Method, device, equipment and storage medium for detecting insulation status of power equipment

Abstract

The present application discloses a method, apparatus, device, and storage medium for detecting the insulation status of power equipment, relating to the technical field of power equipment, and is intended to address the low accuracy of existing detection methods for detecting whether power equipment is insulated. The method comprises: obtaining a partial discharge signal from the power equipment, dividing the partial discharge signal into multiple continuous discharge cycle segments based on the power supply frequency, dynamically identifying and amplitude limiting the partial discharge signal within each discharge cycle segment to obtain a valid pulse signal that retains pulse phase information, performing bottom noise adaptive elimination processing on the valid pulse signal based on a sliding window model to generate a de-noised signal, constructing a phase-resolved partial discharge (PRPD) matrix based on the de-noised signal, and outputting an insulation status diagnosis result based on the PRPD matrix, wherein the PRPD matrix is ​​a two-dimensional matrix in a phase-amplitude space.

Description

Technical Field

[0001] The present application relates to the technical field of power equipment, and in particular to a method, apparatus, device and storage medium for detecting the insulation status of power equipment. Background Art

[0002] Partial discharge (PD) detection is a key technology for assessing the insulation condition of power equipment. By monitoring weak discharge phenomena caused by insulation defects within the equipment, it can effectively warn of insulation degradation risks.

[0003] Currently, insulation testing typically uses phase-resolved partial discharge (PRPD) analysis. By analyzing the distribution characteristics of global discharge pulses, this method can distinguish discharge types and quantify the extent of insulation defects. However, in real-world environments, power equipment generates periodic noise, which can be low-amplitude and mask actual partial discharge signals. This can lead to inconsistent test results and reduce the accuracy of insulation testing. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, equipment and storage medium for detecting the insulation status of power equipment, aiming to solve the problem of low accuracy of existing detection methods in detecting whether power equipment is insulated.

[0005] To achieve the above objectives, this application adopts the following technical solutions:

[0006] The present application provides a method for detecting the insulation status of electric power equipment, the method comprising: obtaining a partial discharge signal of the electric power equipment, and dividing the partial discharge signal into a plurality of continuous discharge cycle segments based on the power supply frequency; dynamically identifying and amplitude limiting the partial discharge signal within each discharge cycle segment to obtain an effective pulse signal, wherein the effective pulse signal is a signal that retains pulse phase information; performing bottom noise adaptive elimination processing on the effective pulse signal based on a sliding window model to generate a noise reduction signal; constructing a PRPD matrix based on the noise reduction signal, and outputting an insulation status diagnosis result based on the PRPD matrix, wherein the PRPD matrix is ​​a matrix in a two-dimensional space of phase and amplitude.

[0007] The insulation status detection method of power equipment provided in the embodiment of the present application first dynamically divides the partial discharge signal based on the power supply frequency, accurately isolates the continuous signal into periodic segments, and effectively avoids periodic noise aliasing; on this basis, through dynamic identification and amplitude limitation processing, the interference of high-amplitude noise pulses on the effective discharge signal is targetedly suppressed, while the pulse phase information is fully retained to ensure the authenticity of the discharge time and space distribution characteristics; then, a sliding window model is used to adaptively eliminate the bottom noise of the processed signal, and through real-time noise modeling and dynamic threshold adjustment, the masking effect of low-amplitude base noise on weak discharge signals is filtered out; finally, a standardized PRPD matrix is ​​constructed based on the phase-amplitude two-dimensional space to eliminate the interference of phase offset on the diagnosis result, thereby improving the detection accuracy of whether the power equipment is insulated.

[0008] In some embodiments, the above-mentioned dynamic identification and amplitude limitation of the local discharge signal within each discharge cycle segment to obtain a valid pulse signal includes: performing Gaussian mixture model identification on the pulse amplitude of the local discharge signal to obtain an abnormal amplitude pulse, where the amplitude of the abnormal amplitude pulse is greater than a preset amplitude; and performing limiting processing and nonlinear filtering processing on the abnormal amplitude pulse to obtain a valid pulse signal.

[0009] Based on this, this application uses a Gaussian mixture model to model the probability distribution of pulse amplitudes, accurately distinguish between noise and effective discharge pulses, combine limiting processing to suppress abnormal amplitude interference, and use nonlinear filtering to eliminate signal distortion after limiting, retain the integrity of phase information, and ensure the authenticity of the spatiotemporal characteristics of subsequent PRPD analysis.

[0010] In some embodiments, the above-mentioned limiting processing and nonlinear filtering processing of the abnormal amplitude pulse to obtain a valid pulse signal include: retaining the phase information of the abnormal amplitude pulse to obtain an initial amplitude pulse; limiting the initial amplitude pulse, and using a median filtering algorithm to smooth the initial amplitude pulse after limiting processing to obtain a valid pulse signal, wherein the window size of the median filtering algorithm is dynamically adjusted according to the signal sampling rate.

[0011] Based on this, the present application dynamically adjusts the median filter window size according to the signal sampling rate to avoid signal detail loss or noise residual problems caused by a fixed window, retains the phase characteristics of the discharge pulse while suppressing high-frequency noise, and improves the effectiveness of subsequent noise reduction signals.

[0012] In some embodiments, the above-mentioned bottom noise adaptive elimination processing of the effective pulse signal based on the sliding window model to generate a noise reduction signal includes: extracting the lowest amplitude area of ​​the effective pulse signal as a noise reference; establishing a noise statistical model within the sliding window based on the noise reference; dynamically calculating the adaptive threshold according to the noise statistical model, and returning the effective pulse signal below the adaptive threshold to zero to obtain a noise reduction signal.

[0013] Based on this, this application constructs a noise reference by extracting the lowest amplitude area of ​​the signal, and combines the sliding window to perform real-time statistical noise mean and variance, dynamically reflecting the time-varying characteristics of the noise base, making the threshold calculation more in line with actual working conditions and enhancing adaptability to complex industrial environments.

