Partial discharge signal detection and segmentation method and system based on wavelet bayesian threshold
By employing wavelet Bayesian thresholding and zero-phase filtering techniques, adaptive noise reduction and automatic segmentation of partial discharge signals from high-voltage cables were achieved. This solved the problems of adaptability and accuracy in partial discharge signal processing in existing technologies, and improved the robustness of signal processing and segmentation accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID SICHUAN ELECTRIC POWER CORP ELECTRIC POWER RES INST
- Filing Date
- 2026-06-18
- Publication Date
- 2026-07-21
Smart Images
Figure CN122430656A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of online power equipment detection technology, and relates to, but is not limited to, a method and system for detecting and segmenting partial discharge signals based on wavelet Bayes threshold. Background Technology
[0002] High-voltage power cables are core equipment in power transmission and distribution networks, widely used in urban power grids, industrial parks, and inter-regional power transmission projects. Their operating status directly determines the safety, stability, and reliability of the power grid. During manufacturing, laying, and long-term service, cables are susceptible to multiple factors such as mechanical damage, changes in environmental temperature and humidity, overvoltage surges, and chemical corrosion, leading to insulation aging and deterioration, which in turn triggers partial discharge. Partial discharge is a core characteristic of early-stage defects in cable insulation. Its signal carries crucial information such as the degree of insulation degradation, defect type, and defect location. Accurate processing of partial discharge signals is a key means of assessing cable insulation status and preventing cable breakdown faults. Currently, the industry commonly uses high-frequency current transformers (HFCTs) for online monitoring. This technology requires no power outages, is easy to install, and causes no equipment damage. However, in complex electromagnetic environments such as substations and cable trenches, the collected partial discharge signals are easily mixed with noise data such as Gaussian white noise, power frequency harmonics, narrowband interference, and random switching pulse interference, resulting in distortion of the effective pulse waveform and a significant reduction in the signal-to-noise ratio, severely hindering subsequent signal analysis and defect diagnosis.
[0003] To address the noise issues in partial discharge signals, various partial discharge denoising algorithms have been implemented in engineering applications, but they generally suffer from technical limitations. For example, Fourier transform filtering is only suitable for suppressing stationary noise and struggles to handle transient non-stationary pulses in partial discharges, easily causing pulse front distortion. Empirical Mode Decomposition (EMD) can achieve adaptive signal decomposition, but it suffers from mode aliasing and endpoint drift problems, and cannot balance denoising effectiveness and signal fidelity in low signal-to-noise ratio scenarios. Variational Mode Decomposition (VMD) effectively solves the mode aliasing problem, but its parameters depend on manual settings, resulting in poor adaptive performance. Furthermore, existing algorithms mostly focus on optimizing a single denoising algorithm, lacking precise pulse front localization methods and standardized automatic partial discharge event segmentation systems. The project uses a fixed threshold to extract signals, which is prone to problems such as noise misjudgment, missed detection of weak pulses, and interference fragments. This not only increases the computational load of fault diagnosis, but also significantly reduces the accuracy of defect identification and location. It is difficult to adapt to the processing needs of low signal-to-noise ratio partial discharge signals on site, and cannot meet the practical requirements of high-voltage cable insulation monitoring. Summary of the Invention
[0004] In view of this, embodiments of the present invention provide a partial discharge signal detection and segmentation method and system based on wavelet Bayesian thresholding, aiming to overcome the technical problems of weak noise reduction adaptability, low wavefront detection accuracy, inability to automatically complete the regular segmentation of partial discharge events, insufficient robustness in strong interference scenarios, and difficulty in adapting to low signal-to-noise ratio partial discharge signals collected by HFCT sensors in the field. By performing adaptive noise reduction, precise wavefront positioning, and automatic event segmentation on the partial discharge signal, the present invention balances noise reduction effect and signal fidelity, providing reliable pre-processing support for cable insulation defect diagnosis and discharge source location, and meeting the practical needs of field engineering.
[0005] The specific technical solutions of this invention are as follows: This invention provides a method for detecting and segmenting partial discharge signals based on wavelet Bayesian thresholding, comprising: The first partial discharge signal was acquired during the test, and the first partial discharge signal was adaptively denoised using a translation-invariant wavelet transform combined with a Bayesian adaptive threshold to obtain the second partial discharge signal. A cascaded fourth-order zero-phase Butterworth high-pass filter is used to filter the second partial discharge signal, removing DC baseline drift and low-frequency narrowband interference signals to obtain the third partial discharge signal; The third partial discharge signal is sequentially subjected to signal amplitude enhancement and amplitude threshold pruning to obtain an effective pulse sequence; The wavefront start position of the effective pulse sequence is identified by using a sliding window energy comparison method, and the pulse end position of the effective pulse sequence is determined by the attenuation trend of the pulse amplitude. The waveform truncation start position is determined based on the wavefront start position and the first extended sampling interval, and the waveform truncation end position is determined based on the pulse end position and the second extended sampling interval. Based on the waveform truncation start position and the waveform truncation end position, the complete effective waveform of partial discharge is truncated, thereby realizing the automatic segmentation of the partial discharge signal.
