Aerodynamic Effect of Winglet on NREL Phase VI Wind Turbine Blade

Abstract

The primary goal in designing wind turbine blades is to maximize aerodynamic efficiency. One promising approach to achieve this is by modifying the blade geometry, with winglets to the tip. Winglets are intended to reduce the strength of the tip vortices, thereby reducing induced drag, increasing torque, and, ultimately, improving the power output of the wind turbines. In this study, computational fluid dynamics (CFD) simulations were utilized to assess the aerodynamic performance of wind turbine blades with and without winglets at various wind speeds (5, 7, 10, 13, 15, 20, and 25 m/s). The results indicate that winglets have a limited effect at low (5 and 7 m/s) and high (20 and 25 m/s) wind speeds due to fully attached and separated flows over the blade surface. However, within the 10–15 m/s range, winglets significantly enhance torque and power output. While this increased power generation is beneficial, it is essential to consider the potential impact of the associated increase in thrust force on turbine stability.

1. Introduction

Wind energy has emerged as a crucial renewable energy source, offering a sustainable and clean alternative to fossil fuels. Its potential for significant contribution is undeniable as the world seeks low-carbon energy solutions. Modern wind turbines harness the kinetic energy from wind and converts it into electricity through a comprehensive system involving rotor blades, a gearbox, and a generator. This method also faces great challenges including cost effectiveness, the use of remote locations, and environmental impacts. The goal in terms of turbine design is to enhance wind turbine efficiency by exploring modifications to their configuration, size, and capacity. A key focus of ongoing research is to enhance the aerodynamic performance of turbine blades.

Winglets, first introduced by NASA [1], have proven to significantly enhance airplanes’ efficiency by reducing wingtip drag. Similarly, the purpose of adding winglets to wind turbine blades is to decrease the induced drag by diverting downwash distribution and thereby increasing power production. Decades of research has explored the application of winglets to enhance wind turbine performance. Numerous studies, both experimental and numerical, have demonstrated the potential benefits of winglets in improving aerodynamic efficiency and thereby increasing power output.

Johansen and Sørensen [2] investigated five winglet designs using CFD and found that adding a winglet can effectively modify the downwash distribution of airflow and increase the power output. Gaunaa and Johansen [3] further demonstrated the superiority of downwind winglets in optimizing the power coefficient (Cp) and the effectiveness of small winglets in increasing Cp. Gertz et al. [4] experimentally verified the positive impact of winglets on power production, reporting increases of 5–8% under varying wind speeds between 6.5 m/s and 9.5 m/s. Ali et al. [5] investigated the effect of winglets on lift-to-drag ratios for a domestic-scale turbine model, finding that upwind winglets improve this ratio, while downwind winglets decrease it. They also highlighted the influence of yaw angle on aerodynamic performance. Tobin et al. [6] investigated a model wind turbine with and without winglets with wind tunnel experiments and achieved 8.2% and 15.0% increases in power and thruster coefficients for turbines with winglets. Khalafallah et al. [7] performed a parametric study of the power output associated with winglets in two orientations with varying cant and twist angles. A maximum increase of 4.39% in the power coefficient was achieved with a winglet with 40° cant and 10° twist angles. Kulak et al. [8] investigated the impact of winglets on small-scale wind turbine blades using both CFD and experimental measurements. Contrary to the early findings of Gaunaa and Johansen [3], their study indicated that pressure-side winglets lead to a significant increase in power output. This discrepancy may be attributed to differences in turbine design, winglet geometry, or flow conditions.

More CFD simulations investigating the potential of winglets to enhance wind turbine performance can be found in the literature. Elfarra et al. [9] introduced an optimized winglet, which achieved a 9% increase in power generation, where CFD solving RANS with the k-ε turbulence model was used for investigation. Hansen and Muhle [10] conducted CFD-based winglet optimization for model horizontal-axis wind turbines (HAWTs) after validation with wind tunnel measurements. Optimized winglets resulted in an approximately 8% increase in power generation. Khaled et al. [11] employed CFD simulations with ANN-aided optimizations to validate the positive effects of winglets on small horizontal-axis wind turbines, observing noticeable enhancements in power and thrust coefficients. Mourad et al. [12] conducted a numerical study on the effect of winglets on a small-scale HAWT and achieved a 6% increase in power coefficient. Garcia-Ribeiro et al. [13] performed a CFD-based parametric analysis on the taper ratio effect of winglets on the New Mexico wind turbine model, and higher efficiency was achieved using a winglet with a lower taper ratio and root chord ratio. Abdelghany et al. [14] investigated the effect of winglet length on the aerodynamic performance of HAWTs. At a fixed cant angle of 90°, the optimum winglet height/length ratio was identified to be about 0.042, resulting in a 6.9% increase in power coefficient.

Winglets have also been explored for vertical-axis wind turbines (VAWTs) using CFD simulations. Studies by Zhang et al. [15], Xu et al. [16], and Dol et al. [17] have shown promising results in terms of increased power output.

Despite extensive research, accurately predicting the impact of winglets on wind turbine performance remains challenging due to the complex interplay of factors such as varying blade geometry, flow conditions, and turbulence modeling. On the other hand, wind tunnel experiments are limited by scale and cost constraints. Therefore, we extended previous studies [18,19] on the reference NREL Phase VI test wind turbine and conducted CFD analyses of tip modification with winglets, focusing on accurately capturing the complex flow interactions and aerodynamic effects. It is worth noting that Menter et al. [20,21] presented the CFD results of flow around the NREL Phase VI turbine blade while reviewing the applications of the Shear Stress Transport (SST) turbulent model. Mo and Lee [22,23] conducted a CFD investigation on the aerodynamic characteristics of the NREL Phase VI wind turbine. More recent studies focused on the same model turbine have been reported by Verma et al. [24], Zhang et al. [25], and Dejene et al. [26].

