Active defense live-line detection method and device for power transformation main equipment

Abstract

The invention is suitable for the technical field of fault detection, and provides an active defense live-line detection method and device for power transformation main equipment, and the method comprises the steps: obtaining a partial discharge signal of the power transformation main equipment; obtaining the correlation between the partial discharge signal and a historical partial discharge signal; wherein the correlation is used for indicating the correlation degree of the partial discharge signals and the historical partial discharge signals with a predetermined fault, and at least two signals synchronously collected in the partial discharge signals participate in correlation judgment; and obtaining a detection result according to the correlation. The problem that in the prior art, the fault recognition rate is low in the mode that analysis is conducted through single data is solved, the fault recognition rate is increased, powerful support is provided for active defense of power transformation main equipment, and the service life of the equipment is further prolonged.

Description

Technical Field

[0001] The present application belongs to the field of detection technology, and in particular, relates to a method and device for active protection against live detection of a main substation equipment. Background Art

[0002] Safe, high-quality and economical transmission of electric energy is a basic requirement for the operation of modern power systems. Transformers and / or reactors (hereinafter referred to as main substation equipment, transformers are used as an example below) are one of the key equipment in power systems and power grid systems.

[0003] Since transformers in power systems need to operate under load for a long time, they are usually more likely to fail than other general equipment. In the current power grid system, there are more and more power grid system upgrades, various cross-regional networking and dispatching. Once a transformer fails, it is easy to cause a chain reaction in a certain part of the power grid. Therefore, daily fault detection and diagnosis for transformers, which can detect possible faults before the transformer fails and carry out repairs, is an important means to maintain the stable operation of the power grid.

[0004] The fault diagnosis method of the transformer is to perform data analysis on the obtained transformer fault information, and judge the predetermined fault of the transformer based on the result of data analysis. In actual engineering, there are six types of diagnostic methods for transformer status: oil chromatography analysis, partial discharge method, infrared temperature measurement method, electrical test method, oil test method and winding deformation test method. The most basic is oil chromatography analysis, which is an effective method for judging transformer faults. Whether it is oil chromatography analysis or other analysis methods, they are all based on one parameter for analysis. Therefore, the inventors found that the method of using single data for analysis in the prior art has a low fault recognition rate. Summary of the invention

[0005] The embodiments of the present application provide a method and device for active protection of live detection of substation main equipment, which can improve the accuracy and reliability of fault detection.

[0006] According to one aspect of an embodiment of the present application, a method for active defense live detection of a substation main equipment is provided, wherein the substation main equipment includes at least one of the following: a transformer, a reactor, and a gas-insulated switchgear (GIS for short); the method is applied to a live detection device including multiple channels, wherein each channel is used to access the signal of a sensor; wherein the partial discharge signal includes at least two signals; the partial discharge signal is obtained by synchronously collecting at least two of the multiple channels; the correlation between the partial discharge signal and the historical partial discharge signal is obtained; wherein the correlation is used to indicate the degree of correlation between the partial discharge signal and the historical partial discharge signal and a predetermined fault, and at least two signals synchronously collected in the partial discharge signal both participate in the judgment of the correlation, and the historical partial discharge signal is a partial discharge signal collected previously, wherein each predetermined fault corresponds to its own historical partial discharge signal, and the historical partial discharge signal includes at least one of the following: a partial discharge signal collected before the fault occurs, a partial discharge signal collected during the fault occurs, and a partial discharge signal collected in a predetermined time period after the fault occurs; and a detection result is obtained according to the correlation.

[0007] Further, the at least two signals are at least two of the following signals: ultrasonic signal, high frequency signal, ultra-high frequency signal, radio frequency partial discharge signal, soundprint signal, vibration signal, bushing end screen signal, pulse current method partial discharge signal, transient ground wave signal, core addition ground current signal; and / or, the sensor includes at least two of the following: ultrasonic sensor, high frequency sensor, ultra-high frequency sensor, radio frequency sensor, soundprint sensor, vibration sensor, bushing end screen sensor, partial discharge input unit, ground wave sensor, ground current sensor.

[0008] Further, obtaining the correlation between the partial discharge signal and the historical partial discharge signal includes: obtaining a signal type of the partial discharge signal; and obtaining the correlation of each signal in the partial discharge signal according to the signal type.

[0009] Further, obtaining the correlation between the local discharge signal and the historical local discharge signal includes: fusing multiple signals in the historical local discharge signal according to the correlation to obtain historical fused data; fusing multiple signals in the local discharge signal according to the correlation to obtain fused data, wherein the local discharge signal is fused in the same manner as the multiple signals in the historical local discharge signal; obtaining the detection result according to the correlation includes: obtaining the detection result according to the fused data and the historical fused data.

[0010] Furthermore, obtaining the detection result according to the fused data and the historical fused data includes: using the historical fused data and the corresponding predetermined fault as training data to train a fault detection model; and inputting the fused data into the fault detection model to obtain the detection result.

[0011] Furthermore, fusing multiple signals in the partial discharge signal according to correlation to obtain fused data includes: obtaining a feature sequence corresponding to each signal in the partial discharge signal under each predetermined fault; dividing the feature sequence corresponding to each signal in the partial discharge signal into a main feature sequence and a secondary feature sequence according to the correlation between the partial discharge signal of each signal type and the predetermined fault and a preset correlation threshold; wherein, if the correlation between a signal and the predetermined fault is less than the correlation threshold, the feature sequence corresponding to the signal is a secondary feature sequence; if the correlation between a signal and the predetermined fault is greater than or equal to the correlation threshold, the feature sequence corresponding to the signal is a main feature sequence; fusing the main feature sequence and the secondary feature sequence to obtain fused data under each predetermined fault.

[0012] Furthermore, obtaining the characteristic sequence corresponding to each signal in the local discharge signal includes: collecting each signal in the local discharge signal at a predetermined frequency to obtain a data sequence corresponding to each signal; when the correlation of the data sequence corresponding to a collected signal is lower than a preset condition, using data enhancement to perform data expansion on the data sequence to obtain an expanded data sequence; and obtaining a corresponding characteristic sequence according to the data sequence and the expanded data sequence.

