Numerous studies have shown that atmospheric conditions affect wind turbine performance; however, some findings have exposed conflicting results for different locations and diverse analysis methodologies. In this study, we explore how the change in wind direction with height (direction wind shear), a site-differing factor between conflicting studies, and speed shear affect wind turbine performance. We utilized lidar and turbine data collected from the 2013 Crop Wind Energy eXperiment (CWEX) project between June and September in a wind farm in north-central Iowa. Wind direction and speed shear were found to follow a diurnal cycle; however, they evolved differently with increasing wind speeds. Using a combination of speed and direction shear values, we found large direction and small speed shear to result in underperformance. We further analyzed the effects of wind veering on turbine performance for specific values of speed shear and found detrimental conditions on the order of 10 % for wind speed regimes predominantly located in the middle of the power curve. Focusing on a time period of ramping electricity demand (06:00–09:00 LT – local time) exposed the fact that large direction shear occurred during this time and undermined turbine performance by more than 10 %. A predominance of clockwise direction shear (wind veering) cases compared to counterclockwise (wind backing) was also observed throughout the campaign. Moreover, large veering was found to have greater detrimental effects on turbine performance compared to small backing values. This study shows that changes in wind direction with height should be considered when analyzing turbine performance.
Wind power generation directly depends on wind speed. Additionally, power depends on atmospheric conditions like static stability, shear, and turbulence (e.g., Bardal et al., 2015; van den Berg, 2008; Kaiser et al., 2007; Rareshide et al., 2009; St. Martin et al., 2016; Sumner and Masson, 2006; Vanderwende and Lundquist, 2012; Wagner et al., 2010; Walter et al., 2009; Wharton and Lundquist, 2012b). Idealized theories state that the power extracted by a wind turbine is a function of the blade element's efficiency (i.e., turbine blade design) and the available power flux through the disk swept by the blades (Burton et al., 2001). However, atmospheric turbine operating conditions diverge from simplified ones used for turbine design. Varying inflow speed and direction profiles, turbulence, transient conditions, and wake effects from upwind turbines alter power production.
Static stability in the lower planetary boundary layer is governed by
temperature gradients that drive or suppress buoyancy (Stull, 1988). Three stability regimes are usually established: stable, neutral, and unstable conditions, corresponding to a stratified, equilibrated, and convective atmosphere, respectively. Several means of quantifying atmospheric stability have been employed in wind energy studies, including the dimensionless wind shear exponent (
Several factors could explain the difference in results between these stability regimes. First, the available data and thus the analysis method differs. Wharton and Lundquist (2012b) segregated power production regimes using wind shear exponents and turbulence intensity in a location with channeled flow. In contrast, Vanderwende and Lundquist (2012) employed the Richardson number and wind shear criteria to quantify local atmospheric stability on a wind farm that could experience directional wind shear. The conflicting results among studies suggest that either additional forcing mechanisms are present or that site-specific factors govern the effects on power production. One likely site-specific factor is the role of directional wind shear, which was not explicitly considered in the above studies but could differ between those sites.
Here we seek to resolve these conflicting results by quantifying the role of directional wind shear in turbine performance. Directional shear is the change of wind direction with height. The three main mechanisms for generating direction shear are the thermal wind, the inertial oscillation, and surface stress. In the meteorological community, “wind veering” is used to describe the clockwise turning of the geostrophic wind with height, while “wind backing” describes the counterclockwise turning of the geostrophic wind with height (Holton and Hakim, 2013). Veering tends to be associated with warm air advection, while backing is associated with cold air advection. Both these terms are associated with deep layers in the atmosphere rather than relatively shallow layers of the atmospheric boundary layer (ABL). Within the boundary layer, veering tends to be more common in the Northern Hemisphere due to the direction of the Coriolis force and the resulting Ekman layer.
