Download full-text PDF

A GUI for urban wind flow CFD analysis of small scale wind applications

Conference Paper (PDF Available)  · September 2015with294 Reads
DOI: 10.13140/RG.2.1.2463.2802
Conference: Conference: Cyberworlds IEEE, At Visby
Bahri Uzunoglu at Uppsala University
  • 16.7
  • Uppsala University
A GUI for urban wind flow CFD analysis of small
scale wind applications
Anders Goude, Bahri Uzuno
˘
glu, Gabriele Giovannini
Department of Engineering Sciences, Division of Electricity
Centre for Renewable Electric Energy Conversion
Uppsala University, The
˚
Angstr
¨
om Laboratory
Box 534, 751 21 Uppsala, Sweden,
e-mail: anders.goude@angstrom.uu.se
e-mail: bahri.uzunoglu@angstrom.uu.se
Javier Magdalena and Antonio Fern
´
andez
SOLUTE Ingenieros
Av. Cerro del guila 3 Edificio 2, Of 3 Izd
28703 San Sebastin de los Reyes, Madrid
e-mail:javier.magdalena@solute.es
e-mail:antonio.fernandez@solute.es
Abstract—In order to have better insight into the physics
of the urban wind turbines, a graphical user interface (GUI)
that employs OpenFOAM flow solver has been developed for
industrial applications by Uppsala University with Spanish en-
gineering company SOLUTE via EU framework as part of the
WINDUR framework 7 project. Urban wind resource assessment
for small scale wind applications present several challenges and
complexities for that are different from large-scale wind power
generation. Urban boundary layer relevant in this regime of
flows have different horizontal profiles impacted by the buildings,
low speed wind regimes, separation and different turbulence
characteristics. This software addresses the project setup and
scientific visualization of the results for right investment decision
needs. Preliminary numerical results will be presented for a
test site in Huesca, Spain where a measurement campaign is
undertaken to validate the Computational Fluid Dynamics (CFD)
results.
KeywordsUrban flow, Small scale wind, Graphical User
Interface (GUI), Computational Fluid Dynamics (CFD), Boundary
Layer, Buildings, OpenFOAM, Reynolds Averaged Navier Stokes
Equation (RANS)
I. INTRODUCTION
The small scale wind mainly corresponds to turbines
installed in rural and isolated areas. As 80 % of European
population lives in cities and the EU Directive 2010/31/EU on
Energy Performance of Buildings requires that Member States
shall ensure that by 31 December 2020 all new buildings are
nearly zero-energy buildings. This is a commercial opportunity
in terms of small scale wind applications while also provides
a motivation to investigate technical challenges related to the
peculiarities of Urban wind regime.
Urbanization brings with itself higher energy demand. Ver-
tically structured generation requires transmission from gener-
ation to urban areas where there is demand. In comparison
to this market structure, one of the advantages of distributed
generation from renewable energy sources such as wind is
local production and consumption. Local consumption of local
wind resources in urban areas can avoid losses, congestion and
infrastructure investment from transmission and distribution
and is preferable on other central power generation approaches
when this approach feasible. If suitable wind regimes can be
found in urban areas, it is possible to have right investment
decisions for small scale wind.
Project design for any wind project should not be too labor
intensive while the cost margins for the project management
employing Computational Fluid Dynamics (CFD) softwares
will require specialized expertise for problem setup and pre-
processing and post-processing tools. This problem which
has been well recognized for large scale wind industry, was
addressed by several commercially developed graphical user
interface (GUI) applications for large scale wind farms. These
GUI applications made use of generalized CFD solvers by
simplifying the labor and expertise required processes. The
usage of tools with similar GUI for small scale wind applica-
tions were limited. A GUI that can process roughness maps,
height contours and that can implement buildings for the urban
boundary layers without intervention of the expert user with
default parameter setup will address some of the aspects of
this gap. The graphical user interface (GUI) approach in this
study provides an user friendly tool based on the open source
CFD solver OpenFOAM, which significantly simplifies the
process of creating the simulation geometry and setting up the
simulation parameters, making the process much faster and
available to non-experts in CFD.
