Emes-Rev Heliostat Wind Load (2021)

Solar Energy 225 (2021) 60–82
Contents lists available at ScienceDirect
Solar Energy
journal homepage: www.elsevier.com/locate/solener
A review of static and dynamic heliostat wind loads

Matthew Emes a, *, Azadeh Jafari a, Andreas Pfahl b, Joe Coventry c, Maziar Arjomandi a
a
Centre for Energy Technology, School of Mechanical Engineering, The University of Adelaide, SA 5005, Australia
b
German Aerospace Center (DLR), Institute of Solar Research, Professor-Rehm-Str. 1, 52428 Juelich, Germany
c
School of Engineering, College of Engineering and Computer Science, Australian National University, Canberra ACT 2601, Australia
A R T I C L E I N F O A B S T R A C T
Keywords: Accurate estimation of the static and dynamic wind loads on heliostats based on detailed measurement and
Heliostat characterisation of turbulence is crucial to avoid structural failure and reduce the cost of the structural heliostat
Wind load components. Wind load predictions for heliostats are not specified in design standards for buildings because of a
Aerodynamics
heliostat’s non-standard shape and the variations of wind velocity and turbulence in the lowest 10 m of the
Atmospheric boundary layer
Turbulence
atmospheric boundary layer (ABL). This paper reviews the static and dynamic wind loads on heliostats in the
most unfavourable operating and stow positions, with a focus on the aerodynamic effects related to the heliostat
structural component geometry, turbulence parameters in the ABL and field spacing. An increased resolution of
field-scale wind measurements at heliostat field sites is recommended to fully characterise the ABL turbulence, as
the high-intensity gusts over shorter durations at heights below 10 m lead to high-amplitude displacements with
larger frequencies than observed in standard building structures. Increased understanding and development of
aerodynamic wind load predictions for heliostats, based on their critical scaling parameters and local wind
conditions, would increase the accuracy of annual field efficiency models through an improved resolution of
operating load data and reduce the capital cost of structural components in power tower plants.
$140/m2 by National Renewable Energy Laboratory (NREL) (Turchi

et al. 2019), with the 2030 DOE target set at USD $50/m2 (Department
1. Introduction of Energy 2017). The most typical heliostat design in the current com
mercial CST plants, such as the 50 MW Khi Solar One heliostat field in
The application of concentrating solar thermal (CST) power tower Fig. 1(a), consists of glass mirror facets supported by steel beams and
technology is emerging as a means for industrial process heating and trusses and a T-shaped pedestal and torque tube with azimuth and
dispatchable renewable electricity production. Thermal energy is elevation drives for tracking (Téllez et al. 2014). Techno-economic
collected by a receiver located at the top of a central tower where solar analysis by Emes et al. (2020a) found that the steel support structure
radiation is concentrated by a large field of heliostats through two-axis components (Fig. 1b) increased their contribution from 18% to 34% of
tracking of the sun. Cumulative installed capacity of power tower plants the total heliostat cost due to increased wind loads with increasing he
increased by five times to approximately 6.3 GW and their levelised cost liostat size from 25 m2 to 150 m2. Furthermore, the total heliostat cost
of electricity (LCOE) decreased by 47% to USD $0.182/kWh between was reduced by 40% and the optimal heliostat size increased from 25 m2
2010 and 2019 (IRENA 2020). During this time, the capacity factor of to 50 m2 by lowering the stow design wind speed from 20 m/s to 10 m/s
deployed commercial-scale power tower plants increased from 30% to (Emes et al. 2015). To achieve the cost reduction targets, innovative
45% through increased power cycle efficiencies operating at high tem designs of the heliostat structural components must be developed to
peratures (Mehos et al. 2017) and increased energy storage capacity reduce their manufacturing and installation cost (Pfahl 2014a; Pfahl
from 5 h to 7.7 h at sites with larger direct solar resources (IRENA 2020). et al. 2017a). This requires a detailed understanding of the flow field
According to projections by IRENA, the LCOE will further decrease to aerodynamics for a reliable estimation of the wind loads on heliostats.
USD $0.07–0.08/kWh for power tower plants commissioned in 2021. Heliostats are exposed to atmospheric wind that imposes unsteady
One promising opportunity to achieve a reduction in the LCOE is by loads on the drives, torque tube, pylon, foundation and mirror trusses.
reducing the heliostat field cost, which contributes approximately Overestimation of the design wind loads increases the capital cost of a
40–50% of the total plant cost (Kolb et al. 2011; Pfahl et al. 2017a). solar plant. The wind-bearing heliostat components are designed for a
Currently the total cost of industrial scale heliostats is estimated as USD
* Corresponding author.
https://doi.org/10.1016/j.solener.2021.07.014
Received 15 April 2021; Received in revised form 27 June 2021; Accepted 7 July 2021
0038-092X/© 2021 International Solar Energy Society. Published by Elsevier Ltd. All rights reserved.
M. Emes et al. Solar Energy 225 (2021) 60–82
Nomenclature lpx Distance to the centre of pressure from the heliostat