[0006] This invention provides a partial discharge signal detection and segmentation system based on wavelet Bayesian thresholding, comprising: The system includes a wavelet noise reduction module, a zero-phase filtering module, a pulse enhancement and screening module, a wavefront identification module, a waveform truncation interval determination module, and a signal segmentation module, wherein: The wavelet denoising module is used to acquire the first partial discharge signal during the test process, and to perform adaptive denoising processing on the first partial discharge signal by combining translation-invariant wavelet transform with Bayesian adaptive thresholding to obtain the second partial discharge signal. The zero-phase filtering module is used to filter the second partial discharge signal by employing a cascaded fourth-order zero-phase Butterworth high-pass filter to remove DC baseline drift and low-frequency narrowband interference signals, thereby obtaining the third partial discharge signal. The pulse enhancement and filtering module is used to sequentially enhance the signal amplitude and prune the amplitude threshold of the third partial discharge signal to obtain an effective pulse sequence; The wavefront identification module is used to identify the wavefront start position of the effective pulse sequence by using a sliding window energy comparison method, and to determine the pulse end position of the effective pulse sequence by the attenuation trend of the pulse amplitude. The waveform truncation interval determination module is used to determine the waveform truncation start position based on the wavefront start position and the first extended sampling interval, and to determine the waveform truncation end position based on the pulse end position and the second extended sampling interval; The signal segmentation module is used to extract the complete effective waveform of partial discharge based on the waveform extraction start position and the waveform extraction end position, thereby realizing the automatic segmentation of the partial discharge signal.
[0007] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: In this embodiment of the invention, the first partial discharge signal is adaptively denoised by employing translation-invariant wavelet transform combined with Bayesian adaptive thresholding to suppress white noise and electromagnetic interference and avoid pseudo-Gibbs effects. Then, a cascaded fourth-order zero-phase Butterworth high-pass filter is used to eliminate DC components and low-frequency noise while fully preserving the pulse leading-edge characteristics. Next, the filtered signal is sequentially subjected to square amplitude enhancement and adaptive thresholding to highlight effective pulses and eliminate minor interferences. Finally, the starting and ending positions of the pulse wavefront are accurately identified through sliding window energy analysis, and the complete partial discharge waveform is automatically extracted by combining the reserved extended sampling interval. Automatic segmentation of partial discharge events can be achieved without relying on manual operation, thereby improving the identification and positioning accuracy of partial discharge signals. Attached Figure Description
[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a flowchart illustrating the partial discharge signal detection and segmentation method based on wavelet Bayes thresholding provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the partial discharge signal detection and segmentation system based on wavelet Bayes threshold provided in an embodiment of the present invention. Detailed Implementation
[0009] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. The following embodiments are used to illustrate the present invention, but are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0010] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.
[0011] It should be noted that the terms "first, second, and third" used in the embodiments of the present invention are only used to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first, second, and third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present invention described herein can be implemented in an order other than that illustrated or described herein.
[0012] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which these embodiments of the invention pertain. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.
[0013] Figure 1 This is a flowchart illustrating a partial discharge signal detection and segmentation method based on wavelet Bayesian thresholding provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the method includes at least the following steps: Step S110: Acquire the first partial discharge signal during the test process, and use translation-invariant wavelet transform combined with Bayesian adaptive thresholding to perform adaptive noise reduction processing on the first partial discharge signal to obtain the second partial discharge signal.
[0014] Specifically, the original partial discharge signal during the test is acquired by a high-frequency current transformer (HFCT). The original partial discharge signal is an analog signal, which is converted by an analog-to-digital converter (AD) to generate the first partial discharge signal x1(t).
[0015] In this embodiment, an HFCT with a bandwidth range of 0.1MHz to 30MHz is selected to collect the original partial discharge signal during the test process. The sampling frequency is adjusted according to the specific bandwidth of the HFCT, and is usually 2.5 to 5 times the highest cutoff frequency of the HFCT.
[0016] Furthermore, after obtaining the first partial discharge signal x1(t), the first partial discharge signal is translated according to the translation step size to obtain the translation signal; then the current translation operation count is recorded. When the current translation operation count is less than or equal to the preset maximum translation count, the adaptive noise reduction processing step is repeated.
