Beruflich Dokumente
Kultur Dokumente
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 20 March 2015
Received in revised form
5 November 2015
Accepted 15 February 2016
Available online 27 February 2016
A procedure for the optimisation of hydrokinetic turbine array layout through surrogate modelling is
introduced. The method comprises design of experiments, computational uid dynamics simulations,
surrogate model construction, and constrained optimisation. Design of experiments are used to build
polynomial and Radial Basis Function surrogates as functions of two design parameters: inter-turbine
longitudinal and lateral spacing, with a view to approximating the capacity factor of turbine arrays
with inline and staggered layouts, each of which having a xed number of turbines. For this purpose, two
scenarios have been used as case studies, considering uniform and non-uniform free-stream ows. The
major advantage of this method in comparison to those reported in the literature is its capability to
analyse different design parameter combinations that satisfy optimality criteria in reasonable computational time, while taking into account complex oweturbine interactions and different turbine types.
2016 Elsevier Ltd. All rights reserved.
Keywords:
Hydrokinetic energy
Turbine array layout
Surrogate based optimisation
1. Introduction
In recent years, in-stream hydrokinetic energy has drawn the
attention of investors around the world. The large amounts of energy found in river ows, tidal channels and ocean currents has
served as a strong motivation for research in optimising hydrokinetic turbine arrays with a view to making their commercial
exploration viable. There are several issues concerning the design
of turbine array layout, such as array size, complex oweturbine
interactions, different turbine types, and environmental impacts,
which are necessary to be taken into account, in order to employ
suitable optimisation methods. Vennell et al. [1] distinguish between two scales of array optimisation: macro and micro array
design scales, where macro-design relates to the total number of
turbines in a farm, the number in each row and the number of rows
in the array. On the other hand, micro design is concerned with the
individual positions of the turbines within the array. This paper
focuses on the aspect of micro design of turbines array
optimisation.
In almost all reported research work on the turbine array layout
optimisation problem, two approaches have been employed. In the
* Corresponding author.
lez-Gorben
~ a), qassim@
E-mail addresses: eduardogg@oceanica.ufrj.br (E. Gonza
peno.coppe.ufrj.br (R.Y. Qassim), pccrosman@ufrj.br (P.C.C. Rosman).
http://dx.doi.org/10.1016/j.renene.2016.02.045
0960-1481/ 2016 Elsevier Ltd. All rights reserved.
46
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
47
48
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
The number of turbines (Nt) are given and xed. For the inline
and staggered arrays 6 and 8 devices are considered, respectively.
Its vertical position (z-axis) is also xed, as well as the position of
the centreline of the turbines. The arrangement of the inline array
consists of two rows with three turbines each. In the case of the
staggered array, there are three rows, with two turbines in the
second row, and three turbines in each of the rst and third rows.
Turbines (T1, T2, , TNt) are numbered from top to bottom and from
left to right, starting at the top left turbine. For an illustration of the
turbine arrangement in each case, see Figs. 3e6.
A constant ow velocity of 2.5 m s1 is prescribed at the inlet for
the regular channel. In the case of the irregular channel, a sinusoidal ow velocity in [ms1] is given as an input signal at the inlet
boundary, see Fig. 7.
The outputs of the computer simulation are averaged surface
integrals of velocity magnitudes for each porous disk; i.e.,
Z
Ud dAT Ud AT
ud;i aT;i
(1)
i1
Fig. 2. Sampling plan using LHD maximin criterion for n 11 and s 2 (blue points)
and 9 extra points (red points). (For interpretation of the references to colour in this
gure legend, the reader is referred to the web version of this article).
n
X
Fig. 3. Top view of the regular channel geometry used for uniform ow CFD simulations. In the gure, the turbine congurations are shown: a) inline layout (top) and b) staggered
layout (bottom) with x1 20D and x2 2D.
Pi
1
rC A U 3
2 PL T d;i
(2)
Fig. 4. Front view of the regular channel geometry used for uniform ow CFD simulations. In the gure, the turbine congurations are shown for: a) inline layout (top) and b)
staggered layout (bottom) with x1 20D and x2 2D.
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
49
Fig. 5. Top view of the irregular channel geometry used for non-uniform ow CFD simulations. In the gure, the turbine congurations are shown for a) inline layout (top) and b)
staggered layout (bottom) with x1 20D and x2 2D.
