Switchgear Detection Method Based on Electrical Parameter Analysis

Abstract

The present application relates to the field of data analysis technology, and specifically to a switchgear detection method based on electrical parameter analysis, including: obtaining initial electrical parameter data in the switchgear detection process; constructing a clustering feature tree based on all initial electrical parameter data; obtaining multiple electrical parameter sequences for each clustering feature; segmenting the electrical parameter sequence to obtain multiple subsequences, and determining the deviation of the electrical data based on the data distribution deviation within the subsequence and the fitting result of the data within the subsequence; determining the aggregation degree of clustering features in non-leaf nodes based on the deviation of the electrical data; obtaining an aggregation feature value based on the aggregation degree; determining the branching factor coefficient of the non-leaf node based on the aggregation feature value combined with the number of layers of the non-leaf node in the clustering feature tree; updating the clustering feature tree based on the branching factor coefficient, and detecting abnormal electrical data in the electrical parameter data collected during the switchgear detection. The present application improves the detection results of abnormal electrical parameter data of the switchgear during the detection process.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis, and in particular to a switch device detection method based on electrical parameter analysis. Background Art

[0002] The detection of switchgear is an important part to ensure the stable and reliable operation of switchgear. At present, the performance detection contents of switchgear include electrical performance detection, insulation resistance test, contact resistance test, etc. When detecting switchgear, it is usually to use monitoring equipment to monitor the electrical parameter data of switchgear under various tests, use oscilloscopes and other measuring equipment to display the waveform of electrical parameter data, and then use the predetermined judgment threshold or experienced personnel to detect whether the switchgear is abnormal. This kind of detection method depends on the rationality of the judgment threshold under various detection contents, and cannot directly measure the fault information of the switchgear.

[0003] At present, a frequently used detection method is to use data analysis to process and analyze the electrical parameter data output by the switch device during the detection process to directly measure whether the switch device has a fault. Among them, cluster detection is one of the commonly used data analysis technologies. For example, the BIRCH (Balanced Iterative Reducing and Clustering using Hierarchies) clustering algorithm is used to process the electrical parameter data when the switch device is detected, and the switch device is analyzed based on the clustering results to determine whether it has a fault. However, when using the BIRCH algorithm, the numerator factor parameters need to be set, resulting in the inability to accurately determine whether the switch device has a fault and obtain accurate detection results when the distribution of electrical parameter data during the switch device detection process is relatively discrete. Summary of the invention

[0004] In order to solve the above problems, the present application provides a switch device detection method based on electrical parameter analysis.

[0005] The switchgear detection method based on electrical parameter analysis of the present application adopts the following technical solutions:

[0006] An embodiment of the present application provides a switch device detection method based on electrical parameter analysis, the method comprising the following steps:

[0007] Obtaining some initial electrical parameter data during the switchgear detection process;

[0008] A clustering feature tree is constructed based on all initial electrical parameter data; several electrical parameter sequences of each clustering feature in each non-leaf node of the clustering feature tree are obtained; each electrical parameter sequence is segmented to obtain several subsequences, and the deviation of each electrical data is determined based on the data distribution deviation in the subsequence and the fitting result of the data in the subsequence; the aggregation degree of each clustering feature in each non-leaf node is determined based on the deviation of the electrical data;

[0009] Obtaining the clustering feature value of each non-leaf node according to the clustering degree of all clustering features in the non-leaf node; determining the branching factor coefficient of the non-leaf node based on the clustering feature value combined with the number of layers of the non-leaf node in the clustering feature tree;

[0010] The clustering feature tree is updated based on the branching factor coefficients of the non-leaf nodes to obtain an optimized clustering feature tree, and abnormal electrical data in the electrical parameter data collected in the switch device detection is determined based on the optimized clustering feature tree.

[0011] Preferably, the step of constructing a cluster feature tree based on all initial electrical parameter data includes the following specific steps:

[0012] The BIRCH clustering algorithm is used to cluster all initial electrical parameter data to obtain a cluster feature tree.

