by Vaclav Smil
In reality, turbine configurations in large wind farms include single strings (sometime with gaps), parallel strings (often close to a perfect grid spacing), multiple strings that are not uniformly oriented, and clusters laid out in irregular fashion, dictated by terrain or microscale differences in wind speed (fig. 3.4). Denholm and co-workers (2009) studied all of these configurations belonging to 172 existing and proposed US large-scale wind projects (with installed capacities greater than 20 MW) in all windy regions of the nation, and their conclusions offer a representative quantification of both typical and extreme land claims of modern wind farms.
Figure 3.4
Layout of a wind farm: 100 1.5-MW turbines in Shiloh, Solano County, California. © Proehl Studios/Corbis.
Denholm and co-workers (2009) used the land total listed in project applications or associated documentation and found that the average total land claim was 34 ± 22 ha/MW, for a power density of 3 ± 1.7 W;/m2 (with the extreme values more than 15 and less than 0.5 W;/m2). Plotting power density as a function of wind farm size does not show any pronounced correlation: most projects with an overall capacity of less than 100 MW have densities of 2-6 W;/m2. For larger projects (more than 300 MW) the density declines a bit, but its spread narrows to 2-4 W;/m2. But all wind power densities cited so far exaggerate the real performance because all were calculated using rated capacities that were not corrected for prevailing capacity factors calculated by using actual annual electricity-generation totals. These rates vary not only among sites but also show significant interannual fluctuations for the same site.
For many early large-scale wind projects of the late 1980s and 1990s, the capacity factors were well below 20%. A detailed examination of the actual record of wind generation in the EU, which has the world's largest concentration of wind power, shows that during the five years between 2003 and 2007, the average capacity factor was less than 21% (Boccard 2009). In 2010 the nationwide means were 25% for the UK and 24% for Spain and Denmark, but only 20% for France and 15% for Germany. In the United States, better siting and better turbine designs have resulted in noticeable longterm gains in average capacity factors. An analysis of 94% of all US projects built between 1983 and 2010 shows the average load factor rising from 25% in 1999 to 33% in 2008, then dropping to about 30% in 2009 and 2010 before rebounding to 33% in 2011 (Wiser and Bolinger 2012).
A year later Wiser and Bolinger (2013) noted that the rate for 2006-2012 (32.1%) was higher than for 2000-2005 (30.3%), but that the trend had not been either as significant or as consistent as expected, and that the 2012 rate was below the peak achieved in 2008, and, most important, the average capacity factors for projects built after 2005 had been stagnant. The explanation is simple: while better turbine designs boosted capacity factors, the locations of many new projects in less consistently windy areas tended to lower it. In 2012 the average wind resource at the height of 80 m was 15% lower than among the turbines built in 1998-1999.
This seemingly irrational location choice makes sense because lower resource quality occurs in locations closer to major markets and readily connected to existing transmission lines-and an inferior capacity factor is compensated for by the savings on high-voltage transmission lines that would be needed to bring electricity from more windy, but more distant, locations. The same consideration has kept Germany's capacity factors quite low. The country has EU's highest wind capacity (about 30% of the total in 2012), much of it in only moderately windy areas, and during the first six months of 2013 German wind turbines had a capacity factor of only 16%, ranging from 11.8% in May to 21.7% in January (Chabot 2013). This poor European performance has been a major reason why all but one of the wind projects rated by Standard & Poor's have fallen from investment grade to speculative grade over time (Standard & Poor's Rating Services 2012).
Taking the most recent US mean of 32% would thus be a proper correction factor for American capacities and would imply-using the previously established average power density of America's large wind farms of 3 ± 1.7 W;/m2-that the power density of the country's wind-driven electricity generation is only 0.96 ± 0.54 We/m2. McDonald and co-workers (2009) used a slightly higher capacity factor of 35% when calculating their range of 1.41.7 We/m2 for the least and the most compact US projects. For the EU, the actual power densities should be calculated with an average capacity of factor of only about 20% (25% for the windier UK and Spain), lowering the rate to just 0.6-0.75 We/m2. This is an order of magnitude lower than the power densities of solar electricity generation and (as I show later in this chapter) only two to three times as high as the densities of the most productive liquid biofuel industries.
