On Icing and Icing Mitigation of Wind Turbine Blades in Cold Climate

The enormous strain on the worldwide energy supply, environmental pollution in developing countries, and global warming etc. have begun to force a transition from nonrenewable energy resources (e.g., fossil fuel and nuclear energy) to renewable energy resources (e.g., biomass, solar, wind, geothermal, ocean, and hydro energy) [1,2]. Renewable resources currently account for approximately 10% of the global energy demand, and will grow to meet up to 60% of the global energy demand by 2050 [1]. Wind energy, mainly applied for electricity generation using wind turbines, is renewable and clean with little adverse effects on the environment. Currently the global cumulative installed wind capacity is more than 300,000 MW with a growth rate of nearly 20% per year, as shown in Fig. 1. The share of global energy contribution of wind power will increase from its current 2.6–18% by 2050 [3]. The installed wind capacity in Europe is more than 40% of the global installed wind power capacity. There are now over 120 GW of installed wind power capacity in the European Union (EU) in 2014. By 2020, the European Wind Energy Association (EWEA) estimates that 230 GW (190 GW onshore and 40 GW offshore) of wind power capacity will be installed in the EU, meeting 15–17% of the EU's electricity demand [4]. By 2050, around 50% of the EU's electricity demand will be provided by wind power [5]. Such growth requires research on wind turbine technology like improvements in design and reliability of wind turbines. Nowadays, large wind farms or wind energy projects are more often implemented in cold climates and/or at higher altitudes mainly due to two reasons. First, numerous cold climate sites around the world offer great wind energy potential and profitable wind power resources. In the north climate regions, available wind power is approximately 10% higher due to increased air density at lower temperatures [6]. Besides, fewer sites with good wind resources are available and the offshore wind development faces higher costs than expected. However, the cold climate poses stricter requirements on wind turbines due to atmospheric icing and low temperatures.

Cold climates refer to sites that may experience significant time or frequency of either icing events or low temperatures outside the normal design limits of standard wind turbines. Figure 2 shows an icing map of Europe [7]. Generally, the degree of icing is severe in north Scandinavia and parts of central Europe such as Germany and Switzerland. Although exact numbers are hard to assess, some 60 GW of installed capacity is located in cold climate areas [8]. Recent interest in offshore wind power development introduces further demands for knowledge in dealing with icing, as turbines installed in the shallow waters in northern Europe and in the coast of New England in the U.S. also face icing conditions. Figure 3 shows examples of ice accretion on wind turbines. Ice accretion on wind turbines, particularly turbine blades, can be detrimental to turbine performance, durability, and the safety of those in the vicinity of operating iced turbines [10]. Specifically, icing accretion mainly affects the following four aspects of wind turbines:

There are additional effects on wind turbines operating in cold climate. Maintenance at low temperatures and icing conditions is more time consuming. Lubricants may lose their viscosity at low temperatures. Cold start might be more difficult.

Up-to-now, needed technologies for wind turbine operating in cold climate, particularly in icing climate, mainly include ice adhesion on wind turbine blades, ice accretion modeling, ice detection, and anti-icing and de-icing methods. In this work, we present a state-of-the-art review on these needed wind turbine technologies.

Atmospheric icing is the accretion of ice or snow on structures which are exposed to the atmosphere. There are three types of atmospheric icing for wind turbines: in-cloud icing (glaze ice, hard rime ice, and soft rime ice), precipitation icing (e.g., wet snow) and frost [12]. In-cloud icing occurs when small and supercooled water droplets impinge on a surface freezing by surface nucleation. Frost appears when water vapor solidifies directly on a cool surface. It often occurs during low wind speed. Frost adhesion can be strong. Table 1 lists typical physical properties of accreted atmospheric icing and the corresponding meteorological parameters controlling atmospheric ice accretion. In general, dry icing results in rime ice containing air bubbles while wet icing always forms glaze ice. Glaze ice is normally a homogeneous transparent hard continuum solid strongly adhered to blades, while rime ice is relatively nonhomogeneous, consisting of ice and trapped air, with less adhesion strength than glaze ice. Figure 4 shows the formation of glaze ice and rime ice, respectively. More specifically, Fig. 5 indicates the separation between glaze ice, hard rime ice, and soft rime ice. The curves are shifted to the left with decreasing supercooled droplet size and increasing liquid water content (LWC).

