Digital measurement method for insulator creepage distance

Abstract

The invention discloses a digital measurement method for the creepage distance of an insulator, which solves the defects in the prior art, and comprises the following steps: 1, obtaining the three-dimensional information of the appearance of the insulator, and generating a three-dimensional grid pattern according to the three-dimensional information; 2, performing foreground and background classification on the three-dimensional grid graph by adopting a neural network model to obtain an insulator foreground grid graph; 3, calculating a symmetry axis of the foreground grid pattern of the insulator, and recognizing the shape of the insulator according to the distance between the boundary point of the foreground grid pattern of the insulator and the symmetry axis; 4, judging starting and ending points of the creepage distance of the insulator by adopting a neural network; and 5, calculating the creepage distance according to the starting point and the ending point of the creepage distance.

Description

Technical Field

[0001] The invention relates to the technical field of insulator measurement, in particular to a digital measurement method for an insulator creepage distance. Background Art

[0002] Insulators are important supporting devices and insulation protection devices for power systems. The creepage distance of insulators refers to the shortest path between two conductive parts or between conductive parts and the protective interface of equipment measured along the insulation surface. The creepage distance of insulators is an important indicator of their quality and plays a vital role in the normal operation, safety and stability of power systems.

[0003] At present, the measurement of creepage distance of insulators mainly relies on manual measurement using a leather tape measure with a scale along the surface of the insulator, or by sticking paper tape on the surface of the insulator and then tearing it off to measure the distance. However, due to the complex shape of the insulator surface, the manual method cannot fit the insulator surface well and cannot ensure that the measurement or sticking direction is parallel to the axis, which will cause large errors in the measurement results. At the same time, this contact measurement has high labor intensity and poses safety hazards. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects in the prior art that the creepage distance of insulators mainly relies on manual measurement, which has high labor intensity, will produce large errors and pose safety hazards, and to provide a digital measurement method for the creepage distance of insulators.

[0005] The purpose of the present invention is achieved through the following technical solutions: A digital measurement method for creepage distance of an insulator comprises the following steps: Step 1, obtaining three-dimensional information of the insulator shape, and generating a three-dimensional grid graphic according to the three-dimensional information; Step 2, using a neural network model to classify the foreground and background of the three-dimensional grid pattern to obtain an insulator foreground grid pattern; Step 3, calculating the symmetry axis of the insulator foreground grid pattern, and identifying the insulator shape according to the distance between the boundary point of the insulator foreground grid pattern and the symmetry axis; Step 4, using a neural network to determine the start and end points of the insulator creepage distance; Step 5: Calculate the creepage distance based on the starting and ending points of the creepage distance.

[0006] Preferably, the neural network model in step 2 is an MLP neural network model, and the construction process of the MLP neural network model includes establishing a foreground point feature vector data set, specifically: The average curvature and Gaussian curvature are calculated in the geodesic field of each vertex of the three-dimensional grid graphics, the geodesic field is based on the grid edge size, the radius of the geodesic field is 10 units, and the edge of the field is irregular; all the foreground points and background points of the three-dimensional grid graphics are used as marking points, the average curvature and Gaussian curvature of each vertex and the average curvature of other adjacent points around the vertex constitute an m-dimensional vector as the feature vector of the store, the feature vector of the background point is marked as 0, and the feature vector of the foreground point is marked as 1.

[0007] Preferably, the construction process of the MLP neural network model also includes construction and training testing, specifically: the MLP neural network model includes an input layer, two hidden layers and an output layer, the feature vector of each annotation point is used as the input feature value of the input layer, the first hidden layer contains i neurons and uses ReLU as an activation function, the second hidden layer contains j neurons and uses ReLU as an activation function, the output layer uses Sigmoid as an activation function, and CrossEntropy is used as a loss function; The foreground point feature vector dataset is divided into a training set and a test set. After setting the training parameters in the MLP neural network model, training is performed based on the training set until the model loss no longer decreases, and then the trained model is tested based on the test set.

