Professional Documents
Culture Documents
Renewable Energy
journal homepage: www.elsevier.com/locate/renene
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 2 March 2015
Received in revised form
10 August 2015
Accepted 6 December 2015
Available online 24 December 2015
In the latest years the wind energy sector experienced an exponential growth all over the world. What
started as a deployment of onshore projects, soon moved to offshore and, more recently to the urban
environment within the context of smart cities and renewable micro-generation. However, urban wind
projects using micro turbines do not have enough prot margins to enable the setup of comprehensive
and expensive measurement campaigns, a standard procedure for the deployment of large wind parks.
To respond to the wind assessment needs of the future smart cities a new and simple methodology for
urban wind resource assessment was developed. This methodology is based on the construction of a
surface involving a built area in order to estimate the wind potential by treating it as very complex
orography. This is a straightforward methodology that allows estimating the sustainable urban wind
potential, being suitable to map the urban wind resource in large areas. The methodology was applied to
a case study and the results enabled the wind potential assessment of a large urban area being consistent
with experimental data obtained in the case study area, with maximum deviations of the order of 10%
(mean wind speed) and 20% (power density).
2015 Elsevier Ltd. All rights reserved.
Keywords:
Urban wind
Resource assessment
Micro-generation
CFD
1. Introduction
Urban wind energy has a large potential to be explored in the
context of smart cities, whether through the installation of small
wind turbines in the domestic sector (building rooftops and surrounding areas), or integrated in the building envelope providing
that they are designed with wind energy exploitation in mind [1].
The wind potential in urban areas is difcult to characterize due to
the high impact of obstacles and structures on the atmospheric
ow. Buildings often cause ow separation, wind speed reduction
and high turbulence on the top and around buildings. Also, in
economic terms, the high costs of wind measurements campaigns
are an important barrier to the development of this sub-sector of
wind energy. Other data sources may be used for the characterization of the wind ow in urban environment, such as databases
and national and regional wind potential atlas. These solutions are
usually based on the application of data from mesoscale models
(MM51 e Fifth Generation Mesoscale Model, WRF e Weather
599
"
!
#
v ruj ui
vp
v
vui vuj
0 0
m
rui uj fi 0 ;
vxj vxj
vxj vxi
vxj
vrui
0
vxi
(1)
(2)
0
where the term rui uj represents the Reynolds tension tensor, that
corresponds to the additional transference of momentum due to
the turbulence uctuations [19,20].
v
v nT vk
PK
Ui k
vxi
vxi sk vxi
(3)
v
v nT v
2
C1 PK C2
Ui
vxi
vxi s vxi
k
k
(4)
where Cm, sk, s, C1 and C2 are known constants, PK is the turbulent kinetic energy production given by Eq. (5), and nT is the kinematic viscosity.
The CFD-Complex model solves the RANS equations using the
nite element method, along with initial conditions, referred to as
the solution's rst estimates. This model may use two types of
solver e namely Segregated and Coupled. The Coupled solver
(Migal) [21] refers to the coupling technique speed e pressure and
to a linear solver that simultaneously updates the speed and
pressure leds across the domain. The linear solver performs only
the rst part of the speed-pressure iteration and is followed by,
Phoenics [22] model that completes the non-linear part of the
iteration. The CFD-Complex process involves the execution of
several modules which includes models that are executed
sequentially in order to obtain the results [18].
The CFD-Urban model also solves the RANS equations and uses
the nite volume method and a turbulence closure. The boundary
conditions and the domain are generated automatically and the
wind proles are obtained according to a theoretical reference
wind. The turbulent kinetic energy is constant and it is evaluated
according to the input roughness data from the simulation domain
[13]. Both models use wind frequency distributions (Weibull distributions and wind rose) as input data.
3. Methodology for the wind resource assessment in urban
environments
The methodology presented in this paper is based on the generation of a digital terrain model which includes the terrain and the
existing buildings as a whole, thus representing an urban digital
terrain model: U-DTM. The urban DTM can be treated as a very
complex terrain and be used as input for a standard wind resource
assessment model (e.g. Wasp, WindSim). This methodology
strongly reduces the computational costs associated with standard
CFD models to simulate groups of buildings; it simplies the geometry of the urban mesh and permits to extend the area of
simulation to a city scale.
