Porcelain insulator defect detection method, device, equipment and medium

Abstract

The method is mainly applied to the technical field of artificial intelligence. The invention discloses a porcelain insulator defect detection method, device, equipment and medium, and the method comprises the steps: generating a temperature feature set corresponding to a to-be-detected insulator string based on the time sequence temperature information of the to-be-detected insulator string, the temperature feature set comprising a first feature set and a second feature set; constructing a position prediction model, and training the position prediction model by using the first feature set; inputting the second feature set into the trained position prediction model, and determining the defect position of the insulator string to be detected through the trained position prediction model to obtain an output result; and generating a linear model according to the second feature set and the output result, and displaying the relationship between the input features of the position prediction model and the output result through the linear model. According to the invention, the thermal imaging technology can be utilized to detect the porcelain insulator, and the accuracy and credibility of the detection result are improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method, device, equipment and medium for detecting defects in porcelain insulators. Background Art

[0002] Porcelain insulators, as an important component of electrical insulation and mechanical support for power equipment, are widely used in transmission lines and substations. Internal and external damage to porcelain insulators will lead to a decrease in the insulation performance of porcelain insulators, which in turn will cause short circuits in power equipment. Therefore, internal and external inspection of porcelain insulators is an important means to determine the insulation performance. With the continuous development of thermal imaging technology, it has become possible to use thermal imaging technology to inspect the inside of porcelain insulators. However, the accuracy of the inspection results of porcelain insulators using thermal imaging technology is low, and the lack of visualization means leads to low credibility of the inspection results. Summary of the invention

[0003] The present invention provides a method, device, equipment and medium for detecting defects of porcelain insulators, which can detect porcelain insulators by using thermal imaging technology, thereby improving the accuracy and credibility of the detection results.

[0004] The present invention provides a method for detecting defects in a porcelain insulator, the method comprising: Based on the time-series temperature information of the insulator string to be measured, a temperature feature set corresponding to the insulator string to be measured is generated, wherein the temperature feature set includes a first feature set and a second feature set; Constructing a location prediction model, and training the location prediction model using the first feature set; Inputting the second feature set into the trained position prediction model, and determining the defect position of the insulator string to be tested by the trained position prediction model to obtain an output result; A linear model is generated based on the second feature set and the output result, and the relationship between the input features and the output result of the position prediction model is displayed through the linear model.

[0005] Further, the time series temperature information includes the current temperature value of each insulator in the insulator string to be tested and the ambient temperature value of the environment where each insulator is located; Based on the time-series temperature information of the insulator string to be measured, a temperature feature set corresponding to the insulator string to be measured is generated, wherein the temperature feature set includes a first feature set and a second feature set, including: Acquiring time-series temperature information of the insulator string to be tested; Based on the time series temperature information, determining a temperature change value of each of the insulators, wherein the temperature change value of the insulator to be measured is a difference between a current temperature value of the insulator to be measured and an ambient temperature value; Based on the temperature change value of each of the insulators, a temperature characteristic information factor of the insulator string to be tested is generated and used as the temperature characteristic of the insulator string to be tested; Return to the step of obtaining the time-series temperature information of the insulator string to be measured, and when a plurality of the temperature characteristics are obtained, generate a temperature characteristic set corresponding to the insulator string to be measured and divide it into a first characteristic set and a second characteristic set.

[0006] Furthermore, the porcelain insulator defect detection method further comprises: Dividing the temperature feature set into a plurality of data clusters, each of the data clusters including a plurality of the temperature features; Determine a first average distance and a second average distance of each temperature feature, wherein the first average distance is an average distance value between the temperature feature and temperature features of the same data cluster, and the second average distance is an average distance value between the temperature feature and temperature features of different data clusters; Determining a profile coefficient of each of the temperature features based on a functional relationship between the first average distance and the second average distance; Based on the result of comparing the silhouette coefficient of each of the temperature features with a preset threshold, a plurality of target temperature features are determined, and a filtered temperature feature set is generated according to the plurality of target temperature features.

[0007] Furthermore, the porcelain insulator defect detection method further comprises: Clustering multiple temperature features in the temperature feature set into multiple classes; determining dispersion values ​​for the plurality of classes; Determining a target class based on a result of comparing the dispersion degree value with a preset dispersion degree value; Each of the target classes is used as a target temperature feature, and a clustered temperature feature set is generated according to a plurality of the target temperature features.