[0013] Preferably, the specific method of the several electrical parameter sequences of each cluster feature in each non-leaf node of the cluster feature tree is as follows:

[0014] Obtain all non-leaf nodes of the clustering feature tree, and determine all clustering features of each non-leaf node respectively;

[0015] All current values, voltage values ​​and impedance values ​​in the cluster corresponding to each clustering feature of each non-leaf node are sorted according to the acquisition time to obtain a current parameter sequence, a voltage parameter sequence and an impedance parameter sequence respectively, and the current parameter sequence, voltage parameter sequence and impedance parameter sequence are all used as an electrical parameter sequence of the clustering feature.

[0016] Preferably, the specific method for determining the deviation of each electrical data is as follows:

[0017] Performing data fitting on the electrical data in each subsequence, and calculating the slope characteristic value of the subsequence based on the fitting result;

[0018] For each electrical data of each electrical parameter sequence, count all subsequences in each electrical parameter sequence that contain data values ​​equal to each electrical data, and calculate the mean of the difference metric distances between the statistically obtained subsequences;

[0019] The ratio of the mean value to the number of non-repeated electrical data values ​​in the statistically obtained subsequence is used as the deviation weight of each electrical data;

[0020] The absolute value of the difference between each electrical data and the mean value of all electrical data in the subsequence where each electrical data is located is calculated, and the product of the absolute value and the deviation weight of each electrical data is taken as the deviation degree of each electrical data.

[0021] Preferably, the slope characteristic value of the subsequence is calculated as follows:

[0022] Arrange the electrical data in the subsequence in chronological order, and obtain a fitting curve corresponding to the arrangement result by using a data fitting method;

[0023] The slope of each electrical data in the subsequence on the fitting curve is calculated respectively, and the average of the slopes of all electrical data in the subsequence is taken as the slope characteristic value of the subsequence.

[0024] Preferably, the degree of aggregation of each clustering feature in each non-leaf node is determined as follows:

[0025] The deviation of the electrical data in all electrical parameter sequences of each clustering feature under the non-leaf node is calculated respectively, the mean of the deviation of all the electrical parameters is negatively mapped, and the negative mapping result is used as the aggregation degree of each clustering feature in the non-leaf node.

[0026] Preferably, the step of obtaining the clustered feature value of each non-leaf node according to the clustering degree of all clustering features in the non-leaf nodes includes the following specific steps:

[0027] Taking the electrical data in each electrical parameter sequence of each cluster feature in each non-leaf node as input, using the LOF algorithm to obtain the abnormality score of each electrical data, then determining the third quartile of the abnormality score of the electrical data in all electrical parameter sequences of all cluster features in each non-leaf node, and taking all the electrical data with abnormality scores greater than the third quartile as a screening data;

[0028] The ratio of the mean of the clustering degree of all clustering features in each non-leaf node to the number of screened electrical data in each non-leaf node is taken as the clustering feature value of each non-leaf node.

[0029] Preferably, the branching factor coefficient of the non-leaf node is determined as follows:

[0030] Set the number of layers of the root node in the clustering feature tree to 1, and increase it downward in steps of 1 to determine the number of layers of each non-leaf node;

[0031] The ratio of the aggregated feature value of each non-leaf node to the sum of the aggregated feature values ​​of all non-leaf nodes in the clustering feature tree is used as the first weight;

[0032] The ratio of the difference between the number of layers of the clustering feature tree and the number of layers of each non-leaf node to the number of layers of the clustering feature tree is used as the level of the number of layers of each non-leaf node;

[0033] The product of the first weight and the level of each non-leaf node is used as the branching factor coefficient of each non-leaf node.

[0034] Preferably, the updating of the cluster feature tree based on the branching factor coefficient of the non-leaf node includes the following specific steps:

[0035] Calculate the product of the difference between the maximum empirical value and the minimum empirical value of the branching factor and the branching factor coefficient of each non-leaf node, and use the sum of the product and the minimum empirical value of the branching factor as the adaptive branching factor of each non-leaf node;

[0036] The adaptive branching factor of each non-leaf node on the clustering feature tree is calculated respectively, and the values ​​are assigned to each non-leaf node on the clustering feature tree in turn, so as to complete the update of the clustering feature tree and obtain the optimized clustering feature tree;

[0037] For each clustering feature in the leaf node on the optimized clustering feature tree, the number of electrical data in the cluster corresponding to each clustering feature is counted respectively, and the first quartile of the number of electrical data in the clusters corresponding to all the clustering features on the optimized clustering feature tree is calculated. The clustering features with a number less than the first quartile are regarded as abnormal clustering features in the optimized clustering feature tree.

