WO2023050564A1

WO2023050564A1 - Layout position detection method for wind turbines, and model training method and device

Info

Publication number: WO2023050564A1
Application number: PCT/CN2021/135497
Authority: WO
Inventors: 曹长健; 焦成柱; 刘晓亚
Original assignee: 新疆金风科技股份有限公司
Priority date: 2021-09-29
Filing date: 2021-12-03
Publication date: 2023-04-06
Also published as: CN115906600A

Abstract

The present application discloses a layout position detection method for wind turbines, and a model training method and device. The layout position detection method for wind turbines comprises: dividing wind turbines in a wind farm into a plurality of sectors, and for a target sector among the plurality of sectors, obtaining current topographic data of the target sector and current wind parameter data of the wind farm; inputting the current topographic data and the current wind parameter data into a trained risk prediction model to obtain a risk prediction result of the target sector; and detecting the layout position of each wind turbine according to the risk prediction result. Therefore, whether the layout position of each wind turbine has a high-frequency vibration risk can be detected according to the correspondence among the predetermined topographic data, the predetermined wind parameter data, and a risk of overrun nacelle acceleration, in combination with the current topographic data of the target sector and the current wind parameter data of the wind farm, and manual determination is not needed, such that the effect of personal experience on a detection result can be avoided, and the accuracy of the detection result is improved.

Description

Wind turbine layout position detection method, model training method and device

Cross References to Related Applications

This application claims priority to the Chinese patent application 202111151153.2 entitled "Wind Turbine Layout Position Detection Method, Model Training Method and Device" filed on September 29, 2021, the entire content of which is incorporated herein by reference .

technical field

This application belongs to the technical field of wind farm site selection, and specifically relates to a wind turbine layout position detection method, model training method and device

Background technique

In wind farms, high-frequency vibrations of wind turbines will affect the safe operation and power generation of wind turbines. Therefore, when selecting a location for a wind turbine, it is usually necessary to detect whether there is a risk of high-frequency vibration at the layout location of the wind turbine. If there is a risk of high-frequency vibration at the layout location of the wind turbine, it may be necessary to change the layout location of the wind turbine.

At present, it is mainly based on manual experience and the terrain of the layout location to detect whether there is a risk of high-frequency vibration of the wind turbine at the layout location, and the accuracy is poor.

Contents of the invention

Embodiments of the present application provide a detection method for a layout position of a wind turbine, a model training method and a device, which can improve the accuracy of a detection result when detecting the layout position of a wind turbine.

In the first aspect, the embodiment of the present application provides a method for detecting the layout position of a wind turbine, the method comprising:

Divide multiple sectors for the wind turbines of the wind farm, and obtain the current terrain data of the target sector and the current wind parameter data of the wind farm for the target sector in the multiple sectors;

Input the current terrain data and current wind parameter data into the risk prediction model that has completed the training to obtain the risk prediction results of the target sector. The risk prediction model is used to represent the corresponding relationship between terrain data, wind parameter data and the risk of cabin acceleration exceeding the limit. The risk prediction results are used to characterize whether the target sector has high-frequency vibration risks;

According to the risk prediction results, the layout position of wind turbines is detected.

In the second aspect, the embodiment of the present application provides a method for training a risk prediction model, the method comprising:

Obtaining training samples, the training samples include the historical terrain data of the target sector in the multiple sectors of each wind turbine of the wind farm and the historical wind parameter data of the wind farm;

Determine the machine learning model used to establish the corresponding relationship between historical terrain data, historical wind parameter data and the risk of cabin acceleration exceeding the limit;

Train machine learning models based on historical terrain data and historical wind parameter data;

If the stop condition is met, the training is stopped, and the risk prediction model that has completed the training is obtained.

In a third aspect, an embodiment of the present application provides a device for detecting the layout position of a wind turbine, the device comprising:

The data acquisition module is used to divide multiple sectors for the wind turbines of the wind farm, and for the target sector in the multiple sectors, obtain the current terrain data of the target sector and the current wind parameter data of the wind farm;

The risk prediction result determination module is used to input the current terrain data and current wind parameter data into the trained risk prediction model to obtain the risk prediction result of the target sector. The risk prediction model is used to represent the terrain data, wind parameter data and cabin acceleration The corresponding relationship of overrun risks, the risk prediction results are used to characterize whether there is high-frequency vibration risk in the target sector;

The detection module is used to detect the layout position of the wind turbine according to the risk prediction result.

In a fourth aspect, the embodiment of the present application provides a training device for a risk prediction model, which includes:

The training sample obtaining module is used to obtain the training sample, the training sample includes the historical terrain data of the target sector in the multiple sectors of each wind turbine of the wind farm and the historical wind parameter data of the wind farm;

The machine learning model determination module is used to determine the machine learning model used to establish the corresponding relationship between historical terrain data, historical wind parameter data and the risk of cabin acceleration exceeding the limit;

The training module is used to train the machine learning model according to historical terrain data and historical wind parameter data;

In a fifth aspect, the embodiment of the present application provides an electronic device, including:

processor;

memory for storing computer program instructions;

When the computer program instructions are executed by the processor, the method for detecting the layout position of the wind turbine according to the first aspect, or the method for training the risk prediction model according to the second aspect is realized.

In the sixth aspect, the embodiment of the present application provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by the processor, the detection of the layout position of the wind turbine as described in the first aspect is realized method, or the method for training the risk prediction model as described in the second aspect.

In the seventh aspect, the embodiment of the present application provides a computer program product. When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device executes the method for detecting the layout position of the wind turbine as described in the first aspect, or The training method of the risk prediction model as described in the second aspect.

The embodiment of the present application provides a layout position detection method, model training method and device of wind turbines, which divides the wind turbines of the wind field into multiple sectors, and obtains the current position of the target sector in the multiple sectors. Terrain data and current wind parameter data of the wind field; input the current terrain data and current wind parameter data into the risk prediction model that has been trained to obtain the risk prediction result of the target sector; according to the risk prediction result, detect the layout position of the wind turbine. That is to say, the embodiment of the present application can detect whether the layout position of the wind turbines is There is a risk of high-frequency vibration without manual judgment, which can avoid the influence of personal experience on the test results, thereby improving the accuracy of the test results.

Description of drawings

FIG. 1 is a flow chart of a method for detecting the layout position of a wind turbine provided in an embodiment of the present application;

Fig. 2 is a top view of a sector provided by an embodiment of the present application;

Fig. 3 is a flow chart of a method for detecting the layout position of wind turbines provided in another embodiment of the present application;

Figure 4 is a schematic diagram of a point provided by the embodiment of the present application;

FIG. 5 is a flowchart of a method for training a risk prediction model provided by an embodiment of the present application;

Fig. 6 is a structural diagram of a layout position detection device for wind turbines provided by an embodiment of the present application;

FIG. 7 is a structural diagram of a training device for a risk prediction model provided by an embodiment of the present application;

FIG. 8 is a structural diagram of an electronic device provided by an embodiment of the present application.

