System, method, and program for predicting when to perform repairs on wind turbine blades

Abstract

To provide a system that enables wind turbine blade repairs to be performed at the appropriate time. [Solution] The present invention provides a system for predicting timing for performing repairs on a wind turbine blade, the system comprising: a receiving means for receiving data representing a surface shape of a wind turbine blade; and a predicting means for predicting the timing based on the data. The predicting means is configured to select a prediction model based on an aspect of a damaged portion of the blade surface represented in the data, and to predict the timing based on the data using the selected prediction model.

Description

[Technical field]

[0001] The present invention relates to a system, method, and program for predicting when repairs should be performed on a blade of a wind turbine. [Background technology]

[0002] The nacelle of a large wind turbine has a height of about 100 m. The circumferential speed of the blade tips of a large wind turbine is about 100 to 120 m / s, and the leading edge, which is the edge of the blade that cuts the wind, wears down by about 20 μm every year. The blades of a large wind turbine are also susceptible to damage from lightning. For this reason, the blades are provided with lightning receptors to channel lightning current to the ground as a lightning protection measure. However, if the lightning receptors fail to conduct electricity due to a malfunction, the blades will be destroyed because they cannot channel the lightning current caused by a lightning strike to the ground. Therefore, the blades of a large wind turbine require regular maintenance, and if there is any damage, they need to be repaired.

[0003] 2. Description of the Related Art A device for performing maintenance on wind turbine blades is known (Patent Document 1). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] International Publication No. 2018 / 155704 Summary of the Invention [Problem to be solved by the invention]

[0005] The objective is to provide a system etc. that enables repairs of wind turbine blades at an appropriate time. [Means for solving the problem]

[0006] The present invention provides a system and the like that predicts the timing for repairing a wind turbine blade using data representing the surface shape of the wind turbine blade.

[0007] The present invention provides, for example, the following items. (Item 1) A system for predicting when to perform repairs on a wind turbine blade, comprising: receiving means for receiving data representative of a surface shape of a blade of a wind turbine; A prediction means for predicting the timing based on the data; A system comprising: (Item 2) The prediction means includes: selecting a predictive model based on the characteristics of the damaged portion of the surface of the blade as indicated in the data; predicting the timing based on the data using the selected predictive model; and The system according to the preceding item, wherein the system is configured to perform the following: (Item 3) The system of any one of the preceding claims, wherein the aspect of the damaged portion includes an erosion pattern of the damaged portion. (Item 4) Selecting the predictive model comprises: selecting a first predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction perpendicular to the surface; selecting a second predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction parallel to the surface; The system according to any one of the preceding claims, comprising at least (Item 5) 2. The system of claim 1, wherein the data representative of the surface shape of the wind turbine blade includes a depth, a width and a position of a damaged portion of the surface of the blade. (Item 6) The receiving means further receives environmental data representing an environment around the wind turbine; The system according to any one of the preceding claims, wherein the prediction means predicts the timing based on the data and the environmental data. (Item 7) The system of any one of the preceding claims, wherein the environmental data includes at least one of wind speed data, rainfall data, raindrop data, UV data, and data representing a location. (Item 8) The receiving means further receives material data representative of a material of at least a portion of the blade; The system of any one of the preceding claims, wherein the prediction means predicts the timing based on the data, the environmental data, and the material data. (Item 9) The system of any one of the preceding claims, wherein the material data includes at least one of paint data representing a material of paint on the surface of the blade, and putty data representing a material of putty on the surface of the blade. (Item 10) 1. A method for predicting when repairs should be made to a blade of a wind turbine, comprising: receiving data representative of a surface profile of a blade of a wind turbine; predicting said timing based on said data; The method includes: (Item 10A) 11. The method according to claim 10, comprising the features according to any one of the preceding claims. (Item 11) A program for predicting a timing for repairing a blade of a wind turbine, the program being executed in a computer system having a processor, the program comprising: receiving data representative of a surface profile of a blade of a wind turbine; predicting said timing based on said data; A program causing the processor to perform a process including the steps of: (Item 11A) 12. The program according to item 11, comprising any one of the features according to the preceding items. (Item 11B) A non-transitory computer-readable storage medium storing the program according to item 11 or item 11A. Effect of the Invention

[0008] According to the present invention, it is possible to provide a system for predicting the timing for repairing wind turbine blades, thereby making it possible to repair wind turbine blades at the appropriate time and reducing the effort and cost involved in repairs. [Brief description of the drawings]

[0009] [Figure 1A] A diagram explaining the procedure for performing maintenance on wind turbine blades. [Figure 1B] A diagram explaining the procedure for performing maintenance on wind turbine blades. [Figure 1C] A diagram explaining the procedure for performing maintenance on wind turbine blades. [Figure 1D] A diagram explaining the procedure for performing maintenance on wind turbine blades. [Diagram 2] FIG. 1 shows an example of a flow of the system 100 of the present invention. [Diagram 3] FIG. 1 shows an example of the configuration of a system 100 according to the present invention. [Figure 4A] A diagram showing an example of data representing the surface shape of a blade plotted on a two-dimensional plane. [Figure 4B] FIG. 1 shows an example of the structure of a neural network model for constructing a prediction model that can be used by the prediction means 120. [Diagram 5] Schematic showing two examples of erosion patterns [Figure 6] FIG. 1 is a diagram showing an example of the configuration of an information processing device 200 that implements the system 100. [Figure 7A] A flow chart showing an example of a process 700 in the system 100 of the present invention. [Figure 7B] A flowchart showing an example of the process of step S702 in process 700. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0011] 1. Wind turbine blade maintenance The inventors of the present invention have been actively researching a device for performing maintenance on wind turbine blades. The device is for autonomously performing maintenance on wind turbine blades.

