Switchgear safety detection method based on big data

Abstract

The present invention discloses a switch cabinet safety detection method based on big data, and the method includes data collection, data preprocessing, data reconstruction, detection data clustering processing and switch cabinet safety detection. The present invention belongs to the field of switch cabinet detection technology, and specifically refers to a switch cabinet safety detection method based on big data. This scheme reconstructs the switch cabinet safety detection data by introducing a comparative learning objective, which helps to accurately identify the fault type of the switch cabinet; improves the alignment of the potential space of the switch cabinet safety detection data through three constraints, thereby improving the accuracy of the switch cabinet safety detection; calculates the local density by introducing multi-scale density and dynamic distance weights, and removes noise points based on an adaptive dynamic threshold, and realizes the processing of the switch cabinet safety detection data based on the gravitational balance state judgment, ensuring that the abnormal state and the normal state can be accurately separated, thereby making the efficiency of switch cabinet fault detection and classification high.

Description

Technical Field

[0001] The present invention relates to the technical field of switch cabinet detection, and in particular to a switch cabinet safety detection method based on big data. Background Art

[0002] The switchgear safety detection method refers to a systematic method of identifying and predicting the operating status and potential faults of the switchgear by collecting, analyzing and processing multi-dimensional data such as electrical, physical, and environmental data during the operation of the switchgear. Its core goal is to ensure the safety and stability of the switchgear and prevent major power accidents. However, the general switchgear safety detection method is unable to distinguish between current anomalies caused by overheating and short circuits caused by voltage fluctuations, and has poor ability to distinguish transitions between different operating states, which leads to low accuracy of switchgear safety detection; the general switchgear safety detection method is unable to distinguish between high-density normal operating data and low-density abnormal areas, and cannot accurately distinguish between temperature abnormal areas and random noise areas, which leads to low accuracy of switchgear safety detection. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a switch cabinet safety detection method based on big data. In view of the fact that the general switch cabinet safety detection method is unable to distinguish between the current anomaly caused by overheating and the short circuit caused by voltage fluctuation, and has poor ability to distinguish the transition between different operating states, which leads to the problem of low accuracy of switch cabinet safety detection, this scheme reconstructs the switch cabinet safety detection data by introducing comparative learning objectives, which helps to accurately identify the fault type of the switch cabinet; through three constraints, the alignment of the potential space of the switch cabinet safety detection data is improved, which can dynamically adapt to the complex switch cabinet safety detection environment, improve the long-term monitoring capability of the overall operating state, and thus improve the accuracy of the switch cabinet safety detection; in view of the fact that the general switch cabinet safety detection method is unable to distinguish between high-density positive Normal operation data and low-density abnormal areas cannot accurately distinguish between temperature abnormal areas and random noise areas, which leads to the problem of low accuracy of switchgear safety detection. This solution introduces multi-scale density and dynamic distance weight to calculate local density, and removes noise points based on adaptive dynamic thresholds to avoid outliers caused by sensor failures affecting the accuracy of abnormal detection and improve the detection ability of local abnormal points in the switchgear; by combining local density and distance weights to calculate the gravity between data points, and based on the gravitational balance state judgment to realize the processing of switchgear safety detection data, automatically capture the local structure of data under different operating states of the switchgear, accurately distinguish high-density areas and low-density areas of switchgear detection, ensure that abnormal states and normal states can be accurately separated, thereby making the efficiency of switchgear fault detection and classification high.

[0004] The technical solution adopted by the present invention is as follows: The switch cabinet safety detection method based on big data provided by the present invention comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: data preprocessing;

[0007] Step S3: data reconstruction;

[0008] Step S4: clustering of detection data;

[0009] Step S5: switch cabinet safety detection.