Detailed ways

By analyzing the operation data of wind turbines, it can be seen that the high-frequency vibration of wind turbines is the main problem affecting the operation of wind turbines, for example, it can affect the operation and maintenance and power generation of wind turbines. The high-frequency vibration of wind turbines is mainly reflected in the form of over-limit acceleration of the nacelle. Therefore, the problem of high-frequency vibration of wind turbines is also the problem of over-limit acceleration of the nacelle.

Based on the current wind farm data with high-frequency vibration units, it can be known that only a few units in the wind farm have high-frequency vibration problems, and the high-frequency vibration problems occur in individual months, and the operation data of wind turbines with high-frequency vibration It can be seen from the analysis that only the acceleration of the nacelle in individual sectors of the wind turbine exceeds the limit. That is, for the wind turbines in the same wind field, the generation of the sector where the acceleration of the nacelle exceeds the limit is related to the terrain and wind parameter data of the sector.

Therefore, in order to detect the layout position of the wind turbine to be built and avoid the problem that the acceleration of the nacelle exceeds the limit in some sectors of the wind turbine to be built, the embodiment of the present application combines the sector's terrain data, wind parameter data and the acceleration of the cabin to exceed the limit. The corresponding relationship between risks, and the layout position of wind turbines are detected to improve the accuracy of the detection results.

The layout position detection method, model training method and device provided by the embodiments of the present application will be described in detail below in conjunction with the accompanying drawings through specific embodiments and application scenarios.

As shown in Figure 1, the layout position detection method of the wind turbine may include the following steps:

S110. Divide a plurality of sectors for the wind turbines of the wind farm, and for a target sector in the plurality of sectors, acquire current terrain data of the target sector and current wind parameter data of the wind farm.

S120. Input the current terrain data and current wind parameter data into the trained risk prediction model to obtain the risk prediction result of the target sector.

Among them, the risk prediction model is used to represent the corresponding relationship between terrain data, wind parameter data and the risk of cabin acceleration exceeding the limit, and the risk prediction result is used to represent whether there is high-frequency vibration risk in the target sector.

S130. Detect the layout position of the wind turbine according to the risk prediction result.

In the embodiment of the present application, the wind turbines of the wind farm are divided into multiple sectors, and for the target sector in the multiple sectors, the current terrain data of the target sector and the current wind parameter data of the wind field are obtained; the current terrain data Input the risk prediction model and the current wind parameter data into the trained risk prediction model to obtain the risk prediction result of the target sector; according to the risk prediction result, the layout position of the wind turbine is detected. That is to say, the embodiment of the present application can detect whether the layout position of the wind turbines is There is a risk of high-frequency vibration without manual judgment, which can avoid the influence of personal experience on the test results, thereby improving the accuracy of the test results.

The above steps are described in detail below, as follows:

In S110, the sector is an area obtained by dividing the terrain where the wind turbine is located with the layout position of the wind turbine as the center.

The embodiment of the present application does not specifically limit the sector division manner, for example, refer to FIG. 2 . In Figure 2, the area formed by 11.25 degrees to the left and right of the north direction of the wind turbine is a sector, that is, sector 0, and then clockwise downward, every 22.5 degrees is a sector, that is, clockwise, No. 0 The next sector of the sector is sector 1 (11.25°-33.75°), so the terrain where the wind turbines are located can be divided into 16 sectors (sector 15). Among them, point O is the center of the circle corresponding to the sector, that is, the layout position of the wind turbine.

The target sector may be one or more sectors among the plurality of sectors that have a risk of dithering. The current terrain data of the target sector is the actual terrain data of the target sector currently collected, and for the same target sector, the obtained terrain data may also be different at different collection times. In the embodiment of the present application, when it is necessary to detect the layout position of the wind turbine, the terrain data of the target sector is collected, and the terrain data is used as the current terrain data of the target sector.

The terrain data can be data reflecting sector terrain information. For example, points that can reflect sector terrain information can be determined, and the current terrain data of the target sector can be determined according to the current data of these points. The embodiment of the present application does not limit the determination process of the points, and the current data of these points may include but not limited to the elevation of these points, the elevation difference between different points, the slope, and the horizontal distance.

The elevation here is the vertical distance from the point to the datum, and the datum can be selected according to actual needs, for example, it can be a horizontal ground. It should be noted that in the embodiment of the present application, each point corresponds to the same reference plane. The elevation difference is the difference between the elevations of two points. For example, for point A and point B, the elevation of point A is h1, and the elevation of point B is h2. Then the height of point A and point B is The elevation difference is h1-h2. The slope is the ratio of the elevation difference between two points and the horizontal distance between the two points. For example, the horizontal distance between point A and point B is x _AB , then the slope of AB is (h1-h2)/x _AB , the slope of BA is (h2-h1)/x _AB .

The current wind parameter data of the wind field is the wind parameter data of the wind field collected at the same time. Exemplarily, the wind parameter data of the wind field may be measured by an anemometer tower in the wind field. A wind measuring tower is a towering tower structure used to measure wind parameter data, that is, a tower-shaped structure used to observe and record airflow near the surface. One wind measuring tower can be installed in a wind field. Under this premise, for different wind turbines in the same wind field, the corresponding wind parameter data are the same.

Exemplarily, the wind parameter data may include but not limited to the shear of wind speed above 10m/s at the hub height (shear_big), the minimum wind shear at the hub height of 6m/s-12m/s wind speed (shear_e_min), the hub height of The average wind shear of 6m/s-12m/s wind speed (shear_e_mean), the proportion of negative wind shear samples of 6m/s-12m/s wind speed at the hub height to all data (shear_min_p), and the turbulence intensity at 10m/s wind speed The maximum value (Tur_max), the SD value of turbulence intensity at 10m/s wind speed (Tur_SD), the average value of turbulence intensity at 10m/s wind speed (Tur_mean), the shear value of cut-in to cut-out average wind speed (Shear), the highest The 90% quantile (direction_transform) of the difference between the layer and the wind direction at 30 meters, the average value of the characteristic turbulence from the wind speed above 8m/s to the cut-out wind speed (Tur_over8), and the largest among the 20 samples with the 10min average wind speed greater than 10m/s The ratio of the value to the average value (Max_min_ratio), the proportion of samples with a standard deviation above 8m/s greater than 20 degrees (Over20_ratio), the maximum wind speed (MAXspeed) of 10min data, the average wind speed (Speed), the number of samples in the sector (Count ), A value of wind speed distribution Weibull fitting and K value of wind speed distribution Weibull fitting.

