A method for predicting power production loss of a wind turbine due to ice formation

Abstract

A method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades of the wind turbine is disclosed. Expected ice formation on the wind turbine blades is simulated, based on generic climate data (1) and using a first ice growth model (2), and a plurality of power coefficient (Cp) curves (5) are generated corresponding to the modified blade profiles (3) representing various degrees of ice formation on the wind turbine blades. Power production loss of the wind turbine due to ice formation on the wind turbine blades is estimated, based on expected local climate data (6) related to a site where the wind turbine is located or is to be located, and on the generated power coefficient (Cp) curves (5).

Description

[0001] A METHOD FOR. PREDICTING POWER PRODUCTION LOSS OF A WIND TURBINE

[0002] DUE TO ICE FORMATION

[0003] FIELD OF THE INVENTION

[0004] The present invention relates to a method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades of the wind turbine. The method according to the invention provides an accurate prediction of the power production loss, regardless of the type of wind turbine, and regardless of the amount of available production data from the wind turbine or from similar wind turbines.

[0005] BACKGROUND OF THE INVENTION

[0006] When wind turbines are located in regions with cold climate, there is a risk that ice may be formed on the wind turbine blades when certain weather conditions are occurring, e.g. including temperatures below 0°C, a certain humidity level and / or the presence of liquid water particles, e.g. in the form of fog, mist, low clouds, etc. Such ice formation on the wind turbine blades alters the aerodynamic profile of the wind turbine blades, thus affecting the aerodynamic properties of the wind turbine blades. This, in turn, affects the ability of the wind turbine blades to extract energy from the wind, and thus the efficiency of the wind turbine, resulting in a power production loss.

[0007] Accordingly, potential ice formation on the wind turbine blades affects the expected power output from a wind turbine, and thus the revenue that can be collected from operation of the wind turbine. In some cases the climate conditions at a wind turbine site may even be such that power production losses due to ice formation on the wind turbine blades makes it unfeasible to operate wind turbines, or at least certain types of wind turbines, at the site. It is therefore desirable to be able to predict or estimate to which extent it can be expected that a given wind turbine at a given site experiences power production loss due to ice formation on the wind turbine blades. Methods for providing such predictions or estimates have been applied, e.g. applying blackbox statistical models for analysing operational data from existing an operating wind turbines. However, the predictions or estimates resulting from these previous methods may be relatively inaccurate. Furthermore, they are not applicable to new wind turbine types where little or no operational data is available.

[0008] DESCRIPTION OF THE INVENTION

[0009] It is an object of embodiments of the invention to provide a method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades, which is more accurate than prior art methods.

[0010] It is a further object of embodiments of the invention to provide a method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades, which accurately applies to any type of wind turbine.

[0011] The invention provides a method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades of the wind turbine, the method comprising the steps of:

[0012] - simulating expected ice formation on the wind turbine blades, based on generic climate data and using a first ice growth model, and generating a plurality of modified blade profiles representing various degrees of ice formation on the wind turbine blades, based on the simulated expected ice formation and on the blade profile of the wind turbine blades without ice formation, generating a plurality of power coefficient (Cp) curves, each corresponding to one of the generated modified blade profiles, obtaining expected local climate data related to a site where the wind turbine is located or is to be located, and

[0013] - estimating power production loss of the wind turbine due to ice formation on the wind turbine blades, based on the local climate data and on the generated power coefficient (Cp) curves.

[0014] Thus, the invention provides a method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades, i.e. power production loss of the kind described above.

[0015] According to the method, expected ice formation on the wind turbine blades is initially simulated. This simulation is based on generic climate data and using a first ice growth model. The climate data applied is generic in the sense that it is not related to any particular location, but rather reflects possible climate or weather conditions that might realistically occur. Thus, the generic climate data will normally be readily available in vast amounts from historical climate and data records. The generic climate data may, e.g., include various combinations of temperature, humidity, precipitation, wind speed, turbulence conditions, gust conditions, etc.

[0016] The simulation further uses a first ice growth model. In the present context the term 'ice growth model' should be interpreted to mean a model reflecting how a layer of ice is formed or grows on a surface under certain ambient conditions. The ice growth model could, e.g., be based on various physical principles governing such ice formation on a surface.

