A switch cabinet partial discharge diagnosis method and system

Abstract

The present application relates to the field of partial discharge diagnosis technology, and specifically to a switchgear partial discharge diagnosis method and system, the method comprising: collecting ultrasonic signals of high-voltage switchgear; dividing the ultrasonic signals into ultrasonic signal segments with a length of one power frequency cycle, and determining the waveform feature vector of each ultrasonic signal segment; determining the first correlation coefficient and the second correlation coefficient of each ultrasonic signal segment; determining the partial discharge feature vector of each ultrasonic signal segment based on the waveform feature vector, the first correlation coefficient, the second correlation coefficient, and the peak value of each ultrasonic signal segment; and diagnosing the partial discharge fault type of the high-voltage switchgear by multi-information fusion based on the ultra-high frequency detection method and the partial discharge feature vector. The present application aims to improve the accuracy of partial discharge fault diagnosis of high-voltage switchgear.

Description

Technical Field

[0001] The present application relates to the technical field of partial discharge diagnosis, and in particular to a method and system for partial discharge diagnosis of a switch cabinet. Background Art

[0002] High-voltage switchgear is a key equipment in the power supply and distribution system. It is not only responsible for controlling and protecting the operation of the power system, but also directly affects the reliability and safety of the power supply. If the high-voltage switchgear fails, it will directly affect the reliability of the power supply. Under the action of operating voltage, current, heat, force, etc., as well as during production, transportation, installation and maintenance, high-voltage switchgear will inevitably produce or leave various insulation defects. These defects gradually develop and cause the electrical strength of the high-voltage switchgear to decrease, leading to failures. Partial discharge diagnosis is an important technical means to prevent sudden power outages caused by insulation failures in high-voltage switchgear. It can improve the reliability of high-voltage switchgear operation and reduce the number of power outages for maintenance and the occurrence of power outage failures.

[0003] With the development of detection technology, the fusion of multi-detection information has become an inevitable trend in the development of fault detection technology. The fusion of multiple detection methods can realize the fault diagnosis of partial discharge of high-voltage switchgear. To accurately diagnose whether there is a partial discharge fault inside the high-voltage switchgear, it is necessary to accurately extract the features of the detection information. At present, neural networks are usually used to directly extract features from detection signals. The performance of neural networks is highly dependent on the quality and diversity of training data. If the training data is insufficient or biased, the trained model may have insufficient generalization ability, resulting in inaccurate feature extraction of detection information and reducing the accuracy of partial discharge fault diagnosis of high-voltage switchgear. Summary of the invention

[0004] In view of the above, it is necessary to provide a switch cabinet partial discharge diagnosis method and system, which improves the accuracy of switch cabinet partial discharge fault diagnosis compared with the traditional switch cabinet partial discharge diagnosis method and system:

[0005] In a first aspect, an embodiment of the present application provides a switch cabinet partial discharge diagnosis method, the method comprising the following steps:

[0006] Collect ultrasonic signals from high-voltage switchgear;

[0007] The ultrasonic signal is divided into ultrasonic signal segments with a length of one power frequency cycle, and a waveform feature vector of each ultrasonic signal segment is determined based on the peak shape of the envelope of each ultrasonic signal segment;

[0008] Determine a first correlation coefficient of any one ultrasonic signal segment based on similarities in distribution characteristics between any one ultrasonic signal segment and all other ultrasonic signal segments;

[0009] Divide each ultrasonic signal segment equally into two ultrasonic signal small segments according to time, form the ultrasonic signal small segments of all ultrasonic signal segments into an ultrasonic signal small segment set, and determine a second correlation coefficient of each ultrasonic signal segment based on the similarity of distribution characteristics between each ultrasonic signal small segment of each ultrasonic signal segment and each ultrasonic signal small segment in the ultrasonic signal small segment set;

[0010] Comprehensively analyzing the waveform characteristic vector, the first correlation coefficient, the second correlation coefficient, and the peak value of each ultrasonic signal segment to determine the partial discharge characteristic vector of each ultrasonic signal segment;

[0011] Based on the ultra-high frequency detection method and the partial discharge feature vector, multi-information fusion is performed to diagnose the partial discharge fault type of the high-voltage switchgear.

[0012] In one embodiment, the process of determining the waveform feature vector is as follows:

[0013] Extracting the envelope of each ultrasonic signal segment, and determining the shape feature vector of the peak where each peak point in each envelope is located based on the peak width and peak height of the peak where each peak point in each envelope is located;

[0014] Based on the shape feature vectors of all peaks of the envelope of each ultrasonic signal segment, a waveform feature vector of each ultrasonic signal segment is determined.