No undisputed model has so far been developed for estimation of icing accretion on wind turbine blades. Due to the nonlinear relationships between the parameters affecting ice accretion, most models include empirical formulas and none can be solved analytically. There are several mathematical models to predict ice accretion [13-18]. The most widely used model today is the Makkonen algorithm [13] for ice accretion. Here, we will describe the Makkonen model briefly. The ice accumulation rate on wind turbine blade depends on the flux density, in other words the cross-sectional area of the blade, the velocity of the droplets and the mass concentration of droplets in the air. Because not all droplets collide with the blade and not all droplets that do collide stick or accrete, three efficiency coefficients were introduced in the Makkonen equationDisplay Formula

where α₁ is the collision efficiency, α₂ is the sticking efficiency, α₃ is the accretion efficiency, w is the mass concentration, V is the droplet velocity relative to the blade, and A is the cross-sectional area relative to the droplet velocity. The left side of Eq. (1) indicates the accumulation rate.

The collision coefficient, α₁, is the ratio of the flux density of droplets hitting the turbine blade to the total flux density. α₁ decreases from 1.0 when droplets are forced around the blade instead of hitting it. Small droplets, large cross sections and low wind velocity reduce α₁. Because α₁ depends on the droplet size, one can safely assume α₁ = 1.0 for wet snow and freezing rain as long as the blade studied is not extremely large. In other words, α₁ normally needs to be considered only for in-cloud icing with small droplets. An empirical equation was suggested by Makkonen [13] to calculate α₁.

The sticking efficiency α₂ represents the efficiency of collection of those droplets that hit the blade, i.e., the ratio of the flux density of the droplets that stick to the blade to that hit the blade. For liquid droplets it can normally be assumed that they do not bounce from the blade surface, and hence α₂ equals 1.0. For ice and snow particles, the ratio of particles that bounce off the object can be a significant share. For dry snow, α₂ is close to or equal to 0. There is presently no detailed theory for the sticking efficiency of wet snow. The available approximation methods of α₂ are empirical equations based on laboratory simulations and some field observations.

The accretion efficiency coefficient, α₃, is the ratio of icing to the flux density of droplets sticking to the blade surface. During dry growth α₃ equals 1.0 because there is no run-off. During wet growth, glaze icing, there is a liquid film on the adhesion surface and α₃ is less than 1.0. The freezing rate can be determined by the rate with which latent heat from freezing can be transported away from the surface. A heat balance over the blade surface can be used to determine the portion that freezesDisplay Formula

where Q_f is the latent heat released during freezing, Q_v is the frictional heating of air, Q_c is the sensible heat loss to air, Q_e is the heat loss due to evaporation, Q_l is the heat loss in warming the impinging supercooled water to the freezing temperature, and Q_s is the heat loss due to radiation. The terms of the heat-balance equation can be parametrized using the meteorological and structural variables. Calculations of the three coefficients, i.e., α₁, α₂, and α₃ are listed in Table 2.

There are several limitations of the Makkonen model [13]. First, it is limited by the fact that it models ice accretion on an ideal cylinder. To model ice accumulation on a wind turbine blade, one needs to break down the structure into small elements, still taking into consideration shadowing effects from other elements and ice growing elements together to form a single element. In addition, the LWC in air, the cloud droplet density and the median volume diameter are required values. However, these parameters are not routinely measured nor does there exist a satisfactorily way to measure them. The high uncertainties in estimation of these parameters tend to give uncertain α₁ values. Further, a difficulty in solving the heat balance equation is the convective heat transfer coefficient. Another limitation of the model is that it takes into account only accumulation of ice. Melting is not included, neither is sublimation.

Nowadays there are several ice accretion codes in the international aircraft icing community such as LEWINT (U.S.), ONERA (France), DRA (UK), CANICE (Canada), and more recently CIRA (Italy) based on the physics of icing. However, the main two codes for ice accretion in wind turbines are TURBICE and LEWINT.

