A fault identification and inspection method and system for a 220 kV substation bus

Abstract

The present invention relates to the technical field of electrical fault detection, and particularly relates to a method and system for fault identification and inspection of a 220 kV substation bus, including: training a primary fault identification model; obtaining the possibility that the fault in the sample is a transformer fault according to the changes in the current data in the bus and branch lines in the sample; obtaining the response duration of each sample; inputting all test samples into the primary fault identification model, and combining the response duration of each test sample to obtain the transformer and bus fault significance of the test samples; constructing a four-dimensional sample space, and obtaining the fault significance weight according to the distribution of the samples in the four-dimensional sample space; and testing the fault type when a fault occurs in the substation according to the fault significance weight. After obtaining the primary fault identification model by traditional methods, the present invention further analyzes the differences between transformer faults and bus faults and sets a correction module to accurately distinguish between transformer faults and bus faults.

Description

Technical Field

[0001] The present invention relates to the technical field of electrical fault detection, and particularly relates to a method and system for fault identification and inspection of a 220 kV substation busbar. Background Art

[0002] In the power system, as the core hub of the power transmission and distribution network, the safe operation of the busbars and transformers in a 220 kV substation directly affects the reliability and stability of the power grid. The busbars play a key role in collecting and distributing electric energy, while the transformers are the core devices for realizing voltage level conversion. Once a fault occurs in either of them, if the fault point cannot be quickly and accurately identified and isolated, it may lead to an expansion of the power outage range and even trigger cascading accidents. However, due to the close electrical connection between the busbar area and the transformer, the electrical quantity characteristics during a fault are similar. Therefore, the traditional substation fault inspection methods are not accurate enough to distinguish between transformer faults and busbar faults. Summary of the Invention

[0003] The present invention provides a method and system for fault identification and inspection of a 220 kV substation busbar to solve the existing problem that the traditional substation fault inspection methods are not accurate enough to distinguish between transformer faults and busbar faults.

[0004] The method and system for fault identification and inspection of a 220 kV substation busbar of the present invention adopt the following technical solutions:

[0005] An embodiment of the present invention provides a method for fault identification and inspection of a 220 kV substation busbar, the method comprising the following steps:

[0006] Obtain a plurality of training samples and a plurality of test samples through the substation, and train a primary fault identification model;

[0007] According to the changes in the current data in the busbar and branch lines in the samples, obtain the affected factor sequences of the busbar and branch lines in the samples; according to the affected factor sequences of the busbar and branch lines in the samples, obtain the affected moments of the busbar and branch lines in the samples; according to the affected moments of the busbar and branch lines in the samples, obtain the possibility that the fault in the sample is a transformer fault;

[0008] Obtain the response duration of each sample; input all the test samples into the primary fault identification model, and combine the response duration of each test sample to obtain the fault characteristics of the transformer and busbar for each test sample; combine the possibility that the fault in the test sample is a transformer fault to obtain the fault significance of the transformer and busbar for the test sample;

[0009] Construct a four-dimensional sample space according to the significance of transformer and bus faults in the test samples and the probability that the fault in the test samples is a transformer fault. Based on the distribution of the samples in the four-dimensional sample space, obtain the neighborhood samples and interference samples of each sample, and then obtain the fault significance weights of all samples, and test the fault types when the substation fails.

[0010] Preferably, the specific method for obtaining the influence factor sequences of the bus and branch lines in the sample according to the changes in the current data in the bus and branch lines in the sample is as follows:

[0011] Preset a local time range ; for any moment in any sample, use the time period from the moment to milliseconds after the moment as the local range of the moment; use the standard deviation of all current data in the local range of the moment in the bus as the influence factor of each moment in the bus.

[0012] Obtain the influence factors of each moment in the bus within the sample, and sort the influence factors of each moment in chronological order to obtain the bus influence factor sequence. Similarly, obtain several branch line influence factor sequences.

[0013] Preferably, the specific method for obtaining the affected moments of the bus and branch lines in the sample according to the influence factor sequences of the bus and branch lines in the sample is as follows:

[0014] For any influence factor in the bus influence factor sequence, use the influence factor as a segmentation point to divide the bus influence factor sequence into two influence factor segments. According to the influence factors in the first and second influence factor segments, combined with the bus influence factor sequence, obtain the influence degree of the bus at the moment corresponding to the influence factor. The specific calculation formula is:

[0015]

[0016] In the formula, represents the influence degree of the bus at the moment corresponding to the influence factor; represents the standard deviation of all influence factors in the bus influence factor sequence; represents the standard deviation of all influence factors in the first influence factor segment; represents the standard deviation of all influence factors in the second influence factor segment; represents the linear normalization function;

[0017] Obtain the degree of influence on the bus at all times; similarly, obtain the degree of influence on each branch line at all times, and take the time when the degree of influence on the bus in the sample is the greatest as the affected time of the bus; similarly, obtain the affected time of each branch line.

[0018] Preferably, the method for obtaining the possibility that the fault in the sample is a transformer fault according to the affected times of the bus and branch lines in the sample specifically includes:

[0019] For any sample, take the temporal distance between the affected time of the bus in the sample and the affected time of each branch line as the characteristic distance of each branch line; and take the branch line whose affected time is before the bus as the abnormal branch line; take the branch line whose affected time is after the bus as the normal branch line.

