Professional Documents
Culture Documents
Energy
journal homepage: www.elsevier.com/locate/energy
a r t i c l e i n f o a b s t r a c t
Article history: As a continuation of authors’ previous work, this work extends and hackles the numerical method for
Received 10 October 2020 wind turbine wakes based on the vortex method, and proposes the Lagrangian actuator model (LAM)
Received in revised form which is used for the representation of the wind turbine rotor under the Lagarangian framework. This
27 April 2021
paper provides two examples of the LAM, the Lagrangian actuator line (LAL) model and the Lagrangian
Accepted 25 May 2021
Available online 1 June 2021
actuator disc (LAD) model, and constructs matching numerical methods for wake predictions respec-
tively. Those methods have high computation efficiency, and the results coincide with the wind tunnel
test data well. Moreover, based on that, a vorticity description framework centered on vortex geometric
Keywords:
Wind turbine wake
structures is established to illustrate wind turbine wake phenomena and explore the wake evolution
Lagrangian framework mechanism.
Actuator model © 2021 Published by Elsevier Ltd.
Vortex method
1. Introduction turbulence, where the velocity field is relatively simple [6]. How-
ever, until now, there has been no clear agreement on the dividing
With the large-scale development of wind farms, the problem of point between them. It is generally considered that the near wake
wake disturbance of wind turbines is becoming more prominent covers the scope from the rotor to 1e3 times the rotor diameter
[1]. A wake is the downstream flow field disturbed by a wind tur- along the downstream distance [4], and far wake refers to the scope
bine. One of its specific demonstrations is the velocity deficit that outside the scope of the near wake (some scholars tend to extend
causes power loss of downstream wind turbines. It is estimated that the near wake scope to the 5 times the rotor diameter [5]).
the power loss caused by the wake effect accounts for 5%e20% of Currently there are two common approaches to forecast the
the total generating capacity at a large wind farm [2]. It is apparent wake velocity deficit, the analytical wake model method and the
that the wake effect among wind turbines plays an important role numerical method based on Computational Fluid Dynamics (CFD)
in the layout optimization and economic evaluation of a wind farm [4].
[3]. The analytical wake model dates back to 1980s. It describes the
Typical wind turbine velocity profiles in a wind turbine wake are distribution of the velocity deficit along the downstream distance
shown in Fig. 1. There is an M-shape wake velocity profile close to in a wake, identifying the optimal distance among wind turbines in
the wind turbine (Zone A). Along the downstream distance, it a wind farm. Under some phenomenological assumptions, the
gradually takes the shape of a “bell” like the Gaussian function, and analytical wake model derives a simple semi-empirical formula
finally recovers to the uniform inflow velocity profile (Zone B). The based on conservation laws of momentum and mass. The model is
two zones are named as the near wake and the far wake, respec- super-fast and low-cost, but failed to consider fluid features and
tively [4,5]. The distinction between them is on the basis of the wake evolution processes. The first analytical wake model is the
complexity of the wake field: the near wake is strongly influenced Jensen wake model, which is also called the top-hat wake model
by the wind turbine rotor, dominated by the pressure, and its ve- [7]. The model assumes that the wake region shows linear expan-
locity field is more complex; the far wake seldom retains rotor sion along the flow direction, and the wake velocity distribution is
geometry and movement information and is controlled by the even and in the shape of a top-hat. With a simple form and high
efficiency, the Jensen model is widely applied in commercial soft-
ware like WAsP, WindFarmer [8]. Since then, the Jensen model has
* Corresponding author.
been improved continuously in many aspects, among which the
E-mail address: shijian@buaa.edu.cn (J. Shi). most important one lies in the revision of the velocity profile shape.
https://doi.org/10.1016/j.energy.2021.121074
0360-5442/© 2021 Published by Elsevier Ltd.