[0017] The adaptive noise reduction process includes: 1) Perform translation signal processing Layer-wise wavelet decomposition is performed to obtain wavelet decomposition coefficients for each layer; these coefficients include wavelet detail coefficients. and wavelet approximation coefficients .in, This indicates the current level of the wavelet decomposition. This represents the maximum number of layers in the wavelet decomposition. ; Indicates the first The wavelet detail coefficients or the first Wavelet approximation coefficients.
[0018] 2) Since high-frequency noise is mainly concentrated in the first layer wavelet detail coefficients and the proportion of partial discharge pulses in the first layer is extremely low, this embodiment uses the Median Absolute Deviation (MAD) method to perform robust noise variance estimation on the first layer wavelet detail coefficients to obtain the global noise variance.
[0019] The formula for calculating the global noise variance is: ; In the formula, Indicates the first The global noise variance of a translation signal, This indicates the median operation. The first level of wavelet decomposition Wavelet detail coefficients, Indicates the first The statistical correction coefficient for a translation signal is typically taken as 0.6745. This statistical correction coefficient is the ratio of the absolute deviation of the median to the standard deviation of the standard normal distribution.
[0020] 3) Based on the wavelet decomposition coefficients of each layer and the global noise variance, the shape parameters and scale parameters corresponding to the wavelet decomposition coefficients of each layer are estimated by using a generalized Gaussian distribution model.
[0021] Specifically, for the first The first layer of wavelet decomposition wavelet detail coefficients Combined with the first Global noise variance of a translation signal The shape parameters of its generalized Gaussian distribution are estimated using the observed coefficients. and scale parameters .in, Shape parameters The estimation process is as follows: A matrix estimation method is used, by calculating the first-order absolute matrix of the wavelet detail coefficients. and second-order absolute matrix Solve , obtained the Shape parameters of layer wavelet decomposition .
[0022] Scale parameters The estimation process is as follows: combining the first Global noise variance of a translation signal From the total variance of the observed coefficients (i.e., the second-order absolute matrix) Separate the signal variance of the translation signal from the data. Then, by using the relationship between the variance and the scaling parameter of the generalized Gaussian distribution, the first... Scale parameters of layer wavelet decomposition .in, "Short for signal" indicates a translation signal. .
[0023] 4) The global noise variance and the shape and scale parameters corresponding to the wavelet decomposition coefficients of each layer are calculated by using the squared error loss function to construct the Bayesian risk of each layer, and the adaptive threshold of each layer is obtained by minimizing the Bayesian risk of each layer.
[0024] The formula for calculating the Bayesian adaptive threshold is: ; In the formula, Indicates the first Bayesian adaptive thresholding for layer wavelet decomposition Indicates the first The global noise variance of a translation signal, Indicates the first Shape parameters of layer wavelet decomposition, Indicates the first The scale parameter of layer wavelet decomposition, Indicates the first Scale parameters of layer wavelet decomposition Power of 1 Indicates the first The first layer of wavelet decomposition Wavelet detail coefficients, This indicates the current level of the wavelet decomposition. , This represents the maximum number of layers in the wavelet decomposition.
[0025] The exponent term of the generalized Gaussian distribution is the core denominator, comprehensively reflecting the first... Energy density and sparsity characteristics of layer wavelet decomposition coefficients.
[0026] 5) The soft thresholding function is used to calculate the wavelet detail coefficients of each layer and the Bayesian adaptive threshold to obtain the wavelet detail coefficients of each layer after thresholding.
[0027] Specifically, after obtaining the Bayesian adaptive threshold for each layer, a soft thresholding function is used to calculate the wavelet detail coefficients and Bayesian adaptive threshold for each layer. This effectively suppresses various types of clutter while preserving the original waveform characteristics of the partial discharge pulse to the greatest extent, thus obtaining the wavelet detail coefficients for each layer after thresholding.
[0028] The soft threshold function is: ; In the formula, Indicates the number of th values after threshold processing The first layer of wavelet decomposition Wavelet detail coefficients.
[0029] 6) Combining the wavelet approximation coefficients and the wavelet detail coefficients of each layer after thresholding, perform wavelet inverse reconstruction operation to obtain the denoised signal after a single cyclic translation.
[0030] 7) When the current translation operation count exceeds the preset maximum translation count, the adaptive noise reduction process is stopped, and the arithmetic average of all the noise-reduced signals is performed to obtain the second partial discharge signal.