elevation axis (m)
A Surface area of heliostat panel (m2) Lxu Integral length scale of longitudinal velocity (m)
AR Aspect ratio (width/height) of heliostat panel =b/c Lxw Integral length scale of vertical velocity (m)
αU Exponent of power law velocity profile MHy Hinge moment about elevation axis of heliostat (Nm)
α Elevation angle of heliostat panel (◦ ) My Overturning moment about base of heliostat pedestal (Nm)
β Angle of attack of wind with respect to heliostat (◦ ) Mz Azimuth moment about vertical axis of heliostat (Nm)
b Width of heliostat panel (m) ρ Density of air (kg/m3)
c Chord length of heliostat panel (m) p Differential pressure between upper and lower surface (Pa)
cFi Coefficient of force Fi where i = x, z r Displacement (mm)
cMi Coefficient of moment Mi where i = Hy, y, z Suu Longitudinal velocity spectrum (m2/s)
δ Atmospheric boundary layer depth (m) Sww Vertical velocity spectrum (m2/s)
δASL Atmospheric surface layer depth (m) θ Mean potential temperature (◦ C)
df Foundation pile depth (m) UH Mean velocity at elevation axis height of heliostat (m/s)
Fx Drag force on heliostat (N) U∞ Freestream velocity in the ABL (m/s)
Fz Lift force on heliostat (N) uτ Friction velocity (m/s)
f Frequency (Hz) x Longitudinal/streamwise direction (m)
Gu Gust factor of wind velocity y Lateral/spanwise direction (m)
H Elevation axis height of heliostat (m) z Height (m)
Iu Turbulence intensity of longitudinal velocity component z0 Logarithmic velocity profile surface roughness height (m)
Iw Turbulence intensity of vertical velocity component
k Von Karman’s constant
serviceability condition with stiffness to minimise local deformations of dimensional aerodynamic coefficients for the drag and lift forces on
the mirror surface during operation at different elevation angles (α > the heliostat surface, and the bending moments about the elevation axis,
0◦ ), and a survivability condition with strength against the maximum vertical axis and base of the pylon, were applied following benchmark
loads during high-wind events (e.g. gust front, storm) when the heliostat wind tunnel studies by Peterka et al. on isolated heliostats. Peterka and
surface is aligned horizontally (α = 0◦ ) in the stow position. The aero Derickson (1992) measured the mean and peak wind load coefficients in
dynamics of these two conditions vary significantly: operating heliostats a simulated ABL with a turbulence intensity Iu = σu /UH = 18%, denoted
are characterised by bluff body features including maximum drag forces as the root-mean-square of the longitudinal velocity component to the
with increasing surface area with respect to the approaching wind and mean wind speed at the elevation axis height H of a square-facet he
vortex shedding from the sharp edges of rectangular heliostat mirrors. liostat model (c = 0.27 m, H = 0.13 m). The forces and moments were
Stowed heliostats are characterised by slender streamlined body fea calculated using high-frequency base force balance measurements on
tures including maximum lift forces in a highly turbulent flow generated the heliostat model (Peterka et al. 1988; Peterka et al. 1989). The
by upstream roughness in the atmospheric boundary layer (ABL). maximum aerodynamic load coefficients on a scaled model heliostat
Furthermore, the dynamic wind loads induced by coupling between the (Peterka et al. 1988; Peterka et al. 1989; Peterka and Derickson 1992)
temporal variations of the wind loads and the dynamic properties of the were reported in the simulated ABL representing an open country terrain
heliostat structure, lead to oscillations of the heliostat surface that im (z0 = 0.03 m) with Iu = 18% and Gu = 1.6 at the heliostat elevation axis
pacts the tracking (mirror orientation) accuracy and optical perfor height. It has been widely acknowledged that the aerodynamic co
mance of the heliostat field. efficients in this benchmark study were reported for a single case,
Evaluation of the maximum wind loads at the appropriate temporal whereas the mean wind speed and turbulence intensity profiles of the
resolution is essential for the cost-effective design of heliostats, since a ABL approaching the heliostat vary significantly with height and surface
wide range of sizes and structural designs is currently deployed in the roughness. The unsteady pressure distribution on the mirror panel due
CST industry. Historically, design wind loads on industrial-scale helio to turbulence in the wind imposes highly fluctuating moments, which
stats incorporated aerodynamic coefficients using scaled models of the can create maximum loads on the heliostat pedestal, foundation and
heliostats in boundary layer wind tunnel experiments. The non- drives. Assessment of the dynamic response of the heliostats under
Fig. 1. Photographs of (a) the 50 MW Khi Solar One heliostat field (Abengoa Solar 2016), and (b) structural heliostat components of the Abengoa Solar heliostat.
Adapted from Advisian Worley Group (2021).
61
unsteady wind loads is necessary for preventing structural failure due to 2. Atmospheric boundary layer modelling
resonance and buffeting (Pfahl et al. 2017a), which may result from the
convergence of the dominant frequency of the wind fluctuations to the The atmospheric boundary layer (ABL) is the lowest 1–2 km of the
natural frequency of heliostat structures in the typical range of 1.6–3 Hz troposphere, where the mechanical properties of the wind are directly
(Gong et al., 2012; Griffith et al., 2015; Vásquez-Arango et al., 2015). influenced by the Earth’s surface (Stull 1988). The lower 100 m of the
Deformations and displacements of the heliostat structural elements ABL, where heliostats and other physical structures including buildings
caused by unsteady pressure distributions and dynamic amplification of and bridges are positioned, is known as the atmospheric surface layer
peak wind loads impacts the ability of heliostats to minimise tracking (ASL). Surface friction and vertical temperature gradient are two
error and spillage losses of solar radiation at the receiver (Arbes et al., important parameters that influence the wind structure in the ASL
2017; Blume et al., 2020), and to withstand strong wind gusts in the stow (Kaimal and Finnigan 1994). Turbulence in the ASL during near-neutral
position at high wind speeds (Emes et al. 2017; Vasquez Arango et al. stability conditions relevant to heliostat design wind speeds is me
2017; Emes et al. 2018; Pfahl 2018; Jafari et al. 2019a). Numerical chanically generated by shear from the terrain surface roughness, with a
methods, such as Large Eddy Simulation (LES), are generally associated negligible impact of the mean potential temperature gradient ∂θ/∂z =
with large computational effort and uncertainties to model the fluctu 0 and the net vertical heat flux w’ θ’ = 0 (Stull 2005). The wind velocity
ating wind loads due to ABL turbulence and the transient response profile in a neutral boundary layer is conventionally modelled as a
characteristics of heliostat structures. RANS methods would be less logarithmic profile in wind engineering applications, such as the ulti
extensive but are not suitable to simulate the upstream turbulence mate design wind loads on heliostats at high wind speeds during storms
structures. Hence, experimental data through wind tunnel and field and gust fronts.
measurements of the ABL turbulence characteristics are usually ob
tained in the design of a heliostat field, for the assessment of operational 2.1. Effect of surface roughness on wind speed and turbulence profiles
performance models and feasibility analyses of power tower systems.
Heliostats in operating positions act as bluff bodies within the ABL, The aerodynamic surface roughness determines the velocity and
where the interaction of their wakes with the incoming highly turbulent turbulence characteristics over a terrain, based on the height and surface
flow results in the aerodynamics of multiple heliostats varying signifi roughness (Simiu and Scanlan 1996). Wind speed is commonly
cantly from a single body. The vortices shed by an upstream heliostat or decomposed into a time-averaged mean component and a fluctuating
the tower can create vibrations and unsteady loads, due to the fluctu turbulent component. The mean velocity profile in the ABL has been
ating turbulence component of wind velocity, on the downstream in- modelled to various degrees of accuracy by the logarithmic law and
field heliostats positioned in the intermediate wake. Due to a blocking power law (Kaimal and Finnigan 1994; Xu 2013), respectively:
effect caused by upstream heliostats, wind tunnel measurements on an ( )
array of heliostats in multiple rows reveal that reducing the distance uτ
U(z) = ln
z
(1)
between heliostats decreases the time-averaged loads on the heliostats k z0
in the inner rows (Peterka et al. 1986). Peterka et al. (1987) In com ( z )αU
parison to a heliostat in the first row, the mean drag force and hinge U(z) = U∞ (2)
δ
moment coefficients on an instrumented heliostat in the fourth row of a
four-row array with low and high field densities were decreased by 10% where U∞ (m/s) is the freestream wind speed, δ (m) is the boundary
to 50%. In comparison to a heliostat in the first row, the peak drag force layer depth, uτ is the friction velocity, κ is von Karman’s constant equal
on the heliostat in the fourth row increased by 40% (Peterka et al. 1987). to 0.4, z0 is the aerodynamic surface roughness height,and αU is the
Hence, the distance between heliostat rows and the layout of heliostat power law exponent that characterises the level of surface roughness.
rows in a field impact the mean and peak wind loads on heliostats The depth δ of the neutrally stratified ABL can vary between a few
differently throughout a field. hundred metres to several kilometres, depending on the surface
This paper presents a review of the literature on the wind loads and roughness of the terrain (Xu 2013). Typical values of z0 for different
aerodynamics of heliostats, with the aim to highlight the key parameters terrains are shown in Fig. 2, varying in scale from millimetres in a very
that impact the accuracy of wind load predictions in the design and flat terrain (e.g. desert) to metres in an urban terrain. The zero-plane
development of industrial-scale azimuth-elevation heliostats. A solid displacement is negligible for small surface roughness lengths, such as
understanding of the wind loads is a major driver to reduce the struc flat and open-country terrains (Cook 1985), where heliostats are usually
tural cost of the heliostat field, without compromising the field effi located. With increased surface roughness and at lower heights in the
ciency and power tower plant performance. Section 2 discusses the ASL, the gradient of the velocity profile increases. Hence, more gusty
temporal and spatial distributions of turbulence, including the state-of- wind conditions occur due to the increasing fluctuating wind speed
the-art experimental modelling techniques for simulation of the ABL in a component due to turbulence close to the surface.
wind tunnel and the similarity requirements for heliostat wind load The power law has been shown to be suitable for modelling the mean
measurements over the range of surface roughness at different field sites. velocity profile at heights around 30–300 m, and thus it is most widely
Section 3 describes the conventional coordinate system of an azimuth- used for study of wind loads on tall buildings and other large civil
elevation heliostat and discusses the effect of the geometry of a helio structures (Xu 2013). Initially derived from the turbulent boundary
stat concentrator and its supporting structure components on the wind layer on a flat plate, the logarithmic law has been demonstrated to be
loads. Field experiment investigations focusing on the dynamic wind most suitable for modelling the mean velocity profile at heights below
load effects on heliostat vibration and tracking error due to the distri 100 m, representing the average depth δASL of the atmospheric surface
bution of surface pressures and wind-induced oscillations are outlined in layer (ASL) (Cook 1997; Li et al. 2010; Sun et al. 2014). The logarithmic
Section 4, followed by a discussion of the wind loads in a heliostat array law provides an accurate velocity profile independent of atmospheric
representing a section of field and the flow around multiple heliostats in stability for heights below 10 m very close to the ground (Kaimal and
Section 5. The key aspects of the literature that are critical to the Finnigan 1994), and is therefore appropriate for modelling the mean
development of wind load design guidelines for heliostats and future velocity profile for study of wind loads on heliostats.
research opportunities for wind load reduction are discussed in Section Statistical parameters of turbulence in the ABL are typically used to
6. determine the wind velocity fluctuations. Turbulence intensity is
representative of the amplitude of velocity fluctuations compared to the
mean velocity, defined as:
62
Fig. 2. Effect of surface roughness on wind velocity profiles in the atmospheric boundary layer.
Adapted from Gilooly and Taylor-Power (2016).
Ii =
σi
(3) (z0 = 0.003 m) to 0.3 in a suburban terrain (z0 = 0.3 m). According to
U the empirical relationships in ESDU 85020 (2001) derived from atmo
where σi is the standard deviation of the velocity component i = u, v, spheric data, σ v /σu and σ w /σu in the ASL are approximately equal to 0.78
w in the longitudinal, lateral, and vertical directions, respectively. and 0.55 at lower heights where z≪δ. The average depth δ of the at
Turbulence in the lowest 10 m of the ASL is anisotropic and the intensity mospheric boundary layer during neutral stability conditions is typically
of the turbulent fluctuations is the largest in the streamwise direction. between 450 m and 600 m, depending on the terrain roughness (Cou
Fig. 3 shows the dependence of the longitudinal (Iu ) and vertical (Iw ) nihan 1975; Xu 2013). Hence, the vertical turbulence intensities in Fig. 3
turbulence intensity on the height z from the ground, and the aero (b) follow a similar trend and are approximately half the magnitude of
dynamic surface roughness height z0 defined in the logarithmic velocity the longitudinal turbulence intensities.
profile in equation 1. The profiles of turbulence intensity are genertaed Turbulence in the atmospheric flow is dependent on the features of
from semi-empirical data in ESDU 85020 (2001) for U = 20 m/s at z = the terrain and varies based on the site of different heliostat fields. With
10 m, with an estimated uncertainty of ± 10% within the full-scale ABL increasing height from the ground, turbulence intensity decreases in
with uniform terrain roughness for an upwind fetch distance of 30 km. Fig. 3 while the integral length scale of turbulence increases in Fig. 4
The level of surface roughness impacts the magnitude and gradient of Iu , (ESDU 85020 2001). The integral length scale of turbulence represents
where the intermediate “open country” terrain (z0 ≈ 0.01–0.05 m) is the average size of the energy-containing eddies within a turbulent
commonly defined in wind engineering study of buildings and helio boundary layer (Emes et al., 2019c). Therefore, based on the height of
stats. For instance, in Fig. 3(a) at z = 6 m that approximates the hinge the heliostats from the ground and the terrain surrounding the heliostat
height of a 120 m2 heliostat, Iu increases from 0.14 in a very flat terrain field, the turbulence intensities and length scales show a very large
Fig. 3. (a) Longitudinal Iu , and (b) verticalIw turbulence intensity profiles in the lower 10 m of ABL for different values of surface roughness height z0 (ESDU 85020
2001). Error bars indicate ± 10% uncertainty of turbulence intensity for equilibrium conditions in the neutral ASL with U = 20 m/s.
63
Fig. 4. (a) Longitudinal and (b) vertical integral length scales of turbulence as a function of height z and surface roughness height z0 (ESDU 85020 2001). Error bars
indicate ± 20% uncertainty of integral length scales for equilibrium conditions in the neutral ASL with U = 20 m/s.
variation in the lowest 10 m of the ASL. Commercial-scale heliostats are the Reynolds number and the turbulence spectra between wind tunnel
manufactured with hinge heights in a typical range between 3 m and 6 experiments and the full-scale condition. The impact of Reynolds num
m. ESDU 85020 (2001) predicts the longitudinal integral length scale Lxu ber similarity can be overcome on sharp-edged models at Reynolds
in Fig. 4(a) to range from 27 m to 63 m in an open country terrain (z0 = numbers above 50,000 (Tieleman 2003). This has been demonstrated by
0.03 m) and from 50 m to 100 m in a very flat terrain (z0 = 0.003 m) the independence of aerodynamic coefficients of heliostats with Rey
with increasing z from 3 m to 6 m. The average longitudinal extent of the nolds number at freestream velocities between 5 m/s and 35 m/s (Pfahl
energetic eddies is therefore typically the same order as the chord length and Uhlemann 2011b). However, the turbulence fluctuations and their
of the heliostat and up to an order of magnitude larger. Eddies that are spectral distribution with the wide range of frequencies in the ABL affect
similar in size to the heliostat panel characteristic length are presumably the wind loads significantly (Jafari et al. 2019b).
responsible for the peak wind loads on heliostats in stow position, as Fig. 5 schematically presents the range of dimensions of a model
turbulence length scales that are comparable with the length scale of the heliostat in three sets of wind tunnel experiments studies (Peterka et al.
structure create a well correlated pressure distribution on the structure 1989; Pfahl et al. 2011a; Emes et al. 2017) and compares the geometric
(Mendis et al. 2007). This is because smaller eddies do not cause high net scaling of a full-scale heliostat and ABL with their respective models in a
pressures that are correlated over the heliostat surface, whereas wind tunnel. These studies measured wind loads, expressed as aero
considerably larger eddies have significantly lower vertical velocity dynamic coefficients of drag, cFx , and lift, cFz , forces, and the moments
fluctuations at the elevation axis height of the heliostat (Pfahl et al. induced at the hinge, cMHy , the foundation, cMy and the vertical azimuth
2015). Furthermore, the vertical component of the fluctuating velocity, axis, cMz , as shown in Fig. 5(b) on heliostat models in stow position and
defined by the vertical integral length scales Lxw in Fig. 4(b), increases inclined at different elevation angles (α) in operating positions:
from 2.2 m to 5.3 m in an open country terrain (z0 = 0.03 m) and from
Fx
4.1 m to 8.3 m in a very flat terrain (z0 = 0.003 m) with increasing hinge cFx = (4)
1/2ρUH2 A
height from 3 m to 6 m. The integral length scales of the vertical velocity
component are similar in magnitude to the heliostat chord length, which Fz
impacts the surface pressure distribution and the maximum hinge cFz = (5)
1/2ρUH2 A
moment on a heliostat in stow position. The interaction of the energetic
eddies with similar sizes to the heliostat (i.e. Lxw /c ≈ 1) are therefore MHy
cMHy = (6)
speculated to be responsible for dynamic effects observed in the field, 1/2ρUH2 Ac
such as aeroelastic flutter and fatigue loads on heliostats.
My
cMy = (7)
1/2ρUH2 AH
2.2. Scaling of heliostat models and turbulence spectra
Mz
The mismatch of scaling ratios, between the ABL thickness and chord cMz = (8)
1/2ρUH2 Ac
length of the heliostat, is an important consideration in wind tunnel
modelling of heliostats due to their small dimensions compared to the where c is the heliostat chord length in the longitudinal (windward)
ABL. It is possible to model heliostats with the same scaling ratio as the direction, H is the elevation axis (hinge) height, and UH is the time-
ABL, due to the due to technological constraints in modelling the averaged wind speed at the height of heliostat elevation axis.
structural details and measurement of the pressure and forces on a he Standard practice in scale-model simulations determines the geo
liostat model. Therefore, heliostats are usually modelled using higher metric scaling ratio of a heliostat model considering the effects of both
scaling ratios. between 1:10 to 1:50. This results in violated similarity of terrain and height, and the spectrum of the simulated boundary layer in
64
Fig. 5. A schematic showing the dimensions of: (a) a full-scale heliostat placed in ABL, (b) a model heliostat in a wind tunnel boundary layer, and the forces and
moments on the heliostat model. The dimensions of the model heliostat are based on the studies in the literature (Peterka et al. 1989; Pfahl et al. 2011a; Emes et al.
2017). The boundary layer thickness is δ, c defines the chord length of the heliostat mirror panel and H is the heliostat hinge height. The subscripts FS and m represent
full scale and model scale, respectively.
a wind tunnel (Cook 1978). Heliostat models were positioned in a the integral length scales of the longitudinal and vertical velocity com
simulated boundary layer with the mean velocity profile and turbulence ponents, respectively. These represent the average size of eddies corre
intensity matched to the ABL in an open-country terrain in wind tunnel sponding to the peak of the turbulence spectrum, which can be
tests. Peterka et al. (1989) tested a heliostat model at a scale of 1:40 in determined semi-empirically from the peak spectral frequency, or from
the Meteorological Wind Tunnel of the Fluid Dynamics and Diffusion the auto-correlation of the fluctuating velocity component (Farell and
Laboratory at Colorado State University. The boundary layer thickness Iyengar 1999).
in their wind tunnel simulation was about 1 m, which compared to the Fig. 6(a) shows a noticeable shift of the longitudinal power spectra to
average ABL thickness in open terrains suggests a scaling factor of 1:350 higher energy levelswith increasing turbulence intensity at a height of
for the ABL. The same scaling challenge was evident for the German 0.3 m within two different wind tunnel boundary layers (Jafari et al.
Aerospace Center (DLR) experiments by Pfahl et al. (2011a), where the 2019a). The shift in the spectral peak to smaller length scales by
heliostat model was at a 1:20 scale. A similar heliostat model scaling matching the turbulence intensity also indicates that the low-frequency
ratio, which was considerably larger than the ABL scaling ratio of part of the spectra cannot be reproduced, as due to the wind tunnel’s
approximately 1:100, was used in the University of Adelaide large-scale restricted cross-section and length, the generation of turbulent eddies is
wind tunnel by Emes et al. (2019a). The difference in scaling ratios is limited. (Peterka et al. 1998; Iyengar and Farell 2001; Banks 2011;
speculated to have led to variations in the reported wind load co Kozmar 2012; De Paepe et al. 2016; Leitch et al. 2016). A similar trend is
efficients for the maximum operational and stow heliostat configura shown for the vertical turbulence spectra in Fig. 6(b), with a shift to
tions at different elevation angles (α) with respect to the horizontal in higher frequencies. Pfahl et al. (2015) suggested that reproducing the
Table 1. This raises uncertainty of the accuracy of the wind load mea vertical power spectrum is important for evaluating the peak wind loads
surements. The similarity of wind tunnel experiment simulations for the on a stowed heliostat, because of the linear relationship found by Ras
evaluation of heliostat wind loads can be verified by an instrumented mussen et al. (2010) between the vertical spectra and the lift forces and
full-scale heliostat prototype at a field site, however such data has been hinge moments on a horizontal flat plate exposed to small vertical tur
scarcely reported in the literature (Jafari et al. 2019b). bulence Iw ≤ 10%. Jafari et al. (2019b) found that turbulent length
The power spectral density function of the wind speed provides scales of the same order as the heliostat’s chord (windward) length and
critical information about the scales of energy-containing turbulent an order of magnitude larger, corresponding to a range of reduced fre
eddies, which is necessary for evaluation of unsteady wind loads on quencies, 0.1 < fc/U < 1, effectively contribute to the unsteady wind
structures. The non-dimensional power spectral density of the velocity loads. Hence, it was proposed by Jafari et al. (2019b) that this range of
fluctuations compares the distribution of turbulence energy in the wind reduced frequencies of the turbulence spectra should be carefully
tunnel boundary layer with that predicted by ESDU 85020 (2001) in the simulated in wind tunnel studies in order to reduce the scaling impact on
ASL through a modified form of the von Kármán (1948) model: the measured peak wind loads and provide accurate wind load pre
dictions on the full-scale structure.
fSuu 4nu
= (1) The discrepancies between wind tunnel and atmospheric turbulence
σ2u (1 + 70.8n2u )5/6 spectra bring into question the reliability of wind load measurements
and whether they correspond to the wind loads on full-scale heliostats.
fSww nw (1 + 755.2n2w )
= (2) However, the formation of turbulent eddies in a wind tunnel is restricted
σ 2w (1 + 283.2n2w )11/6 by the tunnel’s limiting dimensions, therefore the integral length scales
where Suu and Sww are the power spectral density functions of the in the full-scale ABL cannot be replicated (Jafari et al. 2019a). The in
fluctuating streamwise and vertical velocity components, respectively, tegral length scales of the vertical velocity component increase with
and σ2u and σ2w are the streamwise and vertical velocity variances. The height from the ground in wind tunnel and full-scale measurements. In
contrast, the longitudinal length scales are larger near the surface in a
non-dimensional frequency is defined as ni = fLxi /U, where Lxu and Lxw are
Table 1
Comparison of peak operational and stow wind load coefficients reported in the literature.
Wind tunnel experiment Operation Stowα = 0◦ Iu (%)
α = 90 cFx
◦
α = 30 cFx
◦
α = 90 cMy◦
α = 30 cMHy
◦
cFx cFz cMy cMHy
Peterka et al. (1989) 4 2.8 4.35 0.6 0.9 0.6 1 0.2 18