[0014] In some embodiments, the above-mentioned dynamic calculation of the adaptive threshold based on the noise statistical model includes: calculating the mean and standard deviation of the noise statistical model; determining the adaptive threshold based on the mean and standard deviation in combination with an adaptive coefficient, wherein the adaptive coefficient is automatically adjusted as the signal-to-noise ratio changes.

[0015] Based on this, this application dynamically optimizes the adaptive coefficient based on the mean and standard deviation of the noise model, combined with the signal-to-noise ratio, to achieve automatic adjustment of the threshold with signal quality, avoid manual preset deviations, and improve the ability to recognize weak signals in low signal-to-noise ratio scenarios.

[0016] In some embodiments, the above-mentioned construction of a PRPD matrix based on the noise reduction signal includes: performing amplitude normalization and phase alignment processing on the noise reduction signal, and determining the distribution density of the processed noise reduction signal in the phase-amplitude two-dimensional space; based on the distribution density and kernel density estimation method, generating a standardized PRPD grayscale image as the PRPD matrix.

[0017] Based on this, this application eliminates sensor sensitivity differences through segmented amplitude normalization, combines phase alignment to compensate for offset, uses kernel density estimation to smooth the phase-amplitude distribution, and generates a standardized PRPD grayscale image to ensure the consistency of diagnostic results of different devices.

[0018] In some embodiments, the above-mentioned output of the insulation status diagnosis result based on the PRPD matrix includes: extracting the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics of the PRPD matrix; inputting the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics into a multimodal fusion model, and outputting the insulation status diagnosis result; wherein the multimodal fusion model is constructed using a deep convolutional neural network.

[0019] Based on this, this application extracts statistical, texture and time-series multidimensional features from the PRPD matrix, fuses image and numerical features through a deep convolutional network, captures the implicit correlation of discharge patterns, and significantly improves the classification accuracy of insulation defects such as internal discharge and surface discharge.

[0020] The present application provides an insulation status detection device for electric power equipment, which includes: an acquisition unit, used to acquire a partial discharge signal of the electric power equipment, and divide the partial discharge signal into multiple continuous discharge cycle segments based on the power supply frequency; a processing unit, used to dynamically identify and amplitude limit the partial discharge signal in each discharge cycle segment to obtain an effective pulse signal, which is a signal that retains pulse phase information; a generation unit, used to perform bottom noise adaptive elimination processing on the effective pulse signal based on a sliding window model to generate a noise reduction signal; a construction unit, used to construct a PRPD matrix based on the noise reduction signal, and output an insulation status diagnosis result based on the PRPD matrix, which is a matrix in a two-dimensional space of phase and amplitude.

[0021] In some embodiments, the above-mentioned processing unit is specifically used to: perform Gaussian mixture model identification on the pulse amplitude of the partial discharge signal to obtain an abnormal amplitude pulse, where the amplitude of the abnormal amplitude pulse is greater than a preset amplitude; perform limiting processing and nonlinear filtering processing on the abnormal amplitude pulse to obtain a valid pulse signal.

[0022] In some embodiments, the above-mentioned processing unit is specifically used to: retain the phase information of the abnormal amplitude pulse to obtain the initial amplitude pulse; limit the initial amplitude pulse, and use the median filtering algorithm to smooth the initial amplitude pulse after the limiting processing to obtain a valid pulse signal, wherein the window size of the median filtering algorithm is dynamically adjusted according to the signal sampling rate.

[0023] In some embodiments, the above-mentioned generation unit is specifically used to: extract the lowest amplitude area of ​​the effective pulse signal as a noise reference; establish a noise statistical model within the sliding window based on the noise reference; dynamically calculate the adaptive threshold according to the noise statistical model, and zero the effective pulse signal below the adaptive threshold to obtain a noise reduction signal.

[0024] In some embodiments, the above-mentioned generation unit is specifically used to: calculate the mean and standard deviation of the noise statistical model; determine the adaptive threshold based on the mean and standard deviation in combination with the adaptive coefficient, wherein the adaptive coefficient is automatically adjusted as the signal-to-noise ratio changes.

[0025] In some embodiments, the above-mentioned construction unit is specifically used to: perform amplitude normalization and phase alignment processing on the noise reduction signal, and determine the distribution density of the processed noise reduction signal in the phase-amplitude two-dimensional space; based on the distribution density and kernel density estimation method, generate a standardized PRPD grayscale image as a PRPD matrix.

[0026] In some embodiments, the above-mentioned construction unit is specifically used to: extract the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics of the PRPD matrix; input the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics into the multimodal fusion model, and output the insulation status diagnosis results; wherein the multimodal fusion model is constructed using a deep convolutional neural network.

[0027] The present application provides an electronic device, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the above-described method for detecting the insulation status of an electric power device.

[0028] The present application provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal, the terminal executes the above-described method for detecting the insulation status of electric power equipment.

[0029] The present application provides a computer program product comprising instructions, which, when executed by a computer, enables the computer to execute the above-described method for detecting the insulation status of electric power equipment.

[0030] The present application provides a chip, which includes a processor and a communication interface. The communication interface and the processor are coupled, and the processor is used to run computer programs or instructions to implement the above-described method for detecting the insulation status of power equipment.

[0031] Specifically, the chip provided in the embodiment of the present application also includes a memory for storing computer programs or instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0033] Figure 1 A framework diagram of a power equipment insulation status detection system provided in an embodiment of the present application;

[0034] Figure 2 A flow chart of a method for detecting insulation status of electric equipment provided in an embodiment of the present application;

[0035] Figure 3 A structural diagram of an insulation status detection device for electric equipment provided in an embodiment of the present application;

[0036] Figure 4 A structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0038] In the description of this application, it should be understood that the terms "upper," "lower," "left," "right," "front," "back," "inner," "outer," and the like, indicating directions or positional relationships, are based on the directions or relative positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of this application and simplify the description. They do not indicate or imply that the devices or components referred to must have a specific direction, be constructed, or operate in a specific direction. Therefore, they should not be construed as limitations on this application. Unless otherwise specified, the above-mentioned directionality descriptions may be flexibly set in actual application, provided that the relative positional relationships shown in the accompanying drawings are met.

