An Investigation of Atmospheric Icing Effects on Wind Turbine Blade Aerodynamics and Power Output: A Case Study of the NREL 5 MW Turbine

Abstract

This study presents a numerical investigation of the effects of atmospheric icing on the aerodynamic performance and power output of the NREL 5 MW reference wind turbine. In cold climate regions, ice accretion on wind turbine blades significantly alters the airfoil geometry, leading to aerodynamic degradation characterized by increased drag, reduced lift, and substantial power losses. Understanding these effects is therefore essential for reliable performance prediction and efficient turbine operation under icing conditions. To address this problem, numerical simulations were conducted on six representative blade sections using the FENSAP-ICE framework, which integrates flow field calculations, droplet transport, and ice accretion modeling within a unified computational environment. The analyses were performed under different atmospheric icing conditions, considering liquid water content values of 0.22 g/m³ and 0.50 g/m³ and ambient temperatures of −2.5 °C and −10 °C. The median volumetric diameter was fixed at 20 µm, and the icing duration was set to one hour for all cases, allowing for both glaze and rime ice formations to be systematically examined. The results reveal that ice accretion becomes increasingly pronounced toward the blade tip, mainly due to higher relative velocities and increased collection efficiency in the outer sections. Glaze icing conditions produce irregular horn-shaped ice formations and lead to severe aerodynamic degradation, whereas rime ice forms more compact structures near the leading edge and results in comparatively lower performance losses. The degraded aerodynamic coefficients obtained from the iced airfoils were subsequently incorporated into BEM-based power calculations, indicating that total power losses can reach up to 40% under severe icing conditions, with the outer blade sections contributing most significantly to this reduction. Furthermore, an economic assessment based on annual energy losses highlights the substantial impact of atmospheric icing on wind turbine performance and operational costs.

1. Introduction

Wind energy has become one of the most important pillars of the global transition toward sustainable and low-carbon energy systems. Continuous technological advancements, increasing installed capacity, and declining production costs have significantly enhanced the competitiveness of wind power worldwide [1,2]. As a result, wind turbines are increasingly deployed not only in temperate regions but also in cold and complex climates, where harsh environmental conditions can strongly affect turbine performance and reliability. Among these environmental factors, atmospheric icing represents one of the most critical challenges for wind turbine operation, particularly in regions characterized by low temperatures and high humidity levels [3,4,5].

Atmospheric icing occurs when supercooled water droplets present in the air impinge on the blade surface and freeze, leading to ice accretion along the airfoil. This phenomenon has been widely reported to cause severe aerodynamic performance degradation, increased structural loading, safety hazards due to ice shedding, and substantial losses in energy production [6,7,8]. Field observations and experimental studies have shown that icing can reduce annual energy production by several tens of percent in cold climate wind farms, significantly increasing operational costs and threatening the economic viability of wind energy projects [9,10]. Consequently, a detailed understanding of icing-induced aerodynamic and power losses is essential for accurate performance prediction, turbine design optimization, and the development of effective mitigation strategies.

Ice accretion on wind turbine blades alters the local airfoil geometry and surface roughness, directly affecting the aerodynamic characteristics of the blade sections. The presence of ice generally leads to a reduction in lift coefficient, a significant increase in drag coefficient, and an earlier onset of flow separation, ultimately reducing torque generation and power output [11,12,13,14]. The severity of these effects depends strongly on atmospheric conditions such as ambient temperature, liquid water content (LWC), median volumetric diameter (MVD), and relative flow velocity [4,5]. In particular, ice accretion is often more pronounced toward the blade tip, where higher relative velocities and collection efficiencies increase the droplet impingement rate [13,15].

Depending on the thermodynamic conditions during icing, different ice types may form on the blade surface. Glaze ice typically develops at temperatures close to the freezing point under high LWC conditions and is characterized by irregular, horn-shaped geometries caused by partial freezing and water runback [3,7]. These complex ice shapes strongly disturb the flow field and lead to severe aerodynamic penalties. In contrast, rime ice forms at lower temperatures under reduced LWC, where droplets freeze rapidly upon impact, resulting in more compact and smoother ice accretions near the leading edge [3,5]. Although rime ice generally causes less severe aerodynamic degradation than glaze ice, its impact on turbine performance can still be significant, especially when accumulated over extended operating periods [12,16].

Numerous experimental and numerical studies have investigated the effects of atmospheric icing on wind turbine aerodynamics and performance. Early investigations demonstrated that ice accretion on wind turbine blades leads to substantial reductions in lift and pronounced increases in drag, resulting in power losses that can exceed 25–40% under severe icing conditions [13,14,15,17]. Using numerical simulations, Homola et al. [14] analyzed icing effects on the NREL 5 MW wind turbine and reported a significant reduction in the power coefficient due to ice accretion, particularly at higher tip speed ratios. Similarly, Virk et al. [13] showed that ice accumulation is more severe in the outer blade regions because of higher relative velocities and increased collection efficiency, leading to dominant contributions to total power loss.