[0012] With reference to Figures 1A to 1D, the procedure for performing maintenance on wind turbine blades using this device will be described. Figures 1A to 1D (a) show the front of the wind turbine 10, and Figures 1A to 1D (b) show the right side of the wind turbine 10. Figures 1A to 1D (b) show only the blade 11 that is the subject of maintenance, and the other two blades 12, 13 are omitted. The wind turbine shown in Figures 1A to 1D (a) is a wind turbine that rotates clockwise on the drawing, and the straight edge of each blade becomes the leading edge.

[0013] In this specification, the term "wind turbine" refers to a device that obtains power by catching wind. An example of a wind turbine is a wind power generator.

[0014] In this specification, "maintenance" refers to inspection or maintenance of an object. Maintenance includes repair. Examples of maintenance include imaging the surface of an object, inspecting the continuity of the lightning receiving part of the object, cleaning the surface of the object, grinding and / or polishing the surface of the object, applying paint to the surface of the object, applying a material such as putty, adhesive, sealant, etc. to the surface of the object. Repair may include cleaning a damaged part to remove foreign matter, applying a material such as putty, adhesive, sealant, etc. to the damaged part after cleaning, polishing the applied material so as to smoothly connect with the undamaged part, and applying paint to the polished surface or attaching a protective material to the polished surface.

[0015] FIG. 1A shows the device 50 in a preparatory stage before it is attached to the blade 11 of the wind turbine 10.

[0016] Maintenance of the blade 11 is performed with the blade 11 positioned so that it extends vertically downward. This is the same state as the blade is in in the conventional method of performing maintenance by a worker moving over the blade along a rope stretched over the blade, and is therefore easily accepted at existing work sites. When the blade 11 is positioned so that it extends vertically downward, the leading edge is inclined at about 5 degrees with respect to the vertical, as shown in FIG. 1A(b).

[0017] In a preparation stage before the device 50 is attached to the blades 11 of the wind turbine 10, the ropes 20 are installed on the wind turbine 10. Here, the manner in which the ropes 20 are installed on the wind turbine 10 does not matter. The ropes 20 can be installed at any position on the wind turbine 10 by any method.

[0018] In the following, an example will be described in which the two ropes 20 are fixed to the nacelle 14 of the wind turbine 10. Each of the two ropes 20 extends from the nacelle 14 of the wind turbine 10 to the ground, passing near the hub 15. The hub 15 is a member that rotatably couples the blades 11, 12, and 13 to the nacelle 14.

[0019] 1B-1C show the stage of placing the device 50 on the blade 11 of the wind turbine 10. FIG.

[0020] Two ropes 20 extend from the nacelle 14 of the wind turbine 10 to the device 50. The two ropes 20 are connected to the device 50.

[0021] The device 50 includes a moving means for moving along the two ropes 20. The moving means is, for example, a winch. The device 50 includes, for example, two winches, one rope connected to one of the two winches, and the other rope connected to the other of the two winches. One end of the rope 20 is fixed to the ground by, for example, a weight or the like. The device 50 can ascend along the rope 20 by winding up the rope 20 using the winch.

[0022] When the device 50 rises along the rope 20 and reaches approximately the same height as the tip of the blade 11, the spatial position of the device 50 is controlled so that it rests on the tip of the blade 11. The spatial position of the device 50 can be controlled by any method.

[0023] Preferably, the spatial position of the device 50 is controlled by pulling one or both of the two ropes 20 from the ground, since this allows the spatial position of the device 50 to be controlled safely from the ground without requiring any additional equipment.

[0024] As shown in FIG. 1C, once the device 50 reaches above the blade 11, the device 50 can be attached to the blade 11 by any mechanism.

[0025] The device 50 also includes a moving means for moving over the blade 11. The moving means is, for example, a wheel. The device 50 can move over the blade, for example, by using a winch to wind or unwind the rope 20 with the wheel in contact with the blade 11.

[0026] FIG. 1D shows device 50 moving along rope 20 over the leading edge of blade 11.

[0027] When the device 50 moves along the leading edge, the device 50 can maintain a state of being attached to the blade 11. This allows the device 50 to move on the leading edge of the blade 11 without floating up. Also, as described above, since the leading edge of the blade 11 is inclined at about 5 degrees (in some cases, about 1 degree to about 10 degrees) with respect to the vertical, gravity acting on the device 50 acts to press the device 50 against the leading edge of the blade 11, preventing the device 50 from floating up.

[0028] The device 50 performs maintenance on the blade 11 while moving over the leading edge of the blade 11 .

[0029] By controlling the winding and unwinding of the winch, the device 50 can move in the direction in which the rope 20 extends over the leading edge of the blade 11, for example, in the direction from the tip of the blade to the base of the blade and from the base of the blade to the tip of the blade. This allows the device 50 to perform maintenance while moving back and forth over the leading edge of the blade 11. Furthermore, in the unlikely event that the device 50 becomes detached from the blade 11, the rope 20 connected to the winch of the device 50 serves as a lifeline, preventing the device 50 from falling off.