Fig. 6. Front view of the irregular channel geometry used for non-uniform ow CFD simulations. In the gure, the turbine congurations are shown for: a) inline layout (top) and b)
staggered layout (bottom) with x1 20D and x2 2D.
Nt
P
CFj
i
Nt
P
Pi
(3)
Pr;i
where, Pi and Pr,i represent the power output and rated power of
the i-th turbine of a set of Nt turbines, respectively; r depicts water
density, taken to be 998.2 kg m3; nally, CPL depict the power
coefcient expressed in terms of the local velocity Ud. Through the
1-dimensional linear momentum actuator disc theory in an innite
medium an upper bound for the theoretical maximum power
extractable, the known Betz's limit [33], is obtained, with CP 16/
27 (CPL 2). However, it is worth noting that for a turbine in open
channel ows [34], the Betz upper bound may well be exceeded
due to constrained ow conditions, but a series of requirements
have to be achieved [35]. In this study, horizontal axis turbine types
with 20 m rotor diameters are considered to have a rated power of
2.58 MW at a ow velocity of 3 m s1. Detailed information about
the CFD simulations and validation of the porous-jump boundary
6. Surrogate construction
After the sampling plans have been computationally implemented, the next step involves selecting an approximating func^i (x), to use as a surrogate of complex computer
tional form, y
simulations, yi (x), in this case representing the CFj of each turbine
array layout. The majority of metamodels can be dened with a
linear combination of a set of basis functions {B1 (x), B2 (x), , Bn
(x)}. Thus, the general form of a metamodel can be expressed as:
b
y i x b1 B1 x b2 B2 x bl Bl x
(4)
yi x b
y i x i
(5)
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
50
b
y x b
b0
s
X
b
b i xi
i1
s
X
b
b ii x2i
s1 X
s
X
i1
b
b ij xi xj ;
b
y i wT f
i 1; 2;
s
X
b
b i xi
i1
s
X
s
X
b
b ii x2i
s1 X
s
X
i1
b
b 2ij x2i xj
ji1
b
b ij xi xj
i1 ji1
s1
X
s
X
s1
X
b
b 2ji xi x2j
i1 ji1
wi fkx xi k
n
X
wi f xj xi ;
i; j 1; 2; ; n
b
b 3ii x3i ;
i1
y wT F
(7)
Polynomial parameters, b
b i , are calculated through the least
squares procedure [15], which minimises the sum of the square
errors,
8
92
=
n <
L
X
X
T
T
LS y Bb y Bb
Bj xbj
yi
:
;
(8)
j0
(13)
The values of the weights, like the coefcients of the polynomials, may be estimated using the least squares estimator presented in Eq. (10). For that purpose, we have
1
b FT F
FT y
w
(14)
vLS
2y BbT B 0T
vb
(9)
1
b
BT y
b BT B
(10)
cv
(12)
i1
1; 2;
i1
(11)
b
yj
i1
s
X
n
X
i1
i1 ji1
(6)
b
y x b
b0
n
1X
2
y b
yi
n i1 i
n2N
(15)
"
p
1X
2
RMSE
y b
yi
np i1 i
#1=2
(16)
MAX maxjyi b
yij
(17)
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
51
MAX
Nt
X
Pi $t
cost of leasing
For each of the simulations scenarios and layouts, results for
metamodel tting are summarised in Fig. 9. Note that values of
RMSE are normalised by the range of observed data [ymin, ymax] and
the values of MAX represent units of Capacity Factor.
Results for each scenario and surrogate are represented graphically (see Figs. 10e17) as contour plots to help visualize the shape
of the response surfaces.
Subject to
Nt
X
Nt
P
Pi $t
i
Nt
P
Pi
$
f0 x
gi x gi ; i 1; ; m
hj x ji ; j 1; ; k
where the vector x {x1, ,xs} is the set of design variables of the
problem, the function f0: Rn / R is the objective function, the
functions gi and hi: Rn / R, i 1, ,m and j 1, ,k are the
inequality and equality constraint functions, and the constants b1,
,bm and d1, ,dk are the bounds and values of the constraints. In
surrogate optimisation, the objective function is given by the vali^ (x).