[0017] Accordingly, the simulation reflects how and to which extent a layer of ice can be expected to form on the wind turbine blades under various climate conditions defined by the generic climate data. For instance, the simulation may take into account to which extent vapour or water droplets are present under various climate or weather conditions related to temperature, humidity, wind speed, etc., as well as whether or not it can be expected that such droplets condense and stick to the surface of a wind turbine blade under these conditions, thus increasing ice formation. The simulation may further take into account how and to which extent ice previously formed on the wind turbine blade will sublimate, melt or break off under various climate or weather conditions of the kind described above. Thus, the simulation reflects an accumulated ice formation on the wind turbine blades, assuming that the wind turbine has been subjected to certain climate conditions and variations in climate conditions for a given period of time.

[0018] Based on the simulated expected ice formation and on the blade profile of the wind turbine blades without ice formation, a plurality of modified blade profiles are generated. The modified blade profiles represent various degrees of ice formation on the wind turbine blades, e.g. categorised as 'mild', 'moderate', 'severe' and 'most severe'. Since the generation of the modified blade profiles is based on the blade profile of the wind turbine blades without ice formation, the modified blade profiles represent the actual aero-dynamic blade profiles of the wind turbine with various degrees of ice formation, as predicted by the simulation. Thus, the modified blade profiles accurately reflect expected actual aero-dynamic profiles of the wind turbine blades with ice formation thereon as a result of the wind turbine, and thus the wind turbine blades, having been subjected to certain climate conditions.

[0019] Next, a plurality of power coefficient (Cp) curves are generated, each corresponding to one of the generated modified blade profiles. Accordingly, for each of the modified blade profiles, and thus, e.g., for each of the categories of ice formation described above, a Cpcurve is generated. This results in a catalogue or library of Cpcurves representing the aero-dynamic properties of the wind turbine blades under various degrees of ice formation on the wind turbine blades. In particular, the Cpcurves reflects the efficiency of the wind turbine, i.e. the ability of the wind turbine to extract energy from the wind, given that the blade profile of the wind turbine blades has been modified due to various degrees of ice formation thereon. Since the modified blade profiles are accurate, as described above, the generated Cpcurves also accurately reflect the efficiency of the wind turbine under the various degrees of ice formation caused by specified climate and weather conditions. Next, expected local climate data is obtained. The local climate data relates to a site where the wind turbine is located or is to be located. Thus, in the case that the wind turbine has already been erected, the local climate data relates to the site where the wind turbine is located. If the wind turbine has not yet been erected, the local climate data relates to a site where it is intended to erect the wind turbine.

[0020] The expected local climate data may be similar to the generic climate data, in the sense that it may include, e.g., various combinations of temperature, humidity, precipitation, wind speed, turbulence conditions, gust conditions, etc. However, the expected local climate data specifically relate to the site of the wind turbine, and thus specifically reflect climate conditions which the wind turbine may be expected to experience. For instance, the expected local climate data may indicate how many days a year it can be expected or is likely that the wind turbine is subjected to climate conditions which result in increase or decrease in ice formation, respectively, as well as at which parts of the year this is likely to occur.

[0021] Finally, power production loss of the wind turbine due to ice formation on the wind turbine blades is estimated, based on the local climate data and on the generated power coefficient (Cp) curves.

[0022] As described above, the local climate data provides information regarding the climate conditions which the wind turbine can be expected to experience, including to which extent the wind turbine is likely to be subjected to conditions resulting in increase and decrease, respectively, of ice formation on the wind turbine blades. Thus, based thereon it can be determined how large a portion of the operating time the wind turbine blades can be expected to be ice free, and how large a portion of the operating time the wind turbine blades can be expected to have ice formation thereon corresponding to each of the modified blade profiles. Accordingly, it can be determined for how large a portion of the time the wind turbine can be expected to operate according to the respective Cpcurves, including the Cpcurve corresponding to the ice free blade profile and each of the Cpcurves corresponding to the modified blade profiles. This reflects the expected power production of the wind turbine under realistic operating conditions, including to which extent the wind turbine can be expected to provide optimal power production, and to which extent the wind turbine can be expected to operate sub-optimally due to ice formation on the wind turbine blades. Accordingly, the power production loss due to ice formation on the wind turbine blades can be estimated.

