Partial discharge diagnosis method and device for gas insulated switchgear

Abstract

The invention discloses a partial discharge diagnosis method and device for gas insulated switchgear. The method comprises the following steps: acquiring an original partial discharge signal collected for the gas insulated switchgear; performing modal decomposition on the original partial discharge signal to obtain a plurality of target intrinsic modal functions; obtaining a time-frequency spectrum for representing the instantaneous frequency and amplitude of the original partial discharge signal based on the plurality of target intrinsic mode functions; based on the time-frequency spectrum, constructing a phase analysis partial discharge spectrum and a phase analysis pulse sequence spectrum; based on the phase analysis partial discharge map and the phase analysis pulse sequence map, a pre-trained partial discharge diagnosis model is adopted to obtain a partial discharge diagnosis result of the gas insulated switchgear, and the partial discharge diagnosis result is at least used for indicating the partial discharge type of the gas insulated switchgear. According to the invention, the technical problem of low accuracy of partial discharge detection and diagnosis of the gas insulated switchgear (GIS) in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of smart grids, and in particular to a method and device for diagnosing partial discharge of a gas-insulated switchgear. Background Art

[0002] Gas-insulated switchgear (GIS) equipment, an indispensable component of modern power systems, is widely used in high-voltage transmission and distribution. Its safe operation is crucial for ensuring grid stability. However, GIS equipment can be susceptible to internal partial discharge (PD) over long periods of operation, which not only degrades its insulation performance but can also lead to serious electrical failures. Therefore, accurate and timely detection and diagnosis of PD in GIS is crucial for preventing and controlling power system failures.

[0003] Partial discharge detection methods in related technologies, such as frequency domain analysis based on Fourier transform and wavelet transform, can identify discharge signals to a certain extent, but they rely on fixed window lengths and preset basis functions, making it difficult to process nonlinear and non-stationary partial discharge signals. In particular, in complex electromagnetic environments, signal interference and aliasing seriously restrict the accuracy of detection and diagnosis. In addition, the performance and accuracy of image-based diagnostic models in related technologies, such as convolutional neural networks (CNNs), will also be significantly reduced when faced with overlapping pulse signals. In particular, in complex scenarios where multi-source partial discharge signals are processed, the limitations of the methods in related technologies are more prominent. In summary, the related technologies have technical problems such as low accuracy and poor applicability in the detection and diagnosis of partial discharge in gas insulated switchgear (GIS).

[0004] To address the above-mentioned problems, no effective solutions have been proposed so far. Summary of the Invention

[0005] Embodiments of the present invention provide a method and apparatus for diagnosing partial discharge of a gas insulated switchgear, to at least solve the technical problem of low accuracy in detecting and diagnosing partial discharge of a gas insulated switchgear (GIS) in the related art.

[0006] According to one aspect of an embodiment of the present invention, a method for diagnosing partial discharge of a gas-insulated switchgear is provided, comprising: obtaining an original partial discharge signal collected from the gas-insulated switchgear; performing modal decomposition on the original partial discharge signal to obtain multiple target intrinsic mode functions; obtaining a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal based on the multiple target intrinsic mode functions; constructing a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum based on the time-frequency spectrum, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude of the discharge pulse and the phase, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse; based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum, using a pre-trained partial discharge diagnostic model, obtaining a partial discharge diagnostic result of the gas-insulated switchgear, wherein the partial discharge diagnostic result is used to at least indicate the type of partial discharge of the gas-insulated switchgear.

[0007] According to another aspect of an embodiment of the present invention, a partial discharge diagnostic device for a gas-insulated switchgear is provided, comprising: a signal acquisition module for acquiring an original partial discharge signal collected from the gas-insulated switchgear; a decomposition module for performing modal decomposition on the original partial discharge signal to obtain a plurality of target intrinsic mode functions; a time-frequency spectrum acquisition module for obtaining, based on the plurality of target intrinsic mode functions, a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal; a spectrum construction module for constructing, based on the time-frequency spectrum, a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude of a discharge pulse and the phase, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse; and a diagnosis result determination module for obtaining, based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum, a partial discharge diagnosis result of the gas-insulated switchgear using a pre-trained partial discharge diagnosis model, wherein the partial discharge diagnosis result is used to at least indicate the type of partial discharge of the gas-insulated switchgear.

[0008] According to another aspect of an embodiment of the present invention, a non-volatile storage medium is provided. The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded and executed by a processor for any one of the partial discharge diagnosis methods for gas-insulated switchgear.

[0009] According to another aspect of an embodiment of the present invention, an electronic device is also provided, comprising one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors implement any one of the partial discharge diagnostic methods for gas-insulated switchgear.

[0010] According to another aspect of an embodiment of the present invention, a computer program product is provided, comprising a computer program, which implements any one of the steps of the partial discharge diagnosis method for gas-insulated switchgear when the computer program is executed by a processor.

[0011] In an embodiment of the present invention, an original partial discharge signal collected from a gas-insulated switchgear is obtained; a modal decomposition is performed on the original partial discharge signal to obtain a plurality of target intrinsic mode functions; a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal is obtained based on the plurality of target intrinsic mode functions; a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum are constructed based on the time-frequency spectrum, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude of the discharge pulse and the phase, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse; based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum, a pre-trained partial discharge diagnostic model is used to obtain the gas-insulated switchgear. The partial discharge diagnosis results of the switchgear are at least used to indicate the partial discharge type of the gas-insulated switchgear, and the purpose of accurately determining the partial discharge diagnosis results is achieved by collecting the original partial discharge signal, applying intrinsic mode decomposition to obtain multiple modal functions, and then constructing a time-frequency spectrum and phase-resolved partial discharge spectrum (PRPD) and phase-resolved pulse sequence spectrum (PRPS), and finally using a pre-trained partial discharge diagnosis model to analyze these spectra. The partial discharge diagnosis results are improved, thereby achieving the technical effect of improving the accuracy of partial discharge detection and diagnosis of gas-insulated switchgear (GIS), and thus solving the technical problem of low accuracy of partial discharge detection and diagnosis of gas-insulated switchgear (GIS) in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0013] Figure 1 is a flow chart of a partial discharge diagnosis method for a gas insulated switchgear according to an embodiment of the present invention;

[0014] Figure 2 1 is a schematic diagram of an optional target intrinsic mode function acquisition process according to an embodiment of the present invention;

[0015] Figure 3 1 is a schematic diagram of an optional partial discharge noise reduction diagnosis process based on a sliding window attention transformation model according to an embodiment of the present invention;

[0016] Figure 4is a flow chart of an optional partial discharge diagnosis method for gas insulated switchgear according to an embodiment of the present invention;

[0017] Figure 5 2 is a schematic diagram of a partial discharge diagnostic device for a gas insulated switchgear according to an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described 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 efforts should fall within the scope of protection of the present invention.