[0013] Furthermore, fusing the main feature sequence and the secondary feature sequence to obtain the fused data includes: fusing the main feature sequence and the secondary feature sequence according to channel numbers to form a fused data matrix.

[0014] Furthermore, the fault detection model is obtained by training the historical fusion data and the corresponding predetermined faults as training data, including: using the main feature sequence and the secondary feature sequence to train the first fault detection model to obtain a trained first fault detection model; using the secondary feature sequence to train the second fault detection model to obtain a trained second fault detection model; wherein, when the main feature sequence is missing in the fusion data corresponding to the local discharge signal, the second fault detection model is used to output the detection result; when the secondary feature sequence is missing or exists in the fusion data corresponding to the local discharge signal, the first fault detection model is used.

[0015] Furthermore, after obtaining the fault detection result, the active defense detection method for live transformer main equipment also includes: determining the spatial coordinates of the local discharge source by using the ultrasonic signal and the placement position of the sensor; and determining the specific location of the fault based on the spatial coordinates of the local discharge source and the spatial coordinates of the physical structure of the transformer.

[0016] According to another aspect of the embodiment of the present application, there is also provided a device for actively defending against live detection of main substation equipment, comprising: a plurality of channels, wherein each channel is used to access the signal of a sensor, wherein the sensor comprises at least two of the following: an ultrasonic sensor, a high frequency sensor, and an ultra-high frequency sensor; a processor for executing software, wherein the software is used to execute the above-mentioned method for actively defending against live detection of main substation equipment.

[0017] Compared with the related art, the embodiments of the present application have the following beneficial effects:

[0018] The active defense live detection method and device of the substation main equipment of the embodiment of the present application obtains the local discharge signal of the substation main equipment; wherein the local discharge signal includes at least two signals; the local discharge signal is obtained by synchronously collecting at least two of the multiple channels; obtains the correlation between the local discharge signal and the historical local discharge signal; wherein the correlation is used to indicate the degree of correlation between the local discharge signal and the historical local discharge signal and the predetermined fault, at least two signals synchronously collected in the local discharge signal participate in the judgment of the correlation, the historical local discharge signal is the local discharge signal collected before, wherein each predetermined fault corresponds to its own historical local discharge signal, and the historical local discharge signal includes at least one of the following: the local discharge signal collected before the fault occurs, the local discharge signal collected during the fault, and the local discharge signal collected in a predetermined time period after the fault occurs; the detection result is obtained according to the correlation. The problem of low fault recognition rate in the prior art using a single data for analysis is solved, the fault recognition rate is improved, and strong support is provided for the active defense of the substation main equipment, further extending the service life of the equipment.

[0019] The beneficial effects of the above-mentioned second aspect embodiment refer to the beneficial effects of the first aspect embodiment, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0021] Figure 1 This is an application scenario diagram of a transformer active defense live detection device using multi-channel data fusion provided in an embodiment of the present application;

[0022] Figure 2 It is a flow chart of a method for active defense detection of live power of a main substation equipment provided by an embodiment of the present application;

[0023] Figure 3 This is a schematic diagram of a process for obtaining historical fusion data provided by an embodiment of the present application;

[0024] Figure 4 It is a flowchart of actually further obtaining a fault detection result provided by an embodiment of the present application;

[0025] Figure 5 It is a structural schematic diagram of a transformer active defense live detection device with multi-channel data fusion provided in one embodiment of the present application. DETAILED DESCRIPTION

[0026] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0027] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0028] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0029] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0030] References to "an embodiment", "one embodiment" or "some embodiments" described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0031] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings and specific implementation methods. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0032] It should be noted that in the following embodiments, transformers are used as examples for explanation, but the following embodiments can also be applied to other main substation equipment such as reactors and GIS, and their application methods are similar to those of transformers, which will not be described one by one here. The live detection device in the following embodiments uses multiple channels, each of which is used to access the signal of a sensor, and the sensor includes at least two of the following: ultrasonic sensor, high frequency sensor, ultra-high frequency sensor, radio frequency sensor, voiceprint sensor, vibration sensor, bushing end screen sensor, partial discharge input unit, ground wave sensor, ground current sensor; in the following embodiments, ultrasonic sensor, high frequency sensor, and ultra-high frequency sensor are used as examples for explanation, and at least two signals are used for analysis in the following embodiments, which is a data fusion approach. Therefore, in the following embodiments, the live detection device is also referred to as a transformer active defense live detection device with multi-channel data fusion. The so-called active defense here is because the power workers can actively use the detection device for detection, thereby achieving the effect of defense.

[0033] Figure 1 The figure shows an application scenario diagram of a transformer active defense live detection device using multi-channel data fusion according to an embodiment of the present application. Figure 1As shown, in this application scenario, first, the partial discharge signal of the transformer is collected by deploying multiple channels set on the live detection instrument of the transformer active defense live detection device with multi-channel data fusion. Then, the partial discharge signal is processed by using the transformer active defense live detection device with multi-channel data fusion. The result after processing can be used to judge the fault of the transformer. It should be noted that the fault here can be a fault that will occur in the future or a fault that exists currently. There are usually abnormalities before the fault occurs. If these abnormal signals exist in the collected partial discharge information, even if the fault has not occurred yet, it can be judged that a fault will occur in the future, or the probability of a fault in the future. More preferably, the time information of the future fault can also be predicted. After the fault is detected, a predetermined fault for indicating the transformer fault can also be generated according to the collected partial discharge signal.

[0034] Among them, the live detection instrument deployed with a transformer active defense live detection device with multi-channel data fusion can be designed as a multi-functional detection instrument or a portable detection instrument.

[0035] Figure 2 This is a flow chart of a method for active protection of live detection of a main substation equipment provided by an embodiment of the present application, with reference to Figure 2 The active defense live detection method of the main substation equipment includes:

[0036] Step 101, obtaining a partial discharge signal of a main substation equipment (eg, a transformer).

