<<

Characterizing Turbine Inflow and Wakes Daniel A. Pollak through comparison of SODAR and Met Tower SOARS® Summer 2011 Observations—A part of TWICS: The Turbine Science Research Mentors: Wake Inflow Characterization Study Julie Lundquist Writing and Communications Mentor: Abstract Wind offers an inexhaustible domestic Dave Hart energy source with minimal greenhouse gas Contributing Mentors: emissions. To maximize energy generation from Matthew Aitken, Cody Kirkpatrick, wind turbines it is essential to understand the Andy Clifton influence of inflow conditions and wakes on wind turbine energy production. In accordance with Acknowledgements The Significant this goal, the TWICS field campaign was conducted Opportunities in Atmospheric Research and in April and May 2011 at the National Renewable Science (SOARS) Program is managed by the Energy Lab’s (NREL) National Wind Technology University Corporation for Atmospheric Research Center (NWTC) in the complex terrain downwind (UCAR) with support from participating of Colorado’s Front Range mountains. TWICS universities. SOARS is funded by the National employed meteorological monitoring towers and Science Foundation, the National Oceanic and remote sensing systems to provide a three Atmospheric Administration (NOAA) Climate dimensional spatio-temporal illustration of the Program Office, the NOAA Oceans and Human inflow to and wake from a 2.3 MW turbine with a Health Initiative, the Center for Multi-Scale 100 meter rotor diameter. An important step in Modeling of Atmospheric Processes at Colorado analyzing the TWICS data was quantifying the State University, and the Cooperative Institute for performance of the different measurement Research in Environmental Sciences. SOARS is a devices that were used. This research compares partner project with Research Experience in Solid simultaneous measurements taken during TWICS Earth Science for Student (RESESS). by a Second Wind Triton Sodar and from the NREL M2 80 meter meteorological tower, which were located one kilometer apart. During the TWICS 1 Introduction campaign, we found strong linear correlations between wind speed measurements at 50 and 80 1.1 Motivation meters from the sodar and met tower. The high correlation suggests that flow is usually Harnessing the wind to power our homes, homogenous across the NWTC at time scales of businesses, and communities will spring the ten minutes, but that there are also occasional United States towards a cleaner energy future and periods of inhomogeneous flow. Wind speed will help reduce our nation’s carbon footprint. correlations were also found to vary with time of The wind resource in the United States is vast, day. This diurnal variation could represent free and is overhead, ready for use. In order to different conditions at the sodar and tower site encourage exploitation of this amazing resource, because of localized heating and turbulent mixing, the United States Department of Energy (DOE) has but may also be due to changes in sodar called for wind energy to generate 20% of performance as atmospheric stability changes America’s energy need by 2030. To answer this during the course of the day. Results from this call, there is an increased necessity to foster research will feed into future analysis of data strong understanding of the inflow characteristics collected during TWICS and help improve our of the atmosphere in all wind-prone landscapes understanding of turbine performance in the and the characteristics of the wakes that are atmospheric boundary layer. created by the wind turbines. In addition, it is important to understand the intricate interactions between the atmosphere and wind turbines, over all land types and terrain, and the effects that the by the wind industry. (Giordano 2010, Scott et al. subsequent wakes have on downstream turbines. 2010). This study will examine the performance The impetus for this study is based on the need to and consistency of SODAR observations in understand these concepts. complex terrain and in harsh weather conditions. Wind energy has grown rapidly in the (Borgeois et al 2007). United States over the last decade with more than 1,000 wind turbines currently in operation that 1.2 Basics can produce greater than 2.0 Mega-Watts (MW) of power. Additionally, half of new turbines Wind energy is, in essence, a type of solar installed in 2009 had hub heights ranging from 60- energy. Wind is caused by the uneven heating of 100 meters above ground level (AGL) and blades the Earth’s surface (and thus atmosphere) and is extending more than 40 meters out from the hub. modified by the various topography and (DOE Wind & Water Power Program 2010). As landforms that appear on Earth. This wind is a these turbines extend higher in to the atmosphere form of kinetic energy and can be harnessed it has become increasingly necessary to develop through wind turbines. These turbines turn the new ways to measure atmospheric conditions kinetic energy into mechanical energy which can around the turbine. In addition to limitations of then be transformed into electricity through a building meteorological towers that reach the full generator. Turbines come in various sizes, but it is height of modern utility-scale turbines, the data the large utility-sized turbines (over 1 megaWatt from conventional mounted on (MW)) that could help provide bulk power to the these towers only provide data for one specific power grid, propelling the United States towards point and do not show the spatial variability and its 20% wind energy by 2030 goal. (US-DOE-EERE reality of wind speed, wind direction and 2011). turbulence (Lundquist et al. 2009). As a result, development and use of remote