[0020] Take the branch line with the largest characteristic distance among the abnormal branch lines as the suspected fault branch line, and obtain the possibility that the fault in the sample is a transformer fault according to the characteristic distance of the suspected fault branch line and the characteristic distances of each normal branch line. The specific calculation formula is:

[0021]

[0022] In the formula, represents the possibility that the fault in the sample is a transformer fault; represents the standard deviation of the characteristic distances of all normal branch lines; represents the mean value of the characteristic distances of all normal branch lines; represents the characteristic distance of the suspected fault branch line.

[0023] Preferably, the method for obtaining the response duration of each sample specifically includes:

[0024] For any branch line in any sample, record the branch line connected to the same transformer as the corresponding branch line of the branch line; take the temporal distance between the branch line and its corresponding branch line at the affected time as the response duration of the branch line.

[0025] Preferably, the method for inputting all test samples into the primary fault recognition model and obtaining the fault characteristics of the transformer and bus of each test sample in combination with the response duration of each test sample specifically includes:

[0026] Input all test samples into the primary fault recognition model, obtain the degree of transformer fault and the degree of bus fault of each test sample, and obtain the correctly identified samples and the misidentified samples;

[0027] For any test sample, the average response time of all normal branches in the test sample is used as the response time of the test sample; the average response time of all normal branches in all correctly identified samples is used as the reference response time;

[0028] Based on the response time of the test sample and the reference response time, obtain the fault characteristics of the transformer of the test sample; the specific calculation formula is:

[0029]

[0030] In the formula, represents the fault characteristics of the transformer of the test sample; represents the response time of the test sample; represents the reference response time; represents the absolute value operation; represents the linear normalization function;

[0031] Based on the affected time of the bus of the test sample and the protection time of the test sample, obtain the fault characteristics of the bus of the test sample, and the protection time is the time when all protection devices in the historical data of the substation trigger protection actions; the specific calculation formula is:

[0032]

[0033] In the formula, represents the fault characteristics of the bus of the test sample; represents the protection time of the test sample; represents the affected time of the bus of the test sample; represents the exponential function with the natural constant as the base.

[0034] Preferably, the specific method for obtaining the fault significance of the transformer and the bus of the test sample includes:

[0035] For any test sample, based on the fault characteristics of the transformer of the test sample, the fault characteristics of the bus, and combining with the possibility that the fault of the test sample is a transformer fault, respectively obtain the transformer fault significance of the test sample and the bus fault significance of the test sample, and the specific calculation formula is:

[0036]

[0037] In the formula, represents the transformer fault significance of the test sample; represents the bus fault significance of the test sample; represents the possibility that the fault of the test sample is a transformer fault; Represent the fault characteristics of the transformer of the test sample; Represent the fault characteristics of the busbar of the test sample.

[0038] Preferably, the method for constructing a four-dimensional sample space, obtaining the neighborhood samples and interference samples of each sample according to the distribution of the samples in the four-dimensional sample space, and further obtaining the fault significance weights of all samples specifically includes:

[0039] Construct a four-dimensional sample space with the transformer fault degree of the test sample, the busbar fault degree of the test sample, the transformer fault significance of the test sample, and the busbar fault significance of the test sample, and place all test samples into the four-dimensional sample space;

[0040] Furthermore, for any misclassified sample in the four-dimensional sample space; preset a number of neighborhood samples ; Samples in the four-dimensional sample space whose distance from the misclassified sample is less than are used as the neighborhood samples of the misclassified sample; samples in the neighborhood samples of the misclassified sample with different labels from the misclassified sample are used as the interference samples of the misclassified sample;

[0041] For any misclassified sample in the four-dimensional sample space, preset a ratio threshold of interference samples ; When the ratio of the interference samples to the neighborhood samples of the misclassified sample in terms of quantity is greater than , then use the least squares method to perform a linear fit on the samples in the neighborhood samples of the misclassified sample with the same label as the misclassified sample, and obtain the vector form of the fitted straight line as the local feature vector of the misclassified sample , and use the least squares method to perform a linear fit on all correctly classified samples in the four-dimensional sample space, and obtain the vector form of the fitted straight line as the overall feature vector of correctly classified samples ; According to and , obtain the difference in fault significance between the neighborhood samples of the misclassified sample and all test samples, and its specific calculation formula is:

[0042]

[0043] In the formula, represents the difference in fault significance between the neighborhood samples of the misclassified sample and all test samples; represents the overall feature vector of correctly classified samples; represents the local feature vector of the misclassified sample; represents the modulus function;

[0044] Obtain the mean difference in fault significance between the neighborhood samples of all misidentified samples and all test samples, and perform inverse proportional normalization on it. Use the normalized result as the fault significance weight.

[0045] Preferably, the specific method for testing the fault type when the substation fails includes:

[0046] Take the moment when the protection device in the current substation triggers a protection action as the moment to be tested. Take the period from the previous milliseconds before the moment to be tested to the moment to be tested as the period to be tested, and take the current data within the period to be tested as the sample to be tested;

[0047] Input the sample to be tested into the primary fault identification model to obtain the transformer fault degree and bus fault degree of the sample to be tested. According to the transformer fault degree and bus fault degree of the sample to be tested, combined with the fault significance weight, correct the transformer fault degree and bus fault degree of the sample to be tested. The specific calculation formula is:

[0048]

[0049] In the formula, represents the corrected transformer fault degree of the sample to be tested; represents the corrected bus fault degree of the sample to be tested; represents the transformer fault degree of the sample to be tested; represents the bus fault degree of the sample to be tested; represents the fault significance weight; represents the transformer fault significance of the sample to be tested; represents the bus fault significance of the sample to be tested;

[0050] After obtaining the corrected transformer fault degree and corrected bus fault degree of the sample to be tested, take the fault corresponding to the maximum value in the corrected transformer fault degree and corrected bus fault degree as the fault type that occurred in the substation.