[0037] It should be noted that the sensor includes at least two of the following: ultrasonic sensor, high frequency sensor, ultra-high frequency sensor, radio frequency sensor, sound print sensor, vibration sensor, bushing end screen sensor, partial discharge input unit, ground wave sensor, ground current sensor; the partial discharge signal includes at least two signals; the partial discharge signal is obtained by synchronously collecting at least two of the multiple channels. For example, the at least two signals are at least two of the following signals: ultrasonic signal, high frequency signal, ultra-high frequency signal, radio frequency partial discharge signal, sound print signal, vibration signal, bushing end screen signal, pulse current method partial discharge signal, transient ground wave signal, core addition ground current signal. In the following embodiments, at least two of the ultrasonic signal, high frequency signal and ultra-high frequency signal are used as examples for description.

[0038] Exemplarily, the plurality of channels may include a first preset number of ultrasound channels and a second preset number of high frequency channels. For example, the first preset number is 5, the second preset number is 3, and a total of 8 channels. Alternatively, at least one ultrasound channel and at least one ultra-high frequency channel may be included; or at least one high frequency channel and at least one ultra-high frequency channel may be included. The number of these channels may be set according to actual needs. Figure 1 In the device shown, one has 8 channels, which may include at least one ultrasonic channel, at least one high-frequency channel, and at least one ultra-high frequency channel, and the collected partial discharge signal may also include an ultrasonic signal, a high-frequency signal, and an ultra-high frequency signal. In some of the following embodiments, the partial discharge signal including the ultrasonic signal and the high-frequency signal is used as an example for explanation, and the fusion processing method of the partial discharge signal including other signals is similar to it, which will not be repeated here one by one.

[0039] The live detection device may also have a network function, through which the live detection device can be connected to a server, and the server may store the partial discharge signals collected in the past, or the historical partial discharge signals may be stored in the live detection device. It should be noted that in the following embodiments, the training of the artificial intelligence model can be carried out through the server, and the artificial intelligence model can be stored on the server or on the live detection device. When stored on the server, the live detection device needs to connect to the server through the network to interact with the service. If the model is stored in the live detection device, it is not necessary to connect to the server through the network. In the specific implementation, it can be flexibly selected according to different needs, and it will not be repeated here.

[0040] These partial discharge signals can be divided into multiple signal types according to the different signals included. For example, the partial discharge signals of the first signal type include ultrasonic signals and high-frequency signals, the partial discharge signals of the second signal type include ultrasonic signals and ultra-high frequency signals, the partial discharge signals of the third signal type include high-frequency signals and ultra-high frequency signals, and the partial discharge signals of the fourth signal type include ultrasonic signals, high-frequency signals and ultra-high frequency signals. Different signal types correspond to a type of historical data.

[0041] These historical data are collected in advance, and these partial discharge signals may be collected when a fault occurs, or may be collected before a fault occurs, or may even include signal data collected after a fault occurs. The collected partial discharge signals may be a time series of a period of time, for example, a collection is performed every 5 milliseconds, and each collection simultaneously collects the signals of each channel in multiple channels, which is referred to as synchronization in the following embodiments.

[0042] After the acquisition, the local discharge signals corresponding to different predetermined faults (the predetermined faults may include information such as fault type, fault condition, fault characteristics, fault development degree, fault severity, etc., and the fault type is used as an example for explanation below) can be obtained. It should be noted that these local discharge signals need to be processed after being acquired. In the following implementation, the signals that are not processed after acquisition are referred to as initial local discharge signals, because these faults have already occurred, which is equivalent to history at the current time, and therefore are also referred to as historical local discharge signals. The processing of the initial local discharge signals is explained below.

[0043] The initial partial discharge signal of each preset predetermined fault of the transformer is obtained; data cleaning, data standardization and feature extraction operations are performed on the initial historical partial discharge signal to obtain a processed partial discharge signal.

[0044] Among them, data cleaning includes at least one of the following: removing noise from the collected data, removing outliers and removing duplicate data to ensure the accuracy and reliability of the data. Data standardization includes standardizing or normalizing different types of data to eliminate differences in dimensions and numerical ranges to facilitate subsequent data fusion. Feature extraction includes extracting characteristic indicators related to transformer faults from the initial historical partial discharge signal, such as waveform features, spectrum features, etc.

[0045] Generally, each channel will have a designated monitoring point. High-frequency channel monitoring is suitable for capturing discharge activities inside the transformer, such as monitoring faults at the end screen of the transformer bushing, the transformer core and the clamps. Partial discharge events also generate ultrasonic signals, which can be transmitted through the air or transformer oil. Ultrasonic sensors can be installed on the transformer casing to capture these ultrasonic signals. Ultrasonic monitoring is useful for detecting the location of the discharge because the ultrasonic signal decays quickly, which means that the signal source is usually close to the sensor.

[0046] However, if each fault or each designated monitoring point relies solely on the single channel sensor data of the fault detection device during the fault detection process, there are problems such as low detection accuracy and susceptibility to interference. For example, although high-frequency channel monitoring is suitable for capturing discharge activities inside the transformer, when the high-frequency current sensor is interfered with or damaged, it cannot obtain an effective high-frequency signal or cannot obtain a high-frequency signal, and the fault of the monitoring point corresponding to the high-frequency channel will be missed.

[0047] Although other ultrasonic signals are not very sensitive to faults at the monitoring points corresponding to the high-frequency channels, they will also be affected by the faults and change. For example, the performance of local discharges caused by cracks may change. This subtle change can be used to assist in or replace the identification of interfered or damaged channels, thereby avoiding the failure to monitor the occurrence of faults in a timely manner during the long period of sensor replacement and preventing missed faults.