[0008] Preferably, the step 3 is specifically as follows: the symmetry axis of the insulator foreground grid pattern is calculated using the arithmetic mean method, and the boundary points are the six boundary points of the insulator foreground obtained in the directions parallel to and orthogonal to the symmetry axis, namely, up, down, left, right, front and back; the distance parameters between the symmetry axis and the six points are calculated, and the distance parameters are matched with the parameters corresponding to the predicted insulator modeling data according to the principle of closest distance.

[0009] Preferably, the step 4 is specifically as follows: Establish a creepage start and end point feature vector dataset. According to the modeling category of the insulator, extract the creepage start and end points and some non-creeping start and end points in the vertices of the insulator foreground grid figure obtained in step 2 as annotation points. Calculate the average curvature and Gaussian curvature of each annotation point and the vertices around the point to form an n-dimensional vector as the feature vector of each annotation point. The labels corresponding to the feature vectors of the creepage distance starting point, creepage distance end point and non-creeping distance starting and ending points are (1,0,0), (0,1,0), (0,0,0), respectively, and divide them into training set and test set. The MLP neural network is used to build a creepage distance start and end point classifier, and the classifier is trained and tested based on the creepage start and end point feature vector dataset.

[0010] Preferably, if the insulator shape categories are different, different MLP neural networks need to be trained as creepage distance start and end point classifiers matching the insulators.

[0011] Preferably, the step 5 is specifically as follows: Step 5a, matching creepage start and end points, using the KNN nearest neighbor algorithm based on DTW distance to extract matching creepage start and end points, taking feature vectors of different dimensions as time series of different lengths, for each creepage distance starting point, taking the first K creepage distance end points with the smallest DTW distance between its feature vector, if the distance calculated between any creepage distance starting point and any corresponding creepage distance end point meets the creepage distance standard of the insulator of this shape category and the line connecting the creepage distance start and end points is coplanar with the symmetry axis of the insulator, then the creepage distance start and end points are considered to be matched successfully; Step 5b, calculate the creepage distance, and use the Dijkstra algorithm to calculate the geodesic distance between the creepage start and end points. The geodesic distance is the creepage distance.

[0012] The beneficial effects of the present invention are as follows: the method of the present invention targets the influence of environmental noise in the three-dimensional digital graphics of the insulator, constructs an MLP neural network characterized by average curvature and Gaussian curvature, identifies foreground points and background points, and effectively reduces the interference of background points. At the same time, an MLP neural network characterized by average curvature and Gaussian curvature is constructed to improve the accuracy of judging the start and end points of the insulator creepage distance. Finally, the K nearest neighbor method and DTW distance are used to achieve the matching of the start and end points of the insulator creepage distance, and the Dijkstra algorithm is used to calculate the distance between the start and end points of the creepage distance, which ultimately improves the accuracy of the calculation of the insulator creepage distance. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 is a flow chart of the present invention; Figure 2 is a structural diagram of the insulator front-back classifier of the present invention; Figure 3 Schematic diagram of vertices associated with a certain characteristic vector in the three-dimensional grid graph of the insulator. DETAILED DESCRIPTION

[0014] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be more comprehensive and complete and fully convey the concept of the example embodiments to those skilled in the art.

[0015] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0016] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0017] Example: A digital measurement method for creepage distance of insulators, such as Figure 1 As shown, the following steps are included: Step 1, obtaining three-dimensional information of the insulator shape, and generating a three-dimensional grid graphic according to the three-dimensional information; Step 2, using a neural network model to classify the foreground and background of the three-dimensional grid pattern to obtain an insulator foreground grid pattern; Step 3, calculating the symmetry axis of the insulator foreground grid pattern, and identifying the insulator shape according to the distance between the boundary point of the insulator foreground grid pattern and the symmetry axis; Step 4, using a neural network to determine the start and end points of the insulator creepage distance; Step 5: Calculate the creepage distance based on the starting and ending points of the creepage distance.

