A Method and System for Detecting Wind Turbine Tower Deformation Based on Unmanned Aerial Vehicles

Abstract

This invention provides a method and system for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs), relating to the field of wind turbine tower deformation detection technology. The method includes the following steps: scanning a wind turbine tower in a standard state using a UAV equipped with a detection device to form baseline point cloud data of the wind turbine tower; performing M consecutive scans of the wind turbine tower to form M sets of detection point cloud data of the wind turbine tower; performing a first deformation analysis to obtain a first deformation type of the wind turbine tower; if the first deformation type is sudden deformation, acquiring current climate data and performing a second deformation analysis to obtain a second deformation type of the wind turbine tower; if either the first or second deformation type is no deformation, outputting a detection result of no deformation; if the first deformation type is continuous deformation or the second deformation type is deformation, performing a third deformation analysis to output the deformation type and corresponding deformation location. This invention has the advantage of achieving more accurate and reliable deformation detection of wind turbine towers.

Description

Technical Field

[0001] This invention relates to the field of wind turbine tower deformation detection technology, and more specifically, to a method and system for wind turbine tower deformation detection based on unmanned aerial vehicles (UAVs). Background Technology

[0002] Wind power is a common renewable energy source, and the wind turbine tower is an important component of the wind turbine unit. Under the long-term influence of high wind speeds and turbine loads, it is prone to structural deformation. To ensure the reliable operation of the wind turbine, regular deformation inspections of its tower are necessary.

[0003] Traditional wind turbine tower inspection methods rely heavily on manual inspections or the deployment of fixed sensors to acquire critical structural data. However, manual inspections are limited by geographical environment and weather conditions, resulting in low inspection efficiency and safety hazards. Furthermore, the limited number of sampling points makes it difficult to achieve comprehensive, high-precision real-time monitoring of the entire tower structure. Existing technologies have addressed this by employing drones equipped with high-precision lidar or visual sensors to create point cloud models of the tower's shape, which are then used for deformation analysis. However, existing technologies still have limitations. For example, in the operating environment of wind turbines, the data acquisition process for point cloud modeling involves significant disturbances, making it difficult to distinguish between structural deformation and transient disturbances caused by environmental factors during data analysis, leading to a high misjudgment rate.

[0004] Therefore, it is urgent to optimize the wind turbine tower deformation detection method based on UAV point cloud data in order to achieve more accurate and reliable wind turbine tower deformation detection. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs), which achieves more accurate and reliable deformation detection of wind turbine towers.

[0006] This invention is achieved through the following technical solution:

[0007] The method for detecting wind turbine tower deformation based on drones includes the following steps:

[0008] By using a drone equipped with a detection device to scan the wind turbine tower in a standard state, a baseline point cloud data of the wind turbine tower is generated. The standard state is when the wind speed and temperature are both within the corresponding threshold range.

[0009] By using a drone equipped with a detection device to scan the wind turbine tower M times consecutively along the same location or path where the baseline point cloud data is collected, M sets of detection point cloud data of the wind turbine tower are generated.

[0010] Based on the baseline point cloud data and the M group of detection point cloud data, the first deformation analysis is performed to obtain the first deformation type of the wind turbine tower. The first deformation type is no deformation, continuous deformation, or sudden deformation.

[0011] If the first deformation type is sudden deformation, the current climate data is acquired and a second deformation analysis is performed to correct the results of the first deformation analysis and obtain the second deformation type of the wind turbine tower. The second deformation type is no deformation or deformation exists.

[0012] If the first deformation type or the second deformation type is no deformation, the output detection result is no deformation.

[0013] If the first deformation type is continuous deformation or the second deformation type is existing deformation, then based on the reference point cloud data and M sets of detected point cloud data, a third deformation analysis is performed, and the deformation type and the corresponding deformation location are output.

[0014] Preferably, the method for the first deformation analysis is as follows:

[0015] Using the baseline point cloud data as a reference, deformation analysis is performed on each group of the detected point cloud data to obtain the degree of deformation of each group of the detected point cloud data.

[0016] The second deformation type of the wind turbine tower is obtained by comprehensively evaluating the degree of deformation of the detection point cloud data of each group.

[0017] Preferably, the method for performing deformation analysis on each group of the detected point cloud data is as follows: For the m-th group of the detected point cloud data, the following operations are performed: :

[0018] Acquire the detection point cloud data The reference point cloud data is obtained by analyzing key points and the first vectors to each key point during drone scanning. The key points and the second vector to each key point during drone scanning;

[0019] Align and overlap the nth key point of the detected point cloud data with the nth key point of the reference point cloud data. ,get Alignment point cloud map;

[0020] Based on the aligned point cloud map, calculate the probability of suspected deformation of the wind turbine tower under the m-th group of detected point cloud data. ;

[0021] Calculate all The mean and variance;

[0022] If the average value is greater than a preset first average value threshold and the variance is less than a preset first variance threshold, then the first deformation type is determined to be a continuous deformation.