The remainder of this article is organized as follows. Section 2 provides specifications of the baseline wind turbine model and describes the methodology used in this analysis, including geometry modifications, meshes, boundary conditions, and model setups. Section 3 presents the comprehensive results of the CFD simulation with comparisons to experimental measurements. Finally, the conclusions from this study are presented in Section 4.

2. Methodology

This study utilized ANSYS CFX (version 17.0) to conduct CFD simulations to investigate the aerodynamic effects of winglets added to the NREL baseline wind turbine blade. The analysis was divided into three primary stages: preprocessing, simulation, and postprocessing. In the preprocessing phase, ANSYS DesignModeler was utilized to create 3D geometries of both the baseline blade and the blade with winglets. Subsequently, the comprehensive computational domains including the blade surfaces were meshed. A refined mesh was generated near the blade surface to ensure accuracy in boundary layer prediction, with the y⁺ value maintained close to one. During the simulation phase, the ANSYS CFX solver was employed to calculate the flow field surrounding the blades. The SST turbulent model was selected due to its demonstrated efficacy in accurately predicting flow conditions, especially for near-wall flow problems. The calculated results were finally analyzed and visualized using the ANSYS built-in postprocessing tools. This enabled a detailed examination of the aerodynamic performance and the effects of winglets on the turbine blades.

2.1. NREL Phase VI Wind Turbine

The present numerical investigation was based on the NREL Phase VI wind turbine. Figure 1 provides the schematic setup of this baseline wind turbine [27,28,29]. It is a stall-regulated turbine with full-span pitch control and a power rating of 20 kW. The turbine’s blades are both twisted and tapered. The taper begins at 25% of the blade span from the hub, with the chord length of the airfoil decreasing from 0.737 m at 25% span to 0.356 m at the tip. Figure 2 shows the twist angle variation, starting at 22.1° at 25% span and decreasing to −2.0° at the tip. The blade incorporates an NREL S809 airfoil section from the 25% span to the tip. This airfoil section connects to the pitch shaft section at 12% span, with linear segments ensuring a smooth transition between two incongruent contours. The blade pitch axis is located at 30% of the cord length from the leading edge and is centered between the upper and lower surfaces of the blade. The rotor operates at a nominal 72 RPM.

2.2. Aerodynamics

Wind turbines harness the energy of wind through the interactions of their rotor blades with the air. Modern turbine blades, such as the NREL reference model, utilize airfoils to convert wind kinetic energy into electricity. Optimizing blade design for lift and drag is essential to maximize energy output. While wind turbine blades share aerodynamic principles with airplane airfoils, their design differs in that the wind turbines remain stationary. At a given operating condition, a wind turbine’s blades rotate at a constant speed.

Figure 3 illustrates the velocities and force coefficients for a wind turbine airfoil in the 2D plane, where α is the angle of attack (AOA), C_L is the lift force coefficient, C_DP is the pressure drag force coefficient, V_w is the free stream wind speed, V_rot is the blade (rotor) rotational speed, and V_r is the resultant velocity. V_r is also known as the relative wind velocity. Other important force coefficients for an airfoil include the tangential force coefficient C_T, which acts along the chord line, and the normal force coefficient C_N, which acts perpendicular to the chord line toward the suction side. These forces contribute to the toque coefficient C_TQ and the thrust coefficient C_TH in the 2D plane of rotation [27].

2.3. Computational Domains

The computational domain consisted of two regions: a fixed rectangular domain and a rotating cylindrical domain, as illustrated in Figure 4. The rectangular domain, measuring 20 m × 30 m × 20 m, represented the ambient air flow. The cylindrical domain, with a diameter of 12 m and a length of 18 m, enclosed the wind turbine blade.

Inlet boundary conditions were applied to the front face of the rectangular domain, specifying a uniform velocity of 5, 7, 10, 13, 15, 20, or 25 m/s, depending on the simulation case. A pressure outlet boundary condition was applied to the opposite face, maintaining a pressure of 1 atm. Symmetry boundary conditions were applied to the remaining four faces of the rectangular domain. The wind turbine blade was treated as a no-slip wall boundary condition. A rotational velocity of 72 RPM was assigned to the cylindrical domain to simulate the rotor’s motion. The NREL Phase VI wind turbine blade design specification and optimized domain dimensions were taken from previous studies [18].

2.4. Geometry of Blades

The wind turbine blade in this study was modeled using ANSYS DesignModeler. As shown in Figure 5, the twist angles of the sixteen sections along the blade were kept identical to the original NREL Phase VI blade. To investigate the impact of winglets on aerodynamic performance, three winglets with various lengths were added at the 98.2% spanwise location. Each winglet had a fixed inclination angle of 120° (cant angle of 60°) and a 30% taper ratio. A curvature radius of 0.07 m was used to smooth the edge.

It can be seen that Winglet Design-1 (WD-1), Winglet Design-2 (WD-2), and Winglet Design-3 (WD-3) differed only in their lengths. WD-1 had an inclined height of 0.0981 m and a vertical height of 0.12 m. WD-2 was longer, with an inclined height of 0.14 m and a vertical height of 0.156 m. WD-3 was shorter, with an inclined height of 0.06 m and a vertical height of 0.08692 m.