sensing 1.3 TWICS technologies for this purpose have become increasingly important. The two remote sensing The Turbine Wake Inflow and systems used are the SOund Detection and Characteristic Study, or TWICS, was designed to Ranging (SODAR) and the LIght detection and help answer the DOE 20% by 2030 call by ranging (). Similar to a , the providing more information on how the SODAR and LIDAR send out radiation and can atmosphere inflow and wind turbine interact, and determine characteristics of the air through the the characterization of the turbulent structure of speed and time in which the pulses of radiation wind turbine wakes, especially over complex return via the Doppler effect. Instead of emitting terrain. The insight gathered by TWICS can radio waves as weather radars do, the SODAR and improve wind farm site placement and design, LIDAR utilize sound and light waves, respectively. create more robust turbines, and provide These remote sensing techniques allow important input into mesoscale wind models. This for concise measurement of the volume that in turn will lead to more energy efficient wind intersects the turbine inflow or wake. The farms and will increase production and lower the scanning lidar makes it possible to capture these cost of this wind-generated energy. The goals of spatial temporal characteristics of turbine wakes. the project were to 1) assess the structure and (Lundquist et al. 2009; Harris and Hand, 2006). To variability of wakes from multi-MW turbines, 2) to increase confidence from the data of these assess how well mesoscale models perform in remote sensing systems, it is important to prediction turbine inflow in complex terrain, and determine the reliability and accuracy of their 3) To assess how well turbine wake models measurements, especially if they are to replace represent observed structure and variability of the conventional met tower systems that remain wakes (Lundquist et al 2009). the only widely accepted form of measurements The TWICS campaign took place at the many of these areas have great potential for wind National Wind Technology Center (NWTC), part of power generation. the National Renewable Energy Laboratory In 2008, a study sponsored by the (NREL), just south of Boulder, Colorado. The European Union was launched in a hilly location in turbine of interest was a 2.3 MegaWatt (MW) Bosnia-Herzegovina to increase knowledge of the wind turbine which stands 80 meters high and has turbulence intensity and inflow wind blades that extend approximately 50 meters from characteristics in complex terrain. This study took the hub (Figure 1). The TWICS team gathered data wind measurements carried out with a 30 meter from NREL’s M2 met tower, SecondWind’s Triton met tower, a SODAR and a LIDAR and is much like Sodar, NOAA’s (National Oceanic and Atmospheric the present study. (Borgeois et al. 2009) Administration) high resolution Doppler lidar The Bosnia-Herzegovina site, Maligrad, (HRDL) and CU’s (University of Colorado) was selected because it is known for its variability Windcube lidar. All were located upwind of the in wind conditions and its widely fluctuating turbine (Figure 2). During the months of April and meteorological conditions. The main prevailing May 2011, data were taken during times of high- wind that drives these conditions is a katabatic magnitude prevailing wind events,[which here is wind called the Bora wind (in Bosnia and Croatia) 292 degrees] called herein intensive observing and has characteristics of high and large periods (IOPs). Four IOPs occurred during the turbulence intensities. The two motivations of the Spring 2011 study. Before diving into the study were as follows: (1) To assess the intricacies of TWICS, a description of the project’s performance of the SODAR and LIDAR under two foci will be provided in addition to results various and intense wind and meteorological from previous studies. conditions and to (2) analyze the suitability of this site for a future wind farm and determine the 1.4 Characterizing Atmospheric Inflow wind turbine design that would best fit under the local wind conditions (Borgeois et al. 2009). There are two key concepts that with Data were taken from the 30 meter met better understanding can help to create more tower, SODAR and LIDAR for 18 days in November energy efficient wind farms that are located and December of 2007. All three measured wind optimally. The first is understanding the speed and in addition, the SODAR and LIDAR atmospheric inflow characteristics incident on the measured vertical wind profiles and turbulence wind turbine. Give that many wind turbines now intensities. It was found that the SODAR and reach heights of 40-story buildings and are thus LIDAR had cohesive agreement and were able to located in a more complicated and variable part of take high-quality data up to 100m above ground the atmosphere, this need has never been level during most times (Figure 3). An exception greater. With increasing turbine heights, the in- was during events where winds were above 20 situ measurements made by sonic anemometers m/s due to noise interference. Turbulence on meteorological (met) towers become less intensities (TI) calculated from the SODAR and valuable as met tower heights do not exceed LIDAR had biases with winds greater than 15 m/s those of the turbine. As a result, remote sensing but TI were almost uniform between 30 and 100 systems such as the SODAR and LIDAR are used to m ABL. A summary of available data of the SODAR “see” higher into the atmosphere, supplementing measurements at each height level at Maligrad is the sonic measurements. illustrated in Figure 4. (Borgeois 2008) Comparing data from the met tower, SODAR