[0015] In one embodiment, the process of determining the shape feature vector is as follows:

[0016] For each peak point in each envelope, obtain the left peak width, right peak width and peak height of the peak where each peak point is located;

[0017] The calculation method of the left peak width of any peak is as follows: obtain the data point with the smallest ordinate on the left side of the peak point of any peak, record it as the minimum data point on the left side, and take the absolute value of the difference between the peak point of any peak and the minimum data point on the left side as the left peak width of any peak; the calculation method of the right peak width is the same as the calculation method of the left peak width;

[0018] The ratios of the left peak width and the right peak width to the peak height of the peak where any peak point is located are respectively used as the first component and the second component of the shape feature vector of the peak where any peak point is located.

[0019] In one embodiment, the mean of the components at the same position of all shape feature vectors corresponding to each ultrasonic signal segment is used as each component in the waveform feature vector of each ultrasonic signal segment.

[0020] In one embodiment, the process of determining the first correlation coefficient is:

[0021] All ultrasonic signal segments are combined into an ultrasonic signal segment set;

[0022] Calculate the cross-correlation coefficient between any one ultrasonic signal segment in the ultrasonic signal segment set and the remaining ultrasonic signal segments at each time delay by using a cross-correlation function, and record it as a first cross-correlation coefficient;

[0023] Recording the average of the first cross-correlation coefficients between any one ultrasonic signal segment in the ultrasonic signal segment set and the remaining ultrasonic signal segments at all time delays as a first cross-correlation average;

[0024] The average of the first cross-correlation averages between the any one ultrasonic signal segment in the ultrasonic signal segment set and all other ultrasonic signal segments is used as the first correlation coefficient of the any one ultrasonic signal segment.

[0025] In one embodiment, the second correlation coefficient is determined as follows:

[0026] A cross-correlation function is used to calculate the cross-correlation coefficient between each ultrasonic signal small segment in the ultrasonic signal small segment set and the remaining ultrasonic signal small segments at each time delay, which is recorded as a second cross-correlation coefficient; an average of the second cross-correlation coefficients between each ultrasonic signal small segment in the ultrasonic signal small segment set and the remaining ultrasonic signal small segments at all time delays is recorded as a second cross-correlation average;

[0027] Calculate the average of the second cross-correlation mean values ​​between each ultrasonic signal small segment in the ultrasonic signal small segment set and all other ultrasonic signal small segments, and record it as the third cross-correlation mean value;

[0028] The average of the third cross-correlation averages of the two ultrasonic signal small segments of each ultrasonic signal segment is used as the second correlation coefficient of each ultrasonic signal segment.

[0029] In one embodiment, the local discharge characteristic vector is composed of the maximum peak value, waveform characteristic vector and period characteristic vector of the envelope of each ultrasonic signal segment;

[0030] The periodic eigenvector of each ultrasonic signal segment is composed of a first correlation coefficient and a second correlation coefficient of each ultrasonic signal segment.

[0031] In one embodiment, the process of diagnosing the partial discharge fault type of the high-voltage switchgear is:

[0032] The partial discharge feature vectors of all ultrasonic signal segments are combined into a partial discharge feature set of ultrasonic information of the high-voltage switchgear;

[0033] The UHF detection method and principal component analysis method are used to obtain the optimal feature subset of the electromagnetic wave information of the high-voltage switchgear;

[0034] Neural network is used to identify and classify the partial discharge feature set and the optimal feature subset respectively;

[0035] Based on the basic probability distribution, the classification results of the partial discharge feature set and the classification results of the optimal feature subset are respectively corrected;

[0036] The information corrected by the basic probability allocation BPA is fused through the DS evidence theory to obtain the probability of each partial discharge fault.

[0037] Diagnose whether a partial discharge fault occurs in the high-voltage switchgear based on the probability of all types of partial discharge faults.

[0038] In one embodiment, the partial discharge fault with the highest probability is used as the fault type of the high-voltage switchgear. When the probabilities of all types of partial discharge faults are equal or the probabilities of all types of partial discharge faults are less than a preset value, it is determined that the high-voltage switchgear has no fault.

[0039] In a second aspect, an embodiment of the present application further provides a switch cabinet partial discharge diagnostic system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of any one of the above-mentioned switch cabinet partial discharge diagnostic methods when executing the computer program.