[0051] Another embodiment of the present invention provides a fault identification and inspection system for a 220kV substation bus, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of any one of the above-mentioned fault identification and inspection methods for a 220kV substation bus.

[0052] The beneficial effects of the technical solution of the present invention are as follows: training a primary fault recognition model; obtaining the possibility that the fault in the sample is a transformer fault according to the changes in the current data in the bus and branch lines in the sample. Since after a bus fault occurs, the bus fault will affect the current in all branch lines under the bus; and when a transformer fails, the branch line connected to the transformer is first affected by the transformer fault, and then the bus is affected through the branch line. Finally, due to the change of the power balance and impedance distribution of the current system, the affected bus will redistribute the current in each branch line, thereby affecting other branch lines of the bus, so as to obtain the possibility that the fault in the sample is a transformer fault.

[0053] Obtain the response duration of each sample; input all test samples into the primary fault recognition model, and combine the response duration of each test sample to obtain the transformer and bus fault significance of the test samples; construct a four-dimensional sample space, and obtain the fault significance weight according to the distribution of samples in the four-dimensional sample space. Since the current in the branch line affected by the fault will affect other branch lines through the transformer, and it takes a certain amount of time for the current in the branch line to affect other branch lines through the transformer. When a transformer fails, there is a difference in the response time between the corresponding branch lines of the transformer and the response time between normal branch lines, and the fault characteristics of the transformer can be obtained based on this. Since the bus is the main line, when a bus fault occurs, the protection device in the substation will respond quickly, and the fault characteristics of the bus can be obtained based on this; according to the fault significance weight, check the fault type when the substation fails. Whenever the substation fails, input the current data of all branch lines within the local range at the fault moment into the primary fault recognition model. According to the test results of the primary fault recognition model, set a correction module by further analyzing the differences between transformer faults and bus faults, so as to accurately distinguish transformer faults and bus faults and identify the substation fault type. Brief Description of the Drawings

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0055] Figure 1 It is a step flow chart of a fault recognition and inspection method for a 220kV substation bus of the present invention;

[0056] Figure 2 It is an example diagram of the current curve in the bus when a bus fault occurs;

[0057] Figure 3 It is an example diagram of the current curve in the branch line when a bus fault occurs. Detailed implementation manners

[0058] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following will, in combination with the accompanying drawings and preferred embodiments, elaborate in detail on a fault identification and inspection method and system for a 220 kV substation bus proposed according to the present invention, including its specific implementation manners, structures, features, and effects. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0060] The following will specifically describe the specific solutions of a fault identification and inspection method and system for a 220 kV substation bus provided by the present invention with reference to the accompanying drawings.

[0061] Please refer to Figure 1 , which shows a flowchart of the steps of a fault identification and inspection method for a 220 kV substation bus provided by an embodiment of the present invention. The method includes the following steps:

[0062] Step S001: Obtain a number of training samples and a number of test samples from the substation, and train a primary fault identification model.

[0063] It should be noted that, as a fault identification and inspection method for a 220 kV substation bus, in this embodiment, the historical data when the substation fails is specifically used as the training sample to train the fault identification model, and the fault identification model is used to identify and inspect the faults in the bus; also, since it takes a certain amount of time to identify the fault when a fault occurs, the protection device in the substation will trigger a protection action, so the data collected in the period before the protection device triggers the protection action can be used as the training set to train the fault identification model.

[0064] Specifically, mark all the moments when the protection device in the substation triggers the protection action in the historical data as the protection moments, and preset a time range to be analyzed , The specific size of which can be set according to the actual situation by itself, and there is no strict high requirement in this embodiment. In this embodiment, it is described by taking as an example; for any protection moment, the time period from milliseconds before the protection moment to the protection moment is used as the time period to be analyzed for the protection moment; and the current data within the time period to be analyzed for each protection moment is used as the sample;

[0065] Further, all samples are randomly divided into training samples and test samples, and the ratio between the number of training samples and the number of test samples is , and the ratio is the preset ratio of the number of training samples to the number of test samples. and The specific values of can be set according to the actual situation by oneself, and there is no rigid requirement in this embodiment. In this embodiment, take , as an example for description;

[0066] Labels are assigned to all training samples according to the fault types of all training samples. For example, if the fault type of a certain training sample is bus fault, then the label of the training sample is bus fault; all training samples are input into the RNN (Recurrent Neural Network) model to train all training samples. The loss function used is the cross-entropy loss function. Since the specific training process of the RNN model is a well-known prior art, it will not be elaborated in this embodiment. A primary fault identification model is obtained.

[0067] It should be noted that when there is a bus fault and a transformer fault, the current data in the cable will change, that is, the obtained primary fault identification model cannot accurately distinguish between a bus fault and a transformer fault. Therefore, after obtaining the primary fault identification model and using the test samples to view the primary fault identification model, it is necessary to further analyze the differences between the transformer fault and the bus fault to set up a correction module, and through the primary fault identification model and the correction module, to accurately distinguish between the transformer fault and the bus fault.

