Turbine manufacturers are producing new generations of turbines with greater hub heights to reach the higher winds that blow at 80 metres(m) and higher. GE for example is offering its workhorse 1.5 MW turbine at 100 m hub height. The new Enercon E82/3 can be put on a 138 m tower. Vestas' new V112 goes up to 119 m. The rest will follow as more utilities turn to wind as an eventual replacement for fossil fuels.

Taller turbines are the first ingredient to more productive wind farms, but without the right information guiding their deployment, they can end up only marginally more productive than today's standard turbines, yet at a much higher cost.

But with the stakes so high, many developers still rely on 60 m meteorological masts for energy estimates.

While most wind power technology has advanced steadily over the last few years, the basic data collection platform hasn't. The typical hub height of a wind turbine is 80 m or more.

Additionally, the rotor diameter of a commercial wind turbine typically extends from 80 m to 120 m. Developers in North America, however, typically measure wind speed at three heights using anemometers mounted on 60 m towers. That's because erecting met masts of 80 m or more is complex, expensive, and usually requires special permitting. Limited met mast height forces developers to rely more heavily on extrapolation.

However, the higher the turbine tower, the greater the uncertainty and chances for error that go into the siting process. Uncertainty and error lead to inaccurate annual energy production (AEP) estimates. Inaccurate AEP estimates do not encourage banks and other investors to finance large-scale wind products. Clearly, the industry needs more options for collecting higher height data.

Enter remote sensing technology

Remote sensing systems based on established technologies make today's reliance on extrapolation unnecessary.

Analogous to radar, remote sensors use light or sound to detect wind from the ground. They emit light or sound signals in a fixed pattern. The signals encounter variations in the air and reflect back to sensors. The equipment interprets frequency changes in the original signals as Doppler shifts, which reveal whether the winds are moving towards or away from the sensors. Processing the Doppler shifts, elapsed times, and geometries of the emissions and reflections yields a profile of wind conditions.

With much smaller footprints than met towers and portable form factors, many believe remote sensors are more versatile and economical than towers, and measure data at heights up to 200 m. That covers the rotor sweep of higher turbines coming onto the market.

The remote sensors used in the wind industry today employ either sodar (SOund Detection And Ranging) or lidar (LIght Detection And Ranging) technology. Radar isn't used in wind energy applications because it can't resolve wind speeds within a few hundred meters of the ground.

Lidar and sodar also have operational limits. Ambient noise, especially at lower frequencies, and dry, thermally homogenous air affect sodar's performance. Cloud cover, sunlight, and aerosol-free air affect lidars. Sodar can't compete with a hard rain, and lidar can't work through dense fog.

But when these limitations are factored against met towers' inability to collect data at greater heights, remote sensing's advantages still outweighs the disadvantages. Of crucial importance is to give developers data from the 80 m-and-up range, which can then improve the case to bankers and investors.

Case study - Second Wind

The tower at the Boulder Atmosphere Observatory (BAO) in the USA provided Second Wind with an opportunity to validate the performance of its sodar, Triton, against anemometers on a tower that stands 300 m tall – 100 m taller than Triton's highest measured winds as the sodar can normally measure wind up to 200 m.

Built in 1977, the BAO tower is owned and operated by the National Oceanic and Atmospheric Administration's (NOAA) Environmental Technology Laboratory (ETL) and is one of only a handful of 300 m measurement towers in the USA. It was originally created to support atmospheric boundary layer studies and analyse ground-based remote sensing systems, including radar, lidar and sodar.

Located east of Erie, Colorado, the BAO tower stands more than 80 m taller than Colorado's highest office building, the 57-story Republic Plaza. To begin with, two anemometers, one wind vane and one temperature sensor on booms at 200, 150, 100 and 50 m were installed on the BAO tower. The team then set up a Triton unit nearly 350 m away from the tower.

The BAO tower proved to be an excellent site to test at all heights up to 200 m, the upper limit for which most remote sensing systems in the wind industry are designed to provide accurate measurements.

With Triton, Second Wind generally sees correlation to anemometers above .97 on a 0 to 1 scale. Before Triton, one of the most commonly referenced studies of sodar, Gennaro H. Crescenti's A Look Back on Two Decades of Doppler Sodar Comparison Studies in the Bull. Amer. Meteor. Soc., 78, 651–673, 1997, found an average correlation coefficient of .92 between sodar and reference measurements, indicating that Triton offers the industry a much stronger correlation than previous technologies. Comparison of Triton data to BAO tower data shows a correlation coefficient to anemometry of .98 at 50, 100 and 150 m and .96 at 200 m.