Among them, a quantile is a data distribution, specifically the proportion that does not exceed a certain value. Exemplarily, the 90% quantile is that for a certain data, 90% of the data in the data do not exceed a certain specific value. It should be noted that the data of 10m/s, 6m/s-12m/s, 8m/s, 30m, and 90% mentioned above are just examples, and can be adjusted as required in practical applications.

In S120, the risk prediction model is used to represent the corresponding relationship between terrain data, wind parameter data and the risk of cabin acceleration exceeding the limit. The input of the risk prediction model is terrain data and wind parameter data, and the output is the risk prediction result. Exemplarily, the risk prediction result may be yes or no, where "Yes" may indicate that the target sector has a risk of nacelle acceleration exceeding the limit, that is, the wind turbine has a risk of high-frequency vibration at the layout position; "No" It can be indicated that there is no risk of nacelle acceleration exceeding the limit in the target sector, that is, there is no risk of high-frequency vibration of the wind turbine at this layout position.

The embodiment of the present application does not limit the structure of the risk prediction model, and any model that can determine the corresponding relationship between terrain data, wind parameter data and the risk of cabin acceleration exceeding the limit can be used. For example, the XGBoost model can be used, or it can be constructed or selected according to actual needs. Before the application, the risk prediction model may be trained, and the training process of the risk prediction model may refer to the following embodiments.

In this embodiment of the application, the current terrain data and current wind parameter data of the target sector are input into the trained risk prediction model, and the trained risk prediction model can be used to determine whether the target sector has a cabin acceleration exceeding the limit. Risk, that is, whether there is a risk of high-frequency vibration in the layout position of the wind turbine, does not need to be determined manually based on experience, which avoids the unreliability and judgment error of artificial judgment based on experience, and improves the accuracy of the detection results.

In S130, for example, according to the risk prediction result, the layout location of the wind turbine may be detected, specifically, according to the risk prediction result, the detection of the layout location of the wind turbine may include the following steps:

In response to the risk of high-frequency vibration in the target sector, the target sector at the layout position of the wind turbine is closed through the yaw system, or the layout position of the wind turbine is adjusted.

In the embodiment of the present application, if the risk prediction result is yes, that is, the target sector has a risk of high-frequency vibration, in some embodiments, the yaw system may be automatically turned off the target sector. Exemplarily, the target sector is sector 1, and in the case that there is high-frequency vibration risk in sector 1, the yaw system can be turned off in sector 1, even if the yaw system is in the remaining sectors (0 No. sector, No. 2 sector - No. 15 sector) rotation.

In some embodiments, prompt information can also be sent to the user, and the prompt information can include the risk prediction result (Yes), and the adjustment strategy of the layout position determined based on the current terrain data and the current wind parameter data. In this way, the user receives When prompting information, the layout position of the wind turbine can be adjusted according to the adjustment strategy, without manual adjustment based on experience, which simplifies the user's operation and improves the accuracy of the layout position.

If the risk prediction result is no, that is, there is no high-frequency vibration risk in the target sector, for example, when only high-frequency vibration is considered, the layout position can be reserved to indicate that the layout position can be established Wind Turbine.

In this embodiment of the application, the target sector can be predicted based on the trained risk prediction model combined with the current terrain data of the target sector and the current wind parameter data of the wind field, so as to predict whether there is high-frequency vibration in the target sector If the risk prediction result is yes, the target sector can be closed through the yaw system, or the adjustment strategy can be sent to the user, so that the user can adjust the layout position based on the adjustment strategy, which simplifies the manual operation and improves the adjustment efficiency and Adjust the accuracy of the results.

Considering that adjacent sectors will influence each other, in order to improve the accuracy of the detection result, in some embodiments, the above S120 may include the following steps:

In response to the risk of high-frequency vibrations in the target sector, the risk prediction model containing the elements of the adjacent sectors that is trained is completed by inputting the current terrain data of the adjacent sectors adjacent to the target sector and the current wind parameter data of the wind field, To determine whether there is a risk of high frequency vibration in adjacent sectors.

In the embodiment of the present application, when it is determined that the target sector has the risk of the cabin accelerating beyond the limit, it can further judge the sectors adjacent to the target sector to determine whether there is a cabin in the sector adjacent to the target sector. Risk of accelerated overrun. Exemplarily, the target sector is sector 1, sector 0 and sector 2 are adjacent to sector 1, and sector 0 can be further determined if there is a risk of engine cabin acceleration exceeding the limit in sector 1 and whether there is a risk of cabin acceleration exceeding the limit in Sector 2, which can improve the accuracy of the detection results.

In order to obtain the current terrain data of the target sector, in some embodiments, with reference to Fig. 3, the layout position detection method of the wind turbine may include the following steps:

S310. Determine a first point set and a second point set associated with the layout position according to the wind turbine layout position and point determination rules.

S320. Obtain the current wind parameter data of the wind field, the layout position, the current terrain data corresponding to each point in the first point set, and each point in the second point set.

S330. Input the current terrain data and the current wind parameter data into the trained risk prediction model to obtain the risk prediction result of the target sector.

S340. Detect the layout position of the wind turbine according to the risk prediction result.

The processes of S330 and S340 are the same as those of S120 and S130 in FIG. 1 . For details, please refer to the description of S120 and S130 .

The other steps in Figure 3 are described in detail below, specifically as follows:

In S310, the first set of points and the second set of points are set of points that can reflect the terrain information of the target sector.

In some embodiments, the embodiment of the present application uses the first point set to include the first point B, the second point C, the third point D and the fourth point E, and the second point set includes the fifth point Bit B' and sixth point bit C' for example. In order to determine the first set of points and the second set of points, the above S310 may include the following steps:

S3101. Determine the first point from within the target sector according to the relationship between the first elevation of the first candidate point and the second elevation of the layout position.

Wherein, the first candidate point is a point within the target sector whose horizontal distance from the layout position satisfies a first preset condition.

The first elevation is the elevation of the first candidate point, and the second elevation is the elevation of the layout position. Exemplarily, the first preset condition can be that the horizontal distance between the first candidate point and the layout position is greater than or equal to d1, and the size of d1 can be set according to actual needs, for example, it can be set to 100 meters, that is, the first candidate point The horizontal distance between the location and layout location is greater than or equal to 100 meters.

The relationship between the first elevation and the second elevation may include that the first elevation is less than the second elevation, and the first elevation is not less than the second elevation.

