A method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis

Abstract

The application discloses a kind of based on transient signal analysis's mutual inductor dynamic characteristic detection method, belong to mutual inductor detection technical field, comprising: acquisition equipment acquires the first signal data of mutual inductor, and from the first signal data extraction has transient characteristic transient signal data, to the transient signal data extracted pre-processing;Dynamic characteristic parameter extraction is carried out to the transient signal after pre-processing;Dynamic characteristic and the matching mapping relationship of failure mode are established, and failure mode library is constructed;Deviation check is carried out to the dynamic characteristic parameter of transient signal;The dynamic characteristic detection result of mutual inductor is output, and grading early warning is carried out.In the implementation process of the technical scheme of the present application, the transient signal data of mutual inductor is extracted, and different dimensions of dynamic characteristic parameters are extracted according to the transient signal data, so as to judge the fault type of mutual inductor, improve the operation stability of power system.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of smart grid, in particular to a mutual inductor dynamic characteristic detection method based on transient signal analysis. BACKGROUND

[0002] The mutual inductor is a key equipment in the power system, mainly applied in the smart grid, and is divided into two categories of current mutual inductor and voltage mutual inductor, mainly applied in the measurement of current and voltage, current-voltage conversion and other scenes, and is a key equipment to realize the stability of the power system.

[0003] The mutual inductor has static characteristics and dynamic characteristics, wherein the static characteristics mainly refer to the output characteristics when the input signal is stable or changes slowly, mainly reflecting the parameters such as sensitivity and linearity of the mutual inductor, and the dynamic characteristics mainly refer to the output response when the input signal changes rapidly, including key indicators such as response time and frequency response range. In the prior art, there are mature detection methods for the static characteristics of the mutual inductor, such as through steady-state signal test and analog load test, while for the detection of dynamic characteristics, the prior art usually adopts laboratory offline detection method, such as using a standard mutual inductor with known dynamic characteristics as a reference to compare and analyze with the mutual inductor to be tested, or establishing a model of the mutual inductor to be tested through electromagnetic transient simulation software, and judging the dynamic characteristics of the mutual inductor after simulation processing. These methods have been widely applied at present.

[0004] However, in the application process of the above technical solutions, there are still some problems, for example, in the face of complex working conditions, such as high proportion of new energy, flexible direct current transmission, high voltage direct current power grid and other new type power system scenes, the above method will have problems such as limited frequency band range and insufficient response speed, and for some complex scenes of multi-source disturbance, the above detection means is difficult to fully capture the dynamic response characteristics of the mutual inductor, resulting in low detection accuracy and efficiency. SUMMARY

[0005] The technical problem to be solved by the present application is to overcome the defects of the prior art, and to provide a mutual inductor dynamic characteristic detection method based on transient signal analysis, which realizes the fault detection function of the mutual inductor by analyzing the transient signal of the mutual inductor.

[0006] In order to solve the above technical problems, the technical scheme of the present application is: a mutual inductor dynamic characteristic detection method based on transient signal analysis, comprising:

[0007] The acquisition device acquires first signal data of the mutual inductor, and extracts transient signal data with transient characteristics from the first signal data, and pre-processes the extracted transient signal data;

[0008] The dynamic characteristic parameter extraction is performed on the preprocessed transient signal, and the dynamic characteristic parameter extraction includes three dimensions of time domain, frequency domain and time-frequency domain;

[0009] A matching mapping relationship between the dynamic characteristic and the fault mode is established, a fault mode library is constructed, the dynamic parameters of the current transient signal are compared with the standard modes in the fault mode library by combining a matching algorithm, and the fault type is determined through similarity calculation;

[0010] The dynamic characteristic parameters of the transient signal are subjected to deviation checking, and the reference value of the dynamic parameter is corrected according to the rated parameter and the real-time operation data of the transformer;

[0011] The dynamic characteristic detection result of the transformer is output, and a graded early warning is performed, and the dynamic characteristic detection result includes the dynamic characteristic parameter, the fault type determination result and the fault timestamp.

[0012] In the implementation process of the technical scheme of the present application, the transient signal data of the transformer is extracted, and the dynamic characteristic parameters are extracted in different dimensions according to the transient signal data, so as to judge the fault type of the transformer and improve the operation stability of the power system.