dated metamodel, i.e., f0 y
In the turbine array layout optimisation problem for a commercial scale project, a series of constraints types arise (e.g.: economic, environmental, zoning, structural loads, etc.) that need to be
satised. As these constraint types depend on the specic nature of
each project, it is not the objective of this work to present a unique
optimisation problem, but to formulate and include a series of
constraints to serve as a prototype for more complex optimisation
problems. In Li et al. [46], an integrated model for estimating energy cost is presented, which considers many aspects of a tidal
Pr;i $t CFx1 ; x2 $
Nt
X
Pr;i $t
i
Nt
X
Pr;i $t
(19)
8. Surrogate optimisation
Maximise
Subject to :
Nt
X
Pr;i
b
y x$
(18)
t 8760h
(20)
(21)
(22)
h i
Cost of leasing $=year 0:025$= m2 $year $AF m2
(23)
Di 20m
(24)
ci 1; ; NT
(26)
20D x1 30D
(27)
Inline array:
Nt 6
(28)
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
52
Fig. 9. Normalised root mean square error (top) and absolute maximum error (bottom) obtained with the surrogates for each array layout out and scenario.
Lm x1 4x2
(29)
2
(30)
Nt 8
(32)
0:5
5x2
Lm 2 x21 x22
(33)
(34)
Fig. 10. Quadratic polynomial capacity factor surfaces responses for arrays in Channel-1, inline (left) and staggered (right) layouts.
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
53
Fig. 11. Cubic polynomial capacity factor surfaces responses for arrays in Channel-1, inline (left) and staggered (right) layouts.
Fig. 12. Linear RBF capacity factor surfaces responses for arrays in Channel-1, inline (left) and staggered (right) layouts.
(x) with Eq. (3). Constraints (20e27) are common for both types of
array layouts, while constraints (28e31) are specic for the inline
array and constraints (32e37) are particular for the staggered
layout. Constraint (20) denes the time in hours for which the
prot is calculated. Constraints (21e23) set the price of electricity,
O&M costs and project area leasing, respectively. Constraint (24)
species the rotor diameter, D, of the i-th turbine in metres [m].
(35)
satises all conditions is selected. For this purpose, the design parameters are discretised in increments of 0.25D for x1 and of 0.025D
for x2. The results obtained for the above optimisation model for
each modelling scenario and surrogate are summarised in Table 1.
In general, the results of x1 and x2 obtained for each of the
surrogates implemented in common scenarios are very similar. As
54
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
Fig. 13. Thin Plate Spline RBF capacity factor surfaces responses for arrays in Channel-1, inline (left) and staggered (right) layouts.
optimum array congurations obtained for the channel under nonuniform ow are specic for this particular case.
9. Discussion and recommendations
Through this paper, we have presented a different approach for
the turbine array layout optimisation problem. Two variations of
Polynomial and RBF surrogate methods are employed to t an
objective function dependent on two variables: longitudinal and
lateral spacing. The methodology is applied to two different scenarios, considering uniform and non-uniform ow, with the
objective to assess the capacities of each surrogate.
As a consequence of having a greater number of turbine rows,
results from the surrogate tting validation evidence a greater
sensitivity of the response results for staggered layouts than for
inline ones. Similar pattern occurs with the cases with non-uniform
ow, as the irregular channel scenario increases non-linearity in
the objective function response.
For both ow conditions and layouts, Radial Basis Functions
denote an overall better performance than polynomial models in
Fig. 14. Quadratic polynomial capacity factor surfaces responses for arrays in Channel-2, inline (left) and staggered (right) layouts.
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
55
Fig. 15. Cubic polynomial capacity factor surfaces responses for arrays in Channel-2, inline (left) and staggered (right) layouts.
energy turbines array may be adopted to increase the approximation with real conditions. The geometry and rotation of the turbines
under unsteady open channel ow conditions can be modelled. As
the complexity of computational modelling increases, simulations
become more time demanding. It is for these cases that the
methodology presented in this paper becomes even more
appealing, as it provides a surrogate of the underlying model,
which is easier to work with for design space exploration and
sensitivity analysis. The main difculties of the method are to
formulate the problem in terms of a few number of design variables
and to obtain a reliable surrogate using as less as possible computer
simulations.
Future research is directed to implement the SBO approach for
the hydrokinetic array layout problem considering continuous and
discrete design parameters, like space coordinates for turbine
positioning and rotor turbine diameter size, as well as to develop a
more sophisticated optimisation model, including the threedimensional free positioning of turbines and considering environmental constraints as the inuences in sediment dynamics.
Fig. 16. Linear RBF capacity factor surfaces responses for arrays in Channel-2, inline (left) and staggered (right) layouts.
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
56
Fig. 17. Thin Plate Spline RBF capacity factor surfaces responses for arrays in Channel-2, inline (left) and staggered (right) layouts.