In a wind energy context, directional shear causes the incoming wind to be misaligned with the rotor axis over some heights of the rotor swept area (Bardal et al., 2015) as the turbine tends to orient itself into the direction of the wind at hub height. Both veering and backing generate a substantial variation of the horizontal wind speed component orthogonal to the turbine axis altering the energy flux through the rotor and the turbine's capability to extract energy (Wagner et al., 2010). Veering decreases the mean relative wind speed experienced by a clockwise-rotating blade, while backing increases it (Wagner et al., 2010). In contrast, the opposite happens for the angle of attack. The angle of attack is larger for veering and smaller for backing (Wagner et al., 2010). Simulations by Wagner et al. (2010) also show that though backing increases both mean lift and drag over the blade, the resulting tangential force experienced by the rotor is reduced, while it is slightly augmented by veering. The increase in tangential force from wind veering results in a slight increase in power production, whilst wind backing slightly decreases power production.
The existing studies on the effects of atmospheric stability on power production differ in the role of directional wind shear. The Wharton and Lundquist (2012a, b) wind turbines and St. Martin et al. (2016) testing site were surrounded by complex terrain that steered a channeled flow into the turbines and prevented the development of directional wind shear during stable conditions. In contrast, at the Vanderwende and Lundquist (2012) location, complex terrain did not prevent the occurrence of a changing wind direction with height.
Several methodologies have been employed for studying the effects of directional wind shear on turbine performance. Bardal et al. (2015) used measurements from a test site in the coastline of Norway without distinguishing between veering and backing. They found a small reduced power output below rated speeds for directional shear above 0.05
In this present study, the effects of directional wind shear on power production were analyzed by separating the effects of speed shear using data collected in the 2013 Crop Wind Energy eXperiment (CWEX-13) field campaign of a 150 MW onshore wind farm. Section 2 provides an overview of the dataset utilized for this study, which includes turbine power production and wind profiling lidar, and their respective filtering. Section 3 describes the definition of directional wind shear, speed shear, and individual turbine's power curves. Wind shear characterization and its effects on turbine power production are summarized in Sects. 4 and 5.
Schematic view of the wind farm in central Iowa where the CWEX-13 campaign took place. The turbines of interest for this study are marked as A–D.
The Crop Wind Energy eXperiment projects (CWEX) in 2010, 2011, and 2013 explored how wind turbines create changes in microclimates over crops (Rajewski et al., 2013, 2014, 2016), how the diurnal cycle affects wind turbine wakes (Lee and Lundquist, 2017; Rhodes and Lundquist, 2013), and how agricultural cropping and surface management impact wind energy production (Vanderwende and Lundquist, 2016). The 2013 campaign emphasized the impacts of atmospheric conditions like nocturnal low-level jets (Vanderwende et al., 2015) on wind turbine performance and the dynamics of wake variability (Bodini et al., 2017; Lundquist et al., 2014). These data have also been used to test approaches for coupling mesoscale and large-eddy simulation models (Muñoz-Esparza et al., 2017). The CWEX-13 field campaign took place between late June and early September 2013 in a wind farm in north-central Iowa. Measurements from several surface flux stations, a radiometer, three profiling lidars, and a scanning lidar were collected.
The wind farm consisted of 200 wind turbines extending in a parallelogram
with a long axis from the southeast to northwest (Fig. 1). The northernmost 100 turbines were General Electric (GE) 1.5 MW extra-long extended (XLE) models and the southernmost 100 turbines were GE 1.5 MW super-long extended (SLE) model turbines. The land was generally flat with a slope smaller than 0.5
Technical specifications of the turbines studied in the CWEX-13 field campaign (General Electric, 2009).
To quantify wind shear, we relied on data collected from the profiling lidar
Windcube V1, designed by Leosphere, deployed during the CWEX-13 campaign.
This Doppler wind lidar measured vertical profiles of speed and direction at
nominal 1 Hz temporal resolution. It used a Doppler beam swinging (DBS)
approach, obtaining radial wind measurements along four cardinal directions
at an inclination of 62.5
Wind lidar data were available throughout the campaign (Fig. 2a). The 2 min wind speed (80 m height) observations for each day only presented significant shortages (
Lidar
The prevailing wind direction for the recorded period in this wind plant was
primarily south-southwesterly having a mean wind speed of 8.21 m s
Wind rose for lidar hub height (80 m) wind speeds between cut-in and cut-out. The black outline highlights the wind direction sector (southerly) used for subsequent data analysis.