Specifically, in the case that the first elevation is not less than the second elevation, it can be determined that the first point coincides with the layout position, that is, point B coincides with point A.

In the case that the first elevation is less than the second elevation, the first point whose elevation meets the preset elevation can be determined from the first area, wherein the first area is the horizontal distance from the layout position in the target sector to meet the preset distance the area between the points.

Exemplarily, the first area may be an area within the target sector between points whose horizontal distance from point A is greater than d1 and less than d2, optionally, d2=800 meters.

In some embodiments, the first low point in the first area may be determined as the first point. The low point here is the point where the elevation appears an inflection point, that is, the elevation of the points before the low point decreases sequentially, and the elevation of the points after the low point increases sequentially.

For example, refer to FIG. 4 . In the first area, the elevations of the points before point B decrease sequentially, and the elevations of points after point B increase sequentially, that is, point B is the point where the elevation appears an inflection point. Therefore, point Bit B is determined as the first point. Figure 4 takes the first area containing a low point (point B) as an example. In actual application, the first area may contain multiple low points. For example, there are two low points after point B. At this time, the point Bit B (first low) is determined as the first point.

S3102. Determine a second point, a third point, and a fourth point from within the target sector according to the layout position and the first point.

Exemplarily, when the first elevation is smaller than the second elevation, and the horizontal distance between the layout position and the first point is smaller than the first threshold, it is determined from the second area that the elevation meets the second preset condition. point, the second area is the area between the points in the target sector whose horizontal distance from the first point satisfies the third preset condition.

A point within the target sector whose horizontal distance from the second point satisfies the fourth preset condition is determined as the fourth point.

According to the second point and the fourth point, determine the third point, and the third point is located between the second point and the fourth point.

In this embodiment of the present application, for example, the first threshold may be 800 meters. The second preset condition may be that the elevation is the highest in the second area. The third preset condition may be that the horizontal distance from the first point B is greater than d1 and less than d3, optionally, d3=1000 meters. The fourth preset condition may be that the horizontal distance is greater than or equal to d4, for example, d4=1000 meters. In actual application, the first threshold, the second preset condition, the third preset condition and the fourth preset condition can be adjusted as required.

Specifically, when the horizontal distance between AB and AB is smaller than the first threshold, the point with the highest elevation may be selected from the second area as the second point. For example, referring to FIG. 4 , point C is the point with the highest elevation in the second area, so point C may be determined as the second point. Exemplarily, a point with a horizontal distance of 1000 meters from point C, that is, point E in FIG. 4 may be determined as the fourth point.

In the case that the second point and the fourth point are determined, the third point may be determined based on the second point and the fourth point.

Exemplarily, the elevation difference between the second candidate point and the third candidate point can be determined. If the elevation difference is less than the second threshold, the second candidate point is determined to be the third point. If the elevation difference is not less than In the case of the second threshold, it is determined that the third point coincides with the fourth point.

Wherein, the second candidate point and the third candidate point are points between the second point and the fourth point, the elevation of the second candidate point is not less than the elevation of the third candidate point, and the second candidate point The horizontal distance between the position and the third candidate point is a preset distance.

Specifically, it is possible to start from the second point and search with a predetermined step length, for example, it is possible to start from the second point and determine a point whose horizontal distance from the second point is a predetermined step length, and then determine the point and The elevation difference of the second point, when the elevation difference satisfies the second threshold, this point can be determined as the third point, otherwise, start from this point, continue to search with a predetermined step, and determine the next point until the elevation difference between the next point and the previous point is the second threshold, at this time the next point can be determined as the third point. Wherein, the horizontal distance between the next point and the previous point is a predetermined step. Exemplarily, the predetermined step length is 210 meters, and the second threshold is 21 meters.

Exemplarily, referring to Fig. 4, the horizontal distance between point D and point D' is 210 meters, and the elevation difference between point D and point D' is 21 meters, so point D can be determined as the third point.

In some embodiments, when the horizontal distance between AB is not less than the first threshold, it may be determined that the second point, the third point and the fourth point coincide with the first point respectively.

S3103. Determine a second point set according to the layout position, the first point, the second point and the third point.

In some embodiments, the intersection of the line between the layout position and the third point and the first straight line can be determined as the fifth point, and the first straight line is a straight line passing through the first point in the vertical direction ;

The intersection of the line between the layout position and the third point and the second straight line is determined as the sixth point, and the second straight line is a straight line passing through the second point in the vertical direction.

Exemplarily, referring to Fig. 4, point A and point D can be connected, the intersection point B' of the line AD and point B in the vertical direction is determined as the fifth point, and the line AD and point C are vertically The intersection point C' of straight lines in the vertical direction is determined as the sixth point.

It should be noted that the method of determining the first point set and the second point set is not limited to the above-mentioned embodiments, as long as the determined points can reflect the terrain information of the target sector, they can be applied to the embodiments of the present application.

In S320, for example, the terrain data of the target sector may include but not limited to: elevations of A, B, C, D, E, AB slope, BC slope, CD slope, AC slope and AD slope, AB elevation difference , AC elevation difference, BC elevation difference, BB' elevation difference and CC' elevation difference, and the horizontal distance between AB, AC, BC, and CD, where point A is the layout position.

In the embodiment of the present application, preferably, according to the layout position of the wind turbines and the point determination rules, the points reflecting the terrain information of the target sector can be determined, and then the terrain data of the target sector can be determined according to the data of these points. In this way, the terrain information of the target sector can be accurately determined, and the accuracy of the risk prediction result is improved.

When using the risk prediction model to determine the risk prediction results of the target sector, the risk prediction model needs to be trained first to improve the prediction performance of the risk prediction model.

Based on this, the embodiment of the present application also provides a method for training a risk prediction model. For example, referring to FIG. 5 , the method for training the risk prediction model may include the following steps:

S510. Obtain training samples.

Among them, the training samples include the historical terrain data of the target sector in the multiple sectors of each wind turbine of the wind farm and the historical wind parameter data/

S520. Determine a machine learning model used to establish a corresponding relationship between historical terrain data, historical wind parameter data, and risks of cabin acceleration exceeding the limit.

S530. Train the machine learning model according to the historical terrain data and historical wind parameter data; if the stop condition is met, stop the training to obtain a risk prediction model that has completed the training.

In the embodiment of this application, the historical topographic data of the target sector of the wind turbine and the historical wind parameter data of the wind field are used to train the machine learning model to determine the relationship between the historical topographic data, historical wind parameter data and the risk of nacelle acceleration exceeding the limit. In this way, the corresponding relationship can be used to detect the layout position of the wind turbine without manual detection, thereby avoiding the influence of personal experience on the detection results and improving the accuracy of the detection results.