Pfahl et al. (2011a), Pfahl et al. (2015) 3.3 2.1 3.2 0.55 0.43 0.38 0.53 0.18 18
Emes et al. (2019a) 2.25 1.89 2.29 0.21 0.39 0.49 0.43 0.13 13
65
Fig. 6. Comparison of wind tunnel measurements in two simulated ABLs (Jafari et al. 2019a) with the modified von Karman form (ESDU 85020 2001) of non-
dimensional turbulence spectra of the (a) longitudinal fluctuating component of wind speed u, (b) vertical fluctuating component of wind speed w. Simultaneous
matching of both the longitudinal and vertical spectra in the critical range of reduced frequencies cannot be achieved in scaled model wind tunnel experiments.
Similarity of turbulence spectra should be applied to the velocity component that contributes to the unsteady wind loads on the heliostat configuration being
investigated.
wind tunnel boundary layer, where an increased base width of the spires turbulent eddies. Turbulence intensity in the approaching flow is a
generates large vortices through separation. Regardless of the different commonly reported parameter that affects the wind loads on operating
mechanisms that create turbulence in the wind tunnel and the lowest 20 and stowed heliostats. Peterka et al. (1989) studied the mean and peak
m of the full-scale ABL, the increase in Lxu and the decrease in Lxw is also wind loads on a heliostat at different elevation angles in simulated
seen in the ASL’s lower regions due to blocking of the vertical velocity boundary layers at Iu = 14% and Iu = 18%. It was found that with
component near the ground (Jafari et al. 2019a). Pfahl (2018) concluded increasing Iu , the peak lift and drag force coefficients increased for all
that matching the vertical turbulence intensity with full-scale standard elevation angles, α, of the heliostat panel with respect to the horizontal,
data, despite a shift of the streamwise turbulence spectrum to higher with best-fit curves shown by the dashed lines in Fig. 7. The maximum
frequencies in wind tunnel experiments, was appropriate for deter drag force coefficient at α = 90◦ increased from 3 to 4, and the peak lift
mining the lift force and hinge moment measurements on a model-scale force coefficient at α = 30◦ increased from 1.7 to 2.7 by increasing Iu at
stowed heliostat. Hence, the geometric scaling ratio of a heliostat model the heliostat hinge height from 14% to 18%. Furthermore, according to
should be determined according to the turbulence spectrum for the Peterka et al. (1987), the peak lift force coefficient on a heliostat at stow
corresponding full-scale structure, considering the effects of both terrain increased from 0.5 to 0.9 when Iu increased from 14% to 18%. Peterka
and height, and the spectrum of the simulated boundary layer in a wind et al. (1989) discussed that the reason for the increase in the wind loads
tunnel. The geometric scaling ratios for modelling a prototype heliostat was not found in their experiments but it was likely to be was linked to
in an open-country terrain were determined as an example by Jafari the interaction of turbulence and separated shear layers near the plate’s
et al. (2019b). It was found that similarity of the streamwise velocity edge.
spectrum is required to model the unsteady drag force on a vertical Emes et al. (2019a) further investigated the effect of turbulence in
heliostat at α = 90◦ and a 1:20 scale model with larger dimensions tensity on the peak aerodynamic hinge and overturning moment co
showed the closest match to the modified von Karman spectrum (ESDU efficients on a single heliostat model, through an extension of turbulent
85020 2001). In contrast, accurate measurement of the unsteady lift ABLs simulated in previous wind tunnel experiment studies by Peterka
force on a stowed heliostat requires similarity of the vertical turbulence et al. (1989) and Pfahl et al. (2015). The percentages in the legend of
spectrum, which showed the closest match to the von Karman spectrum Fig. 8 indicate the longitudinal turbulence intensity at the hinge height
(ESDU 85020 2001) for a 1:60 model with smaller dimensions. The of the heliostat model for open terrains of a range of roughness heights.
relative contribution of the longitudinal and vertical components of Increased intensity of turbulence of the approaching ABL flow directly
turbulence, for a stowed heliostat and over the range of heliostat oper correlated to increases in the peak moment coefficients. The quasi-
ating conditions, should be further verified through wind tunnel and steady peak values of the force and moment coefficients are deter
full-scale measurements. Since the unsteady longitudinal and vertical mined as the sum of the mean and three-times the standard deviation of
turbulence components are not generated independently using spires the fluctuating moment, with a 99.7% probability of not being exceeded
and roughness elements, this would require investigation of active following a Gaussian distribution (Simiu and Scanlan 1996). As reported
methods of turbulence generation. Analysis of wind loads on full-scale by Peterka and Derickson (1992), there is an approximately linear in
heliostats with respect to the incoming wind turbulence measured crease of the peak coefficients with increasing turbulence intensity at
simultaneously can also verify the scaling effects observed in wind Iu ≥ 10%. The difference between the scaling factors of the model-scale
tunnel experiments to provide a more reliable estimation of wind loads. ABL and heliostat in wind tunnel experiments with respect to their full-
scale counterparts led to variations in the peak wind load coefficients.
2.3. Effect of turbulence intensity and length scales on peak wind loads The relative sizes of the heliostat chord length and the energy-
containing eddies is another important factor influencing the range of
The impact of turbulence on heliostat wind loads has been widely frequencies that contribute to the generation of fluctuating loads.
investigated through systematic wind tunnel experiments in the litera Due to the anisotropic nature of atmospheric turbulence and
ture. Further to the variation of the time-averaged component of the depending on the orientation of the heliostat panel, both streamwise and
wind speed with height and surface roughness in the ABL for the vertical turbulence parameters can be of significance for the wind loads.
determination of design wind speeds on heliostats, the temporal char While in the previous experiments by Peterka et al. (1989) and Emes
acteristics are defined by the intensity of the velocity fluctuations and et al. (2017), all components of turbulence intensity varied during the
the spatial variations are characterised by the integral length scale of experiments, the observed effects on the wind load coefficients were
66
Fig. 7. (a) Peak drag force and (b) lift force coefficients on a heliostat at different elevation angles, α, atIu = 14% andIu = 18%.
Reproduced from Peterka et al. (1989).
Fig. 8. Effect of turbulence intensity Iu (%) and elevation angle α of a heliostat in wind tunnel experiments (Peterka et al. 1989; Pfahl et al. 2015; Emes et al. 2019a)
on: (a) peak hinge moment coefficient, and (b) peak overturning moment coefficient.
only correlated with longitudinal turbulence intensity and the variations was drawn from comparison of the lift force coefficients for two cases in
of vertical turbulence components were not differentiated. Pfahl (2018) the cylinder wake with a heliostat model in a simulated boundary layer
proposed that at stow position, vertical velocity is more decisive for the with Iw = 10% was identical. However, the turbulence in the wake of a
pressure forces on the panel as it acts normal to it and therefore the lift cylinder is dominated by quasi-static vortex shedding with different
force coefficient on a stowed heliostat was suggested to be more closely vertical turbulence profiles and spectral properties than in the ABL.
correlated with vertical turbulence intensity, Iw . The lift force on a Pfahl (2018) suggested that the lift force and hinge moment coefficients
stowed heliostat model in a simulated boundary layer was measured in a in stow position were largely dependent on the vertical turbulence in
series of tests, where Iu and Iw varied in the wake of cylinders of different tensity compared with dissimilarities of the turbulence spectra. Despite
diameters. Fig. 9 shows the peak and root mean square (RMS) lift force changes in the shape of the spectra affecting the pressure distribution, it
coefficients as a function of Iu and Iw . Pfahl (2018) discussed that the was found that the differences in strength and width of the high-pressure
curve-fitted coefficients showed a better match as a function of Iw , and suction region near the heliostat mirror panel’s edge compensate each
therefore, Iw has a stronger effect on the lift force than Iu . This conclusion other regarding these wind load coefficients.
Fig. 9. Effect of turbulence intensity on peak and RMS lift force coefficients on a stowed heliostat for: (a) longitudinal turbulence intensity, Iu , (b) vertical turbulence
intensity, Iw .
Reproduced from Pfahl (2018).
67
Another important parameter which influences the wind loads is the function in equation 10 shows a larger sensitivity of the peak lift force
integral length scale of turbulence in the boundary layer. The ratio of the coefficient to Lxw /c than to Iw . As a result, the influence of the vertical
integral length scale to the heliostat chord length was found to impact velocity turbulent energy’s spatial distribution on the lift force on a
the wind loads on a heliostat at stow position. Emes et al. (2017) studied stowed heliostat is greater than the vertical velocity turbulent energy’s
the effect of changes in Lxu /c by measuring the lift force on stowed he temporal release. In contrast, the smaller exponent of 0.48 indicates that
liostat models of different chord length dimensions in a modelled at the spatial release of longitudinal energy in the investigated range of Lxu /
mospheric boundary layer. They found that the peak lift force coefficient c between 1 and 4, has a relatively smaller effect on the peak drag force
increased with increasing Lxu /c, however both Lxu /c and Lxw /c varied coefficient on a vertical heliostat. Hence, the peak wind loads on he
simultaneously by changing the chord length dimensions of the helio liostats in the ABL can effectively be estimated for these two critical load
stat. By stowing a fixed heliostat size with constant c at different heights cases using the defined turbulence parameter, in terms of the expected
in a simulated ABL, Jafari et al. (2019a) showed that the peak lift co full-scale turbulence intensity and length scales that are a function of the
efficient was more strongly correlated with Lxw /c than Lxu /c. As shown in surface roughness of the terrain in Fig. 3 and Fig. 4, respectively.
Fig. 10, the peak lift coefficient increased by 65% when Lxw /c increased
from 0.3 to 0.5 at a constant Lxu /c = 1. In comparison, only a 10% 3. Heliostat geometry effects on wind loads
reduction in the lift coefficient was observed with increasing Lxu /c from 1
to 1.15 at a constant Lxw /c = 0.5. Hence, this demonstrates that the The wind effects on heliostats are well represented by the bluff body
vertical component of the fluctuating velocity makes a larger contribu aerodynamics of the large reflecting surface inclined at different eleva
tion to the generation of the lift force on a stowed heliostat. The relative tion and azimuth angles during operation of a power tower plant. Fig. 12
influence of the longitudinal and vertical turbulence components on the shows the wind loads on a conventional azimuth-elevation heliostat,
heliostat wind loads at intermediate elevation angles, such as the consisting of an array of rectangular glass facets mounted on tubular
maximum operating lift force and hinge moment at α = 30◦ , should be steel components in a T-shaped configuration to withstand the
considered in future investigations. maximum bending moments about the hinge and the base of the he
The combined effects of intensity and integral length scales of tur liostat pedestal. When inclined at different elevation angles, the gap
bulence on the aerodynamic load coefficients were studied by mea between the lower edge of the heliostat panel and the ground which
surement of the unsteady wind loads on vertical (α = 90◦ ) and stowed (α enlarges as α decreases. The critical scaling parameters that have been
= 0◦ ) heliostats in two simulated ABLs by Jafari et al. (2018) and Jafari investigated in the literature include the aspect ratio of the rectangular
et al. (2019a), respectively. Heliostat models of different chord length heliostat panel in section 3.1, the gaps between the heliostat facets in
dimensions between 0.3 m and 0.8 m at a fixed height H = 0.5 m were section 3.2, and the vertical distance between the elevation axis and the
tested for the maximum drag case on the vertical heliostat. Three chord ground by the pylon height in section 3.3.
length dimensions (c = 0.5, 0.6, 0.7 m) with H/c ratios between 0.2 and
1.3 were tested for the maximum lift case on the stowed heliostat. The 3.1. Aspect ratio
peak drag force coefficient on a vertical heliostat (Fig. 11a) followed a
logarithmic function of the longitudinal turbulence intensity and lon The aspect ratio of the heliostat, defined as the ratio of the width to
gitudinal integral length scale: the height AR = b/c of the panel in Fig. 12(b), has a significant but
[ ( ) ]
Lx
0.48 varying impact on the wind load components on a heliostat. The main
cFx = 1.05ln Iu u +4 (9) components of the heliostat that are exposed to wind effects are the
c
foundation, the pedestal, the panel and the elevation and azimuth
In contrast, the peak lift force on a heliostat at stow position drives. Fig. 13 shows the impact of the aspect ratio of a heliostat panel on
(Fig. 11b) was shown to correlate with a logarithmic function of the the normalised load coefficients for the maximum operating load cases
vertical turbulence intensity and length scale: and in stow position (α = 0◦ ), based on fitted exponential functions of
[ ( ) ] scale-model heliostat measurements in a boundary layer wind tunnel
(Pfahl et al. 2011a). It can be observed that My about the base of the
2.4
Lx
cFz = 0.267ln Iw w + 1.566 (10)
c upright heliostat at α = 90◦ decreases by approximately 30% at AR = 1.5
and by as much as 60% at AR = 3 relative to a square-shaped heliostat
The turbulence parameters in equations 9–10 describe the spatial (AR = 1). A reduction in My and MHy with increasing aspect ratio in
and temporal release of turbulence energy and their effect on the fluc
dicates smaller loads on the elevation drive and that the foundation pile
tuating load coefficients. The larger exponent of 2.4 in the logarithmic
depth and pylon diameter can be reduced. However, the Mz on operating
heliostat and Fz on stowed heliostat increase by 47% and 30%, respec
tively, with increasing AR from 1 to 3. Hence, there is a trade-off be
tween the dimensions of the pedestal with the elevation drive and the
torque tube with the azimuth drive in the heliostat design.
3.2. Facet gap