[0039] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. Throughout this application, unless otherwise specified, "plurality" means two or more.

[0040] In the description of this application, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connected," and "connected" should be understood broadly. For example, they may refer to fixed connections, detachable connections, or integral connections. They may be directly connected, indirectly connected through an intermediary, or internally connected between two components. Those skilled in the art will understand the specific meanings of these terms in this application based on the specific circumstances.

[0041] In some embodiments, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, article, or apparatus that includes the element.

[0042] In some embodiments, words such as "exemplary" or "for example" are used to indicate an example, illustration, or description. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0043] In the description of this specification, specific features, structures, materials or characteristics may be combined in an appropriate manner in any one or more embodiments or examples.

[0044] Currently, PRPD analysis is a commonly used analysis method that can distinguish discharge types and quantify the degree of insulation defects by statistically analyzing the phase-amplitude distribution characteristics of discharge pulses. However, existing PRPD analysis technology faces the following four problems in industrial environments:

[0045] (1) Severe periodic noise interference: Power electronic equipment in industrial environments (such as frequency converters and inverters) generate fixed-frequency periodic noise, whose frequency band overlaps with the partial discharge signal. Traditional PRPD analysis uses global threshold filtering to dynamically suppress such interference, resulting in the real discharge signal being submerged by the noise.

[0046] (2) Bottom noise affects weak signal recognition: The low-amplitude noise floor formed by sensor thermal noise, environmental electromagnetic interference, etc., will mask the phase distribution characteristics of weak partial discharge signals. Especially under low signal-to-noise ratio (SNR) (such as SNR < 5 dB) conditions, manually set static thresholds are prone to missed detection or misjudgment.

[0047] (3) High sensitivity to amplitude and phase: Existing methods rely on the absolute strength of the pulse amplitude and the accuracy of phase alignment. However, in actual measurements, the amplitude is easily affected by factors such as sensor sensitivity and transmission loss, while the phase is offset due to differences in equipment installation, resulting in large fluctuations and poor stability in cross-scenario diagnostic results.

[0048] (4) Artificial thresholds lack adaptability: Noise characteristics change dynamically under complex working conditions. Artificially preset fixed thresholds cannot be adjusted in real time and are difficult to adapt to the impact of different environments (such as high humidity and temperature fluctuations) on the signal baseline.

[0049] In summary, the above problems may lead to insufficient stability and accuracy in the extraction of discharge features by PRPD analysis, which in turn leads to reduced accuracy in detecting the insulation status of power equipment.

[0050] Against this backdrop, and to address the low accuracy of insulation detection in power equipment in related technologies, this application provides a method, apparatus, device, and storage medium for detecting the insulation status of power equipment. By combining periodic noise pulse extraction and amplitude limiting with noise cancellation technology, and using intelligent PRPD for feature extraction, the accuracy of insulation detection in power equipment is improved.

[0051] Figure 1 This is a framework diagram of a power equipment insulation status detection system provided in an embodiment of the present application. The power equipment insulation status detection system 100 includes a signal acquisition module 110, a period segmentation module 120, a noise pulse extraction and limitation module 130, a bottom noise adaptive elimination module 140, a PRPD matrix construction module 150, a feature extraction module 160, a diagnostic model 170, and a result output module 180.

[0052] In some embodiments, the signal acquisition module 110 may acquire partial discharge signals of the power equipment by ultrasonic detection method, ultra-high frequency detection method, pulse current method, etc.

[0053] In some embodiments, the cycle segmentation module 120 can segment the continuous signal into individual cycle segments based on the power frequency (eg, 50 / 60 Hz), and use an improved zero-crossing detection algorithm to ensure accurate cycle boundaries.

[0054] In some embodiments, the noise pulse extraction and limitation module 130 may perform noise pulse identification and pulse limitation processing.

[0055] For example, noise pulse recognition can calculate the statistical distribution of pulse amplitudes in each cycle, establish a pulse classifier based on a Gaussian mixture model (GMM), and identify and mark abnormal amplitude pulses (noise pulses).

[0056] Furthermore, the pulse limiting process can limit the amplitude of the identified noise pulse and retain the pulse phase information. Finally, the signal after smoothing is smoothed using a nonlinear filtering algorithm to obtain a noise-reduced signal.

[0057] In some embodiments, the floor noise adaptive cancellation module 140 may perform noise floor modeling, adaptive threshold calculation, and noise cancellation.

[0058] For example, noise floor modeling can extract the lowest 5% amplitude point of each periodic signal as a noise reference, and establish a noise statistical model within the sliding window (such as calculating the mean + variance).

[0059] For example, the adaptive threshold calculation may dynamically calculate the threshold according to the noise model using formula (1).

[0060] Formula (1)

[0061] Where k is the adaptive coefficient, which is automatically adjusted with the SNR; μ is the noise mean; and σ is the noise standard deviation.

[0062] Exemplarily, noise elimination may perform zeroing processing on signal points below a threshold value and perform smooth transition processing on signals close to the threshold value, thereby preserving the phase distribution characteristics of the signal.

[0063] In some embodiments, the PRPD matrix construction module 150 may perform normalization, phase alignment, and PRPD matrix generation.

[0064] Illustratively, the normalization process may perform amplitude normalization on the processed signal in the interval [0, 1], and then employ piecewise normalization to maintain the relative relationship between different discharge types.

[0065] For example, phase alignment can detect the discharge start phase in each cycle, establish a phase offset compensation mechanism, and ensure phase consistency under different measurement conditions.

[0066] For example, the PRPD matrix can be generated by performing phase-amplitude two-dimensional spatial distribution statistics on the normalized signal, and then using an improved kernel density estimation method to smooth the distribution to generate a standardized PRPD grayscale image as the PRPD matrix.

[0067] In some embodiments, the feature extraction module 160 may extract statistical features, distribution features, shape features, and time series features of the PRPD matrix.