[0023] If the average value is greater than the first average value threshold and the variance is greater than the first variance threshold, then the first deformation type is determined to be a sudden deformation.

[0024] In all other cases, the first deformation type is determined to be no deformation.

[0025] Preferably, the calculation of the probability of suspected deformation of the wind turbine tower under the detection point cloud data of the m-th group is... The method is as follows:

[0026] The probability of suspected deformation of the wind turbine tower under the detection point cloud data of the m-th group is calculated. The method is as follows:

[0027] Feature extraction is performed based on the aligned point cloud map, the first vector, and the second vector of the m-th group of detected point cloud data to obtain the feature values ​​of the m-th group of detected point cloud data. :

[0028] ;

[0029] ;

[0030] in, Let N be the global coordinates of the j-th point in the n-th aligned point cloud map within the m-th group of the detected point cloud data, where N is the number of aligned point cloud maps and J is the total number of points in the detected point cloud data. The distance in the reference point cloud data of the nth aligned point cloud map The global coordinates of the nearest point The threshold is dynamic, and k and b are the training weights and biases obtained by fitting or training based on the training samples, respectively. and Let be the first vector of the detected point cloud data in the nth aligned point cloud map and the second vector of the reference point cloud data in the nth aligned point cloud map, respectively, and let if be the truth function. Determine the modulus length;

[0031] Based on the eigenvalues Calculate the probability of suspected deformation :

[0032] .

[0033] Preferably, the method for performing the second deformation analysis is as follows:

[0034] Obtain the current real-time wind speed. If the real-time wind speed is less than the wind speed threshold, then determine that the second deformation type is deformation.

[0035] If the real-time wind speed is not less than the wind speed threshold, the deformation change is analyzed based on the reference point cloud data and multiple sets of detection point cloud data to determine whether the second deformation type is deformation present or deformation absent.

[0036] Preferably, the method for analyzing deformation changes based on reference point cloud data and multiple sets of detected point cloud data is as follows:

[0037] The highest point along the z-axis in the reference point cloud data is obtained as the first analysis point, and its global coordinates are obtained. ;

[0038] The highest point along the z-axis in the m-th group of detected point cloud data is obtained as the second analysis point, and its global coordinates are acquired. , ;

[0039] Obtain the distance from the first analysis point to the m-th second analysis point respectively. :

[0040] ;

[0041] in, Determine the modulus length;

[0042] Get all The minimum value and variance;

[0043] If the minimum value is less than a preset minimum value threshold and the variance is greater than a preset second variance threshold, then the second deformation type is determined to be no deformation.

[0044] In other cases, the second deformation type is determined to be deformation.

[0045] Preferably, the method for performing the third deformation analysis is as follows:

[0046] The average value of the offset vector, the average value of the offset distance, and the average value of the global coordinates of each point in the M sets of detection point cloud data are obtained. The starting point of the offset vector of a point in the detection point cloud data is the point closest to that point in the reference point cloud data, and the ending point of the offset vector is that point in the detection point cloud data.

[0047] The offset points are filtered according to their degree of offset and grouped according to global coordinates. They are then located according to global coordinates, with each group of offset points representing a deformation position.

[0048] Deformation type analysis is performed on each set of offset points based on the offset vector.

[0049] Preferably, the method of filtering offset points according to the degree of offset and grouping offset points according to global coordinates is as follows:

[0050] Step S401: Obtain points whose average offset value is greater than a preset offset threshold, and form a global offset point group;

[0051] Step S402: Initialize group number v=1;

[0052] Step S403: Randomly select a reference point from the global offset point group, and based on the average value of the global coordinates, find all nearby points in the global offset point group whose distance from the reference point is less than a preset distance threshold.

[0053] Step S404: If the number of nearby points is less than a preset first number threshold, then the reference point is removed from the global offset point group and the process returns to step S403;

[0054] If the number of nearby points is not less than the preset first number threshold, then the reference point and nearby points are put into the vth group of offset points and deleted from the global offset point group, and proceed to the next step S405.

[0055] Step S405: Update the group number v=v+1, and then return to step S403 until the number of remaining points in the global offset point group is less than the preset second quantity threshold.