[0019] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] First, to facilitate understanding of the embodiments of the present invention, some of the terms or nouns involved in the present invention are explained below:

[0021] Gas Insulated Switchgear (GIS) is a combination of high-voltage electrical equipment (such as circuit breakers, disconnectors, current transformers, voltage transformers, busbars, cable terminals, etc.) that is fully or partially enclosed in a grounded metal casing and filled with SF6 gas at a certain pressure as an insulating medium and arc extinguishing medium.

[0022] According to an embodiment of the present invention, an embodiment of a method for diagnosing partial discharge of a gas-insulated switchgear is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0023] Figure 1FIG. 1 is a flow chart of a method for diagnosing partial discharge of a gas insulated switchgear according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0024] Step S102, obtaining an original partial discharge signal collected from a gas-insulated switchgear;

[0025] Optionally, step S102 is the starting point of the entire diagnostic process, aiming to obtain raw electrical signal data from gas-insulated switchgear (GIS) equipment, reflecting the internal partial discharge (PD) condition. These signals can be captured, but are not limited to, by ultra-high frequency (UHF) sensors or other types of sensors installed on the GIS equipment. When partial discharge occurs within the GIS equipment, specific electromagnetic pulses are generated. The sensors can detect these pulses and convert them into electrical signals, which are then recorded and collected. The collected signals typically contain the activity of multiple discharge sources, as well as noise and interference from the environment. Therefore, the collected raw partial discharge signals are complex and non-stationary, manifested by temporal variations in the instantaneous frequency and amplitude of the signals.

[0026] Step S104, performing modal decomposition on the original partial discharge signal to obtain a plurality of target intrinsic mode functions;

[0027] It can be understood that by performing modal decomposition on the original partial discharge signal, the complex original signal can be decomposed into a series of simpler and easier to analyze inherent modes, thereby obtaining multiple target intrinsic mode functions (IMFs) that are simpler and easier to analyze.

[0028] In an optional embodiment, modal decomposition is performed on an original partial discharge signal to obtain a plurality of target intrinsic mode functions, including: adding a plurality of different noise signals to the original partial discharge signal to obtain a plurality of noisy signals, wherein the plurality of noisy signals correspond one-to-one to the plurality of different noise signals; performing multi-order modal decomposition based on the plurality of noisy signals to obtain a plurality of initial intrinsic mode functions, wherein the number of the plurality of initial intrinsic mode functions is greater than the number of the plurality of target intrinsic mode functions; determining correlations between the plurality of initial intrinsic mode functions and the original partial discharge signal; and determining, among the plurality of initial intrinsic mode functions, intrinsic mode functions having correlations greater than a preset correlation threshold as the plurality of target intrinsic mode functions.

[0029] Optionally, multiple noisy signals are generated by adding multiple different noise signals to the original partial discharge signal. These noise signals are randomly generated white noise, and their amplitudes and characteristics need to be adjusted according to the characteristics of the original signal and analysis requirements. Each noisy signal is subjected to multi-order modal decomposition to obtain multiple initial intrinsic mode functions (IMFs). This process is repeated until the decomposed IMFs no longer contain decomposable oscillatory components. This process can produce a larger number of initial IMFs than the final target IMFs. After obtaining these initial IMFs, the correlation between these initial IMFs and the original partial discharge signal needs to be determined. In other words, the similarity and correlation between each initial IMF and the original signal need to be evaluated. Based on a preset correlation threshold, IMFs with a high correlation with the original partial discharge signal are selected from the multiple initial IMFs as target IMFs. These IMFs are more likely to contain valid partial discharge signal features. The preset correlation threshold can be set to an empirical value, such as 0.3, to ensure that the selected target IMFs actually carry partial discharge pattern information.

[0030] Through the above steps, stable and clear target intrinsic mode functions can be extracted from the original partial discharge signal. This approach not only reduces modal aliasing but also helps improve the accuracy of subsequent time-frequency analysis and model diagnosis. By adaptively adding and processing noise signals, the decomposition process is made more robust, effectively separating the intrinsic modes of partial discharge from complex signals, providing a high-quality data foundation for subsequent feature extraction and diagnosis. The key to this process is the noise-assisted set decomposition, which ensures the stability and accuracy of the decomposition results, thereby improving the performance of partial discharge type identification.

[0031] Optionally, the correlation may be calculated using, but not limited to, statistical indicators such as correlation coefficient and kurtosis value, which may reflect the degree of matching between the initial intrinsic mode function and the original partial discharge signal in terms of frequency, amplitude and time series.

[0032] In an optional embodiment, multi-order modal decomposition is performed based on multiple noisy signals to obtain multiple initial intrinsic mode functions, including: taking the first-order modal decomposition in the multi-order modal decomposition as the current-order modal decomposition, and performing the current-order modal decomposition in the following manner: performing empirical mode decomposition on the multiple noisy signals to obtain a set of intrinsic mode functions corresponding to each of the multiple noisy signals, wherein the set of intrinsic mode functions includes a plurality of intrinsic mode functions obtained by decomposing the corresponding noisy signals; based on the set of intrinsic mode functions corresponding to each of the multiple noisy signals, obtaining an average intrinsic mode function corresponding to each of the multiple noisy signals, wherein the average intrinsic mode function is obtained by performing empirical mode decomposition on the eigenmode functions of the noisy signals. The method comprises the following steps: obtaining a plurality of intrinsic mode functions included in a state function set by averaging operation; obtaining a plurality of residual signals based on the average intrinsic mode functions corresponding to the respective noisy signals and the original partial discharge signal, wherein the plurality of residual signals correspond to the plurality of noisy signals one-to-one; using the signals obtained by adding noise to the plurality of residual signals as the new plurality of noisy signals, and using the next-order modal decomposition of the current-order modal decomposition as the new current-order modal decomposition, and repeating the above operation until the last-order modal decomposition in the multi-order modal decomposition is reached; using the average intrinsic mode functions corresponding to the multi-order modal decompositions obtained after the last-order modal decomposition is performed as the plurality of initial intrinsic mode functions.

