Data processing method based on ultrahigh frequency original signal and storage medium

Abstract

The invention discloses a data processing method based on an ultrahigh frequency original signal and a storage medium, and relates to the field of data processing, and the method comprises the steps: carrying out the adaptive gain adjustment collection of the ultrahigh frequency original signal, and enabling the signal amplitude to be stabilized in a preset effective interval through the dynamic adjustment of a gain parameter, acquiring initial acquisition data containing complete discharge characteristics; performing time-frequency domain combined noise reduction preprocessing on the initially acquired data, constructing an adaptive threshold function based on signal frequency domain distribution characteristics, and filtering background noise and non-discharge interference signal components; according to the method, the signal amplitude is dynamically stabilized through adaptive gain adjustment, it is ensured that initial collection data completely contains discharge characteristics, the problem of information loss caused by signal fluctuation is solved, the adaptive threshold is constructed based on frequency domain energy characteristics through time-frequency domain combined noise reduction, background noise and interference components are accurately filtered out, and the signal purity is improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a data processing method and storage medium based on ultra-high frequency raw signals. Background Technology

[0002] Ultra-high frequency (UHF) raw signals are a key basis for cable defect detection. Defects such as cable insulation aging and partial discharge generate UHF electromagnetic radiation. By acquiring raw signals, parameters such as amplitude, spectrum, and pulse characteristics can be analyzed to locate the defect and assess its severity.

[0003] Patent application No. 202510122886.5 discloses a method for denoising ultra-high frequency partial discharge signals, including: Step S1: acquiring the original ultra-high frequency signal; Step S2: performing wavelet decomposition on the acquired original ultra-high frequency signal; Step S3: extracting the upper and lower envelope spectra of the highest-level signal after decomposition using Hilbert transform to obtain the envelope signal, processing the envelope signal using the moving average smoothing method, determining the error band based on the mean of the segments in the smoothed envelope signal where no partial discharge occurred, and determining the specific start and end points of the partial discharge signal based on the intersection of the error band and the smoothed envelope signal; Step S4 The method involves extracting the partial discharge signal from the original UHF signal based on the specific start and end points of the partial discharge signal, performing singular value decomposition on the extracted partial discharge signal, and using clustering to divide the singular values. Step S5: Based on the clustering results, the signal of the non-partial discharge segment is set to zero to obtain the denoised UHF signal. This application aims to solve the problem that "existing denoising methods mainly rely on wavelet transform and singular value decomposition, which are difficult to completely remove various types of noise, and are difficult to accurately define the start and end times of the partial discharge signal in the UHF signal, which will affect the accuracy of subsequent data analysis and may lead to the loss of key information in the denoising process, affecting the accurate analysis of the partial discharge signal".

[0004] However, the application methods of UHF raw signals, which are the key basis for cable defect detection, have not been innovated for a long time.

[0005] To address this, a data processing method and storage medium based on UHF raw signals are proposed. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a data processing method and storage medium based on ultra-high frequency raw signals, which can effectively solve the problems of the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions;

[0008] This invention discloses a data processing method based on ultra-high frequency raw signals, comprising:

[0009] Adaptive gain adjustment is performed on the original UHF signal for acquisition. By dynamically adjusting the gain parameter, the signal amplitude is stabilized within a preset effective range, obtaining initial acquisition data containing complete discharge characteristics. Time-frequency domain joint noise reduction preprocessing is performed on the initial acquisition data. An adaptive threshold function is constructed based on the signal frequency domain distribution characteristics to filter out background noise and non-discharge interference signal components. Pulse characteristic mode decomposition (PMD) is performed on the denoised data. An adaptive mode decomposition algorithm is used to separate the characteristic mode components corresponding to different discharge types, extracting the time-domain pulse width and frequency band energy ratio characteristic parameters of each component. Feature parameter enhancement processing is performed based on the spatial correlation of multiple monitoring points. A spatial correlation matrix of the feature parameters of each monitoring point is constructed, and the decomposed characteristic mode components are weighted and fused to improve the feature signal-to-noise ratio. A dynamic mapping model between UHF features and cable defect types is established. Matching calculations are performed between real-time feature parameters and a dynamically updated defect feature benchmark library to output defect type identification results. Based on the enhanced characteristic mode components, the arrival time of the signal at each monitoring point is extracted. Combined with the signal propagation attenuation coefficient unique to this type of defect in the defect type identification results, the fault point is accurately located by calculating the spatiotemporal coordinate difference of multiple monitoring points.