[0013] Further, the transient signal data with transient characteristics is extracted from the first signal data, which further includes:

[0014] Multi-dimensional feature extraction is performed on the first signal data, and a separate quantization index is established for each dimension, wherein the multi-dimensional feature extraction includes four dimensions of time domain, frequency domain, time-frequency domain and statistical distribution;

[0015] The first signal data is individually analyzed according to the quantization index established for each dimension, and a separate transient feature library is established, and the transient signal screened from each dimension is stored in the corresponding transient feature library;

[0016] The transient signals stored in the transient feature library are subjected to intersection and union set analysis to determine the commonality and difference of the transient signals under each dimension, and the transient signals under each dimension are clustered;

[0017] Based on the intersection and union set analysis result, a comprehensive feature map is generated, which integrates the transient information of each dimension and displays the space-time distribution of the transient signal through visual application.

[0018] Further, establishing a separate quantization index for each dimension includes: establishing multi-scale difference mutation intensity and local extreme value density as quantization indexes for time domain dimension; establishing high-frequency energy proportion index for frequency domain dimension; establishing time-frequency energy concentration and time-frequency gradient for time-frequency domain dimension, calculating the energy proportion of the time-frequency spectrum near the center frequency of the wavelet basis; and establishing probability density function difference index for statistical distribution dimension, and calculating the skewness and kurtosis of the probability density function.

[0019] Further, the multi-scale differential mutation intensity calculates the differential absolute value of the signal data at different time scales, and counts the mutation peak value at each time scale, the mutation peak value refers to the signal point exceeding 3 times the standard deviation of the mean, and the local extreme value density is counted by counting the number of extreme value points in a fixed sliding window, and the local extreme value density is calculated according to the number of extreme value points.

[0020] Further, the wavelet basis center frequency refers to the core frequency used for decomposing the signal in the wavelet transform, and different wavelet basis center frequencies need to be selected to adapt to the characteristics of different signals in different application scenarios.

[0021] Further, in the time domain dimension, the rise time, peak time and decay time constant of the transient signal are calculated, in the frequency domain dimension, the high frequency energy proportion and harmonic distortion rate are calculated to analyze the high frequency component distribution of the transient signal, and in the time-frequency domain dimension, the energy diffusion rate and main frequency drift are extracted through the time-frequency spectrum of the continuous wavelet transform.

[0022] Further, the rise time refers to the time required to rise from 10% to 90% of the steady state value, the peak time refers to the time when the signal reaches the maximum value, and the decay time constant refers to the time required for the transient amplitude to decay to a certain proportion of the initial value.

[0023] Further, the energy diffusion rate refers to the percentage of high frequency energy decay per unit time, and the main frequency drift refers to the change range of the main frequency in the transient process.

[0024] Further, the fault mode library contains dynamic characteristic templates of common types of mutual inductors, and the matching algorithm is a distance-based classifier.

[0025] Further, the deviation check of the dynamic characteristic parameters includes an online parameter calibration and a multi-source data verification process, wherein the online parameter calibration dynamically adjusts the parameter reference value by monitoring the mutual inductor operation data in real time, and verifies whether the dynamic characteristics of the transient signal conform to the physical law through correlation analysis.

[0026] By adopting the above technical scheme, the present application has the following beneficial effects: the transient signal data of the mutual inductor is extracted, and different dimension dynamic characteristic parameters are extracted according to the transient signal data, so as to judge the fault type of the mutual inductor, and improve the operation stability of the power system. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0028] The present application provides a kind of, the person skilled in the art can draw lessons from the content herein, realize by improving process parameters appropriately.It is particularly pointed out that all similar substitutions and changes are obvious to the person skilled in the art, and they all belong to the scope protected by the present application.The method and application of the present application have been described by the preferred embodiments, and the relevant personnel can obviously modify or appropriately change and combine the method and application herein without departing from the content, spirit and scope of the present application to realize and apply the present application technology.

[0029] In order to make the content of the present application more easily understood, the present application will be further described in detail below according to specific embodiments and in conjunction with the accompanying drawings.

