A high quality wind resource assessment (WRA) campaign is the process by which wind farm developers estimate the Annual Energy Production (AEP) of a proposed wind farm. The local terrain and micro weather patterns in the area of a planned wind farm strongly affect the energy in the wind. At the same time, turbine manufacturers have become increasingly interested in site-specific wind quality to determine the suitability of the site for a particular turbine model or rotor size, as increasing turbulence can exacerbate fatigue damage.

Throughout the world many of the best wind sites have been built out, leaving the more challenging sites – those with lower wind speed, higher turbulence, or greater distance to the grid – to be developed.

As a result, the expectations surrounding wind resource data have changed. With more stakeholders involved, tighter access to capital, and more wind farms in production, the data is under higher scrutiny, which has placed greater importance on quality resource assessment campaigns and accurate energy estimates.

A wind resource campaign is typically funded by the wind farm developer and performed by a consulting meteorologist. Every wind energy estimate is based on multiple assumptions, such as the windiness of the measurement period relative to the long-term mean wind speed at the site, and the accuracy of the anemometer.

Each of these assumptions carries its own uncertainty, which results in a cumulative uncertainty for the estimate that is described by a probability distribution. At the centre of this distribution is the P50 estimate, which means there is a 50% chance that the actual energy production will be higher than the estimate (see figure 1).

Also shown is the P90 estimate, above which the actual production will fall with 90% probability – a more conservative estimate.

One of the primary roles of the meteorological consultant is to quantify the uncertainty of each component and determine the shape of the distribution.

Basic wind resource assessment

The basic elements of a wind resource assessment campaign include use of on-site meteorological (met) masts outfitted with redundant anemometers at multiple heights, collecting data for a minimum of one year. While advanced technology to measure the wind is entering the market (see later) this approach still serves as the foundation for a successful wind resource assessment campaign.

Typically, wind speed and wind direction measurements are taken at one or two second samples, and processed by the data logger into 10-minute averages, with minimum, maximum and standard deviation. Turbine manufacturers use the same averaging period to measure the power curve of their turbine, which characterises the actual turbine power output under standard conditions over the range of its operational wind speed.

A frequency distribution of the measured wind speed gives the amount of time spent at each wind speed, which is then multiplied by the power generated by the wind turbine at each wind speed. This gives a resulting curve of estimated energy production, which is then integrated to get the AEP estimate.

Figure 2 shows the power curve for a 1.5 MW wind turbine, the frequency distribution of measured wind, and the resulting energy distribution. It also shows a Weibull curve fit to the actual data, which is sometimes used to determine the time spent at each wind speed instead of the actual wind distribution when, for example, the measured data set is incomplete.

This AEP estimation, however, is incomplete without an accompanying estimate of its uncertainty.

Uncertainty

The total uncertainty of the AEP is the combination of uncertainties of the component parts. Sources of uncertainty include:

• Relative windiness of the measurement period due to inter-annual (year-to-year) wind speed variability;
• Effects of wind shear when extrapolating to hub height;
• Effects of spatial extrapolation from the location of measurement to the location of turbine;
• Losses due to inter-turbine interference (wake effects);
• Losses due to turbine unavailability (down time);
• Wind speed measurement.

Average wind speed varies year to year for any given location. Correcting the measured data for inter-annual variability is typically done using one of several Measure-Correlate-Predict (MCP) techniques. Many airports have a long-term record of wind speed that may serve as a suitable reference. If a good statistical correlation can be established between the measured data and the long-term data, the measured data can be scaled to be consistent with the long-term mean, and thereby improve the accuracy of the estimate.

Wind shear is defined as the change in wind speed and direction with height. When it is not possible or practical to directly measure at hub height, it is necessary to extrapolate the measurements to hub height. Two commonly used models are the Power Law Profile and the Logarithmic Wind Profile.

Both of these models assume that the wind profile across the range of measurements is the same as the profile in the extrapolated area. There are a number of meteorological phenomena such as low-level jets that create periods of time when this assumption is invalid, leading to higher uncertainty in the extrapolation process.