peak
Conventional heliostats are designed with small gaps between the

mirror facets. Wu et al. (2010) found that small gaps have a negligible
impact on the force and moment coefficients through wind tunnel tests
and numerical analysis. However, wider gaps in the mirror panel caused
a larger pressure difference at the edges of the gap at the windward
corners. This led to a 20% increase of the hinge moment on a heliostat at
α = 30◦ , due to a shift of the low-pressure region on the leeward surface
away from the central elevation axis for wind flow along the gap at β =
Fig. 10. Comparison of peak lift force coefficients in stow position for similar 0◦ (Pfahl et al. 2011c). The peak hinge moment at stow position with a
values of Lxu /c and different values of Lxw /c in the ABL with z0 = 0.018 m (Jafari wide gap was also increased due to a similar effect. Peterka and
et al. 2019a). Derickson (1992) stated that the area represented by slits in the mirror
68
Fig. 11. Peak wind load coefficients on a heliostat: (a) drag force coefficient at α = 90◦ as a function of longitudinal turbulence intensity and integral length scale
(Jafari et al. 2018); (b) lift force coefficient at α = 0◦ as a function of vertical turbulence intensity and integral length scale (Jafari et al. 2019a). The dashed lines
indicate the logarithmic relationships in equations 9–10 based on the longitudinal and vertical turbulence parameters, respectively.
Fig. 12. Schematic diagram of the (a) drag and lift forces on the heliostat surface inclined at elevation angle α, (b) hinge, overturning and azimuth moments on the
heliostat components (Emes et al. 2020a).
panel can be considered as a solid surface area up to a ratio of 15%. The to 0.7 for a heliostat with a horizontal primary axis (Téllez et al. 2014).
wind load coefficients were compared with no gap and a heliostat with As shown in in Fig. 14(a), the peak lift coefficient in stow position at H/
two mirror facets separated by a wide gap mirror facets. The total mirror c = 0.5 varies over a range between 0.4 and 0.9, depending on the
area (30 m2 at full scale, modelling scale 1:20), with a gap width of 0.5 m spectral distribution of ABL turbulence (refer to Section 2.2) and the
corresponded to a portion of 8% of the opening. With the exception of ratio of the integral length scales to the scale model heliostat charac
the peak operating hinge moment at α = 30◦ increasing by 20%, there teristic length in different wind tunnel experiments (Emes et al. 2017).
was only a small effect of gap on the wind loads, in agreement with the Measurement of the peak lift force on models with varying pylon heights
findings by Peterka and Derickson (1992). The shielding effect of sup over a range of H/c between 0.2 and 0.8 was used to study the effect of
port structure components contributed to small increases in the drag heliostat hinge height on stow loads. Jafari et al. (2019a) found that the
force in stow position and operating load cases with wind impacting the lift coefficient on a stowed heliostat followed a linear variation with H/c
back surface of the heliostat. Hence, the geometry of a heliostat from 0.5 to 0.2, such as a reduction from 0.3 to 0.2 at Iw = 9% (z0 =
concentrator consisting of facets with narrow gaps can effectively be 0.018 m), and from 0.65 to 0.48 at Iw = 19% (z0 = 0.35 m). The rate of
modelled as a thin flat plate when considering the aerodynamic wind reduction of cFz with decreasing H/c is larger in the ABL with z0 = 0.35
loads on a heliostat, whereas accurate prediction of the dynamic wind m, such that the slope of the linear function at z0 = 0.35 m is three times
loads (refer to Section 4) requires similarity of the structural stiffness larger than for z0 = 0.018 m. Fig. 14(b) shows the peak lift force coef
and mass distribution of the heliostat support structure. ficient on a heliostat at stow, normalised with respect to H/c = 0.5 as a
function of H/c for different values of aerodynamic roughness length z0 .
The peak cFz on a stowed heliostat within the ABL follows a linear
3.3. Pylon height function of H/c that is relatively independent of z0 . This relationship
indicated that the stow lift force can be decreased by up to 80% by
Conventional azimuth-elevation heliostats are commonly designed lowering the stow height of a fixed size panel such that H/c decreases
for a ratio of hinge height to mirror chord length, H/c = 0.5, increasing
69
Fig. 13. Effect of aspect ratio on the normalised heliostat loads for the maximum wind load configurations during operation and stow. Reproduced from the best-fit
power law exponents in Table 3 of Pfahl et al. (2011a).
Fig. 14. Effect of the hinge height to panel chord length ratio H/c on: (a) the peak lift force coefficient at stow based on different wind tunnel studies, (b) peak lift
coefficient normalised with respect to heliostat with H/c = 0.5, as a function of ABL aerodynamic roughness height z0 .
Reproduced from Jafari et al. (2019a).
70
from 0.5 to 0.2 (Jafari et al. 2019a). The pylon height is fixed in thus required to validate the turbulence characteristics at heights below
contemporary heliostat designs (Pfahl et al. 2017a), nevertheless novel 6 m where heliostats are stowed and verify the peak wind load co
concepts such as a carousel heliostat with spindle drive (Pfahl et al. efficients corresponding to the critical operating and stow load cases of
2017b) to lower the heliostat mirror close to the ground in stow during heliostats established in wind tunnel experiments.
high-wind conditions can reduce the maximum wind loads and the cost Emes et al. (2019a) showed that the hinge moment was highly
of a cantilevered heliostat. correlated with the movement of the unsteady centre of pressure from
the central elevation axis, which increased significantly with increasing
4. Dynamic wind effects on heliostat vibrations and tracking turbulence intensity and decreasing elevation angle of the heliostat.
error Through the decomposition of the hinge moment into the net normal
force and the centre of pressure distance, the pressure distributions on
4.1. Heliostat surface pressure distributions the heliostat surface representing the maximum hinge, overturning and
azimuth moments were determined (Emes et al. 2019b). A high-pressure
Dynamic wind load analysis on heliostats has been investigated using region was observed on the operating heliostat surface at α = 30◦ in
transient FEA simulations and experimental data from wind tunnel or Fig. 17(a), leading to the maximum cFz = 2.83 and cMHy = 0.18. Despite
full-scale measurements, such as through surface pressure measure smaller peak values of cFz = 0.42 and cMHy = 0.11 on the stowed he
ments by Gong et al. (2013) on a 1:10 scale model T-shaped heliostat. liostat at α = 0◦ in Fig. 17(b), there was an increased longitudinal (x)
Gong et al. (2013) showed that at the leading edge of the stowed he movement from the central elevation axis (y = 0.4 m) relative to the
liostat mirror surface, substantial negative peak wind pressure co operating heliostat. During operation, an area of high-pressure differ
efficients occurred (see Fig. 15). It is presumed that the turbulent eddies ence on the frontal half of the heliostat surface (α = 30◦ ) and flow
associated with the peaks of the turbulence spectra that are similar in separation at the windward edge of the stowed heliostat surface (α = 0◦ )
size to the chord length of the heliostat mirror have a large impact on the created the highest hinge moment on the torque tube. In contrast, the
maximum lift forces and hinge moments on stowed heliostats (Pfahl maximum azimuth moment during operation (Fig. 17c) corresponded to
et al. 2015). However, the effect of the size of these eddies relative to the the maximum drag coefficient cFx = 2.29 at α = 90◦ ) but with wind
size of the heliostat chord length on the unsteady loads and non-uniform approaching the heliostat at β = 60◦ . Probability distributions of the
pressure distributions on stowed heliostats has not previously been transient load fluctuations followed a Gaussian distribution for most of
investigated. the load cases except the maximum operating azimuth moment (Emes
Pfahl et al. (2014b) showed that the temporal variation of the stow et al. 2020a). In contrast, wind tunnel measurements by Xiong et al.
hinge moment on an 8 m2 heliostat, instrumented with 84 differential (2021) found that the fluctuating shear force at the base of the heliostat
pressure sensors in an open field in Lilienthal in northern Germany pylon followed a Gaussian distribution at α between 0◦ and 20◦ and the
(Fig. 16a), exhibited distinctive peaks over consecutive durations of peak value of the base shear force was most accurately represented by a
approximately one second. This suggests that the 3-second gust wind Generalized Pareto Distribution (GPD) at α between 30◦ and 90◦ . This
speed commonly applied in design codes and standards (ASCE 7-02 suggests that the quasi-steady peak wind loads are generally appropriate
2002; EN 1991-1.4 2010; AS/NZS 1170.2 2011) and recommended by to predict the maximum loads in operating and stow configurations, but
the World Meteorological Organisation (WMO) for wind measurements, extreme value analysis of the fluctuating load distribution should be
can under-estimate the gust wind speed and thus the maximum unsteady considered in operating positions. It should be noted that despite the
wind loads on heliostats. The peak pressure coefficient distribution in smaller peak coefficients on a stowed heliostat, the ultimate design loads
Fig. 16(b) at the instant of the maximum hinge moment, with peak should consider a larger survival wind speed compared to the wind
cMHy = 0.18 and cFz = 1.0 at β = 47◦ and U = 5 m/s indicates a signif speed for calculation of the maximum operating hinge and overturning
icant variation of positive pressure (suction) along the side edges. Pfahl moments.
(2018) discussed that the peak aerodynamic coefficients showed a
general agreement with tabulated values derived in controlled wind
tunnel experiments by Peterka and Derickson (1992) at Iu = 18% and 4.2. Modal vibration analysis and fatigue loads
U = 12.5 m/s. However, the turbulence characteristics of the ABL flow
in the field study by Pfahl (2018) were not reported. Notably the spatial The dynamic response of small-scale structures such as heliostats
similarity of the heliostat chord length (c = 2.5 m) and the integral affects their ability to withstand gusts in the ABL and maintain structural
length scale of the energy-containing turbulent eddies was only esti integrity for their expected design life. As heliostats are slender in shape
mated as Lxu = 3 m at z = H = 2 m, based on extrapolation of semi- and have low natural frequencies less than 10 Hz, the structural com
empirical data (ESDU 85020 2001) in an open country terrain with ponents of heliostats can be exposed to flow-induced vibrations from the
z0 = 0.03 m. High-frequency field measurements of wind velocity are unsteady fluctuating loads caused by turbulence effects. Vortex shed
ding can generate cyclic wind load fluctuations on the elevation and
Fig. 15. Peak pressure coefficient contours on a stowed heliostat at different azimuth angles: (a) β = 0◦ ; (b) β = 90◦ ; (c) β = 180◦ .
Reproduced from Gong et al. (2013).
71
Fig. 16. (a) Field heliostat instrumented with differential pressure sensors in open country terrain (Pfahl 2014a); (b) peak pressure coefficient distribution corre
sponding to the maximum hinge moment cMHy = 0.18 on the stowed heliostat at β = 47◦ ,
reproduced from Pfahl (2018).
Fig. 17. Peak pressure distributions on an instrumented heliostat in a boundary layer wind tunnel with Iu = 13% and Iw = 8%, leading to the maximum: (a) operating
hinge moment, (b) stow hinge moment, (c) azimuth moment (Emes et al. 2019a).
Fig. 18. Heliostat deformed shape for torque tube bending mode 3: (a) FEA simulation of displacement contours with modal frequency of 3.002 Hz, and (b)
experimental hammer excitation test with modal frequency of 3.034 Hz. Adapted from Menicucci et al. (2012). The red lines represent the deformed experimental
mode shape of the five columns from the undeformed reference geometry, the yellow box represents the instrumented facet, and the green line represents the yoke
with measurement locations at the endpoints.
72
azimuth drives in the frequency range between 1 and 5 Hz of a con using spectral analysis of the transient velocity, load and displacements
ventional heliostat, as well as significant vibration and resonance effects following a normal distribution to predict peak values and standard
(Gong et al. 2012; Griffith et al. 2015). Excessive deflections and stresses deviations of moments about principal axes of mirror frame and dis
caused by wind-induced oscillations can lead to structural failure (Jain placements in the normal direction of the mirror surface. Structural
et al. 1996; Mendis et al. 2007). Galloping and torsional flutter tend to failure through overstressing was evaluated by estimating the maximum
occur at frequencies on the order of 1 Hz where the turbulence integral stresses on support structure components, such that the maximum dis
length scales are similar in size to the characteristic length of the he placements due to the dynamic response were calculated to be less than
liostat components, such as the pylon, torque tube and mirror structural 1% of the heliostat chord length (Vasquez Arango et al. 2017).
truss members. A quasi-steady increase in mean velocity occurs when Dynamic wind loads on heliostats have been investigated by fluid
the turbulence scale is increased beyond the order of magnitude of the –structure interaction (FSI), combining transient CFD, FEA simulations
body scale (Nakamura 1993) and the galloping effect becomes negli and modal analysis to link the resolved flow field with the structural
gible when the turbulence scale is decreased below the size of the response. A FSI analysis by Vasquez Arango et al. (2017) showed a
structural member as smaller eddies cannot cause high net pressures pronounced peak at f = 3.8 Hz in the spectral distribution of the over
over the surface (Pfahl et al. 2015). turning moment coefficients on a 2.5 m × 3.22 m heliostat model. In
The equivalent static wind loads have been the subject of most comparison, spectral analysis of the fluctuating azimuth and over
experimental studies, however the dynamic loads due to wind-induced turning moments on a 0.8 m square heliostat model by Emes et al.
displacements are important for determining the heliostat drive units (2020b) in a boundary layer wind tunnel experiment showed a clearly
and support structure components. Dynamic testing of full-scale helio defined peak atf = 7 Hz. Wolmarans and Craig (2019) performed a one-
stats was undertaken by Sandia National Laboratories at the National way FSI modal analysis with scale resolving CFD simulation of a full-
Solar Thermal Test Facility (NSTTF) on a 37 m2 heliostat instrumented scale heliostat to determine the location of maximum stress at two
with triaxial accelerometers, strain gauges and anemometers to evaluate elevation angles. As shown in Fig. 19(a), the maximum von Mises stress
the modal shapes and frequencies (Andraka et al. 2013). Modal tests of occurred near the base of the LH-2 heliostat on the back face of the
the NSTTF heliostat using hammer excitation identified a number of pylon. The dynamic behaviour consisted of back-and-forth motion of the
modes of vibration, including bending of the support structure in modes concentrator due to the large bending moment caused by the maximum
1 and 2, bending of the torque tube in modes 3 (Fig. 18) and 4, and in- frontal area to the oncoming wind at α = 90◦ . In contrast, the maximum
plane and out-of-plane bending of the mirror-truss assemblies (Griffith induced stress decreased and was located at the T-joint between the
et al. 2012; Ho et al. 2012). The natural frequencies derived from torque tube and the pylon at α = 30◦ in Fig. 19(b). Spectral analysis of
experimental measurements showed good agreement with finite the fluctuating stresses indicated dominant frequencies in the 6 Hz range
element analysis (FEA) predictions of the wind-excited dynamic corresponding to the modal frequencies, with increasing side-to-side and
response, such as a modal frequency of 3 Hz corresponding to the first flexural motions of the concentrator at α = 30◦ caused by the peak hinge
torque tube bending mode 3 in Fig. 18. However, higher order modes moment about the torque tube. Although coupled or two-way FSI using
with dependence on the stiffness properties of joints and drive mecha LES is a promising method to investigate dynamic wind loads on he
nisms, such as out-of-plane support structure bending modes, were not liostats, the computational effort with increased accuracy models is very
accurately predicted by the FEA model. Furthermore, the low-frequency high (Pfahl et al. 2017a; Wolmarans and Craig 2019). Consideration of
modes of vibration showed increased damping by 24–120% due to the dynamic amplification of the load fluctuations on the heliostat
aerodynamic damping excited by the wind at speeds of 5–15 m/s components requires further investigation to understand the conditions
compared with the calm winds during the hammer-excited tests. Com that promote the coupling effects between ABL turbulence and modal
parison of the modal frequencies on different heliostat sizes and eleva frequencies of the structure.
tion angles showed that the azimuth drive modal frequency increased
from 1.28 Hz to 2.28 Hz at α = 90◦ and from 1.04 Hz to 1.75 Hz at α = 4.3. Wind-induced tracking error and operational performance
0◦ with increasing heliostat size from 37 m2 to 60 m2 (Ho et al. 2012).
Vásquez-Arango et al. (2015) validated a finite element analysis Ho et al. (2012) investigated two rigid-body vibrational modes at
(FEA) model with hammer-excited experimental modal data, which 1–2 Hz of the 37 m2 NSTTF heliostat correlating to backlash of the
showed that the shapes of vibration corresponding to rigid body modes elevation and azimuth drives in a field experiment test at Sandia Na
of the mirror frame, such as the oscillation about the elevation axis, were tional Laboratories (Ho et al. 2012; Griffith et al. 2015). Furthermore,
excited by fluctuating wind loads. Admittance functions were applied hammer-excited experimental modal analysis showed that the truss
Fig. 19. Maximum von Mises stress contour from a one-way FSI modal analysis of the LH-2 heliostat at (a) α = 90◦ , and (b) α = 30◦ (Wolmarans and Craig 2019).
73
Fig. 20. Heliostat deformed shape for out-of-plane bending mode 2 due to wind excitation that can impact optical performance through deviation of the beam
centroid (Ho et al. 2012).
member to torque tube interfaces due to out-of-plane bending modes total tracking deviation of heliostats. Wind-induced oscillations and
(Fig. 20) were most vulnerable to wind-induced stresses. Maximum deformations at frequencies below 4 Hz in the amplitude spectra
beam deviations of 0.17 m and 1.58 m in the horizontal and vertical (Fig. 21b) most significantly impacted the optical performance of the
directions were observed on the tower target, compared with deviations heliostat at a mean wind speed of 4.8 m/s and turbulence intensity of
of 0.1 m and 0.25 m due to gravity in the absence of wind (Ho et al. 26% (Fig. 21c). To complement the relationships between quasi-static
2012). peak wind loads and ABL turbulence in Section 2, spectral analysis
Dynamic photogrammetry measurements on the 48.5 m2 Stellio correlations between the fluctuating components of the wind velocity
heliostat by Blume et al. (2020) revealed that the wind-induced tracking and the resonant component of the tracking deviations in field in
deviation of 0.44 mrad RMS (Fig. 21a) contained a resonant component vestigations would provide a further insight into the wind-induced os
RMS value an order of magnitude smaller than the combined RMS values cillations that impact the operational performance of a range of full-
of the mean and background components. This tracking deviation scale heliostat prototypes.
caused by the wind contributed to approximately one third of the typical
Fig. 21. (a) Time history of the wind-induced tracking deviation in the lateral (x) and longitudinal (y) directions of the Stellio heliostat concentrator at α = 45◦ and
β = 76◦ ; (b) amplitude spectra of the wind-induced tracking deviations with a low-pass filter and cut-off frequency of 4 Hz; (c) time history of wind speed averaged
over four ultrasonic anemometers on measurement mast at the Jülich DLR field site (Blume et al. 2020).
74
Fig. 22. Different layouts of a heliostat field. (a) A radial heliostat field, Noor III in Morocco. Image from www.masen.ma, (b) a polar heliostat field, PS10 in Spain.
Image from www.eusolaris.eu.
5. Aerodynamics of a heliostat field
Heliostat fields are arranged in rows in a radial (Fig. 22a) or polar