[0010] Furthermore, the adaptive gain adjustment acquisition specifically includes:

[0011] The instantaneous amplitude A(t) of the original UHF signal is acquired in real time, and its deviation from the preset effective interval [Amin, Amax] is calculated. ;

[0012] Based on deviation value and the rate of change of signal amplitude Dynamically adjust gain parameters The formula is adjusted as follows:

[0013] ;

[0014] In the formula: It represents the first derivative of the instantaneous amplitude A(t) of the signal with respect to time t, that is, the instantaneous rate of change of the instantaneous amplitude A(t) of the signal with respect to time t; The gain parameter is from the previous time step; The gain adjustment sensitivity coefficient is within the range of (0, 1); This is the deviation normalization factor; The coefficient for suppressing the rate of change is within the range of (0, 1);

[0015] , Indicates the number of signal sampling points. This represents the sample size for historical deviation, taking the deviation values ​​from the most recent 100 to 500 sampling periods. Represents the bias value of the i-th historical sample, according to The computational logic is used for calculation;

[0016] Specifically, when the signal amplitude deviates significantly from the center of the effective range and needs to be quickly brought back into the range, The larger the value, the better when the signal amplitude is close to the center of the interval or when it is necessary to avoid over-adjustment that could cause oscillation. The smaller the value, the better when the signal amplitude fluctuates drastically and frequent oscillations of the gain parameter need to be avoided. The larger the value, the better when the signal amplitude is stable and the gain adjustment response speed needs to be improved. The smaller the value, the more stable A(t) is within [Amin, Amax]. Keep the current value.

[0017] Furthermore, the construction of the adaptive threshold function includes:

[0018] The time-frequency matrix S(f,t) of the signal is obtained through short-time Fourier transform, and the frequency domain energy concentration index is calculated. ,in, Describe frequency and time respectively;

[0019] based on And noise variance estimate Constructing an adaptive threshold function:

[0020] ;

[0021] In the formula: This represents the number of signal sampling points. This is a threshold adjustment factor, with a value range of (0, 1); The energy concentration sensitivity coefficient, ;

[0022] Among them, the stronger the noise interference and the more dispersed the effective signal frequency domain energy, the better. The larger the value, the higher the proportion of effective signal and the more concentrated the frequency domain energy. The smaller the value, the more significant the difference in energy distribution between different frequency components in the signal; that is, some frequency energy is highly concentrated while other frequency energy is dispersed. The larger the value, the more uniform the energy distribution at each frequency becomes. The smaller the value, the better for the amplitude in the time-frequency matrix. The components are set to zero to achieve noise reduction.

[0023] Furthermore, in the impulse characteristic mode decomposition, the adaptive mode decomposition algorithm includes:

[0024] Initialize the decomposition level n=1, and perform empirical mode decomposition on the denoised data x(t) to obtain the nth order mode component. ;

[0025] calculate Correlation coefficients with various categories of signals in the preset typical discharge signal template library:

[0026] , This represents the k-th type of discharge template signal;

[0027] When the termination condition is met and When the time comes, stop the decomposition;

[0028] in, Represents the nth modal component The maximum value of the correlation coefficient with all categories of template signals in the preset typical discharge signal template library. For the relevant threshold, This represents the energy percentage threshold. This represents the energy of a single-order modal component. This represents the total energy of x(t) after noise reduction. .

[0029] Furthermore, the feature parameter enhancement processing includes:

[0030] Construct an m×m spatial correlation matrix M, where m is the number of monitoring points and the matrix elements are... , Let i be the straight-line distance between monitoring points i and j; The characteristic length of the cable is taken as 1 / 5 to 1 / 3 of the total cable length in the monitoring area; The Pearson correlation coefficient is the characteristic parameter of the two points.

[0031] Based on matrix M, each modal component Perform weighted fusion ;

[0032] in, This represents the enhanced modal components.

[0033] Furthermore, the establishment of the dynamic mapping model includes:

[0034] Build a defect feature benchmark library ,in Let be the frequency band energy proportion feature vector of the k-th type of defect;

[0035] Real-time feature parameters The matching operation with the benchmark library is as follows:

[0036] ;

[0037] In the formula: Let k be the feature vector of the k-th type of defect; Total number of defect types; This represents the energy percentage of the k-th type of defect in the n-th frequency band. The percentage of energy of the signal to be identified in the nth frequency band; The weighted Mahalanobis distance indicates a higher similarity, with smaller distances representing higher similarity. is a diagonal weight matrix; T denotes the transpose operation of a vector;

[0038] Among them, the diagonal weight matrix The diagonal elements are the reciprocal of the variance of the energy proportion characteristic of the k-th type of defect in the i-th frequency band, and the off-diagonal elements are 0. Based on the above formula, the minimum output is obtained. The corresponding defect type is the identification result.