[0031] ; In the formula, x2(t) represents the second partial discharge signal, and x1(t) represents the first partial discharge signal. This indicates the current number of translation operations, and N represents the preset maximum number of translation operations; WTT represents wavelet decomposition, and IWTT represents wavelet inverse reconstruction. denoted as the Bayesian adaptive threshold of the i-th level wavelet decomposition.
[0032] The pseudo-Gibbs effect is suppressed by cyclic translation invariant wavelet decomposition, and the parameters are automatically set based on the adaptive characteristics of Bayesian threshold without manual intervention. While effectively suppressing various clutter, the original waveform characteristics of the partial discharge pulse are preserved to the greatest extent, and finally the noise-reduced second partial discharge signal is obtained.
[0033] In step S120, a cascaded fourth-order zero-phase Butterworth high-pass filter is used to filter the second partial discharge signal, removing DC baseline drift and low-frequency narrowband interference signals to obtain the third partial discharge signal.
[0034] Specifically, after adaptive noise reduction processing of the first partial discharge signal, a small amount of DC component and low-frequency noise may still remain in the second partial discharge signal. Such interference will affect the accuracy of subsequent wavefront identification. Therefore, in this embodiment, after obtaining the second partial discharge signal, a cascaded fourth-order zero-phase Butterworth high-pass filter is used to further filter the second partial discharge signal, remove DC baseline drift and low-frequency narrowband interference signals, and obtain the third partial discharge signal.
[0035] Furthermore, the process of cascading a fourth-order zero-phase Butterworth high-pass filter includes: 1) Based on the sampling frequency of HFCT and the preset digital domain cutoff angular frequency, calculate the simulated high-pass cutoff frequency according to the frequency constraint relationship of zero-phase high-pass filtering.
[0036] The frequency constraint relationship is as follows: ; In the formula, Indicates the preset digital domain cutoff angular frequency. This indicates the simulated high-pass cutoff frequency. This indicates the sampling frequency of HFCT.
[0037] 2) Design a cascaded fourth-order Butterworth high-pass filter based on the analog high-pass cutoff frequency and sampling frequency, and construct the filter transfer function. .
[0038] 3) Based on filter transfer function The second partial discharge signal is subjected to zero-phase bidirectional filtering to remove DC baseline drift and low-frequency narrowband interference signals, thus obtaining the third partial discharge signal.
[0039] Specifically, based on filter transfer function The second partial discharge signal is subjected to forward filtering to obtain a forward filtered signal, and then the signal is processed based on the filter transfer function. The forward-filtered signal is subjected to reverse filtering to complete zero-phase bidirectional filtering, resulting in a third partial discharge signal that eliminates DC components and low-frequency noise.
[0040] By using bidirectional filtering, phase distortion of the signal is avoided during the filtering process, DC components and low-frequency noise interference are effectively eliminated, and the leading-edge details of the partial discharge pulse are fully preserved, resulting in a more pure partial discharge signal.
[0041] Step S130: The third partial discharge signal is sequentially subjected to signal amplitude enhancement and amplitude threshold clipping to obtain an effective pulse sequence.
[0042] Specifically, after obtaining the third partial discharge signal, the difference in amplitude between the effective partial discharge pulse and the residual small interference fluctuations is still not significant enough. Therefore, it is necessary to perform amplitude enhancement and threshold screening on the third partial discharge signal.
[0043] The signal amplitude is enhanced by performing point-by-point squaring on the third partial discharge signal, resulting in an amplitude-enhanced signal that amplifies the amplitude of the effective pulse. , This represents the third partial discharge signal; then, the signal with the largest amplitude is selected for enhancement. And combined with the preset ratio coefficient To obtain the adaptive decision threshold , The interference signals with amplitudes smaller than the adaptive decision threshold in the third partial discharge signal are removed to obtain the effective pulse sequence.
[0044] Effective pulse sequence for: ; In the formula, Indicates an amplitude-enhanced signal. Indicates the adaptive decision threshold. This indicates the effective pulse sequence.
[0045] Step S140: Using a sliding window energy comparison method, the wavefront start position of the effective pulse sequence is identified, and the pulse end position of the effective pulse sequence is determined by the attenuation trend of the pulse amplitude.
[0046] Specifically, the sliding window length L, sliding step size S, and sliding window index m are preset. After obtaining the valid pulse sequence, the sliding window index m is initialized to obtain the current window index, and the current signal energy judgment step is performed based on the current window index. If the change in the current signal energy is not greater than the energy change threshold, the current window index is updated based on the sliding step size, and the current signal energy judgment step is performed again based on the updated current window index. If the change in the current signal energy is greater than the energy change threshold, the starting sampling point position of the sliding window corresponding to the current window index is determined to be the wavefront start position of the valid pulse sequence.