[0030] Additionally, the device 50 can use a scanner to measure the surface shape of the blade 11 while moving over the leading edge of the blade 11. The scanner can be any means capable of measuring the surface shape (or profile) of the blade, such as a laser scanner, an ultrasonic scanner, etc. Alternatively to the scanner, a contact sensor can be used. The scanner can measure the surface shape with a predetermined resolution.

[0031] The surface shape of the blade measured using a scanner can accurately represent the unevenness of the blade surface. For example, if there is a damaged portion on the surface of the blade, data representing the surface shape of the blade 11 obtained by measurement using a scanner can include the depth, position, and width of the damaged portion. This makes it possible to accurately grasp the state of the damaged portion. The data representing the surface shape of the blade 11 obtained by measurement can be used by the system 100 of the present invention.

[0032] For example, when there is a damaged portion on the surface of the blade 11, conventionally, a worker approaches the blade and visually inspects it, or looks at a two-dimensional image taken by a camera, and the worker judges whether or not the damaged portion needs to be repaired based on experience. The two-dimensional image may be, for example, an image taken by a camera mounted on a device that is attached to the blade and performs maintenance, such as the device 50, or may be an image taken by a camera mounted on a device that approaches the blade and observes the blade (for example, a flying device such as a drone). In this case, the accuracy of the judgment could not be guaranteed because it relies on the experience of the worker. In addition, the exact state of the damaged portion may not be known from the two-dimensional image, which also affects the accuracy of the judgment. In addition, in order to obtain the two-dimensional image, it is necessary to periodically attach a device to the blade or bring it close to the blade, which incurs regular costs.

[0033] As a result of extensive research, the inventors of the present invention have developed a system that can predict the timing of blade repairs by utilizing the surface shape (or profile) of the blade. This system can objectively predict the timing of repairs without relying on the experience of a skilled worker. In addition, this system can predict not only whether repairs of damaged parts are currently necessary, but also when they will be necessary in the future, eliminating the need to periodically observe the blade surface and reducing regular costs.

[0034] FIG. 2 illustrates an example of a flow through the system 100 of the present invention.

[0035] In this example, it is assumed that the device 50 described above with reference to Figures 1A to 1D is equipped with a scanner, and data representing the surface shape of the blade 11 is acquired when the device 50 is performing maintenance on the blade 11 of the wind turbine 10.

[0036] For example, suppose that the device 50 is performing maintenance on the blade 11 and finds a small area of ​​damage that does not currently require repair. The small area of ​​damage may appear in the data representing the surface shape of the blade 11.

[0037] In step S1, the device 50 provides data representing the surface shape of the blade 11 to the system 100. The manner in which the device 50 provides the data to the system 100 is not important. For example, the device 50 can transmit the data to the system 100 via wireless communication or wired communication. Alternatively, for example, the device 50 can provide the data to the system 100 via a storage medium (e.g., a removable medium).

[0038] In step S1, in addition to the data representing the surface shape of the blades 11, environmental data representing the environment around the wind turbine 10 (e.g., wind speed, rainfall, raindrops, UV rays, air temperature, water temperature, air pressure, lightning strikes, location, etc. at the site where the wind turbine 10 is built) may also be provided to the system 100, and alternatively or in addition, material data representing the material of at least a part of the blades 11 of the wind turbine 10 (e.g., putty or paint applied to the blades 11) may also be provided to the system 100.

[0039] When data representative of the surface geometry of the blades 11 (and, optionally, environmental data and / or material data) is provided to the system 100, the system 100 predicts, based on the provided data, when repairs should be performed on the blades 11 of the wind turbine 10. In particular, the system 100 predicts when repairs should be performed on damaged areas indicated in the data representative of the surface geometry of the blades 11.

[0040] The system 100 can predict when repairs should be performed in any time unit. For example, the system 100 can predict when repairs should be performed in daily, weekly, biweekly, monthly, bimonthly, six-monthly, or yearly units. The system 100 can select an appropriate prediction model to perform predictions in the desired time unit.

[0041] The timing predicted by the system 100 is provided to the terminal device of the user U in step S2. The manner in which the system 100 provides the output to the terminal device of the user U is not important. For example, the system 100 can transmit the output to the terminal device of the user U by wireless communication or wired communication. Alternatively, for example, the system 100 can provide the output to the terminal device of the user U via a storage medium (e.g., a removable medium). The user U can be, for example, a maintenance company that performs maintenance on the wind turbine 10. Alternatively, the user U can be a business operator that operates the wind turbine 10.

[0042] This allows the user U to know when to repair a small damaged portion that is found during maintenance of the blade 11 and does not currently require repair, and to repair the damaged portion at an appropriate time. This can eliminate the need for regular inspections to see if there is any damaged portion, leading to reduced maintenance costs.

[0043] Furthermore, since the system 100 predicts the timing for repairs based on objective data, even an inexperienced user U can understand when the damaged portion should be repaired.

[0044] In the above example, the device 50 measures the surface shape of the blade 11 while performing maintenance on the blade 11, but the present invention is not limited to this. For example, a means other than the device 50 may measure the surface shape of the blade. For example, a flying device such as a drone equipped with a measuring device such as a scanner may approach the blade 11 and measure the surface shape of the blade. At this time, the measurement data is provided to the system 100 from the flying device.

[0045] The above-described system 100 may have, for example, the configuration described below.

[0046] 2. System configuration for predicting when repairs should be performed FIG. 3 shows an example of the configuration of a system 100 of the present invention.

[0047] The system 100 comprises a receiving means 110 and a predicting means 120 .