[0068] Thus, a primary fault identification model is obtained.

[0069] Step S002: Obtain the affected factor sequences of the bus and the branch lines in the sample according to the changes in the current data in the bus and the branch lines in the sample; obtain the affected moments of the bus and the branch lines in the sample according to the affected factor sequences of the bus and the branch lines in the sample; obtain the possibility that the fault in the sample is a transformer fault according to the affected moments of the bus and the branch lines in the sample.

[0070] It should be noted that this step is to analyze the differences between transformer faults and bus faults in order to better set up the correction module in the subsequent steps; when a bus fault occurs, the bus fault will affect the current in all branches under the bus, that is, when a bus fault occurs, the current in the bus will change suddenly first, and then the current in all branches under the bus will change suddenly; when a transformer fault occurs, the branch connected to the transformer is first affected by the transformer fault, and then the bus is affected through the branch. Finally, due to the change of the power balance and impedance distribution of the current system, the affected bus will redistribute the current of each branch, thereby affecting other branches of the bus, such as Figure 2 、 Figure 3 As shown, the Figure 2 is an example diagram of the current curve in the bus when a bus fault occurs, Figure 3 is an example diagram of the current curve in the branch when a bus fault occurs. Therefore, the degree of influence on the bus can be obtained by the time sequence of the sudden change of the current in the bus and the branch.

[0071] Preferably, in a specific embodiment of the present invention, a local time range is preset , The specific value of can be set according to the actual situation by itself, and this embodiment does not have any requirements for influence. In this embodiment, is taken as an example for description; for any moment in any sample, the time period from the moment to milliseconds after the moment is used as the local range of the moment; the standard deviation of all current data in the local range of the moment in the bus is used as the influence factor of each moment in the bus (if the subsequent time of a certain moment in the sample is less than , the local time range of this moment is not obtained);

[0072] Obtain the influence factor of each moment in the bus within the sample, and sort the influence factors of each moment in chronological order to obtain the bus influence factor sequence. Similarly, obtain several branch influence factor sequences;

[0073] It should be noted that since the current in the cable is stable under normal circumstances, but when the protection device in the substation triggers a protection action, it indicates that a fault has occurred at this time, which means that a fault has occurred in the previous period of time at this moment; when a fault occurs, the current in the cable will change suddenly, making the current data unstable within the local time range. Therefore, the greater the standard deviation of the current data within the local time range, the more likely the cable is affected by the fault at this moment.

[0074] It should be further noted that the cable in this embodiment includes a bus and branch lines, and all the branch lines that appear in this embodiment are the branch lines corresponding to the bus that appears, and all the buses that appear are the buses corresponding to the branch lines that appear; since it takes a certain amount of time for the protection device to identify a fault when it occurs, the time period before the protection moment is recorded as the time period to be analyzed, and there is a moment when a fault occurs in the time period to be analyzed, so there are two parts in the time period to be analyzed, one part is the time period before the fault occurs, and the other part is the time period after the fault occurs; since the current data in the cable remains stable before the fault occurs; and after the fault occurs, the current data in the cable is affected by the fault, resulting in fluctuations in the current in the cable, so the moment when the cable is affected by the fault can be obtained based on this, and a bus and its corresponding several branch lines will be used as an example for the following description.

[0075] Preferably, in a specific embodiment of the present invention, for any affected factor in the bus affected factor sequence, taking the affected factor as a segmentation point, dividing the bus affected factor sequence into two affected factor segments, and based on the affected factors in the first and second affected factor segments, combining the bus affected factor sequence, the degree of influence of the bus at the moment corresponding to the affected factor is obtained, and its specific calculation formula is:

[0076]

[0077] In the formula, represents the degree of influence of the bus at the moment corresponding to the affected factor; represents the standard deviation of all affected factors in the bus affected factor sequence; represents the standard deviation of all affected factors in the first affected factor segment; represents the standard deviation of all affected factors in the second affected factor segment; represents a linear normalization function, and its normalization range is for all affected factors in the affected sequence ;

[0078] The degree of influence of the bus at all moments is obtained; similarly, the degree of influence of each branch line at all moments is obtained, and the moment with the maximum degree of influence of the bus within the sample is used as the affected moment of the bus; similarly, the affected moment of each branch line is obtained.

[0079] It should be noted that the affected moment of the cable represents the moment when the cable is affected by a fault; since the current data of the bus is stable before being affected by a fault and becomes unstable after being affected by a fault, in the time period to be analyzed, each "yes / no" is used as a segmentation point, and the time period to be analyzed is traversed and divided into two affected factor segments. If one of the two affected factor segments is more stable and the other is less stable, it indicates that the bus is more likely to be affected by a fault at this moment. Therefore, the moment with the greatest degree of influence is selected as the affected moment.

[0080] It should be further noted that when a fault occurs in the bus, the bus will simultaneously affect all the branches under it, that is, the affected moments of all the branches under the bus tend to be the same; when a transformer fails, it will first affect the branches connected to the transformer, then affect the bus through the branches, and finally affect other branches through the bus, that is, when a transformer fails, the affected moments of the branches do not tend to be the same, and some branches will be affected first. Therefore, the possibility of the fault being a transformer fault can be obtained based on this.