W. Liu, J. Shi, H. Chen et al. Energy 232 (2021) 121074
2
W. Liu, J. Shi, H. Chen et al. Energy 232 (2021) 121074
X
n
GT rkT þ GR rkR ¼ Gi rki ; (2)
i¼1
S¼ 1 þ S2 ðn þ eÞ2 ; (8)
n¼∞ R
GS
vS;x ¼ S: (9)
2R
Note that
!3 !3
X
þ∞
h2
2 X
þ∞
h2
2
1 þ S2 ðn þ eÞ2 ¼ 1 þ S2 ðn þ 1 eÞ2 :
n¼∞ R n¼∞ R
(10)
Fig. 3. The Lagrangian actuator line model. The symmetry means that the range of e can be narrowed to [0, 0.5].
3
W. Liu, J. Shi, H. Chen et al. Energy 232 (2021) 121074
We hope S f R/hS, which will make vx and vS,x similar in form. On the other hand, Eq. (6) also expresses the unidirectional
Actually, it will be satisfied as long as R/hS is limited within a certain sequence of vortex rings, if the value range of n is changed to n ¼ 0,
scope, as shown in Fig. 5. 1, ,,,, þ ∞. At the same time, note that the value range of the
It is easy to get that when parameter e should expand to [0, 1], considering the disappearance
of the symmetry. According to the Biot-Savart Law, it is easy to get
hS the axial induced velocity vS,x of the sequence at (K, 0, 0)
1; (11)
R
2 2 !32
no matter what value the e is, GS X
þ∞
hS K h K
vS;x ¼ 1 þ ðn þ eÞ2 2 ðn þ eÞ S þ :
2R n¼0 R R R R
2R
Sz : (12) (17)
hS
Similarly, the accumulation part of infinite series is S
Then
X
þ∞ 2 !32
GS GS h2S 2 hS K K
vS;x ¼ Sz : (13) S¼ 1þ ðn þ eÞ 2 ðn þ eÞ þ : (18)
2R hS n¼0
R2 R R R
Let vx ¼ vS,x, then Inspired by the previous work, consider the value changes of
ShS =ð2RÞ with R/hS when the values of K/R and e vary, and make
G GS
¼ : (14) comparison with FðK=RÞ in Eq. (16). The numerical results are
h hS shown in Fig. 6. There are four situations where K/R ¼ 1, 2, 6, 10
(K ¼ 0.5D, 1D, 3D, 5D) respectively. In all these situations, ShS =ð2RÞ
Now, we get two constraints, Eq. (11) and Eq. (14), regarding the
tends to a fixed value FðK=RÞ with the increase of R/hS. It means
sequence of vortex rings.
when the value of R/hS is big enough, vx and vS,x have the same form.
Following a similar process as above descriptions, we can
At this moment, letting vx ¼ vS,x, we can draw the same conclusion
explore the near wake region to look for possible and more strict
as Eq. (14). As shown in Fig. 6, in the regin of concern (K/R 2, i.e.,
constraints. The parameter equation of a semi-infinite length he-
K/D 1), the constraint of R/hS is
lical vortex line is also the Eq. (3). For convenience, let a ¼ q f,
then hS
8 1: (19)
R
>
> hða þ fÞ
>
<x ¼
>
2p
; 0 pt 0 pt 0 pt -0pt.
(15) Note that Eq. (19) is the same as Eq. (11), which means the
>
> y ¼ Rcos a;
>
> constraints of Eq. (11) and Eq. (14) can guarantee that the sequence
:
z ¼ Rsin a; well approximates the axis induced velocity in the region of
concern. At last, there is an interesting numerical phenomenon: as
where a2ð f; þ∞Þ. By integrating the Biot-Savart law, it is easy to shown in Fig. 6(d), when K/R ¼ 10, the curves of ShS =ð2RÞ coincide
get the induced velocity in the x direction vx of the semi-infinite with each other under the situations of e ¼ 0.0 and e ¼ 1.0, e ¼ 0.2
length helical vortex line at any point (K, 0, 0) on the axis and e ¼ 0.8, e ¼ 0.4 and e ¼ 0.6. These coincident curves are the
0 1 same as the corresponding ones in Fig. 5. This is because K/
2 !12 R ¼ 10(K ¼ 5D) is big enough and this condition is close to the
G K G 1B
B1 þ R C
C infinite far region where the symmetry shown in Eq. (10) comes
vx ¼ F ¼ 1þ (16)
h R h 2@ K A into play.