Several studies have focused on distinguishing the aerodynamic consequences of different ice types. Lamraoui et al. [9] and Hochart et al. [18] emphasized that glaze ice, characterized by irregular horn-shaped formations, causes stronger flow separation and more severe aerodynamic penalties compared to rime ice. Ettemaddar et al. [17] reported that glaze icing not only reduces aerodynamic efficiency but also shifts the rated wind speed and alters turbine operational behavior. More recent numerical investigations employing advanced CFD-based icing solvers, such as FENSAP-ICE, have further highlighted the strong sensitivity of aerodynamic degradation to atmospheric parameters, including LWC and ambient temperature [11,19]. Despite these extensive efforts, a significant research gap remains. Many existing studies either focus on a limited number of blade sections without full-span integration, primarily address isolated aerodynamic coefficients, or examine icing without a systematic variation in critical atmospheric parameters [11,13,14,19]. To bridge this gap, the present study aims to provide a comprehensive numerical assessment of atmospheric icing effects on both the aerodynamic behavior of wind turbine blade sections and the resulting power output of the turbine. Six representative sections of the NREL 5 MW reference wind turbine blade are analyzed under a range of atmospheric icing conditions using the FENSAP-ICE framework. Both glaze and rime icing regimes are systematically investigated by varying liquid water content and ambient temperature, while keeping the median volumetric diameter and icing duration constant. The degraded aerodynamic coefficients obtained from the iced airfoil simulations are subsequently integrated into Blade Element Momentum (BEM)-based power calculations to quantify turbine-level power losses.

The main contribution of this work lies in the combined evaluation of local aerodynamic degradation, spanwise ice accretion characteristics, and their direct influence on overall power production. By explicitly linking sectional icing-induced aerodynamic penalties to turbine power output and associated energy losses, this study offers a clearer understanding of the mechanisms governing performance degradation under atmospheric icing conditions. The findings are expected to support more accurate performance prediction, improved cold-climate turbine design, and informed decision-making regarding icing mitigation strategies.

2. Numerical Methodology

2.1. Wind Turbine Blade Sections

The numerical analyses were performed on the NREL 5 MW reference wind turbine, which has been widely adopted in the literature as a benchmark configuration for aerodynamic and performance studies of large-scale wind turbines. The turbine features a three-bladed, upwind rotor with a rated power of 5 MW and a rotor diameter of 126 m, making it representative of modern utility-scale wind turbines operating under a wide range of atmospheric conditions [20]. All fundamental characteristics are given in

Table 1. Fundamental characteristics of the NREL 5 MW wind turbine [20].

Parameter	Value
Rated power	5 MW
Rotor radius	63 m
Hub radius	1.5 m
Blade length	61.5 m
Cut-In, Rated, Cut-Out Wind Speed	3 m/s, 11.4 m/s, 25 m/s
Cut-In, Rated Rotor Speed	6.9 rpm, 12.1 rpm

.

To capture the spanwise variation in ice accretion and its aerodynamic consequences, six representative blade sections were selected along the blade span, extending from the inboard region toward the blade tip, shown in Figure 1. The selected sections were chosen to reflect the geometric and aerodynamic characteristics of the blade from root to tip, where variations in chord length, airfoil shape, and relative flow conditions are significant, as given in

Table 2. Geometric parameters of the NREL 5 MW wind turbine blade sections.

Airfoil	Node Radius R (m)	Twist Angle (°)	Chord Length (m)	Section
Cylinder1	2.8667	13.308	3.542	–
Cylinder1	5.60	13.308	3.854	–
Cylinder2	8.3333	13.308	4.167	–
DU40	11.75	13.308	4.557	–
DU35	15.85	11.480	4.652	Section F
DU35	19.95	9.690	4.458	–
DU30	24.05	9.011	4.249	Section E
DU25	28.15	7.795	4.048	–
DU25	32.25	6.544	3.748	Section D
DU21	36.35	5.361	3.502	–
DU21	40.45	4.188	3.256	Section C
NACA64	44.55	3.125	3.010	–
NACA64	48.65	2.319	2.764	–
NACA64	52.75	1.526	2.518	Section B
NACA64	56.1667	0.863	2.313	–
NACA64	58.90	0.370	2.086	–
NACA64	61.6333	0.106	1.419	Section A

. This approach allows the assessment of local icing effects while maintaining a manageable computational cost.

Table 2 presents the radial position (R), twist angle, chord length, and the corresponding airfoil for each blade section. In addition, the sections A, B, C, D, E, and F used in the analyses are also indicated in the same table.

The selection of multiple blade sections is particularly important for atmospheric icing studies, as ice accretion is strongly influenced by local relative velocity and angle of attack, both of which increase toward the blade tip. Consequently, outer blade sections are expected to experience higher droplet impingement rates and more severe ice accumulation, leading to a dominant contribution to overall aerodynamic degradation and power loss. By analyzing blade sections spanning from the root to the tip region, the present methodology enables a systematic investigation of spanwise icing behavior and its impact on turbine performance.

2.2. Aerodynamic and Power Calculations

The aerodynamic performance of the iced blade sections was evaluated by incorporating the lift and drag coefficients obtained from the icing simulations into a Blade Element Momentum (BEM)-based power calculation framework. For each blade section, the local relative velocity and angle of attack were determined from BEM calculations corresponding to the operating conditions of the NREL 5 MW wind turbine. These parameters account for the combined effects of inflow wind speed, rotor rotation, and radial position along the blade span [13,14].

Blade Element Momentum (BEM) theory is a commonly used approach in aerodynamic analyses and performance predictions for horizontal-axis wind turbines. This theory enables the determination of geometric properties such as rotor diameter, aerodynamic airfoil profiles, chord length, and pitch and twist distributions, in addition to rotor design. The BEM approach is based on combining the axial and angular components of momentum conservation with blade element theory. This approach plays a critical role in both the design and optimization processes by enabling the calculation of torque, power, and turbine performance at different wind speeds [21].