[0023] Due to the accurate simulation of the expected ice formation on the wind turbine blades, and the resulting accurate Cpcurves corresponding to the blade profiles with various degrees of ice formation on the wind turbine blades, the power production loss is also accurately determined. This, e.g., provides an accurate tool for a wind turbine operator for determining whether or not it is feasible to operate a wind turbine of this type at that specific site. Furthermore, the method is applicable to any wind turbine type or any blade profile, regardless of amount of available production data for the wind turbine blade or blade profile, because the relevant ice free blade profile is simply applied when generating the modified blade profiles and the corresponding Cpcurves. Thus, accurate estimates of power production loss can also be obtained for new wind turbine types and / or new blade profiles.

[0024] The method according to the invention may be performed by means of a suitable system, e.g. comprising at least one computer, e.g. having appropriate software running thereon.

[0025] The step of generating a plurality of power coefficient (Cp) curves may comprise using aero-elastic simulations. In the present context, the term 'aero-elastic simulations' should be interpreted to mean simulations relying on aero-elasticity, i.e. which considers inertial, elastic and aero-dynamic forces acting on a structure. Thus, the aero-elastic simulations take into account the interaction between the wind turbine blades and the ambient air.

[0026] Thus, according to this embodiment, the Cpcurves corresponding to the respective modified blade profiles are generated by means of simulations taking such interaction between the wind turbine blades, with the respective modified blade profiles, and the ambient air, e.g. including simulations of the air flow along the modified blade profiles during operation of the wind turbine. For instance, the aero-elastic simulations may be in the form of Computational Fluid Dynamics (CFD) simulations.

[0027] The method may further comprise the step of calibrating the aero-elastic simulations based on historical production data originating from a vast number of operating wind turbines.

[0028] According to this embodiment, the outcome of the aero-elastic simulations is compared to available historical production data, and in the case of discrepancies there between, the aero-elastic simulations are calibrated, e.g. by adjusting relevant parameters applied in the simulation model, so as to obtain consistency between simulated data and corresponding historical production data. Since the historical production data applied for this purpose originate from a vast number of operating wind turbines, it is ensured that the calibration is statistically accurate. The calibration of the aero-elastic simulations ensure that the generated Cpcurves are even more accurate.

[0029] The calibration of the aero-elastic simulations may, e.g., include comparing the resulting Cpcurves with actual production data, in particular power output, originating from wind turbines having been subjected to similar climate conditions as the ones underlying the simulation, and / or with similar ice formation on the wind turbine blades.

[0030] The step of estimating power production loss may comprise simulating expected ice formation on the wind turbine blades based on the local climate data and using a second ice growth model.

[0031] According to this embodiment, a second ice growth model is applied for simulating expected ice formation on the wind turbine blades, under the assumption that the local climate data apply. This may be applied for determining to which extent each of the generated Cpcurves should be applied when estimating the power production loss. The first ice growth model may have a higher resolution than the second ice growth model. According to this embodiment, the first ice growth model, which is used when generating the plurality of Cpcurves based on the generic climate data, is finer and more accurate than the second ice growth model, which is applied when estimating the power production loss based on the local climate data related to the actual site where the wind turbine is located or is to be located. For instance, the spatial resolution of the first ice growth model may be higher than the spatial resolution of the second ice growth model, i.e. the first ice growth model may apply a finer spatial grid of simulation points than the second ice growth model. Thus, according to this embodiment, the outcome of the simulation using the first ice growth model may be expected to more accurately follow the actual weather and climate conditions, as well as the actual ice growth resulting therefrom, than the outcome of the simulation using the second ice growth model. However, the higher resolution has the consequence that the simulation using the first ice growth model is more computationally heavy and time consuming than the simulation using the second ice growth model.