[0038] Preferably, the detecting switch device detecting abnormal electrical data in the collected electrical parameter data comprises the following specific steps:

[0039] Using digital measuring equipment to collect electrical data of the switchgear during performance testing, and adding the collected electrical data to the optimized clustering feature tree;

[0040] After a newly collected electrical parameter data is added to the optimized clustering feature tree, if the newly collected electrical parameter data is classified into the cluster corresponding to the abnormal clustering feature or a new clustering feature is added to the optimized clustering feature tree, the newly collected electrical parameter data is regarded as abnormal electrical parameter data; otherwise, the newly collected electrical parameter data is normal electrical parameter data.

[0041] The beneficial effect of the technical solution of the present application is as follows: after obtaining the initial electrical parameter data of the switching device, the present application constructs a clustering feature tree through the initial electrical parameter data, and then obtains the electrical parameter sequence of the clustering features in the clustering feature tree; the degree of aggregation of each clustering feature in the non-leaf node is obtained through the deviation of each electrical data in each electrical parameter sequence; when obtaining the degree of aggregation, the degree of aggregation of the clustering feature is obtained by combining the deviation and the degree of abnormality of the electrical data, so that the subsequent acquisition of the clustering feature value results of the non-leaf nodes is more accurate, and the detection accuracy of the abnormal electrical parameter data is improved; after obtaining the clustering feature value of the non-leaf node, the branching factor coefficient of each non-leaf node is obtained in combination with the number of layers of the non-leaf node in the clustering feature tree, and then the adaptive branching factor of each non-leaf node is obtained, reducing the need to set fixed molecular factor parameters for the BIRCH clustering algorithm in advance; at the same time, the adaptive branching factor avoids the problem of too high tree depth and too discrete distribution of electrical parameter data during the update of the clustering feature tree, reduces the deviation of the clustering results of the newly added electrical parameter data, and then improves the measurement results of the abnormal electrical parameter data of the switching device, and more accurately measures the fault information of the switching device during the detection process. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 A flowchart of the implementation of the switch device detection method based on electrical parameter analysis provided in an embodiment of the present application;

[0044] Figure 2 A flow chart of obtaining abnormal electrical parameter data in the initial electrical parameter data collected during switch device detection provided in an embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to further explain the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following is a detailed description of the switch device detection method based on electrical parameter analysis proposed in the present application, its specific implementation, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0046] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0047] The specific scheme of the switch device detection method based on electrical parameter analysis provided by the present application is described in detail below with reference to the accompanying drawings.

[0048] See also Figure 1 , which shows an implementation flow chart of a switch device detection method based on electrical parameter analysis provided by an embodiment of the present application, the method comprising the following steps:

[0049] Step S1, obtaining some initial electrical parameter data of the switch device during detection.

[0050] It should be noted that in switchgear testing, electrical parameter analysis is used for switchgear fault diagnosis, performance evaluation, maintenance, etc., and initial electrical parameter data is collected during the performance testing of the switch.

[0051] Specifically, some initial electrical parameter data of the switch are obtained, as follows:

[0052] During the performance test of the switchgear, the current, voltage and impedance of the switchgear are collected by digital measuring equipment. The electrical sensor outputs a current value, a voltage value and an impedance value every 1 second. The current value, voltage value and impedance value output each time are used as an initial electrical parameter data to obtain all the initial electrical parameter data during the switchgear testing process.

[0053] It should be noted that the digital measuring device includes a variety of electrical sensors, such as current sensors and voltage sensors, for collecting initial electrical parameter data. The digital measuring device also includes a CPU for analyzing the initial electrical parameter data, a display screen for displaying the waveform of the initial electrical parameter data, and a prompting device. Other digital measuring devices can be used in other embodiments, and this application does not specifically limit the digital measuring device.

[0054] Step S2: performing data analysis based on the initial electrical parameter data to obtain abnormal electrical data in the initial electrical parameter data collected during the switch device detection.