[0056] Preferably, the method for performing deformation type analysis on each group of offset points based on the offset vector is as follows: For the r-th group of offset points, perform the following operation, where r = 1, 2, ..., R, and R is the total number of groups of offset points obtained by grouping:

[0057] Take any offset point from the r-th set of offset points as the reference offset point, and use the average of its offset vectors as the reference offset vector. ;

[0058] Obtain the angle between the average of the offset vectors of all other offset points in the r-th offset point group and the average of the reference offset vector:

[0059] ;

[0060] in, The average value of the offset vector of the t-th offset point in the r-th offset point group. to the reference offset vector The angle between the reference points, where the t-th offset point is not the reference offset point. Determine the modulus length;

[0061] Get all If the variance is greater than a preset variance threshold, the deformation type is determined to be expansion deformation or fracture deformation; if the variance is not greater than the preset variance threshold, the deformation type is determined to be tilt deformation.

[0062] The present invention also provides a wind turbine tower deformation detection system based on unmanned aerial vehicles (UAVs), applied to the aforementioned UAV-based wind turbine tower deformation detection method, comprising:

[0063] The baseline point cloud data acquisition module is used to scan the wind turbine tower in a standard state using a drone equipped with a detection device to form baseline point cloud data of the wind turbine tower. The standard state is when the wind speed and temperature are both within the corresponding threshold range.

[0064] The detection point cloud data acquisition module is used to scan the wind turbine tower continuously M times using a detection device mounted on a drone, forming M sets of detection point cloud data of the wind turbine tower;

[0065] The first deformation analysis module is used to perform the first deformation analysis based on the benchmark point cloud data and M sets of detection point cloud data to obtain the first deformation type of the wind turbine tower. The first deformation type is no deformation, continuous deformation, or sudden deformation.

[0066] The second deformation analysis module is used to acquire current climate data and perform a second deformation analysis when the first deformation type is sudden deformation, thereby correcting the results of the first deformation analysis and obtaining the second deformation type of the wind turbine tower. The second deformation type is either no deformation or deformation exists.

[0067] The deformation-free output module is used to output the detection result as deformation-free when the first deformation type or the second deformation type is deformation-free.

[0068] The third deformation analysis module is used to perform a third deformation analysis based on the reference point cloud data and M sets of detection point cloud data when the first deformation type is continuous deformation or the second deformation type is deformation. It outputs the deformation type and the corresponding deformation location.

[0069] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0070] This invention establishes benchmark point cloud data of wind turbine towers under standard conditions, providing a precise comparison template for subsequent inspections and effectively improving the accuracy of deformation identification.

[0071] This invention introduces a continuous M-scanning mechanism, which can capture the structural change trend of the tower at different time points. This helps to conduct more levels of deformation analysis through multiple sets of data, determine whether the sudden deformation is caused by environmental factors rather than structural problems, and improve the accuracy of the analysis.

[0072] In performing the first deformation analysis, this invention uses a multi-point alignment mechanism and a vector comparison mechanism for multiple sets of point cloud data to minimize the impact of data acquisition errors on deformation analysis, thereby further improving the reliability of deformation judgment.

[0073] The second deformation analysis of this invention introduces climate data after detecting sudden deformation, extracts the disturbance characteristics of excessive wind speed to achieve further correction analysis of the deformation situation, and avoids misjudgment caused by instantaneous wind force or temperature changes.

[0074] This invention uses drones to carry detection devices, which improves detection efficiency while reducing safety risks, and is easy to promote and implement. Attached Figure Description

[0075] Figure 1 This is a flowchart illustrating the method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs) provided in Embodiment 1 of the present invention.

[0076] Figure 2 This is a schematic diagram of the wind turbine tower deformation detection system based on UAV provided in Embodiment 2 of the present invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0078] Example 1

[0079] This embodiment provides a method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs). See [link / reference]. Figure 1 This includes the following steps:

[0080] Step S1: Scan the wind turbine tower in a standard state using a drone equipped with a detection device to generate reference point cloud data of the wind turbine tower. The standard state is when the wind speed and temperature are both within the corresponding threshold range.

[0081] Step S2: Using a drone equipped with a detection device, the wind turbine tower is scanned continuously M times according to the same location or path of the collected reference point cloud data, forming M sets of detection point cloud data of the wind turbine tower.