[0048] It should be noted that the processing of the above-mentioned historical partial discharge signals can be performed by a server or by a live detection device. After processing, the historical partial discharge signals corresponding to each fault can be obtained. For example, fault A corresponds to a group of historical partial discharge signals, fault B corresponds to a group of historical partial discharge signals, etc. In order to distinguish fault A from fault B, fault A and fault B are referred to as faults of different fault types (i.e., different predetermined faults) in the following implementation. Faults have different fault types, and partial discharge signals also have different signal types. For example, the saved historical partial discharge signals may include partial discharge signals of different fault types corresponding to the first signal type, may also include partial discharge signals of different fault types corresponding to the second signal type, may also include partial discharge signals of different fault types corresponding to the third signal type, and may also include partial discharge signals of different fault types corresponding to the fourth signal type.

[0049] Step 102: Obtain the correlation between the fused data and historical fused data.

[0050] In this step, the correlation with the historical partial discharge signal can be determined according to the partial discharge signals of different signal types. As an optional implementation, the signal type of the partial discharge signal can be obtained, for example, the signal type is one of the following: a first signal type including an ultrasonic signal and a high-frequency signal, a second signal type including an ultrasonic signal and an ultra-high frequency signal, a third signal type including a high-frequency signal and an ultra-high frequency signal, and a fourth signal type including an ultrasonic signal, a high-frequency signal and an ultra-high frequency signal; the correlation of each signal in the partial discharge signal is obtained according to the signal type. Through this optional implementation, the correlation of the corresponding historical partial discharge signal can be found according to different signal types.

[0051] Determine the correlation between the local discharge signal of each signal type and the historical local discharge signal corresponding to different predetermined faults under the signal type. As an optional implementation, since both the local discharge signal and the historical local discharge signal include at least two signals (i.e., at least two of the ultrasonic signal, the high frequency signal and the ultra-high frequency signal), at least two signals in the local discharge signal can also be fused (which can be understood as being processed into a holistic data through an algorithm). In the following implementation, the fused data is referred to as fused data. For determining the correlation, at least two signals in the collected local discharge signal can be fused to obtain the collected fused data (in order to distinguish it from the historical fused data, in the following implementation, the fused data can also be referred to as real-time fused data. It should be noted that "real-time" only represents a relative concept with "history" and has no other special meaning).

[0052] It should be noted that due to the different types of local discharge signals, in an optional embodiment, the signal type of the local discharge signal can be obtained, for example, the signal type is one of the following: a first signal type including an ultrasonic signal and a high-frequency signal, a second signal type including an ultrasonic signal and an ultra-high frequency signal, a third signal type including a high-frequency signal and an ultra-high frequency signal, and a fourth signal type including an ultrasonic signal, a high-frequency signal and an ultra-high frequency signal; and then the correlation of each signal in the local discharge signal is obtained according to the signal type.

[0053] The partial discharge signal includes at least two signals, for example, an ultrasonic signal and a high-frequency signal. These signals are collected at a certain frequency. Therefore, the collected signal can also be called a sequence or a data sequence. The collected sequence can be processed to determine the correlation. The sequence determined by the correlation can be called a feature sequence. Similarly, the historical partial discharge signal is also a sequence. The sequence can calculate the correlation, which can be called a correlation coefficient. For example, the Pearson correlation coefficient can be used as a representation of the correlation.

[0054] The Pearson correlation coefficient is a statistic used to measure the degree of linear correlation between two continuous variables. Its value is between -1 and 1, where 1 represents a perfect positive correlation, -1 represents a perfect negative correlation, and 0 represents no correlation.

[0055] The following is an example of a partial discharge signal including an ultrasonic signal and a high-frequency signal. In step 102, the correlation between the ultrasonic signal and the high-frequency signal in the partial discharge signal and each preset predetermined fault can be calculated by respectively calculating the Pearson correlation coefficient between the sample sequence of the ultrasonic signal and the high-frequency signal and each fault feature. It should be noted that since the historical partial discharge signal is a signal related to the fault, the sequence corresponding to the historical partial discharge signal can also be called a fault feature or a fault sequence or fault data, and the use of the Pearson correlation coefficient to calculate the correlation is only an implementation method. For example, it can also be calculated according to the method of determining the correlation between each type of historical partial discharge signal and the fault feature corresponding to the preset predetermined fault in the following step 202.

[0056] In an optional implementation, considering that each partial discharge signal and each historical partial discharge signal include multiple signals, the sequence corresponding to each signal can also be fused according to the correlation. For example, obtaining the correlation between the partial discharge signal and the historical partial discharge signal includes: fusing multiple signals in the historical partial discharge signal according to the correlation to obtain historical fused data; fusing multiple signals in the partial discharge signal according to the correlation to obtain fused data, wherein the partial discharge signal is fused in the same manner as the multiple signals in the historical partial discharge signal; obtaining the detection result according to the correlation includes: obtaining the detection result according to the fused data and the historical fused data.

[0057] The fused data may correspond to a predetermined fault. For example, then, based on the correlation, the real-time fused data corresponding to each preset predetermined fault is determined, including:

[0058] The characteristic sequence of each type of partial discharge signal is determined in descending order of correlation.

[0059] Exemplarily, characteristic sequences of various types of local discharge signals are determined in order of correlation from high to low, including: determining a data sequence of each type of local discharge signal in order of correlation from high to low; under each preset predetermined fault, using data enhancement technology to perform data expansion on a data sequence with low correlation to obtain an expanded data sequence; based on the expanded data sequence, determining the characteristic sequence of each type of local discharge signal under each preset predetermined fault.

[0060] In an optional embodiment, the real-time fused data can also be divided into a main feature sequence and a sub-feature sequence. It should be noted that the real-time fused data (i.e., fused data) is processed in the same way as the historical fused data. Therefore, the historical fused data can also be divided into a main feature sequence and a sub-feature sequence. For example, for the historical fused data, based on each preset predetermined fault, the feature sequences of various types of local discharge signals are fused to obtain the fused data under each preset predetermined fault.

