Capacity Credit is a measure of how much electricity any new plant can be depended upon to deliver. In the case of wind turbines, which really cannot be depended upon (in industry-speak, they cannot be “dispatched”), it is typically expressed as how much other generation wind can allow to be shut down. Wind energy proponents are eager to demonstrate as large a Capacity Credit as they can, and use traditional statistical techniques to do so. But in the end their logic just doesn’t make much sense. Wind turbines over large geographical areas can all be becalmed at the same time for extended periods of time, telling me that, without massive storage, no traditional plants can be shut down at all. Capacity Credit must not be confused with Capacity Factor, which is what percentage of its “nameplate” capacity a plant actually generates. If the wind turbines were very reliable it would mean the utility wouldn’t have to maintain large reserves – which cost money and produce emissions. Unfortunately “reliability” in the context of wind turbines is not the same as “reliability” for traditional plants.
You’d think the reliability characteristics for traditional nuclear and fossil fuel plants would be well known and predictable, but I discovered a fairly wide range for values, anywhere from 75 – 100%. Hydro might vary depending on rainfall and of course wind energy would vary even more widely depending on the wind. I suspect the variance has to do with the time window involved. Most grid operators have a “day-ahead” operational planning window, where producers make commitments of how much electricity they will deliver over the next 24 hours. For this window all the traditional technologies are extremely reliable, approaching 100% – the fuel is on-hand, the unit is operational, and only a “forced outage” will cause the electricity to not be available. On the other hand, wind and solar make no commitments and the grid operator is obliged, by political fiat, to accept whatever electricity they produce. For longer periods, i.e. yearly, you would include scheduled downtime, and the resulting values would drop from 100%.
I would think the day-ahead reliability numbers would be the most important, as they cause the operator to have actual units in backup, some of them in “spinning reserve”. Unfortunately, I couldn’t find any good data on the reliability of the different generation technologies for just the day-ahead period.
For wind power Capacity Credit has been controversial. Proponents are eager, of course, to show that wind is reliable and can stand on its own as part of a robust grid. So they apply traditional computational techniques to the question, saying that all technologies are unreliable, just to a different degree. The results of this method produce numbers that decline as the penetration of wind increases. At low penetrations the Capacity Credit is close to the Capacity Factor, decreasing as the penetration increases. At a penetration of 20% it drops to somewhere in the 10% range.
Adopting traditional computational techniques for wind strikes me as invalid, and the ISO’s seem unsure of it as well. Ontario’s IESO used to use a flat 10% until November of 2008 when they decided to use the median output (i.e. in the summer – 16%). By definition, half of the time wind will produce less than that, so this seems nonsensical. More recently IESO papers have demonstrated the need for a much lower number, see the chart on p. 21 and the conclusions on p. 22. California originally used statistical techniques for its early wind projects (like Altamont Pass) and came up with numbers in the 20′s. Then the summer of 2006 showed just how risky this method was. A calm spell in Germany tells me that anything above 0% is wishful thinking.
The oft-sited GE Report was undertaken at NY’s expense, and states that wind energy can be used, in part, as base load. As an example of reality impinging on industry studies, GE says in the summer peak days the grid operators should be able to count on a production of 17% of the rated output. As mentioned above, and as detailed here, assuming this high a value is foolhardy.
All the traditional technologies are dispatchable – they (barring a failure) produce when they are told to do so. Wind, on the other hand, just produces what it can and the grid accepts it. (Well, at least until it overwhelms the demand, as sometimes happens in areas with high penetrations at times of low demand.) This difference is so fundamental that I think a new methodology is needed, but I’m not enough of a statistician to figure out what that might be. Proponents often drop back to the geographical dispersion argument – that the wind must always be blowing somewhere. When presented with the facts they then drop back further to arguing about the need for a new grid that can transport lots of power long distances. What they don’t do is to include the substantial costs of this new grid in their economic justifications, nor discuss the substantial environmental impact such a transmission system would have.
In a backward sort of way, I approach this same topic in my Emissions Savings posts. Where the two topics meet is the importance of forecasting. After all, if you can accurately forecast the wind (and thus the power production) your “unreliability” goes down and your Capacity Credit goes up.
Some references are below, and you can also google “capacity credit” (preferably with the quote marks) to get some flavor of the arguments.
- Watts with Wind by David Robinson examines Ontario’s production and how it could never be considered as base generation.
- NY Times article, on grid challenges and the value of forecasting.
- Milligan/NREL Study, a nice technical overview of the issue.
- Tom Adams Presentation, given to the IESO in an attempt to quantify how much geographical diversity adds to the Capacity Credit.
- DOE Figures, especially table 8.2.
- Paul Gipe’s Paper, a sample of proponent thinking.
- BWEA’s Thoughts, another sample of proponent thinking.
- Jon Boone Paper, with a contrary opinion.