[0048] The receiving means 110 is configured to receive data representing the surface shape of the wind turbine blade. The data representing the surface shape of the wind turbine blade is data measured by a measuring means. The measuring means may be, for example, a measuring means such as a scanner provided in the above-mentioned device 50, or a measuring means such as a scanner provided in a measuring device other than the device 50.

[0049] The receiving means 110 may, for example, receive data representing the surface shape of the wind turbine blade from the device 50 or from a measuring device separate from the device 50.

[0050] The data representing the surface shape of the wind turbine blade may be data representing the outer shape of a cross section of the wind turbine blade. The data representing the surface shape of the wind turbine blade may be, for example, point cloud data. The data representing the surface shape of the blade may preferably also represent the relative position of the surface shape of the blade with respect to the measuring means. For example, the data representing the surface shape of the blade may be expressed by position coordinates on a two-dimensional plane with the position of the measuring means as the origin for each of a plurality of points constituting the surface shape in a specific cross section. Thereby, the data representing the surface shape of the blade in a specific cross section may be plotted or drawn in a two-dimensional plane, for example, as shown in FIG. 4A.

[0051] The data representing the surface shape of the wind turbine blade may represent the surface shape of the entire blade or may represent the surface shape of a part of the blade. Preferably, the data representing the surface shape of the wind turbine blade may represent the surface shape of at least a damaged part of the blade. The damaged part may be defined by its depth, width and position. Thus, the data representing the surface shape of the wind turbine blade may include the depth, width and / or position of the damaged part, preferably, the data representing the surface shape of the wind turbine blade may include the depth and width of the damaged part, more preferably, the data representing the surface shape of the wind turbine blade may include the depth, width and position of the damaged part. Here, the width refers to the dimension along or parallel to the surface of the wind turbine blade, and the depth refers to the dimension perpendicular to the surface of the wind turbine blade.

[0052] In data representative of the surface profile of a wind turbine blade, damaged portions can be identified from a comparison with data representative of an ideal surface profile obtained, for example, from a specification, etc. Alternatively, damaged portions can be identified from a comparison with data representative of an undamaged surface profile, for example, estimated from an undamaged portion.

[0053] In addition to the data representative of the surface shape of the wind turbine blades, the receiving means 110 may further receive environmental data representative of the environment around the wind turbine and / or material data representative of the material of at least a part of the wind turbine blades.

[0054] The environmental data representing the environment around the wind turbine may be, for example, data representing the weather conditions around the wind turbine, including, but not limited to, wind speed data, rainfall data, raindrop data, UV data, air temperature data, water temperature data, air pressure data, and lightning data. The wind speed data may be the average wind speed or maximum wind speed within a predetermined period at the location where the wind turbine is built, and may be measured, for example, by an anemometer installed at the wind turbine or in its vicinity. The rainfall data may be the average rainfall or maximum rainfall within a predetermined period at the location where the wind turbine is built, and may be measured, for example, by a rain gauge installed at the wind turbine or in its vicinity. In this specification, "rain" may be a concept that includes fog, hail, and snow in addition to the usual meaning of "rain." The raindrop data represents the size of raindrops falling at the location where the wind turbine is built, or the type of rain falling at the location where the wind turbine is built. The size of the raindrop refers to the diameter of the raindrop when the shape of the raindrop is spherical, and may refer to the maximum size of the raindrop when the shape of the raindrop is not spherical. For example, in the case of drizzle, the size of the raindrop is less than about 0.5 mm, while in the case of normal rain, the size of the raindrop is about 1.0 to 2.0 mm, and in some cases, may exceed about 5.0 mm. The types of rain include, but are not limited to, heavy rain, light rain, drizzle, fog, hail, snow, and the like. The raindrop data may be measured, for example, by a raindrop meter provided at or near the windmill. The UV amount data may be the average amount of ultraviolet rays or the maximum amount of ultraviolet rays within a specified period at the location where the windmill is built, and may be measured, for example, by an ultraviolet meter provided at or near the windmill. The temperature data may be the average temperature, maximum temperature, or minimum temperature within a specified period at the location where the windmill is built, and may be measured, for example, by a thermometer provided at or near the windmill. The water temperature data may be the average water temperature, maximum water temperature, or minimum water temperature within a specified period at the location where the wind turbine is built, if the wind turbine is built offshore, and may be measured, for example, by a thermometer provided at the wind turbine or in its vicinity. The air pressure data may be the average air pressure, maximum air pressure, or minimum air pressure within a specified period at the location where the wind turbine is built, and may be measured, for example, by a barometer provided at the wind turbine or in its vicinity. The lightning strike data may be the frequency with which lightning strikes the wind turbine, or the frequency with which thunderclouds occur within a specified period at the location where the wind turbine is built, and may be measured, for example, by a lightning rod or lightning receiving unit provided at the wind turbine.The data representing these weather conditions may be obtained, for example, from weather data published by organizations such as the Japan Meteorological Agency.

[0055] The environmental data representing the environment around the wind turbine may be, for example, data representing the place where the wind turbine is built (i.e., the location). The data representing the location may include, but is not limited to, whether the location is near the sea (i.e., whether the wind turbine is within the range where seawater splashes and the like can reach by the wind), whether the location is near sandy soil (i.e., whether the wind turbine is within the range where sand and the like can reach by the wind), whether the location is near a forest (i.e., whether the wind turbine is within the range where pollen and the like can reach by the wind), and whether the location is near a factory or power plant (i.e., whether the wind turbine is within the range where soot and the like can reach by the wind). The data representing these locations may be obtained, for example, from map data published by an organization such as the Geospatial Information Authority of Japan.