[0061] Exemplarily, based on each preset predetermined fault, the characteristic sequences of various types of partial discharge signals are fused to obtain real-time fused data under each preset predetermined fault, including:

[0062] Under each preset predetermined fault: according to the correlation between each type of partial discharge signal and the preset predetermined fault and the preset correlation threshold, the feature sequence is divided into a main feature sequence and a secondary feature sequence; the main feature sequence is provided with a main label, and the secondary feature sequence is provided with a secondary label.

[0063] Among them, if the correlation between a certain type of partial discharge signal and the preset predetermined fault is less than the preset correlation threshold, the characteristic sequence corresponding to the certain type of partial discharge signal is a secondary characteristic sequence, and if the correlation between a certain type of partial discharge signal and the preset predetermined fault is greater than or equal to the preset correlation threshold, the characteristic sequence corresponding to the certain type of partial discharge signal is a primary characteristic sequence. For example, for a discharge fault in a liquid medium, the correlation between the high-frequency signal and the predetermined fault is 0.9, and the correlation threshold between the high-frequency signal and the characteristic sequence corresponding to the predetermined fault is 0.3; the correlation between the ultrasonic signal and the predetermined fault is 0.1, and the correlation threshold between the ultrasonic signal and the predetermined fault is 0.4. Then, the characteristic sequence corresponding to the high-frequency signal is the primary characteristic sequence, and the characteristic sequence corresponding to the ultrasonic signal is the secondary characteristic sequence.

[0064] The main feature sequence and the secondary feature sequence are fused to obtain real-time fused data under each preset predetermined fault; the real-time fused data carries the main label and the secondary label.

[0065] Exemplarily, the main feature sequence and the secondary feature sequence may be fused according to the channel number to form a fused data matrix.

[0066] In this embodiment, the selected ultrasonic signal and high-frequency signal are both data related to the fault, but the correlation is high or low. Then the data enhancement technology is used to expand the data sequence with low correlation to improve the characterization ability of such faults in real-time fusion data. For example, in the above example, the correlation between the ultrasonic signal and the discharge fault in the liquid medium was originally 0.02. After data enhancement of the characteristic part that causes slight changes in the ultrasonic signal, the correlation between the ultrasonic signal and the discharge fault in the liquid medium is increased to 0.1. When a certain data with high correlation cannot be obtained normally due to sensor damage, other data sequences with low correlation can still play a role in identification.

[0067] This embodiment collects historical partial discharge signals before each preset predetermined fault of the transformer, analyzes its subtle signs, forms historical fusion data with the help of parameters such as amplitude, phase, frequency, growth rate, duration, acoustic-electric linkage, and positioning of the historical partial discharge signals, establishes an accurate prediction model, and gradually finds out the characteristic laws of transformer faults.

[0068] See also Figure 3 , obtaining historical fusion data may include the following steps:

[0069] Step 201 : obtaining a historical partial discharge signal of each preset predetermined fault of the transformer.

[0070] Exemplarily, the historical partial discharge signal may be obtained from a database of partial discharge signals stored during historical transformer detection.

[0071] Step 202 : determining the correlation between each type of signal in the historical partial discharge signal of each preset predetermined fault and the corresponding preset predetermined fault, and determining the historical fusion data corresponding to each preset predetermined fault based on the correlation.

[0072] The historical fusion data includes the main feature sequence and the secondary feature sequence.

[0073] Exemplarily, in step 202, determining the correlation between various types of historical partial discharge signals and preset predetermined faults includes:

[0074] For a type of preset predetermined fault, the accuracy of the ultrasonic signal and the high-frequency signal in judging the preset predetermined fault can be used to calculate the correlation between the ultrasonic signal and the high-frequency signal in the corresponding historical partial discharge signal and the preset predetermined fault. The higher the accuracy, the stronger the correlation.

[0075] For a class of preset predetermined faults, the correlation between the ultrasonic signal and the high-frequency signal in the corresponding historical partial discharge signal and the preset predetermined fault can also be calculated by respectively calculating the Pearson correlation coefficient between the sample sequences of the ultrasonic signal and the high-frequency signal and the corresponding fault characteristics.

[0076] For a type of preset predetermined fault, the correlation between the ultrasonic signal and the high-frequency signal and the preset predetermined fault can be preset directly according to the existing known sensitivity.

[0077] Next, in step 202, based on the correlation, the historical fusion data corresponding to each preset predetermined fault is determined, including:

[0078] According to the order of correlation from high to low, the characteristic sequence of each type of partial discharge signal corresponding to each preset predetermined fault is determined.

[0079] Exemplarily, characteristic sequences of various types of historical partial discharge signals corresponding to each preset predetermined fault (e.g., preset fault type) are determined in order of correlation from high to low, including: determining data sequences of various types of historical partial discharge signals corresponding to each preset predetermined fault in order of correlation from high to low; under each preset predetermined fault, using data enhancement technology to perform data expansion on data sequences with low correlation to obtain expanded data sequences; based on the expanded data sequences, determining characteristic sequences of various types of partial discharge signals corresponding to each preset predetermined fault.

[0080] The characteristic sequences of various types of historical partial discharge signals corresponding to each preset predetermined fault are fused to obtain historical fusion data under each preset predetermined fault.

[0081] Exemplarily, the characteristic sequences of various types of historical partial discharge signals corresponding to each preset predetermined fault are fused to obtain historical fusion data under each preset predetermined fault, including:

[0082] Under each preset predetermined fault: according to the correlation between each type of historical partial discharge signal and the preset predetermined fault and the preset correlation threshold, the feature sequence is divided into a main feature sequence and a secondary feature sequence; the main feature sequence is provided with a main label, and the secondary feature sequence is provided with a secondary label.

[0083] The main feature sequence and the secondary feature sequence are fused to obtain historical fusion data under each preset predetermined fault; the historical fusion data carries the main label and the secondary label.

[0084] Exemplarily, the main feature sequence and the secondary feature sequence may be fused according to the channel number to form a historical fusion data matrix.

[0085] Step 103, obtaining a detection result according to the correlation.