[0033] Optionally, the original partial discharge signal is combined with different randomly generated white noise sequences to form multiple noisy signals. The first-order modal decomposition is set to the current modal decomposition order. Each noisy signal in the set is subjected to empirical mode decomposition (EMD) to obtain a series of intrinsic mode functions. This process generates an IMF set for each noisy signal, each set containing multiple IMFs obtained by EMD decomposition. Based on the intrinsic mode function set of each noisy signal, its average intrinsic mode function is calculated. This is obtained by adding the values ​​of the corresponding positions of all IMFs in the set and then dividing it by the size of the set (i.e., the number of noisy signals). The calculation of the average IMF can reduce the impact of modal aliasing and improve the stability of the decomposition result. The average IMF obtained above is further subtracted from the original partial discharge signal to obtain the residual signal of the current order. The residual signal can reflect the part of the original signal that has not yet been decomposed. The residual signal is used as the new original signal, and adaptive noise is added to it to generate a new noisy signal set. The adaptive noise here is dynamically adjusted based on the characteristics of the residual signal, aiming to more effectively facilitate the decomposition process and prevent the impact of residual noise on subsequent analysis. This order of modal decomposition is then used as the new "current-order modal decomposition," and the above steps are repeated until the predetermined decomposition order is reached or the residual signal meets the termination criteria (such as becoming a monotonic function or only remaining extreme points). When all orders of modal decomposition are complete, the average intrinsic mode functions obtained from each order of decomposition are aggregated to form multiple initial intrinsic mode functions for subsequent time-frequency analysis and partial discharge diagnosis.

[0034] It should be noted that the method for determining multiple initial intrinsic mode functions in this embodiment can be a modal decomposition method based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) algorithm. This method can overcome the shortcomings of traditional EMD in processing complex signals. By adaptively adding noise and iterative decomposition, it can more finely separate the different components in the original signal. In particular, for nonlinear partial discharge signals, it can effectively extract their inherent oscillation modes, providing a higher-quality signal foundation for subsequent feature analysis and intelligent diagnosis. The above method not only improves the accuracy of partial discharge detection, but also reduces the amount of computation required during the analysis process.

[0035] Optional, Figure 2 FIG. 1 is a schematic diagram of an optional target intrinsic mode function acquisition process according to an embodiment of the present invention, such as Figure 2 As shown, the method flow can be implemented by the following steps:

[0036] S21: Initialization, including:

[0037] Input the original partial discharge signal x(t), set the number of collections N (such as 100 times) and the noise amplitude coefficient sequence {ε k}, the original partial discharge signal is a function of time t, which represents the signal data collected from the GIS equipment without any processing; {ε k} is used to control the intensity of the white noise added to the original signal. Different decomposition processes can correspond to different noise amplitude coefficients.

[0038] S22: Decompose the first-order intrinsic mode function IMF1, that is, perform first-order mode decomposition. The specific implementation process is as follows:

[0039] S22.1, generate a noisy signal set as follows: (i) (t)=x(t)+ε1·ω (i) (t),i=1,2,…,N;

[0040] Where ε1 represents the noise amplitude coefficient corresponding to the first-order modal decomposition; ω (i) (t) is an independently generated white noise sequence; x (i) (t) represents the i-th element in the noisy signal set, that is, the signal after adding noise of a specific intensity.

[0041] Multiple noise-added versions of the original signal, each of which is the result of adding the original signal (x(t)) to an independently generated white noise sequence (n(t)), and each noise sequence is multiplied by a specific noise amplitude coefficient ε k The coefficient ε k The intensity of each noise sequence is controlled so that the noisy signal has a certain degree of randomness and diversity, which helps subsequent signal decomposition algorithms (such as CEEMDAN) to more accurately identify and separate the intrinsic mode functions (IMFs) of the signal.

[0042] S22.2, decompose all noisy signals: for each x (i) (t) Perform empirical mode decomposition (EMD) to obtain the set of first-order IMFs

[0043] S22.3, ensemble averaging, the average eigenmode function corresponding to the first-order modal decomposition is obtained as follows:

[0044]

[0045] S2 2.4, update the residual signal as follows: r1(t)=x(t)-IMF1(t).

[0046] Wherein, r1(t) represents the residual signal obtained after the first-order modal decomposition, and IMF1(t) is the aforementioned IMF1, which represents the average intrinsic mode function obtained after the first-order modal decomposition.

[0047] S23: Decomposing subsequent IMFs, k-th order IMFs (i.e., performing k-th order modal decomposition), specifically including:

[0048] S23.1, generates adaptive noise by the following method

[0049]

[0050] Among them, E k-1 Represents the residual after the first k-1 order EMD decomposition of white noise; ε k represents the noise amplitude coefficient corresponding to the k-th order modal decomposition, E k-1 (ω (i) (t)) represents the residual after the first k-1 order EMD decomposition of white noise.

[0051] S23.2, constructing the noisy residual signal as follows:

[0052]

[0053] Among them, r k-1 (t) represents the residual signal obtained by the k-1th order modal decomposition.

[0054] S23.3, decompose and extract IMF:

[0055] For each Execute EMD to obtain the candidate set of k-th order IMF

[0056] S23.4, ensemble averaging, the average eigenmode function corresponding to the k-th order modal decomposition is as follows:

[0057]

[0058] S23.5, update the remaining signals as follows:

[0059] r k (t) = r k-1 (t)-IMF k (t)

[0060] Among them, r k (t) represents the residual signal obtained after the k-th order modal decomposition, IMF k (t) The aforementioned IMF k, represents the average eigenmode function obtained after the k-th order modal decomposition.

[0061] S24: Termination condition: Repeat step S23 until r k (t) The remaining signal becomes a monotonic function or only extreme points remain, and further decomposition is impossible.

[0062] S25: Retain the IMFs with a correlation coefficient (i.e., correlation) ρ>0.3 with the original signal, give priority to IMFs with high kurtosis values ​​(discharge pulses have high peak characteristics), and give priority to IMF combinations with a cumulative energy share of more than 80%. Generally, IMF3 to IMF7 are selected to obtain multiple target intrinsic mode functions.

[0063] Step S106, obtaining a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal based on the multiple target intrinsic mode functions;

[0064] Optionally, during the modal decomposition process in the previous step, the original partial discharge signal is decomposed into a series of intrinsic mode functions (IMFs). These IMFs represent different intrinsic modes of the signal and play an important role in partial discharge diagnosis because each IMF may contain specific physical information, such as the frequency, amplitude, and shape of the discharge. A time-frequency spectrum is obtained by calculating the instantaneous frequency and amplitude of multiple target intrinsic mode functions. This time-frequency spectrum not only preserves the instantaneous characteristics of the signal but also visually displays the changes in the frequency and intensity of the discharge pulse over time, allowing for clearer identification of the characteristics of the discharge pulse.