[0008] Further, the constructing of the location prediction model and training the location prediction model using the first feature set includes: Generate an objective function of the position prediction model based on the prediction accuracy and model complexity of the position prediction model; Adjusting model parameters of the position prediction model according to the current prediction accuracy of the position prediction model; Updating the current model complexity and the current prediction accuracy of the position prediction model according to the adjusted model parameters; Updating the objective function according to the updated model complexity and the updated prediction accuracy; When the position prediction model is trained using the first feature set, the updated prediction accuracy is used as the current prediction accuracy, and the step of adjusting the model parameters of the position prediction model according to the current prediction accuracy of the position prediction model is returned to execute until the optimal objective function is obtained, and the optimal objective function is an objective function that meets preset conditions.

[0009] Further, the position prediction model includes at least one tree structure, each of the tree structures includes a first node and a second node, and the first node and the second node are both model parameters; The first node is used to generate a new first node or a new second node, and the second node is used to output a prediction result; The adjusting the model parameters of the position prediction model according to the current prediction accuracy of the position prediction model includes: Determine a structure score of the current tree structure based on the current prediction accuracy, the number of first nodes and second nodes of the current tree structure, and the node value of each node; Based on the structural score, a new first node or a new second node is generated on the tree structure to adjust the model parameters of the position prediction model.

[0010] Further, the method comprises: Selecting samples from the second feature set, and grouping the selected samples into a sample subset; When the selected samples are added to the sample subset, a marginal gain value of each sample in the second feature set is determined according to a preset function; Based on the marginal gain value, determining a target sample in the second feature set and using it as the selected sample; When a preset stop condition is reached, the selection of samples in the second feature set is stopped, and the sample subset is used as the optimized second feature set.

[0011] The present invention also provides a porcelain insulator defect detection device, the device comprising: An acquisition module, configured to generate a temperature feature set corresponding to the insulator string to be measured based on the time-series temperature information of the insulator string to be measured, wherein the temperature feature set includes a first feature set and a second feature set; A construction module, used to construct a location prediction model and train the location prediction model using the first feature set; A training module, used for inputting the second feature set into the trained position prediction model, and obtaining an output result after determining the defect position of the insulator string to be tested through the trained position prediction model; A display module is used to generate a linear model based on the second feature set and the output result, and to display the relationship between the input features and the output result of the position prediction model through the linear model.

[0012] The present invention also provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the porcelain insulator defect detection method as described in any one of the above items when executing the computer program.

[0013] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the porcelain insulator defect detection method as described in any one of the above items is implemented.

[0014] The present invention has at least the following beneficial effects: The technical solution of the present application realizes real-time detection of overheating defects of porcelain insulators by real-time monitoring of the time-series temperature information of the insulator string and quickly generating a temperature feature set, and predicts the defect position of the insulator string according to the rules learned from the temperature feature set through the trained position prediction model, thereby achieving real-time detection of overheating defects of porcelain insulators. At the same time, a linear model is generated according to the second feature set and the output results of the model. This linear model can show the relationship between the input features and the output results of the position prediction model, which not only helps to understand how the position prediction model makes predictions based on the input features, but also can identify which features have the greatest impact on the prediction results, thereby realizing visualization of the position prediction model and intuitively displaying the prediction logic and feature importance of the model, which is conducive to enhancing the interpretability and reliability of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solution of the present invention and do not constitute a limitation on the technical solution of the present invention.