The remaining analysis only considers winds with southerly components (wind
direction between 100 and 260
The subset of turbines employed for this study consists of four clockwise-rotating (while looking downwind) GE XLE 1.5 MW variable-blade-pitch wind turbines (see Table 1 for specifications). Power production, nacelle wind speed and blade pitch angles were provided by the wind farm operator as 10 min averages recorded via the supervisory control and data acquisition (SCADA) system of each turbine. To analyze how wind shear impacts power production, turbine underperformance during curtailments was filtered following the blade pitch angle approach of St. Martin et al. (2016). Blade pitch angles are controlled to maximize power production as a function of nacelle-measured wind speed, and large blade pitch angles typically represent curtailed conditions or rapidly changing conditions. Therefore, we discarded 10 min periods with blade pitch angles outside
Power curve based on nacelle-measured wind speed
Turbine- and lidar-recorded data are averaged over different time intervals by their respective data-acquisition systems (2 min for lidar and 10 min for turbine). Matching turbine performance with atmospheric conditions was performed by averaging 2 min lidar measurements for the corresponding 10 min turbine power production period. For example, turbine data for 4 July 2013 from 05:00 to 05:10 LT is matched with the average of five 2 min lidar data measurements corresponding to the same date and time period. As is illustrated in Fig. 5, turbine and lidar data were synchronized for the duration of the campaign.
Time series from 00:00 LT 4 July 2013 to 00:00 LT 7 July 2013 of hub-height wind speeds measured by the cup anemometer on the nacelle of Turbine A
According to the International Electrotechnical Commission's Wind Turbine
Power Performance Standard (IEC, 2005), wind turbine power performance characteristics are determined both by the measured power curve and the
annual energy production. The measured power curve is obtained by
simultaneously collecting data from meteorological variables and turbine
performance over long periods of time. Wind speed is measured at hub height
using cup anemometers mounted on a meteorological mast positioned 2–4 rotor diameters upwind of the turbine, and power output is recorded using a power measurement device (e.g., power transducer) between the wind turbine and the electrical connection. Measurements are averaged over 10 min time
periods. A database for a wide range of wind speeds (0.5 m s
Mean power curve for Turbine C based on 80 m lidar wind speed
measurements overlaying the number of power production cases for each 0.5 m s
Power production for the duration of this campaign reflected persistent
differences from the manufacturer's reference values at wind speeds below 8 m s
As suggested by the histogram in Fig. 6, the frequency of occurrence changed with wind speed, roughly following a Weibull distribution with a shape factor of 2.05 and a 6.9 m s
Directional wind shear is defined as the change in wind direction with height, and speed shear corresponds to the change in the mean horizontal wind speed. One mechanism for generating wind shear is the vertical shear of geostrophic wind referred to as thermal wind. The thermal wind is caused by large-scale horizontal temperature gradients that can be created by sloping terrain, fronts, land–sea interfaces, and large-weather patterns (Stull, 1988). Wind shear overnight is also generated by the inertial oscillation (Blackadar, 1957; Van de Wiel et al., 2010). The inertial oscillation is the rotation in the wind vector in the residual layer caused by a force imbalance at sunset, when mixed layer turbulence ceases. As frictional stress diminishes after sunset, pressure gradients tend to accelerate subgeostrophic winds in the mixed layer back toward geostrophic. Inertia from the counteracting Coriolis force induces an oscillation in the wind vector causing it to become supergeostrophic and to turn clockwise (Northern Hemisphere) with time (Stull, 1988). A third forcing mechanism is frictional drag with the ground. Turbulent momentum fluxes in the boundary layer reduce the actual wind speed near the surface. The Coriolis force, being directly proportional to the wind speed, decreases, creating a force imbalance with the pressure gradients. As a result, the actual surface wind vector is directed across the isobars toward low pressure (Holton and Hakim, 2013).
