Sweeping tower risk prediction method and model training method, device, equipment and medium

Abstract

This application discloses a method, model training method, apparatus, equipment, and medium for predicting tower sweep risk, relating to the field of wind power generation technology. The method includes: first, acquiring the input area to be measured; then, determining the terrain information of the area to be measured based on the correspondence between the area and terrain information; and finally, obtaining the prediction result of the area to be measured based on the terrain information and the tower sweep risk prediction model. The tower sweep risk prediction model is trained based on the terrain information of the target area, wherein the target area is determined based on the wind direction and includes risk areas and normal areas. Compared with predicting tower sweep risk of the area to be measured based on human experience, this method can reduce the subjectivity of manual prediction and the influence of experience, thereby improving the accuracy of the prediction results.

Description

Technical Field

[0001] This application relates to the field of wind power generation technology, and in particular to a method for predicting tower sweep risk, a model training method, an apparatus, equipment, and a computer-readable storage medium. Background Technology

[0002] A wind turbine is a device that converts wind energy into electrical energy, including a nacelle, generator, blades, and tower. The blades are the main component of a wind turbine that absorbs wind energy and are a crucial source of input. Tower swiping refers to the incident where wind causes the blades to collide with the tower. After a tower swiping incident, blade replacement and tower maintenance are required, resulting in substantial losses.

[0003] Currently, the determination of whether a wind turbine will experience a tower sweeping event is mainly based on human experience. This method of prediction is highly subjective and makes it difficult to obtain accurate prediction results. Summary of the Invention

[0004] The purpose of this application is to provide a method for predicting tower sweep risk, a model training method, an apparatus, equipment, and a computer-readable storage medium to improve the accuracy of prediction results.

[0005] Firstly, this application provides a method for predicting tower sweeping risks, including:

[0006] Obtain the input test area;

[0007] Based on the correspondence between region and terrain information, the terrain information of the area to be measured is determined;

[0008] Based on the terrain information of the area to be tested and the tower sweeping risk prediction model, the prediction results of the area to be tested are obtained;

[0009] The tower sweeping risk prediction model is obtained by training based on the terrain information of the target area, which is obtained based on the direction of wind flow and includes risk areas and normal areas.

[0010] Optionally, the terrain information is characterized by elevation values, and the correspondence between the region and the terrain information is characterized by a digital elevation model or topographic map. Obtaining the terrain information of the area to be measured based on the correspondence between the region and the terrain information includes:

[0011] The boundary of the area to be measured is determined, and the digital elevation model or the topographic map is cut using the boundary to obtain the elevation value within the area to be measured.

[0012] Optionally, determining the boundary of the region to be measured includes:

[0013] The area to be tested is divided into multiple sectors. The center line of each sector is used as the direction of wind flow. Based on the direction of wind flow and the center of the area to be tested, the boundary of the area to be tested is generated.

[0014] Optionally, the target area is a shape region corresponding to the calculation method of the tower sweeping risk prediction model.

[0015] Optionally, the shape region is a rectangular region, the wind direction is from the first boundary of the rectangular region through the target position to the second boundary of the rectangular region, the distance between the target position and the second boundary is less than the distance between the target position and the first boundary, the first boundary is parallel to the second boundary, and the wind direction is perpendicular to the first boundary.

[0016] Optionally, the terrain information is represented by elevation values, wherein the elevation values ​​of the target area are represented by a terrain matrix according to a preset rule; the tower sweeping risk prediction model is a model obtained by training based on the terrain matrix.

[0017] Optionally, the danger zone is obtained based on the wind direction corresponding to historical occurrences of tower sweep risks; and

[0018] The normal area is obtained based on the wind direction when there was no risk of tower sweeping in history.

[0019] Optionally, the method further includes:

[0020] When the prediction result indicates that there is a risk of tower sweeping in the area to be tested, the user is prompted with the risk area in the area to be tested and the risk area is displayed on the topographic map;

[0021] When the prediction result indicates that there is no risk of tower sweeping in the area to be tested, the user is prompted that the area to be tested is a normal area and the normal area is displayed on the topographic map.

[0022] Secondly, this application provides a model training method, including:

[0023] The target area is determined based on the direction of wind flow, and the target area includes risk areas and normal areas;

[0024] Based on the correspondence between region and terrain information, the terrain information of the target region is obtained;

[0025] The model is trained using the terrain information of the target area until the model meets the training stopping condition, thus obtaining the tower sweeping risk prediction model.

[0026] Optionally, determining the target area based on the direction of wind flow includes:

[0027] Based on the direction of wind flow when a tower sweep risk occurs, the target area is determined as the risk area; or,

[0028] Based on the direction of wind flow when there is no risk of tower sweeping, the target area is determined to be a normal area.

[0029] Optionally, the risk of tower sweeping includes the distance between the blade and the tower being less than a preset threshold, and the distance between the blade and the tower is obtained by a laser distance sensor.