[0047] The current signal energy determination step includes: Extract the signal data within the corresponding sliding window based on the current window index and use it as the current window data. Calculate the signal energy of the current window data as the current signal energy. Determine whether the mutation amount of the current signal energy is greater than the energy mutation threshold.
[0048] Furthermore, after determining the wavefront start position, the amplitude of the largest sampling point in the effective pulse sequence is determined as the pulse maximum amplitude, and the sampling point position corresponding to the pulse maximum amplitude is the maximum amplitude position; the pulse maximum amplitude is calculated according to the amplitude attenuation ratio to determine the pulse end amplitude; all sampling points in the effective pulse sequence are traversed from the wavefront start position backward, and the position of the first sampling point whose amplitude is less than the pulse end amplitude is determined as the pulse end position.
[0049] Step S150: Determine the waveform truncation start position based on the wavefront start position and the first extended sampling interval, and determine the waveform truncation end position based on the pulse end position and the second extended sampling interval.
[0050] Specifically, to avoid missing key features of the pulse's leading and trailing edges, a certain number of sampling points are reserved forward from the wavefront start position (i.e., the first extended sampling interval) to determine the starting position of the effective waveform truncation. Then, a certain number of sampling points are reserved backward from the pulse end position (i.e., the second extended sampling interval) to determine the ending position of the effective waveform truncation. This achieves automatic segmentation and truncation of the effective partial discharge waveform, resulting in a complete partial discharge event waveform, which facilitates subsequent defect location and fault analysis.
[0051] Furthermore, the formulas for determining the start and end positions of waveform truncation are as follows: ; In the formula, Indicates the starting position of the wavefront. Indicates the end position of the pulse. Indicates the first extended sampling interval. Indicates the second extended sampling interval. Indicates the starting position of waveform truncation. This indicates the end position of waveform truncation.
[0052] Step S160: Based on the waveform truncation start position and waveform truncation end position, the complete effective waveform of partial discharge is truncated to realize the automatic segmentation of the partial discharge signal, so as to segment out the complete partial discharge event. The segmented effective waveform of partial discharge is transmitted to the host computer for storage, which is used for subsequent cable insulation defect location and fault type identification. The host computer supports online parameter adjustment to adapt to the on-site engineering debugging needs.
[0053] The partial discharge signal detection and segmentation method based on wavelet Bayes threshold provided in this embodiment can be widely adapted to various types of high-voltage cable partial discharge signals. Whether it is partial discharge signals of different types of insulation defects simulated in the laboratory or partial discharge signals of cables operating in the complex electromagnetic environment of substations, it can stably achieve noise reduction, accurate wavefront positioning and effective waveform segmentation of partial discharge signals under low signal-to-noise ratio and strong interference conditions, and is suitable for a variety of engineering application scenarios.
[0054] The following describes the above-mentioned partial discharge signal detection and segmentation method based on wavelet Bayes threshold with a specific embodiment. However, it is worth noting that this specific embodiment is only for better illustrating the present invention and does not constitute an improper limitation of the present invention.
[0055] A Rogowski coil type HFCT high-frequency current sensor (bandwidth range 0.1MHz~30MHz) is selected and snapped onto the 110kV high-voltage cable grounding wire to ensure good contact with the grounding wire and avoid signal distortion.
[0056] Using a sampling precision of 16 bits, an operation frequency of no less than 1 GHz, and a sampling frequency of 125 MHz, a 1-second raw partial discharge signal is acquired, with a single-channel data length of [data missing]. The sampling points were used to collect signal amplitudes ranging from -5V to 5V. The signal contained Gaussian white noise, 50Hz power frequency harmonics, and 1MHz~10MHz narrowband electromagnetic interference, with a signal-to-noise ratio of 10dB~15dB. After acquisition, the original partial discharge signal was converted into a first partial discharge signal x1(t). Then, the first partial discharge signal was decomposed into a 6-level wavelet decomposition using a db4 wavelet basis to obtain the wavelet decomposition coefficients of each level. The wavelet decomposition coefficients of each level were then adaptively shrunk according to empirical Bayesian adaptive thresholds (0.02V for level 1, 0.03V for level 2, 0.05V for level 3, 0.08V for level 4, 0.12V for level 5, and 0.15V for level 6) to suppress noise components in each level. Next, the pseudo-Gibbs effect was suppressed by 8 cycles of cyclic translation wavelet decomposition and inverse reconstruction, preserving the characteristics of the partial discharge pulse leading edge to the greatest extent. Finally, the second partial discharge signal (with a signal-to-noise ratio improved to over 35dB) was obtained through wavelet inverse reconstruction.
