Professional Documents
Culture Documents
Small-Scale Wind Turbine
Small-Scale Wind Turbine
Introduction section. Added to this limitation are the difficulties that result
from both geometric scaling and Reynolds number matching. In
Small-scale wind turbines are a fast growing market in wind
fact, Serpa states that wind tunnels have been of limited value in
energy generation. According to the American Wind Energy
wind turbine rotor development except for the gathering airfoil lift
Association (AWEA), 54 small wind turbine models are offered
and drag data [5]. Challenges in scaling make testing in a small
commercially in the U.S. and are used on homes, schools, and
wind tunnel difficult; however, much can still be learned from
commercial and industrial facilities to name a few applications
testing small-scale wind turbine systems. For the small- and
[1]. By the end of 2012, AWEA projected that more than 150,000
microwind turbines, there exists the possibility of scaling both the
small wind turbines would be installed in the U.S. These wind tur-
geometry and the Reynolds number in wind tunnel testing.
bines are generally horizontal axis wind turbines (HAWTs) with
One of the important features to capture when designing small
fixed pitch blades and a direct drive DC generator. Domestically,
wind turbines is that of flow separation and its influence on airfoil
small wind turbines are defined as wind turbines with a capacity
performance data. This can occur at high speeds on a stall-
of 100 kW and lower. Along with this growth in small wind tur-
controlled turbine and on wind turbines operating at low Reynolds
bine installations is the growing trend toward distributed energy
numbers (usually 100,000 and below), especially near the hub.
generation. According to the U.S. Department of Energy, small
Generally speaking, the lack of accuracy of the airfoil data has the
wind turbines accounted for approximately 10.5% of all distrib-
potential to be the largest source of error in design, especially for
uted wind energy or 18.4-MW in 2012 [2]. The year 2013 was
the lower Reynolds number ranges.
expected to be even higher. In the UK, the small wind turbine cat-
For higher Reynolds numbers, usually taken to be greater than
egory is further divided into microwind (0–1.5 kW), small wind
500,000, often airfoil data are generated using computational 2D
(1.5–50 kW), and medium wind (50–500 kW). The UK experi-
models such as XFOIL, a shareware code developed by Drela [6],
enced a 21% increase in the number of installed medium and
and PROFIL, a commercial design code developed by Eppler [7].
small wind turbines in 2012 when compared with 2013 [3] with
At these speeds, the simulations are generally adequate as flow
the most significant increase in the small wind category. World-
will stay attached to the airfoil. For Reynolds numbers below
wide, at the end of 2012, there were 806,000 registered small
100,000, as experienced on small- and microwind turbines, flow
wind turbines, which is a 10% increase over the previous year [4].
separation will occur especially near the blade hub where the rota-
The growth is predominantly in China, the United States, and the
tional speed is lower. Figure 1 shows how airfoil performance,
UK, with China having 70% of the market total. HAWTs account
both lift and drag coefficients, can drastically change with
for the majority of the new wind turbines installed. For the focus
decreasing Reynolds numbers. There is a lack of appropriate
of wind tunnel testing in the context of this paper, only wind
experimental data to use in wind turbine design under these condi-
turbines of the residential size, 1–10 kW with fixed pitch (stall
tions, which is evident in the modeling codes [9]. Predictions of
controlled) and a direct DC generator, will be considered.
performance from existing models can range from 25% to 175%
of the measured values primarily due to the different approaches
to using the 2D airfoil data in the modeling codes, such as blade
Importance of Wind Tunnel Testing. It is important for the
element momentum theory (BEMT) [10]. This variation in per-
future of wind energy that new wind turbine designs be as effi-
formance is dependent on the quality of the 2D airfoil data avail-
cient as possible, to capture as much energy as feasible given the
able from either wind tunnel testing or computational design
constraints of the design. To achieve this efficiency, it is necessary
codes such as XFOIL. The National Renewable Energy Labora-
to involve wind tunnel testing. This testing should include airfoil
tory (NREL) tested their 10 m wind turbine rotor in the NASA
testing for basic lift and drag data as well as the testing of the
Ames wind tunnel and acquired data on 1700 different test condi-
wind turbine system (blades and generator) to optimize efficiency.