[0082] In steps S1-S2, the detection device can be a lidar or ultrasonic sensor. During the scanning process, scanning can be performed on a specific detection area or a global scan can be established directly. It is important to note that regardless of whether the evaluation is local or global, the areas covered by the reference point cloud data and the detection point cloud data are consistent. This consistency can be ensured by controlling the UAV's path and point locations. Preliminary point cloud denoising processing can be performed after data acquisition. The M consecutive scans in step S2 are performed periodically. For example, when scanning a local area, each point is scanned every 5-20 seconds. When establishing global point cloud data, a complete scan can be completed before restarting the scan, repeating this process M times.

[0083] Step S3: Perform the first deformation analysis based on the benchmark point cloud data and the M groups of detection point cloud data to obtain the first deformation type of the wind turbine tower. The first deformation type is no deformation, continuous deformation, or sudden deformation.

[0084] In this embodiment, the method for the first deformation analysis is as follows:

[0085] Using the baseline point cloud data as a reference, deformation analysis is performed on each group of the detected point cloud data to obtain the degree of deformation of each group of the detected point cloud data.

[0086] The second deformation type of the wind turbine tower is obtained by comprehensively evaluating the degree of deformation of the detection point cloud data of each group.

[0087] Specifically, the method for performing deformation analysis on each group of the detected point cloud data is as follows: For the m-th group of the detected point cloud data, the following operations are performed. :

[0088] Acquire the detection point cloud data The reference point cloud data is obtained by analyzing key points and the first vectors to each key point during drone scanning. There are several key points and a second vector to each key point during UAV scanning. Here, the key points of both types of point cloud data are consistent; that is, the nth key point of the detected point cloud data and the nth key point of the reference point cloud data represent the same location on the tower. Several boundary points can be selected as key points. For example, four key points can be set: the first key point is the bottom leftmost and frontmost point of the detected / reference point cloud data; the second key point is the bottom rightmost and rearmost point of the detected / reference point cloud data; the third key point is the top leftmost and rearmost point of the detected / reference point cloud data; and the fourth key point is the top rightmost and frontmost point of the detected / reference point cloud data. All key points can be directly determined and located using three-dimensional global coordinates.

[0089] After obtaining the detection points and their corresponding vectors, the nth key point of the detection point cloud data is first aligned and overlapped with the nth key point of the reference point cloud data. ,get Alignment point cloud map;

[0090] Based on the aligned point cloud map, calculate the probability of suspected deformation of the wind turbine tower under the m-th group of detected point cloud data. ;

[0091] Calculate all The mean and variance;

[0092] If the average value is greater than a preset first average value threshold and the variance is less than a preset first variance threshold, then the first deformation type is determined to be a continuous deformation.

[0093] If the average value is greater than the first average value threshold and the variance is greater than the first variance threshold, then the first deformation type is determined to be a sudden deformation.

[0094] In all other cases, the first deformation type is determined to be no deformation.

[0095] Based on this, the probability of suspected deformation of the wind turbine tower under the detection point cloud data of the m-th group is calculated. The method is as follows:

[0096] The probability of suspected deformation of the wind turbine tower under the detection point cloud data of the m-th group is calculated. The method is as follows:

[0097] Feature extraction is performed based on the aligned point cloud map, the first vector, and the second vector of the m-th group of detected point cloud data to obtain the feature values ​​of the m-th group of detected point cloud data. :

[0098] ;

[0099] ;

[0100] in, Let N be the global coordinates of the j-th point in the n-th aligned point cloud map within the m-th group of the detected point cloud data, where N is the number of aligned point cloud maps and J is the total number of points in the detected point cloud data. The distance in the reference point cloud data of the nth aligned point cloud map The global coordinates of the nearest point The threshold is dynamic, and k and b are the training weights and biases obtained by fitting or training based on the training samples, respectively. and Let be the first vector of the detected point cloud data in the nth aligned point cloud map and the second vector of the reference point cloud data in the nth aligned point cloud map, respectively, and let if be the truth function. Determine the modulus length;

[0101] Based on the eigenvalues Calculate the probability of suspected deformation :

[0102] .

[0103] Based on the scheme in step S3 above, the two point cloud maps are aligned multiple times, and the results are comprehensively evaluated using multiple aligned point cloud maps. As seen in the preceding steps, the key points used for alignment are generally boundary points, because selecting boundary points makes it easier to ensure they are in the same position in different point cloud data. However, boundaries are more susceptible to interference during scanning, making distortion errors more likely. For example, a key point may shift its position due to temporary occlusion or noise. Multiple alignments can balance this error. Therefore, this embodiment chooses to align key points at different positions in the two point cloud maps separately, and then evaluate the number of points with excessive offsets for each alignment. Each alignment makes it easier to focus on the key point and the matching of points near the key point. The smaller the measurement error, the higher the overall consistency of the matching points obtained from multiple alignments.