[0030] Optionally, the terrain information is represented by elevation values, and the correspondence between the region and the terrain information is represented by a digital elevation model or topographic map. Obtaining the terrain information of the target region based on the correspondence between the region and the terrain information includes:

[0031] The boundary of the target area is determined, and the digital elevation model or the topographic map is cut using the boundary to obtain the elevation values ​​within the target area.

[0032] Optionally, before using the terrain information of the target area for model training, the method further includes:

[0033] The elevation values ​​of the target area are adjusted according to preset rules to obtain the terrain matrix of the target area;

[0034] The step of training the model using the terrain information of the target area includes:

[0035] The model is trained using the terrain matrix of the target area.

[0036] Thirdly, this application provides a tower sweeping risk prediction device, comprising:

[0037] The acquisition module is used to acquire the input area to be measured; and to determine the terrain information of the area to be measured based on the correspondence between the area and the terrain information.

[0038] The prediction module is used to obtain the prediction result of the area to be tested based on the terrain information of the area to be tested and the tower sweeping risk prediction model;

[0039] The tower sweeping risk prediction model is obtained by training on the terrain information of a target area that includes risk areas and normal areas. The target area is obtained based on the direction of wind flow and includes risk areas and normal areas.

[0040] Optionally, the terrain information is characterized by elevation values, and the correspondence between the region and the terrain information is characterized by a digital elevation model or a topographic map; the acquisition module is used to determine the boundary of the area to be measured, and use the boundary to cut the digital elevation model or the topographic map to obtain the elevation values ​​within the area to be measured.

[0041] Optionally, the target area is a shape region corresponding to the calculation method of the tower sweeping risk prediction model.

[0042] Optionally, the shape region is a rectangular region, the wind direction is from the first boundary of the rectangular region through the target position to the second boundary of the rectangular region, the distance between the target position and the second boundary is less than the distance between the target position and the first boundary, the first boundary is parallel to the second boundary, and the wind direction is perpendicular to the first boundary.

[0043] Optionally, the terrain information is represented by elevation values, and the elevation values ​​of the target area are represented by a terrain matrix according to preset rules; the tower sweeping risk prediction model is trained based on the terrain matrix.

[0044] Optionally, the dangerous area is obtained based on the wind direction corresponding to the historical occurrence of tower sweep risk; and the normal area is obtained based on the wind direction corresponding to the historical occurrence of no tower sweep risk.

[0045] Optionally, the device further includes a prompting module, which is used to prompt the user about the risk area in the area to be tested and present the risk area on the topographic map when the prediction result indicates that there is a risk of tower sweeping in the area to be tested; and to prompt the user that the area to be tested is a normal area and present the normal area on the topographic map when the prediction result indicates that there is no risk of tower sweeping in the area to be tested.

[0046] Fourthly, this application provides a model training apparatus, comprising:

[0047] The acquisition module is used to determine the target area based on the direction of wind flow. The target area includes risk areas and normal areas. Based on the correspondence between the area and the terrain information, the module acquires the terrain information of the target area.

[0048] The training module is used to train the model using the terrain information of the target area until the model meets the training stopping condition, thus obtaining the tower sweeping risk prediction model.

[0049] Optionally, the acquisition module is used to determine the target area as a risk area based on the wind direction when a tower sweep risk occurs; or, based on the wind direction when a tower sweep risk does not occur, to determine the target area as a normal area.

[0050] Optionally, the risk of tower sweeping includes the distance between the blade and the tower being less than a preset threshold, and the distance between the blade and the tower is obtained by a laser distance sensor.

[0051] Optionally, the terrain information is represented by elevation values, and the correspondence between the region and the terrain information is represented by a digital elevation model or a topographic map. The acquisition module is used to determine the boundary of the target region and use the boundary to cut the digital elevation model or the topographic map to obtain the elevation values ​​within the target region.

[0052] Optionally, the device further includes an adjustment module, which is used to adjust the elevation value of the target area according to a preset rule to obtain a terrain matrix of the target area;

[0053] The training module is used to train the model using the terrain matrix of the target area.

[0054] Fifthly, this application provides a computing device, the computing device including a processor and a memory:

[0055] The memory is used to store computer programs and to transfer the computer programs to the processor;

[0056] The processor is configured to execute any of the methods described above according to instructions in the computer program.

[0057] Sixthly, this application provides a computer-readable storage medium for storing a computer program for performing the method described in any one of the preceding claims.

