Wind farms are now present in almost every corner of North America, operating successfully in a vast range of different wind conditions. Despite this success, the nature of wind – and the highly complex interactions that characterize its behavior – continues to present a challenge to the wind energy industry. For those trying to conduct wind resource assessments and predict the future energy output of wind farms, understanding spatial wind flow patterns has long presented a fundamental obstacle. Indeed, wind flow modeling is widely recognized as one of the largest sources of uncertainty in a wind resource assessment.
Better wind flow predictions can have significant economic benefits. More accurate resource prediction translates to more productive wind farm designs and lower revenue uncertainty, which leads to improved financing terms and better rates of return. New modeling tools and advances in computing power are making this possible, and the industry stands to reap the benefits.
The two most common types of microscale wind flow models used for wind resource assessments are linear and computational fluid dynamics (CFD) models. Linear wind flow models have been in use for many years and continue to be regularly used throughout the wind energy industry. They are computationally efficient; simulations can be completed in seconds on a personal computer. However, these models assume the governing flow equations are linear when, in fact, the real-world flow equations are nonlinear. This can lead to errors in flow predictions. As computers have become faster, wind flow analysts are increasingly turning to nonlinear options – the most popular of these is the CFD model.
CFD has several advantages over linear models, but it is rarely used to its full potential. It is most commonly applied at sites with steep terrain, where CFD offers improved wind flow predictions relative to linear models.
However, simply running any CFD model does not guarantee benefits. Even at sites with simpler terrain, where wind flow variation is supposed to be minimal, linear and CFD models can often struggle to reproduce the wind flow pattern indicated by the measurements. In general, the spatial variation of mean wind speed is much more pronounced than predictions from linear and typical CFD models would suggest. This problem, which is particularly common in the Great Plains and Midwest, relates primarily to atmospheric stability, and improved modeling techniques are helping to solve it.
Atmospheric stability impact
Atmospheric stability has to do with how air temperature varies with height above the ground. Unstable conditions are often associated with the daytime: The sun warms the ground, which in turn, warms the air near the ground, resulting in air that is generally lighter than the air aloft. This creates an unstable cycle where warmer, lighter air from near the ground rises while cooler, heavier air from above descends.
Conversely, stable conditions are often associated with nighttime: When the sun sets, the ground cools as it radiates heat to outer space, cooling the air near the ground. This creates a stable situation where the warmer, lighter air aloft tends to stay aloft while colder, heavier air near the ground tends to stay near the ground. Neutral conditions typically occur briefly around sunrise or sunset.
Atmospheric stability can have a large impact on wind flow, but accounting for stability adds significant complexity to a wind flow model. Most microscale wind flow models used in the industry, both linear and CFD, neglect atmospheric stability and effectively assume the atmosphere to be neutral. These wind flow models are used with the hope that neutral conditions sufficiently represent the average conditions at a given site.
However, experience tells us that this is not the case throughout much of North America. While neutral wind flow models – linear and CFD – can provide reasonable mean wind speed predictions for unstable and neutral flows, these models are not suitable for the prediction of stable flows. Our company research study spanning 132 met tower pairs found that the prediction errors from neutral wind flow models were twice as high when predicting stable flows than when predicting unstable or neutral flows.
For example, the spatial variation of wind speed is much different over a mesa during stable conditions than during neutral conditions. As shown in Figure 1, when the flow is neutral, the wind speeds at the front and back edges of the mesa are similar, but when the flow is stable, the wind speed at the back edge of the mesa is much higher than at the front edge. This trend has been observed at mesas throughout the U.S. Southern Plains, and there is simply no way that a neutral wind flow model can predict the wind speed trends during stable conditions. This is a major problem: At most sites in North America, stable flows represent a significant proportion of the available energy in the wind.
The wind industry has been aware of this problem for some time and has actively sought solutions. In order to tackle this challenge, our company developed a CFD-based wind flow model specifically targeting stable flows. This new model has been validated with 132 met tower predictions spanning 23 sites.
The validation shows that a carefully constructed CFD-based model of stable flow provides significantly improved wind speed predictions. Specifically, when the new stable CFD model is used, the average prediction error is reduced by more than 30%. Figure 2, which shows measurements versus predictions at a site in the Midwest, provides an example of the type of improvement that can be achieved using stability-enabled CFD – even on relatively flat terrain. Accounting for stability in the wind flow model increases the time required to turn around an analysis; however, when calculations are performed in parallel on a modern computing cluster of a few hundred cores, the analysis can be completed well within the time frame of a typical wind resource assessment.
Lower uncertainty, better financing
More accurate wind speed predictions translate to reduced uncertainty in the energy production estimate and can provide substantial financial benefits to the wind farm developer. At a typical site, more accurately capturing variation in atmospheric stability in the wind flow analysis can easily improve the one-year P99 (energy output with 99% probability) by 1% to 2%. In a standard debt financing of a 100 MW project, this level of increase in the P99 can result in $1 million to $2 million of additional debt sizing. Thus, there is a significant financial incentive to use wind flow models with increased accuracy. Extensive validation has shown that substantial accuracy improvements can be consistently achieved and practically implemented by correctly accounting for atmospheric stability in microscale wind flow modeling.
The technology to perform such modeling is now readily available. Indeed, the use of neutral wind flow models exclusively for wind resource assessments should no longer be acceptable to the North American wind industry; a credible wind flow prediction should account for atmospheric stability. w
Industry At Large: Wind Resource Assessment
Wind Modeling Advances Unlock Big Project Savings
By James Bleeg
Extensive validation shows that modeling atmospheric stability provides the key to more accurate wind flow predictions and lower revenue uncertainty.
NAW_body NAW_body_bi NAW_body_b_i NAW_body_bNAW_body_i