[0039] Furthermore, obtaining the signal propagation attenuation coefficient specific to the defect includes:

[0040] Based on the defect type identification result k, the attenuation coefficient is calculated using the following formula. :

[0041] ;

[0042] In the formula: The cable foundation attenuation coefficient; is the attenuation correction factor for the k-th type of defect; For ∈(0,1), the distance decay exponent is ∈(0,1). The straight-line distance from the defect point to the monitoring point; This is the attenuation saturation distance;

[0043] in, The unit is dB / m, which represents the inherent attenuation characteristics of the cable material for ultra-high frequency signals under defect-free conditions. It is determined by the cable type, material, and operating frequency, and is obtained through factory parameters or pre-calibration experiments. Dimensionless, with a value range of (0, 0.8], obtained through statistical analysis of previous experiments with similar defects, and follows the rule that the more severe the defect, the larger the value. Dimensionless, with a value range of (0, 1), for the installation of buried cables The value is greater than that of overhead cables. , This represents the critical distance at which the attenuation rate of an ultra-high frequency signal changes from rapid growth to slow convergence when it propagates in a cable.

[0044] Furthermore, the precise location of the fault point includes:

[0045] Let the coordinates of monitoring point i be... The signal arrival time is The coordinates of the fault point are (x, y), and they follow the following rules:

[0046] ;

[0047] In the formula: This refers to the speed at which ultra-high frequency signals propagate in a cable. To indicate the start time of a cable fault; This is the signal propagation attenuation coefficient; The propagation distance is the value on the left side of the equation;

[0048] in, ∈ , The maximum defect attenuation correction factor. The maximum propagation distance is determined by the severity of the defect type and the shorter the propagation distance. The larger the value, the milder the defect type and the farther the propagation distance. The smaller the value, the more the coordinates (x, y) of the fault point are determined by the least squares solution of the equation system of multiple monitoring points.

[0049] Furthermore, the dynamic mapping model also includes the following real-time optimization steps:

[0050] After each defect identification is completed, the identification confidence level is calculated. , Indicates the minimum matching distance;

[0051] when Update the benchmark library features as follows:

[0052] ;

[0053] In the formula: For updated benchmark library features; To update the features of the previous benchmark library; These are real-time feature parameters; Historical weighting coefficients;

[0054] in, This indicates a preset threshold. When it is necessary to prioritize ensuring the stability of the reference library to adapt to the long-term stable characteristics of cable defects, The larger the value, the better the model's dynamic adaptability to new defects or environmental changes needs to be enhanced. The smaller the value.

[0055] On the other hand, a storage medium storing a computer program, which, when executed by a processor, implements the execution steps of a data processing method based on ultra-high frequency raw signals.

[0056] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects:

[0057] This invention provides a data processing method based on ultra-high frequency raw signals. During the execution of this method, the amplitude of the dynamically stabilized signal is adjusted by adaptive gain to ensure that the initial acquired data completely contains the discharge characteristics, thus solving the problem of information loss caused by signal fluctuations. The time-frequency domain joint noise reduction constructs an adaptive threshold based on the frequency domain energy characteristics to accurately filter out background noise and interference components, thereby improving signal purity. The pulse feature mode decomposition effectively separates the feature modes of different discharge types through dual determination of correlation coefficient and energy ratio, thereby enhancing feature recognition.

[0058] Simultaneously, by combining the weighted fusion of feature parameters with the spatial correlation matrix of multiple monitoring points, the feature signal-to-noise ratio is significantly improved. The dynamic mapping model improves the accuracy and adaptability of defect type identification through weighted Mahalanobis distance matching and real-time updates of the benchmark library. Fault point location is achieved by combining the defect-specific attenuation coefficient with the spatiotemporal coordinate difference of multiple monitoring points. The overall process takes into account the stability of signal acquisition, the accuracy of feature extraction, and the reliability of defect identification and location, effectively supporting cable defect monitoring and diagnosis. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0060] Figure 1 This is a flowchart illustrating a data processing method based on ultra-high frequency raw signals. Detailed Implementation

[0061] 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0062] The present invention will be further described below with reference to embodiments.

[0063] Example:

[0064] This embodiment provides a data processing method based on ultra-high frequency raw signals, such as... Figure 1 As shown, it includes:

[0065] Adaptive gain adjustment is performed on the original UHF signal for acquisition. By dynamically adjusting the gain parameter, the signal amplitude is stabilized within the preset effective range, and initial acquisition data containing complete discharge characteristics is obtained.