[0032] Thus, according to this embodiment, the accurate, but computationally heavy and time consuming, ice growth model is applied when generating the modified blade profiles and the plurality of Cp curves, thus obtaining very accurate Cpcurves for the subsequent estimation of power production loss. This part of the process may be performed separately and up-front, since it is only dependent on the blade profile of the wind turbine blades and on the generic climate data. Therefore, the penalty in the form of computational load and time consumption is acceptable for this part of the process. On the other hand, for the part of the process which is site specific, i.e. depends on the local climate data, the faster and computationally lighter, but less accurate, second ice growth model is applied, thus obtaining the estimated power production loss within an acceptable time and with reasonable processing resources. Since this part of the process further relies on the accurately generated Cpcurves, it can be assumed that the accuracy of the resulting estimated power production loss is still high. Accordingly, a suitable balance is between computational load and time consumption, on the one hand, and accuracy on the other hand is obtained. The method may further comprise the step of calibrating the second ice growth model based on historical production data originating from a vast number of operating wind turbines. According to this embodiment, the second ice growth model is evaluated and adjusted in accordance with actual historical production data originating from actually operating wind turbines. Furthermore, since the historical production data originates from a vast number of wind turbines, the calibration can be assumed to be statistically significant.

[0033] For instance, the historical production data may include production data obtained during periods of time where ice formation was occurring at the wind turbines. In this case the historical production data may include information regarding the duration of such periods of time, the degree of ice formation, the actual power production, as well as the weather conditions prevailing before, during and after these periods of time. The second ice growth model may then be adjusted to match at least the duration of the icing events, i.e. the periods of time where ice formation was occurring. The calibration may comprise adjusting one or more model parameters.

[0034] Alternatively or additionally, the method may further comprise the step of calibrating the power production loss estimation based on historical production data originating from a vast number of operating wind turbines. This is similar to the embodiment described above. However, in this case the power production loss estimation, rather than the second ice growth model, is calibrated.

[0035] The step of calibrating the power production loss estimation may comprise calibrating the first ice growth model. This is similar to the calibration the second ice growth model, and the remarks set forth above in this regard are equally applicable here.

[0036] The step of calibrating the power production loss estimation may comprise adjusting one or more model parameters. The adjustment of the one or more model parameters may, e.g., ensure that the predictions of the applied models match the historical production data, e.g. with regard to duration of icing events, degree of ice formation, power production, etc., under various weather conditions.

[0037] The historical production data may include ice formation detection. This could, e.g., include data indicating that ice formation is occurring on the wind turbine blades, and flagging production data related to such periods of time. The flagged data may then be applied for calibrating the power production loss estimation. The historical production data, in particular the data related to ice formation, may be measured directly at the relevant wind turbines, by means of suitable measurement equipment, e.g. including relevant sensors and / or detectors.

[0038] The method may form part of a siting process. According to this embodiment, the wind turbine has not yet been erected, and the method is performed as part of a planning process, e.g. of a new wind farm or of erection of new wind turbines in an existing wind farm. Estimating the power production loss of a wind turbine due to ice formation on the wind turbine blades as part of a siting process may reveal whether or not it will be feasible to erect a specific type of wind turbine at a specific site with certain climate conditions. Moreover, since the method according to the invention is applicable to any wind turbine type and any blade profile, as long as the relevant blade profile without ice formation is applied when generating the modified blade profiles, this can be accurately obtained, even for new types of wind turbines and / or blade profiles, where available historical data is scarce. The method is therefore highly relevant for siting processes, where new types of wind turbines and / or blade profiles are often applied.

[0039] As an alternative, the method may be performed for an existing wind turbine. In this case the estimated power production loss may, e.g., be used for predicting the annual revenue of the wind turbine, and / or for determining whether or not modifications to the wind turbine and / or to the operation or control of the wind turbine may be relevant.

[0040] The method may further comprise the step of modifying pitch control of the wind turbine blades based on the estimated power production loss. According to this embodiment, the estimated power production loss is taken into account when controlling the wind turbine, e.g. in order to maximize the power output of the wind turbine during icing events. For instance, various pitch control strategies may be selected for the various degrees of ice formation, corresponding to the various modified blade profiles and Cpcurves. Thereby it can be ensured that the pitch control takes the modified aero-dynamical properties of the various modified blade profiles into account, thus maximizing the overall power output of the wind turbine, despite the ice formation on the wind turbine blades.