Urban wind regimes are relevant to urban boundary layer
research which investigates wind profiles and thermally driven
secondary circulations over cities [1] [2] [3] [4]. Some of the
characteristics of urban boundary layer in horizontal direction
involves large roughness elements with stored heat that has
reduced moisture in possible sealed areas that leads to higher
level of turbulence that creates a larger boundary layer [2]
[3] [4]. In this study heat fluxes and heat islands are ignored
however they can be a significant part of flow characteristics in
large city formations where as mentioned above a secondary
circulation can be formed with winds toward city center near
the ground and uprising form there on. Independent urban
block structures at microscale are investigated. Microscale
flow regime level will be more appropriate classification for
the current study. Larger scale simulations where appropriate
boundary conditions for a section block of a larger urban
site block from a complete urban boundary layer will not be
investigated in this study.
Some of the characteristics of urban boundary layer in
vertical direction creates an internal boundary layer with
modified horizontal profile. In this context, the urban boundary
layer can be classified into four main boundary layers in
2015 International Conference on Cyberworlds
978-1-4673-9403-1/15 $31.00 © 2015 IEEE
DOI 10.1109/CW.2015.71
193
2015 International Conference on Cyberworlds
978-1-4673-9403-1/15 $31.00 © 2015 IEEE
DOI 10.1109/CW.2015.71
193
vertical direction [1] [4] [5]. The highest boundary layer is
the geostrophic wind direction that is named Ekman layer.
Below this layer is the third layer what is named Prandtl
layer or surface layer in homogeneous flows that will be
corresponding to constant flux layer (CFL) or inertial sublayer
(IS) in urban flows. For convectively dominated flows, these
two upper layers become one mixing layer. The second layer
and first layer respectively are wake layer that can go up-to five
times the building heights and the bottom urban canopy layer
(UCL) which can go up to average height of the buildings [1]
[4] [5]. The urban-roughness layer (URL) is the term used to
define these third and fourth layers jointly.
The urban-roughness layer profile which is the most rele-
vant profile for wind installations can be approximated by pro-
file laws using logarithmic wind profile for neutral, stable and
unstable stratification. However general formulations for flat
terrain have been noted to be not accurate for complex flows
like urban flows as a result modifications were introduced.
For the Prandtl layer which is the third layer, the following
approximation can be used.
u =
u
κ
ln
z d
z
o
(1)
where u
is characteristic velocity and κ is von karman
constant which is 0.41 and z
0
is roughness length, d is
displacement height which gives the vertical displacement of
the flow for buildings [5] [6] [7] [8] [9] . Herein characteristic
velocity u
can be derived from turbulence fluctuations of
north east and west south components correlated with verti-
cal components or directly by inversion of geostrophic drag
law. The vertical height is denoted by z. Finally u will be
temporarily denote the wind speed here till it is redefined to
be used as west to east wind component in the next section.
This relation has been modified for the second layer which
is wake layer with new parameter α to reflect the impact of
buildings as below
u = α
u
κ
ln
z d
z
o
. (2)
This was further modified for the first layer as an exponential
rule for the bottom urban canopy layer (UCL)
u = u exp
a
z
h
1

(3)
where a is a constant which is dependent on building mor-
phology, z is the vertical height and h is the height of
the building [6] [7] [9]. All of these approximations have
large margins of error in the context of complex flows of
urban flows [6] [7] [9]. As a result, the limitation of this
algebraic relations can be overcomed by using partial differen-
tial equations in this context Computational Fluid Dynamics
(CFD) that will be discussed in the next section. Since the
CFD is labor intensive work, we have developed a graphical
user interface (GUI) for industrial applications that employs
an industry based CFD solver namely OpenFOAM solver
that has been developed by Uppsala University with Spanish
engineering company SOLUTE via EU framework as part
of the WINDUR framework 7 project. For CFD solutions,
steady state equations of fluid mechanics will be used so all
the solutions are time averaged and as discussed in the next
section Reynolds Averaged Navier Stokes (RANS) equations
are going to be used in comparison to Large Eddy Simulation
(LES) which is computationally demanding [10] [11] [12].
RANS use time-averaged NS equations, and LES utilize space-
averaged (filtered) equation [13]. Both approaches produce
turbulence fluctuation terms that need to be modeled. The
main advantage of LES over the computationally more efficient
RANS approach is the increased level of detail it can deliver.
The ability to predict instantaneous flow characteristics and tur-
bulent flow structures is particularly important in simulations
of transient phenomena however the computational demand
might not necessarily practical for financial decision analysis
and business cycle of small scale wind installations while
RANS addresses the problem with much lower computational
resources.