[0081] Preferably, in a specific embodiment of the present invention, for any sample, the distance in time series between the affected moment of the bus in the sample and the affected moment of each branch is used as the characteristic distance of each branch; the branch whose affected moment is before the bus is taken as an abnormal branch; the branch whose affected moment is after the bus is taken as a normal branch;

[0082] Furthermore, the branch with the largest characteristic distance among the abnormal branches is taken as the suspected fault branch. Based on the characteristic distance of the suspected fault branch and the characteristic distances of each normal branch, the possibility that the fault in the sample is a transformer fault is obtained. The specific calculation formula is:

[0083]

[0084] In the formula, represents the possibility that the fault in the sample is a transformer fault; represents the standard deviation of the characteristic distances of all normal branches; represents the mean value of the characteristic distances of all normal branches; represents the characteristic distance of the suspected fault branch.

[0085] It should be noted that the characteristic distance of the branch line is the distance between the affected time of the branch line and the affected time of the busbar. Therefore, when the standard deviation of the characteristic distance of the normal branch line is larger, it means that the time when the normal branch line is affected by the fault is less consistent, that is, the fault is more likely to be a transformer fault. At the same time, when the transformer fails, the branch line connected to the transformer will be affected first, and the branch line affected first will affect other branches through the busbar. Therefore, when the branch line first affected by the fault and the branch line affected subsequently have a larger difference in the time when they are affected by the fault, the fault is more likely to be a transformer fault.

[0086] At this point, the possibility that the fault in the sample is a transformer fault is obtained.

[0087] Step S003: Obtain the response time of each sample; input all test samples into the primary fault identification model, and obtain the fault characteristics of the transformer and bus of each test sample in combination with the response time of each test sample; obtain the significance of the transformer and bus faults of the test sample in combination with the possibility that the fault in the test sample is a transformer fault.

[0088] It should be noted that the possibility that the fault in the time period to be analyzed obtained in step S002 is a transformer fault is obtained only by analyzing the changes in current data in different branches. However, in actual situations, if a serious fault occurs in the bus, the protection device corresponding to the bus triggers the protection action, which in turn causes some branches to fail to trigger the corresponding protection action, making the affected time of the quality current obtained at this time have a large degree of randomness, that is, the affected time is inconsistent, which in turn interferes with the distinction between the bus fault and the transformer fault. Therefore, after the possibility that the fault in the time period to be analyzed is a transformer fault is obtained through step S002, further analysis is required.

[0089] It should be further explained that the current in the branch line affected by the fault will affect other branches through the transformer, and it takes a certain amount of time for the current in the branch line to affect other branches through the transformer. When the transformer fails, the response time between the transformer corresponding branches is different from the response time between normal branches. This can be used as a basis to obtain the fault characteristics of the transformer. Since the busbar is the main line, the protection device in the substation will respond quickly when the busbar fails, and this can be used as a basis to obtain the fault characteristics of the busbar.

[0090] Preferably, in a specific embodiment of the present invention, for any branch line in any sample, a branch line connected to the same transformer as the branch line is recorded as a corresponding branch line of the branch line; a time sequence distance between the branch line and the corresponding branch line of the branch line at the affected moment is taken as the response time length of the branch line;

[0091] Further, input all test samples into the primary fault identification model to obtain the transformer fault degree and bus fault degree of each test sample, and obtain correctly identified samples and misidentified samples;

[0092] For any test sample, take the average response duration of all normal branches in the test sample as the response duration of the test sample; take the average response duration of all normal branches in all correctly identified samples as the reference response duration;

[0093] Obtain the fault characteristics of the transformer of the test sample according to the response duration of the test sample and the reference response duration; the specific calculation formula is:

[0094]

[0095] In the formula, represents the fault characteristics of the transformer of the test sample; represents the response duration of the test sample; represents the reference response duration; represents the absolute value operation; represents the linear normalization function, and its specific normalization range is that of all test samples .

[0096] Further, obtain the fault characteristics of the bus of the test sample according to the affected time of the bus of the test sample and the protection time of the test sample; the specific calculation formula is:

[0097]

[0098] In the formula, represents the fault characteristics of the bus of the test sample; represents the protection time of the test sample; represents the affected time of the bus of the test sample; represents the exponential function with the natural constant as the base; in this embodiment, the model is used to present the inverse proportional relationship and normalization processing, is the input of the model, and the implementer can set the inverse proportional function and normalization function according to the actual situation.

[0099] It should be noted that represents the difference between the response duration of the test sample and the reference response duration. Since when a transformer fails, there is a difference in the response time between the corresponding branches of the transformer and the response time between normal branches, so the larger the value of, the more fault characteristics the transformer of the test sample has. It represents the timing distance between the affected moment of the busbar and the protection moment. The smaller its value, the faster the protection device triggers the protection action, indicating that the busbar of the test sample has more fault characteristics. Further, in combination with the possibility that the fault in the test sample is a transformer fault, the transformer fault significance and the busbar fault significance of the test sample are obtained.

[0100] Preferably, in a specific embodiment of the present invention, for any test sample, according to the fault characteristics of the transformer and the busbar of the test sample, and in combination with the possibility that the fault in the test sample is a transformer fault, the transformer fault significance of the test sample and the busbar fault significance of the test sample are respectively obtained. The specific calculation formula is as follows:

[0101]

[0102] In the formula, represents the transformer fault significance of the test sample; represents the busbar fault significance of the test sample; represents the possibility that the fault in the test sample is a transformer fault; represents the fault characteristics of the transformer of the test sample; represents the fault characteristics of the busbar of the test sample.