[0056] The environmental data representing the environment around the wind turbine preferably includes data representing the location. By taking into account the data representing the location, it is possible to make a prediction that takes into account the presence of physical factors (e.g., seawater splashes, sand, pollen, soot, etc.) that lead to deterioration of the wind turbine blades, and the accuracy of the prediction can be improved. The environmental data representing the environment around the wind turbine preferably includes rainfall data in addition to or instead of the data representing the location. By taking into account the rainfall data, it is possible to make a prediction that takes into account the influence of rain that leads to deterioration of the wind turbine blades, and the accuracy of the prediction can be improved. The environmental data representing the environment around the wind turbine preferably includes wind speed data, particularly average wind speed data, in addition to or instead of the data representing the location and / or the rainfall data. By taking into account the wind speed data, it is possible to make a prediction that takes into account the influence of wind that leads to deterioration of the wind turbine blades, and the accuracy of the prediction can be improved. The environmental data representative of the environment around the wind turbine preferably includes raindrop data in addition to, or instead of, the data representative of the location and / or the rainfall data and / or the wind speed data. By taking into account the raindrop data, it becomes possible to make a prediction that appropriately takes into account the erosive power of raindrops.

[0057] The material data representing at least a part of the material of the wind turbine blade includes, but is not limited to, for example, paint data representing the material of the paint on the surface of the blade, putty data representing the material of the putty on the surface of the blade, protective material data representing the material of the protective material (e.g., film, etc.) on the surface of the blade, and base material data representing the material of the base material of the blade. These material data may be obtained from the past maintenance history or from the specifications of the wind turbine.

[0058] The received data is passed to a prediction means 120 .

[0059] The prediction means 120 is configured to predict when repairs should be performed on the wind turbine blades based at least on data representative of the surface geometry of the wind turbine blades.

[0060] The prediction means 120 can predict the timing for repairing the blades of the wind turbine by using, for example, a prediction model. The prediction model can be a trained model constructed by machine learning. The prediction model can be constructed using any machine learning model. The machine learning model can be, for example, a neural network model.

[0061] FIG. 4B shows an example of the structure of a neural network model for constructing a predictive model that can be used by the predictor 120.

[0062] The neural network model 600 has an input layer, an intermediate layer, and an output layer. The number of nodes in the input layer of the neural network model 600 corresponds to the number of dimensions of the input data (e.g., the number of dimensions of the data received by the receiving means 110). The number of nodes in the output layer of the neural network model corresponds to the number of dimensions of the output data. The output value representing the timing to perform repair of the wind turbine blades may be, for example, one-dimensional. The intermediate layer of the neural network may have any architecture. The intermediate layer may have any number of layers, and each layer of the intermediate layer may have any number of nodes.

[0063] The neural network model may be trained in advance using data representing previously measured surface shapes of the blade and values ​​indicating when previous repairs were made to the blade. The training process is a process of calculating weight coefficients for each node in the intermediate layer of the neural network model using data representing the previously measured surface shapes of the blade and values ​​indicating the interval between when the blade was repaired and when it was measured.

[0064] The learning process is, for example, supervised learning, in which, for example, data representing the measured surface shape of the blade is used as input supervised data, a value indicating the interval between the time when the blade was repaired and the time when the blade was measured is used as output supervised data, and weight coefficients for each node in the intermediate layer of the neural network model are calculated using data for multiple blades or multiple repairs, thereby constructing a predictive model capable of correlating data representing the measured surface shape of the blade with a value indicating the time interval between measurement and repair.

[0065] At this time, by setting the time interval to a desired time unit, a prediction model capable of predicting timing in a desired time unit can be constructed. For example, if it is desired to predict timing in daily units, the time interval can be set to a daily unit. For example, if it is desired to predict timing in monthly units, the time interval can be set to a monthly unit.

[0066] For example, a set of (input teacher data, output teacher data) for supervised learning may be (data representing the surface shape measured from the first blade, a value indicating the interval between when the first blade was repaired and when it was measured), (data representing the surface shape measured from the second blade, a value indicating the interval between when the second blade was repaired and when it was measured), ... (data representing the surface shape measured from the i-th blade, a value indicating the interval between when the i-th blade was repaired and when it was measured), ... etc. When data representing the surface shape measured from a blade to be repaired is input to the input layer of the neural network model that has been trained in this way, a value indicating how often repairs should be made to that blade after measurement is output to the output layer.

[0067] For example, if the receiving means 110 receives, in addition to the data representative of the surface geometry of the wind turbine blades, environmental data representative of the environment around the wind turbine and / or material data representative of the material of at least a part of the wind turbine blades, the predicting means 120 can predict the timing for repairing the wind turbine blades based also on the environmental data and / or the material data. In this case, the predictive model used can be trained to be able to correlate the data representative of the surface geometry of the wind turbine blades, the environmental data and / or the material data, and a value representative of the time interval between measurement and repair.

[0068] In this case, a set of (input teacher data, output teacher data) for supervised learning may be ((data representing the surface shape measured from the first blade, environmental data and / or material data of the first blade), a value indicating the interval between when the repair was performed on the first blade and when the measurement was performed), ((data representing the surface shape measured from the second blade, environmental data and / or material data of the second blade), a value indicating the interval between when the repair was performed on the second blade and when the measurement was performed), ... ((data representing the surface shape measured from the i-th blade, environmental data and / or material data of the i-th blade), a value indicating the interval between when the repair was performed on the i-th blade and when the measurement was performed), ... etc. When data representing the surface shape measured from the blade to be repaired and the environmental data and / or material data of the blade are input to the input layer of the neural network model in which the learning process has been performed in this manner, a value indicating how often repairs should be performed on that blade after measurement is output to the output layer.

