Uptime & downtime

You would be surprised to hear that wind power advocates, long defending their intermittency problem, have turned to the counteroffensive, criticizing the reliability of baseload plants. There are some truly ludicrous claims here. Here's a sampling of the mendacity to which dishonest and deluded wind apologists are resorting to.

Here's Mark Jacobson and Mark Delucchi, in the infamous Scientific American propaganda piece:

[Scientific American] A path to sustainable energy by 2030

A new infrastructure must provide energy on demand at least as reliably as the existing infrastructure. WWS technologies generally suffer less downtime than traditional sources. The average U.S. coal plant is offline 12.5 percent of the year for scheduled and unscheduled maintenance. Modern wind turbines have a down time of less than 2 percent on land and less than 5 percent at sea. Photovoltaic systems are also at less than 2 percent. Moreover, when an individual wind, solar or wave device is down, only a small fraction of production is affected; when a coal, nuclear or natural gas plant goes offline, a large chunk of generation is lost.

And the same in more detail in the accompanying paper, submitted to Energy Policy:

Evaluating the Feasibility of Meeting all Global Energy Needs with Wind, Water, and Solar Power

A WWS infrastructure offers new challenges but also new opportunities with respect to reliably meeting energy demands. On the positive side, WWS technologies generally suffer less downtime than current electric power technologies. For example, the average coal plant in the U.S. from 2000-2004 was down 6.5% of the year for unscheduled maintenance and 6.0% of the year for scheduled maintenance (North American Reliability Corporation, 2009), but modern wind turbines have a down time of only 0-2% over land and 0-5% over the ocean (Dong Energy, et al., 2006, p. 133). Similarly, solar-PV panels have a downtime of around 0-2%. Moreover, there is an important difference between outages of centralized power plants (coal, nuclear, natural gas) and outages of distributed plants (wind, solar, wave): when individual solar panels or wind turbines are down, only a small fraction of electrical production is affected, whereas when a centralized plant is down, a large fraction of the grid is affected.[9]

Here's the US wind industry's PR front, the American Wind Energy Association (AWEA), in one of their propaganda "factsheets":

[AWEA] Wind Power -- Clean AND Reliable

Is wind less “reliable” than conventional generation?

No. Conventional resources occasionally shut down with no notice, and these “forced outages” require operating reserves. For example, a power system that has a 1000 megawatt (MW) nuclear or coal plant will typically keep 1000 MW of other generation available, to be ready to quickly supply electricity if a plant unexpectedly shuts down. The power system can still be operated perfectly reliably in this fashion. Thus, “reliability” is not specific to any single generation facility; rather it is measured on a system-wide basis. Because significant generation reserves are already required to accommodate unexpected changes in electricity supply and demand, in many regions large amounts of wind power can be added to the grid without increasing the total amount of reserves that are needed.

As noted by Jon Brekke, Vice President of Member Services for Great River Energy, a utility that operates in Minnesota and Wisconsin, “Wind energy is a valuable part of our diverse and growing energy portfolio. When partnered with other traditional generation resources, wind energy is an effective way to provide reliable, clean and affordable power to our member cooperatives. Geographic diversity of wind energy helps even out the variability of wind energy in the regional market. In addition, wind farms are typically made up of many individual turbines which reduce the impact of outages. For instance, there are 67 1.5 -MW turbines at our Trimont Wind Farm, so if one is down for maintenance, only 1.5% of the total wind farm's generating capacity is lost.”

They both propone the same claim: that a collection of many wind farms is more reliable than a collection of large baseload plants, because they are smaller and hence an individual outage has a smaller effect. When a single wind turbine goes down, it doesn't knock out 1,000,000 kW of grid power; the damage is lesser.

Additionally, the Jacobson and Delucchi rag claims that wind is "more reliable" than baseload because it has fewer maintenance outages (0-2% onshore, 0-5% offshore). This is a spectular example of cherry-picking: almost all wind outage is due to weather, not turbine maintenance.