[0058] Compared with the prior art, the advantages of the above-mentioned technical solution of this application are as follows:

[0059] This application provides a method for predicting wind turbine sweep risk. The method includes acquiring an input area to be tested, determining the terrain information of the area based on the correspondence between the area and terrain information, and then obtaining a prediction result for the area based on the terrain information and a wind turbine sweep risk prediction model. The wind turbine sweep risk prediction model is trained based on the terrain information of the target area, which is determined based on the wind direction. The target area includes both hazardous and normal areas. Compared with judging the wind turbine sweep risk based on human experience, this method reduces the subjectivity and influence of human prediction, improving the accuracy of the prediction results. Furthermore, after obtaining the prediction result based on the wind turbine sweep risk of the area to be tested, this prediction result can provide a reference for the site selection of wind turbines, thereby avoiding the installation of wind turbines in areas with high wind turbine sweep risk and further reducing losses caused by wind turbine sweeps. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 A schematic diagram of a wind turbine generator provided for an embodiment of this application;

[0062] Figure 2 A flowchart illustrating a method for predicting tower sweeping risks provided in this application embodiment;

[0063] Figure 3A A schematic diagram of a tower sweeping risk prediction interface provided in an embodiment of this application;

[0064] Figure 3B A schematic diagram of another tower sweeping risk prediction interface provided in an embodiment of this application;

[0065] Figure 4 A schematic diagram of a rectangular region provided in an embodiment of this application;

[0066] Figure 5 A schematic diagram of a region of interest provided in an embodiment of this application;

[0067] Figure 6A A schematic diagram of a result interface provided in an embodiment of this application;

[0068] Figure 6B A schematic diagram of a result interface provided in an embodiment of this application;

[0069] Figure 7 A flowchart illustrating a model training method provided in this application embodiment;

[0070] Figure 8 A schematic diagram of a tower sweeping risk prediction device provided in an embodiment of this application;

[0071] Figure 9 This is a schematic diagram of a model training device provided in an embodiment of this application. Detailed Implementation

[0072] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present application.

[0073] The technical terms used in the embodiments of this application will be introduced below.

[0074] A wind turbine is a device that converts wind energy into electrical energy, such as... Figure 1 As shown in the figure, this is a schematic diagram of a wind turbine generator provided in an embodiment of this application. The wind turbine generator includes blades 101 and a tower 102. Due to the action of wind, the blades 101 will rotate, thereby converting wind energy into electrical energy.

[0075] Tower swivel risk refers to the probability that blade 101 will collide with tower 102 during rotation (also known as tower swivel). To reduce tower swivel events, the tower swivel risk of the selected area is usually predicted during the wind turbine site selection process to determine whether to build a wind turbine in that area. Currently, the prediction of tower swivel risk in the selected area is mainly based on human experience, which is subject to subjectivity and uncertainty, making it difficult to obtain accurate prediction results.

[0076] In view of this, embodiments of this application provide a method for predicting tower sweep risk, which can be executed by a first processing device. The first processing device can be a terminal or a server. The terminal includes, but is not limited to, smartphones, tablets, and laptops. The server can be a cloud server, such as a central server in a central cloud computing cluster or an edge server in an edge cloud computing cluster. Alternatively, the server can be a server in a local data center. A local data center refers to a data center directly controlled by the user.

[0077] Specifically, the method includes: a first processing device acquiring the input area to be measured, determining the terrain information of the area to be measured based on the correspondence between the area and terrain information, and then obtaining the prediction result of the area to be measured based on the terrain information of the area to be measured and the tower sweeping risk prediction model. The tower sweeping risk prediction model is trained based on the terrain information of the risk area and the terrain information of the normal area, and the risk area and the normal area are determined based on the direction of wind flow.

[0078] In this method, the predicted results, compared to those determined by human experience, reduce the subjective influence of human experience and are more accurate. Furthermore, using these predictions for wind turbine site selection can effectively reduce the risk of blade sweeping, thereby lowering the maintenance costs of wind turbines.

[0079] To make the technical solution of this application clearer and easier to understand, the method for predicting tower sweeping risks provided in the embodiments of this application will be introduced below from the perspective of the first processing device.

[0080] like Figure 2 As shown, this figure is a flowchart of a tower sweeping risk prediction method provided in an embodiment of this application. The method includes:

[0081] S201: The first processing device acquires the input area to be tested.

[0082] In some examples, the first processing device can present a human-computer interaction interface to the user through a display device, on which the user can input the area to be tested, and then the first processing device can obtain the input area to be tested.

[0083] like Figure 3A As shown in the figure, this is a schematic diagram of a tower sweep risk prediction interface provided in an embodiment of this application. The tower sweep risk prediction interface includes a test area input area 301 and a prediction control 302. The user can input the test area in the test area input area 301, and the first processing device can then acquire the input test area. The user can then click the prediction control 302, and the first processing device predicts the tower sweep risk in the user-input test area based on the user's operation on the prediction control 302 (e.g., clicking).

[0084] In some examples, to improve user input convenience, the user can input only one location coordinate, and the first processing device automatically generates the area to be measured. For example, the user can input longitude and latitude, and the first processing device will automatically generate the area to be measured centered on the location determined by the longitude and latitude. The area to be measured can be rectangular, sector-shaped, circular, etc., and this application does not limit this; those skilled in the art can choose a suitable shape according to actual needs. In other examples, the user can also input multiple location coordinates simultaneously, and thus input multiple areas to be measured simultaneously.