But the comparison must be qualified because of some fundamental differences in the nature of occupancy of the claimed land and the degree of its permanence. Obviously, crops or tree plantations do not allow any other concurrent land use, nor is the area covered by closely spaced PV cells suitable for any other uses. And while the strips between PV arrays in large solar parks are theoretically available for grazing, the coexistence of PV modules and sheep will not be a common occurrence. In contrast, on a large wind farm the land completely excluded from other uses is limited to that covered by turbine pads, transmission infrastructures (including substations), permanent access roads, and service buildings.
The National Renewable Energy Laboratory's wind farm area calculator assumes that this permanent footprint amounts to 0.25 acres per megawatt, that is, 1,000 m2/MW or 1,000 W/m2 (NREL 2013). But according to Denholm and co-workers (2009), direct land claims (permanent and temporary during installations and repairs) average 1 ha/MW, of which 0.7 ha/MW reflects temporary claims during construction and 0.3 ha/MW is permanent land occupation. This last rate translates to a power density of 333 W/ m2, or three times as land-intensive as assumed by the NREL's area calculator, but even so it would be less than 1% (3,000 m2/340,000 m2 = 0.88%) of the total area claimed by an average large American wind farm.
Consequently, 99% of land could be devoted to a variety of agricultural (annual or permanent crops), horticultural (flower beds), or silvicultural (tree and shrub nurseries, Christmas tree plantations) uses, or it could be grazed by domestic animals. Unlike in the case of biofuels or hydro energy, the low power densities of wind-driven electricity generation are of concern not because of relatively large areas of land transformed by energy conversion but because they indicate the spatial limits of wind exploitation: we cannot increase the number of vertical-axis turbines as a way to maximize wind-powered electricity generation within a given area.
This means that in smaller-sized countries with limited wind resources it would not take very long before the continuing exploitation of wind had to move to locations of lower wind quality (lower average speed, lower persistence). But in nations with large territories, many high-quality wind sites that are far away from major urban and industrial areas will remain unexploited until the development of requisite transmission lines connects those relatively high-power-density locations with large electricity markets. In the United States, that would mean building more than a score of longdistance lines on a semicontinental scale, and multiple lines connecting the windy Great Plains with the coasts cannot be completed in a matter of years (Smil 2011).
Noise Effects
Finally, there is one spatial consideration besides the requisite machine spacing that limits the location of large wind turbines: they can overtop crops or grazing animals, but they cannot be sited right next to permanent settlements (as even very large solar parks can) because of the turbine noise. The mechanical component of the noise (from gearboxes) has been reduced by better design, so the concern is about the aerodynamic (whooshing) noise produced by the flow of air around the blades and tall towers. Typical maximum noise levels for human exposure are set at 40 dBA sound intensity (integrated over the 20-20,000 Hz band) with a wind speed of 8 m/s at 10 m height. This usually means buffer zones of about 350 m for large wind farms. At that distance the noise will be 35-45 dB, while a busy offi
ce may register at 60 dB and a quiet bedroom at 20 dB.
Many jurisdictions do not specify decibel levels as they have simply set standard buffers. Scottish planning policy calls for 2 km between wind farms and the edge of cities and villages; Wales has a 500-m recommendation (Regen 2012). But in 2013, after Milton Keynes Borough Council wanted to set a 1.2-km buffer, the UK's High Court of Justice ruled that local councils could not impose their own buffer zones for new wind projects (Royal Courts of Justice 2013). The United States has some 25,000 zoning jurisdictions, with most setback rules set at the country level (USDOE 2011b) and with some buffers being just 150 m from a property line containing a dwelling while others require more than 1 km. A 63-MW farm (21 3-MW Vestas turbines with a 90-m rotor diameter) with regular five by ten-diameter spacing would require 486 ha (about 13 W,/m2), but adding a 1-km noise buffer zone on all sides would increase the claim to 1,786 ha, 3.7 times the spacing claim, and lower the power density to 3.5 W;/m2.