[0068] Exemplarily, statistical features include traditional features such as skewness, kurtosis, and pulse number; distribution features include improved three-dimensional features of phase distribution features (phase distribution entropy Hφ), amplitude distribution features (amplitude distribution entropy Hq), and discharge number features (discharge number entropy Hn); shape features are PRPD pattern features based on image processing; and timing features are correlated with the cycle of discharge pulses.

[0069] In some embodiments, the diagnostic model 170 can perform diagnostic model construction and online diagnosis.

[0070] For example, the diagnostic model can be constructed by using a deep convolutional network to process PRPD image features, and combining statistical features to build a multimodal fusion model, while designing a special loss function.

[0071] For example, online diagnosis can process signals in real time and generate diagnosis results.

[0072] In some embodiments, the result output module 180 can output the type of partial discharge (such as internal discharge, surface discharge, corona discharge, etc.), evaluate the degree of insulation degradation, and issue an early warning.

[0073] This system effectively eliminates periodic and floor noise interference in industrial environments through dynamic noise suppression and adaptive noise reduction technologies, accurately extracting weak partial discharge signals. It then combines segmented normalization and phase compensation mechanisms to standardize the PRPD matrix. Furthermore, it combines multimodal features (statistical, image, and time series) with a dual-channel deep learning model to improve classification accuracy. This, in turn, enhances the accuracy of insulation detection for power equipment.

[0074] Refer to the following Figure 2 The insulation status detection method of electric power equipment provided in an embodiment of the present application is described.

[0075] Figure 2 The method flow chart of the method for detecting the insulation status of electric power equipment provided in the embodiment of the present application, the subject for executing the method may be the above Figure 1 The electric power equipment insulation status detection system shown may also be various devices / modules in the electric power equipment insulation status detection system, such as an integrated circuit or a chip, which is not specifically limited in the embodiments of the present application.

[0076] For example, Figure 2 As shown, the method for detecting the insulation status of electric power equipment provided in the embodiment of the present application may include the following steps S201 to S204:

[0077] S201 : Acquire a partial discharge signal of an electric power device, and divide the partial discharge signal into a plurality of continuous discharge cycle segments based on a power supply frequency.

[0078] In an embodiment of the present application, a signal acquisition device (such as a UHF sensor or a pulse current sensor) can be used to obtain partial discharge signals during the operation of power equipment, and the continuous signal can be divided into independent periodic segments according to the power frequency characteristics of the power system (such as 50 Hz or 60 Hz) for subsequent periodic noise processing.

[0079] In some embodiments, an improved zero-crossing detection algorithm can be used to achieve cycle segmentation. First, the collected signal is bandpass filtered (passband range 30MHz-300MHz) to eliminate low-frequency interference and high-frequency noise. Then, the cycle boundary is determined by detecting the time interval of the signal zero crossing point. When the deviation between consecutive zero crossing intervals and the power supply cycle exceeds 5%, the sliding window correction mechanism is triggered.

[0080] For example, consider a power supply frequency of 50 Hz and a standard cycle of 20 ms. When the sampling rate of the acquired signal is 1 MHz, each cycle corresponds to approximately 20,000 sampling points. The improved zero-crossing detection algorithm sets a threshold of 15% of the effective value of the signal amplitude and employs a sliding average filter combined with a phase-locked algorithm. For example, if a zero-crossing point is detected at the 10,245th sampling point in a given cycle, the change in the signal slope within a sliding window (with a window size of 100 sampling points) is used to determine whether it is a valid zero-crossing point, thus avoiding missegmentation caused by harmonic interference.

[0081] S202 : Dynamically identify and limit the amplitude of the partial discharge signal in each discharge cycle to obtain a valid pulse signal.

[0082] In the embodiment of the present application, the effective pulse signal is a signal that retains pulse phase information.

[0083] In some embodiments, Gaussian mixture model identification may be performed on the pulse amplitude of the partial discharge signal to obtain abnormal amplitude pulses.

[0084] The amplitude of the abnormal amplitude pulse is greater than the preset amplitude.

[0085] In the embodiment of the present application, the preset amplitude may be a manually set value, which may be flexibly adjusted according to actual scenarios. For example, the preset amplitude may be 0.5V.

[0086] Optionally, the preset amplitude may also be determined by statistical analysis of historical noise data, such as taking the 95% quantile of the amplitude distribution in all periodic segments, or dynamically adjusted according to the noise floor of the sensor.

[0087] For example, in an industrial environment, by collecting signals of 100 no-discharge cycles, the 95% quantile of the amplitude is calculated to be 0.5V, so the preset amplitude is set to 0.5V, and pulses greater than this value are preliminarily determined to be abnormal.

[0088] Optionally, a Gaussian mixture model may be used to perform probability modeling on the pulse amplitude distribution within a single cycle, and normal discharge pulses and abnormal noise pulses may be distinguished through statistical characteristics.

[0089] For example, the expectation-maximization (EM) algorithm can be used to train the GMM, dividing the pulse amplitude into two Gaussian components: one corresponding to normal discharge pulses (mean μ1, standard deviation σ1) and the other corresponding to noise pulses (mean μ2>μ1, standard deviation σ2). Bayesian decision theory is used to calculate the probability of each pulse belonging to the noise component, and pulses with a probability greater than a preset amplitude (such as 0.7) are marked as abnormal amplitude pulses.

[0090] Specifically, if the statistical distribution of pulse amplitudes within a certain period conforms to a double-Gaussian model, where the parameters of the normal discharge component are μ1 = 0.2 V and σ1 = 0.05 V, and the parameters of the noise component are μ2 = 0.8 V and σ2 = 0.2 V, and a pulse with an amplitude of 0.9 V is 0.85 and the posterior probability that it belongs to the noise component is greater than 0.7, then the pulse is marked as an abnormal amplitude pulse.

[0091] Furthermore, the abnormal amplitude pulse is subjected to amplitude limiting and nonlinear filtering to obtain a valid pulse signal.

[0092] In some embodiments, the phase information of the abnormal amplitude pulse can be retained to obtain the initial amplitude pulse, which is then limited and smoothed using a median filter algorithm to obtain a valid pulse signal.