Directional shear in this study is calculated as the shortest rotational
path between wind vectors at 40 and 120 m a.g.l., normalized over vertical distance between the measurements. For example, a case with southerly winds at 40 m and westerly winds at 120 m would be calculated as 90
Speed and direction wind shear alter the available power of the air through
the turbine and its ability to extract energy from the wind (Wagner et al., 2010). The available power in the air flowing across a disk is proportional to the projection of the velocity vector over the disk area,
The literature includes a range of different classification thresholds to
analyze and contrast high wind shear and low wind shear to explore their effects
on turbine performance. Bardal et al. (2015) utilized a threshold of 5
A predominance of wind veering was observed in this site compared to wind
backing cases (Fig. 7a). Wind veering occurred more than 77 % of the time and displayed larger numerical mean and maximum values (0.0939 and 1.83
Probability distribution for direction shear
The speed shear probability distribution was bimodal, with a narrow peak
centered around 0 and a broad peak close to 0.4 (Fig. 7b). An increase of speed with height was observed 88.6 % of the time, from which 53 % was above 0.225. Further, 60 % of the recorded data lay above the commonly used
Both direction and speed shear had a tendency to decrease with height. Figure 8a illustrates how wind direction evolved differently through the rotor layer for veering and backing cases. Both clockwise and counterclockwise wind direction rates of change were larger in the lower rotor layer. Directional shear was 1.6 times larger from 40 to 60 m compared to 100 to 120 m a.g.l. for veering and 1.79 times larger for backing. When considering the absolute value of the wind vector rotation, the lower layer (40 to 60 m) experienced an average change in wind direction 1.55 times larger than the upper layer (100 to 120 m). Figure 8b demonstrates how wind speed changed unevenly for positive and negative power law exponents. Negative power law exponent cases only started evidencing decreasing wind speeds with height above 60 m, whereas positive
Wind direction
Direction and speed shear at the test site varied accordingly with time of day (Fig. 9). The correlation between both parameters is 0.9. Nighttime cases showed an evolving surface layer that does not reach equilibrium, as is depicted by consistently increasing directional shear across the rotor layer at an average rate of 0.0304
Diurnal cycle of mean wind direction and speed shear for wind speeds between cut-in and cut-out. Dashed vertical lines indicate sunrise and sunset times for 1 August 2013, the midpoint of the dataset analyzed here. The grey shaded region indicates the morning transition period (06:00–09:00 LT).
Of particular interest is the morning time period (from 06:00 to 09:00 LT – local time) which, according to the US Energy Information Administration (2019), experiences increasing electricity demand in the
US Midwest region. Wind shear presented its largest rate of change during this
time period (Fig. 9). At this time, nearly 50 % of the recorded data between cut-in and cut-out wind speeds were within 5 and 8 m s
Though speed and direction shear varied proportionally throughout the day,
they had opposite monotonical relationships with wind speed. As wind speed
increased, so did speed shear, but direction shear decreased (Fig. 10). Directional shear declined with increasing wind speed for both daytime and nighttime cases. While directional shear at night was generally larger than during the day, in both cases direction shear decreased at a median rate of around 0.0166
Direction (solid line) and speed (dotted line) wind shear variation with 80 m wind speed using each day's sunrise and sunset times of day.
Nighttime shear exceeded that during the day for wind speeds between cut-in
and rated speeds (Fig. 10). Median nighttime directional wind shear was at least 1.8 times as large as daytime cases for wind speeds between cut-in and rated speed. The highest percentage difference occurred near rated wind speeds, where nighttime directional shear was 3.5 times larger than that during the day. Median speed shear during the night was on average 3.2 times larger than during daytime and presented the largest differences near rated speeds (about 4 times larger for wind speeds between 8 and 11 m s
Large directional wind shear tended to occur at wind speeds below 8 m s
Directional wind shear probability density (solid lines) and cumulative (dashed lines) distributions for 1.5 m s
Speed shear distributions changed dramatically for wind speeds above and
below 6.5 m s
Speed shear probability density (solid lines) and cumulative (dashed lines) distributions for 1.5 m s
Both shear parameters were correlated for similar hub-height wind speed
regimes (Fig. 13). The correlation coefficient for increasing speed shear and direction shear values (veering and backing) is 0.9 and for decreasing direction shear and increasing speed shear (veering and backing) is
Mean speed and direction shear relationship for similar hub-height wind speed regimes (1.5 m s
We normalized each power measurement to quantify the role of wind direction
shear and speed shear on turbine power production for different wind speeds.