The U-DTM is then inserted into the CFD-Complex and, by using
synthetic wind data series obtained by mesoscale modeling, which
enables the urban wind potential to be estimated in the absence of
experimental wind data. Since the U-DTM will, in some regions,
smooth the city geometry, and since wind data from numerical
mesoscale modeling usually overestimate the wind potential in
600
Fci 1
vref
(5)
vmodel i
where Fci is the correction factor, vre is the mean wind speed obtained with CFD-Urban and vmodeli is the mean wind speed obtained
with the CFD-Complex with the digital surface model for grid point
i. Depending on the extension of the urban area under analysis,
correction factors are dened for the area surrounding the buildings (area A) and for the building's rooftops (Area B), using the
algorithm described in (6).
vi
(6)
where Fcib is the correction factor obtained for buildings area, Fcig is
the correction factor for an area above the ground and without
buildings and Pi is the grid point i in the mean wind speed spatial
distribution map. For extended urban regions (e.g. area > 5 km2), it
may not be suitable to use the detail expressed in (7) due to the
slow pace and complexity that this procedure will bring to the
method. In this case a mean correction factor can be used for the
whole area (expression 7):
vi Fcim vmodel i
(7)
where Fcim is the mean correction factor obtained from the comparison between the results obtained in both simulations. The same
formulation is performed for power density calculation (W/m2).
The correction factors are then applied to the resource map(s) obtained with the application of standard wind ow models to the UDTM in order to obtain a more accurate spatial description of the
wind resource.
4. Application of the methodology to an urban area
To test the proposed methodology, two neighbor areas were
~o Domingos
selected inside Cascais municipality: Estoril and Sa
administrative regions e where the estimation of the wind potential was performed using the models and approach described in
the previous section. Synthetic wind data series obtained by
mesoscale modeling were used as wind resource input for the
models [23]. In the CFD-Complex approach, the orography of the
city surroundings (10 10 m orography lines) was used together
with a superimposed surface generated from the 3D geometric
information about the buildings in that area, enabling the generation of an urban digital terrain model (4 4m orography lines) as in
Fig. 1. In CFD-Urban, the 3D detailed geometry of the building was
used to model small sets of buildings in the city under study, according to the methodology described in Section 3.
The data obtained from the Portuguese Wind Potential Atlas is
referred to grid points distributed on the areas under study. CFDComplex was used to model the wind speed over the U-DTM and
therefore to obtain wind resource maps. Fig. 2 represents the mean
wind speed and power density at h 10 m obtained with CFD~o Domingos villages, respectively.
Complex in Estoril and Sa
~o Domingos were then
Some regions inside Estoril and Sa
modeled with the natural geometry of the buildings, using the CFDUrban model to establish the correction factors in order to improve
the precision of the U-DTM methodology results.
~o Domingos
Fig. 3 presents the modeled areas of Estoril and Sa
used to establish correction factors in order to obtain the nal wind
resource maps for the urban areas under study.
Figs. 4e7 show the mean wind speed and power density spatial
distribution for the urban areas represented in Fig. 3 for h 10 m. In
order to establish correction factors for the results obtained with
the surface methodology, several locations were selected in the
areas under study and the mean wind speed and power density
results were compared.
~o
Figs. 8 and 9 show the chosen locations in Estoril and Sa
Domingos villages, respectively. Table 1 shows the deviations between the results obtained with both models.
As shown in Table 1, when considering the locations above the
buildings, the deviations are in the order of 2%. For the open nonbuilt areas the deviations are larger, the average being in the order of 13%. The mean deviation for the whole area is 6.65%. It should
be noticed that some of the locations present larger deviations than
others especially in open non-built areas. This is mainly due to the
fact that they correspond to densely urbanized zones, which are
less well described by the U-DTM and where skimming ow has a
strong chance to occur.
The same evaluation was performed for power density (FPI).
Table 2 shows the results obtained for both villages. In this case the
deviations are in the order of 1% for the building rooftops, 23% for
Fig. 1. (a) Extract of the urban area under study, (b) 3D representation of the generated Digital Terrain Model.
601
Fig. 2. Spatial distribution of the mean wind speed and power density in Estoril and Sao Domingos. h 10 m.
Fig. 3. Estoril (top) and Sao Domingos (bottom) areas modeled with CFD-Urban.
Fig. 4. Spatial distribution of the mean wind speed for h 10 m above rooftops in Estoril.
the building-free areas and 10% for the whole analyzed areas.
The nal mean wind speed and power density spatial distribution maps were obtained by the application of the mean correction
factors to the overall area under study, being 6.65% for the mean
wind speed and 10% for the power density. Fig. 10 depicts the nal
maps for the regions under study.
5. Validation and discussion of the results
In order to validate the results obtained with the U-DTM
602
Fig. 6. Spatial distribution of the mean wind speed for h 10 m above rooftops in Sao Domingos.
Fig. 7. Spatial distribution of the power density for h 10 m above rooftops in Sao Domingos.
603
~o Domingos.
Fig. 9. Selected test locations for Sa
Table 1
Comparison results for mean wind speed obtained with both models in Estoril and Sao Domingos (h 10 m).