TURBICE is a panel method based icing program used to simulate the rate and shape of ice accretion. It can be easily combined with the computational fluid dynamics based solver fluent to investigate the complex flow behavior and the aerodynamics of the blade profiles both with and without ice accretions. Homola et al. [19] adopted this combined numerical method to investigate the effects of droplet size variations on ice accretion of a 5 MW pitch controlled wind turbine blade profile (NACA 64618). As shown in Fig. 6, increasing the droplet size increases the area covered by the accreted ice and consequently affects the accreted ice growth and shape due to the larger inertia of larger diameter droplets.

LEWINT integrates the ice accretion code lewice (version 3.2.2) with American Kestrel's user interface, icing analysis tools, and automated plotting. The lewice base is an ice accretion prediction code that applies a time stepping procedure to calculate the shape of an ice accretion. The atmospheric parameters such as temperature, pressure, wind velocity, and the meteorological parameters such as LWC, droplet diameter, and relative humidity are specified as input information and used to simulate the shape of the ice accretion. lewice can perform modeling of both dry and wet (glaze) ice growth. In addition to simulating the ice accretion, LEWICE incorporates a thermal anti-icing function. In this work, LEWINT was used to simulate how a profile of an aerofoil (NACA 63415) for a 2 MW wind turbine reacts in an icing event. Some preliminary results are given in Figs. 7 and 8. It has been found that among the variables that affect ice accretion, the air temperature and wind velocity are the most important ones. The ice accretion is higher at lower temperatures and higher wind velocities. As shown in Fig. 8, there are two different zones for the effect of air temperature. The first zone is in the range between 0 and −10 °C. The characteristic of this zone is that the ice thickness grows rapidly. On the other hand, in the range between −10 and −30 °C the ice thickness seems to be more stable and the growth is less. But this difference might be due to the combination of the different variables. By comparing Figs. 8(a) and 8(b), higher medium volume diameter (MVD) values increase the ice thickness significantly, especially for large angle of attack (AOA) values. By comparing Figs. 8(b) and 8(c), increasing LWC increases the ice thickness greatly, especially at lower temperatures. As for the effect of AOA, at lower MVD values, a lower AOA value gives higher ice thickness. However, at higher MVD values such as 90 μm, the AOA does not affect the ice thickness to any significant amount.

Homola et al. [20] documented 24 direct methods and 5 indirect methods for ice detection. The direct methods detect property changes, such as mass, reflective properties, electrical or thermal conductivity, dielectric coefficient and inductance, caused by ice accretion. Indirect methods detect meteorological data such as humidity, temperature and wind velocity or detect the effects of icing such as power loss. Based on the literature [20-25], the following main points can be extracted:

It is important to note that the methods sensoring electric properties also need lightning protection.

Recently, Owusu et al. [26] proposed to detect ice accretion on wind turbines by measuring the change in capacitance and resistance due to ice accretion between two charged cylindrical probes. When ice accrets on two electrically charged cylindrical probes, the measured capacitance increases while the resistance decreases. As shown in Fig. 9, when the drag force acting on the supercooled water droplets dominates, the droplets tend to follow the air streamlines and go around the cylinders. When inertia of the supercooled water droplets dominates, the droplets deviate from the main stream and collide with the cylindrical probes. As the supercooled droplets collid with the probes and adhere to the surface, the droplets freeze and ice accrets. The ice accreted on each charged cylindrical probe interacts with the electric field resulting in an increase in the measured capacitance. Besides, as ice builds up on the two cylindrical probe surfaces, the air gap between the probes decreases. By interacting with the electric field, the resistance decreases. Therefore, this method can probe ice accretion and the rate of icing. The method can also detect the type of icing as the ice density also affects the decrease rate of the measured resistance. Although this method was validated in an icing wind tunnel, it needs more validation under a wide range of LWCs, temperatures and wind speeds before use in operating wind turbines. A disadvantage of this sensor is that it seems impossible to install the sensor at the blade tip. The measured data located at the nacelle of wind turbines might not indicate the ice accretion on the blades accurately. Obviously, more research on ice detection needs to be done in the future.