[0093] Among them, the window size of the median filtering algorithm is dynamically adjusted according to the signal sampling rate; the window size is dynamically adjusted to adapt to the noise characteristics under different sampling rates, avoiding signal blur or noise residue caused by a fixed window.

[0094] Exemplarily, the window size can be calculated using formula (2).

[0095] Formula (2)

[0096] Where fs is the sampling rate (Hz) and round is the rounding function. When the sampling rate changes, N is adjusted accordingly to ensure that the filter window covers a time length of approximately 10 μs (i.e., a 100 kHz sampling rate corresponds to a window size of 10).

[0097] For example, the amplitude influence of the noise pulse can be suppressed by amplitude limiting processing while retaining its phase information to maintain the spatiotemporal characteristics of the discharge pattern, and then nonlinear filtering can be used to eliminate the signal distortion caused by amplitude limiting.

[0098] Specifically, taking the preset amplitude of 0.5V as an example, for an abnormal pulse with an amplitude of 1.0V, the amplitude can be limited to 1.2 times the preset amplitude (i.e., 0.5V×1.2=0.6V), while retaining the phase information of the abnormal pulse (such as the position of 120° within the cycle). Then, a median filter algorithm with a window size of 10 is used for processing. For example, the median of the 9 sampling points (4 before and after the abnormal pulse) is taken and replaced to obtain a valid pulse signal.

[0099] In this way, this application uses a Gaussian mixture model to model the probability distribution of the pulse amplitude, accurately distinguishes noise from effective discharge pulses, combines limiting processing to suppress abnormal amplitude interference, and uses nonlinear filtering to eliminate signal distortion after limiting, retaining the integrity of the phase information, and ensuring the authenticity of the spatiotemporal characteristics of subsequent PRPD analysis.

[0100] S203 : performing bottom noise adaptive elimination processing on the effective pulse signal based on a sliding window model to generate a noise reduction signal.

[0101] In some embodiments, the lowest amplitude region of the effective pulse signal may be extracted as a noise reference, and then a noise statistical model within the sliding window may be established based on the noise reference.

[0102] For example, for each periodic signal, the 5% data points with the lowest amplitude are extracted as noise parameters. Then, a sliding window with a length of L = 200 periods is used to calculate the mean μ and standard deviation σ of the noise reference within the window, and a noise statistics module (μ, σ²) is established.

[0103] Specifically, let's assume that the amplitude of a valid pulse signal in a certain period ranges from -1V to 1V. We can extract the 5% lowest amplitude points (i.e., points less than -0.8V) as the noise reference. The sliding window contains the noise reference data for the most recent 200 periods. The calculated noise mean μ for the current window is -0.9V, with a standard deviation σ of 0.1V. This indicates that the current noise floor is concentrated around -0.9V, with a fluctuation range of ±0.1V.

[0104] Furthermore, an adaptive threshold can be dynamically calculated according to a noise statistical model, and valid pulse signals below the adaptive threshold are reset to zero to obtain a noise reduction signal.

[0105] In some embodiments, the adaptive threshold may be calculated as follows: first, the mean and standard deviation of the noise statistical model are calculated, and then, based on the mean and standard deviation, the adaptive threshold is determined in combination with the adaptive coefficient.

[0106] The adaptive coefficient automatically adjusts as the signal-to-noise ratio changes. The threshold strictness can be dynamically adjusted based on signal quality, loosening the threshold to retain weak signals at low signal-to-noise ratios and tightening it to enhance noise suppression at high signal-to-noise ratios.

[0107] For example, adaptive threshold adjustment can be achieved by calculating the signal-to-noise energy ratio (SNR) in real time and mapping it to different k value intervals. The SNR is calculated using the signal energy to noise energy ratio within a sliding window with a window length of 100 cycles.

[0108] Specifically, when SNR=3dB (low signal-to-noise ratio), k=2.0 and the threshold is μ+2σ to avoid misjudging weak discharge signals as noise; when SNR=15dB (high signal-to-noise ratio), k=1.0 and the threshold is μ+σ to strictly filter out the noise floor and highlight the discharge signal.

[0109] Exemplarily, the adaptive threshold can be calculated using formula (3).

[0110] Formula (3)

[0111] Where μ is the noise mean, σ is the noise standard deviation, and k is the adaptive coefficient.

[0112] In the embodiment of the present application, the adaptive coefficient k is dynamically adjusted according to the SNR through formula (4).

[0113] Formula (4)

[0114] Specifically, taking the established noise statistical model with μ = -0.85V and σ = 0.05V as an example, if the calculated noise RMS value is 0.1V and the signal RMS value is 0.3V, then SNR = 20lg (0.3 / 0.1) = 9.54dB, which falls within the range of 5dB ≤ SNR < 10dB. Therefore, k = 1.5, and the adaptive threshold is -0.85 + 1.5 × 0.05 = -0.775V. Therefore, signal points below -0.775V can be identified as noise and zeroed to obtain the final noise-reduced signal.

[0115] In this way, this application constructs a noise reference by extracting the lowest amplitude area of ​​the signal, and combines the sliding window to calculate the noise mean and variance in real time, dynamically reflecting the time-varying characteristics of the noise base, making the threshold calculation more in line with actual working conditions and enhancing the adaptability to complex industrial environments.

[0116] S204: construct a PRPD matrix according to the noise reduction signal, and output an insulation status diagnosis result according to the PRPD matrix.

[0117] The PRPD matrix is ​​a matrix in the phase-amplitude two-dimensional space.

[0118] In some embodiments, the above-mentioned PRPD matrix can be constructed in the following manner: first, the noise reduction signal is amplitude normalized and phase aligned, and the distribution density of the processed noise reduction signal in the phase-amplitude two-dimensional space is determined, and then based on the distribution density and kernel density estimation method, a standardized PRPD grayscale image is generated as the PRPD matrix.

[0119] In the embodiment of the present application, amplitude normalization can eliminate amplitude differences under different measurement conditions; phase alignment can compensate for phase offsets caused by installation or environmental factors.