II. T
HEORY AND DISCRETIZATION
For incompressible flows, the general form of the Navier-
Stokes (NS) is given by
∂x
j
[ρu
j
]=0 (4)
∂t
(ρu
i
)+
∂x
j
[ρu
i
u
j
+
ij
τ
ij
]=0, (5)
i =1, 2, 3 (6)
where ρ defines density, p defines hydrostatic pressure, u
j
defines velocity, u
1
= u, u
2
= v, u
3
= w and x
j
spatial
coordinates and the stress tensor τ
ji
depends linearly on the
rate-of-strain tensor S
ij
, μ dynamic viscosity.
τ
ij
=2μS
ij
, (7)
and
S
ij
=
1
2
∂u
i
∂x
j
+
∂u
j
∂x
i
. (8)
In an appropriate time T interval if the velocity u is time
averaged:
u(x)=
1
T
t
0
+T
t
0
u(x, t)dt (9)
The averaged term and the fluctuation term of the Reynolds
decomposition of velocity u =
u + u
that satisfy the
properties
u
=0, and u = u can simplify convective
term in the momentum Equation 12 that can be simplified
as
u
i
u
j
= u
i
u
j
+ u
i
u
j
= u
i
u
j
τ
R
ij
where the fluctuation
term is Reynolds stress tensor τ
R
ij
= u
i
u
j
. After Reynolds
194194
decomposition of velocity, the time-averaged Navier-Stokes
(NS) equation is,
∂x
j
[ρu
j
]=0(10)
∂t
(ρ
u
i
)+
∂x
j
ρ
u
i
u
j
+
ij
τ
ij
+ τ
R
ij
=0, (11)
i =1, 2, 3. (12)
which is also titled Reynolds Averaged Navier-Stokes equation
(RANS) [14].
Turbulence modeling based on k and
The RANS model needs closure. In the derivation of
the time average on momentum equations, six additional
unknowns are introduced via Reynolds stress. To close this,
several approximations exist, one of them is k and model.
These equations carry variable k which is the turbulent kinetic
energy variable which can be derived from RANS equations
by multiplying x, y, z components with u, v, w components
and same will be trued for the other components which will
give the turbulent kinetic energy equations eventually. If the
same derivations is done for the mean kinetic energy K, one
will observe that the k and K equations will also carry a
turbulence production term which is negative in the mean K
and positive in k equations which shows mean kinetic energy
is destroyed and converted turbulence kinetic energy. Also the
dissipation term of k
j
in turbulence kinetic energy equations
gives negative contribution which is caused by the work done
against small eddies by viscous stresses which destroys the
turbulence kinetic energy. If this term is normalized by mass,
it is denoted as . The term can be expanded to another
set of transport equations based on empirical approaches and
this will be define the two equations model of k and . These
equations will be coupled to to RANS equations via turbulence
viscosity term in the Reynolds stress tensor via k and terms
that defined the mixing length with an empirical constant [14].
III. S
IMULATION TOOL
In the present work, a simulation tool has been developed
to allow simulations in urban environments based on the
OpenFOAM solver [15] [16]. The tool is designed to allow
the user to input the geometry data including buildings and
surface roughness. This tool uses the built in mesh generation
tool snappyHexMesh from OpenFOAM, and the tool will both
suggest basic parameters to use for the mesh generation, and
also allow the user to manually change these for better control
of the mesh generation. The tool also handles all boundary
conditions, such as terrain roughness and the atmospheric
boundary layer. It is also capable of running the simulations
and plotting the velocity profiles at chosen evaluation points.
A. Discretization scheme
To describe theses set of equations non-linear transport
flow equations in a computer, discretization needs to be
implemented. For discretization scheme, finite volume method
that divides the domain into a finite set of volumes by
defining a volume integral over the continuous equations can
be implemented. One can create a discrete system of equations
that can be solved by iterative solutions in a computer by linear
algebra [15] [16] [14]. Some of the discretization methods that
are used for divergence, gradient and Laplacian operators are
summarized in a Table I .
For gradient terms, divergence terms and laplacian terms,
the Gauss keyword in Table I denotes the standard finite
volume discretisation of Gaussian integration that interpolates
values from cell centres to face centres.