[0057] After obtaining the second partial discharge signal, a cascaded fourth-order zero-phase Butterworth high-pass filter is used, based on a 125MHz sampling frequency and a preset digital domain cutoff angular frequency. =π / 1000, the simulated high-pass cutoff frequency f is obtained by calculation. c≈19.89kHz; then, a bidirectional filtering method is used to avoid phase distortion, effectively eliminating residual DC components and low-frequency noise in the signal, and completely preserving the details of the pulse leading edge to obtain the third partial discharge signal; next, point-by-point squaring is performed to amplify the amplitude difference between the effective pulse and the small interference fluctuations, and according to the scaling factor =0.15 to calculate the adaptive decision threshold, and remove interference signals with amplitudes smaller than the adaptive decision threshold in the discrete sampling sequence to obtain the effective pulse sequence; finally, a sliding window with a length of 128 sampling points and a sliding step size of 1 sampling point is used to calculate the current signal energy window by window. When the current signal energy is detected to change abruptly from 0 to greater than 0.01V², the position of the starting sampling point of the corresponding window is determined as the wavefront starting position of the effective pulse sequence, with a positioning accuracy of 1 sampling point (8ns).
[0058] After determining the wavefront start position, the amplitude decay trend of the effective pulse sequence is analyzed. When the pulse amplitude decays to 5% of the maximum amplitude, it is determined to be the pulse end amplitude, and the corresponding sampling point position is the pulse end position. Then, based on the wavefront start position and reserving 256 sampling points forward (corresponding to a time of 2.048μs), the waveform truncation start position of the effective waveform is determined. Based on the pulse end position and reserving 512 sampling points backward (corresponding to a time of 4.096μs), the waveform truncation end position of the effective waveform is determined. The complete effective partial discharge waveform is automatically truncated, and the partial discharge event segmentation is completed.
[0059] While this application provides method operation steps as shown in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-inventive labor. The order of steps listed in this embodiment is merely one possible execution order among many and does not represent the only execution order. In actual device or client product execution, the method can be executed in the order shown in this embodiment or the accompanying drawings, or in parallel (e.g., in a parallel processor or multi-threaded processing environment).
[0060] like Figure 2 As shown in the figure, this application also provides a partial discharge signal detection and segmentation system based on wavelet Bayesian thresholding. The system includes a wavelet noise reduction module, a zero-phase filtering module, a pulse enhancement and screening module, a wavefront identification module, a waveform truncation interval determination module, and a signal segmentation module, wherein: The wavelet denoising module is used to acquire the first partial discharge signal during the test process, and to perform adaptive denoising processing on the first partial discharge signal by combining translation-invariant wavelet transform with Bayesian adaptive thresholding to obtain the second partial discharge signal.
[0061] The zero-phase filtering module is used to filter the second partial discharge signal using a cascaded fourth-order zero-phase Butterworth high-pass filter, removing DC baseline drift and low-frequency narrowband interference signals to obtain the third partial discharge signal.
[0062] The pulse enhancement and filtering module is used to sequentially enhance the signal amplitude and prune the amplitude threshold of the third partial discharge signal to obtain an effective pulse sequence.
[0063] The wavefront identification module is used to identify the wavefront start position of the effective pulse sequence by using a sliding window energy comparison method, and to determine the pulse end position of the effective pulse sequence by the attenuation trend of the pulse amplitude.
[0064] The waveform truncation interval determination module is used to determine the waveform truncation start position based on the wavefront start position and the first extended sampling interval, and to determine the waveform truncation end position based on the pulse end position and the second extended sampling interval.
[0065] The signal segmentation module is used to extract the complete effective waveform of partial discharge based on the start and end positions of waveform truncation, thereby realizing the automatic segmentation of partial discharge signals.
[0066] Some modules in the apparatus described in this application can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, classes, etc., that perform a specific task or implement a specific abstract data type. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.
[0067] The apparatus or module described in the above embodiments can be implemented by a computer chip or physical entity, or by a product with a certain function. For ease of description, the above apparatus is described by dividing it into various modules according to their functions. When implementing the embodiments of this application, the functions of each module can be implemented in one or more software and / or hardware. Of course, a module that implements a certain function can also be implemented by combining multiple sub-modules or sub-units.
[0068] The methods, apparatus, or modules described in this application can be implemented in a computer-readable program code manner. The controller can be implemented in any suitable manner, such as a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of a memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code manner, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included within it for implementing various functions can also be considered as structures within the hardware component. Alternatively, the device used to implement various functions can be viewed as either a software module implementing the method or a structure within a hardware component.