[0018] The neural network model in step 2 is an MLP neural network model. The construction process of the MLP neural network model includes establishing a foreground feature vector data set such as Figure 2 As shown, specifically: The mean curvature and Gaussian curvature are calculated in the geodesic field of each vertex of the three-dimensional grid figure. The geodesic field is based on the grid edge size, the radius of the geodesic field is 10 units, and the edge of the field is irregular. All the foreground points and background points of the three-dimensional grid figure are used as annotation points. The mean curvature and Gaussian curvature of each vertex and the mean curvature of other adjacent points around the vertex constitute an m-dimensional vector as the feature vector of the store, such as Figure 3 As shown, the feature vector of the background point is marked as 0, and the feature vector of the foreground point is marked as 1.

[0019] Specifically, in this embodiment, for the calculation of the mean curvature and Gaussian curvature, it is necessary to first calculate the principal curvature of the marked point. For each vertex p on the three-dimensional grid, a geodesic field B around the point p is taken. The geodesic field B is based on the grid edge, and its neighborhood radius is about 10 units. The edge of the field is irregular, and only the points and edges within the field are considered. The following matrix is ​​defined: Among them, |B| represents the area of ​​the geodesic field B, e represents the edge of the grid in the geodesic field B, represents the unit vector in the direction of e, ||e∩B|| represents the length of e∩B, which ranges from 0 to |e|, and β(e) represents the angle between the normals of two normal triangles with e as the common side. Considering the eigenvalues ​​and eigendirections of the above matrix, E p The minimum eigenvalue and eigendirection of (B) can be used as the estimated formula for the normal vector of the mesh surface at vertex p, and the minimum eigenvalue and maximum eigenvalue can be used as the estimated formula for the principal curvatures k1 and k2. Therefore, the average curvature and Gaussian curvature at point p can be obtained by the following formula: k G =k1k2 Among them, k G represents Gaussian curvature, k H represents the mean curvature.

[0020] The construction process of the MLP neural network model also includes construction and training testing, specifically: the MLP neural network model includes an input layer, two hidden layers and an output layer, the feature vector of each marked point is used as the input feature value of the input layer, the first hidden layer contains i neurons and uses ReLU as the activation function, the second hidden layer contains j neurons and uses ReLU as the activation function, the output layer uses Sigmoid as the activation function, and Cross Entropy is used as the loss function; The foreground point feature vector dataset is divided into a training set and a test set in a ratio of 7:3. After setting the training parameters in the MLP neural network model, training is performed based on the training set until the model loss no longer decreases, and then the trained model is tested based on the test set.

[0021] The step 3 is specifically as follows: the symmetry axis of the insulator foreground grid pattern is calculated using the arithmetic mean method, and the boundary points are the six boundary points of the insulator foreground obtained in the parallel and orthogonal directions of the symmetry axis, namely, up, down, left, right, front and back; the distance parameters between the symmetry axis and the six points are calculated, and the distance parameters are matched with the parameters corresponding to the predicted insulator modeling data according to the principle of the closest distance.

[0022] Since insulators have various shapes and materials according to their different use environments, their creepage distances and creepage start and end points are different. The subsequent creepage start and end points need to be analyzed according to the insulator shape. Therefore, it is necessary to first identify the shape of the insulator. In this embodiment, the following steps are specifically included: (1) Calculate the symmetry axis of the insulator foreground, divide the insulator foreground into t height levels along the height direction, and then use the arithmetic mean method to calculate the coordinate value of the center point of each level. The formula is as follows. The symmetry axis of the insulator foreground is obtained through the coordinates of these t center points.

[0023] Among them, (X i , Y i , Z i ) represents the coordinates of the center point of the i-th level, x ij ,y ij ,z ij Represents the coordinates of the jth point at the i-th level, n i is the number of vertices at the i-th level.

[0024] (2) Obtain the upper and lower, left and right, front and back boundary points of the insulator foreground in the parallel direction and the orthogonal direction (two) of the symmetry axis, and calculate the six distance parameters between the symmetry axis and the boundary points.

[0025] (3) Classifying the insulator shapes according to the six distance parameters. The specific method is to use the six distance parameters to match the pre-stored insulator shape data of various types according to the principle of closest distance to obtain the shape category of the current insulator.