[0041] Thus, according to this embodiment, the estimated power production loss obtained by means of the method according to the invention, is applied as an input to the control system of the wind turbine, with the purpose of adjusting the applied pitch strategy in accordance with the modified blade profiles and Cpcurves in the case of expected ice formation. This minimises the power production loss. In addition, the modified pitch control, and the maximised power production resulting therefrom, may be taken into account when the power production loss is estimated.

[0042] BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The invention will now be described in further detail with reference to the accompanying drawings in which

[0044] Fig. 1 is a flow diagram illustrating a method according to an embodiment of the invention,

[0045] Fig. 2 illustrates Cpcurves for various degrees of ice formation on the wind turbine blades of two different types of wind turbines, and

[0046] Fig. 3 illustrates ice growth modelling on a wind turbine blade based on climate data. DETAILED DESCRIPTION OF THE DRAWINGS

[0047] Fig. 1 is a flow diagram illustrating a method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades according to an embodiment of the invention. In a first part of the process, generic climate data 1 related to cold climate is supplied to a first ice growth model 2. The generic climate data 1 is not specifically linked to a specific site or location, but rather reflect climate conditions that may generally occur in regions with cold climate.

[0048] The first ice growth model 2 is applied for performing simulations of ice growth on a wind turbine blade with a given blade profile, given that certain climate conditions defined by the generic climate data 1 is present. The simulations are performed using Computational Fluid Dynamics (CFD), and thus accurately take into account how air moves along the surface of the wind turbine blade and how droplets or moisture in the air condensate on the surface of the wind turbine blade as ice formation. Accordingly, the simulations performed using the first ice growth model 2 can be assumed to very accurately reflect ice formation on the wind turbine blade under given climate and weather conditions.

[0049] The simulations performed by means of the first ice growth model 2 result in a plurality of modified blade profiles 3 representing various degrees of ice formation on the wind turbine blade. Thus, each of the modified blade profiles 3 represents a wind turbine blade with a certain layer of ice formed on the surface thereof. For instance, the modified blade profiles 3 may represent categories of ice formation on the wind turbine blade, such as 'mild', 'moderate', 'severe' and 'most severe'.

[0050] Ice formation on a wind turbine blade affects the aero-dynamic profile of the wind turbine blade, thus modifying the aero-dynamic properties of the wind turbine blade. Accordingly, the ability of the wind turbine blade to extract energy from the wind is affected. Thus, the modified blade profiles 3 reflect what the wind turbine blade looks like with various degrees of ice formation thereon, as well as the aero-dynamic properties thereof, i.e. the modified blade profiles 3 reflect how the various degrees of ice formation affect the aero-dynamic properties of the wind turbine blade.

[0051] The modified blade profiles 3 are supplied to an aero-elastic simulation model 4 which generates a plurality of power coefficient (Cp) curves 5, where each of the Cpcurves 5 corresponds to one of the modified blade profiles 3, and thus to a specific degree of ice formation on the wind turbine blade. Thus, for each of the modified blade profiles 3, the aero-elastic simulation 4 calculates the performance of the modified blade profile 3 as a function of wind speed.

[0052] The steps described above may be performed separately, e.g. centrally, and without any knowledge or information regarding the specific site or location where the wind turbine is located or is to be located. For instance, the steps may be performed 'up-front' before a specific site for the wind turbine has been selected. Accordingly, the computationally heavy and time consuming CFD simulation can be applied without consideration to the processing resources and time available once the site has been selected, and the high resolution associated therewith can therefore be obtained without severe penalty.

[0053] Once a site for the wind turbine has been selected, expected local climate data 6 related to the selected site is obtained. In the case that the wind turbine is an existing wind turbine that has already been erected, the selected site is the site where the wind turbine is actually located. In the case that the wind turbine has not yet been erected, the selected site is a site where it is considered to erect the wind turbine, e.g. a candidate site.