[0010] Furthermore, in step S1, the data collection is to collect historical switch cabinet safety detection data and real-time switch cabinet safety detection data; the historical switch cabinet safety detection data and the real-time switch cabinet safety detection data both include switch cabinet operation status data, switch cabinet temperature monitoring data, switch cabinet humidity monitoring data, environmental parameter data, switch operation data and partial discharge data; the historical switch cabinet safety detection data also includes safety detection results; the safety detection results are used as data labels; data labels are only selected and used as cluster labels and do not participate in data processing.

[0011] Furthermore, in step S2, the data preprocessing is to clean, convert and standardize the collected data; the data cleaning is to process missing values ​​and duplicate values; the data conversion is to convert the data into a vector form; and the standardization is to standardize the data based on the maximum and minimum normalization method.

[0012] Furthermore, in step S3, the data reconstruction specifically includes the following steps:

[0013] Step S31: Data reconstruction: Map the original switch cabinet safety detection data x to the potential representation z through the encoder, and calculate the potential distribution , expressed as: ; Reconstruct the original data from the potential representation z, expressed as: ; In the encoding stage, the contrastive learning objective is introduced to optimize the representation of the latent space through positive and negative sample pairs, and the contrastive learning objective loss function It is expressed as: ;in, is the mean of the latent distribution, generated by the encoder; is the standard deviation of the underlying distribution; is random noise sampled from a standard normal distribution; is the conditional probability distribution of reconstructing the original data; N(·) is a multidimensional Gaussian distribution; I is the identity matrix; is the decoder mapping function parameter; zi and z j is the potential representation of the positive sample pairs of switchgear safety detection data; k is the potential representation of negative samples of switchgear safety detection data; N is the total number of samples; sim(·) is the similarity in the latent space, which is the dot product operation; τ is the temperature coefficient;

[0014] Step S32: Calculate the KL divergence of the prior constraints, expressed as: ; Perform K-means clustering on the original switch cabinet safety detection data and calculate the KL divergence of the clustering constraints, which is expressed as: ; ; Construct the KL divergence of global distribution alignment and calculate the matching degree between the potential distribution and the global prior distribution , expressed as: ;in, is the KL divergence of the prior constraints; is the prior distribution of the latent variable; and is the mean and variance of the ith dimension of the latent distribution; d is the dimension of the latent space; is the KL divergence of the clustering constraint; is the final cluster label of the data point; and are the cluster center and the distribution range of the cluster respectively; is to select the cluster label c of the data point that takes the maximum value in the bracket; and are the standard deviation and mean of the global distribution, respectively;

[0015] Step S33: define the objective function; objective function It is expressed as: ; in, and is the weight coefficient; is the expected value of the data after the potential distribution; is the conditional probability distribution of the reconstructed original data;

[0016] Step S34: Use the gradient descent method to update the encoder parameters and the decoder parameters; a target threshold is set in advance, and when the objective function converges or reaches the target threshold, the final potential representation z is used as the reconstructed switch cabinet safety detection data.

[0017] Furthermore, in step S4, the detection data clustering process specifically includes the following steps:

[0018] Step S41: Calculate local density; introduce multi-scale weights to calculate the local density of each sample point, expressed as: ;in, is the local density of the data point; dist max is the maximum distance between points in the neighborhood; var(dist) is the variance of the distances between points in the neighborhood; nb(·) is the number of neighborhood points; x n is the switch cabinet safety detection data point; r1 is the neighborhood radius; mode(·) is the mode; d avg is the average distance of points in the neighborhood; NNr(·) is the set of data points within the neighborhood radius; x m is a neighborhood data point;

[0019] Step S42: remove noise points; introduce dynamic division and adaptive threshold adjustment, calculate the quartile value, expressed as: , , ; Determine the dynamic density threshold, expressed as: ; ; Identify and remove noise points, expressed as: ; ; Where Q1 and Q3 are the 25th and 75th percentiles of the local density, respectively; IQR(·) is the interquartile range; std(·) is the standard deviation; nedian(·) and mean(·) are the median and mean, respectively; LB min and UB max are the lower and upper limits of the density of dynamic calculation respectively; NP(·) is the set of points identified as noise points; X is the reconstructed switchgear safety detection data set; It is the core data set after removing noise points;