[0103] It should be noted that the transformer fault significance and the busbar fault significance represent the possibility of the fault being a transformer fault and a busbar fault obtained by the primary fault identification model and the difference in the affected moments through different branches. When the possibility that the fault in the test sample is a transformer fault is greater and the fault characteristics of the transformer of the test sample are greater, the test sample is more likely to be a transformer fault; when the possibility that the fault in the test sample is a transformer fault is smaller and the fault characteristics of the busbar of the test sample are greater, the test sample is more likely to be a busbar fault.

[0104] Thus, the transformer fault significance and the busbar fault significance of the test sample are obtained.

[0105] Step S004: According to the transformer and busbar fault significances of the test sample and the possibility that the fault in the test sample is a transformer fault, a four-dimensional sample space is constructed. According to the distribution of the samples in the four-dimensional sample space, the neighborhood samples and interference samples of each sample are obtained, and then the fault significance weights of all samples are obtained, and the fault type when the substation fails is detected.

[0106] It should be noted that the transformer fault significance and bus fault significance obtained in step S003 are characteristics obtained under the ideal condition that a serious fault has occurred on the bus. Therefore, it is necessary to further set up a correction module by analyzing the test results of the primary fault identification model for each test sample. Whenever a fault occurs in the substation, the current data of all branches within the local range at the fault moment are input into the primary fault identification model. According to the test results of the primary fault identification model and combining the transformer fault significance and bus fault significance of all test samples, the substation fault type is identified.

[0107] Preferably, in a specific embodiment of the present invention, a four-dimensional sample space is constructed with the transformer fault degree of the test sample, the bus fault degree of the test sample, the transformer fault significance of the test sample, and the bus fault significance of the test sample, and all test samples are placed in the four-dimensional sample space.

[0108] Furthermore, for any misidentified sample in the four-dimensional sample space, a preset number of neighborhood samples , The specific size of which can be set according to the actual situation and is not strictly required in this embodiment. In this embodiment, it is described by taking as an example; the samples in the four-dimensional sample space whose distance from the misidentified sample is less than are used as the neighborhood samples of the misidentified sample; the samples in the neighborhood samples of the misidentified sample with different labels from the misidentified sample are used as the interference samples of the misidentified sample.

[0109] For any misidentified sample in the four-dimensional sample space, a preset interference sample ratio threshold , The specific size of which can be set according to the actual situation and is not strictly required in this embodiment. In this embodiment, it is described by taking as an example; when the ratio of the number of interference samples to neighborhood samples of the misidentified sample is less than or equal to , the transformer fault degree and bus fault degree of the misidentified sample are respectively equal to the transformer fault significance and bus fault significance of the misidentified sample.

[0110] It should be noted that if the labels of samples are the same, it indicates that the current data in the branch lines and busbars of the samples are similar. Therefore, the more other misidentified samples exist around a misidentified sample, the more interference the model will encounter when verifying and identifying this sample, resulting in incorrect judgment. Thus, when the ratio of the interfering samples to the neighboring samples of the misidentified sample is smaller, it shows that the primary fault identification model does not receive interference from other samples when evaluating this test sample, that is, the primary fault identification model cannot accurately identify the fault type of this sample without being interfered by other samples. Therefore, directly use the transformer fault significance and busbar fault significance of this sample as the transformer fault degree and busbar fault degree of the misidentified sample.

[0111] Furthermore, when the ratio of the interfering samples to the neighboring samples of the misidentified sample is greater than a straight line fitting is performed on the samples with the same label as the misidentified sample among the neighboring samples of the misidentified sample by using the least squares method, and the vector form of the fitting straight line is denoted as the local feature vector of the misidentified sample , and a straight line fitting is performed on all correctly identified samples in the four-dimensional sample space by using the least squares method, and the vector form of the fitting straight line is denoted as the overall feature vector of the correctly identified samples ; According to and , the difference in fault significance between the neighboring samples of the misidentified sample and all test samples is obtained, and its specific calculation formula is:

[0112]

[0113] In the formula, represents the difference in fault significance between the neighboring samples of the misidentified sample and all test samples; represents the overall feature vector of the correctly identified samples; represents the local feature vector of the misidentified sample; represents the modulus function.

[0114] Furthermore, obtain the mean value of the differences in fault significance between the neighboring samples of all misidentified samples and all test samples, and perform inverse proportional normalization on it, and use the normalization result as the fault significance weight of all samples; the specific process of inverse proportional normalization in this embodiment is: 1 minus the linear normalization result of the difference mean value.

[0115] It should be noted that; the more the number of other misrecognized samples around a misrecognized sample, the more the model will be interfered when testing and recognizing this sample, resulting in incorrect judgment. That is, when the number of other misrecognized samples around a misrecognized sample is larger, it indicates that the primary fault recognition model could originally recognize and test the fault type of this sample, but was misrecognized due to interference from other samples. Therefore, the overall feature vector of the correctly recognized sample , and the local feature vector of the misrecognized sample The smaller the difference between them, the more it indicates that the performance of the model on local samples is basically the same as that on global samples, and the accuracy of the fault significance degree feature is relatively high. Therefore, the negative correlation normalization result of the difference mean value of the neighborhood samples of all misrecognized samples and all test samples in terms of fault significance is used as the significance weight.

[0116] It should be further noted that after obtaining the fault significance weight and the primary fault recognition model, the fault type when a fault occurs in the current substation can be accurately recognized according to the fault significance weight and the primary fault recognition model.