[0055] The initial electrical parameter data collected by the switch device during the detection process is used to analyze whether there is an abnormality in the switch device during the detection process, and the clustering feature tree is updated based on the analysis results of the electrical parameter data to detect abnormal electrical data in the initial electrical parameter data. The implementation process of this step is as follows: Figure 2 As shown, it specifically includes step 201, step 202 and step 203:

[0056] Step S201, constructing a cluster feature tree according to all initial electrical parameter data; obtaining the deviation of each electrical data in the electrical parameter sequence of each cluster feature; and obtaining the cluster feature value of each cluster feature according to the deviation.

[0057] Specifically, the BIRCH clustering algorithm is used to construct a clustering feature tree based on all the initial electrical parameter data obtained according to the preset parameters. After the construction is completed, the clustering feature tree (CF-tree) consists of non-leaf nodes and leaf nodes, each of which has multiple clustering features (CF). Since the number of CFs in each non-leaf node is different, and the number of data in each CF is also different. Therefore, when new data is added later, the value of the branching factor of each non-leaf node is also different. For example, if the data in a non-leaf node is relatively clustered, the corresponding branching factor value is relatively large, and the data in a non-leaf node is relatively scattered, then the corresponding branching factor value is smaller.

[0058] Specifically, the specific process of constructing a cluster feature tree based on all initial electrical parameter data is as follows:

[0059] First, all initial electrical parameter data are clustered using the BIRCH clustering algorithm to obtain a clustering feature tree. It should be noted that clustering data using the BIRCH clustering algorithm to obtain a clustering feature tree is a common technology in the field of data processing and will not be described in detail in this embodiment.

[0060] Secondly, first calculate the aggregation degree of the data in each non-leaf node in the clustering feature tree. Since there are multiple CFs in each non-leaf node, the aggregation degree of the data in each non-leaf node is determined by analyzing the aggregation degree of the data in each CF in the non-leaf node and the changing relationship of the aggregation degree between different CFs. After all the data are constructed into a clustering feature tree, the temporal characteristics of the data cannot be effectively represented. Therefore, when calculating the aggregation degree of the data in any CF, it is necessary to uniformly quantify it according to the fluctuation degree of the data in the non-leaf node and the deviation degree of the data in the non-leaf node.

[0061] Furthermore, the electrical parameter sequence of each cluster feature in each non-leaf node of the cluster feature tree is obtained. The specific acquisition method is as follows:

[0062] Get all non-leaf nodes of the clustering feature tree. For any non-leaf node, Take the non-leaf node as an example and get the All clustering features of non-leaf nodes. All current values, voltage values ​​and impedance values ​​in each clustering feature of non-leaf nodes are sorted according to the acquisition time to obtain a current parameter sequence, a voltage parameter sequence and an impedance parameter sequence, and the current parameter sequence, voltage parameter sequence and impedance parameter sequence are all used as an electrical parameter sequence of the clustering feature.

[0063] Furthermore, the deviation of each electrical data in each electrical parameter curve of the electrical parameter curve set is obtained. As an embodiment, the specific obtaining method is as follows:

[0064] First, any non-leaf node of the clustering feature tree is recorded as a target non-leaf node, any clustering feature of the target non-leaf node is recorded as a target clustering feature, and any electrical parameter sequence of the target clustering feature is recorded as a target electrical parameter sequence.

[0065] Secondly, all extreme points of the target electrical parameter sequence are obtained, and the target electrical parameter sequence is divided according to all the extreme points to obtain several electrical parameter subsequences of the target electrical parameter sequence; it should be noted that obtaining extreme points in the curve is a well-known technology, and the specific process will not be repeated here.

[0066] Then, for any subsequence, the electrical data in the subsequence are arranged in chronological order, and the fitting curve corresponding to the arrangement result is obtained by data fitting. The slope of each electrical data in the subsequence on the fitting curve is calculated respectively, and the mean of the slopes of all electrical data in the subsequence is taken as the slope characteristic value of the subsequence.

[0067] The data fitting method includes but is not limited to least squares fitting, curve fitting, polynomial fitting, and kernel method. Preferably, in the embodiment of the present application, the fitting curve corresponding to the subsequence is obtained by least squares fitting, which is a commonly used technology in the field of data processing, and the specific process is not repeated here.

[0068] It should be noted that compared with directly fitting the entire electrical parameter sequence, the electrical data in each subsequence is distributed more concentratedly. If there are mutation points in the electrical parameter sequence or the distribution is more discrete, fitting the entire electrical parameter sequence will cause some data to have a large fitting deviation on the fitting curve, affecting the reliability of subsequent data analysis results.