What is tolerable and what is objectionable? Large weather variations and terrain specifics mean that long-term measurements of noise from wind turbines are needed before actual impacts can be assessed. Such measurements may show that the urban background noise level (above all from road traffic) may dominate at distances beyond 300 m from a wind turbine (Bjorkman 2004). But this is hardly universal. As with other controversial topics involving complex mental and physical human responses to environmental disturbances, the findings of wind turbine noise studies span a wide range of outcomes supporting an equally wide range of views, from those who see a new, full-blown "wind turbine syndrome" of health impacts (Pierpont 2009) to those who dismiss the concern as nothing but "a prime example of a contemporary psychogenic illness" (Chapman 2012, 1).
Such a dismissal is not supported by a great deal of accumulating evidence. There is no need to demonstrate any direct physical effect of noise: noise annoyance can act as a mediator that leads to sleep disturbance and mental distress (Bakker et al. 2012). And there are objective assessments of the effect. Nissenbaum, Aramini, and Harming (2012) studied the effects of wind turbine noise on sleep and health at two sites in Maine and demonstrated that participants living within 1.4 km of a wind farm had worse sleep, were sleepier during the day, and had worse mental component scores than those living farther away. The overall evidence is complex (Roberts and Roberts 2013), but there is no doubt that living too close to large industrial wind turbines can harm human health and that typical symptoms are stress disorder-type diseases acting by indirect pathways (Jeffery, Krogh, and Horner 2013). Dealing with this reality would further dilute the average power density of wind generation from turbines located in more populous regions, including rural areas in many countries in Asia.
At the same time, there are some intriguing possibilities of boosting the typical power densities of wind-driven electricity generation. One of them might be to move away from using the horizontal-axis turbines that dominate the global market to using vertical-axis machines, whose denser spacing can raise the output per unit area. Experimental field tests with six 10-m-tall and 1.2-m-diameter vertical-axis wind turbines with 4.1-m span airfoil blades (a modified version of a commercial model by Windspire Energy) that were spaced 1.65 diameters apart with a footprint of 48.6 m2 demonstrated opportunities for raising average power densities by extracting energy from adjacent wakes and from above the wind farm (Dabiri 2011). With three turbines rotating around their central shaft clockwise and the other three rotating counterclockwise, daily mean power densities with winds above 3.8 m/s ranged between 21 and 47 W/m2 at wind speeds above cut-in speed, and between 6 and 30 W/m2 overall during the three months of testing, an order of magnitude above the power density of horizontal-axis wind turbines.
Another option-one that uses well-developed horizontal-axis turbine designs and that has already been commercially demonstrated in some countries-is to move away from land and set up large offshore wind farms. Power density limitations cannot be avoided (offshore wind turbines still have to be spaced appropriately), but stronger and more reliable offshore winds result in higher capacity factors and remove the common environmental concerns associated with land-based wind electricity generation. Denmark was the first country to build a substantial offshore capacity, with 871 MW installed by the end of 2012; in that year Danish offshore farms had a capacity factor of 44.9% and a lifetime capacity factor of 39.1% (Energi Styrelsen 2013).
But the costs and technical challenges (building longlasting structures in a corrosive environment, the construction of new long-distance, highvoltage transmission lines) are hardly trivial. In 2014 the United States still did not have a single offshore wind farm (although the first ones had been planned during the 1990s), Germany found that its offshore aspirations had become "dramatically problematic" (Dohmen and Jung 2011), and analyses done in both the UK and Germany showed that just connecting an offshore wind farm to the land grid costs more than building an equivalent capacity in gas turbines that can be located virtually anywhere. Even so, the European Wind Energy Association envisages great advances not just for near-shore installations in shallow water but for deep-water turbines far offshore (EWEA 2013).