Normalized performance (
Mean normalized power production (turbines A–D) for all combinations of speed and direction shear that present more than 30 observations. The red line represents the
Small wind backing and small veering showed similar effects on turbine
performance (Fig. 14). Veering below 0.1
Because one of the main differences between the Wharton and Lundquist (2012b) and Vanderwende and Lundquist (2012) studies was the occurrence of directional wind shear at the different sites, we here examined the effect of the shear of wind direction on turbine performance. To separate the effect of speed shear from that of direction shear, we isolated turbine performance that transpired within a 0.1 power law exponent interval and segregated observations using the
Mean normalized power production (turbines A–D) of observations above and below the
Power variability can exert significant impact during morning hours, when
power demand tends to increase. Throughout all power production periods,
from 06:00 to 09:00 LT, 47.6 % of the dataset reported speed and
direction shear combinations above the
Mean normalized power production (turbines A–D) of observations above and below the
Wind shear at the test site showed more veering cases than backing cases (Fig. 7a) and a predominance of wind speed increasing with height (Fig. 7b), as would be expected from the balance between Coriolis, pressure gradient, and frictional forces in the atmospheric boundary layer (Holton and Hakim, 2013). Furthermore, the largest shear values occurred between 40 m and 60 m a.g.l. (Fig. 8), as would also be expected given that turbulent fluxes increase near the surface, causing larger wind vector rotation and speed reduction. However, cases where wind speed decreased between 40 and 120 m evidenced the greatest rate of change between 80 and 100 m above the surface (Fig. 8b). These observations usually took place at low wind speeds during the middle of the day, where a highly convective boundary layer produces near-zero shear in the lower rotor layer.
Shear also depended on time of day (Fig. 9). The observed diurnal pattern is consistent with daily radiative flux cycles. The advent of shortwave radiation from the sun at dawn drives convective air plumes from surface heating causing the largest rate of shear decrease. Rising air parcels transport air with similar zonal and meridional speed components across the rotor layer, decreasing wind shear. As the sun continues to heat the surface through the day, the convective atmosphere is strengthened, and wind direction and speed shear tend to stabilize. Once the shortwave radiative flux ceases at dusk, atmospheric stratification develops, evident from increasing shear values. A previous study in this same site found stable stratification to develop at 19:00 LT and strong veering and speed shear to develop after the evening transition (Lee and Lundquist, 2017). Median nighttime direction and speed (above 4 m s
Though speed and direction shear in the boundary layer have equivalent forcing mechanisms, they displayed opposite relationships with increasing wind speeds (Fig. 10). Convective conditions, typically with low speed shear, usually occurred at low wind speeds (Fig. 12), where large direction shear was more likely (Fig. 11). Large convective eddies cause a fluctuation of the meridional and zonal speed components (large direction shear), but mean horizontal wind speeds remain almost unchanged (small speed shear). Figure 12 suggests that a stratified layer, which entails large wind speed shear, primarily occurred near rated speeds. Decoupled laminar flow through the rotor layer results in low direction shear, whereas winds accelerate toward supergeostrophic speeds (large speed shear).
Nevertheless, a monotonic relationship between speed and direction shear existed for similar hub-height wind speed regimes (Fig. 13). As wind profiles evolved for constant hub-height speeds, both shear parameters developed congruently due to the force balance in the boundary layer. Large surface stress reduces wind speed near the ground and results in cross-isobaric flow toward low pressure. At higher altitudes above ground level, the surface stress is lower, and the wind is geostrophic. In between these heights, the variation of speed and direction with height is described by the Ekman spiral, where wind vectors must increase in magnitude and rotate clockwise (counterclockwise) in the Northern Hemisphere (Southern Hemisphere) to couple friction-driven surface winds with near-frictionless winds aloft (Stull, 1988). Further, when the surface stress decreases following radiative fluxes, inertia causes the wind to accelerate and the Coriolis force turns the wind vector clockwise (Northern Hemisphere) in time (Stull, 1988). The opposite case occurs with increasing surface stresses.