[0030] As Figure 1 shown, the mutual inductor of the present application is mainly applied in smart grid, a mutual inductor dynamic characteristic detection method based on transient signal analysis is proposed, the dynamic characteristics of the mutual inductor are detected, and the fault causes are identified in time, aiming at the problems of limited frequency band range, insufficient response speed and the like in the prior art, the transient signal analysis is carried out to capture the dynamic response characteristics of the mutual inductor under complex working conditions, improve the detection precision and efficiency, and ensure the stable operation of the power system, specifically, the mutual inductor dynamic characteristic detection method comprises the following steps:

[0031] Step 01: The acquisition device acquires the first signal data of the mutual inductor, and extracts the transient signal data with transient characteristics from the first signal data, and pre-processes the extracted transient signal data;

[0032] The mutual inductor is divided into current transformer and voltage transformer, whether it is current transformer or voltage transformer, signal data such as current amplitude, phase change will be generated in the running process, these signal data not only contain transient signals, but also contain steady-state signals, and in the present embodiment, since the transient signal can reflect the dynamic characteristics of the mutual inductor, it is detected, therefore, the transient signal data needs to be extracted from the collected first signal data, and subsequent pretreatment is needed;

[0033] Among them, the acquisition device can be a high-precision data acquisition instrument with multi-channel synchronous sampling function, or a sensor array, which can simultaneously collect multiple signal data, and the acquisition device can be integrated or connected with a data processing center to analyze the collected data subsequently, for example, the acquisition device is connected with an industrial computer or other devices with data processing and analysis function through wired or wireless connection, so as to realize a series of processing processes of mutual inductor signal data, which is not limited in the present embodiment;

[0034] Specifically, extracting the transient signal data with transient characteristics from the first signal data further comprises the following steps:

[0035] Step 101: multi-dimensional feature extraction is performed on the first signal data, and a separate quantitative index is established for each dimension, wherein the multi-dimensional feature extraction includes four dimensions of time domain, frequency domain, time-frequency domain and statistical distribution;

[0036] The essential difference between the transient signal and the steady-state signal of the transformer is the difference in signal mutation type, energy distribution and time-varying law. The traditional separation method usually relies on prior models, but this method requires a large amount of computing resources, and it is difficult to respond in real time in actual application. Therefore, in this embodiment, feature extraction is performed on the differences exhibited by the signal data in different dimensions, and a separate quantitative index is established for each dimension, so as to separate the transient signal from the characteristics of the signal data itself:

[0037] Among them, the multi-dimensional feature extraction respectively contains four dimensions of time domain, frequency domain, time-frequency domain and statistical distribution. The time domain mainly reflects the mutability and non-stationarity of the signal data, such as step change, spike pulse, etc. of voltage and current. The frequency domain mainly reflects the proportion of high-frequency energy of the signal data, and the high frequency is specifically the frequency component in the range of 100Hz to 10kHz. The time-frequency domain mainly judges the energy diffusion mode of the signal data, judges the concentration and dispersion of energy in different time periods, and the statistical distribution mainly judges the probability distribution characteristics of the signal data, such as mean, variance, etc. Through signal analysis in different dimensions, the transient signal in the first signal data can be separated by finally summarizing the analysis results of each dimension. Compared with the traditional modeling method, this method not only saves computing resources, but also can respond in real time. As long as the first signal data collected by the collection device is received, real-time analysis can be performed, and there is no need to wait for the completion of the transmission of all data, thereby improving the data analysis efficiency;

[0038] Establishing a separate quantitative index for each dimension includes: establishing multi-scale difference mutation intensity and local extreme density as quantitative indexes for the time domain dimension, wherein the multi-scale difference mutation intensity calculates the difference absolute value of the signal data in different time scales, and counts the mutation peak value in each time scale. The mutation peak value refers to a signal point that is at least 3 times the standard deviation of the mean. The local extreme density counts the number of extreme points (including maximum and minimum) in a fixed sliding window, and calculates the local extreme density according to the number of extreme points. In a normal case, the extreme density (number of extreme points per unit time) of the transient region is significantly more than that of the steady-state region. Because the steady-state signal has periodicity, the extreme points are evenly distributed, while the transient signal shows burstiness and irregularity, resulting in the concentration of extreme points.

[0039] An energy proportion of high frequency index is established for the frequency domain dimension, and the proportion of the energy of the signal in the high frequency band (above 100 Hz) to the total energy is calculated. Generally, the energy of the steady-state signal is concentrated in the low frequency band, for example, 45 to 55 Hz, while the transient signal contains rapidly changing electromagnetic processes, which will produce additional energy in the high frequency band (above 100 Hz). Therefore, after designing the high frequency energy proportion index, the high frequency energy proportion of the transient signal is significantly higher than that of the steady-state signal, thereby separating the steady-state signal and the transient signal in the frequency domain;

[0040] A time-frequency energy concentration and a time-frequency gradient are established for the time-frequency domain dimension, and the proportion of the energy of the time-frequency spectrum near the center frequency of the wavelet basis is calculated. For example, for a specific frequency component, the higher the time-frequency energy concentration, the more concentrated the energy of the frequency component in a certain time period. The time-frequency energy concentration of the transient signal will decrease significantly over time, while the steady-state signal will remain stable. The time-frequency gradient refers to the calculation of the time gradient and the frequency gradient of the time-frequency spectrum. The time-frequency gradient change of the transient signal will be significantly higher than that of the steady-state signal.