[0040] This application has at least the following beneficial effects:

[0041] The present application analyzes four types of partial discharge faults of the high-voltage switchgear, namely, needle-plate discharge, internal discharge, suspended discharge, and surface discharge, and determines the partial discharge feature vector of the ultrasonic signal in each power frequency cycle based on the peak characteristics, waveform characteristics, and periodic characteristics of the ultrasonic signal of the high-voltage switchgear. The application considers the differences in the maximum peak value, signal waveform, and waveform periodicity in each power frequency cycle between the ultrasonic signals generated by the four types of partial discharge faults of the high-voltage switchgear, namely, needle-plate discharge, internal discharge, suspended discharge, and surface discharge, and improves the discrimination between various partial discharge faults of the high-voltage switchgear.

[0042] Furthermore, the partial discharge feature set of the ultrasonic information of the high-voltage switchgear is obtained based on the partial discharge feature vector, and combined with the optimal feature subset of the electromagnetic wave information of the high-voltage switchgear obtained by the ultra-high frequency detection method, the DS evidence theory is used to fuse multiple detection information to output the probability of each partial discharge fault, and diagnose whether the high-voltage switchgear has a fault. This improves the accuracy of the partial discharge feature set for obtaining the ultrasonic information of the high-voltage switchgear, thereby improving the accuracy of the partial discharge fault diagnosis of the high-voltage switchgear. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are 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 creative work.

[0044] Figure 1 A flowchart of a method for diagnosing partial discharge of a switch cabinet provided in one embodiment of the present application;

[0045] Figure 2 Schematic diagram of the process of determining the waveform characteristic vector;

[0046] Figure 3 This is a schematic diagram of partial discharge fault diagnosis of high-voltage switchgear based on multi-information fusion;

[0047] Figure 4 Figure 2 is a schematic diagram of the diagnostic flow of partial discharge fault. DETAILED DESCRIPTION

[0048] In the description of the embodiments of the present application, words such as "exemplary", "or", "for example" and the like are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "or", "for example" and the like is intended to present related concepts in a concrete manner.

[0049] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. It should be understood that, unless otherwise specified, " / " means or.

[0050] It should also be noted that the terms "first" and "second" in the present application are used to distinguish similar objects rather than to describe a specific order or sequence.

[0051] The specific scheme of a switch cabinet partial discharge diagnosis method and system provided by the present application is described in detail below with reference to the accompanying drawings.

[0052] See also Figure 1 , which shows a flow chart of the steps of a switch cabinet partial discharge diagnosis method provided by an embodiment of the present application, the method comprising the following steps:

[0053] Step 1: Collect ultrasonic signals from high-voltage switchgear.

[0054] Ultrasonic signals of the high-voltage switch cabinet are collected using an ultrasonic sensor. In this embodiment, the ultrasonic signals are collected at the gaps of the metal panel of the high-voltage switch cabinet. As other implementation methods, the ultrasonic signals can be collected at discontinuous parts such as shutters of the metal panel of the high-voltage switch cabinet.

[0055] In this embodiment, the sampling time of the ultrasonic signal is 1 second, the center frequency and sensitivity of the ultrasonic sensor are 40KHz and -65dB respectively. The values ​​of the sampling time, center frequency and sensitivity are preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0056] Since the signals collected by the ultrasonic sensor are usually weak, in order to improve the signal-to-noise ratio of the collected ultrasonic signal, a gain amplifier with amplification and filtering functions is used to amplify and filter the collected ultrasonic signal.

[0057] In this embodiment, the amplification factor of the ultrasonic signal is 60 times. The value of the amplification factor is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0058] Step 2: By analyzing the four types of partial discharge faults of the high-voltage switchgear, namely needle-plate discharge, internal discharge, suspended discharge, and surface discharge, a partial discharge feature vector is constructed based on the maximum peak value, waveform feature vector, and period feature vector of the ultrasonic signal segment of the high-voltage switchgear.

[0059] Partial discharge faults of high-voltage switchgear generally have multiple manifestations, such as needle-plate discharge, internal discharge, suspended discharge, and surface discharge. Due to the differences in discharge positions, dielectric properties, and signal propagation paths, the ultrasonic signals generated by each partial discharge fault of the high-voltage switchgear have obvious differences in the maximum peak value, signal waveform, and periodicity of the waveform in each power frequency cycle. The power frequency is the power grid operating frequency of 50Hz.