[0016] Figure 1 It is a flow chart of the steps of the porcelain insulator defect detection method; Figure 2 is a flow chart of step S101 in the porcelain insulator defect detection method; Figure 3 is a flow chart of step S102 in the porcelain insulator defect detection method; Figure 4 It is a schematic diagram of the working principle of the data processing process when implementing the porcelain insulator defect detection method; Figure 5 It is a schematic diagram of the working principle of the model training process when implementing the porcelain insulator defect detection method; Figure 6It is a schematic diagram of the working principle of the secondary mode selection process when realizing the porcelain insulator defect detection method; Figure 7 It is a schematic diagram of a visualization result obtained by the porcelain insulator defect detection method in an actual scenario; Figure 8 It is a schematic diagram of another visualization result obtained by the porcelain insulator defect detection method in an actual scenario; Fig. 9 It is a structural schematic diagram of a porcelain insulator defect detection device; Fig.10 It is a structural diagram of an electronic device. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0018] In the relevant technical field, porcelain insulators are widely used in transmission lines and substations as an important component of electrical insulation and mechanical support for power equipment. With the extension of working time, its insulation and mechanical properties will decline to a certain extent. The failure of the insulator will cause a short circuit and cause an electrical safety accident. According to relevant research, the causes of the decline in the insulation performance of porcelain insulators are mainly divided into external damage and internal corrosion. Among them, external damage is mainly manifested as damage to the edge of the umbrella skirt and cracking of the steel foot, which can be diagnosed through aerial images or the naked eye, and the relevant technology is already very mature. On the other hand, internal corrosion is manifested as rusty steel caps and internal breakdown. Since its appearance cannot be identified, it needs to be detected by other features. Although the research on porcelain insulator detection technology is constantly developing, it is still in the laboratory stage.

[0019] With the continuous advancement of thermal imaging technology, the inspection of porcelain insulators has become possible. According to our research, when the porcelain insulator is in the early stage of overheating defects (low-value insulators), the temperature of the faulty insulator and its adjacent parts is convex or peak-shaped. At the same time, thermal imaging technology, with its high sensitivity to temperature, provides a new solution for insulator inspection.

[0020] However, the detection method based on thermal imaging technology has defects. First, the detection model lacks visualization means, resulting in weak credibility of the detection results. Second, the detection model network is complex. For tens of thousands or even hundreds of thousands of insulators to be inspected, the computing equipment requirements are extremely high and cannot achieve fast calculations.

[0021] In order to solve the above technical problems, the technical solution of the present application provides a method, device, equipment and medium for detecting defects in porcelain insulators, which can detect porcelain insulators using thermal imaging technology to improve the accuracy and credibility of the detection results. The following are embodiments of the technical solution of the present application.

[0022] Please refer to Figure 1 , Figure 1 It is a flow chart of the steps of the porcelain insulator defect detection method.

[0023] This embodiment provides a method for detecting defects in porcelain insulators, the method comprising: S101. Generate a temperature feature set corresponding to the insulator string to be tested based on time-series temperature information of the insulator string to be tested, where the temperature feature set includes a first feature set and a second feature set.

[0024] S102: construct a location prediction model, and use the first feature set to train the location prediction model.

[0025] S103, inputting the second feature set into the trained position prediction model, and obtaining an output result after determining the defect position of the insulator string to be tested through the trained position prediction model.

[0026] S104: Generate a linear model based on the second feature set and the output result, and use the linear model to demonstrate the relationship between the input features and the output result of the position prediction model.

[0027] In some embodiments, the time-series temperature information includes the current temperature value of each insulator in the insulator string to be tested and the ambient temperature value of the environment where each insulator is located.

[0028] In step S101 of some embodiments, a time series analysis method, such as Fourier transform or wavelet transform, is used to extract frequency domain features from the time series temperature information to form a temperature feature set, or a statistical analysis method is applied, such as calculating the mean, standard deviation, maximum and minimum values ​​of the temperature data to form a temperature feature set.

[0029] In step S102 of some embodiments, a traditional machine learning algorithm, such as a support vector machine (SVM) or a random forest, is used to construct a location prediction model.

[0030] In step S104 of some embodiments, linear regression analysis is used to establish a linear model based on input features and output results, and the linear relationship between the input features and the output results is analyzed. Specifically, a local interpretability model (such as LIME or SHAP) is used in combination with the local interpretability model to provide the specific contribution of each feature to the model prediction, enhance the model interpretability, and thus explain the linear relationship between the input features and the output results. Then, visualization tools such as scatter plots, heat maps, or parallel coordinate plots are used to intuitively display the relationship between the input features and the output results.

[0031] Please refer to Figure 2 , Figure 2 It is a flow chart of step S101 in the porcelain insulator defect detection method.

[0032] In some embodiments, step S101 includes: S201. Obtaining time-series temperature information of an insulator string to be tested.

[0033] S202. Determine a temperature change value of each insulator based on the time-series temperature information, wherein the temperature change value of the insulator to be measured is a difference between a current temperature value of the insulator to be measured and an ambient temperature value.

[0034] S203: Based on the temperature change value of each insulator, generate a temperature characteristic information factor of the insulator string to be measured and use it as the temperature feature of the insulator string to be measured.