[0069] In particular, it is preferable to use raindrop data as the environmental data. For example, even with the same amount of rainfall, the speed at which damage to the blade due to rainfall progresses may differ if the size of the raindrops or the type of rainfall differs. This is because the larger the size of the raindrops, the greater the erosive power they impart to the blade, and the smaller the size of the raindrops, the smaller the erosive power they impart to the blade. By using the raindrop data, it is possible to make a prediction that appropriately takes into account the erosive power of raindrops, and to make a prediction with higher accuracy than when only the amount of rainfall is considered.

[0070] The prediction means 120 may identify a damaged portion of the blade indicated in the data representing the surface shape of the blade, and use an appropriate prediction model according to the state of the damaged portion. That is, the prediction means 120 can select a prediction model based on the state of the damaged portion of the blade, and predict the timing to repair the wind turbine blade using the selected prediction model.

[0071] The aspect of the damaged portion includes, but is not limited to, for example, the erosion pattern of the damaged portion. The erosion pattern refers to a pattern in which the damaged portion expands due to erosion.

[0072] FIG. 5 shows two examples of erosion patterns diagrammatically. In FIG. 5, a cross-sectional view of the blade 11 is shown. In FIG. 5, (1) represents the first stage, (2) represents the second stage, (3) represents the third stage, and (4) represents the fourth stage. The erosion pattern on the left side of FIG. 5 shows the first erosion pattern, and the erosion pattern on the right side of FIG. 5 shows the second erosion pattern. In the first stage, shown as (1), a primary damage site P is present. The first and second erosion patterns can be distinguished according to how the erosion extends from the primary damage site P.

[0073] In the first erosion pattern, shown on the left side of FIG. 5, the primary damage P expands in a direction perpendicular to the surface of the blade 11 (top-bottom direction in FIG. 5) as shown in (3). The expanded damage has a depth > width, typically the depth is at least twice the width, and more typically the depth may be at least three times the width. In the first erosion pattern, erosion expands from the expanded damage in a direction parallel to the surface of the blade 11 (left-right direction in FIG. 5) as shown in (4). This results in a deep and wide damage portion.

[0074] In the second erosion pattern shown on the right side of FIG. 5, the primary damage P expands in a direction parallel to the surface of the blade 11 (left-right direction in FIG. 5) as shown in (2). The expanded damage has a width>depth, typically the width is more than twice the depth, and more typically the width is more than three times the depth. In the second erosion pattern, erosion forms a secondary damage in the expanded damage in a direction perpendicular to the blade surface (up-down direction in FIG. 5) as shown in (3), and the secondary damage expands in a direction parallel to the surface of the blade 11 (left-right direction in FIG. 5). This results in the formation of a deep and wide damaged portion.

[0075] These differences in erosion patterns result in differences in the form of the damaged portion that is formed, and thus the timing for repair may also differ. By identifying these differences in erosion patterns and using an appropriate prediction model, it is possible to predict the timing for repair with a higher degree of accuracy. For example, in the first stage of (1), the loss in power generation efficiency of the wind turbine is, for example, -1%, and the urgency of repair is low, whereas in the second stage of (2), the loss in power generation efficiency of the wind turbine is, for example, -2%, in the third stage of (3), the loss in power generation efficiency of the wind turbine is, for example, -3%, and in the fourth stage of (4), the loss in power generation efficiency of the wind turbine is, for example, -5%, and the urgency of repair gradually increases.

[0076] For example, the predictor 120 can compare the width and depth of the lesion extending from the primary lesion and, if depth>width, identify a first erosion pattern, and, if width>depth, identify a second erosion pattern.

[0077] For example, the prediction means 120 selects a first prediction model for a damaged portion having a first erosion pattern, and predicts the timing to perform repair using the first prediction model. The first prediction model is a trained model that has been trained using the teacher data as described above for the damaged portion having the first erosion pattern.

[0078] For example, the prediction means 120 selects a first prediction model for a damaged portion having a second erosion pattern, and predicts the timing of repair using the first prediction model. The first prediction model is a trained model that has been trained using the teacher data as described above for the damaged portion having the first erosion pattern.

[0079] In the above example, only the first pattern and the second pattern are described, but the erosion patterns are not limited to these. In addition to these erosion patterns, other erosion patterns may be included.

[0080] For example, the prediction means 120 may compare the width and depth of the damaged portion (e.g., the damaged portion expanded from the primary damaged portion), and if the depth>2×width, identify it as the first erosion pattern, if the width>2×depth, identify it as the second erosion pattern, and identify the rest as the third erosion pattern. In this case, the prediction means 120 may select a third prediction model for the damaged portion having the third erosion pattern, and predict the timing to perform repair using the third prediction model. The third prediction model is a trained model that has been trained using the teacher data as described above for the damaged portion having the third erosion pattern.

[0081] The predicted timing may be output from the system 100, for example, provided to personnel performing maintenance on the blade so that the personnel knows when to perform maintenance on the blade. For example, the system 100 may provide an alert (e.g., a visual or audible alert) to the personnel when it is time to perform maintenance on the blade.