Fundamental aerodynamic and geometric parameters such as air density, free-stream velocity, rotor radius, twist angle, number of blades, and the angular velocity of the rotor are determined. Furthermore, it is necessary to assign the induction factors prior to initiating the calculations. According to the literature, the commonly used initial values for the induction factors are $a=0$ for the axial induction factor and $a^{\prime }=0$ for the tangential induction factor [22].

The equations for the axial velocity ($V_{axial}$), tangential velocity ($V_{tangential}$) and the relative velocity ($V_{rel}$), which is the vector sum of these two components, are given in Equations (1), (2) and (3) respectively [22]. Velocity triangles are illustrated in Figure 2.

$$V_{axial}=V_o(1-a)$$

(1)

$$V_{tangential}=\omega r(1+a^{\prime })$$

(2)

$$V_{rel}=\sqrt{{\left[V_o(1-a)\right]}^2+{\left[\omega r(1+a^{\prime })\right]}^2}$$

(3)

where $V_0$ represents the free-stream velocity, $\omega$ denotes the rotational speed of the rotor, and r represents the radial position of the airfoil.

It is important to note that the BEM calculations in this study were intentionally performed at the rated wind speed of 11.4 m/s. While wind turbines naturally operate across a wide range of wind speeds, the rated speed represents the most critical aerodynamic and operational threshold for the NREL 5 MW turbine. At this state, the turbine reaches its maximum aerodynamic loading and extracts its rated power just before the active blade pitch control mechanisms are deployed. Consequently, evaluating the aerodynamic degradation at this specific set-point provides the most critical assessment of ice-induced power penalties, while keeping the computational cost of the 24 parametric scenarios within reasonable bounds.

The sectional aerodynamic coefficients of the iced airfoils were then used to compute the local aerodynamic forces acting on each blade element. Compared to the clean airfoil cases, ice accretion resulted in reduced lift coefficients ($C_L$) and increased drag coefficients ($C_D$), leading to a deterioration of aerodynamic efficiency. The sectional aerodynamic loads were subsequently integrated along the blade span to estimate the overall torque and power output of the turbine under iced conditions. This approach enables a direct assessment of how local icing-induced aerodynamic degradation propagates to turbine-level power losses. The lift force (L) and drag force (D) are defined in Equations (4) and (5), respectively [22].

$$L={\displaystyle \frac{1}{2}}\rho V_{rel}^2{C}_L$$

(4)

$$D={\displaystyle \frac{1}{2}}\rho V_{rel}^2{C}_D$$

(5)

The flow angle ($\varphi$) is defined as the ratio of the axial to the tangential velocity components. The angle of attack ($\alpha$) is defined as the difference between the flow angle and the blade twist angle ($\theta$). The corresponding expressions are given in Equations (6) and (7) respectively [22].

$$\varphi =\arctan \left({\displaystyle \frac{V_o(1-a)}{\omega r(1+a^{\prime })}}\right)$$

(6)

$$\alpha =\varphi -\theta$$

(7)

It should be emphasized that the spanwise variation in the twist angle ($\theta$) directly modifies the local angle of attack ($\alpha$) for each blade element, as governed by Equation (7). Consequently, the distinct geometric inflow conditions and the specific angles of attack caused by these twist variations were inherently incorporated into both the BEM calculations and the boundary conditions of the icing simulations.

Since the CFD-derived aerodynamic polars were obtained for a limited range of operational angles of attack, the Viterna–Corrigan extrapolation method was applied to extend the lift and drag coefficients up to 90°. This robust extrapolation is particularly crucial for the BEM algorithm in post-stall regimes, where severe icing causes early flow separation. For the spanwise integration of these aerodynamic loads, a linear interpolation scheme was utilized to determine the lift and drag coefficients between the specific blade sections analyzed. Furthermore, drag management within the BEM framework was handled by directly incorporating the ice-induced increases in sectional drag coefficients into the torque equations, ensuring that the additional power penalties caused by profile drag were accurately represented across the entire rotor span.

It should be acknowledged that utilizing 2D airfoil data for a 3D rotating wind turbine rotor carries inherent limitations. In actual operation, centrifugal pumping and Coriolis forces induce a phenomenon known as stall delay, where boundary layer separation is naturally postponed compared to static 2D cases. While advanced 3D correction models exist, this study intentionally utilizes 2D sectional data as a conservative baseline. This approach maintains a clear focus on the relative performance degradation specifically caused by the ice accretion geometry, consistent with the established computational icing literature.

To obtain realistic load distributions from the BEM theory, both Prandtl’s tip and hub loss corrections must be incorporated. The fundamental BEM theory assumes a rotor with an infinite number of blades, a simplification that leads to an overestimation of aerodynamic loads near the blade extremities. To rectify this, a combined Prandtl loss factor (F) is utilized, representing the product of the tip ($F_{tip}$) and hub ($F_{hub}$) losses:

$$F=F_{tip}·F_{hub}$$

(8)

where $F_{tip}$ and $F_{hub}$ are calculated as follows:

$$F_{tip}={\displaystyle \frac{2}{\pi }}{\cos }^{-1}\left[\exp \left(-{\displaystyle \frac{B(R-r)}{2r\sin \varphi }}\right)\right]$$

(9)

$$F_{hub}={\displaystyle \frac{2}{\pi }}{\cos }^{-1}\left[\exp \left(-{\displaystyle \frac{B(r-R_{hub})}{2R_{hub}\sin \varphi }}\right)\right]$$

(10)

Here, B is the number of blades, R is the total rotor radius, $R_{hub}$ is the hub radius, r is the local radial position, and $\varphi$ is the local flow angle. This combined factor F is essential for correcting the momentum balance used to determine the axial (a) and tangential ($a^{\prime }$) induction factors.