[0085] In other examples, the first processing device can also present a map to the user, allowing the user to directly select the area to be tested on the map. For example... Figure 3B As shown in the figure, this is a schematic diagram of another tower sweep risk prediction interface provided in an embodiment of this application. The tower sweep risk prediction interface includes a candidate area 303 and a prediction control 304. The candidate area 303 is used to display a map, and the user can extract the area to be tested from the map displayed in the candidate area 303. Thus, the first processing device can obtain the input area to be tested. Then, the user can operate the prediction control 304 (e.g., click), and the first processing device predicts the tower sweep risk in the user-input area based on the user's operation on the prediction control 304.

[0086] It should be noted that the above Figure 3A and Figure 3B This is merely an illustrative example of how the first processing device acquires the area to be tested; those skilled in the art can make alternatives based on their specific needs. For instance, in other embodiments, the first processing device may also acquire the area to be tested from a database via a network or line.

[0087] S202: The first processing device determines the terrain information of the area to be measured based on the correspondence between the area and the terrain information.

[0088] Topographic information can be represented by elevation values, which can be absolute elevation values, representing the vertical distance from a ground point to the elevation reference surface, also known as the elevation of that ground point. In some examples, the correspondence between regions and topographic information can be represented by digital elevation models (DEMs) or topographic maps. That is, based on a DEM, the topographic information corresponding to each region can be obtained, or based on a topographic map, the topographic information of each region can be obtained. Typically, a region includes multiple points, and the topographic information corresponding to the region is represented by the elevation values ​​of multiple points.

[0089] Based on this, after acquiring the area to be measured, the first processing device obtains the terrain information of the area to be measured according to the correspondence between the area and the terrain information. In some examples, the first processing device can first determine the boundary of the area to be measured, and then use the boundary to cut the digital elevation model or topographic map to obtain the elevation value within the area to be measured, thereby obtaining the terrain information of the area to be measured. As an example, the terrain information can be from the 30m horizontal resolution digital elevation model (DEM) dataset of the Shuttle Radar Topography Mission (SRTM).

[0090] In some embodiments, the first processing device can further divide the area to be measured into multiple sectors, using the centerline of each sector as the direction of wind flow, and generating the boundary of the area to be measured based on the direction of wind flow and the center of the area to be measured. The method of dividing the area to be measured into multiple sectors will be described in subsequent embodiments.

[0091] It should be noted that a digital elevation model (DEM) typically refers to a digital simulation of ground topography achieved through limited terrain elevation data, that is, a digital representation of the terrain surface morphology. A DEM can be considered a physical ground model that represents ground elevation using an ordered array of numerical values. Based on a DEM, the aforementioned elevation values ​​can be obtained relatively easily.

[0092] S203: The first processing device obtains the prediction results of the area to be tested based on the terrain information and the tower sweeping risk prediction model of the area to be tested.

[0093] The tower sweeping risk prediction model is trained based on the terrain information of the target area, which includes both risk and normal areas. This model is used to predict the tower sweeping risk of the area to be tested. The training process of the tower sweeping risk prediction model will be described in subsequent embodiments.

[0094] In some examples, the first processing device can divide the area to be measured into multiple sectors. For example, it can divide the area into multiple sectors centered on the user-input location coordinates, such as 16 sectors, one sector per 22.5°. Then, the centerline of each sector can be used as the direction of wind flow, and a region of interest (ROI) can be automatically generated based on the center of the area to be measured. When there are 16 sectors, there are 16 ROIs. In some examples, the ROI can be a rectangular area.

[0095] In some examples, the first processing device can obtain a rectangular area based on the center of the area to be measured and the direction of wind flow, such as... Figure 4As shown in the figure, this is a schematic diagram of a rectangular region provided in an embodiment of this application. Point F can be the center of the region to be measured. The direction from point K to point F is the direction of wind flow. This wind flow direction extends from the first boundary GH of the rectangular region GHIJ through point F to the second boundary JI of the rectangular region GHIJ. The distance from point F to the second boundary JI is less than the distance from point F to the first boundary. The first boundary GH is parallel to the second boundary JI, and the wind flow direction KF is perpendicular to the first boundary GH.

[0096] In some examples, at the center of the known area to be measured, one can proceed forward a first length (e.g., from point F) along the direction of the wind flow. Figure 4 The distance from midpoint F to JI), the second backward length (the distance from point F to GH), and the third leftward length (e.g., the distance from midpoint F to JI). Figure 4 The distance from midpoint F to HI) and the fourth length to the right (e.g.) Figure 4 The parameters are defined using the distance from the midpoint F to GJ, thus obtaining the above... Figure 4 The rectangular region GHIJ is shown in the diagram. The direction of the incoming wind (e.g., the azimuth alpha, which is the angle between point F and true north) can be one of the parameters determining the rectangular region GHIJ. The center of the region to be measured can be characterized by longitude and latitude, where the center can be, for example, the location of a wind measurement tower or wind turbine.