[0020] Step S43: Calculate the gravitational force between data points; expressed as: ; in, ; ; where F(·) is the attraction between data points; a and b are different data points; M(·) is the neighborhood similarity; d(·) is the distance metric; Na and Nb are the neighborhood point sets of data point a and data point b respectively; n3 is the total dimension of the data, k1 is the dimension index; a k1 and b k1 are the eigenvalues ​​of the k1th dimension of data point a and data point b respectively; and are the local densities of data point a and data point b respectively;

[0021] Step S44: Determine the gravitational balance state; for the core data set, start from r=1, gradually expand the neighborhood, calculate the gravitational force between each point and other points; for each point, select the first r points with the largest gravitational force as the neighborhood; check whether the gravitational balance state is achieved; if achieved, record the r value at the end as the natural neighborhood eigenvalue K of the data set, otherwise increase r and continue to expand the neighborhood; the gravitational balance state is determined as: ; When the gravitational equilibrium state is reached, the first K points with the largest gravitational force are used as the initial clustering centers, and the data points are allocated based on the gravitational force between the data points; where x p and x q are any two points in the core data set; n2 is the upper limit of the neighborhood radius; GNNr(·) is the set of points with the largest gravitational attraction in the neighborhood;

[0022] Step S45: clustering iteration; updating the cluster center based on the average position of the cluster, and updating the data point allocation based on the gravitational force between the data points; presetting the maximum number of clustering times, when the clustering converges or reaches the maximum number of clustering times, the clustering of the reconstructed switchgear safety detection data is completed;

[0023] Step S46: Noise point redistribution: For the noise points initially removed, they are redistributed to the nearest cluster according to gravity; expressed as: ;in, is the cluster C for the data point x n The gravitational force of x c is a point in cluster C.

[0024] Further, in step S5, the switch cabinet safety detection is based on the clustering result, and the label with the largest number of historical switch cabinet safety detection data in the cluster is used as the cluster label, and the cluster label corresponding to the real-time switch cabinet safety detection data is used as the detection result of the real-time switch cabinet safety detection data.

[0025] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0026] (1) In view of the fact that general switchgear safety detection methods are unable to distinguish between current anomalies caused by overheating and short circuits caused by voltage fluctuations, and have poor ability to distinguish transitions between different operating states, which leads to low accuracy in switchgear safety detection, this scheme introduces comparative learning objectives to reconstruct the switchgear safety detection data, which helps to accurately identify the fault type of the switchgear; through three constraints, the alignment of the potential space of switchgear safety detection data is improved, which can dynamically adapt to the complex switchgear safety detection environment, improve the long-term monitoring capability of the overall operating status, and thus improve the accuracy of switchgear safety detection.

[0027] (2) In view of the fact that general switchgear safety detection methods are unable to distinguish between high-density normal operating data and low-density abnormal areas, and are unable to accurately distinguish between temperature abnormal areas and random noise areas, which leads to low accuracy of switchgear safety detection, this scheme introduces multi-scale density and dynamic distance weight to calculate local density, and removes noise points based on adaptive dynamic thresholds to avoid outliers caused by sensor failures affecting the accuracy of abnormal detection, thereby improving the ability to detect local abnormal points in the switchgear; by combining local density and distance weight to calculate the gravitational force between data points, and based on the gravitational balance state judgment to realize the processing of switchgear safety detection data, the local structure of the data under different operating states of the switchgear is automatically captured, and the high-density area and low-density area of ​​the switchgear detection are accurately distinguished, ensuring that the abnormal state and the normal state can be accurately separated, thereby making the efficiency of switchgear fault detection and classification high. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic diagram of the flow of the switch cabinet safety detection method based on big data provided by the present invention;

[0029] Figure 2 It is a schematic diagram of the process of step S4.