[0065] In an optional embodiment, a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of an original partial discharge signal is obtained based on a plurality of target intrinsic mode functions, including: performing Hilbert transforms on the plurality of target intrinsic mode functions to obtain analytical signals corresponding to the plurality of target intrinsic mode functions; and obtaining a time-frequency spectrum based on the analytical signals corresponding to the plurality of target intrinsic mode functions.

[0066] Optionally, the Hilbert transform is a mathematical tool for signal analysis that can convert a real signal into its complex analytical form, i.e., an analytical signal. For each target IMF, a Hilbert transform is first performed on it to obtain the corresponding analytical signal. The analytical signal is a complex number consisting of the original signal and its Hilbert transform result, which contains the instantaneous phase and amplitude information of the original signal. Through further analysis of the analytical signals corresponding to multiple target intrinsic mode functions (analysis of instantaneous frequency, amplitude, etc.), the original partial discharge signal is converted into a feature representation in the time-frequency domain, i.e., a real-time spectrum. This feature representation not only contains the frequency and time information of the signal, but also clearly displays the energy distribution of the signal, which is helpful for the identification and diagnosis of partial discharge patterns.

[0067] In an optional embodiment, a time-frequency spectrum is obtained based on analytical signals corresponding to multiple target intrinsic mode functions, including: obtaining instantaneous frequencies and amplitudes corresponding to multiple target intrinsic mode functions based on analytical signals corresponding to multiple target intrinsic mode functions; constructing a time-frequency energy distribution matrix based on the instantaneous frequencies and amplitudes corresponding to multiple target intrinsic mode functions, wherein the time-frequency energy distribution matrix is ​​used to describe the distribution of the instantaneous frequency and amplitude of the signal in time and frequency; and performing integration operation based on the time-frequency energy distribution matrix to obtain the time-frequency spectrum.

[0068] Optionally, first, a Hilbert transform is performed on each target IMF to obtain the corresponding analytical signal. The analytical signal is a complex signal composed of the Hilbert transform of the original signal and the original signal itself. Based on the analytical signal, the instantaneous frequency and instantaneous amplitude can be calculated, where the instantaneous frequency is the derivative of the analytical signal's phase, and the instantaneous amplitude is the modulus of the analytical signal. Next, the instantaneous frequency and amplitude of each target IMF are used to construct a time-frequency energy distribution matrix (Hilbert spectrum). This matrix represents the energy distribution of the signal in the time and frequency domains. Each row represents the energy distribution at a time point, and each column represents the energy contribution of different frequency components at each time point. This method allows for intuitive visualization of the temporal evolution of different frequency components in the signal, which is particularly useful for analyzing nonlinear and nonstationary signals. The final step is to integrate the time-frequency energy distribution matrix to obtain the marginal spectrum, which is the distribution of signal energy over frequency. This constitutes part of the time-frequency spectrum. Constructing the time-frequency spectrum involves integrating the energy distribution of each frequency component to obtain the energy contribution of the entire signal at each frequency point.

[0069] The goal of the above method is to convert the complex raw partial discharge signal into a form that is easier to analyze and understand—a time-frequency spectrum. The Hilbert spectrum can provide a visual representation of the signal energy in the time-frequency domain, thereby clearly identifying different patterns in the signal, especially those specific frequency components associated with partial discharge. This is very important for subsequent spectrum construction and discharge type diagnosis, as it provides key features for distinguishing and identifying different partial discharge patterns, thereby improving the accuracy and speed of diagnosis. In summary, through Hilbert spectrum analysis, we can extract rich information about frequency, time distribution, and energy distribution from the raw signal, providing strong support for further signal analysis and fault diagnosis.

[0070] Optionally, the time-frequency spectrum can be obtained by, but is not limited to, the following methods:

[0071] S31, for each target eigenmode function Perform Hilbert transform to obtain the analytical signal:

[0072] z k (t) = IMF k (t)+j·H[IMF k (t)]

[0073] Among them, z k (t) represents the analytical signal, H[·] is the Hilbert transform operator; j represents the imaginary unit, which is used to expand the real signal to the complex plane to construct the analytical signal; by adding the imaginary part, the IMF k (t) The real signal is converted into a rotation vector on the complex plane.

[0074] S32, instantaneous frequency calculation:

[0075] The phase derivative of the analytical signal is the instantaneous frequency:

[0076] S33, time-frequency spectrum synthesis:

[0077] Combine the instantaneous frequencies and amplitudes of all IMFs to construct the time-frequency energy distribution matrix (Hilbert spectrum):

[0078]

[0079] Among them, A k (t) is IMF k , δ(·) is the Dirac function, and K represents the total number of multiple target intrinsic mode functions.

[0080] S34, Marginal Spectrum and Energy Analysis:

[0081] The marginal spectrum (the distribution of energy over frequency, the real-time spectrum) is obtained by integrating the Hilbert spectrum as follows:

[0082]

[0083] Step S108, constructing a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum based on the time-frequency spectrum, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude of the discharge pulse and the phase, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse;

[0084] Optionally, the Phase Resolved Partial Discharge (PRPD) spectrum is a spectrum that shows the relationship between the discharge pulse amplitude and the system phase. By distributing the discharge pulses according to the system voltage phase, it can reveal the correlation between the discharge and the voltage cycle, thereby providing important information about the discharge type. For example, certain discharge patterns may only appear in a specific phase of the voltage cycle, or their amplitude may change with the phase in a specific pattern. The Phase Resolved Pulse Sequence (PRPS) spectrum is a phase-resolved representation of the discharge pulse in a time series. It focuses on the sequence characteristics of the pulse and can provide deeper information in the frequency domain, such as the repeatability and frequency characteristics of the pulse.

[0085] Optionally, a phase-resolved partial discharge (PRPD) spectrum can be constructed using, but is not limited to, the following methods: First, discharge pulses are identified from the time-frequency spectrum. The presence of a pulse can be determined by setting a threshold, as pulses appear as significant energy peaks in the time-frequency spectrum. A phase value is assigned to each detected discharge pulse. This step utilizes a synchronized voltage signal to map the occurrence of each pulse to a phase position in the voltage signal. Based on the pulse's phase value, the amplitude distribution of the pulses within the phase range is calculated. This typically requires dividing the phase range into multiple intervals, such as dividing a complete AC voltage cycle (360°) into multiple phase segments, and calculating the amplitude and occurrence count of the pulses within each segment. Finally, the calculated pulse amplitude and occurrence count are plotted as a heat map or chromatogram to form a PRPD spectrum. In this spectrum, the horizontal axis represents the phase angle, the vertical axis represents the pulse amplitude, and the color or brightness of the spectrum indicates the frequency of pulses occurring under the same phase and amplitude conditions.