[0060] The ultrasonic signal generated by the needle plate discharge fault is strong, and the waveform of the signal is a trumpet-shaped signal that rises rapidly first and then falls slowly, and the trumpet-shaped waveform of the signal appears once in each power frequency cycle, indicating that the signal has a larger peak value and a stronger 50Hz correlation in each power frequency cycle but does not have a strong 100Hz correlation; the ultrasonic signal generated by the internal discharge fault is weak, and the waveform of the signal is a bell-shaped signal that rises slowly first and then falls slowly, and the bell-shaped waveform of the signal appears twice in each power frequency, indicating that the signal has a smaller peak value and a stronger 100Hz correlation in each power frequency cycle but does not have a strong 50Hz correlation; the ultrasonic signal generated by the suspended discharge fault is strong, and the waveform of the signal is a trumpet-shaped signal, and it appears twice in each power frequency; the ultrasonic signal generated by the surface discharge fault is strong, and the waveform of the signal is a bell-shaped signal, and it appears once in each power frequency.

[0061] Step 2.1, dividing the ultrasonic signal into ultrasonic signal segments with a length of one power frequency cycle, and determining the waveform feature vector of each ultrasonic signal segment based on the peak shape of the envelope of each ultrasonic signal segment.

[0062] Based on the above analysis, the ultrasonic signal is segmented in the time domain to obtain ultrasonic signal segments of the ultrasonic signal. The segmentation length is set to one power frequency cycle, and all ultrasonic signal segments are combined into an ultrasonic signal segment set for subsequent acquisition of the signal characteristics of the ultrasonic signal in each power frequency cycle, including: maximum amplitude, signal waveform and 50Hz correlation. The time domain segmentation of the signal is a well-known technology and will not be repeated in this application.

[0063] The envelope of each ultrasonic signal segment is extracted for subsequent acquisition of the maximum peak value and waveform characteristics of each ultrasonic signal segment, wherein the extraction of the envelope is a well-known technology and will not be described in detail in this application.

[0064] For each peak point in each envelope line, the left peak width, right peak width and peak height of the peak where each peak point is located are obtained, wherein the left peak width of any peak is calculated as follows: obtain the data point with the smallest vertical coordinate on the left side of its peak point on any peak, record it as the smallest data point on the left, and use the absolute value of the difference between the peak point of any peak and the smallest data point on the left as the left peak width of any peak; the right peak width of any peak is calculated as follows: obtain the data point with the smallest vertical coordinate on the right side of its peak point on any peak, record it as the smallest data point on the right, and use the absolute value of the difference between the peak point of any peak and the smallest data point on the right as the right peak width of any peak.

[0065] The shape characteristic vector of the peak at any peak point Q is k=( , ), where W1, W2 and H are the left peak width, right peak width and peak height of the peak where the peak point Q is located, respectively.

[0066] It should be noted that: if the left side of the peak where the peak point Q is located is steeper, that is, The smaller it is, the flatter the right side of the peak where the peak point Q is located is, that is, The larger the value is, the more likely the peak where the peak point Q is located is to be a trumpet-shaped peak that rises rapidly first and then slowly falls. If the left side of the peak where the peak point Q is located is flatter, that is, The larger it is, the flatter the right side of the peak where the peak point Q is located is, that is, The larger it is, the more likely the shape of the peak where the peak point Q is located is to be a bell-shaped peak that rises slowly first and then falls slowly.

[0067] Further, the waveform feature vector K=(a1, a2) of any ultrasonic signal segment B is obtained, where a1 is the mean value of the first component of the shape feature vector of all peaks of the envelope of the ultrasonic signal segment B, and a2 is the mean value of the second component of the shape feature vector of all peaks of the envelope of the ultrasonic signal segment B. The shape feature vector k=( , ), the first and second components are , The schematic diagram of the process of determining the waveform characteristic vector is as follows: Figure 2 shown.

[0068] It should be noted that: if a1 is smaller and a2 is larger, the waveform of the ultrasonic signal segment B is more likely to be a trumpet shape that rises rapidly first and then falls slowly, and the ultrasonic signal segment is more likely to be an ultrasonic signal generated by a needle plate discharge fault or a suspended discharge fault; if a1 is larger and a2 is larger, the waveform of the ultrasonic signal segment B is more likely to be a bell shape that rises slowly first and then falls slowly, and the ultrasonic signal segment is more likely to be an ultrasonic signal generated by an internal discharge fault or a surface discharge fault.

[0069] Step 2.2: determining a first correlation coefficient of any one ultrasonic signal segment based on the similarity of distribution characteristics between any one ultrasonic signal segment and all other ultrasonic signal segments.