Furthermore, standard BEM theory becomes physically invalid when the axial induction factor exceeds the Glauert limit ($a>0.4$), a turbulent wake condition frequently encountered due to the drastically increased drag and reduced lift under severe icing scenarios. To maintain numerical stability and physical accuracy in such high-loading conditions, the Glauert empirical correction was strictly integrated into the iterative algorithm. This ensures that the relationship between the thrust coefficient and the induction factor remains consistent even when the rotor operates within the heavily stalled wake state.

The power equation of the turbine is presented as

$$dP=\omega dQ$$

(11)

where the torque equation of the turbine is

$$dQ=F\sigma \pi \rho {\displaystyle \frac{V_o^2{(1-a)}^2}{{\sin }^2\varphi }}(C_l\sin \varphi -C_d\cos \varphi )r^{2}dr$$

(12)

By applying the degraded aerodynamic characteristics individually to each blade section, the methodology captures the spanwise variation in icing effects, which is particularly important for large-scale wind turbines. As higher relative velocities and angles of attack are encountered toward the blade tip, outer blade sections were found to contribute more significantly to the total power reduction, in agreement with previous numerical and experimental findings reported in the literature [13,14,15,16]. In this study, BEM calculations were performed under steady operating conditions representative of the rated operating point of the NREL 5 MW turbine, assuming identical inflow conditions for both clean and iced configurations to enable a consistent comparison of power losses. Control system dynamics were not considered, and the analysis focuses on aerodynamic performance differences.

2.3. Numerical Framework for Icing Simulations

Atmospheric icing simulations were carried out using the ANSYS FENSAP-ICE 23.1 computational framework, which has been extensively validated and applied in previous icing studies on wind turbines and aerodynamic surfaces [9,11,14]. The framework couples aerodynamic flow field calculations with droplet transport and ice accretion modeling, providing a comprehensive numerical environment for predicting ice formation under specified atmospheric conditions.

The airflow around the blade sections was solved within a Reynolds-averaged Navier–Stokes (RANS) formulation, while turbulent flow effects were modeled using the Spalart–Allmaras turbulence model. This model was specifically chosen for its robustness in handling complex iced geometries and its widespread adoption in wind turbine icing literature as a benchmark approach [11,14,18]. Droplet transport was modeled using an Eulerian approach through the ANSYS DROP3D 23.1 module, allowing the computation of droplet impingement and collection efficiency on the blade surface [12,19]. Ice accretion was subsequently simulated using the ANSYS ICE3D 23.1 module, which is based on an energy balance approach and accounts for phase change, heat transfer, and local thermodynamic conditions during ice growth [3,9].

2.4. Boundary Conditions

The atmospheric icing conditions considered in this study were defined based on parameters commonly used in previous numerical and experimental icing investigations on wind turbines [3,4,5,9,13]. Ice accretion simulations were performed for different combinations of liquid water content (LWC) and ambient temperature in order to represent both glaze and rime icing regimes. Two LWC values, 0.22 g/m³ and 0.50 g/m³, were selected to capture moderate and severe icing conditions, while ambient temperatures of $-2.5$ °C and $-10$ °C were considered to distinguish between wet and dry icing behavior.

The median volumetric diameter (MVD) of the supercooled water droplets was fixed at 20 μm for all cases, consistent with commonly adopted values in wind turbine icing studies and cold climate standards [4,5,9]. This assumption allows a systematic comparison of icing effects by isolating the influence of LWC and temperature. The icing duration was set to one hour for all simulations, providing sufficient time for ice accretion to develop while maintaining computational feasibility.

Based on the selected atmospheric conditions, both glaze and rime ice formations were examined. Glaze icing conditions were associated with higher ambient temperatures and elevated LWC values, leading to partial droplet freezing and water runback along the blade surface. In contrast, rime icing conditions occurred at lower temperatures and reduced LWC, resulting in rapid droplet freezing upon impact and the formation of more compact ice structures near the leading edge [3,9,18]. The aerodynamic simulations were conducted assuming the ice surface as geometrically resolved. Furthermore, the surface roughness induced by the ice accretion was explicitly accounted for using the Shin roughness correlation, as implemented within the FENSAP-ICE framework. These scenarios enable a clear assessment of the influence of ice type on aerodynamic degradation and subsequent power losses.

For each blade section, boundary conditions were defined using the relative velocities and angles of attack obtained from Blade Element Momentum (BEM) calculations corresponding to the operating conditions of the NREL 5 MW wind turbine. These specific local flow parameters, which act as the far-field velocity boundary conditions and are strictly required for accurate droplet trajectory calculations in the Eulerian module, are explicitly detailed in

Table 3. Local flow boundary conditions extracted from BEM calculations for each investigated blade section at the rated wind speed of 11.4 m/s.

Airfoil Profile	Radial Position, r (m)	Relative Velocity (m/s)	Angle of Attack (°)
DU35	19.95	26.30	6.40
DU30	24.05	31.00	5.00
DU25	32.25	40.70	3.60
DU21	40.45	50.50	3.30
NACA64-618	52.75	65.38	4.25
NACA64-618	61.63	76.23	5.76

. The computed flow parameters were applied individually to each section to ensure realistic representation of spanwise variations in aerodynamic loading and ice accretion behavior. This approach allows for the direct comparison of icing severity and its aerodynamic consequences across different blade regions under identical atmospheric conditions.

