Maximizing the profitability of wind projects involves a careful balance of placing turbines in windy locations, minimizing losses and operating costs, and using the available land in the most efficient way possible. In order to achieve this balance, it is important to understand the effect of wake losses, which occur when wind turbines create a wake that reduces the wind available at downwind turbines, thus reducing energy production.
Wakes often represent the largest source of energy loss for a wind project and are one of the most difficult losses to manage after a project is built. Whereas higher-than-expected losses associated with availability or turbine performance may be mitigated through warranties or reduced through changes in maintenance practices, there is no way to reduce turbine wakes, other than to move or remove turbines.
Because wakes cannot be “fixed” after a project is operating, it is critical that project developers understand wakes when the turbine array is being designed. That is the only opportunity to place turbines in a way that balances wind speeds and wakes in order to maximize the wind farm’s energy production.
Although wakes are not yet fully understood, recent studies of wakes in large operating wind farms have advanced industry knowledge of how wakes work and how to model them.
What we learned
It is no surprise that wake models are not 100% accurate. The flow behind a single wind turbine is a complicated fluid-dynamics problem. The combination of wakes behind multiple rows of turbines is even more complex, and considerations such as topography, land cover and atmospheric conditions make it an incredibly complicated physics problem.
Accurate prediction of wake losses would require a multitude of measurement inputs, a complicated set of equations and significant computing power. However, this approach is neither possible nor practical at this time. The input data needed to conduct such modeling is generally not available for most wind projects, and without solid measurement campaigns designed to capture the necessary input data, advanced and computationally intensive models will have limited benefits.
For many projects, the modeled wake loss is adequate in aggregate but has significant errors for certain conditions. For example, some models may overestimate the wake loss in the second turbine row, but the error could be offset by underestimated wake losses in subsequent rows.
Some models will overestimate the wake loss in unstable atmospheric conditions (which often occur during the daytime hours) but underestimate the wake loss in stable conditions (which often occur at night). For this reason, the annual wake-loss estimate for a project is often reasonably accurate.
In some cases, however, the errors do not cancel each other out, but instead result in a bias. Therefore, it is important to understand the conditions that result in model errors so corrections can be made. An accurate aggregate wake-loss estimate is not sufficient for optimizing a project layout and understanding the economics of some projects; accurate estimates for specific conditions should also be considered.
For example, an algorithm cannot truly optimize a turbine layout if the wake-loss estimate is not accurate for each turbine location. Accurate wake modeling is required to correctly determine the trade-off between a tightly spaced turbine layout in the windiest area and a loosely spaced turbine layout that extends into lower-wind-speed areas but experiences less wake loss.
Stable atmospheric conditions typically coincide with low turbulence and abnormal wind-shear profiles. Oftentimes, there is high shear close to the ground but decreasing wind shear on the upper half of the turbine and above the rotor. The combination of low turbulence and decreasing shear with height results in higher wake losses than other atmospheric conditions do because there is less turbulence to mix the wake with more energetic winds and less energetic winds aloft to replenish the wake. Figures 3 and 4 illustrate this concept.
Commercially available wake models typically do not perform well in stable conditions. In these conditions, wake losses extend much farther than typically predicted. Wake losses of 20% to 25% have been observed in stable conditions at a distance of 3 km (approximately two miles or 40 rotor diameters) downwind of just one to two turbine rows, while typical models predicted a loss of 5% to 15% for this case. For this reason, setbacks from nearby wind projects are crucial.
Additionally, stable conditions affect the wake loss internal to a project. For example, Figure 1 and Figure 2 compare measured and modeled wake losses for a large wind project. Wake losses are higher during stable (low turbulence) conditions; the wake loss nearly doubles for very stable conditions compared to neutral and unstable conditions. For context, stable conditions are present approximately 20% to 30% of the time at many North American projects.
Comparing wake-model predictions to measured wake losses shows a consistent underestimation of wake losses during stable conditions. In some cases, this error is offset by overestimates for neutral and unstable conditions, as shown in Figure 2.
However, in other situations, the model underestimates all conditions, as shown in Figure 1. These plots show the results of a single model. However, in evaluations of many different models, no model has performed well in all conditions or for all sites.
One reason for the discrepancy is that wake models typically use turbulence as an input but rarely account for wind-shear conditions that are concurrent with stable conditions. Using information about the wind-shear profile across and above the rotor would allow for a more accurate estimate of the energy available for replenishing the wake. It is important to note that modifying wake models to account for the shear profile across the rotor is only half the battle; the wind-speed profiles must also be measured on-site and, therefore, be available for use in the model.
Much emphasis has been placed on larger-than-predicted wake losses for turbines deep in the array. Although the depth of the array is a significant factor, it is also important to evaluate the true cause of this issue.
Large wake losses can be the result of atmospheric conditions, tight upwind turbine spacing, less-energetic wind-shear profiles, wakes traveling farther than predicted or a combination of these and other factors. It is not only big wind projects that can have large wake losses that are not fully predicted by common models; during certain atmospheric conditions, larger-than-predicted wake losses can also occur in shallow arrays, behind only one or two turbine rows. Therefore, models need to better incorporate the effects of atmospheric conditions – such as turbulence, shear and stability – on wake behavior.
Additionally, modeling efforts should progress toward using inputs that are based on measurable properties that directly impact wake losses. Using a mathematical parameter – such as a roughness length to model the wake of a turbine – is a useful tool but does not have a direct connection to a physical property nor a direct impact on wake losses.
Although some models account for stability, there is no direct link between the stability parameters required by the model and the conditions that are typically measured at a site. Without being able to measure a modeling input, it is difficult to select the appropriate value for the input parameter. Focusing attention on the wind-shear profile and turbulence conditions in relation to measured wake losses is one possible approach.
It is important to design an effective turbine layout before the project is built. This requires a detailed wake analysis that considers the atmospheric conditions at the site, the wake impacts of any nearby and future wind projects, and the turbine location and spacing options. Conventional guidance for the distance between turbines within a row and between rows can be a useful starting point, but a more detailed analysis that addresses the specific characteristics of the site should be undertaken.
For example, if stable conditions occur more often when the wind is coming from the south than when the wind is coming from the northwest, then this should be reflected in the turbine spacing and the orientation of the turbine rows. Similarly, if the site conditions allow for sufficient wake replenishment at the second row, adding a third row may make more sense than putting those turbines elsewhere.
The approach to mitigating future wake losses through land control also deserves a detailed, site-specific wake analysis. A one-size-fits-all land control strategy may be unnecessarily aggressive and costly at one site, but at another site, this approach may leave the owner at risk of future revenue loss due to nearby development. w
Marketplace: Wind Assessment
Navigating The Complexities Of Wake Losses
By Katy Briggs
The wind industry should be careful not to put too much trust in wake-loss models, as several factors can affect their accuracy.
NAW_body NAW_body_bi NAW_body_b_i NAW_body_bNAW_body_i