tions [10]. This data were used to validate the predictive capabil-
Full wind turbine systems are not usually tested in wind tunnels
ities of various codes available to wind turbine blade designers.
due to their large size when compared to the wind tunnel test
After conducting a “blind comparison” of codes with their data,
they concluded that the results were not favorable with wide var-
Contributed by the Advanced Energy Systems Division of ASME for publication
in the JOURNAL OF ENERGY RESOURCES TECHNOLOGY. Manuscript received December
iations between various code predictions. The conditions of the
11, 2014; final manuscript received May 6, 2015; published online June 1, 2015. flow were such that the models were expected to accurately
Assoc. Editor: Ryo Amano. capture the flow characteristics. This implies that more accurate
experimental data are necessary to further the predictive nature of that will be used in wind turbine blade design. Experimental test-
the wind turbine models and to improve the design process, partic- ing by the National Advisory Committee for Aeronautics (NACA)
ularly for low Reynolds number cases. in the 1930 s resulted in the family of airfoils, typically used in the
design of aircraft [28,29]. Early wind turbines began using these
NACA airfoil sections which were considered appropriate because
Overview of Wind Tunnel Testing. Wind turbine testing of of their laminar flow characteristics. For BEMT design, operating
actual systems in a wind tunnel is not usually possible because the at or near the maximum L/D ratio (minimum drag for the lift pro-
large rotor diameters are not physically able to fit into a wind tun- duced) is desirable. From a plot of CL versus CD, generated either
nel or, if it does fit, the rotor disk could cause an unacceptably from wind tunnel data or theoretically determined, it is possible to
large blockage. It is possible to test fixed, full-scale wind turbine locate the maximum L/D ratio directly and, with this value, find the
systems outside a wind tunnel at locations such as the NREL’s appropriate angle of attack for the wind turbine blade design. The
National Wind Technology Center in Golden, CO [11], but this is surface finish of the experimentally tested airfoils must be very
costly. Blanch tested a 3 -m diameter wind turbine mounted to a smooth as an increase in surface roughness and/or freestream turbu-
trailer by towing the trailer along an airport runway [9]. He lence in the experiments can cause differences in the data collected
noticed some variability in the data due to differing environmental when compared with the actual wind turbine performance. Addi-
conditions. In a wind tunnel, both the wind speed and quality are tionally, the blockage corrections applied to the data and the mea-
carefully controlled, eliminating the variability. The largest wind surement methods used in wind tunnel tests can affect the data [9].
turbine testing tunnel in the literature is the NASA Ames More attention is being given to designing airfoils that are
Research Center’s 24.4 m (80 ft) 36.6 m (120 ft) wind tunnel appropriate for wind turbines. Figure 2 shows data for a laminar
[10]. This tunnel enabled testing of a 10 -m diameter wind turbine flow LS-1 airfoil [30] designed for wind turbines. Because the
system (NREL Phase VI test) keeping the blockage to an accepta- local Reynolds number over the blade is a function of the blade
ble level so that corrections were not necessary (<10%). Some radius, experimental airfoil data must be collected for the entire
large wind tunnels do have dimensions of 10 m 10 m but even range of Reynolds numbers expected. NREL has supported
this is not large enough for most wind turbine systems [12]. A research in this area of airfoil design [31]. Other researchers are
trend in the literature shows the emergence of open test sections developing alternative families of airfoils and also testing NREL
for wind tunnels being applied to wind turbine testing [13–15]. airfoils for wind turbine use [32–38]. Future work with wind tur-
The majority of wind tunnels available for testing are smaller and, bine airfoils will include modification of existing airfoils for better
if the blockage constraint of less than 10% is satisfied, this limits performance [30,39].