[0030] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0032] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0033] Example 1, see Figure 1 The present invention provides a switch cabinet safety detection method based on big data, which includes the following steps:

[0034] Step S1: data collection, collecting historical switch cabinet safety detection data and real-time switch cabinet safety detection data;

[0035] Step S2: Data preprocessing, data cleaning, data conversion and standardization of the collected data;

[0036] Step S3: Data reconstruction: reconstruct the switchgear safety detection data by introducing comparative learning objectives; improve the alignment of the potential space of the switchgear safety detection data through three constraints;

[0037] Step S4: clustering of detection data, by introducing multi-scale density and dynamic distance weight to calculate local density, and removing noise points based on adaptive dynamic threshold, by combining local density and distance weight to calculate the gravity between data points, and based on the gravity balance state judgment to achieve the processing of switch cabinet safety detection data;

[0038] Step S5: Switch cabinet safety detection: implementing switch cabinet safety detection based on clustering results.

[0039] Example 2, see Figure 1 , this embodiment is based on the above embodiment. In step S1, the historical switch cabinet safety detection data and the real-time switch cabinet safety detection data both include switch cabinet operation status data, switch cabinet temperature monitoring data, switch cabinet humidity monitoring data, environmental parameter data, switch operation data and partial discharge data; the historical switch cabinet safety detection data also includes safety detection results; the safety detection results are used as data labels; the data labels are only selected and used as cluster labels and do not participate in data processing; the switch cabinet operation status data includes current, voltage, active power, reactive power, power factor and frequency; the switch cabinet temperature monitoring data includes the ambient temperature in the cabinet and the temperature of key parts; the environmental parameter data includes ambient pressure and airflow velocity; the switch operation data includes the frequency of switch actions and the operation time interval; the partial discharge data includes the partial discharge amplitude, the number of partial discharges and the discharge starting phase; the safety detection results include normal operation, slight abnormality, moderate abnormality and severe abnormality.

[0040] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, data cleaning is to process missing values ​​and duplicate values; data conversion is to convert data into vector form; and standardization is to standardize data based on the maximum and minimum normalization method.

[0041] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, data reconstruction specifically includes the following steps:

[0042] Step S31: Data reconstruction: Map the original switch cabinet safety detection data x to the potential representation z through the encoder, and calculate the potential distribution , expressed as: ; Reconstruct the original data from the potential representation z, expressed as: ; In the encoding stage, the contrastive learning objective is introduced to optimize the representation of the latent space through positive and negative sample pairs, and the contrastive learning objective loss function It is expressed as: ;in, is the mean of the latent distribution, generated by the encoder; is the standard deviation of the underlying distribution; is random noise sampled from a standard normal distribution; is the conditional probability distribution of reconstructing the original data; N(·) is a multidimensional Gaussian distribution; I is the identity matrix; is the decoder mapping function parameter; z i and z j is the potential representation of the positive sample pairs of switchgear safety detection data; k is the potential representation of negative samples of switchgear safety detection data; N is the total number of samples; sim(·) is the similarity in the latent space, which is the dot product operation; τ is the temperature coefficient;

[0043] Step S32: Calculate the KL divergence of the prior constraints, expressed as: ; Perform K-means clustering on the original switch cabinet safety detection data and calculate the KL divergence of the clustering constraints, which is expressed as: ; ; Construct the KL divergence of global distribution alignment and calculate the matching degree between the potential distribution and the global prior distribution , expressed as: ;in, is the KL divergence of the prior constraints; is the prior distribution of the latent variable; and is the mean and variance of the ith dimension of the latent distribution; d is the dimension of the latent space; is the KL divergence of the clustering constraint; is the final cluster label of the data point; and are the cluster center and the distribution range of the cluster respectively; is to select the cluster label c of the data point that takes the maximum value in the bracket; and are the standard deviation and mean of the global distribution, respectively;

[0044] Step S33: define the objective function; objective function It is expressed as: ; in, and is the weight coefficient; is the expected value of the data after the potential distribution; is the conditional probability distribution of the reconstructed original data;

[0045] Step S34: Use the gradient descent method to update the encoder parameters and the decoder parameters; a target threshold is set in advance, and when the objective function converges or reaches the target threshold, the final potential representation z is used as the reconstructed switch cabinet safety detection data.