and Similar to the Mostar study, the TWICS LIDAR can assist in determining the accuracy and measurement campaign took place in a location calibration needed for the remote sensing devices. with complex terrain with a strong prevailing wind For wind turbines placed in mountainous or hilly and myriad of meteorological conditions. Just locations, it is advantageous to do such a study, as downwind of the Rocky Mountains, the site experiences many of the meteorological phenomena including downslope wind storms, prevailing wind (Figure 8). In addition, three microbursts, and complex stability effects such as oblique angles to this ER were defined on either density currents and drainage flows from Eldorado side of the ER (-15,-10,-5,5,10,15 degrees). Each canyon located west of the National Wind of these wind farms had eight columns that were Technology Center (NWTC) in Boulder, Colorado. essentially orthogonal to the prevailing wind These unique conditions make the NWTC site an direction. They found that the power deficit appropriate site for a lidar-based study of the induced by wakes reached a minimum threshold turbine-atmosphere interaction (Lundquist et al on downstream turbines. This minimum deficit 2009). value is termed the wake asymptote and in the Barthelmie and Jensen (2010) study was found to 1.5 Turbine Wakes be 60% of the free stream power at the first turbine (Figure 9). This percentage is called the The second concept involves the state of normalized power and is used to describe the the wind directly after it passes by a wind turbine. power deficit caused by these wakes. Simply Since kinetic energy is pulled out of the wind from speaking, normalized power is the percentage of turbines, the wind speed directly downstream of a power output actually generated compared with turbine will be less than the original inflow speed. the maximum possible power. While consistent, The wakes behind the turbine are in essence the column in which the threshold was reached waves of wind speed deficit and are, “similar to depended on the ER value. When looking down the watery wakes behind boats” (NOAA News the ER, the power minimum was reached at the 2011) The wakes expand in both the vertical and turbine in column 2. “For higher angle wind lateral directions (Figure 5). In the case of a directions, the power reduction at the second single, isolated wind turbine, the detrimental turbine was large, but reductions continued for effects of these wakes will be unimportant as each subsequent turbine” (until asymptote). there are no turbines downstream that would (Barthelmie and Pryor et al. 2010) need to harness the wind. These wakes expand Wind farms can be located both on land downstream and eventually dissipate. A study and offshore. Most recent studies have focused completed in Northern Germany in July 2009 used on understanding turbine wakes at large wind lidar to visualize the magnitude and directionality farms that are located offshore. This is largely a of wind turbine wakes as they propagate result of the less turbulent flow that is often seen downstream. Behind these 5 MW turbines wakes, over the shallow marine boundary layer and the (wind speed deficits) extended downstream 400 possibility to simplify the flow moving into the meters and about 100 meters up vertically (Figure wind farm. Given the large potential for wind 6a,b) (Kasler et al. 2010). This study was energy sequestration in mountainous areas, it is conducted in a relatively flat area and thus there is important to note and characterize wakes and still a need to examine wake characterisitcs in atmospheric inflow first for individual turbines in complex terrain. In addition, it was important for such terrain. this study as it fostered more knowledge that could be applied towards the study of multi-MW 1.6 Instrumentation & Current Study wind farms with numerous turbines. The effects of wakes at these wind farms are evident and Having a diverse group of instruments to important to understand. measure characteristics of wind and turbulence is Many previous studies have focused on beneficial and essential to further the large offshore wind farms (Figure 7). In one understanding of atmospheric inflow and turbine study, two offshore Danish wind farms were wakes around wind farms. There are examined—the Nysted and Horns Rev farms. conventional, in-situ measurements from Within each wind farm, an exact row (ER) was meteorological observations towers (the M2 met defined as a row of turbines that follow the tower) and remote measurements from Correlations of wind speed and wind instruments such as sodars and . direction measurements between the M2 met A SOund Detection and Ranging (SODAR) tower and Triton sodar were calculated generally is much like a Doppler radar except that sound using 37-day averages of data and also more waves are used instead of radio waves for particle specifically during the three intensive observing detection. SODAR systems are used to measure periods (IOPs) that were seen during the TWICS wind speeds at multiple heights above ground field campaign. More details will be discussed in level and are therefore help characterize the the forthcoming methods section. lowest layer of the atmosphere—the planetary boundary layer (PBL) and surface layer (SL). They 2 Methods are commonly known as wind profilers. A lidar is very similar to a sodar and radar, but in this case The field component of the TWICS light is employed. Due to light’s shorter campaign took place in April and May 2011 at the wavelengths it can use aerosol particles as National Wind Technology Center in South atmospheric scattering targets. By observing the Boulder, Colorado. This site is