In gradient terms, linear interpolation is used while this can
be limited to face or cell. Cell limited version is chosen. Cell
limiting denotes the limited gradient along a line connecting
adjacent cell centers while face limiting denotes the limited
gradient on the face itself. Gradient limiters otherwise known
as slope limiters are implemented to prevent spurious oscilla-
tions in the solution flow field by enforcing the monotonicity
principle by prohibiting the linearly reconstructed field variable
on the cell faces to exceed the maximum or minimum values
of the neighboring cells. These limiters can be classified as
non-differentiable such as standard limiter, multidimensional
limiter, differentiable limiter. Gradient limiters can be cate-
gorized into two general groups: differentiable limiters and
non-differentiable limiters that contains standard limiter and
multidimensional limiter while minimum and maximum types
of functions for limiting the solution variables are used in this
form for these non-differentiable limiters [15] [16] [14].
For divergence terms, upwind discretization scheme is used
that makes use of flow directional information in interpolation.
Before convergence is reached, divergence of velocity field for
incompressible is not always zero. To include this bound term
within a numerical solution poses boundedness of the solution
variable and promotes better convergence so the bounded term
in Table I implement this [15] [16] [14].
TABLE I: Numerical discretization schemes
Category of mathematical terms Variables Schemes
interpolation schemes all linear
Gradient k cell limited Gauss linear 1
cell limited Gauss linear 1
The rest of variables Gauss linear
Divergence u
j
bounded Gauss linear upwindV
k bounded Gauss linear upwind
bounded Gauss linear upwind
the rest of variables bounded Gauss linear
Laplacian All Gauss linear limited
B. Solution and algorithm control
The discretized equations generated above can be used
linear-solver that solves a set of linear equations. Two of the
solvers are used here. Two solvers are smooth solver and
generalised geometric-algebraic multi-grid (GAMG). Smooth
solver employs well-known linear solvers such as Gauss-
Seidel, Diagonal incomplete-Cholesky (symmetric) and Di-
agonal incomplete-Cholesky with Gauss-Seidel (symmetric)
[15] [16] [14]. The generalised method of geometric-algebraic
multi-grid (GAMG) employs the principle cascade of quick
solution on a mesh starting with a small number of cells and
projecting these solutions onto a finer mesh. The concept uses
an initial guess to obtain an accurate solution on the fine
mesh [15] [16] [14]. This decreases the computational effort by
195195
(a) Graphical user interface for the urban flow software -
building input
(b) Manual editor for mesh refinements
(c) Post processing interface
Fig. 1: Graphical user interface for the urban flow software
solving first on coarser meshes while bypassing the additional
costs of mesh refinement and projection of field data. This
concept can be used as a preconditioner or as an independent
solver. The method used for each field variable is summarized
in Table II. All of these parameters were implemented in
the graphical user interface also giving the user flexibility to
further advances modification of the solver settings.
TABLE II: Solver schemes
Variables Schemes
p GAMG
u
j
smoothSolver
k smoothSolver
smoothSolver
IV. GRAPHICAL USER INTERFACE
Simulation tools, such as OpenFOAM, have the capability
to perform the required simulations, but to properly set up a
simulation in from scratch requires a lot of work and good
knowledge about the simulation software. To make it possible
for a wider audience to use perform these simulations, it is
necessary to provide a user interface which have already been
configured for the chosen set of simulations. This is the main
purpose of the user interface presented in the current work.
This section will therefore describe the GUI and how it can
assist with these difficulties to make it easy to perform urban
simulations.
One initial step of setting up urban simulations is to create
the input geometry, which is the step where a GUI is most
useful, as manual generation of all input geometry files is not
feasible for the normal user. For urban simulations, both the
orography and the building geometries have to be taken into
account and given in the stereolithography (stl) file format,
which is used by OpenFOAM. These stl files contain a set of
triangles that describe the surfaces of the geometry. As this
is not the typical file format used to describe orography, the
GUI has to be able to generate these files from more common
orography data formats. It was chosen to use the traditional file
formats .map and .xyz for the import of orography data. In this
way, data available for other software such as industry standard
Wind resource and energy yield assessment software (WAsP)
.map format can also be used in the current model. These
formats do instead contain height contour lines (or possibly
only point data for the xyz format). From the points specifying
these contour lines, Delaunay triangulation
1
was used to create
a set of triangles, which describe the surface of the terrain and
can be saved in the geometry format used by OpenFOAM. In
the simulations presented in the current work, map files with
countour interval of 5 m were used for the site.
The second part of the geometry is to add the buildings.