[0041] It should be noted that the center frequency of the wavelet basis refers to the core frequency used to decompose the signal in the wavelet transform. In different application scenarios, a suitable wavelet basis center frequency needs to be selected to adapt to the characteristics of different signals. For example, for an exponentially decaying transient signal, the Morlet wavelet basis can be selected, and the center frequency corresponds to the transient dominant frequency. For an oscillatory decaying transient signal, the Daubechies wavelet basis can be selected, and the center frequency corresponds to the oscillatory dominant frequency, ensuring the accuracy of time-frequency analysis.

[0042] A probability density function difference index is established for the statistical distribution dimension, and the skewness and kurtosis of the probability density function are calculated. In the transient signal, due to the inclusion of random disturbances and nonlinear processes, the amplitude distribution will approach a heavy-tailed distribution, such as a Laplace distribution or a Gaussian mixture distribution. The amplitude distribution of the steady-state signal is closer to a normal distribution. By comparing the skewness and kurtosis, the transient and steady-state signals can be effectively distinguished. After calculating the skewness and kurtosis of the probability density function, the quantile interval is determined. The quantile interval refers to the distance between specific quantiles in the probability distribution. The transient signal is greater than the steady-state signal in the quantile interval, and the amplitude fluctuation range is larger, thereby further enhancing the recognition ability of the transient and steady-state signals.

[0043] After establishing separate quantitative indicators for each dimension as described above, the transient and steady-state signals can be identified from different dimensions, and only the characteristics of the signal data itself need to be considered, without relying on external parameters, reducing the risk of misjudgment. Compared with traditional modeling methods, complex parameter adjustment and high computational power requirements are not needed, improving analysis efficiency.

[0044] Step 102: Analyze the first signal data according to the quantification indicators established for each dimension separately, and establish separate transient feature libraries, and store the transient signals screened from each dimension into the corresponding transient feature library;

[0045] When performing multi-dimensional feature extraction, each dimension is analyzed separately, and a transient feature library is established, and the transient signals screened from each dimension are stored in the corresponding transient feature library one by one, ensuring the completeness and independence of the feature library, and facilitating subsequent comprehensive analysis of multiple dimensions;

[0046] Step 103: Perform intersection and union analysis on the transient signals stored in the transient feature library to determine the commonality and differences of the transient signals under each dimension, and cluster the transient signals under each dimension;

[0047] Performing intersection and union analysis on the transient data extracted under different dimensions of the transformer can achieve more comprehensive transient feature analysis. For example, when performing intersection application, transient features common to different dimensions can be screened out, thereby eliminating pseudo-transient signal data in single-dimensional data and improving the accuracy of transient signal identification. In union application, the information dimension can be expanded to capture details missing in a single data set, thereby integrating multi-dimensional features and constructing a more complete transient database. For example, after locking the fault interval through intersection application, the transient differences of each dimension in the interval, such as the energy distribution of CT and PT, are analyzed to accurately determine the fault type. When the fault type is a grounding fault, by comparing the energy distribution differences between CT and PT, it can be accurately determined whether the fault is a metallic grounding fault or an arc grounding fault.

[0048] Step 104: Based on the intersection and union analysis results, generate a comprehensive feature map that integrates transient information from each dimension and displays the temporal and spatial distribution of transient signals through visual application.

[0049] After performing intersection and union analysis, the feature maps of each dimension are superimposed to form a multi-dimensional comprehensive feature map that includes time, frequency, energy, and other multi-dimensional information, which intuitively displays the dynamic changes and spatial distribution of transient signals, facilitating rapid positioning of fault points and analysis of fault types. In addition, the temporal and spatial distribution of transient signals can be obtained through visual application, such as displaying energy distribution of different dimensions through color changes combined with time axis analysis to capture the characteristic changes at the fault instant. Compared with existing technologies, this method can shorten the fault judgment time and reduce the occurrence of misjudgment.