2.5. Mesh Generation

ANSYS Mesh was used to generate the mesh, as shown in Figure 6. Grid sensitivity studies were performed for all three winglet designs. The numbers of grids used were 35.7, 29.3, and 29.4 million, respectively. The number of sublayers used within the inflation layer were 16, 20, and 22, with the height of the first sublayer as 1 × 10⁻⁶ m to satisfy the y⁺ plus value of less than 1. The y⁺ was a dimensionless parameter that measured the distance from the first grid cell to the nearest wall as

$$y^{+}=y·{\displaystyle \frac{u_{*}}{\upsilon }}$$

(1)

where y is the distance from the wall to the nearest grid point and $\upsilon$ is the kinematic viscosity.

For WD-1, the first sub-layer height within the inflation layer was 5 × 10⁻⁶ m, with 16 inflation sub-layers. The face element size of the winglet was 1 × 10⁻² m, while the element sizes of the rectangular and cylindrical domains were 0.4 m and 0.2 m, respectively. Approximately 35.7 million grids were generated for the entire blade domain. The y⁺ value varied from 2.77 at 25 m/s wind speed to 7.72 at 5 m/s wind speed, decreasing with higher wind speeds. For WD-2, the first sub-layer height within the inflation layer was 1 × 10⁻⁶ m with 22, resulting in a y⁺ value below 1. The total number of elements was 29.3 million. A similar approach was used for the WD-3 design, with 20 inflation sub-layers and a first sub-layer height of 1 × 10⁻⁶ m, resulting in 29.4 million grids and a y⁺ value below 1. To achieve the desired y⁺ value, 10 simulation steps were run after each grid generation to evaluate the y⁺ value using the post-processor. Figure 7 illustrates the section-wise mesh generation for the blades.

2.6. Numerical Method

The CFD simulation was carried out using ANSYS CFX, solving the Reynolds-Averaged Navier–Stokes (RANS) equations with the SST turbulence model. The interface between the fixed and rotating computational domains was treated as continuous. The rotating domain was set to rotate at 72 RPM. A high-resolution solver control criterion was employed to ensure accuracy, and the simulation was performed in a steady-state condition.

Wind speeds (5, 7, 10, 13, 15, 20, and 25 m/s) were specified as a uniform velocity at the inlet, depending on the simulation case. To facilitate comparison with NREL experimental measurements, the wind turbine rotor blade span was divided into five sections at 30%, 46.7%, 63.3%, 80%, and 95% spanwise locations. The results for the three blades with winglets and the original NREL Phase VI wind turbine blade were visualized and analyzed using the post-processor in ANSYS.

2.7. SST Turbulence Model

The SST turbulence model is a hybrid model that combines the strengths of the k-ω and the k-ε models. Within the boundary layer, where viscous effects dominate, the model switches to the k-ω model, which provides accurate predictions of near-wall flow. In the free-stream region, where inertial effects are more significant, the model transitions to the k-ε model, which is better suited for capturing large-scale turbulent structures. The SST model is briefly represented as follows [30,31]:

$${\displaystyle \frac{\partial \left(\rho k\right)}{\partial t}}+{\displaystyle \frac{\partial \left(\rho U_{i}k\right)}{\partial x_i}}={\tilde{P}}_k-{\beta }^{*}\rho k\omega +{\displaystyle \frac{\partial }{\partial x_i}}\left[(\mu +{\sigma }_k{\mu }_t){\displaystyle \frac{\partial k}{\partial x_i}}\right]$$

(2)

$${\displaystyle \frac{\partial \left(\rho \omega \right)}{\partial t}}+{\displaystyle \frac{\partial \left(\rho U_i\omega \right)}{\partial x_i}}=\alpha {\displaystyle \frac{1}{{\nu }_t}}{\tilde{P}}_k-\beta \rho {\omega }^2+{\displaystyle \frac{\partial }{\partial x_i}}\left[\left(\mu +{\sigma }_{\omega }{\mu }_t\right){\displaystyle \frac{\partial \omega }{\partial x_i}}\right]+2\left(1-F_1\right)\rho {\sigma }_{\omega 2}{\displaystyle \frac{1}{\omega }}{\displaystyle \frac{\partial k}{\partial x_i}}{\displaystyle \frac{\partial \omega }{\partial x_i}}$$

(3)

Here,

$${\nu }_t={\displaystyle \frac{a_{1}k}{\mathit{max}\left(a_1\omega , S{F}_2\right)}}$$

$$S=\sqrt{2S_{ij}{S}_{ij}}$$

$${\tilde{P}}_k=\mathit{min}\left(P_k, 10·{\beta }^{*}\rho k\omega \right) \mathrm{for} P_k={\mu }_t{\displaystyle \frac{\partial U_i}{\partial x_j}}\left({\displaystyle \frac{\partial U_i}{\partial x_j}}+{\displaystyle \frac{\partial U_j}{\partial x_i}}\right)$$

$$F_1=tanh\left\{{\left\{min\left[max\left({\displaystyle \frac{\sqrt{k}}{{\beta }^{*}\omega y}},{\displaystyle \frac{500\nu }{y^2\omega }}\right),{\displaystyle \frac{4\rho {\sigma }_{\omega 2}k}{C{D}_{k\omega }{y}^2}}\right]\right\}}^4\right\}$$

$$F_2=tanh\left\{{\left[max\left({\displaystyle \frac{2\sqrt{k}}{{\beta }^{*}\omega y}},{\displaystyle \frac{500\nu }{y^2\omega }}\right)\right]}^2\right\}$$

$$C{D}_{k\omega }=max\left(2\rho {\sigma }_{\omega 2}{\displaystyle \frac{1}{\omega }}{\displaystyle \frac{\partial k}{\partial x_i}}{\displaystyle \frac{\partial \omega }{\partial x_i}},{10}^{-10}\right)$$