[0046] By performing the above operations, the general switchgear safety detection method cannot distinguish between current anomalies caused by overheating and short circuits caused by voltage fluctuations, and has poor ability to distinguish transitions between different operating states, which leads to low accuracy of switchgear safety detection. This scheme reconstructs the switchgear safety detection data by introducing comparative learning objectives, which helps to accurately identify the fault type of the switchgear; the alignment of the potential space of switchgear safety detection data is improved through three constraints, which can dynamically adapt to the complex switchgear safety detection environment, improve the long-term monitoring capability of the overall operating status, and thus improve the accuracy of switchgear safety detection.

[0047] Example 5, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S4, the detection data clustering process specifically includes the following steps:

[0048] Step S41: Calculate local density; introduce multi-scale weights to calculate the local density of each sample point, expressed as: ;in, is the local density of the data point; dist max is the maximum distance between points in the neighborhood; var(dist) is the variance of the distances between points in the neighborhood; nb(·) is the number of neighborhood points; x n is the switch cabinet safety detection data point; r1 is the neighborhood radius; mode(·) is the mode; d avg is the average distance of points in the neighborhood; NNr(·) is the set of data points within the neighborhood radius; x m is a neighborhood data point;

[0049] Step S42: remove noise points; introduce dynamic division and adaptive threshold adjustment, calculate the quartile value, expressed as: , , ; Determine the dynamic density threshold, expressed as: ; ; Identify and remove noise points, expressed as: ; ; Where Q1 and Q3 are the 25th and 75th percentiles of the local density, respectively; IQR(·) is the interquartile range; std(·) is the standard deviation; nedian(·) and mean(·) are the median and mean, respectively; LB min and UB max are the lower and upper limits of the density of dynamic calculation respectively; NP(·) is the set of points identified as noise points; X is the reconstructed switchgear safety detection data set; It is the core data set after removing noise points;

[0050] Step S43: Calculate the gravitational force between data points; expressed as: ;in, ; ; where F(·) is the attraction between data points; a and b are different data points; M(·) is the neighborhood similarity; d(·) is the distance metric; Na and Nb are the neighborhood point sets of data point a and data point b respectively; n3 is the total dimension of the data, k1 is the dimension index; a k1 and b k1 are the eigenvalues ​​of the k1th dimension of data point a and data point b respectively; and are the local densities of data point a and data point b respectively;

[0051] Step S44: Determine the gravitational balance state; for the core data set, start from r=1, gradually expand the neighborhood, calculate the gravitational force between each point and other points; for each point, select the first r points with the largest gravitational force as the neighborhood; check whether the gravitational balance state is achieved; if achieved, record the r value at the end as the natural neighborhood eigenvalue K of the data set, otherwise increase r and continue to expand the neighborhood; the gravitational balance state is determined as: ; When the gravitational equilibrium state is reached, the first K points with the largest gravitational force are used as the initial clustering centers, and the data points are allocated based on the gravitational force between the data points; where x p and x q are any two points in the core data set; n2 is the upper limit of the neighborhood radius; GNNr(·) is the set of points with the largest gravitational attraction in the neighborhood;

[0052] Step S45: clustering iteration; updating the cluster center based on the average position of the cluster, and updating the data point allocation based on the gravitational force between the data points; presetting the maximum number of clustering times, when the clustering converges or reaches the maximum number of clustering times, the clustering of the reconstructed switchgear safety detection data is completed;

[0053] Step S46: Noise point redistribution: For the noise points initially removed, they are redistributed to the nearest cluster according to gravity; expressed as: ;in, is the cluster C for the data point x n The gravitational force of x c is a point in cluster C.