[0097] Since the orientation of the regions of interest corresponding to different sectors is different, the first processing device can represent the elevation value of the region of interest through a terrain matrix according to a preset rule. This allows the first processing device to process the terrain information of the area to be measured through the tower sweeping risk prediction model, such as processing the terrain information of all regions of interest within the area to be measured, and thus obtaining the prediction result.

[0098] For example, such as Figure 5As shown in the figure, this is a schematic diagram of a region of interest (ROI) provided in an embodiment of this application. The ROI ABCD is rectangular and includes 12 points to be measured. The boundary AB of the ROI ABCD can be used as the left boundary, and the terrain information of the ROI ABCD can be represented by a terrain matrix. The number of rows in the terrain matrix is ​​the same as the number of points to be measured on boundary AB, and the number of columns in the terrain matrix is ​​the same as the number of points to be measured on boundary BC. In some examples, the elevation value of the first row and first column of the terrain matrix corresponds to the elevation value of point A in the ROI ABCD; the elevation value of the first row and fourth column of the terrain matrix corresponds to the elevation value of point D in the ROI ABCD; the elevation value of the third row and first column of the terrain matrix corresponds to the elevation value of point B in the ROI ABCD; and the elevation value of the third row and fourth column of the terrain matrix corresponds to the elevation value of point C in the ROI ABCD.

[0099] For example, the terrain matrix can be seen in Table 1 below:

[0100] Table 1

[0101]

[0102] Among them, see points V1-V8. Figure 5 .

[0103] It should be noted that the above are merely examples illustrating the preset rules. In other examples, the boundary AD of the region of interest ABCD can also be used as the left boundary. This application does not limit this, and those skilled in the art can choose according to actual needs.

[0104] In some examples, the prediction result can be the probability value of blade sweeping, such as a prediction result of 80%, which indicates that there is an 80% probability of blade sweeping risk in the region of interest. In other examples, the prediction result can also be that blade sweeping will occur in the region of interest or that blade sweeping will not occur in the region of interest.

[0105] In some examples, the first processing device can also present the prediction results to the user. For instance, when the prediction results indicate a risk of tower sweeping in the area to be measured, the first processing device can use a display device to alert the user to the risk area within the area to be measured and present the risk area on a topographic map. In some examples, the risk area can be presented in KML format, GeoJSON format, Esrishapefile format, etc. Figure 6A As shown in the figure, this figure is a schematic diagram of a result interface provided in an embodiment of this application. The result interface includes a risk area 501 and a prompt message 502. The prompt message 502 may be "There is a risk of blade sweeping in the current area".

[0106] When the prediction result indicates that there is no risk of tower sweeping in the area to be tested, the first processing device can prompt the user through the display device that the area to be tested is a normal area and present the normal area on the topographic map. For example... Figure 6B As shown in the figure, this figure is a schematic diagram of another result interface provided by the embodiment of this application. The result interface includes the test area 503 and the prompt information 504. The prompt information 504 may be "There is no risk of blade sweeping in the current area".

[0107] Based on the above description, this application provides a method for predicting wind turbine sweep risk. Compared with judging the wind turbine sweep risk based on human experience, this method can reduce the subjectivity and influence of experience in manual prediction, and improve the accuracy of the prediction results. Furthermore, after predicting the wind turbine sweep risk of the area to be tested, the prediction results can provide a reference for the site selection of wind turbines, thereby avoiding the installation of wind turbines in areas with high wind turbine sweep risk and further reducing the losses caused by wind turbine sweep.

[0108] This application also provides a model training method, which can be executed by a second processing device. The second processing device can be the same as the first processing device, or it can be a different processing device.

[0109] Specifically, the second processing device can determine the target area based on the direction of wind flow. The target area includes risk areas and normal areas. Then, based on the correspondence between the area and the terrain information, the terrain information of the target area is obtained. Finally, the terrain information of the target area is used to train the model until the model meets the training stopping condition, thus obtaining the tower sweeping risk prediction model.

[0110] To make the technical solution of this application clearer and easier to understand, the model training method provided in the embodiments of this application will be introduced from the perspective of the second processing device.

[0111] like Figure 7 As shown, this figure is a flowchart of a model training method provided in an embodiment of this application. The method includes:

[0112] S701: The second processing device determines the target area based on the direction of the incoming wind.

[0113] In some examples, the wind direction can be the direction of wind flow at a target location, which could be, for example, the location of a wind turbine or a wind measurement tower. Based on this, the wind direction at the location of the wind turbine can be determined using the wind vane direction WD and yaw angle θyaw of the wind turbine's nacelle. Specifically, the second processing device can calculate the wind direction based on the following formula:

[0114] alpha=θyaw+WD-180° (1)

[0115] Where alpha represents the direction of the incoming wind.