[0026] The step 4 is specifically as follows: Establish a creepage start and end point feature vector dataset. According to the modeling category of the insulator, extract the creepage start and end points and some non-creeping start and end points in the vertices of the insulator foreground grid figure obtained in step 2 as annotation points. Calculate the average curvature and Gaussian curvature of each annotation point and the vertices around the point to form an n-dimensional vector as the feature vector of each annotation point. The labels corresponding to the feature vectors of the creepage distance starting point, creepage distance end point and non-creeping distance starting and ending points are (1,0,0), (0,1,0), (0,0,0), respectively, and divide the training set and the test set into a ratio of 7:3. The MLP neural network is used to build a creepage distance start and end point classifier, and the classifier is trained and tested based on the creepage start and end point feature vector dataset.

[0027] If the insulator shape categories are different, different MLP neural networks need to be trained as creepage distance start and end point classifiers that match the insulator.

[0028] The step 5 is specifically as follows: Step 5a, matching creepage start and end points, using the KNN nearest neighbor algorithm based on DTW distance to extract matching creepage start and end points, taking feature vectors of different dimensions as time series of different lengths, for each creepage distance starting point, taking the first K creepage distance end points with the smallest DTW distance between its feature vector, if the distance calculated between any creepage distance starting point and any corresponding creepage distance end point meets the creepage distance standard of the insulator of this shape category and the line connecting the creepage distance start and end points is coplanar with the symmetry axis of the insulator, then the creepage distance start and end points are considered to be matched successfully; Specifically, in this embodiment, based on the insulator creepage start and end points obtained in step 4, the Gaussian curvature and spatial coordinates of each creepage start and end point are extracted to form the feature vector of each point, and the creepage distance starting point set A = {X1,...,X i ,...,X m} and the creepage distance end point set B = {Y1,...,Y j ,...,Y n}, X i The characteristic vector representing the starting point of the i-th creepage distance, Y j The eigenvector representing the jth creepage distance end point.

[0029] The KNN nearest neighbor algorithm based on DTW distance is used to extract the matching creepage start and end points. DTW is a method for measuring the similarity between two time series. It is based on the idea of ​​dynamic programming and can overcome the problem of inconsistent length in time series. The feature vectors of different dimensions are regarded as time series of different lengths. For each creepage distance starting point, the first K creepage distance end points with the smallest DTW distance between it and its feature vector are taken. If the distance calculated between any creepage distance starting point and any corresponding creepage distance end point meets the creepage distance standard of the insulator of this shape category and the line connecting the creepage distance start and end points is coplanar with the symmetry axis of the insulator, the creepage distance start and end points are considered to be matched successfully.

[0030] Step 5b, calculate the creepage distance, and use the Dijkstra algorithm to calculate the geodesic distance between the creepage start and end points. The geodesic distance is the creepage distance.

[0031] Creepage distance is the shortest distance between the creepage start and end points along the surface, that is, the geodesic distance between the creepage start and end points. For three-dimensional discrete grids, the geodesic distance between the creepage start and end points is calculated using the Dijkstra algorithm in graph theory, which has high accuracy and strong adaptability.

[0032] First, the three-dimensional mesh of the surface is converted into an undirected weighted graph. The vertices of the graph are the vertices of the mesh, and the weight of the edge of the graph is the distance between the two points connecting the edge. The greedy mechanism is used to find the shortest path between the two points, that is, a set of creepage start and end points (S) are selected from the creepage start and end points obtained in step 5. i ,E j ), from the starting point S i Start by looking for the point p1 with the shortest distance along an edge, and then start from point p1 and look for the point p2 with the shortest distance along an edge, and continue to make greedy selections from the remaining points until the end point E is found. j , determine the shortest path between the creepage starting and ending points. The sum of the lengths of the edges at the vertices of the path is the creepage distance L p .

[0033] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary technical means in the art that are not disclosed in the present application.

[0034] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A digital measurement method for creepage distance of an insulator, characterized in that: The following steps are involved: Step 1, obtaining three-dimensional information of the insulator shape, and generating a three-dimensional grid graphic according to the three-dimensional information; Step 2, using a neural network model to classify the foreground and background of the three-dimensional grid pattern to obtain an insulator foreground grid pattern; Step 3, calculating the symmetry axis of the insulator foreground grid pattern, and identifying the insulator shape according to the distance between the boundary point of the insulator foreground grid pattern and the symmetry axis; Step 4, using a neural network to determine the start and end points of the insulator creepage distance; Step 5: Calculate the creepage distance based on the starting and ending points of the creepage distance.