[0104] In other words, each alignment operation involves geometrically registering a key point region in the detected point cloud data with the corresponding region in the reference point cloud data, fusing these data, improving fault tolerance, and reducing errors caused by individual point cloud offsets due to attitude deviations, occlusions, sensing errors, etc.

[0105] Calculate the probability of suspected deformation of the wind turbine tower based on the detected point cloud data. At that time, the determination was made by integrating M sets of detection point cloud data and the number of excessively offset points in all the uniform point cloud maps of each set of detection point cloud data. The number of excessively offset points was determined by... The expression was calculated, and a dynamic threshold was used. The values ​​of k and b can be obtained through training with data. This dynamic threshold takes into account the distortion caused by the scanning angle deviation, and the threshold is increased appropriately based on this. The angle between the first and second vectors is denoted by . A larger value indicates a larger scanning angle deviation, which in turn leads to a greater distortion deviation. The larger the value, the higher the dynamic threshold; therefore, the threshold is adjusted based on the magnitude of this angle. The feature values ​​obtained from the above calculations... The Sigmoid function is then used to convert the probability into a possible deformity.

[0106] After obtaining the suspected deformation probability of each group of detection point cloud data, classification is performed using the mean and variance. Specifically, if significant deformation is observed at multiple time points and the changes are relatively stable, it is judged as continuous deformation; if deformation is abnormally severe at certain times but the overall fluctuation is large, it is judged as sudden deformation; if there is no significant deviation from the reference point cloud in all scans, it is directly judged as without deformation. Sudden deformation requires further analysis and processing in step S4 to determine whether it is caused by deformation of the structure itself or by an abnormal environment.

[0107] Step S4: If the first deformation type is sudden deformation, acquire the current climate data and perform a second deformation analysis to correct the results of the first deformation analysis and obtain the second deformation type of the wind turbine tower. The second deformation type is no deformation or deformation exists.

[0108] If the first deformation type or the second deformation type is no deformation, the output detection result is no deformation.

[0109] If the first deformation type is continuous deformation or the second deformation type is existing deformation, then based on the reference point cloud data and M sets of detected point cloud data, a third deformation analysis is performed, and the deformation type and the corresponding deformation location are output.

[0110] As a preferred embodiment, the method for performing the second deformation analysis is as follows:

[0111] Obtain the current real-time wind speed. If the real-time wind speed is less than the wind speed threshold, then determine that the second deformation type is deformation.

[0112] If the real-time wind speed is not less than the wind speed threshold, the deformation change is analyzed based on the reference point cloud data and multiple sets of detection point cloud data to determine whether the second deformation type is deformation present or deformation absent.

[0113] Based on this, the method for analyzing deformation changes based on benchmark point cloud data and multiple sets of detected point cloud data is as follows:

[0114] The highest point along the z-axis in the reference point cloud data is obtained as the first analysis point, and its global coordinates are obtained. ;

[0115] The highest point along the z-axis in the m-th group of detected point cloud data is obtained as the second analysis point, and its global coordinates are acquired. , ;

[0116] Obtain the distance from the first analysis point to the m-th second analysis point respectively. :

[0117] ;

[0118] in, Determine the modulus length;

[0119] Get all The minimum value and variance;

[0120] If the minimum value is less than a preset minimum value threshold and the variance is greater than a preset second variance threshold, then the second deformation type is determined to be no deformation.

[0121] In other cases, the second deformation type is determined to be deformation.

[0122] Based on the above scheme, when the first deformation analysis result is sudden deformation, the current wind speed is introduced as an external interference factor for judgment. This can effectively identify false deformation results caused by sudden wind disturbances and other environmental changes, reducing false alarms. By setting a wind speed threshold, when the wind speed is below the threshold, it is directly judged as having real structural deformation, which is suitable for anomaly identification under normal wind turbine operating conditions. When the wind speed is higher, deformation change analysis is then performed, making the detection more adaptable to the real environment and eliminating analysis errors caused by abnormal environments.

[0123] Occasional deformations appearing in data analysis may be caused by environmental instability. This embodiment filters out such cases through further analysis. When performing deformation change analysis, the coordinates of the highest point in the baseline point cloud data and the detection point cloud data are selected as the analysis object because higher areas are more sensitive to wind, and their offsets are highly representative. Then, parameters representing the offset of this analysis object can be used. A change analysis is performed. Wind-induced sporadic deformation generally follows the fluctuations and trends of the wind, and usually has a point in time where it returns to normal. Therefore, if sporadic, recoverable deviations are detected, the minimum value will be less than the preset minimum threshold, and the variance will be greater than the preset second variance threshold. In such cases, it will be directly judged as wind-induced sporadic deformation rather than deformation of the structure itself. Other cases are judged as deformation of the structure itself. This judgment method has a simple calculation logic, based only on the coordinate extraction and distance calculation of the highest point, and has the advantage of low computational load.