[0054] Relevant climate time series 7 are extracted from the local climate data 6, and a suitable resolution, in the form of optimal mesoscale grid points 8, is selected. The extracted local climate time series 7 are supplied to a second ice growth model 9. Similarly to the first ice growth model 2, the second ice growth model 9 is applied for simulating expected ice formation on the wind turbine blades. However, the second ice growth model 9 applies the local climate data 6, 7 rather than the generic climate data 1, and the output of the second ice growth model 9 is therefore specifically related to the site of the wind turbine. The selected mesoscale grid points 8 ensure a reasonable balance between resolution and accuracy at the one hand, and processing load and time consumption on the other hand.

[0055] The output of the second ice growth model 9 includes information regarding to which extent ice formation on the wind turbine blades can be expected under the climate conditions prevailing at the site of the wind turbine, i.e. as specified by the local climate data 6. More particularly, the output of the second ice growth model 9 provides information regarding to which extent ice formation within each of the categories represented by the respective modified blade profiles 3, and thus by the respective Cpcurves 5, can be expected, given the expected climate conditions defined by the local climate data 6.

[0056] The output of the second ice growth model 9 is supplied to a power production loss estimator 10, along with the Cpcurves 5. Based thereon, the power production loss estimator 10 estimates a power production loss of the wind turbine caused by the expected ice formation on the wind turbine blades, as specified by the second ice growth model 9, given that the wind turbine is located at the selected site, i.e. subject to the local climate data 6. Since the power production loss estimator 10 applies the accurate Cpcurves 5, the resulting estimated power production loss can be expected to be very accurate. The estimated power production loss is output as a site specific ice assessment report 11.

[0057] Various parts of the power production loss estimation described above may be calibrated in the manner described below. This could, e.g., include calibrating the first ice growth model 3, the aero-elastic simulations 4, the second ice growth model 9 and / or the power production loss estimator 10.

[0058] Historical production data 12 originating from a vast number of operating wind turbines is obtained, as well as climate data 13 corresponding to the historical production data 12. From the historical production data 12, outliers that appear to relate to icing events, i.e. time periods where ice formation occurs on the wind turbine blades, are identified and labelled, and the production data related thereto is output as ice labelled data 14. The ice labelled data 14 and the climate data 13 is supplied to a first calibrating unit 15 where it is compared to simulations performed by the first ice growth model 2 and / or by the second ice growth model 9 in order to evaluate whether or not the simulations provided by the ice growth models 2, 9 are consistent with the actual production data. In the case of discrepancies, relevant model parameters of the ice growth models 2, 9 are adjusted so as to match the models 2, 9 to the actual production data, thus calibrating the ice growth models 2, 9.

[0059] Furthermore, the actual power production loss 16 experienced by the operating wind turbines during the icing events is estimated from the ice labelled data 14, and this estimated power production loss 16 is supplied to a second calibrating unit 17, along with the output from the first calibrating unit 15 and the Cptables 5. Based thereon, the power production loss estimator 10 is calibrated, and the result 18 is fed back to the power production loss estimator 10. A model diagnostics report 19 regarding the accuracy of the respective models may be output.

[0060] Fig. 2 illustrates simulated Cpcurves for various degrees of ice formation on the wind turbine blades of two different types of wind turbines. For instance, the blade profiles of the wind turbine blades of the two wind turbine types may differ from each other. The Cpcurves of a first wind turbine type are presented in the left panel of Fig. 2, and the Cpcurves of a second wind turbine type are presented in the right panel of Fig. 2. The Cpcurves represent various degrees of ice formation on the wind turbine blades and may, e.g., have been generated in the manner described above with reference to Fig. 1.

[0061] For each wind turbine type, five Cpcurves are shown. The solid line represents the blade profile without ice formation thereon, the short-dashed line represents the blade profile with mild ice formation thereon, the dotted line represents the blade profile with moderate ice formation thereon, the dashed-dotted line represents the blade profile with severe ice formation thereon, and the long- dashed line represents the blade profile with most severe ice formation thereon. Thus, each of the Cpcurves represents a blade profile with a certain degree of ice formation thereon. More particularly, the respective Cpcurves reflect the aero-dynamic properties of the corresponding modified blade profiles, notably the efficiency of the blade profiles in terms of extracting energy from the wind, and thus the ability of the wind turbine to produce power.