2. A digital measurement method for creepage distance of insulators according to claim 1, characterized in that: The neural network model in step 2 is an MLP neural network model. The construction process of the MLP neural network model includes establishing a foreground point feature vector data set, specifically: Calculate the mean curvature and Gaussian curvature in the geodesic field of each vertex of the three-dimensional grid figure, wherein the geodesic field is based on the grid edge size, the radius of the geodesic field is 10 units, and the edge of the field is irregular; All foreground and background points of the three-dimensional grid graphics are taken as marking points. The average curvature and Gaussian curvature of each vertex and the average curvature of other adjacent points around the vertex constitute an m-dimensional vector as the feature vector of the store. The feature vector of the background point is marked as 0, and the feature vector of the foreground point is marked as 1.

3. A digital measurement method for creepage distance of insulators according to claim 2, characterized in that: The construction process of the MLP neural network model also includes construction and training testing, specifically: The MLP neural network model includes an input layer, two hidden layers and an output layer. The feature vector of each annotation point is used as the input feature value of the input layer. The first hidden layer contains i neurons and uses ReLU as the activation function. The second hidden layer contains j neurons and uses ReLU as the activation function. The output layer uses Sigmoid as the activation function and CrossEntropy as the loss function. The foreground point feature vector dataset is divided into a training set and a test set. After setting the training parameters in the MLP neural network model, training is performed based on the training set until the model loss no longer decreases, and then the trained model is tested based on the test set.

4. A digital measurement method for creepage distance of insulators according to claim 1, characterized in that: The step 3 is specifically as follows: the symmetry axis of the insulator foreground grid pattern is calculated using the arithmetic mean method, and the boundary points are the six boundary points of the insulator foreground obtained in the parallel and orthogonal directions of the symmetry axis, namely, up, down, left, right, front and back; the distance parameters between the symmetry axis and the six points are calculated, and the distance parameters are matched with the parameters corresponding to the predicted insulator modeling data according to the principle of the closest distance.

5. The digital measurement method for creepage distance of an insulator according to claim 1 is characterized in that: The step 4 is specifically as follows: Establish a creepage start and end point feature vector data set. According to the modeling category of the insulator, for the insulator foreground grid figure obtained in step 2, extract the creepage start and end points and some non-creeping start and end points in the figure vertices as annotation points, calculate the average curvature and Gaussian curvature of each annotation point and the vertices around the point, and construct an n-dimensional vector as the feature vector of each annotation point. The labels corresponding to the feature vectors of the creepage distance starting point, creepage distance end point, and non-creepage distance starting and ending points are (1,0,0), (0,1,0), and (0,0,0), respectively, and they are divided into training set and test set; The MLP neural network is used to build a creepage distance start and end point classifier, and the classifier is trained and tested based on the creepage start and end point feature vector dataset.

6. A digital measurement method for creepage distance of insulators according to claim 5, characterized in that: If the insulator shape categories are different, different MLP neural networks need to be trained as creepage distance start and end point classifiers that match the insulator.

7. A digital measurement method for creepage distance of insulators according to claim 1, characterized in that: The step 5 is specifically as follows: Step 5a, matching creepage start and end points, using the KNN nearest neighbor algorithm based on DTW distance to extract matching creepage start and end points, taking feature vectors of different dimensions as time series of different lengths, for each creepage distance starting point, taking the first K creepage distance end points with the smallest DTW distance between its feature vector, if the distance calculated between any creepage distance starting point and any corresponding creepage distance end point meets the creepage distance standard of the insulator of this shape category and the line connecting the creepage distance start and end points is coplanar with the symmetry axis of the insulator, then the creepage distance start and end points are considered to be matched successfully; Step 5b, calculate the creepage distance, and use the Dijkstra algorithm to calculate the geodesic distance between the creepage start and end points. The geodesic distance is the creepage distance.

Images