[0069] Here, the deviation of each electrical data in the subsequence is calculated based on the slope characteristic value of each subsequence and the data distribution deviation in the subsequence, which is used to characterize the possibility that each electrical data is abnormal in the subsequence. The calculation formula for the deviation of the jth electrical data is as follows: In the formula, is the deviation weight of the j-th electrical data, n is the number of subsequences in the target electrical parameter sequence that contain elements with the same data value as the j-th electrical data, is the number of non-repeated values ​​of the electrical data in n subsequences, It is the mean of the difference measurement distances between n subsequences; for example, in the subsequences (1, 3, 4, 3) and the subsequence (2, 3, 5), the non-repeated values ​​of the electrical data are 1, 2, 3, 4, 5, then the value of m is 5.

[0070] Among them, the difference measurement distance includes but is not limited to calculating the Euclidean distance, DTW distance, and value variance between subsequences. Preferably, the embodiment of the present application calculates the value variance between subsequences as the difference measurement distance. Value variance is a commonly used technology in the field of data processing, and the specific process will not be repeated.

[0071] Secondly, the deviation of the jth electrical data is determined by combining the difference between the jth electrical data and the rest of the electrical data in its subsequence: In the formula, is the deviation of the jth electrical data in the target electrical parameter sequence, is the normalization function, To take the absolute value, is the data value of the jth electrical data in the target electrical parameter sequence, is the average value of the data values ​​of all electrical data in the subsequence where the jth electrical data in the target electrical parameter sequence is located.

[0072] It should be noted that the fewer the number of subsequences containing the data value of the j-th electrical data in the target electrical parameter sequence, the lower the probability of the j-th electrical data value appearing during the detection of the switching device, and the more likely it is that the measurement data is generated by the abnormality of the switching device. At the same time, the greater the difference between the subsequences containing elements equal to the j-th electrical data value, the more unstable the state of the switching device corresponding to the j-th electrical data is, and the more likely that the switching device is abnormal; and the greater the data value difference between the j-th electrical data and the electrical data in the subsequence to which it belongs, The larger the value of , it means that under the relatively stable distribution, the j-th electrical data still has a high outlier property. The higher the degree to which the data value of the j-th electrical data deviates from the average level, the greater the deviation of the j-th electrical data.

[0073] According to the above process, the deviation of each electrical data in the target electrical parameter sequence is calculated respectively, and the mean deviation of all the electrical parameters is negatively mapped, and the negative mapping result is used as the aggregation degree of each clustering feature in each non-leaf node. The larger the aggregation degree, the higher the degree of aggregation of the electrical data in the corresponding clustering feature. Conversely, the smaller the aggregation degree, the lower the degree of aggregation of the electrical data in the corresponding clustering feature.

[0074] The negative mapping is to ensure that the aggregation of the clustering features is negatively correlated with the mean deviation of the electrical parameter. In one embodiment, the inverse of the mean deviation can be used as the aggregation of the clustering features; in another embodiment, the mean deviation can be used as a function The independent variable is used, and the output function value is used as the aggregation degree of the clustering feature; in other embodiments, the implementer may also use other negative mapping functions to obtain the aggregation degree of the clustering feature, and this application does not make any special limitation on negative mapping.

[0075] Step S202: obtaining a branching factor coefficient of each non-leaf node according to the clustered eigenvalue of the non-leaf node and the number of layers of the non-leaf node in the clustering feature tree; and obtaining an adaptive branching factor of each non-leaf node according to the branching factor coefficient.

[0076] According to the above process, the aggregation degree of each clustering feature in the non-leaf node is obtained respectively. Since a non-leaf node corresponds to multiple clustering features, for all clustering features of any non-leaf node, the higher the aggregation degree of these clustering features and the fewer abnormal data in these aggregation degrees, it means that the aggregation degrees of all clustering features in the non-leaf node are closer, the data in the non-leaf node is more stable, and the corresponding data in the non-leaf node has a higher degree of aggregation.