[0070] The cross-correlation function is used to calculate the cross-correlation coefficient between any one ultrasonic signal segment in the ultrasonic signal segment set and the remaining ultrasonic signal segments at each time delay, which is recorded as a first cross-correlation coefficient; the average of the first cross-correlation coefficients between any one ultrasonic signal segment in the ultrasonic signal segment set and the remaining ultrasonic signal segments at all time delays is recorded as a first cross-correlation average;

[0071] The average of the first cross-correlation averages between the any one ultrasonic signal segment in the ultrasonic signal segment set and all other ultrasonic signal segments is used as the first correlation coefficient of the any one ultrasonic signal segment.

[0072] The calculation formula of the mutual correlation coefficient is: ; In the formula, represents the cross-correlation function between the ultrasonic signal segment x and the ultrasonic signal segment y at a time delay τ; N represents the length of the ultrasonic signal segment x; Represents the value of the nth data in the ultrasonic signal segment x; Indicates the first data; τ represents the time delay between the ultrasonic signal segment x and the ultrasonic signal segment y, the value range of τ is [0, N-1], and τ is an integer.

[0073] It should be noted that: the larger the mutual correlation coefficient between two ultrasonic signal segments at a certain delay time, the more similar the distribution characteristics of the two ultrasonic signal segments at the delay time, and the greater the similarity between the waveforms of the ultrasonic signal segments; therefore, the larger the first mutual correlation mean between the two ultrasonic signal segments at all delay times, the more reliable the analysis of the waveform similarity between the two ultrasonic signal segments; then, when the first correlation coefficient of a certain ultrasonic signal segment is larger, the greater the similarity between the waveforms of the ultrasonic signal segment and all other ultrasonic signal segments, the more likely the waveform of the ultrasonic signal segment is to appear once in each power frequency cycle, and the stronger the 50Hz correlation of the ultrasonic signal segment.

[0074] Step 2.3, divide each ultrasonic signal segment into two ultrasonic signal small segments according to time, and form an ultrasonic signal small segment set of the ultrasonic signal small segments of all ultrasonic signal segments; determine the second correlation coefficient of each ultrasonic signal segment based on the similarity of distribution characteristics between each ultrasonic signal small segment of each ultrasonic signal segment and each ultrasonic signal small segment in the ultrasonic signal small segment set.

[0075] Furthermore, each ultrasonic signal segment in the ultrasonic signal segment set is divided equally according to time to obtain two ultrasonic signal small segments with a length of half the power frequency cycle, which are used to subsequently obtain the signal characteristics of the 100 Hz correlation of the ultrasonic signal within the power frequency cycle of each ultrasonic signal segment, and all the ultrasonic signal small segments obtained after the equal division are combined into an ultrasonic signal small segment set.

[0076] A cross-correlation function is used to calculate the cross-correlation coefficient between each ultrasonic signal small segment in the ultrasonic signal small segment set and the remaining ultrasonic signal small segments at each time delay, which is recorded as a second cross-correlation coefficient; an average of the second cross-correlation coefficients between each ultrasonic signal small segment in the ultrasonic signal small segment set and the remaining ultrasonic signal small segments at all time delays is recorded as a second cross-correlation average;

[0077] Calculate the mean of the second cross-correlation means between each ultrasonic signal small segment and all other ultrasonic signal small segments in the set of ultrasonic signal small segments, record it as the third cross-correlation mean, and take the mean of the third cross-correlation means of two ultrasonic signal small segments of each ultrasonic signal segment as the second correlation coefficient of each ultrasonic signal segment.

[0078] Among them, for two small ultrasonic signal segments, the cross-correlation function between the two small ultrasonic signal segments is calculated according to the same calculation method as the cross-correlation function between the two ultrasonic signal segments. During the calculation process, it is only necessary to replace the ultrasonic signal segment with the small ultrasonic signal segment.

[0079] It should be noted that: the larger the mutual correlation coefficient between two small ultrasonic signal segments at a certain delay time, the more similar the distribution characteristics of the two small ultrasonic signal segments at the delay time, and the greater the similarity between the waveforms of the small ultrasonic signal segments; therefore, the larger the second mutual correlation mean between the two small ultrasonic signal segments, the more reliable the analysis of the waveform similarity between the two small ultrasonic signal segments; the larger the third mutual correlation mean of a small ultrasonic signal segment, the more similar the waveforms of the small ultrasonic signal segment are to all other small ultrasonic signal segments; then, when the second correlation coefficient of a certain ultrasonic signal segment is larger, the waveform of the ultrasonic signal segment appears twice in each cycle, and the 100Hz correlation of the ultrasonic signal segment is stronger.

