Amory Lovins: Renewable Baseload Power

Grist features a running debate between Amory Lovins and commenters on his post, "Stewart Brand’s nuclear enthusiasm falls short on facts and logic." (Brand's talk on his new book Whole Earth Discipline is well worth a listen at the Long Now website.)

One of Lovins' arguments that caught my attention was his discussion of the "baseload myth," which asserts that solar or wind are too intermittent to provide reliable baseload electricity.

From Lovins' paper "Four Nuclear Myths" (pdf):

The manifest need for some amount of steady, reliable power is met by generating plants collectively, not individually. That is, reliability is a statistical attribute of all the plants on the grid combined. If steady 24/7 operation or operation at any desired moment were instead a required capability of each individual power plant, then the grid couldn’t meet modern needs, because no kind of power plant is perfectly reliable. For example, in the U.S. during 2003–07, coal capacity was shut down an average of 12.3% of the time (4.2% without warning); nuclear, 10.6% (2.5%); gas-fired, 11.8% (2.8%). Worldwide through 2008, nuclear units were unexpectedly unable to produce 6.4% of their energy output. This inherent intermittency of nuclear and fossil-fueled power plants requires many different plants to back each other up through the grid. This has been utility operators’ strategy for reliable supply throughout the industry’s history.

Modern solar and wind power are more technically reliable than coal and nuclear plants; their technical failure rates are typically around 1–2%. However, they are also variable resources because their output depends on local weather, forecastable days in advance with fair accuracy and an hour ahead with impressive precision. But their inherent variability can be managed by proper resource choice, siting, and operation. Weather affects different renewable resources differently; for example, storms are good for small hydro and often for windpower, while flat calm weather is bad for them but good for solar power. Weather is also different in different places: across a few hundred miles, windpower is scarcely correlated, so weather risks can be diversified. A Stanford study found that properly interconnecting at least ten windfarms can enable an average of one-third of their output to provide firm baseload power (pdf). Similarly, within each of the three power pools from Texas to the Canadian border, combining uncorrelated windfarm sites can reduce required wind capacity by more than half for the same firm output, thereby yielding fewer needed turbines, far fewer zero-output hours, and easier integration (pdf).

A broader assessment of reliability tends not to favor nuclear power. Of all 132 U.S. nuclear plants built—just over half of the 253 originally ordered—21% were permanently and prematurely closed due to reliability or cost problems. Another 27% have completely failed for a year or more at least once. The surviving U.S. nuclear plants have lately averaged ~90% of their full-load full-time potential—a major improvement for which the industry deserves much credit—but they are still not fully dependable. Even reliably-running nuclear plants must shut down, on average, for ~39 days every ~17 months for refueling and maintenance. Unexpected failures occur too, shutting down upwards of a billion watts in milliseconds, often for weeks to months. Solar cells and windpower don’t fail so ungracefully. ...

To cope with nuclear or fossil-fueled plants’ large-scale intermittency, utilities must install a ~15–20% “reserve margin” of extra capacity, some of which must be continuously fueled, spinning ready for instant use. This is as much a cost of “firming and integration” as is the corresponding cost for firming and integrating windpower or photovoltaic power so it’s dispatchable at any time. Such costs should be properly counted and compared for all generating resources. Such a comparison generally favors a diversified portfolio of many small units that fail at different times, for different reasons, and probably only a few at a time: diversity provides reliability even if individual units are not so dependable.