[0050] The pre-processing of the extracted transient signals includes denoising, normalization, and other processes. For details, refer to the signal data processing methods in existing technologies, which will not be described in detail in this embodiment.

[0051] Step 02: Dynamic characteristic parameter extraction is performed on the preprocessed transient signal, which includes three dimensions of time domain, frequency domain, and time-frequency domain;

[0052] The preprocessed transient signal also needs to further extract dynamic response parameters, so as to quantify the performance of the transformer in the transient process, as the basis for subsequent fault diagnosis. Among them, the dynamic characteristic parameter extraction only includes three dimensions of time domain, frequency domain, and time-frequency domain, and does not include the statistical distribution dimension. Because the statistical distribution dimension more reflects the overall trend, it is difficult to capture the instantaneous change, which is not conducive to accurate diagnosis of transient faults, so it is not considered. Specifically:

[0053] In the time domain dimension, the rise time, peak time, and decay time constant of the transient signal are calculated. The rise time refers to the time required to rise from 10% to 90% of the steady-state value, the peak time refers to the time when the signal reaches the maximum value, and the decay time constant refers to the time required for the transient amplitude to decay to a certain proportion of the initial value, for example, the decay time constant of the excitation surge current is usually 0.1 to 1 second, and the decay time constant of the short-circuit current is smaller, thereby determining the transient type. In the frequency domain dimension, the high-frequency energy proportion and harmonic distortion rate are calculated to analyze the high-frequency component distribution of the transient signal. The high-frequency energy proportion has been calculated in the foregoing process, and the harmonic distortion rate can be obtained according to the calculation method in the prior art. Through harmonic distortion rate analysis, the amount of harmonic components in the signal can be determined, and the fault type can be further identified, for example, the transient signal caused by CT saturation usually contains a large amount of even harmonics (such as 2 times, 4 times), and the transient signal of the iron core loosening may be accompanied by abnormal fluctuations (such as amplitude mutation ± 15% or more) of the fundamental frequency component. In the time-frequency domain dimension, the energy diffusion rate and the main frequency drift are extracted through the time-frequency spectrum of continuous wavelet transform. The energy diffusion rate refers to the percentage of high-frequency energy decay per unit time, and the main frequency drift refers to the change range of the main frequency in the transient process. For example, the main frequency of the transient signal caused by distributed capacitance discharge will quickly drift from several hundred Hz to several tens of Hz, while the main frequency of the lightning impulse transient is basically stable. Through the analysis of these multi-dimensional parameters, different fault types and their evolution processes can be more accurately identified, thereby providing a comprehensive and reliable basis for fault diagnosis and effectively improving the stability and safety of the system.

[0054] Step 03: Establish a matching mapping relationship between dynamic characteristics and fault modes, and construct a fault mode library. The dynamic parameters of the current transient signal are compared with the standard modes in the fault mode library through a matching algorithm, and the fault type is determined through similarity calculation.

[0055] After the dynamic characteristic parameters of the mutual inductor are extracted, fault diagnosis can be performed according to the parameters. In the prior art, the threshold method is usually used for fault diagnosis, that is, the extracted parameters are compared with preset threshold values, and if the parameters exceed the range, it is determined that an abnormality occurs. However, this method has limitations and cannot adapt to complex and variable fault conditions, which can easily lead to misjudgment or missed judgment. Therefore, in the embodiment, a matching mapping relationship between the dynamic characteristics and the fault modes is established, and then the dynamic parameters of the current transient signal are compared with the standard modes in the fault mode library to determine the fault type.

[0056] Specifically, the fault mode library includes dynamic characteristic templates of common types of mutual inductors, for example, an excitation inrush current fault, the characteristic of which is that the decay time constant is greater than 0.5 seconds, the high-frequency energy ratio is greater than 30%, and the odd harmonic distortion rate is greater than 40%; a CT saturation fault, the characteristic of which is that the transient amplitude mutation is greater than 200%, the main frequency is concentrated in 100 to 300 Hz, and the third harmonic ratio in the harmonic is greater than 25%; a core loose fault, the characteristic of which is that the amplitude fluctuation of the fundamental frequency in the transient process is greater than 15%, the time-frequency energy concentration rate decreases by more than 20%, and the main frequency drifts in the range of 50 to 100 Hz; and an edge aging fault, the characteristic of which is that the transient signal contains high-frequency pulses greater than 10 kHz, the pulse duration is less than 1 millisecond, and the quantile interval is greater than twice the steady-state. The above are common fault modes and their characteristic parameters, which can be adjusted or added in actual application. Specifically, reference can be made to the technical standards and practical experience in the field, and the embodiment will not be illustrated one by one.