Icing mitigation techniques include anti-icing and de-icing methods. Anti-icing prevents ice to accrete on the blade while de-icing removes the accreted ice layer from the blade surface. Ice mitigation techniques can also be divided into passive and active techniques. Passive methods take advantage of the physical properties of the blade surface to eliminate or prevent ice, while active methods use external energy that is either thermal, chemical or pneumatic. For a comprehensive view of the anti-icing and de-icing systems (ADIS), please refer to Laakso and Peltola [27], Parent and Ilinca [21], Pallarol et al. [28], and Walsh [29] and others. Some anti-icing and de-icing techniques or systems are described in the literature, e.g., Refs. [30-34]. The main passive and active ice mitigation techniques and their advantages and disadvantages are presented in Table 3. The first five ice mitigation techniques (i.e., chemicals, black paint, special surface coatings, flexible blades, and active pitching) in Table 3 are passive methods while the last five techniques (i.e., air layer, microwave, electrothermal, hot air, and flexible pneumatic boots) are active methods. It might be more efficient to use a combination of passive and active methods. Here, two popular ice mitigation techniques, i.e., electrothermal and hot air, are briefly described. Also a promising technique, which is very hot in research, i.e., surface coatings (ice-phobic coating or hydrophobic coating), is presented.

Electrothermal ADISs consist of electrical heating elements embedded inside the blade or laminated on the surface to heat the blade, especially the leading edge, in order to prevent ice accumulation or remove the ice. Companies like Kelly Aerospace and EcoTEMP have developed electrothermal systems for small aircraft or turbine blades. The electrical heating elements can be used in the form of thermal pads, electrically heated foils, heating wires, and metal or carbon fibers. Recently, Mohseni and Amirfazli [35] proposed implementation of discrete constantan thermal elements with a specific pattern inside the composite airfoil, as shown in Fig. 10(a). More thermal elements/wires can be embedded in the leading edge region (not shown in Fig. 10(a)). In other words, the wires at the leading edge region are embedded closer than those of the area beyond the leading edge region to increase the heat flux at this region. The leading edge area and the top blade surface can be used to indicate if the surface temperature is sufficiently high to prevent ice accretion by the electric heating.

Hot air ADISs work by installing a heater and blower in the root of the turbine blade which blows heated air through a channel that spans the leading edge of the blade [36]. The system is available for ENERCON wind turbines [32]. As shown in Fig. 11, in each rotor blade, a fan heater installed on additional webs near the blade flange heats up the air inside the rotor blade and forces the warm air to flow along the blade's leading edge to the blade tip and then back between the main webs to the blade flange for reheating and recirculation. In this way, the leading edge is heated up to the freezing temperature and allows the accreted ice on the blade to melt. Figure 12 shows the comparison of the energy metering data for wind turbines with and without hot air de-icing at Dragaliden, Sweden. The hot air ice mitigation system produced significantly more energy than the corresponding wind turbine without ice mitigation, with a 50% average increase in power production.

Ice-phobic coating prevents ice from sticking to the surface while hydrophobic coating allows ice removal easily. The underlying mechanism is that these coatings have low adhesion between the surface and the ice/droplets. Thus reducing the ice adhesion on surfaces is the key for anti-icing and de-icing on surfaces. Generally, ice adhesion decreases when the water contact angle increases [37], as shown in Fig. 13, but this trend only applies for smooth surfaces. When the water contact angle on the smooth surfaces or on the smooth coatings is larger than 140–150 deg, the adhesion strength is very low and no ice can form on this kind of surfaces. However, currently it is impossible to have such high contact angles on smooth surfaces. Contact angles higher than 120 deg can only be realized by a surface with low surface energy in combination with a certain structured topography [38]. As superhydrophobic coatings always show a certain roughness with microstructure or nanostructure, they might not be able to reduce ice adhesion and therefore anti-icing efficiently because ice adhesion is strongly linked with surface roughness and microstructure and nanostructure. Figure 14 shows ice adhesion strength for superhydrophilic, hydrophilic, hydrophobic, and superhydrophobic surfaces [39]. The ice adhesion strength on the superhydrophobic and superhydrophilic surfaces is almost the same. Therefore, a superhydrophobic surface cannot reduce the ice adhesion. The ice adhesion strength for the superhydrophobic surface is over 10 times larger than that of the smooth hydrophobic surface. When lowering the temperature, water molecules adsorb at the wall of the surface texture and condensation occurs inside the surface texture. That means, water penetrates partially or even completely into the surface texture. When the liquid water freezes, the ice, and the surface texture are mechanically interlocked and therefore results in a high ice adhesion for the superhydrophobic surface. Figure 15 also shows that superhydrophobic coating does not necessary have a low ice adhesion [40]. The adhesion-reduction-factor (ARF), although with a large deviation, is around unity. It means almost no reduction in ice adhesion for the superhydrophobic surfaces. Therefore, superhydrophobic coating can not be directly used as ice-phobic coatings. However, smooth surfaces with high contact angles are very promising for ice mitigation, but only if these surfaces are reliable and durable.