(Fig. 22b) configuration surrounding a central tower. For an optimum
optical performance, the radial distance between the rows in a heliostat
field typically ranges between a value larger than the chord length of the
mirror panel, x/c > 1 in the inner field rows, to x/c = 8 at the perimeter
of the field (Hui 2011). Heliostats close to the tower have field densities
greater than 40% with smaller shading effects and are typically spaced
less than 20 m apart for a typical heliostat mirror area of 120 m2 (Noone
et al. 2012). With increasing distance from the central tower, the field
density decreases to less than 20% and spacing between heliostats of up
to 45 m at the outer boundary of the field (Pfahl et al. 2011c; Noone et al.
2012). The layout of heliostat fields in power tower plants has been
optimised disregarding of wind load and primarily with respect to the
optical efficiency of the field. However, static wind loads on tandem
Fig. 23. Turbulence intensity in different rows of a heliostat field as a function
heliostats are strongly dependent on the spacing between the heliostat
of heliostat elevation angle (Sment and Ho, 2014).
mirrors, defined by the gap ratio x/c and the heliostat field density
defined as the ratio of mirror area to land area. As wind flows over a
heliostat, a region of disturbed flow is created downstream in its wake. between five rows of the NSTTF heliostat field to measure of the tur
Within the field, the mean flow and turbulence characteristics might be bulence statistics of the flow in the vicinity of the heliostats. Fig. 23
significantly different from the incoming ABL and thus alter the wind shows that an increase of turbulence intensity to more than 50%
loads on heliostats in the field from those on a single heliostat. Hence, downstream of the first and second row of heliostats at α = 90◦ (vertical)
wind loads on heliostats at different in-field positions could be evaluated and 45◦ . For the heliostats in stow (not shown) however, turbulence
given knowledge of differences in flow and turbulence characteristics intensities showed only a small variation in downstream rows and
within a field. This provides a chance to optimise the design and cost of a remained below the maximum turbulence intensity of 20% approaching
heliostat field, with respect to the inner flow field aerodynamics repre the outer row of the field (Sment and Ho 2014).
sented by a combination of ABL turbulence and upstream heliostat Within the boundary layer, the variable shear and turbulence in
wake-generated turbulence. affect the development of the wake of a heliostat and the turbulence
structure in its wake significantly. Jafari et al. (2020a) conducted ve
locity measurements in the wake of a heliostat model placed in simu
5.1. Heliostat wake measurements lated atmospheric boundary layers in the wind tunnel to characterise the
turbulence variations in the heliostat wake. It was found that in the wake
Flow around a heliostat, in the absence of the support structure and of a heliostat, the turbulence properties were significantly different from
the pylon, is resembled by flow around a thin flat plate. As flow passes the atmospheric boundary layer. The results showed a reduction in mean
around a thin flat plate, it separates from the plate at its edges and a low- velocity in the wake, which did not recover over the measured down
pressure region is formed in its immediate downstream. The separated stream distance up to x/c = 8. This was accompanied by an increase in
shear layers then roll up into large scale vortices shedding into the wake. turbulence intensity up to x/c = 4, with a peak at approximately x/c =
Blockage of the flow by the plate and vortex shedding in the wake lead to 1.5 where the streamwise and vertical turbulence intensities increased
a reduction of mean velocity and an increase in turbulence intensity. The by more than 12-times their incoming values at elevation angles of 60◦
alternate shedding of the rolled-up shear layers into the wake creates and 90◦ . Furthermore, it was found that in the wake immediately
oscillations in the flow, characterised by the dominant frequency of downstream of the heliostat, the length scales of turbulence were
vortex shedding. The aerodynamics of multiple heliostats differ from a significantly smaller as the large inflow turbulence length scales were
single heliostat due to the interference of their wakes with each other broken into smaller scales.
and the interaction of the downstream heliostats depending on their The variations of turbulence intensity in the heliostat wake at
arrangement and spacing between them. different streamwise distances indicates the impact of field density on
The profiles of mean velocity and turbulence intensity of the heliostat wind loads. For example, due to the higher turbulence intensity
approaching boundary layer were characterised by Sment and Ho caused by the heliostat wake, the unsteady wind loads in high-density
(2014) using three tri-axial ultrasonic anemometers mounted on a zones of a heliostat field at x/c = 1 − 3 are greater than in low-density
weather tower upstream of a row of instrumented heliostats. Ane zones. This shows the impact of dynamic wind loads for design of he
mometers were also mounted on the heliostats and on portable towers
75
liostats as they are likely to influence the dominant frequencies of the area (GBA), defined as the ratio of the area of upstream blockage pro
unsteady and dynamic loads on heliostats in dense zones of a field. jected to wind direction, including external and internal fences and
Furthermore, despite the reduced mean wind speed within the field, upstream heliostats, over the field ground area. Peterka et al. (1989)
static wind loads such as the hinge moment may increase within the field reports the ratio of the peak drag and lift force coefficients in a field as a
depending on the field density and the elevation angle of heliostats function of GBA as shown in Fig. 25(a–b). The results show cases where
during operation. the peak coefficients are larger than a single heliostat, shaded by red in
Fig. 25(a–b). The reason for increase of wind loads was not explained by
Peterka et al. (1986). Furthermore, the elevation angles and heliostat
5.2. Loads in heliostat field arrays configurations for the presented results were not provided, and it is not
clear for which conditions the wind loads were larger than a single he
The review of the aerodynamics of tandem flat plates and side-by- liostat. Moreover, the results were only presented as a function of GBA,
side flat plates shows that the wake flow around multiple heliostats which includes the effects of both the fence and blockage by upstream
and thus the wind loads on in-field heliostats can differfrom those on a heliostats. Hence, the influence of the fence on the wind loads was not
single heliostat. One of the critical parameters that influences the wind distinguished.
loads is the non-dimensional gap in the longitudinal direction with Peterka et al. (1987) measured the wind loads on 1:60 scale-model
respect to the mirror chord length, x/c, between the heliostats in an heliostats in the fourth row of a four-row arrangement for two
array. Emes et al. (2018) investigated the variation of the stow wind different gap ratios between consecutive rows, x/c = 6.4 and x/c = 3.07,
loads on two tandem heliostats and showed that the peak lift force co representing low- and high-density zones of a heliostat field. The mean
efficient on the second tandem heliostat in stow was up to 7% larger drag force coefficient of a fourth-row heliostat was found to be 12%
than that for the single stowed heliostat for x/c > 1.5. As shown in lower than that of a front-row heliostat at x/c = 3.07. For a higher field
Fig. 24, Jafari et al. (2020b) found that the peak hinge moment coeffi density, the reduction in the mean drag coefficient increased to only
cient on a tandem heliostat increased to 1.5-times that on a single he 32% of that in the first row. In contrast, the peak drag force coefficient
liostat at an elevation angle of 30◦ and more than double at elevation on of a fourth-row heliostat with x/c = 6.4 was found to be 40% larger
angles of 60◦ and 90◦ . Despite the lower mean pressure coefficient on the than that of a front-row heliostat. Pfahl et al. (2011c) measured the wind
tandem heliostat, a region of large-magnitude peak pressure existed at loads on 1:20 scale-models of a four-row tandem arrangement with 30
the leading edge of the panel. Furthermore, analysis of the unsteady m2 mirror area for field densities of 10% and 50% corresponding to gap
pressure distributions showed an increased unsteady centre of pressure ratios (x/c) between the mirrors of 5.5 and 1.5, respectively. Peterka
variation on the second tandem heliostat, specifically at elevation angles et al. (1987) and Pfahl et al. (2011c) found up to 50% reduction in peak
of 30◦ and 60◦ . The unsteady variations of the position of the centre of drag and lift forces on a second heliostat at α = 90◦ compared to the
pressure as a result of the larger turbulence intensity in the wake were front-row heliostat in a tandem arrangement at a field density of 50%.
found increase the mean and peak hinge moment coefficients on the The larger peak drag coefficient may be correlated with an increase in
second heliostat. The large increase of the hinge moment coefficient can longitudinal turbulence intensity of the flow, however the relative
outweigh the reduced wind speed in the wake with respect to the gap contribution of the longitudinal and vertical turbulence components to
between the heliostats and the elevation angle of the heliostat panel. For the lift and hinge moment coefficients on operating heliostats has not
example, at an elevation angle of 30◦ and x/c between 4 and 8, the mean been determined. This highlights the importance of characterisation of
wind speed reduced by less than 10%, while the hinge moment coeffi turbulence in the wake of heliostats and its effect on the wind loads, and
cient was 50% larger than the single heliostat, leading to an increase of measurement of wind loads in a field. Understanding the variations of
between 20% and 50% in the peak hinge moment. Hence, the results wind loads within a heliostat field can help to improve the field design
highlight an opportunity to modify the heliostat design for in-field he with respect to the wind loads. For regions of a field with reduced wind
liostats compared to field-edge heliostats. speed and increased turbulence intensity, the structural stiffness and
In the literature, wind tunnel studies have been performed to study foundation depth of heliostats can be decreased if the dynamic loads are
the influence of fences on the wind loads on heliostats in field ar not overcompensated by an increase in unsteady wind loads.
rangements. Peterka et al. (1986) measured the wind loads on a heliostat In a similar experiment, Pfahl (2018) measured the wind loads on a
placed in an array with perimeter and in-field fences. The configuration heliostat in the fourth row of an array in presence of a fence upstream of
of the heliostat array was chosen based on different regions of a field the first row. The fence had a porosity of 40% and height equal to 1.25
with different densities. Fences with porosities of 0.4, 0.5 and 0.6 and times the heliostat hinge height. Different cases with varied distances
two heights, equal to 0.9 and 1.35 times the heliostat hinge height, were between the heliostat rows and between the fence and the front row
investigated. They found that with addition of the fence, the mean drag were investigated, through which GBA varied between 0.053 and 0.46.
force coefficient on a heliostat at α = 90◦ and a wind direction of 250◦ in Their results in general showed that the maximum wind load coefficients
the third row of an array was reduced from approximately 1 to 0.45. The at operating elevation angles were less than a single heliostat for the
results in Fig. 25(c) were presented as a function of generalised blockage investigated range of GBA. As shown in Fig. 26, the peak lift force co
efficient on a stowed heliostat was up to 25% larger than a single he
liostat for GBA values less than 0.1. The increase in the stow lift force
coefficient was suggested to be related to an increase in vertical velocity
component downstream of the fence. If the entire field is to have a
consistent heliostat design, according to Pfahl (2018), application of
fences therefore may not be beneficial due to the increase of the lift force
in stow position and the negligible impact of the fence on low density
regions of the field. As the results were presented as a function of GBA,
the effect of fence was not differentiated from the effect of blockage by
heliostats at the upstream rows. Pfahl (2018) discussed that the uncer
tainty in the reported results was large due to the limited measurement
cases.
Fig. 24. Peak hinge moment coefficient on a tandem heliostat normalised to a
single heliostat as a function of longitudinal gap spacing x/c between tandem
heliostats at elevation angle α (Emes et al. 2018; Jafari et al. 2020b).
76
Fig. 25. (a) Peak drag and (b) peak lift force coefficients in an array with perimeter and in-field fences normalised with the peak force coefficients on a single
heliostat as a function of generalised blockage area (GBA). The red shaded regions show cases where the peak wind load coefficients are larger compared to a single
heliostat. (c) A schematic of the heliostat array demonstrating the calculation of GBA.
Reproduced from Peterka et al. (1989).
Fig. 26. Peak aerodynamic coefficients as a function of GBA, normalised with respect to GBA = 0 for an isolated heliostat in (a) maximum operating position
(Peterka and Derickson 1992); (b) stow position (Pfahl 2018).
5.3. Wind load mitigation techniques scales. Based on the variation of mean velocity, turbulence intensity and
integral length scale behind the fences, it was estimated that the peak