[0066] Adaptive gain adjustment acquisition specifically includes:

[0067] The instantaneous amplitude A(t) of the original UHF signal is acquired in real time, and its deviation from the preset effective interval [Amin, Amax] is calculated. ;

[0068] Based on deviation value and the rate of change of signal amplitude Dynamically adjust gain parameters The formula is adjusted as follows:

[0069] ;

[0070] In the formula: It represents the first derivative of the instantaneous amplitude A(t) of the signal with respect to time t, that is, the instantaneous rate of change of the instantaneous amplitude A(t) of the signal with respect to time t; The gain parameter is from the previous time step; The gain adjustment sensitivity coefficient is within the range of (0, 1); This is the deviation normalization factor; The coefficient for suppressing the rate of change is within the range of (0, 1);

[0071] , Indicates the number of signal sampling points. This represents the sample size for historical deviation, taking the deviation values ​​from the most recent 100 to 500 sampling periods. Represents the bias value of the i-th historical sample, according to The computational logic is used for calculation;

[0072] Specifically, when the signal amplitude deviates significantly from the center of the effective range and needs to be quickly brought back into the range, The larger the value, the better when the signal amplitude is close to the center of the interval or when it is necessary to avoid over-adjustment that could cause oscillation. The smaller the value, the better when the signal amplitude fluctuates drastically and frequent oscillations of the gain parameter need to be avoided. The larger the value, the better when the signal amplitude is stable and the gain adjustment response speed needs to be improved. The smaller the value, the more stable A(t) is at. At that time, Keep the current value;

[0073] The above adjustment formula is based on the gain parameter of the previous moment, integrating two key pieces of information: signal amplitude deviation and instantaneous rate of change. The deviation reflects the degree to which the current signal deviates from the effective range, while the rate of change reflects the dynamic fluctuation trend of the signal amplitude. Through the synergistic effect of the gain adjustment sensitivity coefficient α and the rate of change suppression coefficient β, precise control of the gain parameter is achieved. When the signal needs to quickly return to the effective range, a larger value of α can accelerate the adjustment speed; when the signal fluctuates violently, a larger value of β can prevent frequent gain oscillations. Simultaneously, the deviation normalization factor is calculated based on the historical deviation sample size, making the quantification of the deviation value more closely match the actual signal fluctuation pattern, ultimately achieving stable acquisition of the signal amplitude within the effective range.

[0074] The initial acquired data is subjected to joint time-frequency domain noise reduction preprocessing. An adaptive threshold function is constructed based on the signal frequency domain distribution characteristics to filter out background noise and non-discharge interference signal components.

[0075] The construction of the adaptive threshold function includes:

[0076] The time-frequency matrix S(f,t) of the signal is obtained through short-time Fourier transform, and the frequency domain energy concentration index is calculated. ,in, Describe frequency and time respectively;

[0077] based on And noise variance estimate Constructing an adaptive threshold function:

[0078] ;

[0079] In the formula: This represents the number of signal sampling points. This is a threshold adjustment factor, with a value range of (0, 1); The energy concentration sensitivity coefficient, ;

[0080] The above formula constructs a threshold based on the time-frequency characteristics of the signal to filter out noise components in the time-frequency matrix. The formula uses the frequency domain energy concentration index and noise variance estimate of the time-frequency matrix as core inputs, and dynamically sets the threshold through a threshold adjustment factor and an energy concentration sensitivity coefficient. When noise interference is strong and the effective signal energy is dispersed, a larger value for the threshold adjustment factor can enhance the noise filtering effect; when the energy distribution of different frequency components in the signal differs significantly, a larger value for the energy concentration sensitivity coefficient can more accurately distinguish between the effective signal and noise. Setting the components in the time-frequency matrix with amplitudes below this threshold to zero can suppress background noise and non-discharge interference to the maximum extent while preserving the characteristics of the effective signal.

[0081] Among them, the stronger the noise interference and the more dispersed the effective signal frequency domain energy, the better. The larger the value, the higher the proportion of effective signal and the more concentrated the frequency domain energy. The smaller the value, the more significant the difference in energy distribution between different frequency components in the signal; that is, some frequency energy is highly concentrated while other frequency energy is dispersed. The larger the value, the more uniform the energy distribution at each frequency becomes. The smaller the value, the better for the amplitude in the time-frequency matrix. The components are set to zero to achieve noise reduction;

[0082] Pulse characteristic mode decomposition is performed on the noise-reduced data. An adaptive mode decomposition algorithm is used to separate the characteristic mode components corresponding to different discharge types, and the time-domain pulse width and frequency band energy ratio characteristic parameters of each component are extracted.