Table 1
Results from the optimisation model.
Flow scenario
Array layout
Surrogate
x1 [D]
x2 [D]
CF [%]
Prot [M$/year]
Uniform
Uniform
Uniform
Uniform
Uniform
Uniform
Uniform
Uniform
Non-uniform
Non-uniform
Non-uniform
Non-uniform
Non-uniform
Non-uniform
Non-uniform
Non-uniform
Inline
Inline
Inline
Inline
Staggered
Staggered
Staggered
Staggered
Inline
Inline
Inline
Inline
Staggered
Staggered
Staggered
Staggered
Quadratic poly.
Cubic poly.
RBF-Linear
RBF-TPS
Quadratic poly.
Cubic poly.
RBF-Linear
RBF-TPS
Quadratic poly.
Cubic poly.
RBF-Linear
RBF-TPS
Quadratic poly.
Cubic poly.
RBF-Linear
RBF-TPS
30.00
30.00
30.00
29.25
10.00
10.00
10.00
11.25
21.50
22.75
21.75
22.50
10.00
10.00
12.50
11.00
2.000
2.275
2.075
2.125
2.500
2.500
2.600
2.225
3.600
4.000
3.725
3.950
4.000
4.000
3.925
3.800
46.50
47.38
46.98
47.82
56.22
56.41
56.51
58.57
64.52
69.83
69.83
71.15
59.88
62.49
61.75
62.51
13,043
13,225
13,161
13,425
21,466
21,538
21,527
22,276
18,258
19,521
19,700
19,930
22,052
23,011
22,311
22,944
Acknowledgements
The corresponding author wishes to acknowledge the postdoctoral fellowship provided by the following funding agency from
Brazil: Conselho Nacional de Desenvolvimento Cientco e Tec gico (CNPq) (Grant number: 500177/2014-7).
nolo
References
[1] R. Vennell, S.W. Funke, S. Draper, C. Stevens, T. Divett, Designing large arrays
of tidal turbines: a synthesis and review, Renew. Sust. Energy Rev. 41 (2015)
454e472, http://dx.doi.org/10.1016/j.rser.2014.08.022.
[2] I.G. Bryden, S.J. Couch, How much energy can be extracted from moving water
with a free surface: a question of importance in the eld of tidal current
energy? Renew. Energy 32 (2007) 1961e1966, http://dx.doi.org/10.1016/
j.renene.2006.11.006.
[3] C. Garrett, P. Cummins, Limits to tidal current power, Renew. Energy 33
(2008) 2485e2490, http://dx.doi.org/10.1016/j.renene.2008.02.009.
[4] R. Vennell, Tuning turbines in a tidal channel, J. Fluid Mech. 663 (2010)
253e67, http://dx.doi.org/10.1017/S0022112010003502.
[5] R. Vennell, Tuning tidal turbines in-concert to maximise farm efciency,
J.
Fluid
Mech.
671
(2011)
587e604,
http://dx.doi.org/10.1017/
S0022112010006191.
[6] R. Vennell, Realizing the potential of tidal currents and the efciency of turbine farms in a channel, Renew. Energy 47 (2012) 95e102, http://dx.doi.org/
10.1016/j.renene.2012.03.036.
[7] S.H. Lee, S.H. Lee, K. Jang, J. Lee, N. Hur, A numerical study for the optimal
arrangement of ocean current turbine generators in the ocean current power
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
lez-Gorben
~ a et al. / Renewable Energy 93 (2016) 45e57
E. Gonza
[18] K.T. Fang, R. Li, Sudjianto, A Design and Modelling for Computer Experiments,
Chapman and Hall CRC Press, 2005.
[19] R.R. Barton, Metamodels for simulation input-output relations, Proc. 24th
Winter Simul. Conf. (1992) 289e299. Article number: DEP LW1174.
[20] G.G. Wang, S. Shan, Review of metamodelling techniques in support of engineering design optimization, J. Mech. Des. 129 (4) (2006) 370e380, http://
dx.doi.org/10.1115/1.2429697.
[21] J. Sacks, W.J. Welch, T.J. Mitchell, H.P. Wynn, Design and analysis of computer
experiments, Stat. Sci. 4 (1989) 409e435.
[22] T.J. Santner, B.J. Williams, W.I. Notz, The Design and Analysis of Computer
Experiments, Springer-Verlag, New York, 2003.