[0069] This application also provides an apparatus, comprising: a processor and a memory for storing processor-executable instructions. When the processor executes the executable instructions, it implements the method described in this application.
[0070] This application also provides a non-volatile computer-readable storage medium storing a computer program or instructions thereon, which, when executed, enables the method described in this application embodiment to be implemented.
[0071] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist independently, or two or more modules can be integrated into one module.
[0072] The aforementioned storage media include, but are not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Cache, Hard Disk Drive (HDD), or Memory Card. The memory can be used to store computer program instructions.
[0073] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary hardware. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, or it can be embodied in the process of data migration. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.
[0074] The various embodiments described in this specification are presented in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. All or part of this application can be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.
[0075] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of this application.
Claims
1. A method for detecting and segmenting partial discharge signals based on wavelet Bayesian thresholding, characterized in that, include: The first partial discharge signal was acquired during the test, and the first partial discharge signal was adaptively denoised using a translation-invariant wavelet transform combined with a Bayesian adaptive threshold to obtain the second partial discharge signal. A cascaded fourth-order zero-phase Butterworth high-pass filter is used to filter the second partial discharge signal, removing DC baseline drift and low-frequency narrowband interference signals to obtain the third partial discharge signal; The third partial discharge signal is sequentially subjected to signal amplitude enhancement and amplitude threshold pruning to obtain an effective pulse sequence; The wavefront start position of the effective pulse sequence is identified by using a sliding window energy comparison method, and the pulse end position of the effective pulse sequence is determined by the attenuation trend of the pulse amplitude. The waveform truncation start position is determined based on the wavefront start position and the first extended sampling interval, and the waveform truncation end position is determined based on the pulse end position and the second extended sampling interval. Based on the waveform truncation start position and the waveform truncation end position, the complete effective waveform of partial discharge is truncated, thereby realizing the automatic segmentation of the partial discharge signal.
2. The partial discharge signal detection and segmentation method based on wavelet Bayes thresholding according to claim 1, characterized in that, The process of adaptively denoising the first partial discharge signal using translation-invariant wavelet transform combined with Bayesian adaptive thresholding to obtain the second partial discharge signal includes: The first partial discharge signal is translated according to the translation step size to obtain the translation signal; Record the current translation operation count. If the current translation operation count is less than or equal to the preset maximum translation count, repeat the adaptive noise reduction processing step. The adaptive noise reduction processing steps include: Perform translation signal Layered wavelet decomposition is performed to obtain wavelet decomposition coefficients for each layer; the wavelet decomposition coefficients include wavelet detail coefficients and wavelet approximation coefficients. The median absolute deviation method is used to perform robust noise variance estimation on the first-level wavelet detail coefficients to obtain the global noise variance. Based on the wavelet decomposition coefficients of each layer and the global noise variance, the shape parameters and scale parameters corresponding to the wavelet decomposition coefficients of each layer are estimated by a generalized Gaussian distribution model. The global noise variance and the shape and scale parameters corresponding to the wavelet decomposition coefficients of each layer are calculated using the squared error loss function to construct the Bayesian risk of each layer, and the adaptive threshold of each layer is obtained by minimizing the Bayesian risk of each layer. The wavelet detail coefficients of each layer are calculated using a soft thresholding function and a Bayesian adaptive threshold, resulting in the wavelet detail coefficients of each layer after thresholding. By combining the wavelet approximation coefficients and the wavelet detail coefficients of each layer after thresholding, wavelet inverse reconstruction operation is performed to obtain the denoised signal after a single cyclic translation. When the current translation operation count exceeds the preset maximum translation count, the adaptive noise reduction process is stopped, and all the noise-reduced signals are arithmetically averaged to obtain the second partial discharge signal.
3. The partial discharge signal detection and segmentation method based on wavelet Bayes thresholding according to claim 2, characterized in that, The formula for calculating the Bayesian adaptive threshold is: ； In the formula, Indicates the first Bayesian adaptive thresholding for layer wavelet decomposition Indicates the first The global noise variance of a translation signal, Indicates the first Shape parameters of layer wavelet decomposition, Indicates the first The scaling parameters of layer wavelet decomposition. Indicates the first Scale parameters of layer wavelet decomposition Power of 1 Indicates the first The first layer of wavelet decomposition Wavelet detail coefficients, This indicates the current level of the wavelet decomposition. This represents the maximum number of layers in the wavelet decomposition.