To ensure strict numerical accuracy and methodological transparency, all solver settings were carefully configured. The aerodynamic flow was modeled assuming a fully turbulent boundary layer, hence no transition model was employed. The droplet trajectories were computed using an Eulerian two-phase flow approach, and the subsequent ice growth was calculated via a single-shot accretion method. Finally, the target convergence criteria for the root-mean-square (RMS) residuals of the continuity, momentum, energy, and droplet equations were all strictly set to $1\times {10}^{-6}$.

2.5. Validation Study

The numerical framework employed in this study was validated by comparing the predicted aerodynamic degradation and power loss trends with previously published numerical and experimental investigations on atmospheric icing of wind turbines. The variations in lift and drag coefficients obtained for iced airfoils show consistent trends with the results reported by NREL [20] and Homola et al. [14], particularly in terms of increased drag, reduced lift, and earlier flow separation under icing conditions.

The clean case validation results for the lift coefficient ($C_L$) are given in Figure 3a. When the deviation values are examined, the difference is 4.48% for 1° angle of attack, 7.94% for 6°and 35.40% for 11°. When the results are compared with the data shared by Homola et al. [14], the differences are 6.85% for about 1° angle of attack, 4.86% for 6° and 2.11% for 11°.

As observed in Figure 3, a divergence between the numerical CFD results and the NREL data begins around an angle of attack of 6°. This discrepancy is a known limitation of fully turbulent 2D RANS simulations (such as the Spalart–Allmaras model used herein). As extensively reported in the literature [23,24,25,26,27], these models tend to over-predict flow attachment and delay boundary layer separation at higher angles of attack compared to experimental wind tunnel data, leading to an overestimation of the maximum lift coefficient. However, the excellent agreement with the numerical results of Homola et al. [14] confirms the reliability of the present computational setup for the subsequent icing analyses.

Figure 3b shows the validation results for the clean case drag coefficient ($C_D$). In contrast, the simulation results demonstrate good agreement with the data reported by Homola et al. [14], which employ the same turbulence model commonly used in the literature.

The DU 21 airfoil was selected for the iced validation study, and simulations were performed at approximately 1°, 6°, and 11° angles of attack. The same mesh resolution and turbulence model parameters employed in the clean case validation were retained to ensure methodological consistency. Icing simulations were conducted using the ANSYS FENSAP-ICE 23.1 framework under the atmospheric conditions, corresponding to a free-stream velocity of 10 m/s, a median volumetric diameter (MVD) of 20 μm, a liquid water content (LWC) of 0.22 g/m³, an air static temperature of −10 °C, and an icing duration of 1 h.

Figure 4a presents the comparison of the lift coefficient values obtained from the present simulations with the data reported by Homola et al. [14]. The deviations between the numerical predictions and the reference data were found to be 0.98% at 1°, 4.25% at 6°, and 6.70% at 11° angle of attack. These results indicate very good agreement across the examined angle-of-attack range, particularly considering the increased flow complexity introduced by ice accretion.

Similarly, the drag coefficient validation was performed using the numerical results reported by Homola et al. [14] as a benchmark. The comparison, illustrated in Figure 4b, shows deviations of 1.12% at approximately 1°, 5.59% at 6°, and 10.68% at 11° angle of attack. Although slightly larger discrepancies are observed at higher angles of attack, the predicted trends and magnitude of drag increase under icing conditions remain consistent with the reference study.

Mesh sensitivity and numerical accuracy were evaluated through preliminary grid refinement analyses and in consideration of the validation studies performed for both clean and iced airfoil cases. The adopted mesh resolution was selected to ensure stable and reliable predictions of aerodynamic coefficients and ice accretion patterns while maintaining computational efficiency. Particular attention was given to resolving the near-wall region and leading-edge gradients to adequately capture flow separation and ice geometry. Therefore, a C-type mesh topology consisting of approximately 1.3 million elements was employed for the turbine blade sections, as illustrated in Figure 5. To ensure numerical stability and minimize discretization errors, stringent mesh quality controls were applied to these generated grids. Quantitatively, over 95% of the computational cells exhibited an orthogonal quality above 0.90, with the absolute minimum maintained strictly above 0.85. Similarly, more than 85% of the cells possessed a skewness below 0.10, with the maximum skewness kept below 0.15 across all simulated sections, indicating a highly structured topology. Furthermore, the maximum aspect ratio inside the boundary layer was limited to 1000, and the first cell height was carefully controlled to approximately $1\times {10}^{-5}$ m. This ensured a non-dimensional wall distance of $y^{+}\le 1.0$ across the airfoil surfaces, strictly satisfying the requirements for the Spalart–Allmaras turbulence model employed in this study.

3. Results

In this study, a total of 24 distinct icing scenarios were simulated to investigate the combined effects of spanwise position, liquid water content (LWC), and ambient temperature on wind turbine performance. The analyses were conducted on six representative blade sections (Sections A through F) defined in the methodology. To systematically evaluate the impact of atmospheric conditions, simulations considered two LWC values (0.22 g/m³ and 0.50 g/m³) representing moderate and severe icing, and two ambient temperatures ($-2.5$ °C and $-10$ °C) corresponding to glaze and rime ice formation, respectively.

To facilitate the presentation of results and ensure clarity in the subsequent discussions, a specific Case ID nomenclature is adopted, as summarized in

Table 4. Nomenclature for the 24 simulated icing scenarios.