located on the Doppler shift of the light scattered by aerosols, foothills of the Rocky Mountains and is subject to the lidar can remotely measure air particles. a variety of weather conditions and phenomena. Sodar measurements are not always able The Triton Sodar, M2 met tower, HDRL Lidar, and to produce complete time series due to factors Windcube lidar were employed in the TWICS such as surrounding noise, clouds, atmospheric campaign to observe inflow characteristics and stability or precipitation that in turn, leads to data indicate the wakes of a 2.3 MW wind turbine. gaps. These effects are filtered out (Borgeois Data were collected during three instances of 2008) Both sodars and lidars are useful in the sustained strong north and north-westerly flow, study of wind inflow and turbine wakes in that hereafter termed intensive observing periods they can, especially in addition to meteorological (IOPs). During these events both in-situ and towers, provide a four dimensional (time being remote observations were taken, but on different the fourth dimension) view. temporal scales, thus re-formatting was In this branch research of TWICS, focus necessary. This will be discussed in a later section. will be given to comparing data from the M2 While both lidars were only operated meteorological tower and the Triton sodar. during IOP events, the M2 met tower and Triton Sodars offer a portable way to characterize wind sodar are continuously operating instruments at speed, direction and turbulence intensity at the NWTC. Therefore it was possible to perform multiple levels above the ground unlike their non- an analysis of both long term trends and IOP stationary met tower counterparts. The goal of events. The 37-day period that was examined for this research is two-fold: 1) to validate Triton the correlation analysis was from April 1 – May 7, sodar data from M2 met tower data using the met 2011. The three IOPs took place on April 14-15, tower as the accepted NWTC site inflow April 22, and April 27-28. instrument, and 2) to characterize turbine inflow using both instruments (Figure 10). A successful 2.1 Instrumentation validation of the sodar data will provide confidence that sodars offer a viable alternative to 2.1.1 2.3 MW Wind Turbine met towers in locations without access to such tower data. Additionally, successful validation will A 2.3 MW turbine (Figure 11) was selected provide a robust dataset that contributes to the for use in TWICS because of its relevance in overall portfolio of measurements included in the industrial-style wind energy applications and analysis within the Turbine Wake Inflow and because observational studies of turbines greater Characterization Study campaign. than 1.3 MW have been few and far between. The turbine hub stands 80 meters tall and has blades of approximately 50 meters. Given that these and has a large diurnal temperature range and turbines are used onshore and offshore, the realm wide range of weather conditions. Figure 12 of applications from the results of this study could provides an elevation profile of the site and be substantial. (Lundquist et al. 2009) displays the terrain located west of the NWTC. To the west of the site is the Eldorado Canyon and 2.1.2 The Triton Sodar: often times the wind is funneled through the canyon as air moves over the continental divide. The Triton sodar is a state-of-the art Given this, an important selection factor for this system that was developed by Second Wind Inc. site is its strong 290 degree prevailing wind located in Somerville, Massachusetts. Triton is a [WNW]. The reliability of this wind provides small, two feet by three feet, device (Figure 11) appropriate siting of the lidar and sodar for that uses, “a hexagonal speaker array to efficiently consistant observation of both the turbine wake focus sound beams to improve signal-to-noise and turbine inflow conditions. The enlarged ratio accuracy and decrease disruption” (Scott et profile of the site in Figure 12 shows the two al. 2010). Triton collects data every few seconds instruments that were used in this study. As the but only permanently records one averaged value diagram depicts the sodar is oriented looking up every ten minutes for data from ten heights and can gather data at only one spatial point. The between 40 meters and 200 meters above ground instruments are located 700 meters apart and are level (40,60,80,100,120, 140,160,180,200). Data upstream of the 2.3 MW turbine. returned at each height include wind speed (both horizontally and vertically), wind direction, and 2.2 Data Filtering turbulence intensity. A summary of the Triton specifics is illustrated in Table 1 below. (Scott et al. To fulfill our research project goal of 2010; Lundquist et al. 2009) correlating observations between M2 and Triton it was imperative to quality control the data so that 2.1.3 The M2 Meteorological Tower only relevant data points were taken into consideration in the analysis. For this reason it The 80 meter M2 meteorological tower was necessary to implement some filters that are (Figure 11) located at the western end of the discussed below and listed in Figure 13. Given that NWTC site acts as the accepted site inflow the M2 met tower was used as the accepted observational instrument and thus the accepted inflow values for the study, most of the filters tool for comparison. Mainly to check the stability were applied to times when M2 did not surpass of the atmospheric layer, the tower takes the established filter specifications. When an measurements of temperature and dew point at observation at M2 was filtered out, the Triton three different heights: 2m,50m and 80m. data was also removed for the same time