[0120] For example, piecewise linear normalization can be used to divide the amplitude into three intervals (such as 0-0.3, 0.3-0.7, and 0.7-1), which are respectively mapped to the interval [0,1] to retain the relative relationship between the amplitudes of different discharge types (such as corona discharge and internal discharge); then the discharge starting phase in each cycle is detected by the cross-correlation method, the phase difference with the reference cycle (such as the first cycle) is calculated, and the phase offset is compensated by linear interpolation. Finally, the Gaussian kernel function K(h) is used to smooth the distribution of the phase-amplitude two-dimensional space. The kernel bandwidth h is set to 1° on the phase axis and 0.05 on the amplitude axis. The kernel density is calculated by formula (5).

[0121] Formula (5)

[0122] in, is the normalized phase-amplitude point, n is the total number of discharge points, is the phase axis kernel bandwidth, is the amplitude axis kernel bandwidth.

[0123] Specifically, consider a discharge pulse with an amplitude of 0.5V and a phase of 60°. After piecewise normalization, the amplitude becomes (0.5-0.3) / (0.7-0.3)=0.5 (falling in the 0.3-0.7 interval). After phase alignment, if the starting phase of the reference cycle is 0° and the starting phase of the current cycle is 5°, the compensated phase becomes 60°-5°=55°. Kernel density estimation uses a Gaussian kernel centered at (55°, 0.5) and weighted summation within the range of ±1° on the phase axis and ±0.05 on the amplitude axis. The smoothed density values ​​are then mapped to grayscale pixel values ​​(0-255).

[0124] It should be noted that the present application adopts piecewise linear normalization in order to set dynamic regions for different discharge types.

[0125] In this way, the present application eliminates the difference in sensor sensitivity through segmented amplitude normalization, combines phase alignment to compensate for offset, uses kernel density estimation to smooth the phase-amplitude distribution, and generates a standardized PRPD grayscale image to ensure the consistency of diagnostic results of different devices.

[0126] In some embodiments, after obtaining the PRPD matrix, the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics of the PRPD matrix can be extracted, and the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics can be input into a multimodal fusion model to output the insulation status diagnosis results.

[0127] Among them, the multimodal fusion model is constructed using a deep convolutional neural network.

[0128] For example, the statistical features can be extracted by calculating the skewness (reflecting the symmetry of the distribution), the kurtosis (reflecting the degree of the spike), and the total number of pulses (TNP) (reflecting the discharge intensity) of the PRPD distribution.

[0129] For example, the distribution features can be extracted based on the improved three-dimensional features Hφ, Hq, and Hn using formula (VI).

[0130] Formula (6)

[0131] in, is the probability density of the phase-amplitude interval.

[0132] For example, the shape feature can be used to extract the contour features of the PRPD pattern (such as geometric parameters such as roundness, aspect ratio, convex hull defects, etc.) through Canny edge detection.

[0133] It should be noted that the Canny algorithm can effectively preserve the edge continuity of the discharge pattern in the PRPD image, and the roundness parameter can distinguish between corona discharge (quasi-circular) and internal discharge (irregular shape).

[0134] For example, the time series characteristics can be obtained by analyzing the correlation of PRPD patterns over 50 consecutive cycles and calculating the phase offset variance, amplitude fluctuation coefficient, etc. between cycles.

[0135] For example, if the statistical characteristics of a PRPD matrix are: skewness = 1.2, kurtosis = 3.8, NP = 200; distribution characteristics Hφ = 2.1, Hq = 1.8, Hn = 2.5; shape characteristics show an elliptical outline with a roundness of 0.7 and an aspect ratio of 1.5; and temporal characteristics with a phase offset variance of 2° and an amplitude fluctuation coefficient of 0.1, then these features can be input into a multimodal fusion model (such as a fusion model of ResNet-18 and fully connected layers). After training with a contrastive loss function, the probability of the output recognition result being "internal discharge" is 0.92.

[0136] Among them, the formula of the loss function can be formula (7).

[0137] Formula (7)

[0138] Among them, N is the number of samples, y is the label (1 means the same class, 0 means different class), is the eigenvector distance, and m is the margin parameter (set to 0.5).

[0139] In the insulation status detection method of power equipment provided in the embodiment of the present application, first, the local discharge signal is dynamically segmented based on the power supply frequency, and the continuous signal is accurately isolated into periodic segments, effectively avoiding periodic noise aliasing; on this basis, through dynamic identification and amplitude limitation processing, the interference of high-amplitude noise pulses on the effective discharge signal is targeted, while the pulse phase information is completely retained to ensure the authenticity of the discharge time and space distribution characteristics; then, a sliding window model is used to adaptively eliminate the bottom noise of the processed signal, and through real-time noise modeling and dynamic threshold adjustment, the masking effect of low-amplitude base noise on weak discharge signals is filtered out; finally, a standardized PRPD matrix is ​​constructed based on the phase-amplitude two-dimensional space to eliminate the interference of phase offset on the diagnosis result, thereby improving the detection accuracy of whether the power equipment is insulated.

[0140] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the method. In order to realize the above functions, the insulation status detection device of the power equipment or the electronic device includes a hardware structure and / or software module corresponding to the execution of each function. It should be easy for those skilled in the art to realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0141] In the embodiment of the present application, the functional modules of the insulation status detection device of the electric equipment or the electronic device can be divided according to the above method. For example, the insulation status detection device of the electric equipment or the electronic device can include functional modules corresponding to the functional divisions, or two or more functions can be integrated into one processing module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical functional division. In actual implementation, there may be other division methods.

[0142] Figure 3 This is a structural diagram of an apparatus for detecting insulation status of electric equipment provided in an embodiment of the present application. The apparatus 300 for detecting insulation status of electric equipment includes: an acquiring unit 301 , a processing unit 302 , a generating unit 303 , and a constructing unit 304 .