Section Labels	LWC [g/m3]	Temperature [°C]	Ice Type	Case ID Format
A, B, C, D, E, F	0.22	−2.5	Glaze	[Section]02225
A, B, C, D, E, F	0.22	−10	Rime	[Section]02210
A, B, C, D, E, F	0.50	−2.5	Glaze	[Section]05025
A, B, C, D, E, F	0.50	−10	Rime	[Section]05010

. Each case is designated by a code combining the section label, LWC value, and temperature. For example, the Case ID A02225 refers to Section A subjected to an LWC of 0.22 g/m³ at a temperature of $-2.5$ °C. Similarly, C05010 denotes Section C under 0.50 g/m³ LWC and $-10$ °C conditions. This naming convention is consistently used throughout the tables and figures in this section.

It should be explicitly noted that the ice accretion profiles and thickness values presented in the subsequent figures and tables do not represent an accretion rate. Rather, they illustrate the cumulative, final ice shapes and maximum thicknesses obtained after a constant and continuous icing exposure time of exactly one hour (3600 s) for all simulated scenarios. The total water mass flux responsible for these accretions is strictly defined by the combination of the specified LWC levels, the section-specific relative flow velocities, and this one-hour duration.

3.1. Ice Accretion Characteristics

The numerical simulations reveal that ice accretion along the wind turbine blade is highly non-uniform and strongly dependent on the spanwise position, ambient temperature, and liquid water content (LWC). The distribution of ice thickness for different blade sections is summarized in

Table 5. Ice thickness for the 0.22 g/m³ and 0.50 g/m³ LWC at −2.5 °C cases.

Case (0.22 g/m3 LWC)	Thickness (m)	Case (0.50 g/m3 LWC)	Thickness (m)
A02225	0.01948	A05025	0.03291
B02225	0.01519	B05025	0.03113
C02225	0.00695	C05025	0.01303
D02225	0.00101	D05025	0.00232
E02225	0.000090	E05025	0.000207
F02225	0.000031	F05025	0.000069

. A consistent trend is observed across all simulated cases: ice accretion severity increases significantly from the blade root toward the tip. Specifically, the maximum ice thickness was consistently recorded at the outermost section (Section A), while negligible accretion was observed at the inboard sections (Sections E and F). This spanwise variation is attributed to the higher local relative velocity and increased collection efficiency at the blade tip, which enhances the droplet impingement rate compared to the slower-moving root sections with larger chord lengths.

The ambient temperature played a decisive role in determining the type and geometry of the accreted ice. At the warmer temperature of $-2.5$ °C, glaze ice formation was observed. As illustrated in Figure 6, glaze ice is characterized by irregular, horn-shaped structures protruding from the leading edge. This formation occurs because the impinging supercooled water droplets do not freeze instantaneously upon impact; instead, a portion of the water forms a film and runs back along the airfoil surface before freezing. This runback effect is particularly pronounced at the stagnation point, leading to the formation of double horns, which severely disrupt the airfoil geometry.

The distinct ice accretion shapes observed across the different blade sections in Figure 6 are directly governed by the spanwise variation in droplet collection rates and impact angles. Because each section possesses a unique twist angle (and consequently a distinct angle of attack, as calculated via the BEM approach), the aerodynamic stagnation point shifts significantly from the root to the tip. This twist-induced shift alters the primary impingement limits of the droplets, causing the ice horns to form at different orientations relative to the chord line. Furthermore, the local collection efficiency ($\beta$) is highly sensitive to the relative flow velocity. Moving from the inboard sections (e.g., Section A) towards the outboard sections (e.g., Section F), the increasing relative velocity enhances the droplet inertia. This higher inertia prevents the droplets from following the deflected airflow, thereby increasing the maximum local collection rate and resulting in more severe and concentrated ice accretions at the outer span, despite the identical far-field atmospheric conditions.

In contrast, at the lower temperature of $-10$ °C, rime ice was the dominant formation type. As shown in Figure 7, rime ice appears as a streamlined, conformal layer that closely follows the original airfoil contour. Due to the rapid freezing of droplets upon impact, there is minimal water runback, resulting in a more compact and smoother ice profile compared to glaze ice. Although the ice thickness is substantial at the leading edge, the absence of aerodynamic horns makes rime ice geometrically less complex than glaze ice.

As observed in Figure 7, an interesting phenomenon occurs at the intermediate blade span, specifically at Section C, where the maximum ice accretion exhibits noticeably less sensitivity to increasing LWC compared to the tip sections. This behavior highlights a localized transition into a strongly heat-transfer-limited (glaze) icing regime. At this radial position, the moderate relative velocity limits the convective heat transfer capacity. When the LWC is increased, the substantial latent heat of fusion released by the freezing droplets cannot be entirely dissipated. Consequently, the excess incoming water fails to freeze instantly at the leading edge and instead forms runback water that spreads downstream. This thermodynamic saturation prevents the primary leading-edge ice thickness from growing proportionally with the LWC, making Section C visually less responsive to LWC variations.

The Liquid Water Content (LWC) significantly influenced the total mass and volume of the accreted ice but did not fundamentally alter the ice type determined by temperature. Increasing the LWC from 0.22 g/m³ to 0.50 g/m³ resulted in a thicker and more extended ice layer for both temperature regimes. Under glaze conditions ($-2.5$ °C), a higher LWC exacerbated the irregularity of the ice shapes, producing larger and more pronounced horns. Under rime conditions ($-10$ °C), the increased LWC primarily led to a uniform expansion of the ice layer around the leading edge without forming complex protrusions. These results indicate that while temperature dictates the accretion mechanism (glaze vs. rime), LWC acts as an amplifier of the icing severity.