[0086] Optionally, a phase-resolved pulse sequence spectrum (PRPS) can be constructed in the following manner, but is not limited to: the discharge pulses are also extracted from the time-frequency spectrum, but the focus here is on the time series characteristics of the pulses, especially the time interval and phase difference between the pulses. The pulses are aligned according to the phase of the system voltage to ensure that each pulse is placed in the correct position corresponding to its phase. At each phase position, the repetitive pattern or sequence characteristics of the pulses are counted, which can be, but is not limited to, the frequency of occurrence of the eight-dimensional pulses, the morphology of the pulse sequence, etc. Based on the statistical results, a distribution spectrum of the pulse sequence in the phase space is drawn. In the PRPS spectrum, the horizontal axis represents the phase angle, the vertical axis represents the time position of the pulse sequence or the frequency characteristics of the pulse, and the color or brightness of the spectrum represents the intensity or number of pulses at that phase and position (or frequency).

[0087] It should be noted that the construction of the above two types of maps (PRPD and PRPS) is to better understand and characterize the characteristics of partial discharge, especially in the phase and time series dimensions. The PRPD map emphasizes the relationship between pulse amplitude and phase, which can help identify the type of discharge; while the PRPS map focuses on the sequence characteristics between pulses, which can analyze the pattern of partial discharge in more detail. By constructing PRPD and PRPS maps, complex time-frequency information can be converted into an intuitive map representation, which is convenient for analysis and diagnosis using visual intelligence algorithms. These maps can not only reveal the amplitude and phase relationship of the discharge pulse, but also provide the time series characteristics of the pulse, providing a solid foundation for accurately identifying the type of partial discharge.

[0088] Step S110 , based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum, a pre-trained partial discharge diagnosis model is used to obtain a partial discharge diagnosis result of the gas-insulated switchgear, wherein the partial discharge diagnosis result is at least used to indicate a partial discharge type of the gas-insulated switchgear.

[0089] It's no secret that PRPD and PRPS patterns carry rich information about the phase and time series of partial discharge pulses. The partial discharge diagnostic model effectively captures and learns the characteristic patterns of discharge types from these patterns, accurately distinguishing different types of partial discharge even in complex environments, thereby providing more accurate diagnostic results. Compared to traditional manual analysis methods, the intelligent diagnostic model rapidly processes input pattern data and instantly delivers diagnostic results, significantly improving the efficiency of GIS equipment diagnostics.

[0090] In an optional embodiment, when the partial discharge diagnostic model is trained based on an initially constructed sliding window attention transformation model, a pre-trained partial discharge diagnostic model is used based on a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum to obtain a partial discharge diagnostic result of a gas-insulated switchgear, including: converting the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum into images of predetermined sizes to obtain a phase-resolved partial discharge image and a phase-resolved pulse sequence image; using an image segmentation module in the partial discharge diagnostic model to perform image segmentation on the phase-resolved partial discharge image and the phase-resolved pulse sequence image to obtain an image segmentation result; using a linear encoding module in the partial discharge diagnostic model to convert the image segmentation result into a multidimensional vector through linear transformation; inputting the multidimensional vector into a sliding window attention transformation module in the partial discharge diagnostic model, performing multi-scale feature extraction and fusion processing through multiple stages to obtain a target feature map; and using an output module in the partial discharge diagnostic model to process the target feature map to obtain a partial discharge diagnostic result.

[0091] Optionally, the phase-resolved partial discharge map and the phase-resolved pulse sequence map are converted into an RGB image format of a predetermined size to obtain a phase-resolved partial discharge image (PRPD Image) and a phase-resolved pulse sequence image (PRPSImage). The image can be adjusted to the input size expected by Swin Transformer, such as 224x224 pixels or larger, to meet the input requirements of the model. The image segmentation module (i.e., PatchPartition module) in the partial discharge diagnosis model is used to segment the phase-resolved partial discharge image and the phase-resolved pulse sequence image to obtain image segmentation results. Each patch is regarded as an independent image region to prepare for subsequent feature extraction. The linear encoding module (i.e., linear encoding Linear Embedding layer) in the partial discharge diagnosis model is used to convert the image segmentation results into a multidimensional vector through linear transformation. This process converts the pixel value of each patch into a vector of fixed length, allowing the model to uniformly process image patches of different sizes. The multidimensional vector is input into the sliding window attention transformer module (i.e., Swin Transformer Block) in the partial discharge diagnosis model. Multi-scale feature extraction and fusion processing are performed in multiple stages to obtain a target feature map. At each stage, the model uses the window-based multi-head self-attention (W-MSA) module and the shifted window-based multi-head self-attention (SW-MSA) mechanism to capture local and cross-window features, completing the efficient extraction and processing of input image features. The target feature map is processed using the output module in the partial discharge diagnosis model to obtain the partial discharge diagnosis result. The output module can include, but is not limited to, one or more fully connected layers (Fully Connected Layer) and a classifier (Classifier), which is used to convert the features extracted from the image into a prediction result of the partial discharge type.

[0092] In the above examples, by converting PRPD / PRPS images into a standard image format and utilizing a pre-trained Swin Transformer-based partial discharge diagnosis model, image segmentation, linear encoding, and multi-scale feature extraction and fusion, efficient and accurate diagnosis of partial discharge types in gas-insulated switchgear (GIS) equipment is achieved. This approach cleverly combines the advantages of signal processing and deep learning, significantly improving the intelligence level and diagnostic accuracy of partial discharge detection.

[0093] In an optional embodiment, before obtaining the partial discharge diagnosis results of the gas-insulated switchgear using a pre-trained partial discharge diagnosis model based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum, the method further includes: obtaining multiple groups of images and the partial discharge types corresponding to the multiple groups of images, wherein each group of images includes a phase-resolved partial discharge image obtained by interpreting a historical partial discharge signal and a corresponding phase-resolved pulse sequence image; using the multiple groups of images as input and the partial discharge types corresponding to the multiple groups of images as sample labels, an initially constructed sliding window attention transformation model is trained to obtain the partial discharge diagnosis model.

[0094] Optionally, multiple sets of historical partial discharge signal data are acquired, each set consisting of PRPD and PRPS images analyzed from the signals, along with known partial discharge type labels corresponding to these images. This historical data, serving as the foundation of the training set, contains a rich set of partial discharge patterns and is a key source of model learning and generalization capabilities. The collected sets of PRPD and PRPS images are paired with corresponding partial discharge type labels to form multiple sets of training images and corresponding sample labels. These images and labels are used to train an initial Swin Transformer model. This model, based on the Transformer architecture, utilizes windowed self-attention (W-MSA) and shifted windowed self-attention (SW-MSA) mechanisms, along with a hierarchical design, to simultaneously focus on local and global information when processing partial discharge images, effectively balancing computational efficiency and model performance. The training process involves adjusting model parameters and minimizing the model's loss function when predicting partial discharge types through optimization methods such as backpropagation and gradient descent. After training is complete, the resulting partial discharge diagnosis model is saved. This model can be deployed on the cloud or terminal devices to process new PRPD and PRPS images in real time or batch, thereby quickly and accurately diagnosing the type of partial discharge in gas-insulated switchgear.