period Additionally there are six anemometers that take regardless of its value. Only data in time periods wind speed and direction observations at 2m, 5m, where the wind direction was between 225 and 10m, 20m, 50m, 80m. Observations of wind 360 degrees was accepted for further analysis, speed variances and turbulence are also taken. focusing on events when the M2 and Triton were (Lundquist et al. 2009) upwind of the turbine. To create the plots of wind direction, only time periods in which the wind speed was greater than 3.5 meters per second (just under eight miles per hour) were used in 2.2 Site Layout and Typical Conditions analysis. Calm winds have insignificant wind direction values, as they are subject to many The study was performed at the National micro-scale processes. Wind Technology Center on Colorado’s Front The one filter that was initiated on raw Range in South Boulder. The climate is semi-arid Triton data involved the filtering scheme specified by Second Wind (Walls 2008). This was done with the goal of creating a reliable and scientifically 3.1 Triton Quality with Height sound dataset in which to compare to M2. For Triton data analysis Second Wind recommended a Second Wind puts out a quality factor for wind speed quality factor of greater than 95%. each Triton reading at all heights. This quality This quality factor is a function of the signal to factor is largely reflective on the signal-to-noise noise ratio and was developed by Second Wind. ratio of the acoustic pulses sent out from the Furthermore, a wind speed quality factor of Triton. The signal-to-noise ratio is simply a ratio greater than 95% will be dubbed high quality data. of the desired signal to the background noise Examination of data quality by height is discussed present, or, a ratio of the number of beams send in the results section of this paper. Similar to how out to the number of beams that arrive back at the Triton data was filtered out based on M2 the sodar. For each of the three intensive observations during same time periods, M2 data observing periods (IOPs) a filled contour plot was removed alongside non-high quality Triton illustrating Triton’s quality by height were made data. (Figures 14a,b,c). The dark red color indicates Triton quality of greater than 90 percent and the 2.3 Data Preparation for Analysis blues indicate quality of lower than 30 percent. To perform data analysis, Second Wind recommends M2 data was recorded once per minute. a data quality value of greater than 95 percent To compare M2 and Triton data, it was necessary (CITATION). By looking at these plots, most to average all ten, one-minute observations for heights from 80 meters and below have >95 every ten minute period to create one averaged percent quality such that an analysis can be made. value. Averaging wind directions can prove to be 90 percent quality extends roughly up to 100 a dangerous task given the wrap-around factor in meters. Looking at time-height cross degrees of wind speed measurements. To section plots of raw triton wind speed mitigate this problem, whenever wind directions measurements could have led to some irrational were below 40 degrees, we added 360 degrees to conclusions. Figure 15 displays such a chart for the value. The values were then averaged over IOP #1. Looking at the raw data you might the 10-minute period. If the resulting average hypothesize there to be a low level jet. The black value was greater than 360 degrees, then a lines indicate the upper bounds of 95 and 50 subtraction of 360 degrees was performed. This percent quality. Therefore it is impossible to process prevented erroneous wind direction consider the low-level jet hypothesis. values. Given the high Triton data quality at these To ensure consistency across both heights, it was conceivable to do a correlation datasets, it was also vital to use the same time analysis of the 50 and 80 meter observations zone so that there would not be a data offset. The between the M2 met tower and the Triton sodar. Triton sodar data was in Coordinated Universal In some cases though, 90 and 95 percent quality is Time (UTC) which during the summer months is lower than the aforementioned heights. For these seven hours ahead of local mountain daylight times of lower quality, the filtering process time. Data from M2 was in local mountain time discussed above excludes such data. To gain an and was the time zone we selected for use in this insight into the overall performance of the Triton study. We therefore converted the UTC time in comparison to the Sodar, we analyze 37 days- stamp on the Triton data to mountain daylight worth of data, from April 1 through May 7, 2011. time. After the data filtering and quality control process it was acceptable to begin the analysis 3.2 Long-term correlations process. Scatterplots comparing observations of 3 Results & Discussion wind speed and wind direction from both the M2 and Triton at both 50 and 80 meters were reduced number of data points that passed generated to gauge the overall performance and through the data filtering process. At 80 meters, correlation between the two instruments. For the especially during certain parts of the day, it is entire 37 day period, there were slightly over possible that the boundary layer is stratified and 5,300 raw, unfiltered data points. We filtered the decoupled. Sodars have trouble in these air data so that only M2 observations of wind masses and thus the two instruments could be directions between 225 degrees (SW) and 360 detecting different air masses and thus since degrees(N) were plotted. The data is only filtered further from the surface, there is more decoupling for M2 given that it is the baseline inflow