ID
Estoril
Sao Domingos
Area 1
1
2
3
4
5
6
7
8
9
10
11
Area 2
Area 3
Area 1
Area 2
CFD-Urban
CFD-Complex
CFD-Urban
CFD-Complex
CFD-Urban
CFD-Complex
CFD-Urban
CFD-Complex
CFD-Urban
CFD-Complex
4.7
5.11
5.29
4.65
5.69
4.9
3.79
3.37
3.38
3.46
3.77
5.18
5.53
5.48
5.09
5.26
5.09
5.40
5.50
5.20
5.40
5.20
5.03
5.11
5.24
4.6
5.23
5.71
4.13
4.35
4.1
4.51
4.88
5.53
5.64
5.36
5.52
5.44
5.4
5.65
5.42
5.4
5.35
5.47
5.33
5.03
5.14
4.86
5.05
5.04
4.59
4.49
4.52
4.8
4.68
4.95
4.97
4.89
4.96
4.95
4.93
4.93
5.08
5.01
4.84
4.91
5.16
4.14
4.66
5.10
5.06
4.72
4.63
4.52
4.85
4.73
e
4.66
4.53
4.67
4.8
4.69
4.73
4.81
4.66
4.59
4.76
e
5.06
4.61
4.23
4.51
4.67
4.81
4.64
5.09
4.98
5.07
5.05
4.45
4.52
4.75
4.72
5.02
4.6
4.87
4.86
4.72
4.47
4.82
604
Table 2
Comparison results for power density obtained with both models in Estoril and Sao Domingos(h 10 m).
ID
Estoril
Sao Domingos
Area 1
1
2
3
4
5
6
7
8
9
10
11
Area 2
Area 3
Area 1
Area 2
CFD-Complex
CFD-Urban
CFD-Complex
CFD-Urban
CFD-Complex
CFD-Urban
CFD-Complex
CFD-Urban
CFD-Complex
CFD-Urban
102
115
149
98
184
123
53
39
43
44
55
141
169
163
159
145
139
161
165
150
156
145
131
136
149
100
149
190
73
84
73
95
120
167
178
151
165
158
156
178
157
157
150
159
158
129
137
118
132
126
101
91
90
115
106
119
123
115
118
117
118
116
128
124
113
118
133
71
103
131
134
143
133
129
124
148
e
100
91
99
109
102
104
109
99
94
107
e
127
92
74
120
97
106
65
96
85
92
93
84
88
101
99
121
93
109
111
101
87
110
Fig. 10. Spatial distribution of the mean wind speed and power density, h 10 m in Estoril and Sao Domingos after the application of the correction factors.
Table 3
Comparison between estimated mean wind speed values (V) and Power density (PD) with measurements for both villages.
a)Estoril
~o Domingos
a)Sa
b)Estoril
b)S~
ao Domingos
c)Estoril
c)S~
ao Domingos
Correction
factor for V [%]
V [m/s]
CFDcomplex
H 10 m
V [m/s] LIDAR
H 10 m
98
87
93.4
93.4
100
100
5.07
4.65
4.83
4.99
5.17
5.34
4.88
e
4.88
e
4.88
e
V [m/s] Tires
H 10 m
4.47
e
4.47
4.47
Deviation
for V [%]
Correction factor
for FPI [%]
PD [W/m2]
CFD- complex
H 10 m
PD [W/m2]
LIDAR
H 10 m
PD [W/m2]
Tires
H 10 m
Deviation
for PD [%]
3.89
4.08
1.02
10.6
5.94
19.7
101
77
90
90
100
100
142
102
127
120
141
133
161
e
161
e
161
e
e
123
e
123
e
123
11.8
17.07
21.1
2.43
12.4
8.13
Fig. 11. Vertical mean wind speed proles estimated by the CFD-ORO model (with and
without mean correction factor) and measured by the LIDAR.
605
[2] N.G. Mortensen, L. Landberg, I. Troen, E.L. Petersen, Wind Atlas Analysis and
Application Program (WASP), vol. 1, 1983. Getting started, Roskilde.
~ es, P. Costa, A. Estanqueiro, A rst methodology for wind energy
[3] T. Simo
resource assessment in urbanised areas in Portugal, in: 2009 European Wind
Energy Conference Proceedings, Marseille, 2009.
[4] WindSim, WindSim, Version 4.9.1.24234, copyright 1997-2009, WindSim
AS, 2009.
[5] Meteodyn, UrbaWind, Version 1.5.0.0, copyright , Meteodyn, 2008, p. 2008.
[6] P. Costa, P. Miranda, A. Estanqueiro, Development and validation of the Portuguese Wind Atlas, in: 2006 European Wind Energy Conference Procedings,
Athens, 2006.
