Using remote sensing to reduce uncertainty in wind project energy estimates

Accurate energy production estimates are key to successfully financing a wind power project. The most important input for an energy estimate is a wind resource assessment, during which the future performance of a planned wind project is commonly evaluated based on historic wind data combined with one or more years of on-site measurements. Consultants who evaluate the wind resource for project financing quantify the uncertainty that stems from various site conditions and wind resource assessment methods. When other factors are equal, a more thorough resource-assessment campaign makes it possible to reduce uncertainty and obtain more favorable financing, which increases the project’s value.

This article looks at reducing uncertainty and increasing project value by adding incremental measurements to a wind resource campaign. A cost-benefit analysis conducted by DNV KEMA Energy & Sustainability1 compared the costs of additional measurements and calculated the resulting increase in project value from each measurement method. In the study, additional measurements from sodars and met towers are shown to provide cost-effective ways of increasing project value.

Wind resource assessment
The classical approach to wind resource assessment uses anemometers and wind vanes mounted on met masts to measure wind speed and direction. A recent class of wind measurement technology, remote sensing, has been introduced to the commercial wind industry on a wide scale. Remote sensing technology includes sodar (sonic detection and ranging) and lidar (light detection and ranging) which are a class of instruments that measure wind characteristics by emitting sound or light waves and use their echoes and reflections to calculate wind speed and direction at various heights above ground.

Wind project developers are typically not choosing between met masts and remote sensing systems. Rather, they are choosing whether or not to supplement mast-based campaigns with remote sensing. Unlike met masts, which measure at discrete points, sodars and lidars typically measure across the turbine rotor area and provide horizontal and vertical wind speeds and wind direction at more heights. This higher data coverage can be a valuable part of the wind resource assessment.

Financing
Wind project financing takes many forms, including direct equity investments, tax-equity investments, and project debt. Most projects have some combination of these. Project debt and the change in energy estimate uncertainties impact the amount of leverage a project may carry. Leveraging a higher level of debt generally provides a higher return on the direct equity portion of the project financing.

Wind resource uncertainties drive the probability distribution of expected energy production for a project. The illustration One and 20 year energy production shows 1-year and 20-year probability distributions for a typical project. As seen in the blue 20-year curve, the P50 energy production level is the central energy production estimate; the probability of producing more than this amount of energy over the 20-year expected life of the project is 50%. The P99 level represents an energy production value that has a 99% probability of exceedance over the life of the project. The red one-year curve represents the values for any one year within the 20-year project life. The shape of these distributions is determined by the energy assessment uncertainties and the variability of the wind resource. The wider the spread between the P99 and the P50, the more uncertain the energy estimate.

Lenders typically size project debt at a level that can be serviced even at the P99 energy production level. For example, the debt size may be calculated with a debt service coverage ratio (DSCR) of 1.0 times the one-year P99 cash flows. In other words, at a given interest rate, what annual debt payments can be made, or “covered,” by the project with the annual cash flow that is generated if the project produced energy at the one-year P99 level? With this common method of debt sizing, the wind resource uncertainties that drive the P99 energy production value have an important role in the project financing.

A strategic approach to reducing uncertainty
To make the greatest impact on the project financing by reducing wind resource uncertainty, a project developer must evaluate the uncertainties and focus on reducing the largest uncertainty categories. The average and typical uncertainty values are presented in the table Sources of uncertainty and the illustration Average of uncertainty from 200 preconstruction assessments. These are based on a survey of about 200 North American pre-construction energy estimates of utility-scale wind farms conducted by DNV KEMA. However, for any given project, the distribution of uncertainties will be different based on the size of the project, complexity of the terrain, height of the proposed turbines, duration of the wind measurement campaign, availability of historical wind data, and other factors. For example, the charts in Distribution of energy uncertainties illustrate a project in which the variation in wind resource across the project is not well characterized by the on-site measurements and therefore there is a large spatial variation uncertainty.

Large uncertainty categories disproportionately impact the overall uncertainty (and thus the difference between the P50 and the P99 energy estimates) because uncertainty categories are combined in a non-linear manner, by taking the square root of the sum of the individual uncertainties squared.

To get the most value out of their investments, developers should focus additional effort on reducing the largest areas of uncertainty in their projects.

Reducing uncertainty through measurement
Choices of measurement height, location, duration, and the number of measurement locations, affect uncertainty in wind resource assessment. Uncertainty is reduced mainly through additional measurements. Measurement options for reducing vertical extrapolation and spatial variability uncertainty are discussed below.