[0083] In impulse eigenmode decomposition, adaptive mode decomposition algorithms include:

[0084] Initialize the decomposition level n=1, and perform empirical mode decomposition on the denoised data x(t) to obtain the nth order mode component. ;

[0085] calculate Correlation coefficients with various categories of signals in the preset typical discharge signal template library:

[0086] , This represents the k-th type of discharge template signal;

[0087] When the termination condition is met and When the time comes, stop the decomposition;

[0088] in, Represents the nth modal component The maximum value of the correlation coefficient with all categories of template signals in the preset typical discharge signal template library. For the relevant threshold, This represents the energy percentage threshold. This represents the energy of a single-order modal component. This represents the total energy of x(t) after noise reduction. ;

[0089] In the above settings and formulas, the correlation coefficient calculation is used to measure the similarity between the nth-order mode component obtained from the empirical mode decomposition of the denoised data and the signals of each category in the preset typical discharge signal template library. The termination condition controls the decomposition process through two key indicators: first, the maximum correlation coefficient between the nth-order mode component and all categories of template signals in the template library must not be lower than the correlation threshold, ensuring that the decomposed component has clear discharge signal characteristics; second, the proportion of the energy of a single-order mode component to the total energy of the denoised data must not be lower than the energy proportion threshold, ensuring that the component contains sufficient signal energy. When both conditions are met simultaneously, the decomposition stops, thereby accurately separating the characteristic mode components corresponding to different discharge types, laying the foundation for subsequent feature extraction.

[0090] Feature parameter enhancement processing is performed based on the spatial correlation of multiple monitoring points. A spatial correlation matrix of feature parameters of each monitoring point is constructed, and the decomposed feature modal components are weighted and fused to improve the feature signal-to-noise ratio.

[0091] Feature parameter enhancement processing includes:

[0092] Construct an m×m spatial correlation matrix M, where m is the number of monitoring points and the matrix elements are... , Let i be the straight-line distance between monitoring points i and j; The characteristic length of the cable is taken as 1 / 5 to 1 / 3 of the total cable length in the monitoring area; The Pearson correlation coefficient is the characteristic parameter of the two points.

[0093] Based on matrix M, each modal component Perform weighted fusion ;

[0094] in, Indicates the enhanced modal components;

[0095] A dynamic mapping model between UHF characteristics and cable defect types is established. The model is matched with real-time feature parameters and a dynamically updated defect feature benchmark library to output defect type identification results.

[0096] The establishment of the dynamic mapping model includes:

[0097] Build a defect feature benchmark library ,in Let be the frequency band energy proportion feature vector of the k-th type of defect;

[0098] Real-time feature parameters The matching operation with the benchmark library is as follows:

[0099] ;

[0100] In the formula: Let k be the feature vector of the k-th type of defect; Total number of defect types; This represents the energy percentage of the k-th type of defect in the n-th frequency band. The percentage of energy of the signal to be identified in the nth frequency band; The weighted Mahalanobis distance indicates a higher similarity, with smaller distances representing higher similarity. is a diagonal weight matrix; T denotes the transpose operation of a vector;

[0101] The above formula introduces a diagonal weight matrix to assign differentiated weights to the energy proportion characteristics of different frequency bands. The diagonal elements of the weight matrix are the reciprocals of the variance of the corresponding frequency band's energy proportion characteristics. The smaller the variance (the more stable the feature), the greater the weight of the frequency band, making the matching process more focused on the contribution of stable features. The smaller the calculated weighted Mahalanobis distance, the higher the similarity between the real-time feature and the defect feature of that type. Finally, the defect type corresponding to the smallest distance is output as the identification result, achieving accurate matching of cable defect types.

[0102] Among them, the diagonal weight matrix The diagonal elements are the reciprocal of the variance of the energy proportion characteristic of the k-th type of defect in the i-th frequency band, and the off-diagonal elements are 0. Based on the above formula, the minimum output is obtained. The corresponding defect type is the identification result;

[0103] The calculation of the spatial correlation matrix elements comprehensively considers the straight-line distance between monitoring points and the Pearson correlation coefficient of the feature parameters. The closer the distance and the higher the correlation coefficient, the larger the matrix element value, thus quantifying the spatial correlation of feature parameters at different monitoring points. When weighted fusion of modal components based on this matrix, each monitoring point's modal component receives a corresponding weight according to its correlation strength with other monitoring points; the stronger the correlation, the higher the proportion of the feature in the fusion result. This fusion method effectively integrates complementary information from multiple monitoring points, weakens the influence of noise from a single monitoring point, thereby improving the signal-to-noise ratio of the feature modal components and enhancing the stability and reliability of the feature parameters.