[0035] S204, returning to step S201, when multiple temperature features are obtained, generating a temperature feature set corresponding to the insulator string to be tested and dividing it into a first feature set and a second feature set.

[0036] It is understandable that according to relevant research, in the micro-heat state, the temperature of the defective part in the insulator string is in a convex parabola relationship with the adjacent part. Therefore, in order to focus the model on the adjacent temperature change component, the temperature data is highlighted. Specifically, the temperature gradient coefficient (i.e., the temperature characteristic information factor) is constructed using the adjacent temperature data of the insulator, and then a normalization operation is performed.

[0037] A specific implementation of this embodiment is to calculate the temperature change value of the insulator by the following formula: (1) in, The temperature change value of the i-th insulator of the insulator string to be inspected, is the on-site temperature value of the i-th insulator of the insulator string to be inspected, is the reference temperature value (ie, the ambient temperature value) of the i-th insulator of the insulator string to be inspected.

[0038] Based on formula (1), the temperature characteristic information factor of each insulator in the insulator string to be tested is obtained according to formula (2).

[0039] (2) in, is the kth relative temperature diagnostic information, is the maximum value of the insulator string data set to be checked, It is the minimum value of the insulator string data set to be checked.

[0040] Please refer to Figure 3 , Figure 3 It is a flow chart of step S102 in the porcelain insulator defect detection method.

[0041] In some embodiments, step S102 includes: S301. Generate an objective function of the position prediction model based on the prediction accuracy and model complexity of the position prediction model.

[0042] S302: Adjust model parameters of the position prediction model according to the current prediction accuracy of the position prediction model.

[0043] S303: Update the current model complexity and current prediction accuracy of the position prediction model according to the adjusted model parameters.

[0044] S304: Update the objective function according to the updated model complexity and the updated prediction accuracy.

[0045] S305. When the position prediction model is trained using the first feature set, the updated prediction accuracy is used as the current prediction accuracy, and the process returns to step S302 until an optimal objective function is obtained. The optimal objective function is an objective function that meets preset conditions.

[0046] In this embodiment, the location prediction model can be a fusion Markov and multi-class machine learning model, which combines the Markov model with multiple machine learning algorithms, such as decision tree (DT), random forest (RF), k-nearest neighbor (kNN) and support vector machine (SVM). The Markov model predicts the probability of the next state based on the current state, while the machine learning model provides the ability to classify the location data.

[0047] In some embodiments, the location prediction model uses an optimized decision tree model. The location prediction model includes at least one tree structure, each tree structure includes a first node and a second node, the first node and the second node are both model parameters; the first node is used to generate a new first node or a new second node, and the second node is used to output a prediction result.

[0048] In some embodiments, step S302 includes: Based on the current prediction accuracy, the number of first nodes and second nodes of the current tree structure, and the node value of each node, the structural score of the current tree structure is determined; based on the structural score, a new first node or a new second node is generated on the tree structure to adjust the model parameters of the position prediction model.

[0049] See also Figure 4 In a specific embodiment, the objective function of the position prediction model can be expressed by formula (3): (3) Among them, l is the loss function of the true value of the i-th sample and the predicted value of the sample, and is the complexity of the kth tree. In the recursive process, the kth tree will be trained on the basis of the k-1th tree, so (4) is obtained.

[0050] (4) in, is the sample of the jth tree.

[0051] When training the kth tree, the sum of the model complexities of the previous k−1 trees can be regarded as a constant, and equation (5) is obtained: (5) In some embodiments, the objective function of the location prediction model is further optimized.

[0052] In a specific embodiment, the objective function is optimized by taking the second-order derivative of the Taylor technique.

[0053] Specifically, the second-order expansion of the Taylor series is shown in formula (6): (6) Based on equation (5), the first-order and second-order derivatives are as follows: (7) (8) Substituting equations (6), (7) and (8) into equation (5), we get the Taylor series conversion equation (9): (9) Next, simplify equation (9) into equation (10): (10) The decision tree super clustering algorithm uses a greedy algorithm to process the selection of the optimal tree structure. It constructs the structure score through formula (11), dynamically selects whether to establish a tree node based on the structure score, and finally outputs the optimal tree structure of the model. Formula (11) is as follows: (11) In the formula , g i The sum of the left and right subtrees of , h i The sum of the left subtree and the right subtree of , γ and λ are parameters.