The above steps are described in detail below, as follows:

In S510, multiple wind turbines may be included in the same wind farm, and different wind turbines may use the same division method when dividing sectors. For example, an area formed by 11.25 degrees to the left and right of the true north direction of the wind turbine can be used as a sector, and then clockwise downwards, every 22.5 degrees can be used as a sector.

In the embodiment of the present application, the historical terrain data of the target sector of each wind turbine in the wind farm and the historical wind parameter data of the wind farm are used as training samples to train the risk prediction model, which can increase the diversity of samples and improve the training effect of the model .

It should be noted that for different wind turbines in the same wind farm, the sectors with the risk of exceeding the acceleration of the nacelle can be different. For example, for the No. For wind turbine No. 11, the No. 4 sector (56.25°-78.75°) has the risk of nacelle acceleration exceeding the limit. Therefore, the target sectors of different wind turbines in the training sample can be different.

The historical terrain data is the terrain data of the target sector in the historical time period, and the historical wind parameter data is the wind parameter data of the wind field in the same historical time period. For different wind turbines in the same wind farm, their wind parameter data are the same. The specific content of the terrain data and the wind parameter data can be referred to the above-mentioned embodiments, and for the sake of brevity, details are not repeated here.

In S520, the machine learning model is a model used to establish a corresponding relationship between historical topographical data, historical wind parameter data, and the risk of cabin acceleration exceeding the limit. The embodiment of the present application does not limit the type of the machine learning model, and any model that can establish the corresponding relationship between the historical terrain data, the historical wind parameter data and the risk of cabin acceleration exceeding the limit can be used.

For example, a machine learning model can be selected from existing models. In order to ensure the reliability and stability of the machine learning model, a 5-fold cross-validation method can be used for model selection, which can ensure the reliability and stability of the selected machine learning model Exemplarily, XGBoost can be selected as the machine learning model to be trained.

In S530, for example, historical terrain data and historical wind parameter data can be input into the above-mentioned machine learning model, and the above-mentioned machine learning model outputs a sample risk prediction result, wherein the sample risk prediction result can be determined by the machine learning model according to the target sector The acceleration of the sample cabin is determined. For example, the a% quantile of the acceleration of the sample cabin in the target sector can be determined. If the a% quantile is not lower than the preset value, the sample risk prediction result is yes, otherwise it is no. Exemplarily, a=90, and the preset value is 0.049.

The sample nacelle acceleration can be determined by a machine learning model based on historical terrain data and historical wind parameter data, and the specific determination process is not limited in this embodiment of the present application.

The stop condition is the condition for stopping the training of the above machine learning model. Exemplarily, the stop condition can be that the number of training times reaches the preset number, or the loss value of the sample nacelle acceleration output by the machine learning model and the nacelle acceleration of the wind turbine in the historical time period becoming steady. In this way, a trained risk prediction model can be obtained.

In order to improve the effect of the risk prediction model, before using the training samples to train the machine learning model, you can first optimize the parameters of the machine learning model, and then use the training samples to train the machine learning model after parameter optimization, which can improve the training effect of the model .

Based on this, in some embodiments, after S520, the method may further include the following steps:

The parameters of the machine learning model are optimized by means of grid search, and the value of the model evaluation index (Area Under Curve, AUC) as an evaluation of the predictive performance of the machine learning model is obtained.

Among them, grid search is a parameter tuning method, which can be searched by selecting a small finite set of model hyperparameters, and then arranging and combining the possible values of these hyperparameters to generate all possible combination results, and get " grid".

AUC is defined as the area under the ROC curve. The ROC curve is based on a series of different binary classification methods (cutoff value or threshold), with the true positive rate (sensitivity) as the ordinate, and the false positive rate (1-specificity) The curve plotted for the abscissa reflects the ability of the model to classify imbalanced samples.

A set of hyperparameters corresponds to an AUC value. The larger the value of AUC, the better the classification effect of the model and the better the performance. In the embodiment of the present application, the above-mentioned training samples can be divided into a training set and a verification set, and the above-mentioned machine learning model can be trained by using the training set to obtain the AUC value (training AUC) of each group of hyperparameters for the training set, and then use the verification set to verify Perform verification to obtain the AUC value (verification AUC) of each group of hyperparameters for the verification set, and use the hyperparameter with the largest AUC value in the verification AUC as the parameter optimization result of the above machine learning model.

After parameter optimization, the machine learning model after parameter optimization can be trained using training samples to obtain a trained risk prediction model, which can improve the training effect of the risk prediction model and further improve the accuracy of the risk prediction results.

In order to improve the training efficiency of the model, in some embodiments, the above S530 may include the following steps:

According to the wind direction corresponding to the historical wind parameter data within a predetermined time, determine the first target sector, where the first target sector is a sector in the target sector;

Preprocessing the historical wind parameter data to obtain the first historical wind parameter data;

A machine learning model is trained according to the historical terrain data of the first target sector and the first historical wind parameter data.

In the embodiment of the present application, the wind direction of the historical wind parameter data may or may not belong to the target sector. If the wind direction of the historical wind parameter data within a predetermined time period belongs to the target sector, the sector to which the wind direction belongs may be determined as the first target sector.

Exemplarily, the target sectors are sector 0 and sector 1, and the wind direction of the historical wind parameter data within a predetermined time period belongs to sector 1, then sector 1 may be determined as the first target sector. As another example, if the wind direction of the historical wind parameter data within the predetermined time period belongs to sector 0, then sector 0 may be determined as the first target sector.

If the wind direction of the historical wind parameter data within the predetermined time does not belong to the target sector, as an option, the target sector may be abandoned, that is, the training samples do not include the historical terrain data and historical wind parameter data of the target sector.

In the embodiment of the present application, the target sector is screened using the wind direction of the historical wind parameter data within a predetermined period of time to obtain the first target sector that matches the wind direction. When the above machine learning model is used, the accuracy of the model training results can be improved.

In some embodiments, considering that the above-mentioned wind parameter data has many dimensions, dimension reduction processing may be performed on the above-mentioned wind parameter data, so as to improve the training efficiency of the model.

Considering that the correlation between some wind parameter data and terrain data is small, that is, the impact on the high-frequency vibration risk of the unit is small, in some embodiments, the historical wind parameter data can be preprocessed to obtain the first historical wind parameter data.

Exemplarily, feature screening may be performed on historical wind parameter data to extract historical wind parameter data with specific wind parameter characteristics; standardization processing is performed on the extracted historical wind parameter data to obtain first historical wind parameter data.

Specific wind parameter features can be features that are highly correlated with terrain data. Exemplarily, specific wind parameter features can include but not limited to shear_big, shear_e_mean, Tur_max, Tur_mean, speed, Over20_ratio, direction_transform, Max_min_ratio, etc., each parameter The meaning of can refer to the above-mentioned examples.