[0124] As a preferred embodiment, the method for performing the third deformation analysis is as follows:

[0125] First, obtain the average value of the offset vector, the average value of the offset distance, and the average value of the global coordinates of each point in the M sets of detection point cloud data. The starting point of the offset vector of a point in the detection point cloud data is the point closest to that point in the reference point cloud data, and the ending point of the offset vector is that point in the detection point cloud data.

[0126] The average value of the offset vector reflects the directional trend of the offset at that point, while the average value of the offset distance reflects the intensity of the deformation at that point. The average value of the corresponding global coordinates is used to locate the point, which facilitates subsequent spatial clustering.

[0127] Next, offset points are filtered based on their degree of offset and grouped according to global coordinates. Each group of offset points represents a deformation location. In this step, the filtering is primarily based on the average offset distance, selecting points greater than a preset offset threshold to form a group of points exhibiting offset. These points are considered to have significant deformation characteristics. Then, clustering and grouping are performed based on global coordinates. The preferred method for filtering offset points based on their degree of offset and grouping them according to global coordinates is as follows:

[0128] Step S401: Obtain points whose average offset value is greater than a preset offset threshold, and form a global offset point group;

[0129] Step S402: Initialize group number v=1;

[0130] Step S403: Randomly select a reference point from the global offset point group, and based on the average value of the global coordinates, find all nearby points in the global offset point group whose distance from the reference point is less than a preset distance threshold.

[0131] Step S404: If the number of nearby points is less than the preset first number threshold, the reference point is deleted from the global offset point group and the process returns to step S403. This step is to discard the points because the number of points in a cluster is too small to form a clustered region, and it is directly judged as an error caused by detection, etc.

[0132] If the number of nearby points is not less than the preset first number threshold, then the reference point and nearby points are put into the vth group of offset points and deleted from the global offset point group, and proceed to the next step S405.

[0133] Step S405: Update the group number v=v+1, and then return to step S403 until the number of remaining points in the global offset point group is less than the preset second quantity threshold.

[0134] Through the above grouping, each group of offset points represents a suspected deformation area, which means that the location where deformation occurs has been determined.

[0135] Finally, deformation type analysis is performed on each group of offset points based on the offset vector. A preferred method is to perform the following operation on the r-th group of offset points, where r = 1, 2, ..., R, and R is the total number of groups of offset points obtained:

[0136] Take any offset point from the r-th set of offset points as the reference offset point, and use the average of its offset vectors as the reference offset vector. ;

[0137] Obtain the angle between the average of the offset vectors of all other offset points in the r-th offset point group and the average of the reference offset vector:

[0138] ;

[0139] in, The average value of the offset vector of the t-th offset point in the r-th offset point group. to the reference offset vector The angle between the reference points, where the t-th offset point is not the reference offset point. Determine the modulus length;

[0140] Get all If the variance is greater than a preset variance threshold, the deformation type is determined to be expansion deformation or fracture deformation; if the variance is not greater than the preset variance threshold, the deformation type is determined to be tilt deformation.

[0141] Since the average value of the offset vector can reflect the directional trend of the point offset, based on the characteristics of various deformations, if the variance of the included angle is greater than the preset variance threshold, it indicates that the offset direction in this area is relatively dispersed, and it is judged as expansion deformation or fracture deformation; if the variance of the included angle is not greater than the threshold, it indicates that the offset direction is consistent, and it is judged as tilt deformation.

[0142] The above solution enables multi-dimensional identification and precise spatial positioning of wind turbine tower deformation, making it particularly suitable for complex scenarios involving tilting or structural fractures occurring at different inspection intervals. Individual point cloud comparison analysis ensures that every deformation point can be identified.

[0143] Example 2

[0144] This embodiment provides a UAV-based wind turbine tower deformation detection system, applied to the aforementioned UAV-based wind turbine tower deformation detection method. (See reference...) Figure 2 ,include:

[0145] The baseline point cloud data acquisition module is used to scan the wind turbine tower in a standard state using a drone equipped with a detection device to form baseline point cloud data of the wind turbine tower. The standard state is when the wind speed and temperature are both within the corresponding threshold range.