Now, return to the original problem. Assume that the helical
vortex line and the vortex ring sequence are released by a vortex
Compared to Eq. (5), there is an extra factor FðK=RÞ in the formula line generator and a vortex ring generator respectively in the uni-
above. form and stable background flow. The rotation period of the rotor is
T and the releasing time interval of the vortex ring generator is TS,
ignoring the interaction between vortices,
h ¼ TV0 ; (20)
hS ¼ TS V0 ; (21)
R
TS ; (22)
V0
TS
GS ¼ G: (23)
T
Take the equal sign in Eq. (22), then the parameter configuration of
Fig. 5. The relationship between ShS/(2R) and R/hS. the vortex ring generator is
4
W. Liu, J. Shi, H. Chen et al. Energy 232 (2021) 121074
Fig. 6. The relationship between ShS/(2R) and R/hS when the values of K/R and e change.
R
TS ¼ ; (24)
V0
RG
GS ¼ ; (25)
V0 T
the vortex filament method [34], the vortex sheet method [35] and parameters, the vortex ring circulation G, the vortex core radius rc,
the vortex volume method [36]. As for the wake issue of wind the vortex ring radius rh and the central axial position x.
turbines, the vortex filament is closer to the vortex structure in an The control equation is the NeS equation (the vorticity transport
actual wake intuitively. For that reason, this paper chooses the equation)
vortex filament as the discrete element, hoping a good balance
between computation precision and efficiency. vu
¼ ðV , VÞu þ ðu , VÞV þ nD,u; (26)
A wind turbine wake prediction method using the LAL model vt
and the vortex filament element was introduced in authors’ pre-
where u is the vorticity, V is the velocity and n is the coefficient of
vious article where it was called the VFWM [31]. This work will
kinematic viscosity. Three on the right hand side represent the
discuss a numerical method based on the LAD model. Similar to the
convection, stretching and diffusion of vorticity respectively.
VFWM, this method with obviously higher computation efficiency
The induced velocity of a vortex ring can be calculated through
is deeply inspired by the Pseudo-Implicit Predictor-Corrector (PIPC)
discretizing the ring into some linear vortex filaments and accu-
method of helicopter aerodynamics [37].
mulating their induced velocities.
Under the velocity gradient, the vortex ring radius can be
3.1. Vortex ring
stretched to rh þ Drh. At this moment, the vortex core radius will
have a corresponding change. Suppose the fluid is incompressible,
A vortex ring is a round vortex filament end to end as shown in
using the mass conservation law, then the stretching effect on the
Fig. 8, where G is the vortex ring circulation, rc is the vortex core
core radius can be written as
radius, rh is the vortex ring radius and O(x, y, z) is the central point.
In the stable background flow field, if the wind turbine rotor is
2p2 r 2c rh ¼ 2p2 ðrc Drc Þ2 ðrh þ Drh Þ (27)
treated as a vortex ring generator, the velocity field inducted by the
vortex ring sequence in the wake must be symmetrical to the
Then
central line. This means the vortex rings can only move along the
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
central line or rotate around the central line or stretch evenly in this Drc rh
velocity field, but it can not offset, slant or twist. This feature ¼1 : (28)
rc rh þ Drh
guarantees the reasonableness of using the vortex ring as the basic
unit. A vortex ring in the wake can be completely expressed by four The vortex core growth under the viscosity and turbulent effects
can be described by following equations [31].