[0095] This example aims to leverage historical partial discharge data and the advanced features of the Swin Transformer to construct a high-precision partial discharge diagnosis model. This model can not only process complex PRPD / PRPS images and identify subtle partial discharge characteristics, but also adapt to different partial discharge patterns, improving the automation and accuracy of diagnosis. This is of great significance for GIS equipment status monitoring and fault warning in power systems, helping operators to promptly detect equipment defects, reduce the occurrence of power accidents, and ensure the safe and stable operation of the power grid.

[0096] Optional, Figure 3This is a schematic diagram of an optional partial discharge noise reduction diagnosis process based on a sliding window attention transformation model according to an embodiment of the present invention, specifically including:

[0097] S61: First, the PRPS / PRPD atlas image is converted into an [H, W, C] image by adjusting, padding or cropping, where H and W pixels are equal, which are 224, 256 or 384, and the RGB channel is 3. This is consistent with the image size of the training set image input when training the model, and then input into the image block partitioning module for block partitioning;

[0098] For S62, take the input image size of [224, 224, 3] as an example. Set the convolution kernel size to [4, 4]. After the PatchPartition module, the image shape changes from [H, W, 3] to [H / 4, W / 4, 48]. The LinearEmbeding layer linearly transforms the channel data of each pixel from 48 to C, that is, the image shape changes from [H / 4, W / 4, 48] to [H / 4, W / 4, C]. When C is 96, the input is 56×56×96, and the output is 28×28×192.

[0099] S63 divides the input feature map into multiple small feature blocks, and uses 4 stages (i.e., Stage 1 to Stage 4) to construct feature maps of different sizes. That is, the feature map is downsampled through different stages. Every time the sample is downsampled by 2 times, the channel needs to be doubled. After splicing, each size is 28×28, the channel becomes 4C, and then the 4C is converted to 2C through linear transformation.

[0100] S64, when training classification tasks, the input image needs to have a category label. By using the sliding window attention transformation module Swin Transformer Block structure, a multi-level feature map is generated to classify the image vector.

[0101] Among them, the specific implementation of the Swin Transformer Block model is achieved by means of two Swin Transformer methods, and this method appears in groups and is used in pairs, namely Two Swin Transformer Blocks, first using a W-MSA (window self-attention mechanism) structure and then using a SW-MSA (sliding window self-attention mechanism) structure. Therefore, the number of times the Swin Transformer Block is stacked is an even number (because it is used in pairs). This embodiment uses the Swin Transformer algorithm to implement the specific logic. Through the hierarchical structure of multiple Stages, higher-level features are gradually extracted. Each Stage will halve the size of the feature map and double the number of channels. Figure 3 As shown in the figure on the right, a multi-layer perceptron (MLP) and layer normalization (LayerNorm) are set in the Two SwinTransformer Blocks. Among them, the MLP is composed of a stack of fully connected layers of multiple hidden layers. In the Swin Transformer, MLP is mainly used to: map the input multi-dimensional vector to another multi-dimensional space. In this process, the dimension of the vector may change, such as first increasing the dimension through a linear layer, and then reducing it back to its original size through another linear layer. This feature mapping helps the model learn more complex data representations. Between each hidden layer of the MLP, a nonlinear activation function (such as ReLU, GELU, etc.) can also be inserted to break the linear relationship, enhance the learning ability of the network, and enable it to fit nonlinear complex functions.

[0102] LayerNorm is a data normalization technique used during neural network training to accelerate convergence and improve stability. LayerNorm normalizes the output features of each layer by calculating the mean and variance of each dimension of the feature and then normalizing it to stabilize the feature distribution around zero mean and unit variance. Through normalization, LayerNorm helps alleviate the vanishing or exploding gradient problem in deep neural networks, making the model more stable when training large networks. After normalization, the model can learn faster because the inputs of each layer are more numerically consistent, reducing instability during training.

[0103] In the Swin Transformer, MLP and LayerNorm are primarily used for the output processing of each Transformer Block. Specifically, the output of each Block is first normalized using LayerNorm, then enters the MLP for feature mapping and conversion, and finally undergoes LayerNorm processing again to ensure the stability of the next layer's input. This design not only enhances the model's ability to learn the features of partial discharge images, but also ensures the model's stability and efficiency throughout the training process. By introducing LayerNorm and MLP, the Swin Transformer can better process the multi-scale features in partial discharge atlas images, effectively combining local and global information, and thus demonstrating higher accuracy and robustness in partial discharge diagnosis.

[0104] Through the above steps S102 to S110, it is possible to acquire the original partial discharge signal, apply intrinsic mode decomposition to obtain multiple modal functions, and then construct a time-frequency spectrum and a phase-resolved partial discharge spectrum (PRPD) and a phase-resolved pulse sequence spectrum (PRPS). Finally, a pre-trained partial discharge diagnostic model is used to analyze these spectra to accurately determine the partial discharge diagnosis results, thereby achieving the technical effect of improving the accuracy of partial discharge detection and diagnosis of gas insulated switchgear (GIS), and thus solving the technical problem of low accuracy of partial discharge detection and diagnosis of gas insulated switchgear (GIS) in related technologies.

[0105] Based on the above embodiments and optional embodiments, the present invention proposes an optional implementation mode: Figure 4 FIG. 1 is a flow chart of an optional method for diagnosing partial discharge of a gas-insulated switchgear according to an embodiment of the present invention. Figure 4 As shown, the method includes:

[0106] S1, obtain the online original partial discharge signal x(t) of the GIS equipment and set the sampling frequency and sampling step size;

[0107] S2, performing modal decomposition on the original partial discharge signal using the CEEMDAN algorithm to obtain a number of initial intrinsic mode functions (IMFs), screening out valid IMFs (i.e., target IMFs) and retaining them. The specific implementation process is the same as that of the previous embodiment and will not be repeated here;

[0108] S3, performing Hilbert-Huang Transform (HHT transform) on each selected target IMF to calculate its instantaneous frequency and amplitude to obtain a time-frequency spectrum. The specific implementation is the same as that in the previous embodiment and will not be repeated here;

[0109] S4, performing phase synchronization, pulse detection and parameter extraction on the time-frequency spectrum to obtain the extraction results;

[0110] S5, generating a PRPD / PRPS map based on the extraction results and converting it into a PRPD / PRPS image;

[0111] S6: Input the PRPS / PRPD atlas image after the noise reduction process in steps S2 to S5 into the built sliding window attention transformation model Swin Transformer model. The specific implementation method is the same as that in the previous embodiment and will not be repeated here;

[0112] S7, output the discharge type prediction result.