[0054] The model complexity is controlled by the number of leaf nodes, node values ​​and tree depth, as shown in formula (12): (12) Among them, T is the number of leaf nodes and ω is the leaf node value.

[0055] Based on formula (10), combined with formula (12), we get (13): (13) (14) (15) Substituting (14) and (15) into equation (13), equation (13) can be simplified to equation (16): (16) The optimal leaf node value of the tree can be obtained by formula (16): (17) Combining the above formulas, we can deduce that the minimum value of the loss function is: (18) In some embodiments, a porcelain insulator defect detection method further includes: Clustering multiple temperature features in a temperature feature set into multiple classes; determining dispersion values ​​of multiple classes; determining a target class based on a result of comparing the dispersion value with a preset dispersion value; taking each target class as a target temperature feature, and generating a temperature feature set after clustering processing based on multiple target temperature features.

[0056] In some embodiments, a porcelain insulator defect detection method further includes: The temperature feature set is divided into multiple data clusters, each data cluster includes multiple temperature features; the first average distance and the second average distance of each temperature feature are determined, wherein the first average distance is the average distance value between the temperature feature and the temperature features of the same data cluster, and the second average distance is the average distance value between the temperature feature and the temperature features of different data clusters; based on the functional relationship between the first average distance and the second average distance, the contour coefficient of each temperature feature is determined; based on the result of comparing the contour coefficient of each temperature feature with a preset threshold, multiple target temperature features are determined, and a screened temperature feature set is generated based on the multiple target temperature features.

[0057] It is understandable that for tens of thousands or even hundreds of thousands of temperature characteristic information factors, it is important to have both training speed and accuracy. Therefore, based on the traditional decision tree model, this embodiment first simplifies redundant features through a super clustering algorithm, and then establishes a deteriorated insulator position prediction method based on decision tree super clustering according to its timing characteristics.

[0058] Specifically, the degree of dispersion of each class can be measured by two indicators, s(i) and CH. For a data set D of n temperature characteristic information factors (i.e., a temperature feature set), D is divided into k clusters. For each object o in D, the average distance a(o) between it and other objects in the cluster to which it belongs is calculated, and the minimum average distance b(o) between it and the cluster to which it does not belong is calculated. The silhouette coefficient is calculated for each object in the data set and the average value is taken as the quality measure of the clustering. The value range of the silhouette coefficient is [-1,1]. The closer to 1, the better the clustering quality, and the closer to -1, the worse the clustering quality.

[0059] (19) Calculate the average distance a between sample i and other samples in the same cluster i , a i The smaller it is, the more likely sample i should be clustered into this cluster. i Become the intra-cluster dissimilarity of sample i. Calculate the distance from sample i to another cluster C j The average distance b of all samples i , called sample i and cluster c j The dissimilarity of . It is defined as the inter-cluster dissimilarity of sample i: b i = min{b i1 ,b i2 ,......b ik},s i If it is close to 1, it means that the clustering of sample i is reasonable, and s i If s is close to -1, it means that sample i should be classified into another cluster. i If it is approximately 0, it means that sample i is on the boundary of two clusters.

[0060] (20) Where n is the number of clusters, k is the current class, trB(k) is the trace of the between-class dispersion matrix, and trW(k) is the trace of the within-class dispersion matrix. The larger the CH, the tighter the clusters themselves and the more dispersed the clusters are, which means better clustering results.

[0061] In some embodiments, when si>0.8 and CH>20 of a certain temperature feature, it can be considered as a redundant feature and similar feature parts can be removed.

[0062] In some embodiments, a porcelain insulator defect detection method further includes: Samples are selected from the second feature set, and the selected samples are grouped into a sample subset; when the selected samples are added to the sample subset, the marginal gain value of each sample in the second feature set is determined according to a preset function; based on the marginal gain value, the target sample is determined in the second feature set and used as the selected sample; when the preset stopping condition is reached, the sample selection in the second feature set is stopped, and the sample subset is used as the optimized second feature set.

[0063] It is understandable that the existing porcelain insulator detection usually adopts a "black box model" (that is, training through the network and directly outputting the results) to prevent the model from having small sample local fitting phenomena. At the same time, because porcelain insulators are components in the power field, based on the reliability of their detection results, the weights of their network judgments need to be visualized.