In order to avoid the impact of extreme data on the training result, for example, the embodiment of the present application performs standardized processing on the extracted historical wind parameter data of a specific wind parameter feature to obtain the first historical wind parameter data.

Exemplarily, the extracted historical wind parameter data can be standardized by the following formula:

Among them, x ^* is the historical wind ginseng data after the standardization of a specific wind ginseng feature, that is, the first historical wind ginseng data, x is the historical wind ginseng data before the standardization of the specific wind ginseng feature, x _max and x _min are the The historical maximum value and historical minimum value of a specific wind parameter characteristic within a predetermined period of time.

After the first target sector and the first historical wind parameter data are determined, the terrain data of the first target sector can be merged with the first historical wind parameter data, and then the above-mentioned machine learning model can be trained with the combined data, which can improve Model training efficiency and training effect.

In the embodiment of this application, by establishing a machine learning model and using training samples to train the machine learning model, the trained machine learning model (risk prediction model) can automatically determine the corresponding risk based on the input terrain data and wind parameter data. Whether there is a risk of cabin acceleration exceeding the limit in the sector, the automatic detection of the layout position is realized, and the unreliability and judgment error of artificial judgment based on experience are avoided.

Considering the mutual influence between adjacent sectors, in order to improve the training effect of the model, in some embodiments, as an option, the historical terrain data of the sectors adjacent to the first target sector can also be obtained, and the training The above machine learning model.

Specifically, the above-mentioned "training machine learning model according to historical terrain data and historical wind parameter data" may include the following steps:

According to the historical terrain data of the first target sector, the historical terrain data of adjacent sectors adjacent to the first target sector, and the first historical wind parameter data, a machine learning model including elements of adjacent sectors is trained.

Exemplarily, the first target sector is sector 1, and the sectors adjacent to the first target sector include sector 0 and sector 2, then the Historical terrain data. Then, use sector 0 (the sector adjacent to the first target sector), sector 1 (the first target sector) and sector 2 (the sector adjacent to the first target sector) The historical terrain data and the first historical wind parameter data train the above machine learning model, which can improve the training effect of the model. Based on the same idea, the embodiment of the present application also provides a device for detecting the layout position of a wind turbine. The device for detecting the layout position of a wind turbine provided in the embodiment of the present application will be described in detail below with reference to FIG. 6 .

As shown in Figure 6, the layout position detection device of the wind turbine may include:

The data acquisition module 61 is used to divide a plurality of sectors for the wind turbines of the wind farm, and for a target sector in the multiple sectors, acquire the current terrain data of the target sector and the current wind parameter data of the wind farm;

The risk prediction result determination module 62 is used to input the current terrain data and current wind parameter data into the risk prediction model that has completed the training to obtain the risk prediction result of the target sector. The risk prediction model is used to represent the terrain data, wind parameter data and engine room The corresponding relationship of the risk of acceleration exceeding the limit, the risk prediction result is used to indicate whether there is a high-frequency vibration risk in the target sector;

The detection module 63 is configured to detect the layout position of the wind turbine according to the risk prediction result.

The layout location detection method of the wind turbines provided in the embodiment of the present application divides the wind turbines of the wind field into multiple sectors, and for the target sector in the multiple sectors, obtains the current terrain data of the target sector and the current wind speed of the wind field. Parameter data; input the current terrain data and current wind parameter data into the risk prediction model that has been trained to obtain the risk prediction results of the target sector; according to the risk prediction results, detect the layout position of the wind turbine. That is to say, the embodiment of the present application can detect whether the layout position of the wind turbines is There is a risk of high-frequency vibration, and no manual judgment is required, so that the influence of personal experience on the test results can be avoided, and the accuracy of the test results can be improved.

In some embodiments, the detection module 63 is specifically used for:

In some embodiments, the layout position detection device of the wind turbine may also include:

The point set determination module is used to divide multiple sectors for the wind turbines of the wind farm in the acquisition module 61, and for the target sector in the multiple sectors, obtain the current terrain data of the target sector and the current wind parameter data of the wind farm Before, according to the layout position and point determination rules of the wind turbine, determine the first point set and the second point set associated with the layout position;

Obtain module 61, specifically for:

Obtain the current terrain data corresponding to the layout position, each point in the first point set, and each point in the second point set.

In some embodiments, the first point set includes a first point, a second point, a third point and a fourth point;

Point set determination module, including:

The first determination unit is used to determine the first point from within the target sector according to the relationship between the first elevation of the first candidate point and the second elevation of the layout position, the first candidate point being within the target sector A point whose horizontal distance from the layout position satisfies the first preset condition;

The second determination unit is used to determine the second point, the third point and the fourth point from the target sector according to the layout position and the first point;

The third determining unit is configured to determine the second point set according to the layout position, the first point, the second point and the third point.

In some embodiments, the second determination unit is specifically used to:

When the first elevation is less than the second elevation, and the horizontal distance between the layout position and the first point is less than the first threshold, determine the second point whose elevation satisfies the second preset condition from the second area, the first The second area is the area between the points in the target sector whose horizontal distance from the first point satisfies the third preset condition;

Determine the point in the target sector whose horizontal distance from the second point satisfies the fourth preset condition as the fourth point;

In some embodiments, the second set of spots includes a fifth spot and a sixth spot;

The third determination unit is specifically used for:

The intersection point of the connection line between the layout position and the third point and the first straight line is determined as the fifth point, and the first straight line is a straight line passing through the first point in the vertical direction;

The wind turbine layout position detection device provided in the embodiment of the present application can realize the various processes in the embodiments of the wind turbine layout position detection method shown in Fig. 1-4, and will not be repeated here to avoid repetition.

Based on the same idea, an embodiment of the present application also provides a training device for a risk prediction model. The following describes in detail the training device for a risk prediction model provided in the embodiment of the present application with reference to FIG. 7 .

As shown in Figure 7, the training device of the risk prediction model may include:

The training sample obtaining module 71 is used to obtain the training sample, the training sample includes the historical terrain data of the target sector in the multiple sectors of each wind turbine of the wind farm and the historical wind parameter data of the wind farm;

The machine learning model determination module 72 is used to determine the machine learning model used to establish the corresponding relationship between historical terrain data, historical wind parameter data and the risk of cabin acceleration exceeding the limit;

The training module 73 is used to train the machine learning model according to historical terrain data and historical wind parameter data;

In some embodiments, the training module 73 includes:

A determining unit, configured to determine a first target sector according to a wind direction corresponding to historical wind parameter data within a predetermined time period, where the first target sector is a sector in the target sector;

A preprocessing unit, configured to preprocess the historical wind parameter data to obtain the first historical wind parameter data;

The training unit is used to train the machine learning model according to the historical terrain data and the first historical wind parameter data of the first target sector.