[0143] Among them: the acquisition unit 301 is used to obtain the partial discharge signal of the power equipment and divide the partial discharge signal into multiple continuous discharge cycle segments based on the power supply frequency; the processing unit 302 is used to dynamically identify and limit the amplitude of the partial discharge signal in each discharge cycle segment to obtain a valid pulse signal, which is a signal that retains the pulse phase information; the generation unit 303 is used to perform bottom noise adaptive elimination processing on the effective pulse signal based on the sliding window model to generate a noise reduction signal; the construction unit 304 is used to construct a PRPD matrix based on the noise reduction signal and output the insulation status diagnosis result based on the PRPD matrix. The PRPD matrix is ​​a matrix in the two-dimensional space of phase and amplitude.

[0144] In some embodiments, the processing unit 302 is specifically used to: perform Gaussian mixture model identification on the pulse amplitude of the partial discharge signal to obtain an abnormal amplitude pulse, where the amplitude of the abnormal amplitude pulse is greater than a preset amplitude; and perform limiting processing and nonlinear filtering processing on the abnormal amplitude pulse to obtain a valid pulse signal.

[0145] In some embodiments, the above-mentioned processing unit 302 is specifically used to: retain the phase information of the abnormal amplitude pulse to obtain the initial amplitude pulse; limit the initial amplitude pulse, and use the median filtering algorithm to smooth the initial amplitude pulse after the limiting processing to obtain a valid pulse signal, wherein the window size of the median filtering algorithm is dynamically adjusted according to the signal sampling rate.

[0146] In some embodiments, the above-mentioned generation unit 303 is specifically used to: extract the lowest amplitude area of ​​the effective pulse signal as a noise reference; establish a noise statistical model within the sliding window based on the noise reference; dynamically calculate the adaptive threshold according to the noise statistical model, and zero the effective pulse signal below the adaptive threshold to obtain a noise reduction signal.

[0147] In some embodiments, the generating unit 303 is specifically configured to: calculate the mean and standard deviation of the noise statistical model; determine an adaptive threshold based on the mean and standard deviation in combination with an adaptive coefficient, wherein the adaptive coefficient is automatically adjusted as the signal-to-noise ratio changes.

[0148] In some embodiments, the construction unit 304 is specifically used to: perform amplitude normalization and phase alignment processing on the noise reduction signal, and determine the distribution density of the processed noise reduction signal in the phase-amplitude two-dimensional space; based on the distribution density and kernel density estimation method, generate a standardized PRPD grayscale image as a PRPD matrix.

[0149] In some embodiments, the above-mentioned construction unit 304 is specifically used to: extract the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics of the PRPD matrix; input the statistical characteristics, distribution characteristics, shape characteristics and time series characteristics into the multimodal fusion model, and output the insulation status diagnosis results; wherein the multimodal fusion model is constructed using a deep convolutional neural network.

[0150] In the insulation status detection device for electric power equipment provided in the embodiment of the present application, first, the local discharge signal is dynamically segmented based on the power supply frequency, and the continuous signal is accurately isolated into periodic segments, effectively avoiding periodic noise aliasing; on this basis, through dynamic identification and amplitude limitation processing, the interference of high-amplitude noise pulses on the effective discharge signal is targeted, while the pulse phase information is completely retained to ensure the authenticity of the discharge time and space distribution characteristics; then, a sliding window model is used to adaptively eliminate the bottom noise of the processed signal, and through real-time noise modeling and dynamic threshold adjustment, the masking effect of low-amplitude base noise on weak discharge signals is filtered out; finally, a standardized PRPD matrix is ​​constructed based on the phase-amplitude two-dimensional space to eliminate the interference of phase offset on the diagnosis result, thereby improving the detection accuracy of whether the electric power equipment is insulated.

[0151] Regarding the apparatus in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0152] Figure 4 This is a structural diagram of an electronic device provided in an embodiment of the present application. Figure 4 As shown, the electronic device 400 includes but is not limited to: a processor 401 and a memory 402 .

[0153] The memory 402 is used to store executable instructions of the processor 401. It is understandable that the processor 401 is configured to execute instructions to implement the method for detecting the insulation status of the power equipment in the above embodiment.

[0154] It should be noted that those skilled in the art can understand that Figure 4 The electronic device structure shown in the figure does not limit the electronic device, and the electronic device may include Figure 4 More or fewer components may be shown, or certain components may be combined, or the components may be arranged differently.

[0155] The processor 401 is the control center of the electronic device. It connects the various parts of the entire electronic device using various interfaces and lines. By running or executing software programs and / or modules stored in the memory 402 and accessing data stored in the memory 402, it performs various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. The processor 401 may include one or more processing units. Optionally, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly handles wireless communications. It is understood that the modem processor may not be integrated into the processor 401.

[0156] Memory 402 can be used to store software programs and various data. Memory 402 may primarily include a program storage area and a data storage area. The program storage area may store an operating system and application programs required by at least one functional module (such as a determination unit, a processing unit, etc.). Furthermore, memory 402 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0157] In an exemplary embodiment, a computer-readable storage medium including instructions is further provided, such as a memory 402 including instructions. The above instructions can be executed by the processor 401 of the electronic device 400 to implement the method for detecting the insulation status of the power equipment in the above embodiment.

[0158] In actual implementation, Figure 3 The steps performed by the acquisition unit 301, the processing unit 302, the generation unit 303 and the construction unit 304 in Figure 4 The processor 401 in the embodiment calls the computer program stored in the memory 402. The specific execution process can be referred to the description of the method part in the above embodiment, which will not be repeated here.

[0159] Optionally, the computer-readable storage medium may be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0160] In an exemplary embodiment, the present application also provides a computer program product including one or more instructions, which can be executed by the processor 401 of the electronic device to complete the method for detecting the insulation status of the power equipment in the above embodiment.

[0161] It should be noted that when the instructions in the above-mentioned computer-readable storage medium or one or more instructions in the computer program product are executed by the processor of the electronic device, the various processes of the above-mentioned method embodiment are implemented and the same technical effect as the above-mentioned method can be achieved. To avoid repetition, they will not be repeated here.