3.2. Aerodynamic Performance Degradation

The alteration of the blade geometry due to ice accretion leads to severe degradation in aerodynamic performance, manifested as a reduction in the lift coefficient ($C_L$) and a substantial increase in the drag coefficient ($C_D$). The spanwise variations of these aerodynamic coefficients under different icing conditions are presented in Figure 8.

Consistent with the ice accretion characteristics, the aerodynamic penalties were most severe at the outboard sections (Sections A and B), where the relative velocity and ice thickness are highest. In contrast, the inboard sections (Sections E and F) exhibited minimal deviation from the clean airfoil performance due to negligible ice accumulation.

Under glaze icing conditions ($T_{amb}=-2.5$ °C), the formation of horn-shaped ice at the leading edge acted as a major disturbance to the flow field. As shown in Figure 8a, this resulted in a sharp decrease in $C_L$, particularly for the 0.50 g/m³ LWC case. The horns effectively acted as spoilers, forcing the flow to separate prematurely from the suction side of the airfoil. This separation significantly reduced the suction peak near the leading edge, thereby diminishing the lift generation capability. In comparison, rime ice ($T_{amb}=-10$ °C) caused a more moderate reduction in $C_L$. Since rime ice conforms more closely to the airfoil shape without forming large protrusions, the flow remained attached over a larger portion of the chord, resulting in less severe lift losses compared to the glaze cases.

3.3. Power Loss Analysis

To quantify the impact of the observed aerodynamic degradation on the overall energy conversion efficiency, the modified lift and drag coefficients obtained from the FENSAP-ICE simulations were integrated into the BEM-based power calculation framework. This integration enables the translation of local spanwise performance penalties into global turbine power losses.

The calculated power losses for all investigated icing cases are summarized in

Table 6. Summary of total power losses calculated via BEM for different atmospheric icing conditions. The highest loss is observed under high-LWC glaze ice conditions.

Icing Case (LWC = 0.22 g/m3)	Power Loss (%)	Icing Case (LWC = 0.50 g/m3)	Power Loss (%)
T = −2.5 °C (Glaze)	34.96	T = −2.5 °C (Glaze)	40.76
T = −10 °C (Rime)	31.28	T = −10 °C (Rime)	36.87

. The results demonstrate a clear correlation between the ice type (driven by ambient temperature) and the severity of power reduction. Under moderate liquid water content conditions (LWC = 0.22 g/m³), the glaze ice formed at $-2.5$ °C resulted in a total power loss of 34.96%. In comparison, the rime ice formed at $-10$ °C caused a comparatively lower reduction of 31.28%. This difference confirms that the irregular, horn-shaped geometries associated with glaze ice disrupt torque generation more severely than the streamlined profiles typical of rime ice.

Increasing the Liquid Water Content (LWC) from 0.22 g/m³ to 0.50 g/m³ consistently exacerbated the power losses across both temperature regimes. For the glaze ice scenario, the power loss increased from 34.96% to 40.76%, representing the worst-case scenario among all investigated conditions. Similarly, for the rime ice scenario, the loss rose from 31.28% to 36.87%. It is worth noting that while doubling the LWC leads to greater ice mass, the increase in power loss is not linearly proportional. This suggests that the initial alteration of the airfoil shape by ice accretion contributes to the majority of the aerodynamic penalty, with subsequent ice growth intensifying the effect.

The magnitude of these losses—ranging from approximately 31% to 41%—highlights the critical role of the outer blade sections. Since the outboard regions (Sections A and B) operate at the highest relative velocities and leverage the longest moment arms, they contribute the majority of the aerodynamic torque. Consequently, the severe icing and aerodynamic degradation observed at the blade tip (as detailed in Section 3.2) have a disproportionately large impact on the total power output of the turbine.

To explicitly quantify this localized aerodynamic degradation and its contribution to the overall deficit, the spanwise distribution of the total power loss was extracted from the BEM calculations. As detailed in

Table 7. Spanwise section contributions to the total aerodynamic power loss under severe glaze icing conditions (T_amb = −2.5 °C, LWC = 0.50 g/m³).

Radial Position, r (m)	Contribution to Total Power Loss (%)
11.75	0.00
19.95	1.19
24.05	2.08
32.25	10.89
40.45	19.47
52.75	44.11
61.63	22.27
Total	≈100.0

, the power penalty is not evenly distributed along the blade. Because torque generation is governed by the radial distance and swept area, the inner blade sections ($r<25$ m) contribute negligibly to the overall power loss. Conversely, the outboard region ($r>40$ m) accounts for more than 85% of the total deficit. The maximum contribution to the power loss occurs at $r=52.75$ m (44.11%), rather than at the absolute blade tip, which perfectly reflects the physical load alleviation caused by Prandtl tip-loss effects.

3.4. Economic Analysis

To translate the aerodynamic performance degradation into practical economic indicators, an assessment of annual energy production (AEP) losses was conducted. This evaluation considers a representative onshore wind farm located in Northern Finland, classified as an Icing Class 4 region. Based on IEA Wind TCP classifications, an instrumental icing duration of 1500 h per year (approximately 17% of the annual operation) was assumed [28,29]. The economic impact was calculated using an average electricity market price of 46 €/MWh and a regional capacity factor of 30.1% [30].

To ensure the reproducibility of the economic assessment, the annual energy production and subsequent financial losses were calculated using explicit operational formulations. The baseline annual energy production for the clean rotor is defined as

$$AE{P}_{\mathrm{clean}}=P_{\mathrm{rated}}·CF·8760$$

(13)

where $P_{\mathrm{rated}}$ is the rated power of the turbine (5 MW), $CF$ represents the average annual capacity factor, and 8760 denotes the total number of hours in a year. Accordingly, the baseline $AE{P}_{\mathrm{clean}}$ for the NREL 5 MW reference turbine is calculated as 13,183.8 MWh/year.