[0086] There are many ways to obtain the detection results, for example, the historical fusion data and the corresponding predetermined faults are used as training data to train a fault detection model; the fusion data is input into the fault detection model to obtain the detection result. The above-mentioned fault model can be pre-trained, and therefore, it can also be called a preset fault detection model. The preset fault detection model can be implemented by, for example, a support vector machine, a random forest, a deep learning model, etc., and the model parameters are optimized by methods such as cross-validation to improve the accuracy of fault detection.

[0087] In one optional implementation manner described above, a primary feature sequence and a secondary feature sequence are used. Therefore, as an optional manner, the fault detection model described above includes a first fault detection model and a second fault detection model, which will be described below.

[0088] The first fault detection model is trained by using the main feature sequence and the secondary feature sequence of the feature sequence of the historical partial discharge signal to obtain the trained first fault detection model.

[0089] The second fault detection model is trained by using the sub-feature sequence of the feature sequence of the historical partial discharge signal to obtain a trained second fault detection model. The sub-feature sequence can be stored in a separate file before data fusion, and when training the second fault detection model, the sub-feature sequence and the preset predetermined fault are input into the second fault detection model; the sub-feature sequence can also be extracted from the historical fusion data by using the sub-label, and then the sub-feature sequence and the preset predetermined fault are input into the second fault detection model.

[0090] The real-time fusion data obtained each time will be stored in the database as historical fusion data, and the preset fault detection model will be automatically updated based on the updated historical fusion data.

[0091] This embodiment trains the fault detection models in two situations and performs targeted training on data with different characteristics to improve the accuracy of fault identification.

[0092] In this step, the fault detection result is obtained based on the real-time fusion data and the preset fault detection model.

[0093] Exemplarily, the process of obtaining fused data based on partial discharge signals is consistent with the process of obtaining fused data based on historical partial discharge signals. This step is different from step 202 after the fused data is obtained in that it is necessary to identify whether the real-time fused data includes a main feature sequence or a secondary feature sequence. For example, when a sensor that obtains data of the main feature sequence is disturbed or damaged, the main feature sequence does not exist in the real-time fused data. Therefore, after obtaining the real-time fused data, it is necessary to identify the real-time fused data.

[0094] Therefore, see Figure 4 :

[0095] Step 1031 , based on the real-time fusion data, determine a target fault detection model from the trained first fault detection model and the trained second fault detection model.

[0096] Exemplarily, the main label and sub-label of the real-time fusion data are obtained. When the main label and sub-label of the real-time fusion data exist at the same time, it means that the channels with high correlation and low correlation corresponding to the current predetermined fault are working normally. In this case, the trained first fault detection model is selected as the target fault detection model.

[0097] When the main label of the real-time fusion data does not exist and only the secondary label exists, it means that the channel with high correlation corresponding to the current predetermined fault is not working properly, while the channel with low correlation is working normally. In this case, the trained second fault detection model is selected as the target fault detection model.

[0098] In addition, if the secondary label of the real-time fusion data does not exist and only the main label exists, since the main feature sequence is highly correlated with the current fault, the absence of the secondary feature sequence will not have a significant impact on the recognition result, and the first fault detection model is still selected as the target fault detection model.

[0099] Step 1032: Perform detection based on the target fault detection model to obtain a fault detection result.

[0100] The target fault detection model outputs a fault detection result, and the fault detection result includes a predetermined fault.

[0101] For example, after determining the predetermined fault, the approximate location range of the fault can be obtained. There may be multiple locations of the bushing end screen, the core, the clamp, and the winding grounding lead. If the precise fault location is required, further analysis is required. Then, after obtaining the fault detection result, the active defense live detection method of the substation main equipment also includes:

[0102] Based on the fault detection results, the spatial coordinates of the local discharge source are determined using ultrasonic signals and the placement of the sensor; the specific location of the fault is determined based on the spatial coordinates of the local discharge source and the spatial coordinates of the physical structure of the transformer.

[0103] Using ultrasonic signals and the placement of sensors, the spatial coordinates of the local discharge source are determined, including:

[0104] The moment when the local discharge source emits an ultrasonic signal is selected as the timing starting point, and ultrasonic signals of different receiving durations from at least three channels are received. The three ultrasonic sensors are placed in fixed and non-collinear positions. The time when the ultrasonic signal of each channel is received from its respective receiving point is measured, and the spatial position of the point where the object to be measured is obtained by combining the ultrasonic propagation velocity.

[0105] For example, when receiving three-channel ultrasonic signals, because the receiving point of the ultrasonic signal is fixed, their coordinates in the three-dimensional space coordinate system are known, namely (x1, y1, z1), (x2, y2, z2) and (x3, y3, z3), and the ultrasonic propagation rate k is also known. If the coordinates of the local discharge source are assumed to be (x, y, z), the coordinates of the local discharge source can be calculated by the following equation.

[0106] (x1-x) 2 +(y1-y) 2 +(z1-z) 2 =k 2 t1 2

[0107] (x2-x) 2 +(y2-y) 2 +(z2-z) 2 =k 2 t2 2

[0108] (x3-x) 2 +(y3-y) 2 +(z3-z) 2 =k 2 t3 2

[0109] Among them, t1, t2 and t3 are the time taken for the ultrasonic signals of the three channels to propagate from the local discharge source to the respective receiving points.

[0110] After the coordinates of the local discharge source are determined, they are matched with the spatial coordinates of the physical structure of the transformer to determine the specific location of the fault. The spatial coordinates of the physical structure of the transformer, such as the location of the bushing end screen, the location of the core, the location of the clamp, and the location of the winding grounding lead, can be pre-stored in the system.

[0111] Exemplarily, after determining the predetermined fault or the specific location of the fault, the active defense live detection method of the substation main equipment further includes:

[0112] An early warning is issued based on the fault detection results and the specific location of the fault, and the fault detection results and the specific location of the fault are visualized.