In this model, all constants are computed by a blend of the corresponding constants from the k-ω and k-ε models. Any constant ϕ can be determined by

$$\varphi =F_1{\varphi }_1+\left(1-F_1\right){\varphi }_2$$

The constants for this SST model were

$${\alpha }_1={\displaystyle \frac{5}{9}},{ \alpha }_2=0.44,$$

$${\sigma }_{k1}=0.85, { \sigma }_{\omega 1}=0.5, { \beta }_1=0.075,$$

$${\sigma }_{k2}=1.0, { \sigma }_{\omega 2}=0.856, { \beta }_2=0.0828,$$

$${\beta }^{*}=0.09, a_1=0.31.$$

3. Results and Discussion

CFD simulations were conducted on the original NREL blade and the modified blades with different winglet designs (WD-1, WD-2, and WD-3). CFD calculations were compared to NREL experimental measurements. To provide a comprehensive comparison, the blade span was divided into five sections (30%, 46.7%, 63.3%, 80%, and 95%) for a detailed analysis. Key aerodynamic characteristics, including pressure and velocity coefficients, flow contours, pressure coefficient distribution, tangential and normal force coefficient curves, and torque and thrust forces, were examined.

3.1. Variation of AOA over the Blade Span

Previous studies have focused on the S809 airfoil to investigate the complex flow patterns around wind turbine blades under a wide range of angles of attack, including the stall region and the variation of angle of attack along the blade span. Understanding these flow phenomena is crucial for optimizing blade design and predicting performance under various operating conditions. Figure 8 plots the lift and drag coefficients at different angles of attack, while Figure 9 illustrates the spanwise angle of attack distribution under various wind speeds. In both graphs, the dynamic stall region is highlighted, indicating where the fluid begins to separate from the blade surface.

3.2. Pressure Coefficient

The pressure coefficient (C_p) is a dimensionless number that represents the ratio of static and free stream pressure difference to dynamic pressure. It depends on the AOA, relative wind speed, and pressure at each section. Negative C_p values were expected on the suction side of the blade.

Figure 10 shows the pressure coefficient contours for the blades with winglets and the original NREL blade at a wind speed of 5 m/s. At this wind speed, the entire blade remained in the attached flow region, with a small range of AOA. As a result, a uniformly distributed pressure coefficient was observed at each section for all four blades.

Similar observations were made at a wind speed of 7 m/s, as the entire blade remained in the transition region (pre-stall). However, a slightly wider low-pressure coefficient region was observed compared to the 5 m/s case, as shown in Figure 11. The pressure distribution remained relatively uniform across all five sections. Meanwhile, a stronger low-pressure region was observed at the suction side trailing edge for the blade modified with WD-1 and the original NREL blade.

Figure 12 compares the pressure coefficient contours for the three winglet designs and the original NREL blade at a wind speed of 10 m/s. The 30% span section remained in the deep stall region, with a larger low-pressure coefficient on the suction surface and a weaker high-pressure coefficient on the leading-edge pressure side. All winglet cases exhibited a larger low-pressure zone. The blade sections at 46.7% and 63.3% span were in the dynamic stall region (20° AOA). The 80% and 95% blade sections were in the pre-stall region. In general, the winglets cases had larger low-pressure zones. At 95% span, the winglet cases, especially WD-3, showed a larger low-pressure zone, indicating more attached flow. This suggested a higher total pressure force, potentially leading to increased torque and power.

At a wind speed of 13 m/s, approximately 73% of the blade span was in the deep stall region, with the remainder in the dynamic stall region. Figure 13 illustrates the pressure coefficient contour for this wind speed. Compared to the original NREL blade at the 30% span section, the blades with winglets exhibited smaller low-pressure regions. WD-1 had a stronger low-pressure region at the suction side of the 46.7% span section. A weak low-pressure zone was observed for all cases at the 63.3% span section. At 80% and 95% span, the sections remained in the dynamic stall region, with larger low-pressure coefficient zones for the winglet designs. This suggested that flow attachment was enhanced with the winglet designs, particularly WD-3.

Figure 14 illustrates the pressure coefficient contour at a wind speed of 15 m/s. Approximately 90% of the blade span from the hub remained in the deep stall region, with the rest in the dynamic stall region. Consequently, the flow was completely separated from the blade up to a span of 90%. Due to the higher AOA, the most intense low-pressure region was observed at the 30% span, and larger low-pressure regions were seen for WD-1 and the original NREL blade. Similar to the 13 m/s case, a sudden drop in the pressure coefficient intensity occurred at 63.3% span, with larger low-pressure regions for the blades with winglets. The low-pressure intensity decreased up to 80% span for all blade shapes. However, at 95% span, a stronger low-pressure region was observed, with significantly larger zones for all blades with winglets, especially WD-3.