[0116] This application does not specifically limit the shape of the target region. To facilitate model training, the target region can be a shape corresponding to the operation method of the tower sweeping risk prediction model. The operation method of the tower sweeping risk prediction model can be 2D convolution operation, and based on this, the target region can be a rectangular region.

[0117] The second processing device determines the target region in a similar way to the first processing device determine the region of interest; see the above embodiments for details. Figure 4 This will not be elaborated upon here.

[0118] In some examples, the target area includes both risk areas and normal areas. The second processing device determines the target area as a risk area based on the wind direction when a tower sweep risk occurs at the target location; similarly, it can determine the target area as a normal area based on the wind direction when a tower sweep risk does not occur at the target location.

[0119] The risk of tower sweeping at the target location can be that the distance between the wind turbine blades and the tower at the target location is less than a preset threshold. This distance can be obtained based on a laser distance sensor installed on the wind turbine, such as a laser clearance sensor.

[0120] Based on this, the second processing device can obtain a large amount of data related to the target location (such as the location of the wind turbine), the direction of incoming flow when a tower sweep risk occurs, and the direction of incoming flow when a tower sweep risk does not occur, thereby obtaining a large number of risk areas and normal areas.

[0121] In some examples, 400 wind turbines equipped with laser clearance sensors can be continuously monitored for a year to obtain the wind direction when the distance between all blades and the tower is less than a preset threshold, i.e., when a tower sweep risk occurs. To ensure data richness and model robustness, 4000 wind direction data points for tower sweep risk scenarios can be obtained. Furthermore, to ensure a balance between positive and negative samples in the training data, 4000 wind direction data points for scenarios where a tower sweep risk does not occur can be randomly obtained. Then, risk areas and normal areas are determined separately.

[0122] S702: The second processing device obtains the terrain information of the target area based on the correspondence between the area and the terrain information.

[0123] The second processing device obtains the terrain information of the target area based on the correspondence between the area and the terrain information in a similar manner to S202 described above, and will not be repeated here. The second processing device can acquire the terrain information of risk areas and normal areas. Among them, risk areas can be used as negative samples, and normal areas can be used as positive samples.

[0124] S703: The second processing device trains the model based on the terrain information of the target area until the model meets the training stop condition, thus obtaining the tower sweeping risk prediction model.

[0125] In some embodiments, the second processing device can also filter the target area. For example, a circle can be drawn with the target location as the center, and then divided into multiple sectors, such as one sector for every 22.5°, resulting in 16 sectors. Taking 40 data points of tower sweep risk occurring at the target location as an example, 19 of them fall in sector 1, 20 fall in sector 2, and only 1 falls in sector 3. The other 13 sectors do not have any data points of tower sweep risk. Therefore, the tower sweep risk data point corresponding to sector 3 may be a false alarm caused by the laser clearance sensor or other reasons. Thus, this data point can be removed to ensure the accuracy of the sample data and avoid the negative impact of poor sample data on subsequent model training.

[0126] The terrain information of the target area can be represented by the elevation values ​​of the target area. In some embodiments, the second processing device can adjust the elevation values ​​of the target area according to preset rules to obtain a terrain matrix of the target area. The specific adjustment method is as follows... Figure 5 The method shown is similar and will not be repeated here.

[0127] As mentioned above, the second processing device can obtain 4,000 positive samples (e.g., terrain matrix corresponding to sectors in normal areas) and 4,000 negative samples (e.g., terrain matrix corresponding to sectors in risk areas). 3,200 samples are randomly selected from the positive and negative samples respectively as the training set, and the rest are used as the test set for model training.

[0128] During model training, the second processing device inputs N L×W terrain matrices as training samples, where L can be the number of rows in the terrain matrix and W can be the number of columns in the terrain matrix. The sample labels are one-dimensional vectors with N elements.

[0129] In some examples, the model can be an 8-layer deep convolutional neural network, where layers 1, 3, and 5 are convolutional layers, layers 2, 4, and 6 are pooling layers, and layers 7 and 8 are fully connected layers. The convolutional kernels in the convolutional layers are 3×3 with a stride of 1, and the pooling layers use max pooling with a stride of 2. The convolutional layers and the first fully connected layer use ReLU as the activation function, while the last fully connected layer uses the Sigmoid function. To avoid overfitting on the test set, a 20% random dropout of neurons is implemented in the 7th fully connected layer.

[0130] The model can be trained using TensorFlow 2.0 as the computational framework, stochastic gradient descent as the optimization algorithm, and binary cross-entropy as the loss function. Training is performed for 100 epochs, iterating through the training samples in each epoch until the model meets the training stopping condition, thus obtaining the tower sweeping risk prediction model. The training stopping condition can be that the model's accuracy on the test set meets a requirement, such as an accuracy higher than 75%. If the training stopping condition is not met, the model can be optimized by adjusting the neural network structure and hyperparameters until the training stopping condition is met again.