The wind load reduction on in-field heliostats, in wind tunnel ex drag force on a heliostat at the vertical position could be reduced by 48%
periments by Peterka et al. (1986) and Pfahl (2018), compared to a with utilisation of a wire mesh fence with a porosity of 0.46 using the
single heliostat were presented as a function of GBA, such that the effects developed relationships in Jafari et al. (2018). it was predicted that the
of both the fence and blockage by upstream heliostats were not distin peak lift force on a stowed heliostat could be reduced by 53% behind a
guished. Hence, the manipulation of inflow ABL turbulence by the fence wire mesh fence with a porosity of 0.46 based on the correlation given in
and its effectiveness in wind load reduction were not reported. Turbu Jafari et al. (2019a), as shown in Fig. 27(b). With increasing the porosity
lence properties downstream of mesh fences (Fig. 27b) of various mesh of the wire mesh fence to 0.75, the reduction in peak drag and lift forces
opening widths and porosities were determined from experimental could only reach 19% and 15%, respectively. The measurement of forces
measurements in a wind tunnel (Jafari et al. 2021). It was found that on a heliostat behind the fence can further verify these estimated peak
with application of fences with porosities between 0.46 and 0.75, an load reductions derived from the turbulence intensities and length scales
inflow streamwise turbulence intensity of 12.5% could be reduced to reductions due to the fence. For such a method to be employed in a
between 8.8% and 9.9%. Furthermore, a significant reduction in the heliostat field, further study is necessary in the future. to determine the
integral length scale of turbulence was shown immediately downstream optimum geometric parameters of the mesh fence, including its height
of the fences and grew afterwards with increasing the downstream dis and distance to the heliostats.
tance, with the longitudinal length scale remaining 25% and the vertical Wind load reduction by fences may be more appropriate for helio
length scale remaining 21% below the inflow level for the fences with stats of smaller dimensions, due to the increased material cost of larger
porosities between 0.46 and 0.64. Through comparison of the turbu fences that would be required for a field of large-scale heliostats.
lence reduction behind wire mesh fences with different porosities and Although fences at the perimeter of the field have been shown to have
mesh opening widths, it was found that porosity was the main factor negligible impact on the forces on heliostats with increasing distance
which determined the reduction in turbulence intensity and length into the field, a modification in the design of perimeter fences to
77
Fig. 27. Wind load mitigation techniques using (a) heliostat edge-treatment devices (Pfahl et al. 2013), and (b) a heliostat field perimeter porous wire mesh fence
(Jafari et al. 2021). The plots show the effect of wire mesh fence porosity on the predicted reduction of (c) the peak drag force/coefficient on a heliostat at α = 90◦ ,
(d) the peak lift force/coefficient on a heliostat at α = 0◦ (Jafari et al. 2021).
heliostat edge-mounted devices may reduce the wind loads on in-field 1 Hz at a height of 10 m. Standard wind velocity data at automatic
heliostats. The high overturning moments on a stowed heliostat are weather stations (Bureau of Meteorology 2020; National Climatic Data
due to the vertical velocity component of the turbulent flow separating Center 2020) are not obtained at a sufficient frequency to reliably
at the leading edge, which creates suction on the other side of the mirror determine the longitudinal and vertical turbulence intensities that
and a high-pressure difference between the upper and lower heliostat impact the maximum heliostat wind loads (Blackmon 2014). Long-span
surfaces. Wind tunnel experiments by ToughTrough indicated that cable-supported bridges are sensitive to peak gusts of a duration of the
fence-like “spoilers” (Fig. 27a) can reduce separation and suction near order of 2–3 s (Xu 2013), whereas stowed heliostats are exposed to
the leading edge in stow position, leading to 40% wind load reduction shorter duration gusts of approximately 1 s (Pfahl 2018). Hence, it is
and 30% weight reduction of heliostat support structure (Pfahl et al. expected that the relevant gust period for a heliostat is shorter than that
2014b). A disadvantage of such flow manipulator devices is the addi of a building and thus the dynamic response and vibrational mode
tional maintenance cost to clean the mirrors and the shading of the shapes of heliostats are different. The collection of high frequency (i.e.
mirrors. second) wind velocity and solar radiation data at concentrating solar
power plant sites over an extended duration (e.g. years) would increase
6. Discussion the accuracy of annual field efficiency models through an improved
resolution of operating load data. The transient nature of the ABL should
6.1. Resolution of heliostat field measurements in plant performance therefore be accounted for in the design of a heliostat field, including the
models wind load predictions and the assessment of operational performance
models.
Typical meteorological year (TMY) data contains wind and solar
radiation data averaged over a duration of one hour as an input to 6.2. Assessment of critical aerodynamic load cases of a heliostat
annual solar field efficiency models. In practice during operation of a
power tower field, however, heliostats are stowed based on a 3-second Design wind load codes and standards provide aerodynamic shape
gust wind speed (Price et al. 2020). Second-generation heliostats were factors, external pressure coefficient and design external pressure,
defined by Murphy (1980) with specifications for gust wind speeds of 22 aerodynamic (drag) force coefficient cF and the centre of pressure dis
m/s and 40 m/s at a 10-m height for the maximum operational and stow tance lpx /c from the windward edge of simple-shaped structures based
survival design conditions, respectively. This is the same as the 3-second on a characteristic length c of the structure. For example, Chapter 5 -
gust wind speed stated in design wind guidelines and norms for build Wind Loads of ASCE 7–02 (2002) provides a range of tables containing
ings and other physical structures with natural frequencies smaller than the design pressures for solid freestanding walls, solid signs and
78
monoslope roofs with tilt angles 10-30◦ (in increments of 5◦ ) and aspect thickness of the pedestal and torque tube can be reduced for operating
ratio of the cross-sectional roof area varying between 1/5 and 5. conditions at low wind speed sites, whereas the torque tube is a critical
Furthermore, the IEC 61400-1 (2005) wind turbine design standard component that can lead to structural failure in stow position at
provides guidance on the static and dynamic loads on wind turbine increased wind speeds. Wind tunnel experiments have shown that for
components, considering the effects of turbulence intensity and length measurement of the unsteady drag force on a heliostat at α = 90◦ ,
scales and the variation of average and gust wind speeds across the rotor similarity of the streamwise velocity spectrum is required and a model
plane. However, heliostats have a non-standard shape that does not with larger dimensions (i.e. smaller scaling ratio) can be used. In
conform to conventional shapes of buildings (ASCE 7-02 2002; EN 1991- contrast, accurate measurement of the unsteady lift force on a stowed
1.4 2010; AS/NZS 1170.2 2011) and rooftop solar panels (ASCE/SEI 7- heliostat requires similarity of the vertical turbulence spectrum, which
16 2016) associated with corner vortices and separation at the leading can only be achieved for a model with smaller dimensions or larger
edge of the building roof (Kopp et al. 2012). The thin plate and tubular scaling ratio (Jafari et al. 2019b). The relative contribution of the lon
geometries of heliostat facets, support beams, torque tube and pedestal gitudinal and vertical components of turbulence, for a stowed heliostat
are not applicable to the design procedures outlined for buildings and and over the range of heliostat operating conditions with varying gap
wind turbines in terms of their size, shape and position within the lowest between the lower heliostat edge and the ground, should be further
10 m of the ABL. This can lead to under-estimation of the peak loads on verified through wind tunnel and full-scale measurements. Analysis of
heliostats, such as in stow position due to the large dynamic response wind loads on full-scale heliostats with respect to the incoming wind
caused by near-surface gust events (Durst 1960; Mendis et al. 2007). turbulence measured simultaneously can verify the scaling effects
Table 2 shows the maximum operating wind load configurations, in observed in wind tunnel experiments to provide a more reliable esti
terms of the elevation and azimuth angles that result in the peak wind mation of wind loads.
load coefficients reported in wind tunnel measurements in Table 1 (refer
to Section 2.2). For wind approaching an upright heliostat at α = 90◦ 6.3. Modal analysis of heliostat vibrations and wind-induced
from the front (β = 0◦ ) or back (β = 180◦ ), the maximum drag force Fx displacements
on the concentrator leads to the maximum overturning moment My at
the base of the heliostat pylon for design of the foundation. Similarly, on Measurements of local deformations and displacements on full-scale
a heliostat inclined at α = 30◦ , the maximum lift force on the heliostat heliostats have provided an insight into the dynamic wind loads, such as
panel leads to a maximum hinge moment MHy about the elevation axis of vibrations and fatigue loads on drive units and support structure com
the heliostat in operation that impacts the design of the torque tube and ponents. Modal analyses have been conducted in the literature both
elevation drive. The maximum load case for the azimuth drive is the computationally and experimentally to determine the mode shapes and
moment about the vertical axis of an upright operating heliostat (α = frequencies of a heliostat structure. Low-frequency vibrational modes
90◦ ) with wind approaching from an oblique angle β = 60◦ and 120◦ . corresponding to quasi-static sway motion of the heliostat subjected to
The wind load coefficients found by Peterka et al. (1989) apply to one time-averaged loads can be accurately reproduced by numerical simu
case of the ABL with limited information on the turbulence spectra and lations. However, modes that are dependent on the stiffness and
length scales, particularly in the vertical turbulence component that is damping of joints, such as elevation and azimuth drives, are most
crucial to the maximum wind loads in stow position. The maximum accurately characterised through full-scale experiments and two-way
wind loads on heliostats often considered wind impacting the front of fluid–structure interaction that captures the gust spectrum range
the heliostat at β = 0◦ , however the maximum wind loads on a heliostat (~1–2 Hz) of the fluctuating load distribution caused by backlash or slop
at β = 180◦ can be larger and the presence of an upstream heliostat in the gear drives (Griffith et al. 2012; Ho et al. 2012). High-amplitude
influences the spectral peak of pressure variations in operating positions dynamic response of the pylon and support structure was less likely to be
(Yu et al. 2019). The number of working conditions for azimuth- impacted by the shedding of vortices from the heliostat structure
elevation heliostat configurations can be reduced from 130 to 13 (Wolmarans and Craig 2019), although this applies to a broad range of
through the application of uniform design method and regression frequencies depending on the wind speed and the heliostat size (Ho et al.
analysis to all wind load coefficients (Xiong et al. 2019). The contribu 2012). Hence, the heliostat structure should be designed to avoid wind
tion of spectral energy in the turbulent eddies to wind loads and the loads that cause high-amplitude or high-cycle counts in the drive com
resulting aerodynamic effects on heliostat geometry over a larger range ponents that result from resonant effects due to convergence of the
of orientations has been investigated in more detail in recent wind modal frequencies with the gust frequencies of energy-containing eddies
tunnel experiments (Pfahl et al. 2015; Emes et al. 2017; Emes et al. in the approaching wind and the vortex shedding frequency from up
2019a; Jafari et al. 2019a). stream heliostats. Further work is still necessary to examine dynamic
Prediction of the design loads on heliostats should allow for the wind loads on heliostats positioned inside the heliostat field. For
maximum operating cases and stow cases, due to both the scaling pa instance, field measurements can provide validation points to comple
rameters of individual components and the level of ABL turbulence ment numerical studies to investigate load amplification factors asso
represented by the surrounding terrain. The influence of the heliostat ciated with different operational wind speeds and turbulence
concentrator aspect ratio (Pfahl et al. 2011a) and the pylon height (Emes characteristics over an increasing range of heliostat orientations and
et al. 2017; Jafari et al. 2019a) have a large effect on the maximum structural designs.
aerodynamic coefficients, whereas small gaps between mirror facets
have a negligible impact on the pressure distribution and the wind loads 6.4. Dynamic wind effects on operational heliostat tracking error
(Wu et al. 2010). Structural reliability of the heliostat components
through stress analysis by Benammar and Tee (2019) suggested that the Wind engineering design standards do not account for the dynamic
effects of heliostats, such as a dynamic response or amplification factor
in AS/NZS 1170.2 (2011) for slender buildings and large permanent