met between observations from the M2 vs. Triton. tower for the National Wind Technology Center site. This filter is used since this is generally the prevailing wind and also when both instruments 3.2.2 Overall Wind Direction Correlations are upwind of the turbine. An additional filter was used when considering wind direction. The filter A strong linear relationship is noticed in discarded wind direction data for when winds the wind direction correlation plots between were below 3.5 meters per second (just below Triton and M2 (Figures 19,20), however there are eight miles per hour). It was implemented to a few clusters of outliers. Currently, R values for prevent calm wind and thus inappropriate wind 50 meter and 80 meter wind direction were 0.595 directions from being included in the analysis. and 0.687 respectively. We now believe that After filtering there were 2683 data points taken these clusters and subsequently lower R values into consideration in the 50 meter wind speed are a result of anomalous beam propagation and correlation plot and 1853 data points in the 80 we are currently working to quantify and correct meter wind speed correlation plot. The number these errors. (Note the different axis scales; of wind direction points are subject to further different scales are used given the 225-360 degree analysis given the need for another filter/quality filter used for the M2 observations). The control of the data. development of further quality controlled plots should yield better correlations. 3.2.1 Overall Wind Speed Correlations These scatterplots, that take all filtered Scatterplots reveal that wind speed data points into consideration, provide a good correlations between the Triton sodar and M2 overview of Triton’s general success when met tower were good with a Pearson correlation compared to the M2 met tower located 700 coefficient (R) value at 50 meters of 0.9635 (Figure meters away. To gain insight into intricacies 16) and R value at 80 meters of 0.9391 (Figure 17). behind these general trends, correlations were As illustrated on these wind speed correlation made that evaluated diurnal changes in wind plots, there are strongly visible linear relationships speed correlation based on the time of day. Also between the two datasets but also a rather large wind speed correlation based on wind direction spread expanding around the linear trend line. were determined. This variability is quantified by the statistical root mean squared error that, for 50 and 80 meter speed, were 1.56 and 1.81 respectively. 3.2.3 Wind Speed Correlations by Time of Day: A similar met tower and sodar correlation study at a wind farm in Texas yielded a much For each day of the study period (37 days) more tight linear fit (Scott et al. 2010)(Figure 18) we plotted the data from each hour, in local time. The slightly lower correlation and higher Figure 21 displays the wind speed correlation RMSE for 80 meter wind speed is possibly the values by hour at 50 (dashed line) and 80 meters result of higher variability in this part of the (solid line). The left y-axis is the R correlation boundary layer and also potentially due to the value by hour and the right y-axis represents the is a clear and substantial difference in wind number of data points for each of the time period direction as you go from 50 meters to 80 meters. bins listed on the x-axis. The number of data This potentially indicates wind veer. points is represented by the red (50m) and blue (80m) bars. The two most noticeable features in 3.3 Intensive Observing Periods the graph are 1) the lower correlations seen in 80 meter speed versus 50 meter speed, and 2) the The three intensive observing periods lower correlations seen in the early and mid- occurred on April 14-16, April 22 and April 27-28, morning hours and the higher correlations seen in 2011. They were characterized by strong the afternoon. Possible explanations for this northwesterly or westerly flow onto the NWTC trend include diurnal changes in atmospheric site. For these events we plotted wind speed and stability, average wind speeds, and number of wind direction at 50 and 80 meters as seen by filtered data points. both M2 and Triton. We found similar trends The atmospheric boundary layer tends to across each of the IOP events. In figure 24 we use be stably stratified in the morning which can IOP#3, April 27-28, as an example. The first plot increase signal to noise errors from the Triton, simply shows the wind direction measured by thus lowering its performance. Additionally, this each instrument at each height. As expected the could also be one reason behind the lower winds centered around westerly and quantity of filtered data points during the morning northwesterly. and late evening hours. Figures 22a,b show that Figures 25 and 26 show IOP #3 wind speed the average wind speed is also less at both 50 and at 50 and 80m for both instruments. The black 80 meters during the morning which could line and errorbars represent M2 measurements potentially allow for more diversion of the flow and standard deviation and the red line indicates from directly across the site. Lastly, it is also Triton measurements. At 50 meters Triton stays possible that during the morning hours, as the sun within the M2 standard deviation error bars. At begins heating the ground, differing conditions at 80 meters though, Triton has slightly higher the M2 and Triton could be exhibited as a result of measurements in comparison to M2 and is often localized heating and turbulent mixing. outside the upper bound of the error-bars. This indicates that at 80 meters the triton 3.2.4 Wind Speed Correlations by Wind Direction overestimates the wind speed.