[0104] The dynamic mapping model also includes the following real-time optimization steps:

[0105] After each defect identification is completed, the identification confidence level is calculated. , Indicates the minimum matching distance;

[0106] when Update the benchmark library features as follows:

[0107] ;

[0108] In the formula: For updated benchmark library features; To update the features of the previous benchmark library; These are real-time feature parameters; Historical weighting coefficients;

[0109] in, This indicates a preset threshold. When it is necessary to prioritize ensuring the stability of the reference library to adapt to the long-term stable characteristics of cable defects, The larger the value, the better the model's dynamic adaptability to new defects or environmental changes needs to be enhanced. The smaller the value;

[0110] The above logic and formulas only initiate updates when the identification confidence level is not lower than a preset threshold, ensuring the reliability of the benchmark library update. During the update process, the new benchmark library features are obtained by weighted fusion of historical features and real-time feature parameters before the update. The value of the historical weight coefficient can be flexibly adjusted: when the value is large, the stability of historical features is preserved first, which is suitable for long-term stable cable defect feature scenarios; when the value is small, the influence of real-time features is enhanced, improving the model's dynamic adaptability to new defects or environmental changes. Through this dynamic update mechanism, the benchmark library can continuously adapt to changes in actual working conditions and maintain the long-term accuracy of defect identification.

[0111] Based on the enhanced characteristic modal components, the arrival time of the signal at each monitoring point is extracted. Combined with the signal propagation attenuation coefficient unique to this type of defect in the defect type identification results, the fault point is accurately located by calculating the spatiotemporal coordinate difference of multiple monitoring points.

[0112] The acquisition of the signal propagation attenuation coefficient unique to defects includes:

[0113] Based on the defect type identification result k, the attenuation coefficient is calculated using the following formula. :

[0114] ;

[0115] In the formula: The cable foundation attenuation coefficient; is the attenuation correction factor for the k-th type of defect; For ∈(0,1), the distance decay exponent is ∈(0,1). The straight-line distance from the defect point to the monitoring point; This is the attenuation saturation distance;

[0116] in, The unit is dB / m, which represents the inherent attenuation characteristics of the cable material for ultra-high frequency signals under defect-free conditions. It is determined by the cable type, material, and operating frequency, and is obtained through factory parameters or pre-calibration experiments. Dimensionless, with a value range of (0, 0.8], obtained through statistical analysis of previous experiments with similar defects, and follows the rule that the more severe the defect, the larger the value. Dimensionless, with a value range of (0, 1), for the installation of buried cables The value is greater than that of overhead cables. , This represents the critical distance at which the attenuation rate of an ultra-high frequency signal changes from rapid growth to slow convergence when it propagates in a cable.

[0117] The above formula is used to calculate the propagation attenuation coefficient of UHF signals under specific defect types, comprehensively reflecting the influence of inherent cable characteristics, defect severity, and propagation distance on signal attenuation. The cable base attenuation coefficient reflects the inherent attenuation characteristics of the cable material in a defect-free state; the defect attenuation correction factor increases with the severity of the defect, quantifying the additional impact of defects on signal attenuation; the distance attenuation index is set according to the cable laying method (buried or overhead), with a larger value for buried cables to reflect a more complex propagation environment; the ratio of propagation distance to attenuation saturation distance describes the changing trend of the attenuation rate, which tends to level off after exceeding the saturation distance. Through the synergistic effect of these parameters, the formula can accurately characterize the signal propagation attenuation law under different scenarios, providing a reliable attenuation parameter basis for fault location.

[0118] Precise fault location includes:

[0119] Let the coordinates of monitoring point i be... The signal arrival time is The coordinates of the fault point are (x, y), and they follow the following rules:

[0120] ;

[0121] In the formula: This refers to the speed at which ultra-high frequency signals propagate in a cable. To indicate the start time of a cable fault; This is the signal propagation attenuation coefficient; The propagation distance is the value on the left side of the equation;

[0122] in, ∈ , The maximum defect attenuation correction factor. The maximum propagation distance is determined by the severity of the defect type and the shorter the propagation distance. The larger the value, the milder the defect type and the farther the propagation distance. The smaller the value, the more accurately the fault point coordinates (x, y) are determined by least squares solution of the multi-monitoring point equation system;

[0123] The above formula is constructed based on the relationship between the arrival times of signals from multiple monitoring points and their spatiotemporal coordinates, and is used to solve for the precise coordinates of the fault point. The left side of the equation represents the straight-line distance from the fault point to the monitoring point, while the right side expresses the signal propagation distance as a function of propagation speed, propagation time (the difference between the arrival time and the fault initiation time), and attenuation coefficient—the larger the attenuation coefficient (the more severe the defect and the closer the distance), the closer the calculated actual propagation distance is to the measured characteristics after signal attenuation. By solving the equation system constructed from multiple monitoring points using least squares, the measurement error of a single monitoring point can be effectively offset, and the coordinates (x, y) of the fault point can be determined, achieving precise location of the cable fault point.