d 2
r ¼ 4aðn þ a1 G þ nt Þ; (29)
dt c
rffiffiffi
2 A pffiffiffi
nt ¼ kD; (30)
3 pK 1:5
0
k ¼ Dk þ k0 ; (31)
x 0:64
Dk ¼ min 0:506V 20 a1:665 I0:065
0 ; 0:506V 20 a2 ; (32)
D
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a ¼ 0:5ð1 1 CT Þ; (33)
^
ind;r ðxi;l;k ; rhi;l;k Þ ¼ εVind;r ðxi;l;k ; rh i;l;k Þ þ ð1 εÞVind;r ð~
xi;l;k ; r~hi;l;k Þ;
n1 n1
V
(39)
1
~ xi;l1;k1 þ Dt V0 þ
xi;l;k ¼ ~ V ðxn1 ; r n1 Þ þ
4 ind;x i;l;k h i;l;k
where Vind,x denotes the axial component of the induced velocity at xðI; h; 0Þ ¼ 0; (42)
a certain point, Vind,r denotes its radial component.
Let xi,l,k denote the central point axial coordinate x(Ii, hl, xk) of the
rh ðI; h; 0Þ ¼ rh;shed ðIÞ; (43)
vortex ring (Ii, hl, xk) in simplified form and rhi;l;k denote the ring radius
rh(Ii, hl, xk). Owing to the symmetry, we might as well utilize the flow
velocity at the point ðxi;l;k ; rhi;l;k ; 0Þ on the vortex ring (Ii, hl, xk) to rc ðI; h; 0Þ ¼ rc;shed ðIÞ: (44)
compute the translation and stretching effects under the flow field. Let
In these equations, rh,shed(I) is the initial radius of the tip vortex or
Vind;x xi;j;k ; rhi;l;k denote the axial component of this velocity and the root vortex which can be calculated through BEM theory.
Vind;r ðxi;j;k ; rhi;l;k Þ denote the radial component. Imitating the PIPC rc,shed(I) is the initial core radius of the tip vortex or the root vortex
method, a steady solution to the formula above can be constructed: which is usually 10% of the average chord length of the blade and
the numerical result is not sensitive to the value.
1 ^ Owing to the inherent periodicity, the wake vortex system will
xni;l;k ¼ xni;l1;k1 þ Dt V0 þ V ind;x ðxi;l;k ; rhi;l;k Þ þ
4 return to the former condition after TS,T, then the natural periodic
boundary condition is
^ ^ (36)
V ind;x ðxi;l;k1 ; rhi;l;k1 Þ þ V ind;x ðxi;l1;k ; rhi;l1;k Þ þ
xðI; h; xÞ ¼ xðI; h þ TS;T ; xÞ; (45)
i
^
V ind;x ðxi;l1;k1 ; rhi;l1;k1 Þ ;
rh ðI; h; xÞ ¼ rh ðI; h þ TS;T ; xÞ: (46)
1 ^ The discretization and initial condition and periodic boundary
rh ni;l;k ¼ rh ni;l1;k1 þ Dt V ind;r ðxi;l;k ; rhi;l;k Þ þ
4 condition of the whole computational domain are shown in Fig. 10.
For convenience in the signification, we make TS,T ¼ 5TS,R ¼ 10Dt.
^ ^
ind;r ðxi;l;k1 ; rhi;l;k1 Þ þ V ind;r ðxi;l1;k ; rhi;l1;k Þ þ
V (37)
The sequence of vortex rings with equal distance can be used to
initiate the first iteration. Iteration convergence can be identified
^
V ind;r ðxi;l1;k1 ; rhi;l1;k1 Þ ; through the formula below
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
XXX 2
n
2
RMS ¼
1 n1
xni;l;k xn1 i;l;k þ r h i;l;k
r h i;l;k
; (47)
imax lmax kmax D i l k
7
W. Liu, J. Shi, H. Chen et al. Energy 232 (2021) 121074
4. Verification
Fig. 12. Comparison with the Jensen wake model and the improved Jensen wake model.
Fig. 13. Geometric characteristics Parameters under the LAL and the LAD model.
9
W. Liu, J. Shi, H. Chen et al. Energy 232 (2021) 121074
Acknowledgements
References turbine near-wakes and les modeling with new porous disc approach. En-
ergies 2021;14(8):2101.
[20] Abdelsalam AM, Boopathi K, Gomathinayagam S, Kumar SSHK, Ramalingam V.
[1] Park J, Law KH. Layout optimization for maximizing wind farm power pro-
Experimental and numerical studies on the wake behavior of a horizontal axis
duction using sequential convex programming. Appl Energy 2015;151(151):
wind turbine. J Wind Eng Ind Aerod 2014;128:54e65.
320e34.
[21] Sanderse B, Pijl VDS, Koren B. Review of computational fluid dynamics for
[2] Barthelmie RJ, Pryor SC. An overview of data for wake model evaluation in the
wind turbine wake aerodynamics. Wind Energy 2011;14(7):799e819.
virtual wakes laboratory. Appl Energy 2013;104:834e44.
[22] Kuo J, Rehman D, Romero DA, Amon CH. A novel wake model for wind farm
[3] Gao X, Yang H, Lu L. Optimization of wind turbine layout position in a wind
design on complex terrains. J Wind Eng Ind Aerod 2018;174:94e102.
farm using a newly-developed two-dimensional wake model. Appl Energy
[23] Crespo A, Hernandez J, Frandsen S. Survey of modelling methods for wind
2016;174:192e200.
turbine wakes and wind farms. Wind Energy 1999;2(1):1e24.
[4] Vermeer LJ, Sorensen JN, Crespo A. Wind turbine wake aerodynamics. Prog
[24] Ammara I, Leclerc C, Masson C. A viscous three-dimensional differential/
Aero Sci 2003;39(6):467e510.
actuator-disk method for the aerodynamic analysis of wind farms. J Solar
[5] Mo J, Choudhry A, Arjomandi M, Lee Y. Large eddy simulation of the wind
Energy Eng-Trans Asme 2002;124(4):345e56.
turbine wake characteristics in the numerical wind tunnel model. J Wind Eng
[25] Sorensen JN, Shen WZ. Numerical modeling of wind turbine wakes. J Fluids
Ind Aerod 2013;112:11e24.
€l F, Mann J, Ott S, Sørensen JN, et al. Eng-Trans Asme 2002;124(2):393e9.
[6] Larsen GC, Madsen Aagaard H, Bingo
[26] Shen WZ, Zhang JH, Sorensen JN. The actuator surface model: a new
Dynamic wake meandering modeling. Tech. Rep. Risø National Laboratory;
NaviereStokes based model for rotor computations. J Solar Energy Eng-Trans
2007.
Asme 2009;131(1):011002.
[7] Jensen NO. A note on wind turbine interaction. Tech. Rep. Risø National
[27] Rajagopalan RG, Klimas PC, Rickerl TL. Aerodynamic interference of vertical
Laboratory; 1983.
axis wind turbines. J Propul Power 1990;6(5):645e53.
[8] Barthelmie RJ, Larsen GC, Frandsen ST, Folkerts L, Rados K, Pryor SC, et al.
[28] Gao ZT, Li Y, Wang TG, Ke ST, Li DS. Recent improvements of actuator line-
Comparison of wake model simulations with offshore wind turbine wake
large-eddy simulation method for wind turbine wakes. Appl Math Mech
profiles measured by sodar. J Atmos Ocean Technol 2006;23(7):888e901.
2021;42(4):1e16.
[9] Chamorro LP, Porteagel F. A wind-tunnel investigation of wind-turbine wakes:
[29] Adedipe TA, Chaudhari A, Kauranne T. Impact of different forest densities on
boundary-layer turbulence effects. Boundary-Layer Meteorol 2009;132(1):
atmospheric boundary-layer development and wind-turbine wake. Wind
129e49.