In addition to looking at correlations at 4 Conclusions & Next Steps various times of the day it was also of interest to see any changes in correlation based on wind The TWICS project is helping the United direction bins of ten degrees each from 230 States reach its 20% wind energy by 2030 goal degrees to 360 degrees. by researching and improving the Figure 23 shows the Pearson R correlation understanding of turbine inflow and wakes, and value of wind speeds within each ten degree this sub-study is contributing to the TWICS goals sector and also the number of points that were by determining which instruments provide within each of the 13 bins. Overall, correlations insightful datasets. Through a validation study of within each of these bins were very good with all the Triton sodar to the 80-meter M2 R values above 0.95. The best correlations were meteorological tower on the National Wind realized for wind directions of around 260 to 320 Technology Center site, we found that the sodar degrees which surround the 292 degree does prove to be a valuable dataset. directional angle between the Triton and M2. At this site, downstream of Colorado’s Interesting to note is that the peak in number of Front Range, the Triton had reduced quality above points by bin also coincides with the bin with 80 meters. Even though this line demarking high highest correlations. Also interesting is that there quality is lower than in some previous studies Barthelmie, R. J., Jensen, L. E., 2010: Evaluation of (Borgeois et al 2009), it still allowed for a wind farm efficiency and wind turbine comparison of wind speed and wind direction data wakes at the Nysted offshore wind farm. at 50 and 80 meters from both Triton and M2. Wind Energy, 13, 573-586, Strong linear correlations of wind speed, R values doi:10.1002/we.408. greater than 0.93, existed at 50 and 80 meters. Wind speed correlations were found to Barthelmie, R. J., Pryor, S. C. et al, 2010: vary with time of the day with weaker correlations Quantifying the Impact of Wind Turbine occurring the morning and stronger correlations in Wakes on Power Output at Offshore Wind the afternoon. Different conditions at Triton and Farms. Journal of Atmospheric and M2 from localized heating and turbulent mixing Oceanic Technology, 27, could be one factor behind these trends. Another doi:10.1175/2010JTECHA1398.1 potential explanation for this trend is that there are changes in Triton performance due to diurnal Bourgeois, S., 2008: Documentation and results of changes in atmospheric stability; there is the SODAR and LIDAR Measurements at decreased data quality and consequently there the Maligrad site in Bosnia and are less data points that can be used for the Herzegovina: Measurement Campaign correlation analysis. from October 30th 2007 to February 4th There were also decent linear correlations 2008 (SODAR) and November 20th 2007 to of wind direction but outliers were present in December 10th 2007 (LIDAR). METEOTEST, clusters and further quality control is required to 22pp. determine the validity of these specific findings. Looking at the IOPs there did seem to be an Bourgeois, S., Winkelmeier, H., Meissner, C., 2009: overestimation of wind speed at 80 meters by Turbulence intensity and high wind speeds Triton. However, it is important to keep in mind above complex terrain: measurements that the instruments were 700 meters apart and and CFD-modelling, 4pp. were thus subject to possible differences due to localized heating and turbulent mixing. DOE Wind and Water Power Program, Some next steps within this research Wind Power Today, 2010, Downloaded include: examining reasons for the weak wind from direction correlation, examining a larger data set http://www1.eere.energy.gov/windandhy (more time periods) to provide more robust dro/pdfs/47531.pdf correlation results for the time of day and wind direction bin correlations, looking at other Giordano, S., 2010: Eliminating Uncertainty with available metrics such as turbulence intensity and Sodar. Wind Systems Magazine, February atmospheric stability , and lastly comparing Triton 2010, 50-52. and M2 data with the Windcube lidar and NOAA High Resolution Doppler Lidar that were also Harris, M., Hand, M., Wright, A., 2006: operated during TWICS. Lidar for Turbine Control. NREL Technical Overall, correlations of wind speed Report /TP-500-39154, 55pp. between the M2 met tower and Triton sodar were good and validate that the Triton dataset is a valuable piece of information within the TWICS Kasler, Y., Rahm, S., Simmet, R., Kuhn, M., 2010: project and would also be useful in future Wake Measurements of a Multi-MW Wind research and industry use. Turbine with Coherent Long-Range Pulsed Doppler Wind Lidar, Journal of Atmospheric and Oceanic Technology, Works Cited Sept 2010, 1529-1532, doi:10.1175/2010JTECHA1483.1.