[0146] The detection point cloud data acquisition module is used to scan the wind turbine tower continuously M times using a detection device mounted on a drone, forming M sets of detection point cloud data of the wind turbine tower;

[0147] The first deformation analysis module is used to perform the first deformation analysis based on the benchmark point cloud data and M sets of detection point cloud data to obtain the first deformation type of the wind turbine tower. The first deformation type is no deformation, continuous deformation, or sudden deformation.

[0148] The second deformation analysis module is used to acquire current climate data and perform a second deformation analysis when the first deformation type is sudden deformation, thereby correcting the results of the first deformation analysis and obtaining the second deformation type of the wind turbine tower. The second deformation type is either no deformation or deformation exists.

[0149] The deformation-free output module is used to output the detection result as deformation-free when the first deformation type or the second deformation type is deformation-free.

[0150] The third deformation analysis module is used to perform a third deformation analysis based on the reference point cloud data and M sets of detection point cloud data when the first deformation type is continuous deformation or the second deformation type is deformation. It outputs the deformation type and the corresponding deformation location.

[0151] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs), characterized in that, Includes the following steps: By using a drone equipped with a detection device to scan the wind turbine tower in a standard state, a baseline point cloud data of the wind turbine tower is generated. The standard state is when the wind speed and temperature are both within the corresponding threshold range. By using a drone equipped with a detection device to scan the wind turbine tower M times consecutively along the same location or path where the baseline point cloud data is collected, M sets of detection point cloud data of the wind turbine tower are generated. Based on the baseline point cloud data and the M group of detection point cloud data, the first deformation analysis is performed to obtain the first deformation type of the wind turbine tower. The first deformation type is no deformation, continuous deformation, or sudden deformation. If the first deformation type is sudden deformation, the current climate data is acquired and a second deformation analysis is performed to correct the results of the first deformation analysis and obtain the second deformation type of the wind turbine tower. The second deformation type is no deformation or deformation exists. If the first deformation type or the second deformation type is no deformation, the output detection result is no deformation. If the first deformation type is continuous deformation or the second deformation type is existing deformation, then based on the reference point cloud data and M sets of detected point cloud data, a third deformation analysis is performed, and the deformation type and the corresponding deformation location are output. The method for the first deformation analysis is as follows: Using the baseline point cloud data as a reference, deformation analysis is performed on each group of the detected point cloud data to obtain the degree of deformation of each group of the detected point cloud data. Based on a comprehensive evaluation of the deformation degree of each group of detected point cloud data, the second deformation type of the wind turbine tower is obtained. The method for performing deformation analysis on each group of the detected point cloud data is as follows: For the m-th group of the detected point cloud data, perform the following operations. : Obtain N aligned point cloud maps; Based on the aligned point cloud map, calculate the probability of suspected deformation of the wind turbine tower under the m-th group of detected point cloud data. ; Calculate all The mean and variance; If the average value is greater than a preset first average value threshold and the variance is less than a preset first variance threshold, then the first deformation type is determined to be a continuous deformation. If the average value is greater than the first average value threshold and the variance is greater than the first variance threshold, then the first deformation type is determined to be a sudden deformation. In all other cases, the first deformation type is determined to be no deformation; The method for performing the second deformation analysis is as follows: Obtain the current real-time wind speed. If the real-time wind speed is less than the wind speed threshold, then determine that the second deformation type is deformation. If the real-time wind speed is not less than the wind speed threshold, the deformation change is analyzed based on the benchmark point cloud data and multiple sets of detection point cloud data to determine whether the second deformation type is deformation present or deformation absent. The method for performing the third deformation analysis is as follows: The average value of the offset vector, the average value of the offset distance, and the average value of the global coordinates of each point in the M sets of detection point cloud data are obtained. The starting point of the offset vector of a point in the detection point cloud data is the point closest to that point in the reference point cloud data, and the ending point of the offset vector is that point in the detection point cloud data. The offset points are filtered according to their degree of offset and grouped according to global coordinates. They are then located according to global coordinates, with each group of offset points representing a deformation position. Deformation type analysis is performed on each set of offset points based on the offset vector.

2. The method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The method for obtaining N aligned point cloud maps is as follows: Acquire the detection point cloud data The reference point cloud data is obtained by analyzing key points and the first vectors to each key point during drone scanning. The key points and the second vector to each key point during drone scanning; Align and overlap the nth key point of the detected point cloud data with the nth key point of the reference point cloud data. ,get Alignment point cloud map.