Table 2 structures (H ≤ 200 m) with natural frequencies less than 1 Hz. To avoid
Critical operating load cases of an azimuth-elevation heliostat.
structural excitation due to buffeting and torsional galloping, the natural
Maximum aerodynamic coefficient α(◦ ) β(◦ ) frequency of a long inclined flat plate (i.e. solar array) is recommended
F x , My 90 0, 180
to be greater than 5 Hz. Hence, the essential scaling parameters of the
Fz , MHy 30 0, 180
heliostat structure and the aerodynamic loads on the tubular compo
Mz 90 60, 120
nents were shown to be very sensitive to the variations of turbulence
intensity and length scale with height and surface roughness in the ABL
79
(Emes et al., 2020c,a). Although square-mirrored heliostats are less 1) In wind standards, turbulence intensity and integral length scale
likely to be exposed to torsional vibrations, the ratio Lx /c of the integral profiles are only given for heights above three metres. However,
length scales in the longitudinal and vertical directions to the heliostat there is demand for smaller heliostats as they are advantageous for
chord length significantly affects the peak wind loads on heliostats in high-temperature applications, such as hydrogen production due to
operating and stow positions (Emes et al. 2017; Jafari et al. 2019a). lower astigmatism losses. Therefore, field investigations of the wind
Based on the common sizes of heliostat mirrors that are currently characteristics between one and three metres height for typical solar
manufactured, Lxu /c ≈ 6.5 in an open country terrain with z0 = 0.03 m sites would be beneficial.
(ESDU 85020 2001). However, Lxu /c decreases with increasing surface 2) The maximum operational loads and the stow survival loads have
roughness to Lxu /c = 5.5 at z0 = 0.05 m and Lxu /c = 4.5 at z0 = 0.1 m. To been defined by the heliostat orientation with respect to the wind. It
reduce the maximum wind loads as Lxu /c and Lxw /c approach unity, a is most important to model the range of reduced frequencies of the
heliostat of fixed mirror chord length can be stowed at a lower elevation turbulence spectrum that contribute to the unsteady forces on he
axis height H that is closer to the ground (Pfahl et al. 2017b) through a liostats in wind tunnel experiments in order to reduce the scaling
reduction of H/c and Lxu /c. effect on the measured peak wind loads and accurately reproduce the
wind loads on the full-scale structure. These maximum heliostat load
6.5. Variation of wind loads on heliostats throughout a field cases were referenced to design wind speeds and turbulence in
tensities at a constant height, such as the standard reference height of
Due to the variation in heliostat orientations across a field with 10 m in wind load codes and standards. An increased resolution of
respect to the wind, the aerodynamic loads on some heliostats in field-scale wind measurements is essential to understand the effect of
favourable orientations can be reduced with respect to the maximum surface roughness on the peak aerodynamic coefficients at a range of
load cases in the field. Statistical correlation of wind speed and DNI data heliostat field sites to fully characterise the longitudinal and vertical
with heliostat tracking angles at the Plataforma Solar de Almeria (PSA) turbulence intensities and length scales that impact the maximum
CESA-I field by Emes et al. (2020c) indicated that a stowing strategy wind loads for operating serviceability and stow survivability
based on wind speed and direction can increase the annual operating considerations.
time of the heliostat field by 6% with increasing stow design wind speed 3) Scaling factors and relationships have been derived in scale-model
from 6 m/s to 12 m/s. For an assumed 10-minute stow transition from wind tunnel experiments that account for the variation in wind
operating positions of the heliostat field, a stowing strategy that allowed loads due to geometry effects, such as the aspect ratio, mirror chord
“protected” heliostats with reduced wind loads at β = 90 ± 15◦ to length and pylon height from a baseline square-mirror azimuth-
continue to operate at wind speeds larger than 10 m/s was investigated. elevation heliostat. Further investigations should focus on the in
Emes et al. (2020c) found this strategy achieved an additional 24 GWh fluence of wind direction and heliostat shape due to changes in
of thermal energy collected annually by heliostat field operation during aspect ratio, and the effect of the gap between the lower heliostat
periods that would conventionally stow the entire field. It is therefore edge and the ground on the aerodynamic coefficients.
apparent that there is a potential to increase the operating performance 4) Dynamic wind loads and modal analysis of local deformations of
through consideration of wind load distributions and “smart” stowing heliostat components was most effectively investigated in field en
strategies of the heliostat field to maximise the energy yield of a power vironments with the mechanical and structural properties of a full-
tower plant. scale heliostat. Due to the large range of heliostat sizes and struc
Porous fences were found by Jafari et al. (2021) to reduce the tur tural types, the design wind loads are commonly estimated using a
bulence intensity and integral length scales by 20–25% relative to the combination of peak aerodynamic coefficients and appropriate load-
incoming ABL, but the material cost of perimeter fences for large he response correlations from finite element models at the relevant
liostats and their area of influence into a heliostat field remained a design wind speeds. Dynamic amplification factors for alternative
research question. Other methods to reduce the wind loads on heliostats heliostat designs to a conventional azimuth-elevation tracking
positioned at the inner rows of a field include the attachment of “edge configuration (e.g. spinning axis, tilt-roll) should be further investi
treatment” devices to the heliostat, such as to mitigate the impact of gated, such as the lowering of the mirror closer to the ground in stow
vortex shedding from the leading and trailing edges. Alternatively, the position and resonance effects in the transition to stow due to in
installation of a series of slender plate or rod large-eddy break-up creases of wind speed at intermediate operating angles.
(LEBU) devices at the perimeter of a heliostat field can reduce the effect 5) Systematic experimental studies in small-scale boundary layer wind
of the energetic turbulent eddies in the ABL on the heliostat field tunnel measurements have effectively simulated the aerodynamics
operation. Characterisation of the flow and wind loads using these and quasi-static wind loads through investigation of the critical
methods are required for an improved understanding of their effec scaling parameters of isolated, tandem and arrays of heliostats over a
tiveness. A techno-economic analysis of the cost-effectiveness of fences range of wind turbulence conditions in the ABL. Wind loads on the
in heliostat fields is required to assess the sensitivity of reduced loads structural heliostat components, such as bending moment reactions
and heliostat capital cost with respect to the increased land area and to be resisted by the drives, torque tube and foundation, have been
material cost of the fence construction. characterised through scale-model testing in wind tunnel experi
ments. The variation of wind-induced displacements due to opera
7. Conclusions tional wind loads on in-field heliostats has been related to the vortex
shedding and vibrational modes, but simultaneous load and wake
There has been an extensive range of studies on heliostat aero measurements can provide understanding on how the field spacing
dynamic wind loads in the literature. The aerodynamic coefficients form and orientation affects the operational performance of individual
a basis for the design wind loads on isolated heliostats, which were heliostats throughout the field. Instrumenting arrays of heliostats in
shown to depend on the geometric parameters of the heliostat, along different rows within a field would also be highly beneficial to better
with wind speed and turbulence parameters in the atmospheric understand the relative contribution of heliostat-generated wake
boundary layer (ABL). The following major conclusions can be drawn turbulence and incoming ABL turbulence on the heliostat field
from the literature to further develop the understanding of the aero aerodynamics, wind load distributions and wind-induced tracking
dynamic wind loads on heliostats: errors during operation of a field.
6) It is postulated that the total cost of the heliostat field is conservative
as all heliostats are designed based on the maximum wind load co
efficients on a single heliostat, while the loads on heliostats in
80
various rows vary across the field. Heliostat wind loads in arrays Cook, N.J., 1985. The designer’s guide to wind loading of building structures, Part 1:
Background, damage survey, wind data and structural classification. Building
have presented wind load reductions on in-field heliostats based on
Research Establishment, Garston, UK.
the concept of GBA, however the fence’s independent effect was not Cook, N.J., 1997. The Deaves and Harris ABL model applied to heterogeneous terrain.
distinguished from the impact of upstream heliostat blockage. Un Journal of Wind Engineering and Industrial Aerodynamics 66, 197–214.
derstanding the variation of wind loads within a heliostat field Counihan, J., 1975. Adiabatic atmospheric boundary layers: a review and analysis of
data from the period 1880–1972. Atmospheric Environment 9 (10), 871–905.
through the systematic analysis of independent wind load reduction De Paepe, W., Pindado, S., Bram, S., Contino, F., 2016. Simplified elements for wind-
methods can help to improve the field design with respect to the tunnel measurements with type-III-terrain atmospheric boundary layer.
wind loads. Characterisation of the flow and wind loads using Measurement 91, 590–600.
Department of Energy, 2017. “The SunShot 2030 Goals”, DOE/EE-1501. Solar Energy
favourable methods to reduce heliostat wind loads, such as perimeter Technologies Office, USA.
and in-field fences and edge treatment devices, should independently Durst, C.S., 1960. Wind speeds over short periods of time. Meteorological Magazine 89
assess their cost-effectiveness and feasibility in power tower plants. (1960), 181–186.
Emes, M., Jafari, A., Arjomandi, M., 2020b. Wind load design considerations for the
elevation and azimuth drives of a heliostat. AIP Conference Proceedings 2303 (1),
There is a strong case for the development of design guidelines for 030013.
wind load predictions on full-scale heliostats that account for the effects Emes, M., Jafari, A., Collins, M., Wilbert, S., Zarzalejo, L., Siegrist, S., Arjomandi, M.,
2020c. Stowing Strategy for a Heliostat Field Based on Wind Speed and Direction. In:
of ABL turbulence based on the scaling of the heliostat structural com Proc., Proceedings of the 26th SolarPACES 2020 International Conference.
ponents and field layout. Such guidelines can benefit the operational Emes, M.J., Arjomandi, M., Ghanadi, F., Kelso, R.M., 2017. Effect of turbulence
performance of the plant and the material costs of manufacturing based characteristics in the atmospheric surface layer on the peak wind loads on heliostats
in stow position. Solar Energy 157, 284–297.
on the local wind conditions below heights of 10 m at different sites.
Emes, M.J., Arjomandi, M., Kelso, R.M., Ghanadi, F., 2019c. Turbulence length scales in
Accurate prediction of the maximum wind loads in real-scale operating a low-roughness near-neutral atmospheric surface layer. Journal of Turbulence 20
conditions provide greater confidence in field efficiency and power (9), 545–562.
tower plant performance models, which enhances the reliability of Emes, M.J., Arjomandi, M., Nathan, G.J., 2015. Effect of heliostat design wind speed on
the levelised cost of electricity from concentrating solar thermal power tower plants.
techno-economic analyses of the solar field operation and structural Solar Energy 115, 441–451.
design of the heliostat components. Emes, M.J., Ghanadi, F., Arjomandi, M., Kelso, R.M., 2018. Investigation of peak wind
loads on tandem heliostats in stow position. Renewable Energy 121, 548–558.
Emes, M.J., Jafari, A., Coventry, J., Arjomandi, M., 2020a. The influence of atmospheric
Declaration of Competing Interest boundary layer turbulence on the design wind loads and cost of heliostats. Solar
Energy 207, 796–812.
The authors declare that they have no known competing financial Emes, M.J., Jafari, A., Ghanadi, F., Arjomandi, M., 2019a. Hinge and overturning
moments due to unsteady heliostat pressure distributions in a turbulent atmospheric
interests or personal relationships that could have appeared to influence boundary layer. Solar Energy 193, 604–617.
the work reported in this paper. Emes, M.J., Jafari, A., Ghanadi, F., Arjomandi, M., 2019b. A method for the calculation
of the design wind loads on heliostats. AIP Conference Proceedings 2126 (1),
030020.
Acknowledgements EN 1991-1.4 (2010), Actions on structures, Part 1-4: General actions - Wind actions,
Eurocode, Brussels.
The authors acknowledge the support from the Australian Solar ESDU 85020 (2001), Characteristics of atmospheric turbulence near the ground, Part II:
single point data for strong winds (neutral atmosphere), Engineering Sciences Data
Thermal Research Institute (ASTRI) and funding provided by the
Unit, London.
Australian Renewable Energy Agency (ARENA) Grant 1-SRI002. Farell, C., Iyengar, A.K., 1999. Experiments on the wind tunnel simulation of
atmospheric boundary layers. J Wind Eng Ind Aerodyn 79 (1), 11–35.
References Gilooly, S. and Taylor-Power, G. (2016), Physical Modeling of the Atmospheric Boundary
Layer in the University of New Hampshire’s Flow Physics Facility, Honours Thesis,
307.