[0113] In summary, the active defense live detection method for the main equipment of the transformer provided in the embodiment of the present application can more comprehensively capture various fault characteristics that may occur inside the transformer by comprehensively analyzing different types of partial discharge signals, and can respectively analyze the correlation between each type of partial discharge signal and these predetermined faults in combination with a plurality of preset predetermined faults (for example, fault types), so as to achieve comprehensive coverage of various potential faults of the transformer, not limited to detecting a single type of fault, and can simultaneously identify multiple different types of faults, thereby enhancing the comprehensiveness of fault detection. Through data fusion technology, the partial discharge signals of different channels are integrated into fused data containing the main feature sequence and the secondary feature sequence, which effectively reduces the redundancy of the data, improves the efficiency of subsequent model processing, avoids missed detection, improves the accuracy and reliability of fault detection, and provides strong support for the active defense of the transformer, further extending the service life of the equipment.

[0114] Data enhancement technology is used to expand data sequences with low correlation to improve the representation ability of such faults in fused data, and to perform auxiliary or alternative identification of disturbed or damaged channels to avoid the failure to monitor the occurrence of faults in a timely manner during the long period of sensor replacement and prevent missed faults.

[0115] See also Figure 5 The embodiment of the present application provides a transformer active defense live detection device with multi-channel data fusion, which executes the above-mentioned active defense live detection method for substation main equipment. The software in the device may include: an acquisition module 301, a data analysis and fusion module 302 and a detection result output module 303.

[0116] The acquisition module 301 is used to acquire a partial discharge signal of a main substation equipment; wherein the partial discharge signal includes at least two signals; and the partial discharge signal is acquired by synchronously acquiring at least two of the multiple channels;

[0117] The data analysis and fusion module 302 is used to obtain the correlation between the partial discharge signal and the historical partial discharge signal; wherein the correlation is used to indicate the correlation degree between the partial discharge signal and the historical partial discharge signal and the predetermined fault, at least two signals synchronously collected in the partial discharge signal participate in the judgment of the correlation, the historical partial discharge signal is the partial discharge signal collected previously, wherein each predetermined fault corresponds to its own historical partial discharge signal, and the historical partial discharge signal includes at least one of the following: a partial discharge signal collected before the fault occurs, a partial discharge signal collected during the fault occurs, and a partial discharge signal collected in a predetermined time period after the fault occurs;

[0118] The detection result output module 303 is used to obtain the detection result according to the correlation.

[0119] The software is used to execute various implementations of the above method embodiments, which are described below.

[0120] Exemplarily, the data analysis and fusion module 302 includes a correlation analysis module and a data fusion module. The data fusion module is used to:

[0121] Determine the characteristic sequence of each type of partial discharge signal in descending order of correlation;

[0122] Based on each preset predetermined fault, the characteristic sequences of various types of partial discharge signals are fused to obtain real-time fused data under each preset predetermined fault.

[0123] Exemplarily, the characteristic sequences of various types of partial discharge signals are determined in descending order of relevance, including:

[0124] Determine the data sequence of each type of partial discharge signal in descending order of correlation;

[0125] Under each preset predetermined fault, data enhancement technology is used to expand the data sequence with low correlation to obtain the expanded data sequence;

[0126] Based on the expanded data sequence, a characteristic sequence of each type of partial discharge signal under each preset predetermined fault is determined.

[0127] Exemplarily, based on each preset predetermined fault, the characteristic sequences of various types of partial discharge signals are fused to obtain real-time fused data under each preset predetermined fault, including:

[0128] For each preset scheduled fault:

[0129] According to the correlation between each type of partial discharge signal and the preset predetermined fault and the preset correlation threshold, the feature sequence is divided into a main feature sequence and a secondary feature sequence; the main feature sequence is provided with a main label, and the secondary feature sequence is provided with a secondary label;

[0130] The main feature sequence and the secondary feature sequence are fused to obtain real-time fused data under each preset predetermined fault; the real-time fused data carries the main label and the secondary label.

[0131] Exemplarily, the transformer active defense live detection device with multi-channel data fusion also includes a model training module.

[0132] Before obtaining the fault detection results based on the real-time fusion data and the preset fault detection model, the model training module is specifically used to:

[0133] Acquire a historical partial discharge signal of each preset predetermined fault of the transformer;

[0134] Determine the correlation between each type of signal in the historical partial discharge signal of each preset predetermined fault and the corresponding preset predetermined fault, and determine the historical fusion data corresponding to each preset predetermined fault based on the correlation; the historical fusion data includes a main feature sequence and a secondary feature sequence;

[0135] The historical fusion data is used to train the fault detection model to obtain a trained preset fault detection model.

[0136] Exemplarily, the preset fault detection model includes a first fault detection model and a second fault detection model;

[0137] The preset fault detection model is trained by using the historical fusion data to obtain the trained preset fault detection model, including:

[0138] Using the main feature sequence and the secondary feature sequence, training a first fault detection model to obtain a trained first fault detection model;

[0139] The secondary feature sequence is used to train the second fault detection model to obtain a trained second fault detection model.

[0140] Exemplarily, after obtaining the fault detection result, the transformer active defense live detection device with multi-channel data fusion further includes a fault location module.

[0141] The fault location module is used to:

[0142] Based on the fault detection results, the spatial coordinates of the partial discharge source are determined using the ultrasonic signal and the placement of the sensor. The specific location of the fault is determined based on the spatial coordinates of the partial discharge source and the spatial coordinates of the physical structure of the transformer.

[0143] Exemplarily, the transformer active defense live detection device with multi-channel data fusion also includes a visualization module.

[0144] The visualization module is specifically used for:

[0145] An early warning is issued based on the fault detection results and the specific location of the fault, and the fault detection results and the specific location of the fault are visualized.

[0146] Exemplarily, in the signal acquisition module 301, the multiple channels include a first preset number of ultrasound channels and a second preset number of high-frequency channels.

[0147] For example, the first preset number is 5, the second preset number is 3, and there are 8 channels in total.