[0077] Furthermore, the electrical data in each electrical parameter sequence of each cluster feature in each non-leaf node is used as input, and the LOF algorithm (Local Outlier Factor) is used to obtain the abnormal score of each electrical data. Then, the abnormal scores of the electrical data in all electrical parameter sequences of all cluster features in each non-leaf node are counted, and the third quartiles of all the abnormal scores are determined respectively. All the electrical data with abnormal scores greater than the third quartile are used as a screening data, and each screening data is considered to be the detection data that may be collected in each non-leaf node when the switch device is in an abnormal state. Among them, the calculation of quartiles is a commonly used technology in the field of data processing, and the specific process will not be repeated.

[0078] Secondly, the clustering feature value of each non-leaf node is obtained according to the clustering degree of all clustering features in the non-leaf nodes. As an embodiment, the specific calculation method is as follows: In the formula, is the mean of the aggregation degree of all clustering features in each non-leaf node, The number of electrical data filtered in each non-leaf node; is the aggregated eigenvalue of non-leaf nodes.

[0079] It should be noted that The smaller the value, the fewer outliers there are in all the data contained in the non-leaf node, and the higher the aggregation degree of the corresponding non-leaf node. The aggregation degree of all clustering features in the non-leaf node is combined to obtain the aggregation feature value of the non-leaf node.

[0080] It should be noted that the above-mentioned clustering eigenvalues ​​of each non-leaf node are obtained. Since the non-leaf nodes are at different levels in the clustering feature tree, the levels of their corresponding clustering clusters are also different. For example, the closer the non-leaf node is to the root node, the higher the level of its corresponding clustering cluster. In the subsequent setting of the adaptive branching factor, the larger the clustering eigenvalue of the non-leaf node and the higher the corresponding clustering cluster level, the larger the corresponding adaptive coefficient should be.

[0081] Specifically, according to the clustered eigenvalues ​​of the non-leaf nodes and the number of layers of the non-leaf nodes in the clustering feature tree, the branching factor coefficient of each non-leaf node is obtained. As an embodiment, the specific calculation method is as follows: In the formula, is the branching factor coefficient of the i-th non-leaf node, is the aggregate eigenvalue of the i-th target non-leaf node, is the sum of the aggregated eigenvalues ​​of all non-leaf nodes in the clustering feature tree, H, They are respectively the number of layers of the clustering feature tree and the number of layers of the i-th non-leaf node in the clustering feature tree.

[0082] It should be noted that the number of layers where the root node in the cluster feature tree is located is 1, and the number of layers of each non-leaf node is determined by the step size of 1. The larger the value of is, the denser the distribution of electrical data in the i-th non-leaf node is, and the larger the corresponding branching factor coefficient is. It indicates the level of the i-th non-leaf node in the clustering feature tree, indicating that the closer the i-th non-leaf node is to the root node in the clustering feature tree, the larger its value and the larger the branching factor coefficient should be.

[0083] Furthermore, according to the above process, the adaptive branching factor of each non-leaf node is calculated respectively. The higher the clustering degree of the non-leaf node, the larger the branching factor coefficient, and the larger the corresponding adaptive branching factor. By adjusting the branching factor, when new data is added later, it can avoid excessive addition of data as new CF, reducing the depth and complexity of the clustering feature tree.

[0084] Specifically, the adaptive branching factor of each non-leaf node is obtained according to the branching factor coefficient. As an embodiment, the specific calculation method is as follows: In the formula, is the adaptive branching factor of the i-th non-leaf node, is the branching factor coefficient of the i-th non-leaf node, , They are the maximum and minimum empirical values ​​of the branching factor respectively.

[0085] Step S203: updating the clustering feature tree according to the adaptive branching factor to obtain abnormal electrical data.

[0086] Specifically, the adaptive branching factor of each non-leaf node on the clustering feature tree is calculated respectively, and is assigned to each non-leaf node on the clustering feature tree in turn, so as to complete the update of the clustering feature tree and obtain the optimized clustering feature tree.

[0087] It should be noted that the existing method of updating the clustering feature tree by adaptively adjusting the branching factor of each non-leaf node in the clustering feature tree is not described in detail.