At the wind speeds of 20 m/s and 25 m/s, the entire blade span was in the dynamic stall region and the flow was fully separated. So, the pressure coefficient contours for the original NREL blade and the modified blades with winglets were nearly identical. A high-intensity low-pressure zone was observed at the 30% span for both wind speed cases, indicating high AOAs. The intensity of low-pressure zones decreased along the spanwise direction, and similar pressure coefficient distributions were observed at each span section for both wind speeds. Figure 15 and Figure 16 show the pressure coefficient contours for all blades. The addition of winglets had a negligible effect on pressure coefficient contours at high wind speeds.

3.3. Velocity Contour

Flow attachment and separation could be visualized using velocity coefficient contours. The velocity coefficient was defined as the ratio of the velocity of the stationary frame to the relative velocity. The relative velocity of the air at the blade was given by

$$V_r=\sqrt{V_{rot}^2+V_w^2}$$

(4)

Relative velocity varied across different blade sections, influenced by pitch angle and wind speed. Figure 9 highlights the region of potential flow separation in red.

The primary goal of adding a winglet to the blade was to minimize the flow separation and reduce tip vortex strength. Improved flow attachment can positively impact rotor torque. This subsection compares the velocity coefficient contours at five blade span locations under different wind speeds for the original NREL blade and the three blades modified with winglets.

At a wind speed of 5 m/s, the AOA range was small, and the entire blade span remained in the attached flow region. Figure 17 shows the velocity contours of all blades at different span sections. The flows remained predominantly attached, with minimal differences between the original NREL and the other three winglet designs.

Figure 18 compares the velocity contours for the four blades at a wind speed of 7 m/s. At this wind speed, the entire blade span stayed in the transition region, indicating potential flow transitions. Figure 18 reveals a transitional effect at the suction surface trailing edge of the 30.0% span section only for WD-1. From the 46.7% span section to the remaining sections, a suction side trailing edge transition was observed for both cases, but it was slightly stronger for WD-1. Similar results were observed for the three blades with winglets, but the transitional effect was more significant for WD-3, which could negatively impact torque.

At a wind speed of 10 m/s, differences in the velocity contours emerged between the original NREL blade and the WD-1 blade. As shown in Figure 9, the blade span passed through three regions: deep stall (approximately up to 46.7% span), dynamic stall (46.7% to 73% span), and transition (remaining portion of the blade).

The most significant differences in the velocity contours occurred in the dynamic stall region due to the addition of winglets. Figure 19 compares the velocity contours for these four types of blades at 10 m/s. At the 30% span section, WD-2 and WD-3 had smaller separations on the suction side toward the trailing edge compared to the original NREL blade. At the 46.7% and 63.3% span sections, WD-1 and the original blade exhibited stronger trailing edge separations, while WD-2 and WD-3 had minimal separations. Since the 80% and 95% span sections were in the pre-stall region, flow attachment was generally observed for all cases. However, WD-1 and the original blade showed some transitions at 89% span. The large low-pressure zones observed for the blades with winglets in the relevant pressure contours aligned with the increased flow attachment.

Figure 20 shows the velocity contours of the blades under a wind speed of 13 m/s. Approximately 73% of the blade span from the hub stayed in the deep stall region, with the remainder in the dynamic stall region. Significant variations were expected for the 80.0% and 95.0% span sections. Comparing the original blade to the WD-1 blade, the 30% span section showed weaker flow separation on the suction surface for WD-1. However, at the 46.7% and 63.3% span sections, flow separation was stronger for the blade with the winglet. Conversely, the blade with the winglet exhibited more flow attachment at both the 80% and 95% span sections.

Among the three blades with winglets, up to 63.3% span, WD-2 and WD-3 exhibited weaker flow separations on the suction side than WD-1. At the 80% span, WD-3 showed more flow attachment. At the 95% span, the flow remained attached at the suction side for all winglet cases. These observations aligned with the larger low-pressure regions for all winglet cases, which could potentially impact torque output for these blades.

Figure 21 shows the velocity contours for these blades under a wind speed of 15 m/s, with observations like those at 13 m/s. Approximately 90% of the blade span was in the deep stall region, with the remainder in the dynamic stall region. At the 30% span section, a thin layer of flow separation was observed for all blades. Up to 80% span section, similar strong flow separations were observed for the original blade and the winglet designs. However, at the 95% span section, the winglet cases exhibited significantly weaker flow separations at the suction side near the trailing edge compared to the original blade. Among the three winglet designs, WD-1 and WD-2 showed weak separations, while WD-3 had mostly attached flow. These findings aligned with the pressure contours and suggested that the winglet designs could potentially improve torque output.

At a wind speed of 20 m/s, the entire blade span remained in the deep stall region, with fully separated flow. Figure 22 compares the velocity contours of these blades at this speed. At the 30% span section, the blades with winglets exhibited more flow separation. However, the velocity contours for the remaining sections were nearly identical across all blade designs. This suggested that the effect of adding winglets was negligible at a wind speed of 20 m/s.

Figure 23 illustrates the velocity contours for these blades under a wind speed of 25 m/s. Similar to the 20 m/s case, the entire blade span remained in the dynamic stall region, with fully separated flow. Consequently, the addition of winglets had a limited impact at high wind speeds.