[23] D. Bursztyn, D.M. Steinberg, Comparison of designs for computer experiments, J. Stat. Plan. Inference 136 (2006) 1103e1119, http://dx.doi.org/
10.1016/j.jspi.2004.08.007.
[24] M.D. McKay, R.J. Beckman, W.J. Conover, A comparison of three methods for
selecting values of input variables in the analysis of output from a computer
code, Technometrics 21 (1979) 239e245, http://dx.doi.org/10.2307/1268522.
[25] F.A.C. Viana, G. Venter, V. Balabanov, An algorithm for fast generation of
optimal Latin hypercube designs, Int. J. Num. Meth. Eng. 82 (2010) 135e156.
[26] M. Johnson, L. Moore, D. Ylvisaker, Minimax and maximin distance design,
J. Stat. Plan. Inference 26 (1990) 131e148, http://dx.doi.org/10.1016/03783758(90)90122-B.
[27] J.L. Loeppky, J. Sacks, W.J. Welch, Choosing the sample size of a computer
experiment: a practical guide, Technometrics 51 (4) (2009) 366e376, http://
dx.doi.org/10.1198/TECH.2009.08040.
[28] K. Palmer, M. Realff, Metamodelling approach to optimization of steady-state
owsheet simulations: model generation, Chem. Eng. Res. Des. 80 (7) (2002)
760e772, http://dx.doi.org/10.1016/10.1205/026387602320776830.
[29] ANSYS Inc, ANSYS FLUENT User's Guide. Release 14.5, USA, 2012.
[30] X. Sun, Numerical and Experimental Investigation of Tidal Current Energy
Extraction, Ph.D. Thesis, University of Edinburgh, UK, 2008, http://hdl.handle.
net/1842/2756.
[31] L. Bai, R.G. Spencer, G. Dudziak, Investigation of the inuence of array
arrangement and pacing on tidal energy converter (tec) performance using a
3-dimensional cfd model, Proc. 8th Eur. Wave Tidal Energy Conf. Upps. Swed.
(2009) 654e660.
[32] S.R. Turnock, A.B. Phillips, J. Banks, R. Nicholls-Lee, Modelling tidal current
turbine wakes using a coupled RANS-BEMT approach as a tool for analysing
power capture of arrays of turbines, Ocean Eng. 38 (2011) 1300e1307, http://
dx.doi.org/10.1016/j.oceaneng.2011.05.018.
[33] A. Betz, Das maximum der theoretisch moglichen Ausnutzung des Windes
durch Windmotoren, Z. fr Das. gesamte Turbinenwes. 26 (1920) 307e309.
[34] G.T. Houlsby, S. Draper, M. Oldeld, Application of Linear Momentum Actuator
Disc Theory to Open Channel Flow, Tech rep no 2296-08, University of Oxford,
Oxford, UK, 2008.
[35] R. Vennell, Exceeding the Betz limit with tidal turbines, Renew. Energy 55
(2013) 277e285, http://dx.doi.org/10.1016/j.renene.2012.12.016.
~ a, Um modelo de otimiza~
tico para a geraa
~o de
[36] E.G. Gorben
ao matema
trica a partir de correntes hidrodina
^micas, PhD Thesis, COPPE/UFRJ,
energia ele
2013 (In Portuguese), http://www.oceanica.ufrj.br/intranet/teses/2013_
Doutorando_Eduardo_Gonzalez_Gorbena_Eisenmann.pdf.
~ o na
~o linear para simu[37] M.I. Santos, Constru~
ao de metamodelos de regressa
cnica de Lisla~
ao de acontecimentos discretos, PhD Thesis, Universidade Te
boa, 2002 (In-portugues).
[38] E.P. Box, N.R. Draper, Empirical Model Building and Response Surfaces, Wiley,
NewYork, 1987.
[39] A.I.J. Forrester, A.J. Keane, Recent advances in surrogate-based optimization,
Prog.
Aerosp.
Sci.
45
(2009)
50e79,
http://dx.doi.org/10.1016/
j.paerosci.2008.11.001.
[40] B.J.C. Baxter, The Interpolation Theory of Radial Basis Functions, PhD Thesis,
Trinity College Cambridge University, 1992.
57
[41] G.B. Wright, The Interpolation Theory of Radial Basis Functions, PhD Thesis,
University of Colorado, 2003.