For simple building shapes, the GUI can directly generate the
required geometry files, but to support advanced shapes, it is
suitable to allow import of CAD drawings as well. As it is
common that CAD software can export drawings in the stl
format, and this format is hence used as a possible input format
in the GUI. One key aspect of setting up the urban simulations
is to add the buildings to the correct geometric positions. A
CAD drawing will normally only contain the shape of the
building, but it is up to the GUI to modify this geometry data to
place the buildings at their correct positions, and also rotate the
buildings according to their real positions. The GUI supports
both moving buildings graphically by drag and drop, and by
manually giving building coordinates. One further step that is
required from the GUI is to calculate the ground level at each
building position and place the buildings at the correct height.
A snapshot of the GUI for building input is given in Figure 1a.
As the geometry files used for OpenFOAM are in the stl
format, which uses single precision, the final step of the gener-
ation of geometry files for the GUI is to shift all geometries to
make the all simulation domains at zero, instead of using map
coordinates. This is to limit the effects of truncation errors,
1
http://www.codeproject.com/Articles/587629/A-Delaunay-triangulation-
function-in-C
196196
which otherwise could remove all details from the building
data.
Mesh generation in the GUI and solver is managed by
OpenFOAM structured and unstructured mesh generators. The
basic mesh generated by structured mesh generator is further
refined by unstructured mesh generator for complex geometries
such as buildings. The GUI will provide a suggested set of
default refinement parameters, to simplify for the user, but
for the advanced user, the mesh refinement parameters can be
manually edited (Figure 1b ). Similar to the building input, the
GUI will also allow the user to select given positions, where
the velocity should be calculated. This information will also
be used to automatically refine the mesh around these points
of interest.
The next step of setting up the simulations is to give
other simulation parameters, such as surface roughness and
velocity profile. Surface roughness can be included, either as
a constant value, or by providing roughness data in either the
map format, or as structured xyz files. The velocity is given as
a logarithmic profile, where the velocity at a reference height
can be given, which together with the roughness data, will
provide the velocity profile.
The next step of the for the GUI is to take care of
all steps in the OpenFOAM execution, to allow the user to
run simulations by pushing a single button. The GUI will
first call the mesh generation parts of OpenFOAM. When
the mesh is generated, the boundary conditions has to be
updated with respect to the new mesh. First, the ground level
for the logarithmic velocity profile has to be updated for
each individual position, to make the logarithmic profile start
from the correct position throughout the domain. If varying
roughness data is provided, this also has to be updated to give
each ground position the correct roughness value. Then the
GUI should call OpenFOAM to run the simulations. Finally,
the GUI should extract the simulation results at the given points
of interest. All these steps are handled automatically without
user influence. The GUI will provide support for running both
single simulations, and sector sweeps, where the wind direction
is changed between the simulations to generate data for wind-
rose plots.
The final part of the GUI is to provide a simple post-
processing interface where the simulation results for the points
of interest can be seen. This is illustrated in Figure 1c,
where the turbulent kinetic energy k is plotted for the upwind
building.
V. N
UMERICAL RESULTS AND SIMULATION PARAMETERS
Huesca site in Spain that is displayed in in Figure 2a shows
the characteristics of the orography and buildings involved
for numerical development. The measurement campaign at
3.5m based on the design considerations of wind turbine
that is being developed is in progress on this site. There are
two measurement stations on this site for validation purposes
and our vertical profiles are going to be profiled form these
positions for a fixed wind speed.
The orography was imported from map files for the site
with contour interval of 5 m. The buildings were drawn
using traditional CAD software and imported into the GUI.
(a) EDIF 3 and EDIF 4 in Huesca that are being simu-
lated.
(b) The buildings in Huesca that have met masts namely
EDIF 3 and EDIF 4 in urban flow simulator.
Fig. 2: The simulation set-up and visualization in urban flow
software a) EDIF 3 and EDIF 4 buildings that are being
simulated in Huesca. b) Graphical user interface for the urban
flow software. c) Velocity magnitude field view from Paraview
for the EDIF 3 and EDIF 4 buildings that have met masts.
To avoid excessive computational effort, only the important
buildings were drawn in detail, and the remaining buildings
were simplified to avoid having too many mesh points to
describe these buildings. All buildings were created with 10
m extra height, and were then shifted 10 m into the ground
to account for possible slopes in the terrain, which otherwise
could give empty space below parts of the building.