[0054] By performing the above operations, the general switch cabinet safety detection method cannot distinguish between high-density normal operating data and low-density abnormal areas, and cannot accurately distinguish between temperature abnormal areas and random noise areas, which leads to low accuracy of switch cabinet safety detection. This solution introduces multi-scale density and dynamic distance weight to calculate local density, and removes noise points based on adaptive dynamic thresholds to avoid outliers caused by sensor failures affecting the accuracy of abnormal detection, thereby improving the detection capability of local abnormal points in the switch cabinet; the gravity between data points is calculated by combining local density and distance weight, and the switch cabinet safety detection data is processed based on the gravitational balance state judgment, the local structure of the data under different operating states of the switch cabinet is automatically captured, and the high-density area and low-density area of ​​the switch cabinet detection are accurately distinguished, ensuring that the abnormal state and the normal state can be accurately separated, thereby making the efficiency of switch cabinet fault detection and classification high.

[0055] Example 6, see Figure 1 This embodiment is based on the above embodiment. In step S5, the switch cabinet safety detection is based on the clustering result. The label with the largest number of historical switch cabinet safety detection data in the cluster is used as the cluster label, and the cluster label corresponding to the real-time switch cabinet safety detection data is used as the detection result of the real-time switch cabinet safety detection data; when the detection result is moderately abnormal or severely abnormal, the relevant personnel are warned.

[0056] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0057] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0058] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A switch cabinet safety detection method based on big data, characterized in that: The method comprises the following steps: Step S1: data collection, collecting historical switch cabinet safety detection data and real-time switch cabinet safety detection data; Step S2: data preprocessing; Step S3: data reconstruction; Step S4: clustering of detection data; Step S5: switch cabinet safety detection, implementing switch cabinet safety detection based on clustering results; Step S4 includes step S41: calculating local density; introducing multi-scale weights to calculate the local density of each sample point, expressed as: ;in, is the local density of the data point; dist max is the maximum distance between points in the neighborhood; var(dist) is the variance of the distances between points in the neighborhood; nb(·) is the number of neighborhood points; x n is the switch cabinet safety detection data point; r1 is the neighborhood radius; mode(·) is the mode; d avg is the average distance of points in the neighborhood; NNr(·) is the set of data points within the neighborhood radius; x m is a neighborhood data point; In step S3, the data reconstruction specifically includes the following steps: Step S31: Data reconstruction: Map the original switch cabinet safety detection data x to the potential representation z through the encoder, and calculate the potential distribution , expressed as: ; Reconstruct the original data from the potential representation z, expressed as: ; In the encoding stage, the contrastive learning objective is introduced to optimize the representation of the latent space through positive and negative sample pairs, and the contrastive learning objective loss function It is expressed as: ;in, is the mean of the latent distribution, generated by the encoder; is the standard deviation of the underlying distribution; is random noise sampled from a standard normal distribution; is the conditional probability distribution of reconstructing the original data; N(·) is a multidimensional Gaussian distribution; I is the identity matrix; is the decoder mapping function parameter; z i and z j is the potential representation of the positive sample pairs of switchgear safety detection data; k is the potential representation of negative samples of switchgear safety detection data; N is the total number of samples; sim(·) is the similarity in the latent space, which is the dot product operation; τ is the temperature coefficient; Step S32: Calculate the KL divergence of the prior constraints, expressed as: ; Perform K-means clustering on the original switch cabinet safety detection data and calculate the KL divergence of the clustering constraints, which is expressed as: ; ; Construct the KL divergence of global distribution alignment and calculate the matching degree between the potential distribution and the global prior distribution , expressed as: ;in, is the KL divergence of the prior constraints; is the prior distribution of the latent variable; and is the mean and variance of the ith dimension of the latent distribution; d is the dimension of the latent space; is the KL divergence of the clustering constraint; is the final cluster label of the data point; and are the cluster center and the distribution range of the cluster respectively; is to select the cluster label c of the data point that takes the maximum value in the bracket; and are the standard deviation and mean of the global distribution, respectively; Step S33: define the objective function; objective function It is expressed as: ;in, and is the weight coefficient; is the expected value of the data after the potential distribution; is the conditional probability distribution of the reconstructed original data; Step S34: Use the gradient descent method to update the encoder parameters and the decoder parameters; a target threshold is set in advance, and when the objective function converges or reaches the target threshold, the final potential representation z is used as the reconstructed switch cabinet safety detection data.