Box 3.4
Power of a water turbine
Water Power and Hydro Generation
The energy flux of water turning a turbine (P) depends on the rate of water flow (Q in m3/s) and on the hydraulic head (h, in m), the distance through which the water falls before hitting the blades, and the actually delivered power will be also a function of the turbine efficiency (ii); water density and the acceleration of gravity remain identical (p = 1 g/cm3, that is, 1,000 kg/ m3, and g = 9.81 m/s'). A large, highly efficient (87%) turbine working under a head of 118 m and receiving a water flow of 700 m3/s will generate about 700 MW of electricity:
These numbers are the actual specifications for each of the 20 turbines installed in Itaipu, still the world's largest hydro station in terms of annual electricity generation, located on the Parana River between Brazil and Paraguay. China's Sanxia (Three Gorges, completed in 2009) Dam has a 60% higher installed capacity (22.5 GW, compared to Itaipu's 14 GW) but a much lower capacity factor (about 50%, compared to Itaipu's rate of close to 80%), and hence it typically delivers only about 85% of Itaipu's power (Chincold 2013; CTGC 2013). With little difference in the high efficiency of modern turbines and with p and g constant, the power densities of modern hydroelectricity generation will range widely along the continuum governed by the extremes of Q and h. At one extreme are the projects that rely on massive water flows with small generating heads created by low dams. The Yacyreta project on the Parana between Paraguay and Argentina is an excellent example of this category: its hydraulic head is just 22 m, but its low dam creates a reservoir of 1,600 km2 to supply enough water to install 3.1 GW and to generate 20 TWh/year (that is, 2.29 GW and a high load factor of nearly 75%).
The other extreme is represented by the tall dams pioneered in the Alps and creating high generating heads. The world's tallest dam is the 300-m Nurek on the Vakhsh in Tajikistan, built during the Soviet era and operating since 1980; an even taller, 335-m dam is considered for an upstream location on the same river, but so far the Rogun project has no financing. In some cases of high-head stations there may be no reservoir at all; these projects merely divert part of a mountain stream's water flow into a steeply falling conduit (underground tunnel or aboveground pipes) that leads it to turbines located far below. And there is yet another kind of river-run station (streaming systems), exemplified by Manitoba Hydro's Limestone project on the Nelson River, where a low dam (hydraulic head at 27.6 m) creates virtually no reservoir but the flow (regulated by three dams upstream) suffices to support an installed capacity of 1.34 GW (Manitoba Hydro 2013a). But, not surprisingly, for the Three Gorges Dam, the world's largest hydroelectric project, both Q and h are substantial.
Rated heads determine the type of turbine uses. Francis turbines can work with both low and high heads (10-300 m), Pelton and Turgo impulse designs
are used for high heads (higher than 100 m), and Kaplan machines, with adjustable guide vanes, are installed on projects with low to medium high heads (2-70 m). Rated speeds can be up to 1,500 rpm for Francis and Pelton turbines, half that rate for Kaplan machines; and maximum unit ratings can be in excess of 500 MW. Plant structures (the dam, including the spillway, the powerhouse, and for some projects also ship locks or fish ladders) and associated infrastructures (switchyard, access roads) make almost always only very small claims compared to the land submerged by a reservoir.
Power Densities of Large Hydro Projects
Dams built in the lower reaches of major rivers create large, even enormous, relatively shallow reservoirs. Lake Volta, impounded by the Akosombo Dam in Ghana, covers 8,502 km2, or 3.6% of the territory of Ghana, and is the largest man-made reservoir in the world. The Churchill Falls hydroelectric project, on the Churchill River in Labrador, reaches almost 7,000 km2 and created the second largest man-made reservoir in the world. The Kuybyshev Reservoir, on the Volga was created by the Zhiguli Dam and is the third largest reservoir by surface area, having a maximum extent of 6,500 km2. Lake Volta, created by the Akosombo Dam, is thus nearly as large as Puerto Rico (about 8,900 km2), and the Kuybyshev Reservoir is larger than Brunei (nearly 5,300 km2). The power densities of these projects (leaving aside the comparatively minor areas required for dams, associated structures, transformers, high-voltage connectors, and approach roads) are almost invariably less than 1 W;/m2: Akosomobo, with an installed capacity of 912 MW and a head of 68.9 m (Volta River Authority 2013), rates just 0.11 W;/m2, and the Churchill Falls project rates less than 0.8 We/m2 (fig. 3.5).