The conclusions of the research demonstrate that not only does sodar-based remote sensing provide accurate data at heights measured by typical met masts, but it also works very well all the way up to 200 m, well beyond the top height of the turbine blade.

Towering uncertainty

Many believe that the effect of not having data from heights greater than 60 m is introducing uncertainty into the wind farm siting process.

There are two sources of uncertainty when using the traditional approach of extrapolating 40-60 m data up to an 80 m hub height. The first is in extrapolation itself. Extrapolating is legitimate science and can often be accurate to within a reasonable certainty, but in wind power a lot of profitability can be lost in reasonable certainties.

The second is wind speed variations across the rotor span, which affects energy production.

Second Wind conducted a study of 111 data sets collected with its Triton. The sites were categorised as either hilly/ridgeline terrain, flat terrain, forested, or coastal. Instead of comparing a tower to a Triton, the study compared what happens when data from 40-60 m is extrapolated to 80 m, compared to measured data at 80 m. The results illustrated the error and uncertainty that can arise when data collected at 40-60 m is used to estimate the winds at a hub height of 80 m.

First, Second Wind used the two most common extrapolation methods, the power law profile and the log law profile, to estimate annual energy production (AEP) at each of the sites. The power law profile is a simpler method that assumes neutral stability and does not take surface roughness into account. Both methods produce a shear exponent that can be used to calculate an estimated wind speed for various heights above the measured wind speeds. However, no extrapolation method can capture the variation in shear profiles that occur at different times of day, seasons, or wind directions. Using the two profiling methods, Second Wind calculated the AEP for each site.

As previously mentioned, the extrapolation/estimation process yields two sources of uncertainty. The first is from wind speed extrapolation, and the second is from wind speed variations across the span of the rotor. Because of the effects of wind shear, which are more pronounced over a larger rotor, basing energy production estimates on the hub height wind speeds can affect the accuracy of the energy projection.

For this reason, the measured Triton data was also used in two ways. The measured 80 m wind speeds were used to estimate AEP. To explore the effects of variations across the rotor, ‘equivalent’ 80 m wind speeds were found by taking a weighted average over the diameter of the rotor, and then the annual energy production was estimated.

The results

The results of comparing each of the four AEP calculation methods for the range of sites produced some interesting conclusions. Overall, the power law profile extrapolation method tended to overestimate energy production in comparison with calculations done using measured 80 m wind speeds. The log law extrapolation method did not show this overall trend. Uncertainty was also higher with the power law profile method, at 4.2%, but still significant with the log law method.

The uncertainty in the AEP estimates from extrapolation varied significantly depending on the terrain where the measurements were made. Not surprisingly, the uncertainty was lowest at flat sites – about 2%. At forested sites, where surface roughness has a major effect on wind shear, the power law on average significantly overestimated the AEP, with an average error of 5.8%. The estimates using log law more closely matched the estimates from measured wind speeds, but still had an uncertainty of 5%.

At hilly sites and on ridgelines, both the power and log law yielded energy estimates that were, on average, 1.9% and 1.5% high and with fairly high uncertainties of 3.6% and 2.5%, respectively.

In complex terrain, both extrapolation methods yielded energy estimates that were less than 2% high, with uncertainties of 3.6% (power law method) and 2.5% (log law method).

Using the measured wind speed instead of the equivalent wind speed led to an over-prediction. The average error was found to be 1.3% with an uncertainty of 1.4%. This was most pronounced at forested sites.

These variations of a few percentage points here and there represent thousands of kilowatt-hours over a wind farm's life. They can easily be the difference between a highly profitable wind farm that encourages investors to support more projects, and a disappointing project that sours investors on wind power. Tower-based sensors will never be able to collect data at the heights needed for accurate AEP predictions at 80 m and higher.

Remote sensing can measure at those heights accurately and economically. The industry doesn't fully accept remote sensing data, according to Pike Research Renewable Energy Analyst Peter Asmus, but it's inevitable that they will, he believes:

“At present, leading engineering analysts working on behalf of banks and financial institutions do not accept a sole reliance upon sodar data,” he wrote on 29 July 2010. “However, this will change as the technology gains a bigger market share and is proven and validated by a third party analysis of sodar data with actual past performance of wind farms.”

About the author:

Susan Giordano is General Manager of Second Wind based in Somerville, Massachusetts.

Renewable Energy Focus, Volume 11, Issue 6, November-December 2010, Pages 18-20