2. The switch cabinet safety detection method based on big data according to claim 1 is characterized in that: In step S4, the detection data clustering process further includes the following steps: Step S42: remove noise points; introduce dynamic division and adaptive threshold adjustment, calculate the quartile value, expressed as: , , ; Determine the dynamic density threshold, expressed as: ; ; Identify and remove noise points, expressed as: ; ; Where Q1 and Q3 are the 25th and 75th percentiles of the local density, respectively; IQR(·) is the interquartile range; std(·) is the standard deviation; nedian(·) and mean(·) are the median and mean, respectively; LB min and UB max are the lower and upper limits of the density of dynamic calculation respectively; NP(·) is the set of points identified as noise points; X is the reconstructed switchgear safety detection data set; It is the core data set after removing noise points; Step S43: Calculate the gravitational force between data points; expressed as: ;in, ; ; where F(·) is the attraction between data points; a and b are different data points; M(·) is the neighborhood similarity; d(·) is the distance metric; Na and Nb are the neighborhood point sets of data point a and data point b respectively; n3 is the total dimension of the data, k1 is the dimension index; a k1 and b k1 are the eigenvalues ​​of the k1th dimension of data point a and data point b respectively; and are the local densities of data point a and data point b respectively; Step S44: Determine the gravitational balance state; for the core data set, start from r=1, gradually expand the neighborhood, calculate the gravitational force between each point and other points; for each point, select the first r points with the largest gravitational force as the neighborhood; check whether the gravitational balance state is achieved; if achieved, record the r value at the end as the natural neighborhood eigenvalue K of the data set, otherwise increase r and continue to expand the neighborhood; the gravitational balance state is determined as: ; When the gravitational equilibrium state is reached, the first K points with the largest gravitational force are used as the initial clustering centers, and the data points are allocated based on the gravitational force between the data points; where x p and x q are any two points in the core data set; n2 is the upper limit of the neighborhood radius; GNNr(·) is the set of points with the largest gravitational attraction in the neighborhood; Step S45: clustering iteration; updating the cluster center based on the average position of the cluster, and updating the data point allocation based on the gravitational force between the data points; presetting the maximum number of clustering times, when the clustering converges or reaches the maximum number of clustering times, the clustering of the reconstructed switchgear safety detection data is completed; Step S46: Noise point redistribution: For the noise points initially removed, they are redistributed to the nearest cluster according to gravity; expressed as: ;in, is the cluster C for the data point x n The gravitational force of x c is a point in cluster C.

3. The switch cabinet safety detection method based on big data according to claim 2 is characterized in that: In step S1, the historical switch cabinet safety detection data and the real-time switch cabinet safety detection data both include switch cabinet operation status data, switch cabinet temperature monitoring data, switch cabinet humidity monitoring data, environmental parameter data, switch operation data and partial discharge data; the historical switch cabinet safety detection data also includes safety detection results; the safety detection results are used as data labels; data labels are only selected and used as cluster labels and do not participate in data processing.

4. The switch cabinet safety detection method based on big data according to claim 1 is characterized in that: In step S5, the switch cabinet safety detection is based on the clustering result, and the label with the largest number of historical switch cabinet safety detection data in the cluster is used as the cluster label, and the cluster label corresponding to the real-time switch cabinet safety detection data is used as the detection result of the real-time switch cabinet safety detection data.

5. The switch cabinet safety detection method based on big data according to claim 1 is characterized in that: In step S2, the data preprocessing is to clean, convert and standardize the collected data; the data cleaning is to process missing values ​​and duplicate values; the data conversion is to convert the data into a vector form; and the standardization is to standardize the data based on the maximum and minimum normalization method.

Images