While Equation (13) provides the baseline production, the energy loss caused specifically by icing conditions ($AE{P}_{\mathrm{loss}}$) is estimated using the following expression:

$$AE{P}_{\mathrm{loss}}=P_{\mathrm{rated}}·CF·h_{\mathrm{ice}}·\eta$$

(14)

where $h_{\mathrm{ice}}$ represents the estimated annual icing frequency (1500 h/year), and $\eta$ is the total power loss ratio obtained from the BEM calculations for each respective icing condition. Finally, the annual economic loss ($C_{\mathrm{loss}}$) is quantified by

$$C_{\mathrm{loss}}=AE{P}_{\mathrm{loss}}·c_{\mathrm{elec}}$$

(15)

where $c_{\mathrm{elec}}$ represents the average wholesale electricity market price (46 €/MWh).

The deviation from the baseline under icing conditions is summarized in

Table 8. Annual energy production losses and estimated economic impact for a 5 MW turbine under different icing scenarios (assuming 1500 h/year icing duration and 46 €/MWh electricity price).

Temp. (°C)	LWC (g/m3)	Power Loss (η)	ΔAEP (MWh/Year)	Economic Loss (€/Year)
−2.5	0.22	0.3496	789.25	36,305
−2.5	0.50	0.4076	920.15	42,327
−10.0	0.22	0.3128	706.14	32,482
−10.0	0.50	0.3687	832.34	38,287

. The analysis reveals that the aerodynamic penalties discussed in Section 3.2 result in substantial financial losses. Under moderate glaze icing conditions ($T_{amb}=-2.5$ °C, LWC = 0.22 g/m³), the turbine experiences an annual energy loss of approximately 789 MWh, corresponding to a revenue reduction of €36,305 per turbine per year. When the icing severity increases (LWC = 0.50 g/m³), the financial loss escalates to €42,327. Even under rime icing conditions ($T_{amb}=-10$ °C), which are aerodynamically less severe than glaze ice, the economic losses remain significant, ranging from €32,482 to €38,287 annually.

These figures demonstrate that atmospheric icing is not merely an aerodynamic concern but a critical economic challenge. For a typical wind farm consisting of multiple turbines, these individual losses would accumulate to a substantial financial deficit.

Furthermore, as established by the spanwise power loss distribution (Table 7), over 85% of this deficit originates from the outboard region. This explicitly dictates that active ice protection systems (IPSs) should economically prioritize the outer 30% of the blade span to mitigate the vast majority of financial penalties, clearly justifying the investment in active anti-icing or de-icing systems for ensuring the economic viability of wind energy projects in cold climates.

While the present study provides a comprehensive parametric assessment of icing effects and their influence on power production, certain methodological limitations should be acknowledged. First, the numerical icing simulations were conducted on 2D extracted blade sections under steady operating conditions. Although this approach enables the high-resolution evaluation of 24 distinct scenarios with reasonable computational cost, it inherently neglects 3D spanwise cross-flow effects and atmospheric variability. Second, the current computational framework assumes a rigid blade and a continuous ice accretion process. Consequently, it does not account for aeroelastic deformations, active control system responses (e.g., blade pitching), or dynamic ice shedding events. Furthermore, transient aerodynamic phenomena such as dynamic stall are not modeled. Dynamic stall is particularly significant for iced airfoils, where massive flow separation is highly unsteady and can drastically alter aerodynamic loads. Therefore, the results presented herein represent comparative performance trends rather than exact operational predictions. Future studies employing full 3D transient simulations coupled with fluid–structure interaction (FSI), unsteady aerodynamic models, and probabilistic Uncertainty Quantification (UQ) frameworks are recommended to further elucidate these complex dynamic phenomena in cold-climate wind energy operations.

4. Conclusions

This study presented a comprehensive numerical investigation into the effects of atmospheric icing on the aerodynamic performance and power output of the NREL 5 MW reference wind turbine. By utilizing a coupled CFD-BEM approach within the FENSAP-ICE framework, ice accretion was simulated on six representative blade sections under varying liquid water content and ambient temperature conditions, representing both glaze and rime icing regimes.

The key findings of this research are summarized as follows:

• The formation of glaze ice horns induced severe flow separation, leading to a drastic reduction in lift and a substantial increase in drag. Rime ice, despite its considerable thickness, resulted in comparatively lower aerodynamic penalties due to its smoother profile. The degradation was most critical at the blade tip (Section A), which is the primary contributor to torque generation.
• Turbine power output was significantly impacted by icing. The study revealed that glaze icing conditions are more detrimental than rime icing. The total power loss ranged from 31.28% for moderate rime ice to a maximum of 40.76% for severe glaze ice (LWC = 0.50 g/m³). These results highlight that LWC acts as an amplifier of performance degradation, while temperature determines the accretion mechanism.
• The estimated annual energy production losses for a single turbine operating in a cold climate region (e.g., Northern Finland) ranged between 789 MWh and 920 MWh, corresponding to financial losses of approximately €32,000 to €42,000 per year.

This study demonstrates that atmospheric icing—particularly glaze ice at the blade tip—poses a severe threat to wind turbine efficiency and economic viability. The substantial power losses quantified herein underscore the critical necessity of implementing effective active ice protection systems (IPSs), with a strategic focus on the outer 30% of the blade span where aerodynamic performance is most vulnerable.