[0080] Step 2.4, comprehensively analyzing the waveform characteristic vector, the first correlation coefficient, the second correlation coefficient, and the peak value of each ultrasonic signal segment to determine the partial discharge characteristic vector of each ultrasonic signal segment.

[0081] The periodic characteristic vector T=(T(50), T(100)) of the ultrasonic signal segment B is obtained, wherein T(50) represents the first correlation coefficient of the ultrasonic signal segment B, and T(100) represents the second correlation coefficient of the ultrasonic signal segment B.

[0082] It should be noted that: if T(50) is larger and T(100) is smaller, the 50Hz correlation of the ultrasonic signal segment B is stronger and the 100Hz correlation is weaker, the waveform of the ultrasonic signal has more periodic distribution characteristics, and the waveform of the ultrasonic signal segment is more likely to appear once in each power frequency cycle, then the ultrasonic signal segment B is more likely to be an ultrasonic signal generated by a needle-plate discharge fault or a surface discharge fault; and if Ti(50) is smaller and Ti(100) is larger, the 50Hz correlation of the ultrasonic signal segment B is weaker and the 100Hz correlation is stronger, the waveform of the ultrasonic signal has more periodic distribution characteristics, and the waveform of the ultrasonic signal segment B is more likely to appear twice in each power frequency cycle, then the ultrasonic signal segment B is more likely to be an ultrasonic signal generated by an internal discharge fault or a suspended discharge fault.

[0083] All peak points of the envelope of any ultrasonic signal are extracted, and the maximum amplitude among the obtained peak points is taken as the maximum peak value of any ultrasonic signal segment. The smaller the maximum peak value, the more likely it is that any ultrasonic signal segment is an ultrasonic signal generated by an internal discharge fault of a high-voltage switchgear. The larger the maximum peak value, the more likely it is that any ultrasonic signal segment is an ultrasonic signal generated by a needle-plate discharge fault, a suspended discharge fault, or a surface discharge fault.

[0084] The local discharge characteristic vector β=(F,K,T) of the ultrasonic signal segment B is obtained, where F, K, and T represent the maximum peak value, waveform characteristic vector, and period characteristic vector of the ultrasonic signal segment B, respectively.

[0085] Step 3: diagnose the type of partial discharge fault of the high-voltage switchgear by fusing multiple information based on the ultra-high frequency detection method and the partial discharge feature vector.

[0086] The same calculation method as the partial discharge feature vector β=(F, K, T) is used to calculate the partial discharge feature vector of each ultrasonic signal segment, and all the obtained partial discharge feature vectors are combined into a partial discharge feature set of ultrasonic information of the high-voltage switchgear.

[0087] This application realizes the fault diagnosis of partial discharge of high-voltage switchgear based on the ultra-high frequency detection method and the partial discharge feature set of ultrasonic information of high-voltage switchgear. Figure 3 shown.

[0088] Specifically, a back propagation neural network (BPNN) is used to identify and classify the local discharge feature set, and the classification results are passed to the basic probability assignment BPA (Basic probability assignment) of the DS fusion block for information correction. The loss function of BPNN is the mean square error. BPNN is a well-known technology and will not be described in detail in this application. Among them, the process of acquiring training samples of the BPNN neural network is: randomly extract a preset number of ultrasonic signal samples from the historically collected ultrasonic signals, and for the ultrasonic signal samples, use the same calculation method as the local discharge feature vector β to extract the local discharge feature vector from the ultrasonic signal sample, and use the local discharge feature vector extracted from the ultrasonic signal sample as the training sample of the BPNN neural network. Training based on the features extracted from the ultrasonic signal samples can help the model identify key information in the data.

[0089] In this embodiment, the value of the preset number is 2000. The value of the preset number is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0090] The electromagnetic wave information of the high-voltage switchgear is collected by using the ultra-high frequency detection method, and the discharge phase distribution characteristic information of the electromagnetic wave information is extracted in the phase analysis mode, which is also called the PRPD (Phase Resolved Partial Discharge) fingerprint, to obtain the PRPD spectrum information, and then the PRPD spectrum information is subjected to feature extraction to obtain the feature set of the PRPD spectrum information. The feature set of the PRPD spectrum information is subjected to feature subset extraction and data simplification by the principal component analysis method (PCA) to obtain the optimal feature subset of the electromagnetic wave information of the high-voltage switchgear, and the optimal feature subset is identified and classified by BPNN, and the classification result is passed to the basic probability allocation BPA of the DS fusion block for information correction. Among them, the ultra-high frequency detection method and the principal component analysis method are well-known technologies, and this application will not repeat them. Among them, the training sample of the BPNN for identifying and classifying the optimal feature subset is: the optimal feature subset of the electromagnetic wave information collected historically.