[7] Dong Li, Shujie Wang, Peng Yuan, A Review of micro wind turbines in the built
environment, in: Power and Energy Engineering Conference (APPEEC), 2010
Asia-Pacic
1,
2010,
pp.
1e4,
http://dx.doi.org/10.1109/
APPEEC.2010.5448223, 4, 28e31 March 2010.
[8] K. Syngellakis, P. Clement, J. Cace, Wind Energy Integration in the Urban
Environment - WINEUR, 2005. U.K.
[9] S. Gryning, E. Batchvarova, Measuring meteorology in urban areas e some
progress and many problems, in: Meteorological and Air Quality Models,
Springer, Berlin, 2009, pp. 125e132.
[10] J.L. Acosta, S.Z. Djokic, Assessment of renewable wind resources in UK urban
areas, in: MELECON 2010-2010 15th IEEE Mediterranean Electrotechnical
Conference, 2010, pp. 1439e1444 [Online], http://ieeexplore.ieee.org/xpls/
abs_all.jsp?arnumber5476217&tag1.
[11] R. Britter, T. Bentham, Spatially averaged ow within obstacle arrays, Atmos.
Environ. 37 (2003) 2037e2043.
[12] M. Kanda, Review of Japanese urban models and a scale model experiment, in:
Meteorological and Air Quality Models for Urban Areas, Springer, Berlin, 2009,
pp. 39e46.
[13] A. Kalmikov, G. Dupont, K. Dykes, C. Chan, Wind power resource assessment
in complex urban environments. MIT campus case-study using CFD analysis,
in: AWEA 2010 Windpower Conference, May 23-26, 2010, pp. 1e28.
[14] L. Ledo, P.B. Kosasih, P. Cooper, Roof mounting site analysis for micro-wind
turbines, Renew. Energy 36 (2011) 1379e1391.
[15] M. Tewari, et al., Impact of coupling a microscale computational uid dynamics model with a mesoscale model on urban scale contaminant transport
and dispersion, Atmos. Res. 96 (2010) 656e664.
[16] J. Baik, J. Kim, Urban ow and dispersion simulation using a CFD model
coupled to mesoscale model, Am. Meteorological Soc. (2009) 1667e1681.
[17] Tristan Wallbank, WindSim Validation Study - CFD Validation in Complex
Terrain, WindSim AS, Tnsberg, 2008 [Online], http://www.windsim.com/
documentation/papers_presentations/thesis/080512trw%20WindSim%
20Write%20Up%20-%20Validation%20study.pdf.
[18] Arne R. Gravdahl, Meso Scale Modeling with a Reynolds Averaged NavierStokes Solver. Assessment of wind resources along the Norwegian coast, in:
31th IEA Experts Meeting, State of the Art on Wind Resource Estimation,
RISOE, 1998. Roskilde.
[19] N.O. Jensen, N.E. Bush, Atmospheric turbulence, in: Engineering Meteorology,
Elsevier, Amsterdam, 1982, pp. 179e231.
[20] J.R. Holton, An Introduction to Dynamic Meteorology, third ed., Elsevier Science (USA), San Diego, USA, 1972.
[21] Michel Ferry, New features of the MIGAL solver, in: 9th Phoenics User Conference, 2002, pp. 1e24. Moscow, http://www.cham.co.uk/PUC/PUC_Moscow/
MRFDC.pdf [Online].
[22] CHAM Ltd, PHEONICS - Your Gateway to Successful CFD, 2013 [Online], http://
www.cham.co.uk/.
[23] P. Costa, P. Miranda, A. Estanqueiro, Development and validation of the Portuguese wind Atlas, in: EWEC 2006-European Wind Energy Conference 2006,
2006, pp. 1e9. Athens.
[24] S. Milashuk, W.A. Crane, Wind speed prediction accuracy and expected errors
of RANS equations in low relief inland terrain for wind resource assessment
purposes, Environ. Model. Softw. 26 (4) (Apr. 2011) 429e433 [Online], http://
www.sciencedirect.com/science/article/pii/S136481521000263X.
[25] S.M. Weekes, A.S. Tomlin, Evaluation of a semi-empirical model for predicting
the wind energy resource relevant to small-scale wind turbines, Renew. Energy 50 (2013) 280e288.
[26] Islam Abohela, Neveen Hamza, Steven Dudek, Effect of roof shape, wind
direction, building height and urban conguration on the energy yeld and
positioning of roof mounted wind turbines, Renew. Energy 50 (Feb. 2013)
1106e1118
[Online],
http://www.sciencedirect.com/science/article/pii/
S0960148112005381.