[0062] For both of the illustrated wind turbine types it can be seen that each of the modified blade profiles, representing various degrees of ice formation on the wind turbine blades, results in lower power production than the blade profile without ice formation thereon, in particular at low wind speeds. Thus, ice formation on the wind turbine blades adversely affects the aero-dynamic properties of the wind turbine blades and reduces the efficiency of the wind turbine blades, thus resulting in power production loss. It can further be seen that this effect becomes more significant as the degree of ice formation increases.

[0063] Finally, it can be seen that the second wind turbine type illustrated in the right panel of Fig. 2 appears to be more sensitive to ice formation on the wind turbine blades than the first wind turbine type illustrated in the left panel of Fig. 2. This indicates that the first wind turbine type may be better suited for operation in cold climate conditions than the second wind turbine type.

[0064] Fig. 3 illustrates ice growth modelling on a wind turbine blade based on climate data. The lower panel illustrates estimated increase and decrease in ice formation on the wind turbine blades during a winter season, based on expected local climate data, and using an ice growth model. Data points above zero indicate that ice is building up on the wind turbine blades, i.e. an increase in ice formation, and data points below zero indicate that ice is melting, sublimating or breaking off, i.e. a decrease in ice formation. The value of the data points indicates a rate at which the ice formation is increased or decreased. The upper panel illustrates the corresponding accumulated ice formation on the wind turbine blades.

[0065] The accumulated ice formation is categorised into 'non-iced' 20, 'mild' 21, 'moderate' 22, 'severe' 23 and 'most severe' 24. Thus, when estimating power production loss of the wind turbine due to ice formation on the wind turbine blades, the estimated accumulated ice formation illustrated in the upper panel of Fig. 3 provides information regarding to which extent the various Cpcurves illustrated in Fig. 2 should be applied for estimating the expected power output of the wind turbine in the course of a year or a winter season. Based thereon, the expected power production loss can be estimated.

Claims

CLAIMS1. A method for predicting power production loss of a wind turbine due to ice formation on the wind turbine blades of the wind turbine, the method comprising the steps of:- simulating expected ice formation on the wind turbine blades, based on generic climate data (1) and using a first ice growth model (2), and generating a plurality of modified blade profiles (3) representing various degrees of ice formation on the wind turbine blades, based on the simulated expected ice formation and on the blade profile of the wind turbine blades without ice formation,- generating a plurality of power coefficient (Cp) curves (5), each corresponding to one of the generated modified blade profiles (3),- obtaining expected local climate data (6) related to a site where the wind turbine is located or is to be located, and- estimating power production loss of the wind turbine due to ice formation on the wind turbine blades, based on the local climate data (6) and on the generated power coefficient (Cp) curves (5).

2. A method according to claim 1, wherein the step of generating a plurality of power coefficient (Cp) curves (5) comprises using aero-elastic simulations (4).

3. A method according to claim 2, further comprising the step of calibrating the aero-elastic simulations (4) based on historical production data (12) originating from a vast number of operating wind turbines.

4. A method according to any of the preceding claims, wherein the step of estimating power production loss comprises simulating expected ice formation on the wind turbine blades based on the local climate data (6) and using a second ice growth model (9).

5. A method according to claim 4, wherein the first ice growth model (2) has a higher resolution than the second ice growth model (9).

6. A method according to claim 4 or 5, further comprising the step of calibrating the second ice growth model (9) based on historical production data (12) originating from a vast number of operating wind turbines.

7. A method according to any of the preceding claims, further comprising the step of calibrating the power production loss estimation (10) based on historical production data (12) originating from a vast number of operating wind turbines.

8. A method according to claim 7, wherein the step of calibrating the power production loss estimation (10) comprises calibrating the first ice growth model (2).

9. A method according to claim 7 or 8, wherein the step of calibrating the power production loss estimation (10) comprises adjusting one or more model parameters.

10. A method according to any of claims 6-9, wherein the historical production data (12) includes ice formation detection (12).

11. A method according to any of the preceding claims, wherein the method forms part of a siting process.

12. A method according to any of the preceding claims, further comprising the step of modifying pitch control of the wind turbine blades based on the estimated power production loss.

Images