[0091] The information corrected by the basic probability allocation BPA is fused through the DS evidence theory to output the probability of each partial discharge fault. DS fusion is a well-known technology and will not be described in detail in this application.

[0092] The partial discharge fault with the highest probability is taken as the fault type of the high-voltage switchgear. When the probabilities of the four partial discharge faults, namely, needle plate discharge, internal discharge, suspension discharge, and surface discharge, are equal or the probabilities of the four partial discharge faults are all less than the preset values, it is determined that the high-voltage switchgear has no fault. The schematic diagram of the partial discharge fault diagnosis process is shown in the figure. Figure 4 shown.

[0093] In this embodiment, the preset value is 50%. The preset value is preset manually and can be set by the implementer. This application does not impose any special restrictions.

[0094] Based on the same inventive concept as the above method, an embodiment of the present application also provides a switch cabinet partial discharge diagnostic system, including a memory, a processor, and a computer program stored in the memory and running on the processor, and when the processor executes the computer program, the steps of any one of the above switch cabinet partial discharge diagnostic methods are implemented.

[0095] In summary, the present application analyzes four types of partial discharge faults of the high-voltage switchgear, namely, needle-plate discharge, internal discharge, suspended discharge, and creeping discharge, and determines the partial discharge feature vector of the ultrasonic signal in each power frequency cycle based on the peak characteristics, waveform characteristics, and periodic characteristics of the ultrasonic signal of the high-voltage switchgear. The application considers the differences in the maximum peak value, signal waveform, and waveform periodicity in each power frequency cycle between the ultrasonic signals generated by the four types of partial discharge faults of the high-voltage switchgear, namely, needle-plate discharge, internal discharge, suspended discharge, and creeping discharge, thereby improving the discrimination between various partial discharge faults of the high-voltage switchgear.

[0096] Furthermore, the partial discharge feature set of the ultrasonic information of the high-voltage switchgear is obtained based on the partial discharge feature vector, and combined with the optimal feature subset of the electromagnetic wave information of the high-voltage switchgear obtained by the ultra-high frequency detection method, the DS evidence theory is used to fuse multiple detection information to output the probability of each partial discharge fault. Finally, based on the fault rule, it is diagnosed whether the high-voltage switchgear has a fault, which improves the accuracy of the partial discharge feature set of the ultrasonic information of the high-voltage switchgear, and thus improves the accuracy of the partial discharge fault diagnosis of the high-voltage switchgear.

[0097] The flowchart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to the embodiment of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. In some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. In the description corresponding to the flowchart and the block diagram in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in a different order from the order disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified functions or actions, or may be implemented by a combination of dedicated hardware and computer instructions.

[0098] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the basic features of the present application. Therefore, no matter from which point of view, the above embodiments of the present application should be regarded as exemplary and non-restrictive.