[0088] Secondly, for the clustering features in the leaf nodes on the optimized clustering feature tree, the number L of electrical data in the cluster corresponding to the clustering feature is counted respectively. The larger the number L of electrical data, the larger the number of electrical data with the same data features collected during the detection of the switch device, and the more likely the corresponding clustering feature is the data collected when the switch device is normal. The number L corresponding to all the clustering features on the optimized clustering feature tree is obtained, and the first quartile of all the numbers L is calculated. The clustering features with a number L less than the first quartile are regarded as abnormal clustering features in the optimized clustering feature tree, and the electrical data in the cluster corresponding to the abnormal clustering feature is regarded as abnormal electrical data.

[0089] Step S3: adding the collected data during the switch device detection process to the optimized clustering feature tree to determine whether the switch device is abnormal.

[0090] It should be noted that the above process updates the cluster feature tree through the adaptive branching factor of the non-leaf node, and uses the optimized cluster feature tree to detect abnormal electrical data. For the switchgear, the digital measuring equipment is also used to collect the electrical data of the switchgear during the performance detection process, and the collected electrical data is added to the optimized cluster feature tree to determine whether the switchgear has an abnormality during the detection process.

[0091] Specifically, when a newly collected electrical parameter data is added to the optimized clustering feature tree, if the newly collected electrical parameter data is classified into the cluster corresponding to the abnormal clustering feature or a new clustering feature is added to the optimized clustering feature tree, the newly collected electrical parameter data is regarded as abnormal electrical parameter data; otherwise, the newly collected electrical parameter data is normal electrical parameter data.

[0092] It should be noted that when abnormal electrical parameter data is detected, it means that there is an abnormality in the switch device during the detection process, and the prompt device emits a prompt sound. Subsequently, engineers can perform corresponding maintenance on the switch device by analyzing the abnormal electrical parameter data to ensure that the switch device operates normally.

[0093] It should be noted that the above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

[0094] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A switchgear detection method based on electrical parameter analysis, characterized in that: The method comprises the following steps: Obtaining some initial electrical parameter data during the switchgear detection process; A clustering feature tree is constructed based on all initial electrical parameter data; several electrical parameter sequences of each clustering feature in each non-leaf node of the clustering feature tree are obtained; each electrical parameter sequence is segmented to obtain several subsequences, and the deviation of each electrical data is determined based on the data distribution deviation in the subsequence and the fitting result of the data in the subsequence; the aggregation degree of each clustering feature in each non-leaf node is determined based on the deviation of the electrical data; Obtaining the clustering feature value of each non-leaf node according to the clustering degree of all clustering features in the non-leaf node; determining the branching factor coefficient of the non-leaf node based on the clustering feature value combined with the number of layers of the non-leaf node in the clustering feature tree; The clustering feature tree is updated based on the branching factor coefficients of the non-leaf nodes to obtain an optimized clustering feature tree, and abnormal electrical data in the electrical parameter data collected in the switch device detection is determined based on the optimized clustering feature tree.

2. The switch device detection method based on electrical parameter analysis according to claim 1, characterized in that: The specific steps of constructing a cluster feature tree based on all initial electrical parameter data are as follows: The BIRCH clustering algorithm is used to cluster all initial electrical parameter data to obtain a cluster feature tree.

3. The switchgear detection method based on electrical parameter analysis according to claim 1, characterized in that: The specific method of obtaining several electrical parameter sequences of each cluster feature in each non-leaf node of the cluster feature tree is as follows: Obtain all non-leaf nodes of the clustering feature tree, and determine all clustering features of each non-leaf node respectively; All current values, voltage values ​​and impedance values ​​in the cluster corresponding to each clustering feature of each non-leaf node are sorted according to the acquisition time to obtain a current parameter sequence, a voltage parameter sequence and an impedance parameter sequence respectively, and the current parameter sequence, voltage parameter sequence and impedance parameter sequence are all used as an electrical parameter sequence of the clustering feature.

4. The switchgear detection method based on electrical parameter analysis according to claim 1, characterized in that: The specific method for determining the deviation of each electrical data is as follows: Performing data fitting on the electrical data in each subsequence, and calculating the slope characteristic value of the subsequence based on the fitting result; For each electrical data of each electrical parameter sequence, count all subsequences in each electrical parameter sequence that contain data values ​​equal to each electrical data, and calculate the mean of the difference metric distances between the statistically obtained subsequences; The ratio of the mean value to the number of non-repeated electrical data values ​​in the statistically obtained subsequence is used as the deviation weight of each electrical data; The absolute value of the difference between each electrical data and the mean value of all electrical data in the subsequence where each electrical data is located is calculated, and the product of the absolute value and the deviation weight of each electrical data is taken as the deviation degree of each electrical data.