Kelley, N.D., Jonkman, B. J., Scott, G.N., Pichugina, Y.L., 2007: Comparing Pulsed Doppler LIDAR with SODAR and Direct Measurements for Wind Assessment. Proc. AWEA WindPower 2007 Conference and Exhibition, Los Angeles, California, National Renewable Energy Lab, 21pp.

Lundquist, J., Banta, R., Pichugina, Y., Kelley, N., 2009: Integration of turbine inflow and wake observations from a 2-micron lidar into a wind energy forecasting model- A Proposal in response to the DOE call from EERE/WHTP on 20% Wind by 2030: Overcoming the Challenges. LLNL/NOAA/NREL, 33pp.

Scott, G., Elliott, D., Schwartz, M., 2010: Comparison of Second Wind Triton Data with Meteorological Tower Measurements. NREL Technical Report /TP-550-47429, 12 pp.

U.S. Department of Energy – Energy Efficiency and Renewable Energy, Wind and Water Power Program, 2011: How Wind Turbines Work. [Available online at http://www1.eere.energy.gov/windandhy dro/wind_how.html]

Walls, September 2008: Viability of Sodar for Long-Term Research Assessment, Wind Tech International, 4, 6.

Wharton, S., Lundquist, J., 2011: Assessing atmospheric stability and its impacts on rotor-disk wind characteristics at an onshore wind farm  submitted

Figures: Figure 1 Figure 3: : Time series of the data availability of the SODAR measurements at Maligrad for each measurement level. Plausible data are displayed in green, rejected values in red.

A turbine at the National Wind Technology Center south of Boulder, Colo. High Resolution (Credit: CIRES) Figure 4: : Amount of available data of the SODAR measurements at each level at Maligrad Figure 2:

Figure 5: Figure 6:

Normally invisible, wind wind wakes take shape in the clouds behind the Horns Rev offshore wind farm west of Denmark. High Resolution (Credit: Vattenfall

(a) Results for the LOS component of the wind vector for an elevation scan through the rotor blades at night. The wind was blowing from the northeast parallel to the laser beam. Wind direction and position of the wind turbine are indicated in the figure. The white vertical line indicates the rotor disc. (b) Azimuth scan at night covering the ambient wind field of several wind turbines, including the M5000-1 prototype. The vertical white lines indicate the position of the rotor discs of M5000-1 and M5000-2.

the case studies examined here. ER denotes a direct down the row wind direction with minimum distance between columns of turbines.

Figure 7:

Figure 8:

Layout of the Nysted wind farm. (bottom) Layout of the Horns Rev wind farm. Gray lines shown on the two wind farm layouts are the directions used for

Figure 9:

Figure 10:

Figure 11:

Figure 12:

Table 1:

Table 1:

Figure 13:

Figures 14a,b,c: Triton quality with height

Figure 15:

Figure 16:

Fi

Figure 17:

Figure 18: Texas Case Study of Triton vs. Met Scott, Elliott, Schwartz 2010

Figure 19:

Figure 20:

Figure 21:

Figure 22a,b:

Figure 23:

Figure 24:

Figure 25:

Figure 26: