1. Introduction
The wake region forms behind the rotor when air flowing through the running wind turbine has a major effect on the safe operation of the wind turbines downstream. In particular, the variation of the turbulence intensity and the wind speed in this region are important factors to consider. The power loss of a large offshore wind farm caused by the wake effect can be as high as 10–20% [
1], so, the wake is an important factor that affects the safe operation of the wind turbine and benefits of the wind farm.
In the early stage of wake effect research, wind tunnel experiments and field measurements were used as the main experimental methods. An early wind tunnel experimental study was carried out at the National Aeronautical Research Institute (Swedish: Flygtekniska Försöksanstalten Abbreviated; FFA) and there were several more wind tunnel tests in recent years [
2]. Some important onshore field measurements can be found in Tjæreborg [
3], Nibe [
4] and Sexbierum [
5]. The most famous offshore wind farm measurements are from Horns Rev [
6] and Nysted. With the advancement of technology, advanced measurement methods are constantly appearing and the measurements of wind turbine wakes can be obtained by various methods, such as the supervisory control and data acquisition (SCADA) system, sodar and lidar [
7,
8,
9,
10]. But not all the data are public. It is difficult to solve the problem with the experimental method only, because it requires huge amounts of money and time but is unable to provide flow details, although the experimental method is credible.
With the rapid development of computer technology, this problem has been solved by the computational fluid dynamics (CFD) method. Simulation methods of the wind turbine wakes have been continually developing [
11]. The engineering “wake model” was established earlier and has a simple structure and short calculation time. The most famous Jensen model, which is also the basic model of the commonly used commercial wind resources software WAsP, was proposed to evaluate the wind turbine wakes on flat terrain [
12]. The eddy viscosity model appeared as another analytical wind turbine wakes model and was used in the commercial wind resources software WindFarmer belonging to the Garrad Hassan company [
13]. These models are based on assumptions of the flow behavior, such as axisymmetry and self-similarity. They work fast and are suitable for engineering applications, but they cannot predict wake accurately. The blade model can be constructed directly in the flow field and used to precisely calculate the load of the wind turbine and other aspects related to wind turbine performance [
14]. Qian Y.R. used this method to predict the air-load with different wind speeds under yaw conditions [
15]. However, this method needs more grids around the blade and a lot of computing resources; therefore, it does not apply to multiple wind turbines or wind farm scales. The actuator line model simplifies the blade into a line which is rotated as the blade, then constructs a functional relationship between the aerodynamic data of the blade’s local section airfoil and body force, so that it can simulate the interaction between the blade and the inflow [
16]. The actuator line model, combined with the large eddy simulation (LES) method, was used to investigate the near wake region and explain the reason why the velocity deficit and turbulence intensity increase nearby the rotor [
17]. Storey R.C. coupled the model with multi-physics method in order to research the wind turbine performance under transient conditions [
18]. The actuator line model also can demonstrate the wake of the wind turbine over complex terrain and reduce the cost of wind farm simulation to a certain extent [
19], but it still takes a lot of computing resources. The AD method can solve this problem by equating the rotor to a penetrable disc to greatly reduce the calculation time [
20].
Wind turbines work in the atmospheric boundary layer (ABL). So, not only the wind turbine but also the flow field in the ABL should be simulated well to improve the numerical accuracy of the wake. To some extent, the flow field can be precisely predicted in the ABL by using the LES method. Nilsson K. simulated the power production of the Lillgrund wind farm using the LES method [
21]. The wake of the horizontal and vertical axis wind turbine was imitated by Martínez-Tossas L.A. and Shamsoddin S. also using the LES method [
22,
23]. However, this takes a long time and requires more computing resources. The Reynolds Averaged Navier Stokes (RANS) method can balance the numerical accuracy of the wake simulation and the need for time and computing resources.