[0113] It should be noted that the CEEMDAN-HHT and Swin Transformer-based PD noise reduction diagnostic method provided in this embodiment, compared to traditional online diagnostic methods, avoids the shortcomings of traditional signal analysis methods (such as Fourier transform and wavelet transform) that rely on window selection and preset basis functions, making them difficult to handle non-stationary signals. It significantly reduces the computational complexity of PD signal decomposition, is less dependent on noise, and is more adaptable to local signal characteristics. The constructed intelligent algorithm model can process super-resolution images, save computational resources, and can focus on both global and local information, effectively balancing computational efficiency and model performance.

[0114] This embodiment also provides a partial discharge diagnostic device for a gas-insulated switchgear. This device is used to implement the aforementioned embodiments and preferred implementations, and details already described will not be repeated. As used below, the terms "module" and "device" may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented using software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0115] According to an embodiment of the present invention, there is also provided an embodiment of a device for implementing the above-mentioned partial discharge diagnosis method for gas-insulated switchgear. Figure 5 FIG. 1 is a schematic structural diagram of a partial discharge diagnostic device for a gas insulated switchgear according to an embodiment of the present invention. Figure 5 As shown, the partial discharge diagnostic device for gas-insulated switchgear includes: a signal acquisition module 500, a decomposition module 502, a time-frequency spectrum acquisition module 504, a spectrum construction module 506, and a diagnosis result determination module 508, wherein:

[0116] The signal acquisition module 500 is used to acquire the original partial discharge signal collected from the gas-insulated switchgear;

[0117] The decomposition module 502 is connected to the signal acquisition module 500 and is used to perform modal decomposition on the original partial discharge signal to obtain multiple target intrinsic mode functions;

[0118] The time-frequency spectrum acquisition module 504 is connected to the decomposition module 502 and is used to obtain a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal based on multiple target intrinsic mode functions;

[0119] A spectrum construction module 506 is connected to the time-frequency spectrum acquisition module 504 and is used to construct a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum based on the time-frequency spectrum, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude and phase of the discharge pulse, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse;

[0120] The diagnostic result determination module 508 is connected to the spectrum construction module 506 and is used to obtain a partial discharge diagnostic result of the gas-insulated switchgear based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum using a pre-trained partial discharge diagnostic model, wherein the partial discharge diagnostic result is used to at least indicate the type of partial discharge of the gas-insulated switchgear.

[0121] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0122] It should be noted that the signal acquisition module 500, decomposition module 502, time-frequency spectrum acquisition module 504, spectrum construction module 506, and diagnosis result determination module 508 correspond to steps S102 to S110 in the embodiment. The examples and application scenarios implemented by these modules and corresponding steps are the same, but are not limited to the contents disclosed in the above embodiment. It should be noted that the above modules, as part of the device, can be run on a computer terminal.

[0123] It should be noted that the optional or preferred implementation of this embodiment can be found in the relevant description in the embodiment, which will not be repeated here.

[0124] The above-mentioned partial discharge diagnostic device for gas-insulated switchgear may further include a processor and a memory. The above-mentioned signal acquisition module 500, decomposition module 502, time-frequency spectrum acquisition module 504, graph construction module 506, diagnosis result determination module 508, etc. are all stored in the memory as program modules, and the processor executes the above-mentioned program modules stored in the memory to implement corresponding functions.

[0125] The processor includes a core, which retrieves corresponding program modules from memory. There can be one or more cores. Memory may include non-permanent memory in a computer-readable medium, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0126] According to an embodiment of the present application, an embodiment of a non-volatile storage medium is also provided. Optionally, in this embodiment, the non-volatile storage medium includes a stored program, wherein, when the program is executed, the device containing the non-volatile storage medium is controlled to execute any of the aforementioned methods for diagnosing partial discharge in gas-insulated switchgear.

[0127] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group, and the non-volatile storage medium includes a stored program.

[0128] Optionally, when the program is running, the device where the non-volatile storage medium is located is controlled to execute any one of the above-mentioned partial discharge diagnosis methods for gas-insulated switchgear.

[0129] According to an embodiment of the present application, a processor embodiment is further provided. Optionally, in this embodiment, the processor is configured to run a program, wherein the program, when running, executes any one of the aforementioned methods for diagnosing partial discharge of a gas-insulated switchgear.

[0130] According to an embodiment of the present application, an embodiment of a computer program product is also provided. When executed on a data processing device, the computer program product is suitable for executing a program that initializes any one of the steps of the partial discharge diagnosis method for gas-insulated switchgear described above.

[0131] An embodiment of the present invention provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: obtaining an original partial discharge signal collected from a gas-insulated switchgear; performing modal decomposition on the original partial discharge signal to obtain multiple target intrinsic mode functions; obtaining a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal based on the multiple target intrinsic mode functions; constructing a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum based on the time-frequency spectrum, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude of the discharge pulse and the phase, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse; and obtaining a partial discharge diagnosis result of the gas-insulated switchgear based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum using a pre-trained partial discharge diagnosis model, wherein the partial discharge diagnosis result is used to at least indicate the type of partial discharge of the gas-insulated switchgear.

[0132] The above sequence of the embodiments of the present invention is for description only and does not represent the superiority or inferiority of the embodiments.

[0133] In the above embodiments of the present invention, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0134] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the above modules can be a logical function division. In actual implementation, there may be other division methods, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, modules or indirect coupling or communication connection of modules, which can be electrical or other forms.

[0135] The modules described above as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed across multiple modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment.

[0136] In addition, the functional modules in various embodiments of the present invention may be integrated into a single processing module, or each module may exist physically separately, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0137] If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable non-volatile storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a non-volatile storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned non-volatile storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, and other media that can store program codes.

[0138] The above are only preferred embodiments of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for diagnosing partial discharge of a gas-insulated switchgear, characterized in that: include: Obtain the original partial discharge signal collected for gas-insulated switchgear; Performing modal decomposition on the original partial discharge signal to obtain a plurality of target intrinsic mode functions; Based on the multiple target eigenmode functions, a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal is obtained; Based on the time-frequency spectrum, a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum are constructed, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude of the discharge pulse and the phase, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse; Based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum, a pre-trained partial discharge diagnostic model is used to obtain a partial discharge diagnostic result of the gas-insulated switchgear, wherein the partial discharge diagnostic result is at least used to indicate a partial discharge type of the gas-insulated switchgear.