3. The method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The probability of suspected deformation of the wind turbine tower under the detection point cloud data of the m-th group is calculated. The method is as follows: Feature extraction is performed based on the aligned point cloud map, the first vector, and the second vector of the m-th group of detected point cloud data to obtain the feature values ​​of the m-th group of detected point cloud data. : ； ； in, Let N be the global coordinates of the j-th point in the n-th aligned point cloud map within the m-th group of the detected point cloud data, where N is the number of aligned point cloud maps and J is the total number of points in the detected point cloud data. The distance in the reference point cloud data of the nth aligned point cloud map The global coordinates of the nearest point The threshold is dynamic, and k and b are the training weights and biases obtained by fitting or training based on the training samples, respectively. and Let be the first vector of the detected point cloud data in the nth aligned point cloud map and the second vector of the reference point cloud data in the nth aligned point cloud map, respectively, and let if be the truth function. Determine the modulus length; Based on the eigenvalues Calculate the probability of suspected deformation : 。 4. The method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The method for analyzing deformation changes based on reference point cloud data and multiple sets of detected point cloud data is as follows: The highest point along the z-axis in the reference point cloud data is obtained as the first analysis point, and its global coordinates are obtained. ; The highest point along the z-axis in the m-th group of detected point cloud data is obtained as the second analysis point, and its global coordinates are acquired. , ; Obtain the distance from the first analysis point to the m-th second analysis point respectively. : ； in, Determine the modulus length; Get all The minimum value and variance; If the minimum value is less than a preset minimum value threshold and the variance is greater than a preset second variance threshold, then the second deformation type is determined to be no deformation. In other cases, the second deformation type is determined to be deformation.

5. The method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The method for filtering offset points based on their degree of offset and grouping them according to global coordinates is as follows: Step S401: Obtain points whose average offset value is greater than a preset offset threshold, and form a global offset point group; Step S402: Initialize group number v=1; Step S403: Randomly select a reference point from the global offset point group, and based on the average value of the global coordinates, find all nearby points in the global offset point group whose distance from the reference point is less than a preset distance threshold. Step S404: If the number of nearby points is less than a preset first number threshold, then the reference point is removed from the global offset point group and the process returns to step S403; If the number of nearby points is not less than the preset first number threshold, then the reference point and nearby points are put into the vth group of offset points and deleted from the global offset point group, and proceed to the next step S405. Step S405: Update the group number v=v+1, and then return to step S403 until the number of remaining points in the global offset point group is less than the preset second quantity threshold.

6. The method for detecting wind turbine tower deformation based on unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The method for performing deformation type analysis on each group of offset points based on the offset vector is as follows: For the r-th group of offset points, perform the following operation, where r = 1, 2, ..., R, and R is the total number of groups of offset points obtained by grouping: Take any offset point from the r-th set of offset points as the reference offset point, and use the average of its offset vectors as the reference offset vector. ; Obtain the angle between the average of the offset vectors of all other offset points in the r-th offset point group and the average of the reference offset vector: ； in, The average value of the offset vector of the t-th offset point in the r-th offset point group. to the reference offset vector The angle between the reference points, where the t-th offset point is not the reference offset point. Determine the modulus length; Get all If the variance is greater than a preset variance threshold, the deformation type is determined to be expansion deformation or fracture deformation; if the variance is not greater than the preset variance threshold, the deformation type is determined to be tilt deformation.

7. A UAV-based wind turbine tower deformation detection system, applied to the UAV-based wind turbine tower deformation detection method according to any one of claims 1-6, characterized in that, include: The baseline point cloud data acquisition module is used to scan the wind turbine tower in a standard state using a drone equipped with a detection device to form baseline point cloud data of the wind turbine tower. The standard state is when the wind speed and temperature are both within the corresponding threshold range. The detection point cloud data acquisition module is used to scan the wind turbine tower continuously M times using a detection device mounted on a drone, forming M sets of detection point cloud data of the wind turbine tower; The first deformation analysis module is used to perform the first deformation analysis based on the benchmark point cloud data and M sets of detection point cloud data to obtain the first deformation type of the wind turbine tower. The first deformation type is no deformation, continuous deformation, or sudden deformation. The second deformation analysis module is used to acquire current climate data and perform a second deformation analysis when the first deformation type is sudden deformation, thereby correcting the results of the first deformation analysis and obtaining the second deformation type of the wind turbine tower. The second deformation type is either no deformation or deformation exists. The deformation-free output module is used to output the detection result as deformation-free when the first deformation type or the second deformation type is deformation-free. The third deformation analysis module is used to perform a third deformation analysis based on the reference point cloud data and M sets of detection point cloud data when the first deformation type is continuous deformation or the second deformation type is deformation. It outputs the deformation type and the corresponding deformation location.

Images