[0064] See also Figure 5 In a specific implementation of this embodiment, for any set The set of all samples (V is the second feature set) and any element ,have: (twenty one) Where F is the importance of the selected sample subset in explaining the model behavior. This part uses formula (21) to obtain representative samples to obtain the sample subset, such as Figure 5 shown.

[0065] Based on the above optimized second feature set, this embodiment generates multiple adjacent samples by randomly perturbing the original sample x to form a set of adjacent sample sets X', and obtains the predicted value y' of the adjacent samples in the DCT model (degraded insulator position prediction model). At the same time, minimize (21) to obtain the fitted linear model g: (twenty two) in, is a deteriorated insulator position prediction model; g is a local linear model, and the model coefficient is positively correlated with the deteriorated insulator position prediction model characteristics mentioned in the present invention. It is a measure The loss function for the difference in prediction between y and g; is the regularization term; It is a weight function.

[0066] It can be understood that this embodiment uses a secondary selection to obtain a representative insulator data set (the optimized second feature set), and then obtains a local linear model through formula (22), and by analyzing the coefficients of the local linear model, it reflects the contribution of the input features of the deteriorated insulator position prediction model to the prediction results. In addition to avoiding the selection of similar explanation instances and iteratively adding samples with the highest marginal coverage gain, this method can also effectively reduce the construction time of the explanation model.

[0067] See also Figure 6 In a specific embodiment, the defect location of the porcelain insulator and the model visualization effect are tested according to the porcelain insulator defect detection method.

[0068] In this embodiment, in order to verify the superiority of the algorithm, the temperature data of 1400 strings of 110 kV deteriorated insulator strings and 100 strings of normal insulator strings at different positions were selected for the experiment. The ratio of the training set, test set and validation set in this experiment is 8:1:1.

[0069] The hardware is divided into servers and edge devices. The server is equipped with a 2.4 GHz dual-memory processor and an NVIDIA GeForce RTX 3090 image processor, which can well complete the tasks of data storage and data processing. The edge device uses the DJI Meridian M300 RTK as the platform and uses Cloudcrest iCrest 2.0 to build the edge processor.

[0070] In order to ensure that the system reaches the optimal state during operation, it is necessary to adjust the hyperparameters that play a major role in the direct torque control algorithm. We combine grid search and 5-fold cross validation method.

[0071] In order to test the performance of the SLIME algorithm, we conducted an ablation experiment on the visualization algorithm. The first string of faulty insulators was selected for the experiment, and the performance of the random selection method and the random sampling method were compared.

[0072] First, by calculating the gradient and difference value of the time series temperature data, the time series temperature difference is highlighted, and the subsequent training iteration time is accelerated. Then, a decision tree super clustering DTC model is constructed, and redundant features are screened through super clustering. The decision tree of the greedy algorithm is used as the core to build the prediction model. Finally, based on the decision tree super clustering DTC model and the validation set data, the local interpretable model is selected through the sub-model for visual analysis.

[0073] Finally, the visualization of classification features is as follows Figure 7 and Figure 8 As shown in the figure, the relative temperature NO1 value has the greatest effect on the gain of pattern classification, followed by the relative temperature NO9, NO12 and NO2 values, but these three values ​​have little effect on pattern classification, and the value of NO13 is negative for the classification of model effects. As can be seen from the figure, the weight of NO1 still accounts for the maximum gain of model classification feature recognition, and the values ​​of NO12 and NO13 are no longer the gain of classification recognition, and the influence ratio increases.

[0074] By comparing the data of three iterations, it can be found that the porcelain insulator defect detection method using the SLIME algorithm in the present invention is superior to the porcelain insulator defect detection method using the existing LIME algorithm in terms of model convergence speed and detection effect.

[0075] Any of the above embodiments can monitor the time series temperature information of the insulator string in real time, quickly generate a temperature feature set, and predict the defect position of the insulator string according to the rules learned from the temperature feature set through the trained position prediction model, thereby realizing real-time detection of overheating defects of porcelain insulators. At the same time, a linear model is generated based on the second feature set and the output results of the model. This linear model can show the relationship between the input features and the output results of the position prediction model, which not only helps to understand how the position prediction model makes predictions based on the input features, but also can identify which features have the greatest impact on the prediction results, thereby realizing visualization of the position prediction model, intuitively displaying the model's prediction logic and feature importance, and helping to enhance the interpretability and reliability of the model.