[42] N.V. Queipo, R.T. Haftka, W. Shyy, T. Goel, R. Vaidyanathan, P. Kevin Tucker,
Surrogate-based analysis and optimization, Prog. Aerosp. Sci. 41 (1) (2005)
1e28, http://dx.doi.org/10.1016/j.paerosci.2005.02.001.
[43] T. Hastie, R. Tibshirani, J. Friedman, The Elements of Statistical Learning,
second ed., Springer, New York, 2009.
[44] R.H. Myers, D.C. Montgomery, C.M. Anderson-Cook, Response Surface Methodology Process and Product Optimization Using Designed Experiments, third
ed., John Wiley & Sons, New York, 2009.
[45] A. Forrester, A. Sobester, A. Keane, Engineering Design via Surrogate Modelling: a Practical Guide, John Wiley & Sons, 2008.
[46] Y. Li, B.J. Lence, S.M. Calisal, An integrated model for estimating energy cost of
a tidal current turbine farm, Energy Convers. Manag. 52 (2011) 1677e1687,
http://dx.doi.org/10.1016/j.enconman.2010.10.031.
[47] IEA, Energy prices and taxes: quarterly statistic, Int. Energy Agency 4 (2014),
http://dx.doi.org/10.1787/16096835.
[48] IRENA, Renewable energy technologies: cost analysis series, Int. Renew. Energy Agency (2012) 1.
[49] Maine DEP, Information sheet: regulation of tidal and wave energy projects,
Dep. Environ. Prot. (2010). DEP LW1174. Available online in: http://www.
maine.gov/dep/water/dams-hydro/is_tidal_wave_reg.html.
[50] S. MacDougall, Financial evaluation and cost of energy, in: S. MacDougall,
J. Colton (Eds.), Community and Business Toolkit for Tidal Energy Development, Acadia Tidal Energy Institute, 2012, pp. 131e145. Available online in:
http://tidalenergy.acadiau.ca/tl_les/sites/tidalenergy/resources/Documents/
Toolkit/Module_7Sweb.pdf.
[51] G. Allan, M. Gilmartin, P. McGregor, K. Swales, Levelised costs of wave and
tidal energy in the UK: cost competitiveness and the importance of banded
renewables obligation certicates, Energy Policy 39 (2011) 23e39, http://
dx.doi.org/10.1016/j.enpol.2010.08.029.
[52] C.M. Johnstone, D. Pratt, J.A. Clarke, A.D. Grant, A techno-economic analysis of
tidal energy technology, Renew. Energy 49 (2013) 101e106, http://dx.doi.org/
10.1016/j.renene.2012.01.054.
[53] Y. Li, L. Willman, Feasibility analysis of offshore renewables penetrating local
energy systems in remote oceanic areas e a case study of emissions from an
electricity system with tidal power in Southern Alaska, Appl. Energy 117
(2014) 42e53, http://dx.doi.org/10.1016/j.apenergy.2013.09.032.
[54] P. Venkataraman, Applied Optimization with Matlab Programming, John
Wiley & Sons, 2001.
[55] S.M.F. Rodrigues, P. Bauer, J. Pierik, Modular approach for the optimal wind
turbine micro siting problem through CMA-ES algorithm, in: Proceedings of
the 15th annual conference companion on Genetic and evolutionary
computation, ACM, 2013, pp. 1561e1568.
[56] J. Zhang, S. Chowdhury, A. Messac, L. Castillo, A response surface-based cost
model for wind farm design, Energy Policy 42 (2012) 538e550, http://
dx.doi.org/10.1016/j.enpol.2011.12.021.
[57] J. Zhang, S. Chowdhury, A. Messac, Uncertainty quantication in surrogate
models based on pattern classication of cross-validation errors, in: Proceedings of the 14th AIAA/ISSMO Multidisciplinary Analysis and Optimization
Conference, Indianapolis, USA, 2012.
[58] A. Mehmani, W. Tong, S. Chowdhury, A. Messac, Surrogate-based particle
swarm optimization for large-scale wind farm layout Design, in: Proceedings
of the 11th World Congress on Structural and Multidisciplinary Optimization,
Sydney, Australia, 2015.
[59] A. Dean, M. Morris, J. Stufken, D. Bingham, Handbook of Design and Analysis
of Experiments, Chapman and Hall/CRC, 2015.
[60] B.G. Husslage, G. Rennen, E.R. van Dam, D. den Hertog, Space-lling Latin
hypercube designs for computer experiments, Optim. Eng. 12 (4) (2011)
611e630, http://dx.doi.org/10.1007/s11081-010-9129-8.