3.4. Flow Conditions

Velocity contours over the whole domain were used to investigate the wake formation for all four blades under different wind speeds. The velocities on the horizontal plane at the hub height of the wind turbine were plotted to visualize the flow conditions downstream of the turbine. The wake formation indicated the potential power extraction from the wind. Irregularities in the wake could influence blade torque, and the flow conditions downstream of the blade could provide insights into the torque comparison between different blade designs. Figure 24, Figure 25, Figure 26 and Figure 27 compare the wake formations for the baseline blade and three modified blades with different winglets.

At a wind speed of 5 m/s, the flow remained fully attached in all cases. Large wakes were generated as the air passed over the blades and a uniform flow pattern was observed.

At 7 m/s, the blades were in the transition region (pre-stall). The flow pattern was nearly uniform near the tips and wake formation began further downstream compared to that with a wind speed of 5 m/s.

Differences emerged among the four blades at 10 m/s, with wake formation starting further downstream. Two symmetric wake zones formed behind the turbine. The baseline blade generated two strong wakes that gradually weakened downstream; WD-1 and WD-3 had smaller wake formations, potentially impacting torque. WD-2 exhibited larger and stronger wake formations.

At a wind speed of 13 m/s, a stagnation point was observed at the suction surface of all blades. Another stagnation point near the tip was visible for the winglet cases, indicating that the winglets influenced wake formation. The wakes for the winglet cases were closer together compared to the baseline blade, which had a relatively weaker wake.

Similar observations held for the wind speed of 15 m/s, with larger flow separations and denser wakes observed for the blades with winglets compared to the results of the baseline blade.

At 20 m/s, the flow was fully separated from the blade suction surface, and less wake generation was observed for all the cases. This reduced wake generation suggested lower torque values for the blades.

At 25 m/s, even greater flow separation occurred on the blade suction side for all cases. This resulted in fewer and weaker wakes that were farther from the blade.

3.5. Chordwise Pressure Distribution

To further understand the aerodynamic effect of adding winglets, chordwise pressure distributions around the blades were determined. Pressure coefficients (C_p) were plotted as a function of non-dimensional distance along the chord (x/c), where x is the distance along the chord and c is the chord length. The x/c ratio value ranged from 0 (leading edge) to 1 (trailing edge). Negative C_p values, indicating suction pressure, are plotted on the upper part of the graph, while positive C_p values (pressure side) are on the lower part. Figure 28, Figure 29, Figure 30, Figure 31, Figure 32, Figure 33 and Figure 34 show the chordwise pressure distributions for selected span sections (r/R) across different blades and wind speeds.

At a wind speed of 5 m/s, as shown in Figure 28, minimal variations were observed among the different blade designs. The CFD simulations for all four blades yielded nearly identical pressure distributions, which were in good agreement with NREL’s experimental measurements. A slight discrepancy in the upper curve was noted between the WD-1 and the other blades for chord ratios between 0.4 and 0.5. Some discrepancies were also observed at the trailing edge for all the span locations. The negative C_p values exceeded −1 in all cases and C_p distributions across the span were similar at this wind speed.

At a wind speed of 7 m/s, the entire blade remained in the transition range of the pre-stall region. The C_p distributions from the CFD simulations of all four blades were in reasonably good agreement with the experimental results for the baseline blade. Minor discrepancies were observed at the trailing edge for different tip modifications. Notably, WD-2 and WD-3 exhibited lower C_p values on the pressure surface (lower part) near the trailing edge compared to WD-1 and the baseline blade.

At a wind speed of 10 m/s, as shown in Figure 9, the blade was expected to experience three flow regions: deep stall (up to nearly 46.7% span), dynamic stall (46.7% to 73% span), and transition (remaining span).

Figure 30 shows fairly good agreement between the CFD simulations and experimental measurements for most sections (30.0%, 63.3%, 80.0%, and 95.0% radial distance, r/R). However, a significant discrepancy occurred in the 46.7% section. The experimental measurements indicated a deep stall region, while the CFD simulations suggested a different flow scenario, possibly due to the limitations in turbulence models predicting stall delay.

Overall, the blades with winglets exhibited higher negative C_p values compared to the baseline blade at all sections. WD-3 had the largest negative C_p values over most of the chord, except near the trailing edge, where the blades of WD-2 and WD-3 experienced smaller negative C_p values compared to the baseline blade and WD-1.

Figure 9 indicates that at the wind speed of 13 m/s, approximately 73% of the blade span from the hub was in the deep stall region, with the remaining portion in the dynamic stall region. Flow separation was expected at the 73% section, where a 20° AOA occurred. Figure 31 shows good agreement between the CFD calculations and experimental measurements for most sections, except for the 30% section on the suction side. The numerical calculations overpredicted the negative C_p from the leading edge to the 0.1 chord ratio, while significantly underpredicting C_p in the chord range from 0.1 to 0.8. The blades with winglets exhibited significantly larger negative C_p values at the 80% and 95% radial sections. WD-1 had the largest negative C_p value at the trailing edge. At the 80% radial section, WD-3 had the largest negative C_p from the leading edge to the 0.8 chord. Overall, WD-3 showed relatively larger negative C_p values on the suction sides for all radial sections at this wind speed.