[0162] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete the full classification or partial functions described above.

[0163] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, 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, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0164] The units described as separate components may or may not be physically separate, and the components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0165] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0166] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application, or the part that contributes to the prior art, or the entire classification part or part of the technical solution can be embodied in the form of a software product. The software product is stored in a storage medium and includes a number of instructions for enabling a device (which can be a single-chip microcomputer, chip, etc.) or a processor to execute the entire classification part or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, ROM, RAM, disk or optical disk, etc., various media that can store program code.

[0167] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for detecting the insulation status of an electric power device, characterized in that: The method comprises: Acquire a partial discharge signal from the power equipment, and segment the partial discharge signal into a plurality of consecutive discharge cycle segments using an improved zero-crossing detection algorithm based on the power frequency, triggering a sliding window correction mechanism when the deviation between consecutive zero-crossing intervals and the power cycle exceeds a preset threshold; Performing Gaussian mixture model identification on the pulse amplitude of the partial discharge signal to obtain an abnormal amplitude pulse, where the amplitude of the abnormal amplitude pulse is greater than a preset amplitude; The abnormal amplitude pulse is subjected to a limiting process and a nonlinear filtering process to obtain a valid pulse signal, wherein the valid pulse signal is a signal that retains the pulse phase information, wherein the nonlinear filtering adopts a median filtering algorithm, and the size of the median filtering window is dynamically adjusted according to the signal sampling rate to ensure that the time length covered by the filtering window is stable; Extracting the lowest amplitude region of the effective pulse signal as a noise reference, and establishing a noise statistical model within a sliding window based on the noise reference; Dynamically calculating an adaptive threshold according to the noise statistical model, and performing zeroing processing on valid pulse signals below the adaptive threshold to obtain a noise reduction signal, wherein the adaptive threshold is determined based on the noise mean, the noise standard deviation, and an adaptive coefficient. The adaptive coefficient is dynamically adjusted according to the signal-to-noise ratio, relaxing the threshold to retain weak signals when the signal-to-noise ratio is low, and tightening the threshold to enhance noise suppression when the signal-to-noise ratio is high; After amplitude normalization and phase alignment processing are performed on the noise reduction signal, the phase-amplitude two-dimensional spatial distribution is smoothed based on the kernel density estimation method to generate a standardized phase-resolved partial discharge PRPD grayscale image as a PRPD matrix, and the insulation status diagnosis result is output according to the PRPD matrix. The amplitude normalization adopts piecewise linear normalization, and the phase alignment adopts the cross-correlation method to compensate for the phase offset.

2. The method according to claim 1, characterized in that The performing of amplitude limiting and nonlinear filtering on the abnormal amplitude pulse to obtain the valid pulse signal includes: retaining the phase information of the abnormal amplitude pulse to obtain an initial amplitude pulse; The initial amplitude pulse is subjected to a limiting process, and the initial amplitude pulse after the limiting process is smoothed using a median filtering algorithm to obtain the effective pulse signal. The window size of the median filtering algorithm is dynamically adjusted according to the signal sampling rate.

3. The method according to claim 1, characterized in that The dynamically calculating the adaptive threshold according to the noise statistical model includes: Calculating the mean and standard deviation of the noise statistical model; The adaptive threshold is determined based on the mean and standard deviation in combination with an adaptive coefficient, and the adaptive coefficient is automatically adjusted as the signal-to-noise ratio changes.

4. The method according to claim 1, wherein Outputting the insulation status diagnosis result according to the PRPD matrix includes: Extracting statistical features, distribution features, shape features and time series features of the PRPD matrix; Inputting the statistical features, the distribution features, the shape features and the time series features into a multimodal fusion model, and outputting the insulation status diagnosis result; Among them, the multimodal fusion model is constructed using a deep convolutional neural network.

5. A device for detecting insulation status of electric power equipment, characterized in that: The device comprises: an acquisition unit, configured to acquire a partial discharge signal of the power equipment, and segment the partial discharge signal into a plurality of consecutive discharge cycle segments using an improved zero-crossing detection algorithm based on the power frequency, and trigger a sliding window correction mechanism when a deviation between consecutive zero-crossing intervals and the power cycle exceeds a preset threshold; A processing unit for: Performing Gaussian mixture model identification on the pulse amplitude of the partial discharge signal to obtain an abnormal amplitude pulse, where the amplitude of the abnormal amplitude pulse is greater than a preset amplitude; The abnormal amplitude pulse is subjected to a limiting process and a nonlinear filtering process to obtain a valid pulse signal, wherein the valid pulse signal is a signal that retains the pulse phase information, wherein the nonlinear filtering adopts a median filtering algorithm, and the size of the median filtering window is dynamically adjusted according to the signal sampling rate to ensure that the time length covered by the filtering window is stable; Generates a unit for: Extracting the lowest amplitude region of the effective pulse signal as a noise reference, and establishing a noise statistical model within a sliding window based on the noise reference; Dynamically calculating an adaptive threshold according to the noise statistical model, and performing zeroing processing on valid pulse signals below the adaptive threshold to obtain a noise reduction signal, wherein the adaptive threshold is determined based on the noise mean, the noise standard deviation, and an adaptive coefficient. The adaptive coefficient is dynamically adjusted according to the signal-to-noise ratio, relaxing the threshold to retain weak signals when the signal-to-noise ratio is low, and tightening the threshold to enhance noise suppression when the signal-to-noise ratio is high; A construction unit is used to perform amplitude normalization and phase alignment processing on the noise reduction signal, then smooth the phase-amplitude two-dimensional spatial distribution based on a kernel density estimation method, generate a standardized PRPD grayscale image as a PRPD matrix, and output an insulation status diagnosis result according to the PRPD matrix. The amplitude normalization adopts piecewise linear normalization, and the phase alignment adopts a cross-correlation method to compensate for phase offset.

6. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the method according to any one of claims 1 to 4.

7. A computer-readable storage medium storing instructions, characterized in that: When a computer executes the instruction, the computer performs the method according to any one of claims 1 to 4.

Images