[0076] Please refer to Fig. 9 , Fig. 9 It is a structural schematic diagram of a porcelain insulator defect detection device.

[0077] This embodiment also provides a porcelain insulator defect detection device, which includes: The acquisition module 401 is used to generate a temperature feature set corresponding to the insulator string to be tested based on the time-series temperature information of the insulator string to be tested, where the temperature feature set includes a first feature set and a second feature set.

[0078] The construction module 402 is used to construct a location prediction model and train the location prediction model using the first feature set.

[0079] The training module 403 is used to input the second feature set into the trained position prediction model, and obtain an output result after determining the defect position of the insulator string to be tested through the trained position prediction model.

[0080] The display module 404 is used to generate a linear model according to the second feature set and the output result, and to display the relationship between the input features and the output result of the position prediction model through the linear model.

[0081] It will be appreciated by those skilled in the art that all or some of the steps and devices in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transient medium). As known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. As is well known to those skilled in the art, communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0082] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0083] An embodiment of the present application also provides an electronic device, the electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, any of the above porcelain insulator defect detection methods is implemented.

[0084] refer to Fig.10 , Fig.10 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes: The processor 501 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application; The memory 502 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 502 can store operating devices and other applications. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 502, and the processor 501 is called to execute the porcelain insulator defect detection method of the embodiment of this application; Input / output interface 503, used to implement information input and output; Communication interface 504, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.); A bus 505 that transmits information between various components of the device (e.g., the processor 501, the memory 502, the input / output interface 503, and the communication interface 504); The processor 501 , the memory 502 , the input / output interface 503 and the communication interface 504 are connected to each other in communication within the device via the bus 505 .

[0085] It can be understood that the contents of the above method embodiments are all applicable to the electronic device embodiment, the functions specifically implemented by the electronic device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0086] An embodiment of the present application also provides a computer-readable storage medium, which stores a program executable by a processor. When the program executable by the processor is executed by the processor, it is used to implement the porcelain insulator defect detection method as described in any one of the above specific embodiments.

[0087] An embodiment of the present application also discloses a computer program product, including a computer program or computer instructions, wherein the computer program or computer instructions are stored in a computer-readable storage medium, a processor of a computer device reads the computer program or computer instructions from the computer-readable storage medium, and the processor executes the computer program or computer instructions, so that the computer device executes the porcelain insulator defect detection method as described in any of the previous embodiments.

[0088] It can be understood that the contents of the above method embodiments are all applicable to the present storage medium embodiments, the functions specifically implemented by the present storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0089] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, device, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices. It should be understood that in the present application, "at least one (item)" refers to one or more, and "a plurality" refers to two or more.

[0090] In the several embodiments provided in the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0091] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0092] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0093] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store program codes.

[0094] Although the description of the present application has been quite detailed and specifically describes several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but should be regarded as providing a broad possible interpretation of these claims by reference to the attached claims, taking into account the prior art, so as to effectively cover the intended scope of the present application. In addition, the above description of the present application is based on the embodiments foreseeable by the inventor, and its purpose is to provide a useful description, and those non-substantial changes to the present application that have not yet been foreseen may still represent equivalent changes to the present application.

Claims

1. A method for detecting defects in porcelain insulators, characterized in that: The method comprises: Based on the time-series temperature information of the insulator string to be measured, a temperature feature set corresponding to the insulator string to be measured is generated, wherein the temperature feature set includes a first feature set and a second feature set; Constructing a location prediction model, and training the location prediction model using the first feature set; Inputting the second feature set into the trained position prediction model, and determining the defect position of the insulator string to be tested by the trained position prediction model to obtain an output result; A linear model is generated based on the second feature set and the output result, and the relationship between the input features and the output result of the position prediction model is displayed through the linear model.