The standard
k-ε turbulence model is widely used in Computing Wind Engineering. In 2009, Europe organized a blind comparison based on the Bolund Island Wind Measurement Project which used high frequency data from 35 anemometers and provided a unique database designed to validate micro-scale flow models [
24]. With its great influence, more than 40 groups from around the world participated in the blind comparison with over 50 model predictions [
25,
26]. The results showed that the CFD method for solving the RANS equation had as smaller calculation error and was superior to other model results, thus, it was obviously superior to the traditional linear model. In this blind comparison, the ten results of the model with the smallest errors are shown, the
k-ε turbulence model accounts for more than half of them and showed higher accuracy [
27], so that the
k-ε turbulence model was found better for simulating in the ABL.
In summary, the wind turbine wakes can be studied by the AD method combined with the
k-ε turbulence model. The combined method has been used by the commercial codes WindSim and Fuga as their basic CFD wake model [
28,
29,
30]. The flat topographic wind resources have almost been exhausted, wind power projects in complex terrain and offshore have become the mainstream of development. The flow field of wind turbines was simulated by using the combined CFD and AD method in complex terrain [
31]. Along with the CFD method, experiments were taken to investigate the performance of the wind farm operation in complex terrain [
32]. The interaction of multiple wind turbine wakes has also been studied [
33]. Three
k-ε turbulence models were used to evaluate the wind energy for a highly complex terrain by Dhunny A.Z. [
34]. The interaction of atmospheric flow, wind turbine wakes and ground surface has been taken into consideration under different situations [
35]. The wakes and power of the offshore wind farm can also be investigated by using the combined CFD and AD method [
36]. The standard
k-ε turbulence model overestimates the wake recovery, so a
k-ε model which include a parameter sensitive to high velocity gradients was established to improve the simulation accuracy [
37,
38]. The purpose of adding the dissipation rate source term into the
k-ε turbulence model by Kasmi A.E. was to solve this problem [
39]. The improved model is also used for optimizing the layout of the wind farm [
40].
In order to investigate the wake accurately, a modified k-ε turbulence model is proposed in which both the turbulent kinetic energy source term and the dissipation rate source term are added. The major improvement is that it coordinates the generation and dissipation of turbulent kinetic energy better than only adding the dissipation rate source term. The parameter C4ε, which obeys a parabolic distribution, is used based on a theoretical analysis. The body force distribution on the AD is also used, instead of a constant value which is used in the classical AD method. The results are consistent, with the measurements, and better reflect the relative velocity distribution of the wake.
2. Actuator Disc Model
The wind turbine is represented by a simplified AD method; it equates the wind turbine to a penetrable disc, the air is free to pass through, but encounters some drag [
41]. To some extent, the effect of the rotor on the inflow can be reflected by this hypothetical model. In this paper, the AD method is derived from one-dimensional momentum theory (
Figure 1).
It is assumed that the inflow velocity is
u1, and the axial thrust of the rotor is:
where
ρ is the air density and
CT is the thrust coefficient. Therefore, the thrust of the AD in equations is:
where
Su is the source term of the rotor, ∆x is the thickness of the AD.
The thrust is mainly determined by the thrust coefficient and inflow velocity. It is easy to obtain the inflow velocity with a single wind turbine. However, it becomes difficult to determine in the case of multiple wind turbines, so changes need to be made as follows:
where
a is the induction factor,
uD is the wind speed at the AD [
42]. The Equation (3) is not consistent with the experiment when the value of
a is too large, so Glauert’s correction and Shen’s correction are popular countermeasures [
43,
44]. The Equation (3) will have non-physical rationality when
a > 0.5, in the present model, the parameter
CT represents the actual performance of the rotor, so the maximum value of
a calculated by
CT will not exceed 0.5.
The momentum source term of the rotor is changed as follows:
where
Cx = 4
a/(1−
a) is the drag coefficient of the AD. So, the momentum source term is linked to the wind speed on the AD and it is more suitable for predicting the flow.
The nacelle can also be simplified as a part of the AD and its drag to the incoming flow can be calculated as the flow drag of the solid. The thrust on the flow is:
where
CD is the drag coefficient of the nacelle, which varies from 0.8 to 1.2 [
45]. In this paper,
CD takes a value of 1.0.