[0117] Preferably, in a specific embodiment of the present invention, the moment when the protection device in the current substation triggers a protection action is used as the moment to be tested, and the time from milliseconds before the moment to be tested to the moment to be tested is used as the time period to be tested, and the current data within the time period to be tested is used as the sample to be tested;

[0118] The sample to be tested is input into the primary fault recognition model to obtain the transformer fault degree and bus fault degree of the sample to be tested. According to the transformer fault degree and bus fault degree of the sample to be tested, combined with the fault significance weight, the transformer fault degree and bus fault degree of the sample to be tested are corrected. The specific calculation formula is:

[0119]

[0120] In the formula, represents the corrected transformer fault degree of the sample to be tested; represents the corrected bus fault degree of the sample to be tested; represents the transformer fault degree of the sample to be tested; represents the bus fault degree of the sample to be tested; represents the fault significance weight; represents the transformer fault significance of the sample to be tested; represents the bus fault significance of the sample to be tested.

[0121] After obtaining the transformer fault degree of the corrected sample to be tested and the bus fault degree of the corrected sample to be tested, the fault with the larger corrected fault degree is taken as the fault type that occurs in the substation.

[0122] Another embodiment of the present invention provides a fault identification and inspection system for a 220 kV substation bus, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements a fault identification and inspection method for a 220 kV substation bus in steps S001 to S004.

[0123] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A fault identification and inspection method for a 220kV substation busbar, characterized in that: The method comprises the following steps: Obtain a number of training samples and a number of test samples through the substation, and train a primary fault recognition model; According to the changes in the current data of the busbar and branch in the sample, the affected factor sequence of the busbar and branch in the sample is obtained; according to the affected factor sequence of the busbar and branch in the sample, the affected time of the busbar and branch in the sample is obtained; according to the affected time of the busbar and branch in the sample, the possibility that the fault in the sample is a transformer fault is obtained; Obtain the response time of each sample; input all test samples into the primary fault identification model, and obtain the fault characteristics of the transformer and busbar of each test sample in combination with the response time of each test sample; obtain the significance of the transformer and busbar faults of the test sample in combination with the possibility that the fault in the test sample is a transformer fault; According to the significance of transformer and busbar faults of the test samples and the possibility that the faults in the test samples are transformer faults, a four-dimensional sample space is constructed. According to the distribution of samples in the four-dimensional sample space, the neighborhood samples and interference samples of each sample are obtained, and then the fault significance weights of all samples are obtained, and the fault type when the substation fails is tested; The specific method of obtaining the response time of each sample includes: For any branch in any sample, the branch connected to the same transformer as the branch is recorded as the corresponding branch of the branch; the timing distance between the branch and the corresponding branch of the branch at the affected moment is taken as the response duration of the branch.

2. A 220kV substation bus fault identification and inspection method according to claim 1, characterized in that: The method of obtaining the sequence of factors affecting the busbars and branches in the sample according to the changes in the current data in the busbars and branches in the sample includes: Preset a local time range ; For any time in any sample, convert the time to the time after The time period of milliseconds is taken as the local range of the moment; the standard deviation of all current data in the local range of the moment in the bus is taken as the affected factor at each moment in the bus; The affected factors at each moment in the bus within the sample are obtained, and the affected factors at each moment are sorted in chronological order to obtain a bus affected factor sequence, and similarly, several branch affected factor sequences are obtained.

3. A 220kV substation bus fault identification and inspection method according to claim 1, characterized in that: The specific method of obtaining the affected moments of the busbars and branches in the sample according to the affected factor sequence of the busbars and branches in the sample is as follows: For any affected factor in the bus affected factor sequence, the affected factor is used as a segmentation point to divide the bus affected factor sequence into two affected factor segments. According to the affected factors in the first and second affected factor segments, combined with the bus affected factor sequence, the affected degree of the bus at the time corresponding to the affected factor is obtained. The specific calculation formula is: In the formula, Indicates the degree of influence of the busbar at the corresponding moment of the affected factor; Represents the standard deviation of all affected factors in the affected factor sequence of the bus; represents the standard deviation of all affected factors in the first affected factor segment; represents the standard deviation of all affected factors in the second affected factor segment; represents the linear normalization function; The affected degree of the busbar at all times is obtained; similarly, the affected degree of each branch line at all times is obtained, and the moment when the affected degree of the busbar in the sample is the largest is taken as the affected moment of the busbar; similarly, the affected moment of each branch line is obtained.

4. A 220kV substation bus fault identification and inspection method according to claim 1, characterized in that: The method of obtaining the possibility that the fault in the sample is a transformer fault according to the affected moments of the busbar and the branch line in the sample includes: For any sample, the time series distance between the affected time of the main line and the affected time of each branch line in the sample is used as the characteristic distance of each branch line; and the branch line whose affected time is before the main line is used as an abnormal branch line; and the branch line whose affected time is after the main line is used as a normal branch line; The branch with the largest characteristic distance among the abnormal branches is taken as the suspected fault branch. According to the characteristic distance of the suspected fault branch and the characteristic distance of each normal branch, the possibility that the fault in the sample is a transformer fault is obtained. The specific calculation formula is: In the formula, Indicates the possibility that the fault in the sample is a transformer fault; represents the standard deviation of the characteristic distances of all normal branches; represents the mean of the characteristic distances of all normal branches; Indicates the characteristic distance of the suspected fault branch line.