[0082] The above-described system 100 can be implemented as, for example, any information processing device. The information processing device may be, for example, a server device or a user's terminal device (for example, a personal computer, a tablet, a smartphone, etc.).

[0083] FIG. 6 shows an example of the configuration of an information processing device 200 in which the system 100 is implemented.

[0084] The information processing device 200 includes an interface unit 210, a processor unit 220, and a memory unit 230. The information processing device 200 can be connected to a database unit 300.

[0085] The interface unit 210 exchanges information with the outside of the information processing device 200. The processor unit 220 of the information processing device 200 can receive information from the outside of the information processing device 200 via the interface unit 210, and can transmit information to the outside of the information processing device 200. The interface unit 210 can exchange information in any format. For example, the above-mentioned device 50 or the measurement device can communicate with the information processing device 200 via the interface unit 210. For example, when the information processing device 200 is a server device, a user's terminal device can communicate with the information processing device 200 via the interface unit 210.

[0086] The processor unit 220 executes the processing of the information processing device 200 and controls the operation of the entire information processing device 200. The processor unit 220 reads out a program stored in the memory unit 230 and executes the program. This makes it possible for the information processing device 200 to function as a system that executes desired steps. The processor unit 220 may be implemented by a single processor or by multiple processors.

[0087] The memory unit 230 stores a program required to execute the processing of the information processing device 200, data required to execute the program, and the like. The memory unit 230 may store a program for causing the processor unit to execute a process of predicting the timing to repair the blades of a wind turbine (for example, a program for realizing the process shown in Figs. 7A and 7B described later). Here, there is no limit to how the program is stored in the memory unit 230. For example, the program may be pre-installed in the memory unit 230. Alternatively, the program may be installed in the memory unit 230 by being downloaded via a network. Alternatively, the program may be stored in a non-transitory computer-readable storage medium. The memory unit 230 may be implemented by any storage means.

[0088] For example, a prediction model used to predict the timing for repairing a blade of a wind turbine may be stored in the database unit 300. The processor unit 220 can, for example, select and acquire a prediction model stored in the database unit 300 and predict the timing using the prediction model.

[0089] The database unit 300 may store training data for constructing a prediction model. The training data may be, for example, data representing a previously measured surface shape of a blade and data indicating when the blade was repaired. In addition to these, the training data may include previously measured environmental data and / or material data.

[0090] The receiving means 110 of the system 100 may be implemented by an interface unit 210 and / or a processor unit 220. The predicting means 120 of the system 100 may be implemented by the processor unit 220.

[0091] In the example shown in FIG. 6, the database unit 300 is provided outside the information processing device 200, but the present invention is not limited to this. At least a part of the database unit 300 can be provided inside the information processing device 200. In this case, at least a part of the database unit 300 may be implemented by the same storage means as the storage means implementing the memory 230, or may be implemented by a storage means different from the storage means implementing the memory unit 230. In any case, at least a part of the database unit 300 is configured as a storage unit for the information processing device 200. The configuration of the database unit 300 is not limited to a specific hardware configuration. For example, the database unit 300 may be configured as a single hardware component, or may be configured as a plurality of hardware components. For example, the database unit 300 may be configured as an external hard disk device of the information processing device 200, or may be configured as a cloud storage connected via a network.

[0092] Each component of the information processing device 200 described above may be composed of a single hardware part or may be composed of multiple hardware parts. When composed of multiple hardware parts, the manner in which each hardware part is connected does not matter. Each hardware part may be connected wirelessly or by wire. It is also within the scope of the present invention to configure the processor unit 220 by an analog circuit rather than a digital circuit. The system 100 is not limited to a specific hardware configuration. The configuration of the system 100 is not limited to the one described above as long as it can realize its functions.

[0093] 3. Processing in the system to predict when repairs should be performed 7A shows an example of a process 700 in the system 100 of the present invention. In this example, a case will be described in which the system 100 is implemented by an information processing device 200. The process 700 is executed in the processor unit 220 of the information processing device 200.

[0094] In step S701, the processor unit 220 receives data representing the surface shape of the wind turbine blade. The processor unit 220 can receive data from outside the information processing device 200 via the interface unit 210. For example, the processor unit 220 can receive data from a measuring means (for example, a measuring means provided in the device 50, or a measuring means provided in a device other than the device 50) via the interface unit 210.

[0095] The processor unit 220 may further receive environmental data representing the environment around the wind turbine and / or material data representing the material of at least a part of the wind turbine blade, in addition to the data representing the surface shape of the wind turbine blade. For example, the processor unit 220 can receive environmental data from a measuring device provided on or near the wind turbine via the interface unit 210. For example, the processor unit 220 can receive environmental data and / or material data from a publicly available database via the interface unit 210.

[0096] In step S702, the processor unit 220 predicts when the wind turbine blade should be repaired based on the data received in step S701. The processor unit 220 can predict when the wind turbine blade should be repaired using a prediction model. For example, the data received in step S701 can be input to a prediction model that can correlate data representing the surface shape of the blade with a value indicating the time interval between measurement and repair, thereby causing the prediction model to output the time when the blade should be repaired.

[0097] In step S701, if environmental data and / or material data are received in addition to the data representing the surface shape of the wind turbine blades, the processor unit 220 can input the data representing the surface shape of the blades and the environmental data and / or material data into the predictive model, thereby causing the predictive model to output the timing for blade repair.