As a focus point: here's some data I uncovered last year -- the generating statistics of wind in Germany:

[Capacity Factor (this blog)] Section 17

This is the sum of four major utilities' wind generation, for all of Germany over the month of January 2009. According to wikipedia statistics there could be at most 24 GW capacity, and according to the data there is at least 16 GW represented; I'm not sure exactly how much capacity was online at the time, but it's between 16-24 GW and probably towards the higher figure.

And for comparison, here's a similar capacity of US nuclear power -- 25.06 GW -- NRC's Region I, over the same time period (with the disclaimer that this performance is better than usual, as I will show shortly, because there were no refuelling outages):

This is the sum generation of the 26 NRC Region I reactors, according to power levels from NRC status data, and power capacities (net) from IAEA data. (That is: where a reactor is 1,000 MWe capacity, and recorded as running at 90% power level, I add 900 MWe). I've restricted to this 25 GW subset of the data to make a fair comparison with the 16-24 GW of wind capacity in the German data.

It's spectacular isn't it? First, when Mark Jacobson says wind energy has "less downtime" than traditional sources -- which as you can see is, um, not quite the case -- none of that obvious, obvious downtime you see on the graph counts. It doesn't count because it's not turbine downtime, but weather downtime. The "downtime" Jacobson boasts about is less than 5%, which is "less than traditional sources". Oy vey.

Now the really interesting lie, is that a collection of many wind farms is more reliable than a collection of large baseload plants, because they are smaller and hence an individual outage has a smaller effect.

You can see from the power graphs this is obviously false. But why is it false? One obvious reason is that the uptimes of wind power are much lower than other sources (contrary to the egregious misdirection of Jacobson et. al). But an even more crucial reason is the correlations between plants. Probabilistically, a grid of N power plants -- if they are entirely independent -- will never fail all at once, and their aggregate will act as a very reliable system. When plants go down, they go down at different times (with extremely high probability), and the rest of the system can be relied to back them up. As it happens, none of the four outages (A,B,C,D -- two unplanned) in the nuclear graph coincided, and more broadly they will never all fail at the same time, like wind turbines do. There's little or no correlation. (Actually I suspect it may be better than uncorrelated - they may be anticorrelated, when flexible-schedule maintenance between units can be intentionally staggered, or shifted forward or delayed to counteract conditions at other plants.) Handwavingly, with independent units, the variability (as standard deviation of the probability distribution) will tend to fall as 1/sqrt(N), according to the central limit theorem.

Whereas, for complete and utter contrast, wind power is massively correlated. It is correlated with diurnal cycles, it is correlated with seasonal patterns, it is correlated with weather fronts over thousands of miles of space... completely separate wind farms are all correlated together, rising and failing in synchrony. You clearly see this in action in the German national data -- 20,000 turbines, scattered hundreds of miles apart, all failing at the same time. That's the key: correlation. When my quoted wind advocates say

Moreover, when an individual wind, solar or wave device is down, only a small fraction of production is affected;

(Jacobson & Delucchi) or

In addition, wind farms are typically made up of many individual turbines which reduce the impact of outages. For instance, there are 67 1.5 -MW turbines at our Trimont Wind Farm, so if one is down for maintenance, only 1.5% of the total wind farm's generating capacity is lost.

(AWEA) -- they've failed to understand the whole point of wind intermittency: that they don't fail one a time, but all at once.

I mentioned that the nuclear performance is unusually good, because there are no refuelling outages included. The US nuclear capacity factor is around 91%, although on the short term it ranges from 80-100% (dark blue):

(Same NRC data, now for a 12-month period.) I've added a gray line for the total US electricity consumption for each month (arbitrarily normalized; source is EIA). The key feature missing the January data is refueling outages, which cluster in spring and fall. I'm pretty sure this is deliberate; they are in the months of lowest demand (according to the gray line); a sort of load following. You see dozens of unplanned outages over the year as well, but clearly they are totally uncorrelated, and the collection is infinitely more reliable than wind farms.

To see the individual outages, I've drawn a rudimentary graphic (I don't what it's called) of the 26 Region I reactors and their power levels. The timescale is the same, the year 2009:

So you see individual features like refueling episodes (~1 month, at ~2 year intervals (so half the reactors refuel in 2009), clustered at low seasonal demand), as well as very short outages which in total amount to little.

Update (3/16): for another example of this lie see Sovacool.