Remote sensors, devices that can measure the wind from the ground up to and beyond hub height, help reduce uncertainty from extrapolation by directly measuring the wind speed and direction at multiple heights across the full rotor diameter.

Typical wind farms have a variety of terrain elevations, surface features, and roughness; and each turbine may have a slightly different wind resource. The turbines also need to be placed to minimise wake losses that reduce turbine production.

To solve this problem, micrositing models are typically used to optimise the location of each turbine, thereby maximising total energy generation while taking into account distance from the grid interconnection, noise impacts on neighbouring locations, and visual impact. The data collected during the resource assessment campaign is a critical input to this micrositing phase.

Wind measurement data from multiple sites can significantly reduce the uncertainty of the results. For this reason, it is often cost effective to invest in additional met towers up front so that uncertainty of the modelling, and therefore the AEP estimation, can be reduced, diminishing the overall risk and the cost of capital for the project.

When measuring wind speed with a met tower, the tower itself disturbs the air flow and interferes with the measurement, which is a source of uncertainty. For tubular towers, it is recommended to mount the anemometer at least 8 to 10 tower diameters away from the tower centre line.

For lattice towers, mounting the anemometer at least 7 face widths from the tower centreline is recommended. Even with that boom length, when the anemometer is directly downwind of the tower, the measurement is not useful.

For this reason, a redundant sensor is typically mounted at the same height, and the booms are oriented 45 to 90 degrees on either side of the prevailing wind direction so that all wind directions have an unobstructed sensor. Sensor pairs at each level also reduce uncertainty by providing the ability to cross check each sensor for good quality.

The impact of remote sensors

Remote sensing technology has evolved significantly in recent years, providing additional tools that complement traditional anemometry. The two most prominent remote sensing technologies are LIDAR (Light Detection and Ranging) and SODAR (Sonic Detection and Ranging).

Table 1 above shows a side-by-side comparison of the two technologies. The industry is still learning how to apply remote sensing to resource assessment, as well as to power performance measurement of wind turbines and to wind forecasting.

Compared to a met tower, remote sensing has the advantage of measuring up to blade tip and in all three dimensions at once. Remote sensors are also more portable than a met tower, which is useful for directly measuring more locations across a potential site.

However, remote sensors typically have lower data availability due to limitations in the current technology, suggesting a tower is required to obtain a full data set for at least one year. Also, the accuracy of the data is reduced in complex terrain where high turbulence and flow curvature conflicts with the assumption of homogeneous flow through the measured volume.

Remote sensing is also used for offshore wind development, where the return on investment is particularly favourable relative to installing offshore met towers.

Remote sensors measure a volume of wind across the entire rotor diameter while anemometers measure one or more points in the flow that may or may not be within the rotor diameter.

This fundamental difference has raised questions about which method gives a better representation of the wind as it relates to wind turbine performance. However, it is clear even now that remote sensing will play an increasingly important role in wind resource assessment as research in how to apply its measurements is conducted.

Using the previous example AEP estimation of 25.1 GWh per year (see figure 2), a notional sample calculation is given below to show how the use of remote sensing to reduce uncertainty related to wind shear improves the P90 estimate (while the P50 estimate remains the same).

In this example, the uncertainty due to wind shear is reduced because measurements are taken directly at hub height. Also, the uncertainty due to spatial extrapolation is reduced because the remote sensor has measured the wind at several locations around the wind farm site, improving the accuracy of the micrositing model. As can be seen in Table 2 and Figure 3, a small reduction in AEP uncertainty (1.7%) results in a substantial increase in the P90 estimate (561 MWh).

Looking to the future of wind resource assessment

Reducing uncertainty in wind resource assessment directly translates into a higher P90 estimate, which can lead to a lower cost of capital, more favourable terms of a power purchase agreement, and higher return on investment.

The use of remote sensing is one way developers and others have been able to realise these gains. Technological improvements, further research into the application of remote sensing data, and innovations in other areas of resource assessment will drive further reductions in uncertainty, leading to more accurate annual energy production estimates.

Brendan Taylor is the Engineering Manager for Research and Development and Steven Clark is a Mechanical Engineer with NRG Systems.

For references please contact the editor.