[0124] A storage medium storing a computer program, which, when executed by a processor, implements the execution steps of a data processing method based on ultra-high frequency raw signals.

[0125] In this embodiment, the above method stabilizes signal acquisition quality through adaptive gain adjustment, effectively filters interference by combining time-frequency domain joint noise reduction, accurately extracts discharge features through mode decomposition, enhances feature signal-to-noise ratio through spatial correlation, achieves accurate defect type identification through dynamic mapping model, and further precisely locates faults by combining defect-specific attenuation coefficients with spatiotemporal calculations from multiple monitoring points. Real-time optimization of the benchmark library can also continuously improve adaptability, significantly improving the accuracy and efficiency of cable defect monitoring.

[0126] Based on the method in the above embodiments, an application example of this method is shown below:

[0127] When conducting partial discharge detection on a 110kV cable line, a power company successfully identified the defect type and accurately located the fault point using a data processing method based on ultra-high frequency raw signals. The specific implementation process is as follows:

[0128] I. Adaptive Gain Adjustment Acquisition

[0129] The testing equipment acquires the original UHF signal from the cable in real time, with a preset effective amplitude range of [missing information]. When the instantaneous amplitude of the signal at a certain moment is collected... When, calculate its deviation from the effective interval. (Based on the center of the interval, 0.5V). Meanwhile, the calculated instantaneous amplitude change rate dA(t) / dt is 0.3V / ms, and the gain parameter at the previous moment... .

[0130] The gain parameters are dynamically adjusted based on the deviation value and rate of change: Because the current signal amplitude deviates significantly from the center of the interval, a rapid correction is required; therefore, the gain adjustment sensitivity coefficient α is set to 0.7 (within the range of 0,1). Due to the drastic fluctuations in signal amplitude, to avoid frequent gain oscillations, the rate of change suppression coefficient β is set to 0.6 (within the range of 0,1). The deviation normalization factor ƣ is calculated to be 0.4 using historical deviation values ​​from the most recent 300 sampling periods. After substituting these values ​​into the adjustment formula, the new gain parameter G(t) is calculated to be 1.5. After adjustment, the signal amplitude corrects to 0.6V, stabilizing within the effective range.

[0131] II. Time-Frequency Domain Joint Noise Reduction Preprocessing

[0132] A short-time Fourier transform was performed on the initial acquired data to obtain the signal time-frequency matrix S(f,t). The calculated frequency domain energy concentration index is high, indicating that the effective signal energy is mainly concentrated in the 500MHz-800MHz frequency band; at the same time, the noise variance is estimated to be 0.02.

[0133] An adaptive threshold function was constructed based on energy concentration and noise variance: Since the current noise interference is weak and the effective signal energy is concentrated in the frequency domain, the threshold adjustment factor was set to 0.3 (within the range of 0, 1); the energy distribution at each frequency differed significantly, so the energy concentration sensitivity coefficient was set to 1.2 (>0). The threshold calculated using the function was 0.15. Components with amplitudes < 0.15 in the time-frequency matrix were set to zero, successfully filtering out background noise and non-discharge interference signals.

[0134] III. Pulse Characteristic Mode Decomposition

[0135] Empirical mode decomposition (EMD) is performed on the denoised data, with the initial decomposition level n=1. The first-order mode component is compared with a preset library of typical discharge signal templates (including templates for corona discharge, surface discharge, and levitation discharge), and the correlation coefficient is calculated. The correlation coefficient between the first-order component and the corona discharge template is 0.5, which is below the correlation threshold of 0.7, so decomposition continues.

[0136] When the decomposition reaches n=3, the correlation coefficient between the third-order mode component and the surface discharge template is 0.85 (≥0.7), and the energy of this component accounts for 92% of the total energy of the denoised data (≥ energy percentage threshold 0.9), satisfying the termination condition, and the decomposition is stopped. The time-domain pulse width of this component is extracted to be 80ns, and the frequency band energy percentage characteristic parameters are (500-600MHz accounts for 35%, 600-700MHz accounts for 45%, and 700-800MHz accounts for 20%).

[0137] IV. Feature Parameter Enhancement Processing

[0138] Three monitoring points were set up for this test, and a 3×3 spatial correlation matrix M was constructed. The calculated straight-line distance between monitoring points 1 and 2 is 50m, the distance between 1 and 2 and 3 is 80m, and the distance between 2 and 3 is 60m. The characteristic length L of the cable is taken as 1 / 4 of the total cable length in the monitoring area, i.e., 60m. The Pearson correlation coefficients of the characteristic parameters of each monitoring point were calculated, with the correlation coefficients between monitoring points 1 and 2 being 0.8, 1 and 3 being 0.7, and 2 and 3 being 0.85. These coefficients were then substituted into the matrix element calculation formula to construct the complete matrix.