[0057] After the fault mode library is constructed, the dynamic parameters of the current transient signal need to be compared with the standard modes in the library by using a matching algorithm to determine the fault type. The matching algorithm is a distance-based classifier, for example, a K-nearest neighbor algorithm, which selects a number of modes with the smallest distance as candidate fault types by calculating the Euclidean distance between the to-be-tested sample and each standard mode in the library, or a support vector machine algorithm, which classifies by finding an optimal hyperplane. Finally, the fault type is determined. Specifically, reference can be made to the prior art, and the embodiment will not be described in detail.

[0058] Step 04: Deviation check on the dynamic characteristic parameters of the transient signal, the reference value of the dynamic parameters is corrected according to the rated parameters and real-time operation data of the mutual inductor.

[0059] In order to avoid the drift of dynamic parameters caused by the change of the operating state of the mutual inductor, for example, the parameter deviation caused by temperature rise, real-time calibration of the dynamic characteristic parameters of the transient signal is also needed, and the reference value of the dynamic parameters is corrected to maintain the accuracy and reliability of the diagnosis and ensure the accuracy of fault identification is not disturbed by external factors, thereby improving the overall stability and operating efficiency of the system.

[0060] Specifically, the deviation check of dynamic characteristic parameters includes online parameter calibration and multi-source data verification process. The online parameter calibration dynamically adjusts the parameter reference value by monitoring the operating data of the transformer in real time. For example, when the primary current increases from 5A to 30A, the excitation inductance of the CT will decrease, resulting in a shortened decay time constant. At this time, the calculation formula of the decay time constant needs to be adjusted according to the real-time load rate to adapt to the actual operating conditions. The multi-source data verification synchronously collects auxiliary signals such as the primary side bus current and the secondary side load current of the transformer, and verifies whether the dynamic characteristics of the transient signals conform to the physical laws through correlation analysis. For example, when the CT saturation fault occurs, the correlation coefficient of the primary side current and the secondary side output current should be significantly reduced (less than 0.7), while under normal working conditions, the correlation coefficient should be greater than 0.95. If the verification fails, it means that there is a deviation in the dynamic characteristic parameters, which needs to be recalibrated or adjusted to match the algorithm.

[0061] Step 05: Output the dynamic characteristic detection results of the transformer and perform hierarchical early warning. The dynamic characteristic detection results include dynamic characteristic parameters, fault type determination results, and fault time stamps.

[0062] After the above steps, the dynamic characteristic detection results of the transformer can be generated and output. The detection results include dynamic characteristic parameters such as rise time, decay time constant, and high-frequency energy proportion, as well as fault type determination results and fault occurrence time stamps. The fault type determination results include excitation inrush fault, CT saturation fault, and corresponding confidence levels. The fault occurrence time stamps record the specific time of fault occurrence, such as fault occurrence time, duration, and corresponding waveform segments, so that the maintenance personnel can quickly locate the fault and effectively handle it.

[0063] At the same time, hierarchical early warning is performed according to the severity of the fault. For example, if the core loosening causes the secondary side output to fluctuate violently, it is considered as a first-level early warning. If the CT saturation causes the differential current to exceed the limit, it is considered as a second-level early warning. If early signs of insulation aging occur, such as occasional high-frequency pulses, it is considered as a third-level early warning. Through different levels of early warning prompts, the maintenance personnel can take targeted preventive measures to avoid the expansion of the fault and ensure the safe and stable operation of the system.

[0064] Each level of early warning event can be set as needed, including early warning threshold, response time, handling suggestions, etc., to ensure the accuracy and timeliness of the early warning information. Each level of early warning event generates a log automatically after occurrence and pushes it to the operation and maintenance platform for real-time monitoring and subsequent analysis. The log content covers early warning level, triggering condition, handling status, and other detailed information.