Energy 2020;23(5):1165e80.
[10] Gaumond M, Rethore P, Ott S, Pena A, Bechmann A, Hansen KS. Evaluation of
[30] Meng H, Li L, Zhang J. A preliminary numerical study of the wake effects on
the wind direction uncertainty and its impact on wake modeling at the horns
the fatigue load for wind farm based on elastic actuator line model. Renew
rev offshore wind farm. Wind Energy 2014;17(8):1169e78.
Energy 2020;162:788e801.
[11] Bastankhah M, Porteagel F. A new analytical model for wind-turbine wakes.
[31] Liu W, Liu W, Zhang L, Sheng Q, Zhou B. A numerical model for wind turbine
Renew Energy 2014;70:116e23.
wakes based on the vortex filament method. Energy 2018;157:561e70.
[12] Ishihara T, Qian G. A new Gaussian-based analytical wake model for wind
[32] Dasari T, Wu Y, Liu Y, Hong J. Near-wake behaviour of a utility-scale wind
turbines considering ambient turbulence intensities and thrust coefficient
turbine. J Fluid Mech 2019;859:204e46.
effects. J Wind Eng Ind Aerod 2018;177:275e92.
[33] Beale JT, Majda AJ. Vortex methods. i. convergence in three dimensions. Math
[13] Sun H, Yang H. Study on an innovative three-dimensional wind turbine wake
Comput 1982;39(159):1e27.
model. Appl Energy 2018;226:483e93.
[34] Leonard A. Computing three-dimensional incompressible flows with vortex
[14] Gao X, Li B, Wang T, Sun H, Yang H, Zhao F. Investigation and validation of 3d
elements. Annu Rev Fluid Mech 1985;17(1):523e59.
wake model for horizontal-axis wind turbines based on filed measurements.
[35] Brady M, Leonard A, Pullin DI. Regularized vortex sheet evolution in three
Appl Energy 2020;260:114272.
dimensions. J Comput Phys 1998;146(2):520e45.
[15] Sun H, Yang H. Numerical investigation of the average wind speed of a single
[36] Carley M. A triangulated vortex method for the axisymmetric euler equations.
wind turbine and development of a novel three-dimensional multiple wind
J Comput Phys 2002;180(2):616e41.
turbine wake model. Renew Energy 2020;147:192e203.
[37] Bagai A, Leishman JG. Rotor free-wake modeling using a pseudo-implicit
[16] Keane A, Aguirre PEO, Ferchland H, Clive P, Gallacher D. An analytical model
techniquedincluding comparisons with experimental data. J Am Helicopter
for a full wind turbine wake. J Phys Conf 2016;753(3):032039.
Soc 1995;40(3):29e41.
[17] Masaru H, Yuji O, Hiroaki K. Numerical studies of flows around a wind turbine
[38] Li Q, Murata J, Endo M, Maeda T, Kamada Y. Experimental and numerical
equipped with a flanged diffuser shroud by using an actuator-disc model.
investigation of the effect of turbulent inflow on a horizontal axis wind tur-
J Web Eng 2006;73(108):335e8.
bine (part i: power performance). Energy 2016;113:713e22.
[18] Sedaghatizadeh N, Arjomandi M, Kelso R, Cazzolato B, Ghayesh MH. Modelling
[39] Li Q, Murata J, Endo M, Maeda T, Kamada Y. Experimental and numerical
of wind turbine wake using large eddy simulation. Renew Energy 2018;115:
investigation of the effect of turbulent inflow on a horizontal axis wind tur-
1166e76.
bine (part ii: wake characteristics). Energy 2016;113:1304e15.
[19] Uchida T, Yoshida T, Inui M, Taniyama Y. Doppler lidar investigations of wind
11