At a wind speed of 15 m/s, approximately 90% of the blade span was in the deep stall region, with the remaining portion in the dynamic stall region. Figure 32 shows good agreement between the CFD simulations and experimental measurements, with significant discrepancies only at the 30% and 63.3% radial sections. At the 30% radial section, the numerical calculation overpredicted the negative C_p on the suction side from the leading edge to 0.2 chord ratio, while underpredicting C_p from 0.3 to the trailing edge. At the 63.3% radial section, CFD underpredicted the negative C_p on the suction side across the entire chord length. The effect of adding winglets was still evident near the tailing edge at this speed. Significant differences among the blades were observed in the C_p plots at the 30% and 95% radial sections. At 30%, WD-3 had the largest negative C_p on the suction side from the leading edge to 0.16 x/c but the smallest negative C_p from 0.16 to 0.6 x/c. Near the tip (95% radial section), the blades with winglets had larger negative C_p values on the suction side compared to the baseline blade, with WD-3 having the largest negative C_p value.

At wind speeds of 20 and 25 m/s, flow separations occurred across the entire blade span, placing the entire blade in the deep stall region. CFD results underpredicted negative C_p values closer to the hub (30% and 46.7% radial sections, as shown in Figure 33 and Figure 34). However, the overall agreement between the CFD and experiment results was reasonable. The closer toward the hub, the larger the negative C_p on the suction side, as observed at the 30% section. While CFD results suggested a slight difference near the trailing edge, the effect of winglets on negative C_p was negligible for most of the chord length.

3.6. Tangential and Normal Forces

Wind turbine blades experience varying pressures as they interact with wind, resulting in a resultant aerodynamic force. The force can be divided into tangential (F_T) and normal (F_N) components, as shown in Figure 4. F_T is the projection of the resultant force along the chord line, while F_N is that perpendicular to the chord. These forces are often analyzed in their non-dimensional forms: the tangential force coefficient (C_T) and the normal force coefficient (C_N).

At a wind speed of 5 m/s, the AOAs across the span were small and relatively uniform. Figure 35 shows little difference between the CFD calculations and experimental measurements, with the former slightly underpredicting the forces. The blades with winglets experienced slightly lower forces than the baseline blade, but the difference was negligible.

At a wind speed of 7 m/s, the AOAs increased across the span, but the entire blade remained in the pre-stall region. Figure 36 shows the tangential and normal force coefficient distribution at this speed. The CFD and experimental results were in good agreement, with the blades with winglets experiencing slightly lower tangential and normal forces.

At a wind speed of 10 m/s, differences between the CFD and experimental results were observed in the tangential force, particularly from the leading edge to 63.6% of the chord (Figure 37). However, the normal forces showed good agreement.

The experimental measurements indicated a significant dip in the tangential force distribution at 46.7% chord length for the baseline blade, suggesting a deep stall region. This dip was not captured in the CFD results. Nevertheless, the aerodynamic effect of the winglets was evident in the tangential force distribution, with WD-3 exhibiting the largest C_T value at this speed. Compared to lower wind speeds (5 and 7 m/s), both tangential and normal force coefficients increased with the wind speed.

For a wind speed of 13 m/s, the tangential and normal force distribution curves over the blade are shown in Figure 38. The CFD results were in fairly good agreement with the experimental measurements. Compared to lower wind speeds (5, 7, and 10 m/s), the 13 m/s plot exhibits distinct “U-“ or “V-“shaped distributions for tangential and normal forces, respectively.

The tangential force distribution showed a U-shaped curve, with C_T dropping rapidly to the minimum values near 46.6% and 63.3% of the span before increasing to the second-highest value at 80% span. This aligned with the deep stall observed in pressure distribution at those sections. CFD overpredicted the tangential force across the entire span. Negative C_p values were found at 46.6% span in both the experimental results of the baseline blade and the CFD simulations of the blades with winglets. Normal force coefficients at this speed were significantly larger than tangential force coefficients. CFD underpredicted C_N values compared to experimental measurements.

The effects of adding winglets on the tangential and normal force coefficients were more noticeable toward the blade tip.

At a wind speed of 15 m/s, both the tangential and normal forces exhibited U-shaped distributions over the span, as seen in Figure 39. The CFD and experimental results showed fair agreement at this speed, with CFD overpredicting tangential force and underpredicting normal force.

Similar to the 13 m/s case, the C_T value dropped rapidly to a minimum at 46.6% span before recovering near 95% span. The effect of the winglets was more pronounced toward the tip in both C_T and C_N plots. The C_N values near the hub were significantly larger than those at 13 m/s.

At higher wind speeds (20 and 25 m/s), the entire blade remained in the deep stall region. Figure 40 and Figure 41 show similar force distribution patterns, but they differ from those at lower wind speeds. C_T and C_N decreased from the hub toward the tip, with a slight bump in C_T at 63.3% span. CFD overpredicted tangential force and underpredicted normal force compared to the experimental measurements. Adding winglets significantly reduced tangential force over most of the span except the tip area, while slightly lowering normal force.

3.7. Torque and Thrust Force

Figure 42 compares the result for the torque generated by the blades with winglets, the CFD-simulated original NREL blade, and the NREL experimental measurement across seven wind speeds. Higher torque values generally lead to increased power generation, primarily influenced by the tangential force coefficient.

At 5 m/s, torque values for all four blade configurations were similar. This was because the entire blade remained in the attached flow region, with small AOAs over the blade span. Consequently, the flow remained attached, and tangential force coefficients were comparable. However, WD-3 exhibited slightly lower torque.

Similarly, at 7 m/s, most of the flow remained attached, and tangential force coefficients were slightly lower for the winglet cases, resulting in slightly lower torque.