the size of the turbine that can be tested. The measurements and Several studies have attempted to gather experimental airfoil
experiments in controlled conditions (MEXICO) tests were done data for low Reynolds numbers and compare these data to simula-
in the DNW facility with a 4.5 -m diameter turbine in a tions. McGhee et al. [40] studied the Eppler 387 (E387) airfoil at
9.5 m 9.5 m wind tunnel test section with no corrections [16]. Reynolds numbers from 60,000 to 460,000 and presented both
Sicot et al. tested a 1.3 -m diameter turbine in a 4 m 4 m test sec- experimental and simulation data using PROFIL. A good compar-
tion with no blockage corrections [17]. Smaller wind tunnels of ison was found between the theoretical and experimental results
the sort more common in research universities are also useful for for the higher Reynolds numbers, but at lower Reynolds numbers
wind turbine system testing. Burdett and Van Treuren have suc- PROFIL was unable to accurately predict the increase in drag
cessfully tested systems in a wind tunnel with dimensions of resulting from a laminar separation bubble. The laminar separa-
0.61 m 0.61 m and wind turbine diameters of 0.5 -m [18,19]. tion on low Reynolds numbers airfoils was studied by Hu and
Other testing has been done on small-scale models in wind tunnels Yang [41]. They document the behavior of this separation bubble
of 1 m 1 m or less [20–26]. Blockage is always a consideration for a Reynolds number of 70,000 using both particle image veloc-
under these small-scale conditions. Traditionally, tunnel block- imetry and surface pressure measurements. They also found that
ages of 6–7.5% or even up to 10% have been acceptable with no the separation bubble moves forward as the angle of attack is
corrections needed, according to Howell et al. [24]. Scaling down increased. Being able to predict this separation bubble and its
rotating machinery typically leads to losses in efficiency which movement is critical to predicting airfoil performance for small-
must be understood [27]. scale wind turbines. A number of studies have been done on air-
foils for small-scale wind application [17,42–48]. A wide range of
Results and Discussion: Wind Tunnel Testing Reynolds numbers have been tested and/or simulated with these
studies. For Reynolds numbers above 200,000, agreement
Airfoil Characteristics. Much of the literature concerning between experimental data and PROFIL was generally good.
wind tunnel testing of wind turbines involves experiments to However at 100,000 and below, where laminar separation can
determine the lift and drag coefficients of airfoil cross sections dominate the flow over an airfoil, the agreement was considered
fair [33]. Selig and McGranahan [47] have shown that drag behav- If low Reynolds number or poststall data are desired, experi-
ior is highly dependent on Reynolds number, especially under low mental testing must be performed because neither PROFIL nor
Reynolds number conditions. Thus, there is a need for additional XFOIL can accurately predict aerodynamic performance due to
airfoil analysis, both experimental and simulated, at the low the flow separation that can be experienced over large portions of
operational Reynolds numbers for small-scale wind turbines to the airfoil surface under these low flow conditions. While compu-
further refine blade design. tational fluid dynamics (CFD) could be an alternative to experi-
Two notable sources of airfoils designed for the low Reynolds mental testing, it must first be validated experimentally before the
numbers that small-scale wind turbines experience are the NREL theoretical model can be extrapolated to conditions not currently
and the University of Illinois, Urbana-Champaign (UIUC). NREL predicted. CFD does not presently model transition from laminar
has developed two families of airfoils for turbines under 10 -m in to turbulent flow well, which is a critical phenomenon with a sepa-
diameter with each family addressing a specific diameter wind ration bubble. Since an accurate, dependable model does not exist,
turbine. The S833–S835 have been designed for 1–3 m turbines experimental testing is still required.