[0098] The predicted timing is output to the outside of the information processing device 200. For example, it is provided to a worker performing maintenance on the blade, so that the worker can know when to repair the blade. For example, the information processing device 200 may provide an alert (e.g., a visual or audible alert) to the worker when it is time to repair the blade.

[0099] In step S702, as described below with reference to FIG. 7B, a prediction model may be selected according to the state of the damaged portion of the blade, and prediction may be performed using the selected prediction model.

[0100] FIG. 7B illustrates an example of the process of step S702 in process 700.

[0101] In step S7021, the processor unit 220 identifies the type of damaged portion on the surface of the blade from the data representing the surface shape of the wind turbine blade received in step S701.

[0102] The processor unit 220 first identifies a damaged portion from the data representing the surface shape, and then identifies the aspect (e.g., depth and width, or shape, etc.) of the identified damaged portion.

[0103] In step S7022, the processor unit 220 determines whether the erosion pattern of the damaged portion is the first pattern or the second pattern based on the aspect of the damaged portion identified in step S7021. If it is determined that the erosion pattern of the damaged portion is the first pattern, the process proceeds to step S7023. If it is determined that the erosion pattern of the damaged portion is the second pattern, the process proceeds to step S7024.

[0104] In step S7023, the processor unit 220 selects the first prediction model as the prediction model to be used. The first prediction model is a trained model that has been trained using data obtained from a blade having a damaged portion with a first erosion pattern as training data.

[0105] In step S7024, the processor unit 220 selects the second prediction model as the prediction model to be used. The second prediction model is a trained model that has been trained using data obtained from a blade having a damaged portion with a second erosion pattern as training data.

[0106] In step S7025, the prediction model selected in step S7023 or step S7024 is used to predict the timing of repairs based on the data received in step S701.

[0107] In this way, by taking differences in erosion patterns into account when predicting the timing of repairs, it is possible to make predictions with greater accuracy.

[0108] 7A and 7B, steps are described as being performed in a particular order, but the order shown is only an example and the order in which steps are performed is not limited thereto. Steps may be performed in any order that is logically possible. Also, other steps may be performed in addition to or instead of the steps shown.

[0109] In the example described above with reference to Figures 7A and 7B, the processing of each step shown in Figures 7A and 7B can be realized by the processor unit 220 and the program stored in the memory unit 230 of the information processing device 200 that implements the system 100. However, the present invention is not limited to this. At least one of the processing of each step shown in Figures 7A and 7B may be realized by a hardware configuration such as a control circuit.

[0110] The present invention is not limited to the above-mentioned embodiment. It is understood that the scope of the present invention should be interpreted only by the claims. It is understood that a person skilled in the art can implement an equivalent scope based on the description of the present invention and technical common sense from the description of the specific preferred embodiment of the present invention. [Industrial Applicability]

[0111] The present invention is useful for providing a system for predicting when repairs should be performed on wind turbine blades. [Explanation of symbols]

[0112] 10 windmill 11, 12, 13 Blades 14 Nacelle 15 Hub 20 Rope 50 equipment 100 Systems 110 Receiving means 120 Prediction Methods

Claims

1. A system for predicting when to perform repairs on a wind turbine blade, comprising: receiving means for receiving data representative of a surface shape of a blade of a wind turbine; A prediction means for predicting the timing based on the data; The prediction means comprises: selecting a predictive model based on an erosion pattern of a damaged portion of a surface of the blade as indicated in the data, wherein selecting the predictive model comprises: selecting a first predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction perpendicular to the surface; selecting a second predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction parallel to the surface; and predicting the timing based on the data using the selected predictive model; and A system configured to:

2. The system of claim 1 , wherein the data representative of a surface profile of the wind turbine blade includes a depth, a width, and a location of a damaged portion of the surface of the blade.

3. The receiving means further receives environmental data representing an environment around the wind turbine; The system of claim 1 , wherein the predicting means predicts the timing based on the data and the environmental data.

4. The system of claim 3 , wherein the environmental data includes at least one of wind speed data, rainfall data, raindrop data, UV data, and data representative of a location.

5. The receiving means further receives material data representative of a material of at least a portion of the blade; The system of claim 3 , wherein the predicting means predicts the timing based on the data, the environmental data, and the material data.

6. The system of claim 5 , wherein the material data includes at least one of paint data representative of a material of a paint on the surface of the blade, and putty data representative of a material of a putty on the surface of the blade.

7. 1. A method for predicting when repairs should be made to a blade of a wind turbine, comprising: receiving data representative of a surface profile of a blade of a wind turbine; predicting said timing based on said data; and predicting the timing includes: selecting a predictive model based on an erosion pattern of a damaged portion of a surface of the blade as indicated in the data, wherein selecting the predictive model comprises: selecting a first predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction perpendicular to the surface; selecting a second predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction parallel to the surface; and predicting the timing based on the data using the selected predictive model; and A method comprising:

8. A program for predicting a timing for repairing a blade of a wind turbine, the program being executed in a computer system having a processor, the program comprising: receiving data representative of a surface profile of a blade of a wind turbine; predicting said timing based on said data; and predicting the timing by causing the processor to perform a process including selecting a predictive model based on an erosion pattern of a damaged portion of a surface of the blade as indicated in the data, wherein selecting the predictive model comprises: selecting a first predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction perpendicular to the surface; selecting a second predictive model when the damaged portion has an erosion pattern that extends from a primary damage portion in a direction parallel to the surface; and predicting the timing based on the data using the selected predictive model; and Including, the program.

Images