership entering service in 2018 would cost $138/MWh. In other words, the CEC staff found that nuclear power is 86% more expensive than an equivalent natural gas power plant. That’s really strange, because at the time the model was prepared, the U.S. Nuclear Regulatory Commission had 27 applications for new reactors either in review or in pre-application discussions. I think intellectual honesty should have required the CEC staff to apply a gut check to CoG’s results. If nuclear power were truly so expensive, why would so many utilities and public utility commissions in other states be wasting their time and ratepayer money with very expensive NRC applications?

Even still, the triple digit result for the cost of a CCGT plant is absurd. I haven’t come across any analysis that puts the cost that high. For comparison, the average wholesale power price in the CAISO territory was about $35/MWh in 2009, or $38/MWh in dollars-inflation adjusted to 2014. In general, the wholesale power price represents the variable operating cost (i.e. fuel) of the marginal generator (nearly exclusively natural gas in California at any given hour). Even if you assume that all the other costs are recovered by means other than wholesale power markets (e.g., ancillary service markets, bilateral capacity contracts with utilities), a wholesale power price of $38 does not come anywhere close to the levelized fuel cost that CoG estimates for a CCGT, $95/MWh.

I tried checking the latest version of CoG (from 2014), but I found basically the same results. I think the explanation is that the natural gas fuel price forecast simply assumes some rate of exponential growth well above the rate of inflation for the infinite future. Obviously, the events of 2009 revealed this to be a laughably false assumption. But I’m not hear to defend natural gas. The natural gas industry has enough highly paid lobbyists active in California to do that. I just want to make sure there isn’t some underlying flaw in CoG that precludes its use as a reasonable estimator of LCOE. As we shall see, there are quite a few minor and major errors here and there. The extent to which they affect every technology is not something I have the free time or energy to explore for now, but a later post in this series will address these matters.

A Preliminary Correction

The very first error in CoG that stuck out to me when I first began looking at in 2013 was the listed gross electrical capacity for the AP1000: “960 MW.” 960 MW? There is no citation for where this number comes from! The name of the AP1000 is deliberately intended to advertise that the net electrical capacity of the reactor is at least 1000 MW (this will vary with the choice of cooling technology). A more accurate capacity rating should have been readily available to CEC staff and their grossly overpaid consultants in 2009. In the “Final Safety Evaluation Report Related to Certification of the AP1000 Standard Design” (NUREG-1793), dated September 2004 (!!!), the staff of the NRC note in the introduction: “The AP1000 nuclear reactor design is a pressurized water reactor with a power rating of 3415 megawatts thermal (MWt) and an electrical output of at least 1000 megawatts electric (MWe).” However, CoG is structured in such a manner that the selection of gross capacity does not affect the $/MWh result. The model scales linearly with the choice of gross capacity.

The next error I noticed was in the item for “plant-side uses & losses” (My dad refers to these as “house loads,” and in other instances I have seen the term “parasitic load” used.) This refers to the electricity used by the plant itself for lighting, office equipment, running pumps in the plant, and so on. CoG assigns 2.2% of gross capacity to handle house loads, meaning 97.8% of the remaining gross capacity is available to export to the grid. This turns out to be one of the very few generous but erroneous assumptions that CoG makes about the AP1000. IAEA’s Power Reactor Information System shows that the four AP1000 reactors currently under construction in the United States all are rated at 1250 MW gross and 1117 MW net. This implies that a house load factor of 8.9%.

Because of the well-established economies of scale arising in construction and operation of multi-reactor plants, a more typical case will be 2500 MW gross and 2234 MW net for a dual-reactor plant. While the choice of gross capacity does not have a direct bearing on the calculation of the cost estimate, the degree of difference between net and gross capacity is important, as the capital and fixed O&M costs in CoG are denominated in gross capacity, while the plant’s revenue is a product of the net capacity. This distinction is important to be aware of because estimates of capital and fixed operating costs are much more frequently denominated in terms of net capacity than gross capacity. I will be employing this correction in future posts.