Abengoa Solar (2016), Abengoa, IDC and Khi Community Trust commence commercial Gong, B., Li, Z., Wang, Z., Wang, Y., 2012. Wind-induced dynamic response of Heliostat.
operation of Khi Solar One, the first solar tower plant in Africa, https://www.abe Renewable Energy 38 (1), 206–213.
ngoa.com/web/en/noticias_y_publicaciones/noticias/historico/2016/02_febrero/a Gong, B., Wang, Z., Li, Z., Zang, C., Wu, Z., 2013. Fluctuating wind pressure
bg_20160205.html, Last Accessed 19 March 2021. characteristics of heliostats. Renewable energy 50, 307–316.
Advisian Worley Group (2021), Khi Solar One 50MW solar power tower plant, https Griffith, D.T., Moya, A.C., Ho, C.K., Hunter, P.S., 2012. Structural Dynamics Testing and
://www.advisian.com/en/case-studies/khi-solar-one-50mw-solar-power-tower-pla Analysis for Design Evaluation and Monitoring of Heliostats. In: Proc., ASME 2011
nt, Last Accessed 19 March 2021. 5th International Conference on Energy Sustainability.
Andraka, C.E., Christian, J.M., Ghanbari, C.M., Gill, D.D., Ho, C.K., Kolb, W.J., Moss, T. Griffith, D.T., Moya, A.C., Ho, C.K., Hunter, P.S., 2015. Structural dynamics testing and
A., Smith, E.J., Yellowhair, J., 2013. Sandia Capabilities for the Measurement, analysis for design evaluation and monitoring of heliostats. Journal of Solar Energy
Characterization, and Analysis of Heliostats for CSP. Sandia National Lab.(SNL-NM, Engineering 137 (2), 021010.
Albuquerque, New Mexico. Ho, C.K., Griffith, D.T., Sment, J., Moya, A.C., Christian, J.M., Yuan, J.K., Hunter, P.S.,
Arbes, F., Wöhrbach, M., Gebreiter, D., Weinrebe, G., 2017. Towards high efficiency 2012. Dynamic Testing and Analysis of Heliostats to Evaluate Impacts of Wind on
heliostat fields. In: Proc., AIP Conference Proceedings. Optical Performance and Structural Fatigue. Proc, SolarPACES, Marrakech.
AS/NZS 1170.2 (2011), “Structural Design Actions - Part 2: Wind actions”, Standards Hui, T.M. 2011, Design and optimization of heliostat field using spinning-elevation sun
Australia and Standards New Zealand, Sydney. tracking method based on computational analysis, Master thesis, UTAR.
ASCE 7-02 (2002), Minimum design wind loads for buildings and other structures, IEC 61400-1 (2005), Wind turbines–part 1: Design requirements, Geneva.
American Society of Civil Engineers, Reston, Virginia. IRENA, 2020. Renewable power generation costs in 2019. International Renewable
ASCE/SEI 7-16 (2016), “Wind loads on building appertenances and other structures: Energy Agency, Abu Dhabi.
main wind force resisting system (directional procedure)”, American Society of Civil Iyengar, A.K.S., Farell, C., 2001. Experimental issues in atmospheric boundary layer
Engineers, Reston, Virginia. simulations: Roughness length and integral length scale determination. Journal of
Banks, D., 2011. Measuring peak wind loads on solar power assemblies. In: Proc., 13th Wind Engineering and Industrial Aerodynamics 89 (11), 1059–1080.
International Conference on Wind Engineering. Jafari, A., Emes, M., Cazzolato, B., Ghanadi, F., Arjomandi, M., 2020a. Turbulence
Benammar, S., Tee, K.F., 2019. Structural reliability analysis of a heliostat under wind characteristics in the wake of a heliostat in an atmospheric boundary layer flow.
load for concentrating solar power. Solar Energy 181, 43–52. Physics of Fluids 32 (4), 045116.
Blackmon, J.B., 2014. Heliostat drive unit design considerations–Site wind load effects Jafari, A., Emes, M., Cazzolato, B., Ghanadi, F., Arjomandi, M., 2020b. An experimental
on projected fatigue life and safety factor. Solar energy 105, 170–180. investigation of unsteady pressure distribution on tandem heliostats. AIP Conference
Blume, K., Röger, M., Schlichting, T., Macke, A., Pitz-Paal, R., 2020. Dynamic Proceedings 2303 (1), 030022.
photogrammetry applied to a real scale heliostat: Insights into the wind-induced Jafari, A., Emes, M., Cazzolato, B., Ghanadi, F., Arjomandi, M., 2021. Wire mesh fences
behavior and effects on the optical performance. Solar Energy 212, 297–308. for manipulation of turbulence energy spectrum. Experiments in Fluids 62 (2), 30.
Bureau of Meteorology (2020), Climate data services, http://www.bom.gov.au/climate/ Jafari, A., Ghanadi, F., Arjomandi, M., Emes, M.J., Cazzolato, B.S., 2019a. Correlating
data/, Last Accessed 1 April 2020. turbulence intensity and length scale with the unsteady lift force on flat plates in an
Cook, N.J., 1978. Determination of the model scale factor in wind-tunnel simulations of atmospheric boundary layer flow. Journal of Wind Eng and Ind Aero 189, 218–230.
the adiabatic atmospheric boundary layer. Journal of Wind Engineering and Jafari, A., Ghanadi, F., Emes, M.J., Arjomandi, M., Cazzolato, B.S., 2018. Effect of Free-
Industrial Aerodynamics 2 (4), 311–321. stream Turbulence on the Drag Force on a Flat Plate. Australasian Fluid Mechanics
Conference, Adelaide.
81
Jafari, A., Ghanadi, F., Emes, M.J., Arjomandi, M., Cazzolato, B.S., 2019b. Measurement Pfahl, A., Buselmeier, M., Zaschke, M., 2011c. Determination of wind loads on heliostats.
of unsteady wind loads in a wind tunnel: scaling of turbulence spectra. Journal of In: Proc., Proceedings of the 17th SolarPACES Conference.
Wind Eng and Ind Aero 193, 103955. Pfahl, A., Coventry, J., Röger, M., Wolfertstetter, F., Vásquez-Arango, J.F., Gross, F.,
Jain, A., Jones, N.P., Scanlan, R.H., 1996. Coupled flutter and buffeting analysis of long- Arjomandi, M., Schwarzbözl, P., Geiger, M., Liedke, P., 2017a. Progress in heliostat
span bridges. Journal of Structural Engineering 122 (7), 716–725. development. Solar Energy 152, 3–37.
Kaimal, J.C., Finnigan, J.J., 1994. Atmospheric Boundary Layer Flows: Their Structure Pfahl, A., Gross, F., Liedke, P., Hertel, J., Rheinländer, J., Mehta, S., Vásquez-Arango, J.
and Measurement. Oxford University Press. F., Giuliano, S. and Buck, R. (2017b) Reduced to Minimum Cost: Lay-Down Heliostat
Kolb, G.J., Ho, C.K., Mancini, T.R., Gary, J.A., 2011. “Power Tower Technology Roadmap with Monolithic Mirror-Panel and Closed Loop Control. In: Proc., SolarPACES 2017,
and Cost Reduction Plan”, SAND2011-2419. Sandia National Laboratories, Santiago.
Albuquerque, USA. Pfahl, A., Randt, M., Holze, C., Unterschütz, S., 2013. Autonomous light-weight heliostat
Kopp, G.A., Farquhar, S., Morrison, M.J., 2012. Aerodynamic mechanisms for wind loads with rim drives. Solar Energy 92, 230–240.
on tilted, roof-mounted, solar arrays. Journal of Wind Engineering and Industrial Pfahl, A., Randt, M., Meier, F., Zaschke, M., Geurts, C., Buselmeier, M., 2015. A holistic
Aerodynamics 111, 40–52. approach for low cost heliostat fields. Energy Procedia 69, 178–187.
Kozmar, H., 2012. Physical modeling of complex airflows developing above rural Pfahl, A., Uhlemann, P., 2011b. Wind loads on heliostats and photovoltaic trackers at
terrains. Environmental Fluid Mechanics 12 (3), 209–225. various Reynolds numbers. Journal of Wind Engineering and Industrial
Leitch, C.J., Ginger, J., Holmes, J., 2016. Wind loads on solar panels mounted parallel to Aerodynamics 99, 964–968.
pitched roofs, and acting on the underlying roof. Wind and Structures 22 (3), Price, H., Mehos, M.S., Cable, R., Kearney, D., Kelly, B., Kolb, G., Morse, F., 2020. CSP
307–328. plant construction, start-up, and O&M best practices study. In: Proc., AIP Conference
Li, Q.S., Zhi, L., Hu, F., 2010. Boundary layer wind structure from observations of a 325 Proceedings.
m tower. Journal of Wind Engineering and Industrial Aerodynamics 98, 818–832. Rasmussen, J.T., Hejlesen, M.M., Larsen, A., Walther, J.H., 2010. Discrete vortex method
Mehos, M., Turchi, C., Vidal, J., Wagner, M., Ma, Z., Ho, C., Kolb, W., Andraka, C., simulations of the aerodynamic admittance in bridge aerodynamics. Journal of Wind
Kruizenga, A., 2017. Concentrating solar power Gen3 demonstration roadmap. Engineering and Industrial Aerodynamics 98 (12), 754–766.
National Renewable Energy Laboratory (NREL, Golden, Colorado. Simiu, E., Scanlan, R.H., 1996. Wind effects on structures: fundamentals and applications
Mendis, P., Ngo, T., Haritos, N., Hira, A., Samali, B., Cheung, J., 2007. Wind loading on to design. John Wiley & Sons.
tall buildings. EJSE Special Issue: Loading on Structures 3, 41–54. Sment, J., Ho, C., 2014. Wind patterns over a heliostat field. Energy Procedia 49,
Menicucci, A.R., Ho, C.K. and Griffith, D.T. (2012) High Performance Computing for 229–238.
Static and Dynamic Analyses of Heliostats for Concentrating Solar Power. In: Proc., Stull, R.B., 1988. An introduction to boundary layer meteorology. Kluwer Academic,
Proceedings of the World Renewable Energy Forum (WREF 2012), Denver, Dordrecht, Netherlands.
Colarodo. Stull, R.B., 2005. The Atmospheric Boundary Layer. University of British Columbia,
Murphy, L.M., 1980. “Wind loading on tracking and field-mounted solar collectors”, Vancouver, Canada.
SERI-TP-632-958. Solar Energy Research Institute, Golden, USA. Sun, H., Gong, B., Yao, Q., 2014. A review of wind loads on heliostats and trough
Nakamura, Y., 1993. Bluff-body aerodynamics and turbulence. J Wind Eng Ind Aerodyn collectors. Renewable and Sustainable Energy Reviews 32, 206–221.
49 (1), 65–78. Téllez, F., Burisch, M., Villasente, Sánchez, M., Sansom, C., Kirby, P., Turner, P., Caliot,
National Climatic Data Center (2020), Climate data online, http://www.ncdc.noaa. C., Ferriere, A., Bonanos, C.A., Papanicolas, C., Montenon, A., Monterreal, R. and
gov/cdo-web/datasets, Last Accessed 1 April 2020. Fernández, J. (2014), “State of the Art in Heliostats and Definition of Specifications”,
Noone, C.J., Torrilhon, M., Mitsos, A., 2012. Heliostat field optimization: A new 609837; STAGE-STE Project, Madrid.
computationally efficient model and biomimetic layout. Solar Energy 86 (2), Tieleman, H.W., 2003. Wind tunnel simulation of wind loading on low-rise structures: A
792–803. review. Journal of Wind Engineering and Industrial Aerodynamics 91 (12),
Peterka, J.A., Bienkiewicz, B., Hosoya, N., Cermak, J.E., 1987a. Heliostat mean wind 1627–1649.
load reduction. Energy 12 (3–4), 261–267. Turchi, C.S., Boyd, M., Kesseli, D., Kurup, P., Mehos, M.S., Neises, T.W., Sharan, P.,
Peterka, J.A., Derickson, R.G., 1992. “Wind load design methods for ground-based Wagner, M.J., Wendelin, T., 2019. “CSP Systems Analysis-Final Project Report”,
heliostats and parabolic dish collectors”, SAND92-7009. Sandia National National Renewable Energy Laboratory (NREL). Golden, Colorado.
Laboratories, Albuquerque, New Mexico. Vásquez-Arango, J.F., Buck, R., Pitz-Paal, R., 2015. Dynamic Properties of a Heliostat
Peterka, J.A., Hosoya, N., Bienkiewicz, B., Cermak, J.E., 1986. “Wind load reduction for Structure Determined by Numerical and Experimental Modal Analysis. Journal of
heliostats”, SERI/STR-253-2859. Colorado State University, Fort Collins, USA. Solar Energy Engineering 137 (5), 051001.
Peterka, J.A., Hosoya, N., Dodge, S., Cochran, L., Cermak, J.E., 1998. Area-average peak Vasquez Arango, J.F., Pitz-Paal, R. and Breuer, M. (2017), Dynamic wind loads on
pressures in a gable roof vortex region. Journal of Wind Engineering and Industrial heliostats, PhD Thesis, Lehrstuhl für Solartechnik (DLR).
Aerodynamics 77–78, 205–215. von Kármán, T., 1948. Progress in the Statistical Theory of Turbulence. Proceedings of
Peterka, J.A., Tan, L., Bienkiewcz, B. and Cermak, J.E. (1987), Mean and peak wind load the National Academy of Sciences 34 (11), 530–539.
reduction on heliostats, Technical Report for Colorado State University. Wolmarans, J.R., Craig, K., 2019. One-way fluid-structure interaction of a medium-sized
Peterka, J.A., Tan, Z., Bienkiewicz, B., Cermak, J., 1988. “Wind loads on heliostats and heliostat using scale-resolving CFD simulation. Solar Energy 191, 84–99.
parabolic dish collectors: Final subcontractor report”, SERI/STR-253-3431. Solar Wu, Z., Gong, B., Wang, Z., Li, Z., Zang, C., 2010. An experimental and numerical study
Energy Research Institute, Golden, Colorado. of the gap effect on wind load on heliostat. Renewable Energy 35 (4), 797–806.
Peterka, J.A., Tan, Z., Cermak, J.E., Bienkiewicz, B., 1989. Mean and peak wind loads on Xiong, Q., Li, Z., Luo, H., Zhao, Z., 2019. Wind tunnel test study on wind load coefficients
heliostats. Journal of Solar Energy Engineering 111 (2), 158–164. variation law of heliostat based on uniform design method. Solar Energy 184,
Pfahl, A., 2014a. Survey of heliostat concepts for cost reduction. Journal of Solar Energy 209–229.
Engineering 136 (1). Xiong, Q., Li, Z., Luo, H., Zhao, Z., Jiang, A., 2021. Study of probability characteristics
Pfahl, A., 2018. Wind loads on heliostats and photovoltaic trackers. PhD thesis. and peak value of heliostat support column base shear. Renewable Energy 168,
Technische Universiteit Eindhoven. 1058–1072.
Pfahl, A., Brucks, A., Holze, C., 2014b. Wind load reduction for light-weight heliostats. Xu, Y.L., 2013. Wind Effects on Cable-Supported Bridges. John Wiley & Sons, Singapore.
Energy Procedia 49, 193–200. Yu, J.S., Emes, M.J., Ghanadi, F., Arjomandi, M., Kelso, R.M., 2019. Experimental
Pfahl, A., Buselmeier, M., Zaschke, M., 2011a. Wind loads on heliostats and photovoltaic investigation of peak wind loads on tandem operating heliostats within an
trackers of various aspect ratios. Solar Energy 85, 2185–2201. atmospheric boundary layer. Solar Energy 183, 248–259.
82

Emes-Rev Heliostat Wind Load (2021)

Uploaded by

Copyright:

Available Formats

Emes-Rev Heliostat Wind Load (2021)

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Emes-Rev Heliostat Wind Load (2021)

Uploaded by

Copyright:

Available Formats

Solar Energy 225 (2021) 60–82

Contents lists available at ScienceDirect

A review of static and dynamic heliostat wind loads

$140/m2 by National Renewable Energy Laboratory (NREL) (Turchi

Nomenclature lpx Distance to the centre of pressure from the heliostat

Peterka et al. (1989) 4 2.8 4.35 0.6 0.9 0.6 1 0.2 18

3.2. Facet gap

Conventional heliostats are designed with small gaps between the

5. Aerodynamics of a heliostat field

Heliostat fields are arranged in rows in a radial (Fig. 22a) or polar

You might also like