At 10 m/s, the blade encountered a 20° AOA (onset of complete separation) at 46.7%, where discrepancies in tangential force coefficients were observed between the NREL experimental and CFD results. Thus, the torque value of the original blade was comparatively overpredicted. The CFD results for all blades suggested higher tangential force coefficients, leading to higher torque values for all winglet cases compared to the baseline blade. Among the winglet cases, WD-2 exhibited the highest tangential force coefficient value at 95% span, resulting in the highest torque value at 10 m/s.

At 13 m/s, the blade encountered a 20° AOA at 73% span, and discrepancies in tangential force coefficients were observed between the NREL experimental and CFD results. CFD overpredicted tangential force in most sections, leading to slightly overpredicted torque values. The tangential force coefficients of the three winglets were overpredicted at 30%, 80%, and 95% span, resulting in higher torque values compared to the baseline blade. However, the torque increase was less significant than that at 10 m/s. WD-2 again exhibited the highest torque value.

At 15 m/s, the blade encountered a 20° AOA at 90% span. CFD overpredicted the tangential force before 90% span and underpredicted it at 95% compared to the NREL experimental data. Torque values were similar for both cases. The tangential force coefficients were higher for the winglet cases at 30%, 80%, and 95% span, leading to higher torque compared to the NREL CFD and experimental results. WD-3 exhibited the highest torque.

At 20 m/s, the entire blade was in the deep stall region. CFD torque values for the original blade were nearly identical to the experimental results. The winglet cases exhibited lower torque values due to increased flow separation and lower tangential force coefficients. All three winglet cases had similar torque values.

At 25 m/s, the flow was completely separated. The torque values of all cases were underpredicted compared to the experimental measurements, likely due to the increased flow separation caused by the winglets.

At high wind speeds, winglets had a negligible beneficial effect on torque.

Figure 43 compares the thrust forces for blades with winglets, the CFD-simulated original NREL blade, and the NREL experimental measurements at various wind speeds. The thrust force was primarily influenced by the normal force coefficient (C_N). Higher C_N values generally led to higher thrust force. However, increased thrust can negatively impact the stability of a wind turbine.

At low wind speeds of 5 and 7 m/s, all blade configurations exhibited similar thrust force and stability, with minimal variations. This was likely due to the attached flow and the low normal force coefficients.

At medium wind speeds (10 to 15 m/s), blades with winglets showed significantly increased thrust force, potentially compromising stability.

At high wind speeds (20 and 25 m/s), the flow was completely separated, and the entire blade was in deep stall. CFD underpredicted normal force coefficients compared to the experimental data, leading to lower thrust force values for all winglet cases. WD-3 exhibited the lowest thrust force. This reduction in thrust at high wind speeds could improve stability for blades with winglets, particularly WD-3. However, further analysis and experimental validation would be necessary to confirm this.

By conducting detailed comparisons across various metrics, this study aimed to provide a comprehensive understanding of the aerodynamic effects of winglets on wind turbines. Our findings indicate that the impact of winglets varies significantly across different wind speeds:

At low wind speeds (5 and 7 m/s), the angle of attack (AOA) remains low across the entire blade span, maintaining attached flow to the blade surface. Consequently, the aerodynamic effects of winglets are negligible.

At 10 m/s, winglets enhance the aerodynamic performance in this region, resulting in increased torque and thrust. The WD-2 winglet configuration yielded the highest torque, approximately 24% greater than the baseline.

At 13 m/s, winglets again improve performance, particularly near the winglet tip, where increased flow attachment leads to higher torque and thrust. The WD-3 configuration provided the highest torque, although the overall increase was smaller than at 10 m/s.

At 15 m/s, winglets delay flow separation near the tip, resulting in significant torque and thrust increases. The WD-3 configuration again offered the highest torque, with a 35% increase compared to the baseline.

At high wind speeds (20 m/s and 25 m/s), the entire blade experiences significant flow separation. While winglets can still influence the flow field, their impact on overall performance is limited. In some cases, winglets may even slightly reduce torque and thrust due to increased separation.

4. Conclusions

In this study, CFD simulations were conducted to investigate the aerodynamic effects of winglets based on the NREL Phase VI wind turbine model. Three winglet designs with varying lengths were analyzed across a wind speed range of 5 to 25 m/s.

From the presented results, the following can be concluded:

• Winglets demonstrate significant performance gains at moderate wind speeds (10–15 m/s) by delaying flow separation. In particular, at 10 m/s, the increases in torque for blades with WD-1, WD-2, and WD-3 were approximately 19%, 23%, and 22%, respectively; at 13 m/s, the increases in torque for blades WD-1, WD-2, and WD-3 were approximately 21%, 21%, and 25%, respectively; and at 13 m/s, the increases in torque for blades WD-1, WD-2, and WD-3 were approximately 29%, 26%, and 35%, respectively.
• Winglets demonstrate limited impact at low and high wind speeds. At low wind speeds, the flow remains attached, and winglets have minimal impact. At high wind speeds, extensive flow separation limits the effectiveness of winglets.

It can also be seen that while 2D airfoil analysis can provide valuable insights into the flow physics around wind turbine blades, it is essential to consider three-dimensional effects, such as spanwise variations in angle of attack, three-dimensional flow separation, and vortex formation, to accurately predict the performance of wind turbines. Comprehensive CFD simulations coupled with detailed visualizations of pressure and velocity distributions offer a powerful tool for understanding complex flow physics and optimizing winglet designs.

Future research should focus on optimizing winglet design, exploring the impact of different geometric parameters (such as length, cant angle, twist angle, and taper ratio), and investigating the integration of winglets with other blade modifications.