2. A porcelain insulator defect detection method according to claim 1, characterized in that: The time series temperature information includes the current temperature value of each insulator in the insulator string to be tested and the ambient temperature value of the environment where each insulator is located; Based on the time-series temperature information of the insulator string to be measured, a temperature feature set corresponding to the insulator string to be measured is generated, wherein the temperature feature set includes a first feature set and a second feature set, including: Acquiring time-series temperature information of the insulator string to be tested; Based on the time series temperature information, determining a temperature change value of each of the insulators, wherein the temperature change value of the insulator to be measured is a difference between a current temperature value of the insulator to be measured and an ambient temperature value; Based on the temperature change value of each of the insulators, a temperature characteristic information factor of the insulator string to be tested is generated and used as the temperature characteristic of the insulator string to be tested; Return to the step of obtaining the time-series temperature information of the insulator string to be measured, and when a plurality of the temperature characteristics are obtained, generate a temperature characteristic set corresponding to the insulator string to be measured and divide it into a first characteristic set and a second characteristic set.

3. A porcelain insulator defect detection method according to claim 2, characterized in that: The method further comprises: Dividing the temperature feature set into a plurality of data clusters, each of the data clusters including a plurality of the temperature features; Determine a first average distance and a second average distance of each temperature feature, wherein the first average distance is an average distance value between the temperature feature and temperature features of the same data cluster, and the second average distance is an average distance value between the temperature feature and temperature features of different data clusters; Determining a profile coefficient of each of the temperature features based on a functional relationship between the first average distance and the second average distance; Based on the result of comparing the silhouette coefficient of each of the temperature features with a preset threshold, a plurality of target temperature features are determined, and a filtered temperature feature set is generated according to the plurality of target temperature features.

4. A method for detecting defects in porcelain insulators according to claim 1, characterized in that: The method further comprises: Clustering multiple temperature features in the temperature feature set into multiple classes; determining dispersion values ​​for the plurality of classes; Determining a target class based on a result of comparing the dispersion degree value with a preset dispersion degree value; Each of the target classes is used as a target temperature feature, and a clustered temperature feature set is generated according to a plurality of the target temperature features.

5. A porcelain insulator defect detection method according to claim 1, characterized in that: The step of constructing a location prediction model and training the location prediction model using the first feature set includes: Generate an objective function of the position prediction model based on the prediction accuracy and model complexity of the position prediction model; Adjusting model parameters of the position prediction model according to the current prediction accuracy of the position prediction model; Updating the current model complexity and the current prediction accuracy of the position prediction model according to the adjusted model parameters; Updating the objective function according to the updated model complexity and the updated prediction accuracy; When the position prediction model is trained using the first feature set, the updated prediction accuracy is used as the current prediction accuracy, and the step of adjusting the model parameters of the position prediction model according to the current prediction accuracy of the position prediction model is returned to execute until the optimal objective function is obtained, and the optimal objective function is an objective function that meets preset conditions.

6. A method for detecting defects in porcelain insulators according to claim 5, characterized in that: The position prediction model includes at least one tree structure, each of the tree structures includes a first node and a second node, and the first node and the second node are both model parameters; The first node is used to generate a new first node or a new second node, and the second node is used to output a prediction result; The adjusting the model parameters of the position prediction model according to the current prediction accuracy of the position prediction model includes: Determine a structure score of the current tree structure based on the current prediction accuracy, the number of first nodes and second nodes of the current tree structure, and the node value of each node; Based on the structural score, a new first node or a new second node is generated on the tree structure to adjust the model parameters of the position prediction model.

7. A porcelain insulator defect detection method according to claim 1, characterized in that: The method comprises: Selecting samples from the second feature set, and grouping the selected samples into a sample subset; When the selected samples are added to the sample subset, a marginal gain value of each sample in the second feature set is determined according to a preset function; Based on the marginal gain value, determining a target sample in the second feature set and using it as the selected sample; When a preset stop condition is reached, the selection of samples in the second feature set is stopped, and the sample subset is used as the optimized second feature set.

8. A porcelain insulator defect detection device, characterized in that: The device comprises: An acquisition module, configured to generate a temperature feature set corresponding to the insulator string to be measured based on the time-series temperature information of the insulator string to be measured, wherein the temperature feature set includes a first feature set and a second feature set; A construction module, used to construct a location prediction model and train the location prediction model using the first feature set; A training module, used for inputting the second feature set into the trained position prediction model, and obtaining an output result after determining the defect position of the insulator string to be tested through the trained position prediction model; A display module is used to generate a linear model based on the second feature set and the output result, and to display the relationship between the input features and the output result of the position prediction model through the linear model.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the porcelain insulator defect detection method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the porcelain insulator defect detection method according to any one of claims 1 to 7 is implemented.

Images