The combined effect of speed and direction shear on turbine performance
displayed a linear threshold (given in Eq. 5) that separates under- and
overperformance at this wind farm (Fig. 14). Several models have shown that power law exponents between 0 and 0.33 result in lower available power over the whole rotor area (e.g., Antoniou et al., 2009; Bardal et al., 2015). Also, as the wind vector turns with height the magnitude of the projected velocity decreases following a cosine function, causing a reduction in available power. Here, we found slight overperformance for wind shear combinations below the
Large wind veering combined with small speed shear resulted in wind turbine
underperformance (Fig. 14). In contrast, overperformance occurred for large speed shear and small changes in wind direction with height. Observations exceeding the
For a given value of directional shear, as quantified in 0.1
For a given value of speed shear, as quantified in 0.1 power law exponent
intervals, increasing the directional shear resulted in turbine power
depletion at this wind farm (Fig. 14). Normalized performance revealed a negative trend (around
Small wind backing was found to have similar effects as small wind veering. The change in energy flux through the rotor disk and turbine blades' efficiency appeared to be minor for these low direction shear conditions. Our dataset only evidenced statistically distinct power production between veering and backing for two speed shear ranges, suggesting the power asymmetries found by Walter et al. (2009) and Wagner et al. (2010) did not occur at these low shear conditions. Moreover, the small mean backing numerical values for these speed shear ranges indicate additional forcing mechanisms were in place for these underperformance observations. Not enough large backing observations were recorded to compare turbine performance against large veering atmospheric conditions.
In distinguishing the effects of high- and low-direction shear using the
When considering power law exponents between 0.2 and 0.3, we found direction
shear to exert a larger impact on power production in the middle of the
partial load regime than near cut-in or rated speeds (Fig. 15a). Most observations within this speed shear range took place between 6.5 and 8 m s
Focusing on a period of rapidly increasing electricity demand (06:00 to 09:00 LT) exposed the fact that directional shear's detrimental effects
preferentially occurred during this time. Mean direction and speed shear were 0.196
The substantial power reductions and number of cases affected by the change of wind direction with height in this wind farm make directional wind shear effects critical to consider in wind resource assessment, grid integration studies, and wind turbine control algorithm design. Large veering values affected turbine performance for small and large speed shear, suggesting that aerodynamic efficiency reductions dominate the increase in energy flux over the rotor disk caused by increasing speed shear values. In addition, the fact that large directional shear undermined power production here also provides an explanation for how turbine operation was undermined for stable atmospheric conditions in the Vanderwende and Lundquist (2012) study. Turbine overperformance for stratified channeled flow conditions in St. Martin et al. (2016) and Wharton and Lundquist (2012b) studies was likely augmented by low direction shear due to channeled flow in those regions.
The present work has provided insight into the impact of wind veer on
clockwise-rotating wind turbines' performance for different wind shear conditions. Recent simulations suggest that the direction of turbine
rotation interacts with wind veer to affect wake structures (Englberger et al., 2019). To understand the impact of the rotational direction of a wind turbine on performance; however, future field studies and simulations should incorporate counterclockwise-rotating wind turbines. Further work regarding directional wind shear in offshore locations should also be pursued. A preliminary wind resource assessment on the coast of Massachusetts by Bodini et al. (2019) demonstrated large changes in wind direction with height. Average values of 0.1
We have created an online archive for the lidar data used for creating lidar-only plots with a Readme.txt file explaining the data at
JKL conceived the initial concept and guided the data collection. MSG performed and extended the data analysis and wrote the initial manuscript draft. Both authors contributed to the discussion of the results and extensively revised the manuscript.
The authors declare that they have no conflict of interest.
The CWEX project was supported by the National Science Foundation under the State of Iowa EPSCoR grant 1101284. The role of the University of Colorado Boulder in CWEX-13 was supported by the National Renewable Energy Laboratory. The authors thank NextEra Energy for providing the wind turbine power data. This work was authored (in part) by the National Renewable Energy Laboratory, operated by Alliance for Sustainable Energy, LLC, for the US Department of Energy (DOE) under contract no. DE-AC36-08GO28308. Funding provided by the US Department of Energy Office of Energy Efficiency and Renewable Energy Wind Energy Technologies Office. The views expressed in the article do not necessarily represent the views of the DOE or the US Government. The US Government retains and the publisher, by accepting the article for publication, acknowledges that the US Government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this work, or allow others to do so, for US Government purposes.
The authors express appreciation to the University of Colorado and Iowa State University teams who deployed the lidars and surface flux stations during CWEX-13. Julie K. Lundquist was supported an agreement between the University of Colorado and NREL (grant no. APUP UGA-0-41026-125).
This paper was edited by Gerard J. W. van Bussel and reviewed by Rozenn Wagner and one anonymous referee.