2. The method according to claim 1, characterized in that The performing modal decomposition on the original partial discharge signal to obtain a plurality of target intrinsic mode functions includes: adding a plurality of different noise signals to the original partial discharge signal to obtain a plurality of noisy signals, wherein the plurality of noisy signals correspond one-to-one to the plurality of different noise signals; Performing multi-order modal decomposition based on the multiple noisy signals to obtain a plurality of initial intrinsic mode functions, wherein the number of the multiple initial intrinsic mode functions is greater than the number of the multiple target intrinsic mode functions; Determining correlations between the multiple initial intrinsic mode functions and the original partial discharge signal; Among the multiple initial eigenmode functions, eigenmode functions having a correlation greater than a preset correlation threshold are determined as the multiple target eigenmode functions.

3. The method according to claim 2, characterized in that The performing multi-order modal decomposition based on the multiple noisy signals to obtain multiple initial intrinsic mode functions includes: The first-order modal decomposition in the multi-order modal decomposition is used as the current-order modal decomposition, and the current-order modal decomposition is performed in the following manner: Performing empirical mode decomposition on the multiple noisy signals to obtain an intrinsic mode function set corresponding to each of the multiple noisy signals, wherein the intrinsic mode function set includes multiple intrinsic mode functions obtained by decomposing the corresponding noisy signals; Obtaining, based on the eigenmode function sets corresponding to the multiple noisy signals, an average eigenmode function corresponding to each of the multiple noisy signals, wherein the average eigenmode function is obtained by averaging multiple eigenmode functions included in the eigenmode function set of the noisy signal; Based on the average intrinsic mode functions corresponding to the multiple noisy signals and the original partial discharge signal, a plurality of residual signals are obtained, wherein the plurality of residual signals correspond to the multiple noisy signals one-to-one; adding noise to the plurality of residual signals as a plurality of new noisy signals, and taking the next-order modal decomposition of the current-order modal decomposition as the new current-order modal decomposition, and repeatedly performing the above operation until the last-order modal decomposition in the multi-order modal decomposition is reached; The average intrinsic mode functions corresponding to the multiple-order modal decompositions obtained after performing the last-order modal decomposition are used as the multiple initial intrinsic mode functions.

4. The method according to claim 1, wherein The method is based on the plurality of target eigenmode functions, Obtaining a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal, including: Performing Hilbert transform on the multiple target intrinsic mode functions to obtain analytical signals corresponding to the multiple target intrinsic mode functions; The time-frequency spectrum is obtained based on analytical signals corresponding to the multiple target eigenmode functions.

5. The method according to claim 4, characterized in that The obtaining the time-frequency spectrum based on analytical signals respectively corresponding to the plurality of target intrinsic mode functions includes: Based on analytical signals corresponding to the plurality of target intrinsic mode functions, respectively, obtaining instantaneous frequencies and amplitudes corresponding to the plurality of target intrinsic mode functions; Based on the instantaneous frequencies and amplitudes corresponding to the multiple target intrinsic mode functions, respectively, a time-frequency energy distribution matrix is ​​constructed, wherein the time-frequency energy distribution matrix is ​​used to describe the distribution of the instantaneous frequency and amplitude of the signal in time and frequency; An integration operation is performed based on the time-frequency energy distribution matrix to obtain the time-frequency spectrum.

6. The method according to claim 1, characterized in that In a case where the partial discharge diagnosis model is trained based on an initially constructed sliding window attention transformation model, obtaining a partial discharge diagnosis result of the gas-insulated switchgear based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum using the pre-trained partial discharge diagnosis model includes: Converting the phase-resolved partial discharge map and the phase-resolved pulse sequence map into images of predetermined sizes to obtain a phase-resolved partial discharge image and a phase-resolved pulse sequence image; Using the image segmentation module in the partial discharge diagnostic model to segment the phase-resolved partial discharge image and the phase-resolved pulse sequence image to obtain an image segmentation result; Using the linear encoding module in the partial discharge diagnostic model, the image segmentation result is converted into a multi-dimensional vector through linear transformation; Inputting the multidimensional vector into a sliding window attention transformation module in the partial discharge diagnosis model, performing multi-scale feature extraction and fusion processing in multiple stages to obtain a target feature map; The target characteristic graph is processed using an output module in the partial discharge diagnosis model to obtain the partial discharge diagnosis result.

7. The method according to claim 6, characterized in that Before obtaining a partial discharge diagnosis result of the gas-insulated switchgear based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum using a pre-trained partial discharge diagnosis model, the method further includes: Acquiring multiple sets of images and the partial discharge types corresponding to the multiple sets of images, wherein each set of images includes a phase-resolved partial discharge image obtained by interpreting a historical partial discharge signal and a corresponding phase-resolved pulse sequence image; The multiple groups of images are used as input, and the partial discharge types corresponding to the multiple groups of images are used as sample labels. The initially constructed sliding window attention transformation model is trained to obtain the partial discharge diagnosis model.

8. A partial discharge diagnostic device for a gas-insulated switchgear, characterized in that: include: A signal acquisition module, used to acquire the original partial discharge signal collected from the gas-insulated switchgear; a decomposition module, configured to perform modal decomposition on the original partial discharge signal to obtain a plurality of target intrinsic mode functions; A time-frequency spectrum acquisition module, configured to obtain a time-frequency spectrum for characterizing the instantaneous frequency and amplitude of the original partial discharge signal based on the multiple target intrinsic mode functions; a spectrum construction module, configured to construct a phase-resolved partial discharge spectrum and a phase-resolved pulse sequence spectrum based on the time-frequency spectrum, wherein the phase-resolved partial discharge spectrum is used to characterize the relationship between the amplitude of the discharge pulse and the phase, and the phase-resolved pulse sequence spectrum is used to characterize the time series characteristics of the partial discharge pulse; a diagnostic result determination module, configured to obtain a partial discharge diagnostic result of the gas-insulated switchgear based on the phase-resolved partial discharge spectrum and the phase-resolved pulse sequence spectrum using a pre-trained partial discharge diagnostic model, wherein the partial discharge diagnostic result is used to at least indicate a partial discharge type of the gas-insulated switchgear.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executed by the partial discharge diagnosis method for gas-insulated switchgear according to any one of claims 1 to 7.

10. An electronic device, characterized in that: The method comprises one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the partial discharge diagnosis method for gas-insulated switchgear according to any one of claims 1 to 7.

Images