The simulation domain was chosen to extend up and
downstream 1000 m and crosstream 800 m from each of the
two positions where met masts are installed. This gives a
domain size of 2126 x 1602 m. The simulation height is chosen
to extend 1000 m above the highest point of the terrain. By
using the meshing model of OpenFOAM, the domain is first
divided into uniform base mesh, in this case 28 x 21 x 14
elements. Then, for the important regions, the mesh is refined.
In the current method, the mesh was refined close to the terrain
and close to the buildings. Mesh refinements up to 8 times (i.e.
each block is divided into 8 smaller boxes 8 times) were used,
although the majority of the 6.9 million mesh cells were at
refinement level 6.
197197
(a) The mesh used in the simulations.
(b) The wind speed magnitudes.
Fig. 3: Simulation set-up a) the mesh used in the simulations
b) the wind speed magnitudes.
The boundary conditions used was to use the logarithmic
boundary layer for the inlet velocity since the flow simulations
were always started from flat terrain where displacement height
is d =0. Here, the asymptotic flow velocity was set to 10 m/s
at 1000 m height from EDIF 4 to EDIF3 West to East direction.
The surface roughness was estimated to be z
0
=0.05 which
corresponds to a ground with bushes. This was chosen from
studying satellite images of the surrounding terrain. Rough
wall boundary conditions were also used for the terrain to
include the effects of surface roughness here as well. The
preliminary velocity profiles on the roof of the buildings are
presented in Figure 4a and 4b which shows the significant
impact of one building on the downstream building profiles
for meteorological mast locations which are marked with red
in the figures where vertical profiles are analyzed.
VI. C
ONCLUSIONS
To meet the needs of the wind industry, a novel GUI with
several numerical solver, discretization and mesh generation
options via OpenFOAM open source for a user friendly envi-
ronment is implemented. The approach combines advantages
of open source solver approach with a user friendly interface.
The preliminary results are presented for a site in Spain namely
Huesca where a measurement campaign is in progress to
validate the numerical results. The final work will include
0 1 2 3 4 5 6
0
2
4
6
8
10
Horizontal wind speed
Height from the roof of the building
Last building EDIF 3
First building EDIF 4
(a) The velocity profile for EDIF 3 and EDIF 4 when both buildings
are configured in simulations.
0 2 4 6 8
0
5
10
15
20
Horizontal wind speed
Height from the roof of the building
Both buildings EDIF 3 AND EDIF 4
One building EDIF 3
(b) The EDIF 3 velocity profile with EDIF 3 with and without EDIF 4.
Fig. 4: Velocity profile comparisons for meteorological mast
locations which are marked with red in the figures where
vertical profiles are analyzed a) the velocity profile with two
buildings EDIF 3 and EDIF 4 same as above b) the velocity
profile for EDIF 3 simulation only without EDIF 4.
the validation exercise with the measurement campaign. The
software addresses microscale urban flow regimes in urban
roughness layer where complex geometries of buildings create
several physical phenomena where simple profiles might not
be sufficient.
R
EFERENCES
[1] E. Plate, “Urban climates and urban climate modelling: An
introduction, in Wind Climate in Cities, ser. NATO ASI
Series, J. Cermak, A. Davenport, E. Plate, and D. Viegas, Eds.
Springer Netherlands, 1995, vol. 277, pp. 23–39. [Online]. Available:
http://dx.doi.org/10.1007/978-94-017-3686-2
2
[2] J. Hidalgo, V. Masson, A. Baklanov, G. Pigeon, and L. Gimeno,
Advances in urban climate modeling, Annals of the New York
Academy of Sciences, vol. 1146, no. 1, pp. 354–374, 2008. [Online].
Available: http://dx.doi.org/10.1196/annals.1446.015
198198
[3] M. KANDA, “Progress in urban meteorology :a review, Journal of the
Meteorological Society of Japan. Ser. II, vol. 85B, pp. 363–383, 2007.
[4] M. Roth, “Review of atmospheric turbulence over cities,
Quarterly Journal of the Royal Meteorological Society, vol.