[0148] Of course, there is no limitation on the number and type of channels, and other related channels and the number of channels of UHF, temperature, and gas content can also be expanded on the basis of the present invention.

[0149] The beneficial effects of the embodiment of the transformer active protection live detection device with multi-channel data fusion mentioned above refer to the beneficial effects of the embodiment of the active protection live detection method for substation main equipment, which will not be repeated here.

[0150] It should be noted that, although several units / modules or sub-units / modules of the transformer active defense live detection device with multi-channel data fusion are mentioned in the above detailed description, this division is only exemplary and not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided into multiple units / modules to be embodied.

[0151] In addition, although the operations of the method of the present application are described in a specific order in the drawings, this does not require or imply that the operations must be performed in this specific order, or that all the operations shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps.

[0152] The present application also provides a computer-readable storage medium, in which computer execution instructions are stored. When a processor executes the computer execution instructions, the active defense live detection method for substation main equipment provided in the above embodiment of the present application is implemented.

[0153] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0154] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for active defense detection of live power of main substation equipment, characterized in that: The substation main equipment includes at least one of the following: a transformer, a reactor, and a gas-insulated fully enclosed combined electrical appliance; the method is applied to a live detection device including a plurality of channels, wherein each channel is used to access a signal of a sensor, and the method includes: Acquire a partial discharge signal of a main substation equipment; wherein the partial discharge signal includes at least two signals; the partial discharge signal is acquired by synchronously collecting at least two of the multiple channels; Obtaining the correlation between the partial discharge signal and the historical partial discharge signal; wherein the correlation is used to indicate the degree of correlation between the partial discharge signal and the historical partial discharge signal and the predetermined fault, at least two signals synchronously collected in the partial discharge signal both participate in the determination of the correlation, the historical partial discharge signal is a partial discharge signal collected previously, wherein each predetermined fault corresponds to its own historical partial discharge signal, and the historical partial discharge signal includes at least one of the following: a partial discharge signal collected before the fault occurs, a partial discharge signal collected during the fault occurs, and a partial discharge signal collected in a predetermined time period after the fault occurs; A detection result is obtained according to the correlation.

2. The method according to claim 1, characterized in that The at least two signals are at least two of the following signals: ultrasonic signal, high frequency signal, ultra-high frequency signal, radio frequency partial discharge signal, sound print signal, vibration signal, bushing end screen signal, pulse current method partial discharge signal, transient ground wave signal, core addition grounding current signal.

3. The method according to claim 2, characterized in that Acquiring the correlation between the partial discharge signal and the historical partial discharge signal includes: Acquiring a signal type of the partial discharge signal; The correlation of each signal in the partial discharge signal is obtained according to the signal type; wherein the sensor includes at least two of the following: an ultrasonic sensor, a high frequency sensor, an ultra-high frequency sensor, a radio frequency sensor, a sound print sensor, a vibration sensor, a casing end screen sensor, a partial discharge input unit, a geomagnetic wave sensor, and a ground current sensor.

4. The method according to claim 1, characterized in that: Acquiring the correlation between the partial discharge signal and the historical partial discharge signal comprises: fusing multiple signals in the historical partial discharge signal according to the correlation to obtain historical fused data; fusing multiple signals in the partial discharge signal according to the correlation to obtain fused data, wherein the partial discharge signal and the multiple signals in the historical partial discharge signal are fused in the same manner; Obtaining the detection result according to the correlation includes: obtaining the detection result according to the fusion data and historical fusion data.

5. The method according to claim 4, characterized in that Obtaining the detection result according to the fused data and the historical fused data includes: The historical fusion data and the corresponding predetermined faults are used as training data to train a fault detection model; The fused data is input into the fault detection model to obtain the detection result.

6. The method according to claim 5, characterized in that The step of fusing multiple signals in the partial discharge signal according to correlation to obtain fused data includes: Under each predetermined fault, obtaining a characteristic sequence corresponding to each signal in the partial discharge signal; According to the correlation between the partial discharge signal of each signal type and the predetermined fault and a preset correlation threshold, the feature sequence corresponding to each signal in the partial discharge signal is divided into a main feature sequence and a secondary feature sequence; wherein, if the correlation between a signal and the predetermined fault is less than the correlation threshold, the feature sequence corresponding to the signal is the secondary feature sequence; if the correlation between a signal and the predetermined fault is greater than or equal to the correlation threshold, the feature sequence corresponding to the signal is the main feature sequence; The main feature sequence and the secondary feature sequence are fused to obtain fused data under each predetermined fault.

7. The method according to claim 6, characterized in that Acquiring a characteristic sequence corresponding to each signal in the partial discharge signal comprises: Collect each signal in the partial discharge signal at a predetermined frequency to obtain a data sequence corresponding to each signal; When the correlation of a data sequence corresponding to a collected signal is lower than a preset condition, data enhancement is used to perform data expansion on the data sequence to obtain an expanded data sequence; A corresponding feature sequence is obtained according to the data sequence and the expanded data sequence.

8. The method according to claim 6, characterized in that The main feature sequence and the secondary feature sequence are fused to obtain the fused data, including: The main feature sequence and the secondary feature sequence are fused according to the channel number to form a fused data matrix.

9. The method according to claim 6, characterized in that The historical fusion data and the corresponding predetermined faults are used as training data to train a fault detection model, including: Using the main feature sequence and the secondary feature sequence to train the first fault detection model to obtain a trained first fault detection model; The second fault detection model is trained by using the secondary feature sequence to obtain a trained second fault detection model; wherein, when the main feature sequence is missing in the fused data corresponding to the local discharge signal, the second fault detection model is used to output the detection result; and when the secondary feature sequence is missing or exists in the fused data corresponding to the local discharge signal, the first fault detection model is used.

10. An active protection live detection device for main substation equipment, characterized in that: include: Multiple channels, each channel is used to access the signal of a sensor; A processor, used to execute software, wherein the software is used to execute the active defense live detection method for substation main equipment as described in any one of claims 1 to 9.

Images