[0131] like Figure 8 As shown in the figure, this is a schematic diagram of a tower sweeping risk prediction device provided in an embodiment of this application. The tower sweeping risk prediction device includes:

[0132] The acquisition module 801 is used to acquire the input area to be measured; and to determine the terrain information of the area to be measured based on the correspondence between the area and the terrain information.

[0133] The prediction module 802 is used to obtain the prediction result of the area to be tested based on the terrain information of the area to be tested and the tower sweeping risk prediction model;

[0134] The tower sweeping risk prediction model is obtained by training on the terrain information of a target area that includes risk areas and normal areas. The target area is obtained based on the direction of wind flow and includes risk areas and normal areas.

[0135] Optionally, the terrain information is characterized by elevation values, and the correspondence between the region and the terrain information is characterized by a digital elevation model or a topographic map; the acquisition module is specifically used to determine the boundary of the area to be measured, and use the boundary to cut the digital elevation model or the topographic map to obtain the elevation values ​​within the area to be measured.

[0136] Optionally, the target area is a shape region corresponding to the calculation method of the tower sweeping risk prediction model.

[0137] Optionally, the shape region is a rectangular region, the wind direction is from the first boundary of the rectangular region through the target position to the second boundary of the rectangular region, the distance between the target position and the second boundary is less than the distance between the target position and the first boundary, the first boundary is parallel to the second boundary, and the wind direction is perpendicular to the first boundary.

[0138] Optionally, the terrain information is represented by elevation values, and the elevation values ​​of the target area are represented by a terrain matrix according to preset rules; the tower sweeping risk prediction model is specifically trained based on the terrain matrix.

[0139] Optionally, the dangerous area is obtained based on the wind direction corresponding to the historical occurrence of tower sweep risk; and the normal area is obtained based on the wind direction corresponding to the historical occurrence of no tower sweep risk.

[0140] Optionally, the device further includes a prompting module, which is used to prompt the user about the risk area in the area to be tested and present the risk area on the topographic map when the prediction result indicates that there is a risk of tower sweeping in the area to be tested; and to prompt the user that the area to be tested is a normal area and present the normal area on the topographic map when the prediction result indicates that there is no risk of tower sweeping in the area to be tested.

[0141] like Figure 9 As shown in the figure, this is a schematic diagram of a model training device provided in an embodiment of this application. The model training device includes:

[0142] The acquisition module 901 is used to determine the target area based on the direction of wind flow, the target area including risk areas and normal areas; and to acquire the terrain information of the target area based on the correspondence between the area and terrain information.

[0143] Training module 902 is used to train the model using the terrain information of the target area until the model meets the training stopping condition, thereby obtaining the tower sweeping risk prediction model.

[0144] Optionally, the acquisition module 901 is specifically used to determine the target area as a risk area based on the wind direction when a tower sweep risk occurs; or, based on the wind direction when a tower sweep risk does not occur, to determine the target area as a normal area.

[0145] Optionally, the risk of tower sweeping includes the distance between the blade and the tower being less than a preset threshold, and the distance between the blade and the tower is obtained by a laser distance sensor.

[0146] Optionally, the terrain information is represented by elevation values, and the correspondence between the region and the terrain information is represented by a digital elevation model or a topographic map. The acquisition module 901 is specifically used to determine the boundary of the target region, and use the boundary to cut the digital elevation model or the topographic map to obtain the elevation values ​​within the target region.

[0147] Optionally, the device further includes an adjustment module, which is used to adjust the elevation value of the target area according to a preset rule to obtain a terrain matrix of the target area;

[0148] The training module 902 is specifically used to train the model using the terrain matrix of the target area.

[0149] This application provides a computing device, the device including a processor and a memory:

[0150] The memory is used to store computer programs and to transfer the computer programs to the processor;

[0151] The processor is used to execute the methods described in the above embodiments according to the instructions in the computer program.

[0152] This application provides a computer-readable storage medium for storing a computer program that performs the methods described in the above embodiments.

[0153] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. The device embodiments described above are merely illustrative. The units and modules described as separate components may or may not be physically separate. Furthermore, some or all of the units and modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0154] The above description is only a specific embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for predicting tower sweeping risks, characterized in that, include: Obtain the input test area; Based on the correspondence between region and terrain information, the terrain information of the area to be measured is determined; Based on the terrain information of the area to be tested and the tower sweeping risk prediction model, the prediction results of the area to be tested are obtained; The tower sweep risk prediction model is obtained by training on the terrain information of the target area, which is obtained based on the wind direction. The target area includes risk areas and normal areas. The risk areas are obtained based on the wind direction corresponding to historical tower sweep risks, and the normal areas are obtained based on the wind direction corresponding to historical no tower sweep risks. The target area is a shape area corresponding to the calculation method of the tower sweeping risk prediction model. The shape area is a rectangular area. The direction of the wind flow is from the first boundary of the rectangular area through the target position to the second boundary of the rectangular area. The distance between the target position and the second boundary is less than the distance between the target position and the first boundary. The first boundary is parallel to the second boundary, and the direction of the wind flow is perpendicular to the first boundary.