In some embodiments, the preprocessing unit is specifically used for:

Perform feature screening on historical wind parameter data to extract historical wind parameter data with specific wind parameter characteristics;

The extracted historical wind parameter data is standardized to obtain the first historical wind parameter data.

In some embodiments, the training device of the risk prediction model may also include:

The parameter optimization module is used to optimize the parameters of the machine learning model by means of grid search after the training module 71 trains the machine learning model according to the historical terrain data and the first historical wind parameter data of the first target sector, The value of AUC as a model checking index for evaluating the predictive performance of the machine learning model is obtained.

In some embodiments, the training module 73 is specifically used for:

According to the historical terrain data of the first target sector, the historical terrain data of adjacent sectors adjacent to the first target sector, and the first historical wind parameter data, a machine learning model including elements of adjacent sectors is trained. The risk prediction model training device provided in the embodiment of the present application can implement each process in the risk prediction model training method embodiment shown in FIG. 5 , and details are not repeated here to avoid repetition.

Based on the same idea, an embodiment of the present application also provides an electronic device, which may be a mobile electronic device or a non-mobile electronic device.

As shown in FIG. 8, the electronic device may include a processor 81 and a memory 82 for storing computer program instructions.

The processor 81 may include a central processing unit (Central Processing Unit, CPU), or a specific integrated circuit (Application Specific Integrated Circuit, ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

Memory 82 may include mass storage for data or instructions. By way of example and not limitation, memory 82 may include a hard disk drive (Hard Disk Drive, HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a Universal Serial Bus (Universal Serial Bus, USB) drive or two or more Combinations of multiple of the above. In one example, memory 82 may include removable or non-removable (or fixed) media, or memory 82 may be a non-volatile solid-state memory. In one example, the memory 82 may be a read only memory (Read Only Memory, ROM). In one example, the ROM can be mask programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), electrically rewritable ROM (EAROM), or flash memory or both. A combination of one or more of the above.

The processor 81 reads and executes the computer program instructions stored in the memory 82 to realize the method in the embodiment shown in FIGS. The effect is described for brevity and will not be repeated here.

In an example, the electronic device may further include a communication interface 83 and a bus 84 . Wherein, as shown in FIG. 8 , the processor 81 , the memory 82 , and the communication interface 83 are connected through a bus 84 and complete mutual communication.

The communication interface 83 is mainly used to realize the communication between various modules, devices and/or devices in the embodiments of the present application.

Bus 84 includes hardware, software, or both, and couples the various components of the electronic device to each other.

The electronic device divides a plurality of sectors for the wind turbines of the wind farm, and for a target sector among the multiple sectors, after obtaining the current terrain data of the target sector and the current wind parameter data of the wind farm, it can execute the The method for detecting the layout position of the wind turbine, thereby realizing the method for detecting the layout position of the wind turbine described in conjunction with FIGS. 1-4 and the device for detecting the layout of the wind turbine described in FIG. 6 .

After the electronic device obtains the training samples, it can also execute the risk prediction model training method in the embodiment of the present application, thereby realizing the risk prediction model training method described in conjunction with FIG. 5 and the risk prediction model training device described in FIG. 7 .

In addition, in combination with the method for detecting the layout position of wind turbines or the method for training the risk prediction model in the above embodiments, the embodiments of the present application may provide a computer storage medium for implementation. Computer program instructions are stored on the computer storage medium; when the computer program instructions are executed by a processor, any method for detecting the layout position of a wind turbine or the method for training a risk prediction model in the above-mentioned embodiments is implemented.

In addition, in combination with the method for detecting the layout position of wind turbines in the above embodiments, or the method for training the risk prediction model, the embodiments of the present application may provide a computer program product for implementation. When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device executes the method for detecting the layout position of the wind turbine in the above embodiment, or the method for training the risk prediction model in the above embodiment.

It is to be understood that the application is not limited to the specific configurations and processes described above and shown in the figures. For conciseness, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present application is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order of the steps after understanding the spirit of the present application.

The functional blocks shown in the above structural block diagrams may be implemented as hardware, software, firmware or a combination thereof.

It should also be noted that the exemplary embodiments mentioned in this application describe some methods or systems based on a series of steps or devices. However, the present application is not limited to the order of the above steps, that is to say, the steps may be performed in the order mentioned in the embodiment, or may be different from the order in the embodiment, or several steps may be performed simultaneously.

Aspects of the embodiments of the present application are described above with reference to flowcharts and/or block diagrams of methods, apparatuses (systems) and computer program products according to the embodiments of the present application. It will be understood that each block of the flowchart and/or block diagrams, and combinations of blocks in the flowchart and/or block diagrams, can be implemented by computer program instructions.

The embodiments of the present application have been described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementations. The above-mentioned specific implementations are only illustrative and not restrictive. Those of ordinary skill in the art will Under the inspiration of this application, without departing from the purpose of this application and the scope of protection of the claims, many forms can also be made, all of which belong to the protection of this application.

Aspects of the embodiments of the present application are described above with reference to flowcharts and/or block diagrams of methods, apparatuses (systems) and computer program products according to the embodiments of the present application. It will be understood that each block of the flowchart and/or block diagrams, and combinations of blocks in the flowchart and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine such that execution of these instructions via the processor of the computer or other programmable data processing apparatus enables Implementation of the functions/actions specified in one or more blocks of the flowchart and/or block diagrams. Such processors may be, but are not limited to, general purpose processors, special purpose processors, application specific processors, or field programmable logic circuits. It can also be understood that each block in the block diagrams and/or flowcharts and combinations of blocks in the block diagrams and/or flowcharts can also be realized by dedicated hardware for performing specified functions or actions, or can be implemented by dedicated hardware and combination of computer instructions.