[37] and the S822–S823 have been designed for 3–10 m turbines
[38]. These airfoils have been designed theoretically using PRO-
FIL in the Reynolds numbers range of 150,000–800,000. Among Airfoil Testing in a Wind Tunnel. While airfoil testing has
other design criteria, these airfoils were designed to minimize the become relatively standardized, many differences exist in wind
effect of leading-edge roughness by simulating the flow transition tunnel sizes that have been used to measure airfoil data. Large
that occurs under those conditions. tunnels, such as the tunnel at the University of Illinois Urbana-
Additional airfoils have been designed by Selig et al. and these Champaign, have test sections as large as 0.914 m 1.22 m
as well as other low Reynolds numbers airfoils have been tested (3 ft 4 ft). These types of tunnels employ several different tech-
experimentally over Reynolds numbers range of 30,000–50,000 niques to measure the forces on the airfoil. A common method
[47,49–51]. In Refs. [49] and [50], airfoils were tested with both used is the force balance, or sting, which measures lift and drag
normal flow conditions and tripped flow conditions (to simulate directly. Some tunnels do measure lift and drag using a side mount
leading-edge roughness). For most airfoil testing, however, data system which allows the flow to be more two-dimensional (2D)
were only taken at Reynolds numbers as low as 100,000. Addi- than using a sting. Other experiments measure drag downstream
tionally, the data were never taken poststall, which is necessary of the model by characterizing the wake and using the momentum
for the design of stall-regulated wind turbines. deficit approach. What is evident is that experimental data
Fig. 3 Typical small-scale wind tunnel [18] Fig. 4 ACP Spyderfoam airfoil with epoxy finish [52]
scaled up or down exactly in proportion to the change in radius. the radius, as seen in Table 1 for rotational speed. As can be seen
Petersen [59] stated that small variations in the radius, such as in Table 2, the freestream velocity must increase by the geometric
extensions of the blade supports at the hub, cause the scaling pro- scaling factor, and the rotational velocity must increase by the
jections to be inaccurate. Associated with the scaling would be the square of the scaling factor. Because of the relationships derived
increased weight of the blades and the loads generated on the from Reynolds number matching, scaling down for wind tunnel
blades themselves. This does not always follow the same geomet- testing in this manner can lead to impractical freestream and rota-
ric similarity. Moe states, for example, that increasing the size by tional velocities.
20% would increase the power production by 50% while increas- The result is that, to test the full-scale Reynolds number on a
ing the weight by 73% [57]. Increasing the size of the wind tur- smaller scale, it is necessary to test at tip speeds above the
bine using geometric scaling will usually find that costs do not
scale in the same proportion and will become an increasingly
larger part of the overall energy production costs [57]. It is sug-
gested that scaling a wind turbine up past a certain size will even-
tually not be possible and will result in a constant response level,
regardless of size [57]. Manwell et al. [58] state that “geometric
similarity is maintained to the extent possible,” recognizing there
may be cases where geometric similarity is not achievable.
Reynolds number matching is a test condition that is desired in
wind turbine testing. Without Reynolds number matching, the aer-
odynamic conditions between the two turbines are not truly identi-
cal. However, for Reynolds numbers greater than 500,000, airfoil
performance does not change considerably with Reynolds num-
ber, so Reynolds number matching is less significant. Most
medium to large turbines ( > 20 -m) operate at higher Reynolds
numbers and this effect does not need to be taken into account [5].
At Reynolds numbers below 500,000, however, “extreme and un-
usual behavior is caused by anomalous transition, separation, and
bubble formation phenomena” [5]. Burdett and Van Treuren have
investigated the case for scaling when Reynolds number matching
is included [18]. Table 2 shows that values, such as the rotational
velocity, scale as one over the radius squared instead of one over
Fig. 7 Eppler 387 airfoil coefficient of (a) lift and (b) drag at a
Fig. 6 Airfoil and test section with force balance [52] Reynolds number of 100,000 [53]
Radius R R1/R2 ¼ sc
Chord c c1/c2 ¼ sc
Kinematic viscosity 1 =2 ¼ 1
Freestream velocity U U1/U2 ¼ 1/sc
Rotational velocity X X1/X2 ¼ 1/sc2
Rotor power PRotor PRotor1 =PRotor2 ¼ 1=sc
Torque s s1 =s2 ¼ sc
PRotor sX
CP;M ¼ ¼ (18)
PFreestream;M 1
qpR2 UM
3
2
PRotor sX
CP ¼ ¼ CP;M BF3 ¼ (19)
PFreestream;M 1
qpR2 U3
2
Conclusions
Small-scale wind turbines are being installed at an ever increas-
Fig. 15 Corrections to S823 airfoil data at Re 5 200,000 [52] ing rate and it is important to have good experimental data upon