4. The partial discharge signal detection and segmentation method based on wavelet Bayes thresholding according to claim 1, characterized in that, The second partial discharge signal is filtered using a cascaded fourth-order zero-phase Butterworth high-pass filter to remove DC baseline drift and low-frequency narrowband interference signals, resulting in a third partial discharge signal, including: The analog high-pass cutoff frequency is obtained based on the sampling frequency and the preset digital domain cutoff angular frequency. Based on the simulated high-pass cutoff frequency and the sampling frequency, a cascaded fourth-order Butterworth high-pass filter is designed, and the filter transfer function is constructed. The second partial discharge signal is subjected to zero-phase bidirectional filtering based on the filter transfer function to remove DC baseline drift and low-frequency narrowband interference signals, thereby obtaining the third partial discharge signal.
5. The partial discharge signal detection and segmentation method based on wavelet Bayes thresholding according to claim 1, characterized in that, The step of sequentially enhancing and thresholding the third partial discharge signal to obtain an effective pulse sequence includes: The third partial discharge signal is subjected to point-by-point squaring operation to complete the signal amplitude enhancement processing, resulting in an amplitude-enhanced signal; The signal with the largest amplitude is selected and combined with a preset scaling factor to obtain an adaptive decision threshold; By removing interference signals with amplitudes smaller than the adaptive decision threshold from the third partial discharge signal, an effective pulse sequence is obtained.
6. The partial discharge signal detection and segmentation method based on wavelet Bayes thresholding according to claim 1, characterized in that, The method of identifying the wavefront start position of the effective pulse sequence using a sliding window energy comparison and determining the pulse end position of the effective pulse sequence by the attenuation trend of the pulse amplitude includes: Set the sliding window length, sliding step size, and sliding window index; The sliding window index is initialized to obtain the current window index, and the current signal energy determination step is performed based on the current window index; If the change in the current signal energy is not greater than the energy change threshold, then update the current window index based on the sliding step size, and execute the current signal energy judgment step again based on the updated current window index. If the sudden change in the current signal energy is greater than the energy sudden change threshold, then the starting sampling point position of the sliding window corresponding to the current window index is determined to be the wavefront starting position of the effective pulse sequence. The current signal energy determination step includes: The signal data within the corresponding sliding window is extracted based on the current window index and used as the current window data. The signal energy of the current window data is then calculated as the current signal energy. Determine whether the sudden change in the current signal energy is greater than the energy sudden change threshold.
7. The partial discharge signal detection and segmentation method based on wavelet Bayes thresholding according to claim 6, characterized in that, Determining the pulse end position of the effective pulse sequence by the attenuation trend of the pulse amplitude includes: The maximum amplitude of the sampling point in the effective pulse sequence is determined as the pulse maximum amplitude value, and the sampling point position corresponding to the pulse maximum amplitude value is the maximum amplitude value position. The maximum amplitude of the pulse is calculated based on the amplitude attenuation ratio to determine the pulse termination amplitude; Traverse all sampling points in the effective pulse sequence from the wavefront start position backward, and determine the position of the first sampling point whose amplitude is less than the pulse end amplitude as the pulse end position.
8. A partial discharge signal detection and segmentation system based on wavelet Bayes thresholding, characterized in that, The system includes a wavelet noise reduction module, a zero-phase filtering module, a pulse enhancement and screening module, a wavefront identification module, a waveform truncation interval determination module, and a signal segmentation module, wherein: The wavelet denoising module is used to acquire the first partial discharge signal during the test process, and to perform adaptive denoising processing on the first partial discharge signal by combining translation-invariant wavelet transform with Bayesian adaptive thresholding to obtain the second partial discharge signal. The zero-phase filtering module is used to filter the second partial discharge signal by employing a cascaded fourth-order zero-phase Butterworth high-pass filter to remove DC baseline drift and low-frequency narrowband interference signals, thereby obtaining the third partial discharge signal. The pulse enhancement and filtering module is used to sequentially enhance the signal amplitude and prune the amplitude threshold of the third partial discharge signal to obtain an effective pulse sequence; The wavefront identification module is used to identify the wavefront start position of the effective pulse sequence by using a sliding window energy comparison method, and to determine the pulse end position of the effective pulse sequence by the attenuation trend of the pulse amplitude. The waveform truncation interval determination module is used to determine the waveform truncation start position based on the wavefront start position and the first extended sampling interval, and to determine the waveform truncation end position based on the pulse end position and the second extended sampling interval; The signal segmentation module is used to extract the complete effective waveform of partial discharge based on the waveform extraction start position and the waveform extraction end position, thereby realizing the automatic segmentation of the partial discharge signal.
9. An electronic device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and executable by the processor. When the processor executes the computer program, it implements the partial discharge signal detection and segmentation method based on wavelet Bayes threshold as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the partial discharge signal detection and segmentation method based on wavelet Bayes threshold as described in any one of claims 1 to 7.