Reducing vertical extrapolation uncertainty
Wind turbines used in new projects are typically taller and have larger rotors than previous turbine models. Typical turbine hub heights are 80 to 100m, with some in Europe reaching to 140m. Rotors are also larger, commonly spanning 80 to 120-m in diameter. A typical 60-m met mast cannot measure winds at heights covered by the blades of modern utility-scale turbines. However, within the North American wind market, 60-m met masts are commonly used because of the additional permitting required to install a mast taller than 60m.

To estimate wind speeds across the turbine rotor, the wind industry typically extrapolates wind-speed measurements from lower heights because cross-rotor measurements are not commonly available. Information about the surrounding terrain, vegetation, and the atmospheric conditions are often used in the extrapolation. The vertical extrapolation methods are rooted in the assumption that data gathered at lower heights represents conditions at higher heights. However, this is often not the case and it leads to error and uncertainty in the extrapolated wind speeds. A study by Second Wind of 111 data sets compared measured data to extrapolated data; it found uncertainties in annual energy production calculations ranging from 3.0 to 4.2%.2

The uncertainty associated with vertical extrapolation (wind shear) can be reduced in two ways: by measuring at higher heights which reduces the need to extrapolate, and by validating the shear extrapolations with measurements from sodars, lidars, and taller met masts. For example, if measurements have been made using a 60-m met tower for three years, deploying a sodar nearby for an additional year to measure at heights from 40 to 140m can yield a better understanding of the wind shear profile. Hub-height sodar data can be used directly in an energy assessment without vertical extrapolation and to identify unusual shear patterns across the turbine rotor.

Spatial variation uncertainty
Additionally, the sodar data can help the analyst identify possible sources of error when data are extrapolated horizontally to other locations. Spatial variation (wind flow modeling) uncertainty results from the differences in wind characteristics across a project site due to terrain, surface roughness, and other elements. Measurements are typically taken at a few locations within the project area, but the wind resource must be evaluated at all locations where wind turbines will be deployed. For example, a project spread over many miles may measure the wind resource at 5 locations but plan to install wind turbines at 80 locations.

To evaluate the wind resource across the site, the wind industry relies on wind-flow modeling. There are several different types of such models, each based on empirical models or simplifications of physical equations, including linear flow models, non-linear flow or computational fluid dynamics models, and dynamic mesoscale atmospheric simulation models. Studies have found significant error and uncertainty in these wind flow models. For many projects, the greatest share of uncertainty in wind resource assessment is the spatial variation (wind flow modeling) category. The most effective way to reduce this uncertainty is through additional measurements to better characterize the wind flow across the site.

Cost benefit analysis
A previous white paper by DNV KEMA evaluated costs and benefits of two scenarios to reduce wind resource uncertainties: a tall turbine scenario and a complex terrain scenario.

Tall Turbine Scenario: A project is planned to have 100-m hub height turbines. With only 60-m mast measurements, there is a large vertical extrapolation uncertainty. To reduce this uncertainty, the existing 60-m met mast measurements are augmented with a short-term sodar measurement, a long-term sodar measurement, or a 100-m met mast. For these cases, DNV KEMA determined that incremental measurement investments of $38,000 to $155,000 can yield increased debt sizes of $2.0 million to $4.4 million when a project is financed.

Complex Terrain Scenario: A project located in complex terrain has too few measurements to fully characterize the variation of the wind across the site. Therefore, there is a large spatial variation uncertainty. To reduce this uncertainty, the existing 60-m met mast measurements are augmented with an additional 60-m mast, additional sodar measurement, and the addition of both a sodar and a 60-m mast. For these cases, DNV KEMA determined that incremental measurement investments of $40,000 to $99,000 can yield increased debt sizes of $1.9 million to $3.9 million when a project is financed.

Conclusion
Thorough resource assessment campaigns are important for understanding expected energy production and increasing project value. Sources of uncertainty should be evaluated and measurement campaigns should be adapted to reduce the largest uncertainty category to have the largest benefit. A cost-benefit analysis showed that additional remote sensing or met mast measurements provide increased project returns and larger debt size. To determine the most appropriate measurement campaign for any particular project, a cost-benefit analysis should be conducted based on the specific circumstances of the project. For more information and a cost-benefit analysis of different technologies see: www.secondwind.com/dnv-kema-white-paper. WPE