126, no. 564, pp. 941–990, 2000. [Online]. Available:
http://dx.doi.org/10.1002/qj.49712656409
[5] S. Emeis, Wind Energy Meteorology: Atmospheric Physics
for Wind Power Generation, ser. Green Energy and
Technology. Springer Berlin Heidelberg, 2012. [Online]. Available:
https://books.google.se/books?id=YdM9FQhPfdYC
[6] S. L. Walker, “Building mounted wind turbines and their
suitability for the urban scalea review of methods of
estimating urban wind resource, Energy and Buildings,
vol. 43, no. 8, pp. 1852 1862, 2011. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0378778811001101
[7] M. A. Heath, J. D. Walshe, and S. J. Watson, “Estimating the
potential yield of small building-mounted wind turbines, Wind
Energy, vol. 10, no. 3, pp. 271–287, 2007. [Online]. Available:
http://dx.doi.org/10.1002/we.222
[8] D. R. Drew, J. F. Barlow, and S. E. Lane, “Observations of
wind speed profiles over greater london, uk, using a doppler
lidar, Journal of Wind Engineering and Industrial Aerodynamics,
vol. 121, no. 0, pp. 98 105, 2013. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0167610513001645
[9] L. Jian-Zhong, L. Hui-Jun, and Z. Kai, “New expressions for the
surface roughness length and displacement height in the atmospheric
boundary layer, Chinese Physics, vol. 16, no. 7, p. 2033, 2007.
[Online]. Available: http://stacks.iop.org/1009-1963/16/i=7/a=038
[10] A. Zhang, C. Gao, and L. Zhang, “Numerical simulation
of the wind field around different building arrangements,
Journal of Wind Engineering and Industrial Aerodynamics,
vol. 93, no. 12, pp. 891 904, 2005. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0167610505000966
[11] M. Tutar and G. Oguz, “Computational modeling of wind flow around
a group of buildings, International Journal of Computational Fluid
Dynamics, vol. 18, no. 8, pp. 651–670, Dec. 2004. [Online]. Available:
http://dx.doi.org/10.1080/10618560412331291160
[12] J.-J. Kim and J.-J. Baik, A numerical study of the effects of
ambient wind direction on flow and dispersion in urban street canyons
using the {RNG} k turbulence model, Atmospheric Environment,
vol. 38, no. 19, pp. 3039 3048, 2004. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S1352231004001773
[13] M. Tutar and G. Oguz, “Large eddy simulation of wind flow
around parallel buildings with varying configurations, Fluid Dynamics
Research, vol. 31, no. 56, pp. 289 315, 2002. [Online]. Available:
http://www.sciencedirect.com/science/article/pii/S0169598302001272
[14] H. Versteeg and W. Malalasekera, An Introduction to
Computational Fluid Dynamics: The Finite Volume Method.
Pearson Education Limited, 2007. [Online]. Available:
https://books.google.se/books?id=RvBZ-UMpGzIC
[15] H. G. Weller, G. Tabor, H. Jasak, and C. Fureby, A tensorial
approach to computational continuum mechanics using object-oriented
techniques, Comput. Phys., vol. 12, no. 6, pp. 620–631, Nov. 1998.
[Online]. Available: http://dx.doi.org/10.1063/1.168744
[16] OpenFOAM Documentation. [Online]. Available:
http://cfd.direct/openfoam/user-guide/
199199
Conference Paper
May 2017
    Vortex induced vibrations (VIV) for energy extraction have been revisited in last years by both marine power and wind power communities. Even though vortex induced vibrations have been focus of research for many years, energy extraction from vortex induced vibrations is relevantly new field which needs more detailed investigation and modeling. To this end, there has been recent experimental... [Show full abstract]
    Article
      Based on FLUENT soft platform with Reynolds Stress Model (RSM) and Large Eddy Simulation (LES), the wind pressure distributions on structures, such as, dome, cylinder, catenoid, saddle and elliptic hyperbolic paraboloid were obtained in steady and unsteady numerical simulation. The auto-regressive model was adopted to generate the time history of pulse wind speed. The time series satisfied the... [Show full abstract]
      Article
      May 2007 · Wind Energy
        The wind profile in the urban boundary layer is described as following a logarithmic curve above the mean building height and an exponential curve below it. By considering the urban landscape to be an array of cubes, a method is described for calculating the surface roughness length and displacement height of this profile. Firstly, a computational fluid dynamics (CFD) model employing a k-ε... [Show full abstract]
        Article
          Cities increasingly face the necessity to choose among possible policy measures to counter urban heat and then to justify them. The contingent valuation method (CVM) is a frequently applied approach to estimate the monetary value of environmental policy measures. The authors scrutinized the appropriateness of the CVM to estimate the value of trees and drinking fountains as measures against... [Show full abstract]
          Discover more