2. The method according to claim 1, characterized in that, The terrain information is represented by elevation values, and the correspondence between the region and the terrain information is represented by a digital elevation model or topographic map. The step of obtaining the terrain information of the area to be measured based on the correspondence between the area and the terrain information includes: The boundary of the area to be measured is determined, and the digital elevation model or the topographic map is cut using the boundary to obtain the elevation value within the area to be measured.

3. The method according to claim 2, characterized in that, Determining the boundary of the region to be measured includes: The area to be tested is divided into multiple sectors. The center line of each sector is used as the direction of wind flow. Based on the direction of wind flow and the center of the area to be tested, the boundary of the area to be tested is generated.

4. The method according to claim 1, characterized in that, The terrain information is represented by elevation values, wherein the elevation values ​​of the target area are represented by a terrain matrix according to a preset rule; The tower sweeping risk prediction model is a model obtained by training the terrain matrix.

5. The method according to any one of claims 1-4, characterized in that, The method further includes: When the prediction result indicates that there is a risk of tower sweeping in the area to be tested, the user is prompted with the risk area in the area to be tested and the risk area is displayed on the topographic map.

6. A model training method, characterized in that, include: Based on the direction of wind flow when a tower sweep risk occurs, the target area is determined as the risk area; Alternatively, based on the wind direction when no tower sweep risk occurs, the target area is determined as a normal area; the target area is a shape area corresponding to the calculation method of the tower sweep risk prediction model, the shape area is a rectangular area, the wind direction is from the first boundary of the rectangular area through the target position to the second boundary of the rectangular area, the distance between the target position and the second boundary is less than the distance between the target position and the first boundary, the first boundary and the second boundary are parallel, and the wind direction is perpendicular to the first boundary; Based on the correspondence between region and terrain information, the terrain information of the target region is obtained; The model is trained using the terrain information of the target area until the model meets the training stopping condition, thus obtaining the tower sweeping risk prediction model.

7. The method according to claim 6, characterized in that, The risk of tower sweeping includes the distance between the blade and the tower being less than a preset threshold, and the distance between the blade and the tower is obtained by a laser distance sensor.

8. The method according to claim 6, characterized in that, The terrain information is represented by elevation values, and the correspondence between the region and the terrain information is represented by a digital elevation model or topographic map. Obtaining the terrain information of the target region based on the correspondence between the region and the terrain information includes: The boundary of the target area is determined, and the digital elevation model or the topographic map is cut using the boundary to obtain the elevation values ​​within the target area.

9. The method according to claim 8, characterized in that, Before using the terrain information of the target area for model training, the method further includes: The elevation values ​​of the target area are adjusted according to preset rules to obtain the terrain matrix of the target area; The step of training the model using the terrain information of the target area includes: The model is trained using the terrain matrix of the target area.

10. A tower sweeping risk prediction device, characterized in that, include: The acquisition module is used to acquire the input test area; Based on the correspondence between region and terrain information, the terrain information of the area to be measured is determined; The prediction module is used to obtain the prediction result of the area to be tested based on the terrain information of the area to be tested and the tower sweeping risk prediction model; The tower sweeping risk prediction model is trained based on terrain information of a target area, which includes risk and normal areas. The target area is determined by the direction of wind flow and includes both risk and normal areas. The risk area is determined by the direction of wind flow when tower sweeping risks occurred in the past, and the normal area is determined by the direction of wind flow when tower sweeping risks did not occur in the past. The target area is a rectangular region corresponding to the calculation method of the tower sweeping risk prediction model. The direction of wind flow is from the first boundary of the rectangular region through the target location to the second boundary of the rectangular region. The distance between the target location and the second boundary is less than the distance between the target location and the first boundary. The first boundary is parallel to the second boundary, and the direction of wind flow is perpendicular to the first boundary.

11. A model training device, characterized in that, include: The acquisition module is used to determine the target area as the risk area based on the direction of the wind when a tower sweep risk occurs; Alternatively, based on the wind direction when no tower sweep risk occurs, the target area is determined as a normal area; the target area is a rectangular area corresponding to the calculation method of the tower sweep risk prediction model, the wind direction is from the first boundary of the rectangular area through the target location to the second boundary of the rectangular area, the distance between the target location and the second boundary is less than the distance between the target location and the first boundary, the first boundary and the second boundary are parallel, and the wind direction is perpendicular to the first boundary; the terrain information of the target area is obtained according to the correspondence between the area and the terrain information; The training module is used to train the model using the terrain information of the target area until the model meets the training stopping condition, thus obtaining the tower sweeping risk prediction model.

12. A computing device, characterized in that, include: The computing device includes a processor and memory: The memory is used to store computer programs and to transfer the computer programs to the processor; The processor is configured to execute the method according to any one of claims 1-9 according to instructions in the computer program.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store a computer program for performing the method according to any one of claims 1-9.

Images