Claims

A method for detecting the layout position of a wind turbine, comprising:

Divide a plurality of sectors for the wind turbines of the wind farm, and for a target sector in the plurality of sectors, acquire current terrain data of the target sector and current wind parameter data of the wind farm;

Inputting the current terrain data and the current wind parameter data into the trained risk prediction model to obtain the risk prediction result of the target sector, the risk prediction model is used to characterize the terrain data, wind parameter data and cabin acceleration The corresponding relationship of overrun risk, the risk prediction result is used to characterize whether the target sector has high-frequency vibration risk;

According to the risk prediction result, the layout position of the wind turbine is detected.
The method according to claim 1, wherein the detecting the layout position of the wind turbine according to the risk prediction result comprises:

In response to the risk of high-frequency vibration in the target sector, the target sector of the layout position of the wind turbine is closed through the yaw system, or the layout position of the wind turbine is adjusted.
The method according to claim 1, wherein the detecting the layout position of the wind turbine according to the risk prediction result further comprises:

In response to the risk of high-frequency vibrations in the target sector, the training is completed by inputting the current topographic data of the adjacent sectors adjacent to the target sector and the current wind parameter data of the wind field, including the adjacent sectors. The risk prediction model of the elements of the zone is used to determine whether there is high-frequency vibration risk in the adjacent sector.
The method according to any one of claims 1-3, wherein the wind turbines for the wind farm are divided into multiple sectors, and for a target sector in the multiple sectors, the target sector is obtained Before the current terrain data and the current wind parameter data of the wind field, the method also includes:

Determining a first set of points and a second set of points associated with the layout position according to the layout position and point determination rules of the wind turbine;

The acquisition of the current terrain data of the target sector includes:

Obtain the current terrain data corresponding to the layout position, each point in the first point set, and each point in the second point set.
The method according to claim 4, wherein said first point set comprises a first point, a second point, a third point and a fourth point;

The determining the first point set and the second point set associated with the layout position according to the layout position and point determination rules includes:

According to the relationship between the first elevation of the first candidate point and the second elevation of the layout position, determine the first point from within the target sector, the first candidate point being the target sector A point whose horizontal distance from the layout position satisfies the first preset condition;

determining the second point, the third point and the fourth point from within the target sector according to the layout position and the first point;

The second point set is determined according to the layout position, the first point, the second point and the third point.
The method according to claim 5, wherein, according to the layout position and the first point, the second point, the third point and the fourth point are determined from the target sector ,include:

When the first elevation is less than the second elevation, and the horizontal distance between the layout position and the first point is less than a first threshold, determine that the elevation from the second area satisfies the second preset The second point of the condition, the second area is the area between the points in the target sector whose horizontal distance from the first point satisfies the third preset condition;

determining a point within the target sector whose horizontal distance from the second point satisfies a fourth preset condition as the fourth point;

The third point is determined according to the second point and the fourth point, and the third point is located between the second point and the fourth point.
The method according to claim 5, wherein the second point set includes a fifth point and a sixth point;

The determining the second point set according to the layout position, the first point, the second point and the third point includes:

Determining the intersection of the line between the layout position and the third point and the first straight line as the fifth point, and the first straight line passes through the first point in a vertical straight line in direction;

Determining the intersection of the line between the layout position and the third point and the second straight line as the sixth point, and the second straight line passes through the second point in the vertical direction straight line.
A training method for a risk prediction model, comprising:

Acquiring training samples, the training samples including the historical topographic data of the target sector in the multiple sectors of each wind turbine of the wind farm and the historical wind parameter data of the wind farm;

Determine the machine learning model used to establish the corresponding relationship between historical terrain data, historical wind parameter data and the risk of cabin acceleration exceeding the limit;

training the machine learning model according to the historical terrain data and the historical wind parameter data;

If the stop condition is met, the training is stopped, and the risk prediction model that has completed the training is obtained.
The method according to claim 8, wherein said training said machine learning model according to said historical terrain data and said historical wind parameter data comprises:

Determining a first target sector according to a wind direction corresponding to historical wind parameter data within a predetermined time period, where the first target sector is a sector in the target sector;

Preprocessing the historical wind parameter data to obtain the first historical wind parameter data;

The machine learning model is trained according to the historical terrain data of the first target sector and the first historical wind parameter data.
The method according to claim 9, wherein said preprocessing the historical wind parameter data to obtain the first historical wind parameter data comprises:

Perform feature screening on the historical wind ginseng data to extract historical wind ginseng data with specific wind ginseng characteristics;

The extracted historical wind parameter data is standardized to obtain the first historical wind parameter data.
The method according to any one of claims 8-10, wherein after said determining the machine learning model used to establish the corresponding relationship between historical terrain data, historical wind parameter data and the risk of cabin acceleration exceeding the limit, said method Also includes:

The parameters of the machine learning model are optimized by means of grid search, and the value of the model evaluation index AUC used to evaluate the predictive performance of the machine learning model is obtained.
The method according to claim 8, wherein said training said machine learning model according to said historical terrain data and said historical wind parameter data further comprises:

Determining a first target sector according to a wind direction corresponding to historical wind parameter data within a predetermined time period, where the first target sector is a sector in the target sector;

Preprocessing the historical wind parameter data to obtain the first historical wind parameter data;

According to the historical terrain data of the first target sector, the historical terrain data of the adjacent sectors adjacent to the first target sector, and the first historical wind parameter data, train the machine learning model.
A layout position detection device for a wind turbine, comprising:

A data acquisition module, configured to divide a plurality of sectors for the wind turbines of the wind farm, and for a target sector in the plurality of sectors, acquire the current terrain data of the target sector and the current wind parameters of the wind farm data;

A risk prediction result determination module, configured to input the current terrain data and the current wind parameter data into the trained risk prediction model to obtain the risk prediction result of the target sector, and the risk prediction model is used to characterize the terrain data, wind parameter data, and the corresponding relationship between the risk of cabin acceleration exceeding the limit, and the risk prediction result is used to characterize whether there is a high-frequency vibration risk in the target sector;

The detection module is configured to detect the layout position of the wind turbine according to the risk prediction result.
A training device for a risk prediction model, comprising:

A training sample acquisition module, configured to acquire a training sample, the training sample including the historical topographic data of the target sector in the multiple sectors of each wind turbine in the wind farm and the historical wind parameter data of the wind farm;

The machine learning model determination module is used to determine the machine learning model used to establish the corresponding relationship between historical terrain data, historical wind parameter data and the risk of cabin acceleration exceeding the limit;

A training module, configured to train the machine learning model according to the historical terrain data and the historical wind parameter data;

If the stop condition is met, the training is stopped, and the risk prediction model that has completed the training is obtained.
An electronic device comprising:

processor;

memory for storing computer program instructions;

When the computer program instructions are executed by the processor, the wind turbine layout position detection method according to any one of claims 1-7 is realized, or the method according to any one of claims 8-12 is realized. Methods for training risk prediction models.
A computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method for detecting the layout position of a wind turbine according to any one of claims 1-7 is implemented, Or the training method of the risk prediction model as described in any one in claim 8-12.
A computer program product, characterized in that, when the instructions in the computer program product are executed by the processor of the electronic device, the electronic device executes the layout of the wind turbine according to any one of claims 1-7 A position detection method, or a method for training a risk prediction model according to any one of claims 8-12.