[0065] The above-described specific embodiments further illustrate the technical problems solved by the present application, technical solutions and beneficial effects, and it should be understood that the above-described are only specific embodiments of the present application and are not intended to limit the present application, and any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting dynamic characteristics of a transformer based on transient signal analysis, characterized in that: The method comprises the following steps: The acquisition device acquires the first signal data of the mutual inductor, extracts transient signal data with transient characteristics from the first signal data, and pre-processes the extracted transient signal data; Dynamic characteristic parameter extraction is performed on the pre-processed transient signal, which includes three dimensions of time domain, frequency domain and time-frequency domain; A matching mapping relationship between dynamic characteristics and fault modes is established, and a fault mode library is constructed, the dynamic parameters of the current transient signal are compared with the standard modes in the fault mode library by combining a matching algorithm, and the fault type is determined by similarity calculation; The dynamic characteristic parameters of the transient signal are subjected to deviation checking, and the reference value of the dynamic parameter is corrected according to the rated parameter and the real-time operation data of the mutual inductor; The dynamic characteristic detection result of the mutual inductor is output, and a hierarchical early warning is performed, the dynamic characteristic detection result includes dynamic characteristic parameters, fault type determination result and fault timestamp; The transient signal data with transient characteristics is further extracted from the first signal data, which comprises the following steps: Multi-dimensional feature extraction is performed on the first signal data, and a separate quantization index is established for each dimension, wherein the multi-dimensional feature extraction includes four dimensions of time domain, frequency domain, time-frequency domain and statistical distribution; The first signal data is analyzed separately according to the quantization index established for each dimension, and a separate transient feature library is established, and the transient signals screened from each dimension are stored in the corresponding transient feature library; The transient signals stored in the transient feature library are subjected to intersection and union set analysis to determine the commonality and difference of the transient signals under each dimension, and the transient signals under each dimension are clustered; Based on the intersection and union set analysis result, a comprehensive feature map is generated, which integrates the transient information of each dimension, and the time and space distribution of the transient signal is displayed through visual application; For each dimension, a separate quantization index is established, which includes: for the time domain dimension, multi-scale differential mutation intensity and local extreme value density are established as quantization indexes; for the frequency domain dimension, a high-frequency energy proportion index is established; for the time-frequency domain dimension, time-frequency energy concentration and time-frequency gradient are established, the energy proportion of the time-frequency spectrum near the wavelet base center frequency is calculated; for the statistical distribution dimension, a probability density function difference index is established, and the skewness and kurtosis of the probability density function are calculated.

2. The method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis according to claim 1, characterized in that: The multi-scale differential mutation intensity calculates the differential absolute value of the signal data at different time scales, and counts the mutation peak value at each time scale, the mutation peak value refers to the signal point which is more than 3 times the standard deviation of the mean value, the local extreme value density is calculated by counting the number of extreme value points in a fixed sliding window, and the local extreme value density is calculated according to the number of extreme value points.

3. The method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis according to claim 1, characterized in that: The wavelet base center frequency refers to the core frequency used for decomposing the signal in wavelet transform, in different application scenarios, different wavelet base center frequencies need to be selected to adapt to the characteristics of different signals.

4. The method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis according to claim 1, characterized in that: In the time domain dimension, the rise time, peak time and decay time constant of the transient signal are calculated; in the frequency domain dimension, the high-frequency energy proportion and harmonic distortion rate are calculated to analyze the high-frequency component distribution of the transient signal; in the time-frequency domain dimension, the energy diffusion rate and main frequency drift are extracted through the time-frequency spectrum of continuous wavelet transform.

5. The method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis according to claim 4, characterized in that: Wherein the rise time refers to the time required to rise from 10% to 90% of the steady-state value, the peak time refers to the time when the signal reaches the maximum value, and the decay time constant refers to the time required for the transient amplitude to decay to a certain proportion of the initial value.

6. The method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis according to claim 4, characterized in that: Wherein the energy diffusion rate refers to the percentage of high-frequency energy attenuation per unit time, and the main frequency drift refers to the change range of the main frequency during the transient process.

7. The method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis according to claim 1, characterized in that: The fault mode library contains dynamic characteristic templates of common types of transformers, and the matching algorithm is a distance-based classifier. The dynamic characteristic templates include excitation inrush fault, CT saturation fault, core looseness fault, and edge aging fault.

8. The method for detecting dynamic characteristics of a mutual inductor based on transient signal analysis according to claim 1, characterized in that: The deviation check of dynamic characteristic parameters includes online parameter calibration and multi-source data verification process. The online parameter calibration dynamically adjusts the parameter reference value by real-time monitoring of transformer operation data, and verifies whether the dynamic characteristics of the transient signal conform to the physical law through correlation analysis.

Images