5. The switchgear detection method based on electrical parameter analysis according to claim 4, characterized in that: The calculation method of the slope characteristic value of the subsequence is as follows: Arrange the electrical data in the subsequence in chronological order, and obtain a fitting curve corresponding to the arrangement result by using a data fitting method; The slope of each electrical data in the subsequence on the fitting curve is calculated respectively, and the average of the slopes of all electrical data in the subsequence is taken as the slope characteristic value of the subsequence.

6. The switchgear detection method based on electrical parameter analysis according to claim 4, characterized in that: The method for determining the aggregation degree of each clustering feature in each non-leaf node is as follows: The deviation of the electrical data in all electrical parameter sequences of each clustering feature under the non-leaf node is calculated respectively, the mean of the deviation of all the electrical parameters is negatively mapped, and the negative mapping result is used as the aggregation degree of each clustering feature in the non-leaf node.

7. The switchgear detection method based on electrical parameter analysis according to claim 4, characterized in that: The specific steps of obtaining the clustering feature value of each non-leaf node according to the clustering degree of all clustering features in the non-leaf nodes are as follows: Taking the electrical data in each electrical parameter sequence of each cluster feature in each non-leaf node as input, using the LOF algorithm to obtain the abnormality score of each electrical data, then determining the third quartile of the abnormality score of the electrical data in all electrical parameter sequences of all cluster features in each non-leaf node, and taking all the electrical data with abnormality scores greater than the third quartile as a screening data; The ratio of the mean of the clustering degree of all clustering features in each non-leaf node to the number of screened electrical data in each non-leaf node is taken as the clustering feature value of each non-leaf node.

8. The switchgear detection method based on electrical parameter analysis according to claim 7, characterized in that: The branching factor coefficient of the non-leaf node is determined as follows: The number of layers of the root node in the clustering feature tree is set to 1, and the number of layers of each non-leaf node is determined by increasing the number of layers downward in steps of 1. The ratio of the aggregated feature value of each non-leaf node to the sum of the aggregated feature values ​​of all non-leaf nodes in the clustering feature tree is used as the first weight; The ratio of the difference between the number of layers of the clustering feature tree and the number of layers of each non-leaf node to the number of layers of the clustering feature tree is used as the level of the number of layers of each non-leaf node; The product of the first weight and the level of each non-leaf node is used as the branching factor coefficient of each non-leaf node.

9. The switchgear detection method based on electrical parameter analysis according to claim 8, characterized in that: The specific steps of updating the clustering feature tree based on the branching factor coefficient of the non-leaf node are as follows: Calculate the product of the difference between the maximum empirical value and the minimum empirical value of the branching factor and the branching factor coefficient of each non-leaf node, and use the sum of the product and the minimum empirical value of the branching factor as the adaptive branching factor of each non-leaf node; The adaptive branching factor of each non-leaf node on the clustering feature tree is calculated respectively, and the values ​​are assigned to each non-leaf node on the clustering feature tree in turn, so as to complete the update of the clustering feature tree and obtain the optimized clustering feature tree; For each clustering feature in the leaf node on the optimized clustering feature tree, the number of electrical data in the cluster corresponding to each clustering feature is counted respectively, and the first quartile of the number of electrical data in the clusters corresponding to all the clustering features on the optimized clustering feature tree is calculated. The clustering features with a number less than the first quartile are regarded as abnormal clustering features in the optimized clustering feature tree.

10. The switchgear detection method based on electrical parameter analysis according to claim 1, characterized in that: The specific steps of detecting abnormal electrical data in the electrical parameter data collected during the detection of the switch device include the following: Using digital measuring equipment to collect electrical data of the switchgear during performance testing, and adding the collected electrical data to the optimized clustering feature tree; After a newly collected electrical parameter data is added to the optimized clustering feature tree, if the newly collected electrical parameter data is classified into the cluster corresponding to the abnormal clustering feature or a new clustering feature is added to the optimized clustering feature tree, the newly collected electrical parameter data is regarded as abnormal electrical parameter data; On the contrary, the newly collected electrical parameter data is normal electrical parameter data.

Images