Wind turbine scales up in both blade size and power capacity since its introduction. Scaling up turbines to lower costs has been effective so far. Now the blade size is around 125 m, and the power capacity is around 5 MW. The blade size will continue to increase up to 250 m with corresponding power capacity of 20 MW [42] (Fig. 16). Scaling up in blade size presents new challenges for ice detection and ice mitigation. For example, the ice sensors are normally mounted on the nacelle of the turbine. However, as the blade size increases, it becomes more and more necessary to mount ice sensors on the blade instead of the nacelle because the blade has a large swept area and the blade tips are more likely to reach the low clouds. The ice data measured on the nacelle cannot show the actual ice accretion on the blade any more. Besides, as the blade size increases, the anti-icing and de-icing methods may also need to scale up and require more validation over a wide range of meteorological parameters. The efficiency of some ice mitigation systems, such as hot air ADIS, will become lower and therefore these ice mitigation systems need to be improved or new cost-effective ice mitigation techniques need to be developed in the future. In addition, as the blade size increases, wind turbines should be equipped with lightning protection in combination with ice sensors or ice mitigation techniques. It is also very important to pay attention to ice throw to ensure safety as the blade size increases.

Other needed work in the future includes accurate icing maps, development of ice accretion models applicable for wind turbine blades, reliable ice detection methods, remote sensing of the icing conditions, cost-effective anti-icing and de-icing techniques, the influence of land-shape on the energy production potential of wind farm sites [43] and icing, etc. The effects of climate change on meteorological parameters and icing on wind turbines also need to be performed [44].

As a renewable energy, wind energy continues to grow globally. Large wind farms are more often implemented in cold climates due to larger wind energy potential. However, the cold climate poses stricter requirements on wind turbines due to atmospheric icing and low temperatures. Icing accretion affects the aerodynamics, leads to mass imbalance, and therefore reduces power production.

There are mainly three types of atmospheric icing for wind turbines: in-cloud icing (glaze ice, hard rime ice, and soft rime ice), precipitation icing (e.g., wet snow) and frost. Their formation depends on meteorological parameters such as air temperature, wind speed, liquid content in air, droplet size, etc. Normally, a denser ice type shows higher ice adhesion compared to a lighter one. The widely used model for ice accretion is the Makkonen model [13]. However, its applicability for wind turbine blade is limited by several factors such as that it was developed based on an ideal cylinder and it needs input parameters which unfortunately have high uncertainties.

The main two codes for computational modeling of ice accretion on wind turbine blades are TURBICE and LEWINT. LEWINT was used to simulate how a profile of an aerofoil (NACA 63415) for a 2 MW wind turbine reacts in an icing event. It has been found that among the variables that affect ice accretion, the air temperature and wind velocity are the most important. The ice accretion is higher at lower temperatures and higher wind velocities.

Three basic requirements for ice detection on large wind turbine blades are sensor positions mounted on blade tip, high sensitivity to detect small accretion, and ability to detect ice over a large area specifically for glaze icing. Reliable ice detectors need to be developed to correctly and timely activate de-icing systems.

The main passive and active anti-icing and de-icing techniques and their advantages and disadvantages were presented in this work. Electrothermal and hot air techniques are popular nowadays while special surface coating is promising for ice mitigation. It might be more effective to use a combination of passive and active ice mitigation techniques. As wind turbine blades scale up, the efficiency of some ice mitigation systems, such as hot air ADIS, becomes lower and therefore these ice mitigation systems need to be improved or new cost-effective ice mitigation techniques need to be developed in the future.

Financial support from the Swedish National Research Council and the Swedish Energy Agency was gratefully acknowledged.