5. A 220kV substation bus fault identification and inspection method according to claim 1, characterized in that: The specific method of inputting all test samples into the primary fault identification model and obtaining the fault characteristics of the transformer and busbar of each test sample in combination with the response time of each test sample is as follows: Input all test samples into the primary fault identification model, obtain the transformer fault degree and busbar fault degree of each test sample, and obtain the correctly identified samples and the incorrectly identified samples; For any test sample, the average response time of all normal branches in the test sample is used as the response time of the test sample; The average response time of all normal branches in all correctly identified samples is taken as the benchmark response time; Acquire a fault feature of a transformer of the test sample according to a response time of the test sample and a reference response time; The specific calculation formula is: In the formula, representing a fault characteristic of the transformer of the test sample; Indicates the response time of the test sample; Indicates the baseline response time; Represents absolute value operation; represents the linear normalization function; According to the affected time of the busbar of the test sample and the protection time of the test sample, the fault characteristics of the busbar of the test sample are obtained, and the protection time is the time when all protection devices in the historical data of the substation trigger the protection action; the specific calculation formula is: In the formula, Indicates the fault characteristics of the busbar of the test sample; Indicates the protection moment of the test sample; represents the affected moment of the busbar of the test sample; Represents an exponential function with a natural constant as base.

6. A 220kV substation bus fault identification and inspection method according to claim 1, characterized in that: The specific method of obtaining the transformer and bus fault significance of the test sample includes: For any test sample, according to the fault characteristics of the transformer and the bus of the test sample, combined with the possibility that the fault of the test sample is a transformer fault, the transformer fault significance of the test sample and the bus fault significance of the test sample are obtained respectively, and the specific calculation formula is: In the formula, Indicates the transformer fault significance of the test sample; Indicates the busbar fault significance of the test sample; Indicates the possibility that the fault of the test sample is a transformer fault; representing a fault characteristic of the transformer of the test sample; Indicates the fault characteristics of the busbar of the test sample.

7. A 220kV substation bus fault identification and inspection method according to claim 5, characterized in that: The four-dimensional sample space is constructed, and according to the distribution of samples in the four-dimensional sample space, the neighborhood samples and interference samples of each sample are obtained, and then the fault significance weights of all samples are obtained, including the specific method of: A four-dimensional sample space is constructed based on the transformer fault degree of the test sample, the bus fault degree of the test sample, the transformer fault significance of the test sample, and the bus fault significance of the test sample, and all the test samples are placed in the four-dimensional sample space; Furthermore, for any recognition error sample in the four-dimensional sample space; Preset a neighborhood sample size ; The distance between the sample in the four-dimensional sample space and the recognition error sample is less than as the neighborhood samples of the misidentified samples; and using the samples in the neighborhood samples of the misidentified samples that have different labels from the misidentified samples as the interference samples of the misidentified samples; For any misidentified sample in the four-dimensional sample space, a threshold of interference sample ratio is preset. ; When the ratio of the interference samples of the identified error samples to the neighboring samples is greater than When , the least squares method is used to perform straight line fitting on the samples in the neighborhood of the misidentified sample with the same label as the misidentified sample, and the vector formula of the fitted straight line is recorded as the local feature vector of the misidentified sample. , and use the least squares method to perform straight line fitting on all correctly identified samples in the four-dimensional sample space, and obtain the vector formula of the fitting line, which is recorded as the overall feature vector of the correctly identified samples ;according to and , obtain the difference in fault significance between the neighborhood samples of the identified error samples and all test samples, and the specific calculation formula is: In the formula, Indicates the difference in fault significance between the neighborhood samples of the identified error sample and all test samples; Represents the overall feature vector of the correct sample identified; A local feature vector representing the recognition error sample; represents the modulo function; The mean difference in fault significance between the neighborhood samples of all misidentified samples and all test samples is obtained, and an inverse proportional normalization process is performed on them. The normalized result is used as the fault significance weight.

8. A 220kV substation bus fault identification and inspection method according to claim 1, characterized in that: The specific method of checking the fault type when a fault occurs in the substation includes: The time when the protection device in the current substation triggers the protection action is taken as the time to be tested, and the time before the time to be tested is taken as the time before the time to be tested. milliseconds to the time to be tested is used as the time period to be tested, and the current data in the time period to be tested is used as the sample to be tested; The samples to be tested are input into the primary fault identification model to obtain the transformer fault degree and bus fault degree of the samples to be tested. According to the transformer fault degree and bus fault degree of the samples to be tested and combined with the fault significance weight, the transformer fault degree and bus fault degree of the samples to be tested are corrected. The specific calculation formula is: In the formula, Indicates the corrected transformer fault degree of the sample to be tested; Indicates the corrected busbar fault degree of the sample to be tested; Indicates the transformer fault degree of the sample to be tested; Indicates the busbar fault degree of the sample to be tested; represents the fault significance weight; Indicates the transformer fault significance of the sample to be tested; Indicates the significance of busbar fault of the sample to be tested; represents the linear normalization function; After obtaining the corrected transformer fault degree of the sample to be tested and the corrected bus fault degree of the sample to be tested, the fault corresponding to the maximum value of the corrected transformer fault degree and the corrected bus fault degree is taken as the fault type occurring in the substation.

9. A 220 kV substation bus fault identification and inspection system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of a fault identification and inspection method for a 220kV substation busbar are implemented as described in any one of claims 1 to 8.

Images