Claims

1. A switch cabinet partial discharge diagnosis method, characterized in that: The method comprises the following steps: Collect ultrasonic signals from high-voltage switchgear; The ultrasonic signal is divided into ultrasonic signal segments with a length of one power frequency cycle, and a waveform feature vector of each ultrasonic signal segment is determined based on the peak shape of the envelope of each ultrasonic signal segment; Determine a first correlation coefficient of any one ultrasonic signal segment based on similarities in distribution characteristics between any one ultrasonic signal segment and all other ultrasonic signal segments; Divide each ultrasonic signal segment equally into two ultrasonic signal small segments according to time, form the ultrasonic signal small segments of all ultrasonic signal segments into an ultrasonic signal small segment set, and determine a second correlation coefficient of each ultrasonic signal segment based on the similarity of distribution characteristics between each ultrasonic signal small segment of each ultrasonic signal segment and each ultrasonic signal small segment in the ultrasonic signal small segment set; Comprehensively analyzing the waveform characteristic vector, the first correlation coefficient, the second correlation coefficient, and the peak value of each ultrasonic signal segment to determine the partial discharge characteristic vector of each ultrasonic signal segment; Based on the ultra-high frequency detection method and the partial discharge feature vector, multi-information fusion is performed to diagnose the partial discharge fault type of the high-voltage switchgear; The process of determining the waveform feature vector is as follows: Extracting the envelope of each ultrasonic signal segment, and determining the shape feature vector of the peak where each peak point in each envelope is located based on the peak width and peak height of the peak where each peak point in each envelope is located; Determining a waveform feature vector of each ultrasonic signal segment based on shape feature vectors of all peaks of the envelope of each ultrasonic signal segment; The process of determining the shape feature vector is as follows: For each peak point in each envelope, obtain the left peak width, right peak width and peak height of the peak where each peak point is located; The calculation method of the left peak width of any peak is as follows: obtain the data point with the smallest ordinate on the left side of the peak point of any peak, record it as the minimum data point on the left side, and take the absolute value of the difference between the peak point of any peak and the minimum data point on the left side as the left peak width of any peak; the calculation method of the right peak width is the same as the calculation method of the left peak width; The ratios of the left peak width and the right peak width to the peak height of the peak where any peak point is located are respectively used as the first component and the second component of the shape feature vector of the peak where any peak point is located; The determination process of the first correlation coefficient is: All ultrasonic signal segments are combined into an ultrasonic signal segment set; Calculating the cross-correlation coefficient between any one ultrasonic signal segment in the ultrasonic signal segment set and the remaining ultrasonic signal segments at each time delay by using a cross-correlation function, recorded as a first cross-correlation coefficient; Recording the average of the first cross-correlation coefficients between any one ultrasonic signal segment in the ultrasonic signal segment set and the remaining ultrasonic signal segments at all time delays as a first cross-correlation average; taking the average of the first cross-correlation averages between the any one ultrasonic signal segment in the ultrasonic signal segment set and all other ultrasonic signal segments as the first correlation coefficient of the any one ultrasonic signal segment; The determination process of the second correlation coefficient is: A cross-correlation function is used to calculate the cross-correlation coefficient between each ultrasonic signal small segment in the ultrasonic signal small segment set and the remaining ultrasonic signal small segments at each time delay, which is recorded as a second cross-correlation coefficient; an average of the second cross-correlation coefficients between each ultrasonic signal small segment in the ultrasonic signal small segment set and the remaining ultrasonic signal small segments at all time delays is recorded as a second cross-correlation average; Calculate the average of the second cross-correlation mean values ​​between each ultrasonic signal small segment in the ultrasonic signal small segment set and all other ultrasonic signal small segments, and record it as the third cross-correlation mean value; The average of the third cross-correlation averages of the two ultrasonic signal small segments of each ultrasonic signal segment is used as the second correlation coefficient of each ultrasonic signal segment.

2. A switch cabinet partial discharge diagnosis method according to claim 1, characterized in that: The mean values ​​of the components at the same positions of all shape feature vectors corresponding to each ultrasonic signal segment are used as the components in the waveform feature vector of each ultrasonic signal segment.

3. A switch cabinet partial discharge diagnosis method according to claim 1, characterized in that: The local discharge characteristic vector is composed of the maximum peak value, waveform characteristic vector and period characteristic vector of the envelope of each ultrasonic signal segment; The periodic eigenvector of each ultrasonic signal segment is composed of a first correlation coefficient and a second correlation coefficient of each ultrasonic signal segment.

4. A switch cabinet partial discharge diagnosis method according to claim 1, characterized in that: The process of diagnosing the partial discharge fault type of the high-voltage switchgear is as follows: The partial discharge feature vectors of all ultrasonic signal segments are combined into a partial discharge feature set of ultrasonic information of the high-voltage switchgear; The UHF detection method and principal component analysis method are used to obtain the optimal feature subset of the electromagnetic wave information of the high-voltage switchgear; Neural network is used to identify and classify the partial discharge feature set and the optimal feature subset respectively; Based on the basic probability distribution, the classification results of the partial discharge feature set and the classification results of the optimal feature subset are respectively corrected; The information corrected by the basic probability allocation BPA is fused through the DS evidence theory to obtain the probability of each partial discharge fault. Diagnose whether a partial discharge fault occurs in the high-voltage switchgear based on the probability of all types of partial discharge faults.

5. A switch cabinet partial discharge diagnosis method according to claim 1, characterized in that: The partial discharge fault with the highest probability is taken as the fault type of the high-voltage switchgear. When the probabilities of all types of partial discharge faults are equal or the probabilities of all types of partial discharge faults are less than a preset value, it is determined that the high-voltage switchgear has no fault.

6. A switch cabinet partial discharge diagnostic system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the switch cabinet partial discharge diagnosis method as described in any one of claims 1 to 5 are implemented.

Images