[0139] Based on matrix M, the modal components of the three monitoring points are weighted and fused. After fusion, the signal-to-noise ratio of the feature modal components is increased from 20dB to 35dB, and the feature clarity is significantly improved.

[0140] V. Defect Identification in Dynamic Mapping Model

[0141] The constructed defect feature benchmark library contains frequency band energy proportion feature vectors for five types of defects. The enhanced real-time feature parameters (35% for 500-600MHz, 45% for 600-700MHz, and 20% for 700-800MHz) are then matched with the benchmark library.

[0142] When calculating the weighted Mahalanobis distance, the elements in the diagonal weight matrix W are taken as the reciprocal of the variance of the energy proportion of each frequency band corresponding to various defects. The calculated weighted Mahalanobis distance between the real-time features and the surface discharge feature vector is the smallest (value 0.12), therefore the output defect type identification result is surface discharge. The confidence level for this identification is c=0.92 (≥ preset threshold 0.8). The real-time feature parameters and the original surface discharge features in the benchmark library are weighted and updated using a historical weight coefficient of 0.7 (prioritizing stability). After the update, the benchmark library features are more closely aligned with the current detection environment.

[0143] VI. Precise location of the fault point

[0144] The coordinates of three monitoring points are (0,0), (100,0), and (50,80), and the arrival times of the signals at each monitoring point are extracted to be 1.2μs, 1.8μs, and 1.5μs, respectively. Since the defect type is surface discharge, its attenuation correction factor Kk is calculated to be 0.6 based on previous experiments; the cable foundation attenuation coefficient is 0.02dB / m, the distance attenuation index μ is taken as 0.7 (higher values ​​are used for buried cables), and the attenuation saturation distance do = 200m.

[0145] Based on the propagation speed of UHF signals in the cable (taken as 0.2 m / ns), a system of equations was constructed to determine the fault point coordinates (x, y). Considering the signal propagation attenuation coefficient (the attenuation coefficient is larger at close range due to severe surface discharge defects), the fault point coordinates were obtained as (48, 32) through least squares solution. On-site excavation verified that surface discharge traces were present in the cable insulation layer at this location, with a positioning error of less than 5 m.

[0146] In summary, the methods described in the above embodiments, during execution, dynamically stabilize the signal amplitude through adaptive gain adjustment to ensure that the initial acquired data completely contains discharge characteristics, thus solving the problem of information loss caused by signal fluctuations. Time-frequency domain joint noise reduction constructs an adaptive threshold based on frequency domain energy characteristics to accurately filter out background noise and interference components, improving signal purity. Pulse feature mode decomposition effectively separates the feature modes of different discharge types through dual determination of correlation coefficient and energy ratio, enhancing feature recognition. At the same time, it combines the weighted fusion of feature parameters with the spatial correlation matrix of multiple monitoring points to significantly improve the feature signal-to-noise ratio. The dynamic mapping model improves the accuracy and adaptability of defect type identification through weighted Mahalanobis distance matching and real-time updates of the benchmark library. Fault point location combines the defect-specific attenuation coefficient with the spatiotemporal coordinate difference of multiple monitoring points to achieve precise location. The overall process takes into account the stability of signal acquisition, the accuracy of feature extraction, and the reliability of defect identification and location, effectively supporting cable defect monitoring and diagnosis.

[0147] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention 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 of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A data processing method based on a very high frequency raw signal, characterized by, The method comprises the following steps: Adaptive gain adjustment collection is performed on the ultra-high frequency original signal, the gain parameter is dynamically adjusted, the signal amplitude is stabilized in the preset effective interval, the initial collection data containing complete discharge characteristics are obtained; Joint denoising preprocessing in time and frequency domains is performed on the initial collection data, an adaptive threshold function is constructed based on the signal frequency domain distribution characteristics, and background noise and non-discharge interference signal components are filtered out; Pulse characteristic mode decomposition is performed on the denoised data, an adaptive mode decomposition algorithm is used to separate characteristic mode components corresponding to different discharge types, and time-domain pulse width and frequency band energy proportion characteristic parameters of each component are extracted; Based on the spatial correlation of multiple monitoring points, characteristic parameter enhancement processing is performed, a spatial correlation matrix of characteristic parameters of each monitoring point is constructed, the decomposed characteristic mode components are weighted and fused to improve the signal-to-noise ratio of the characteristic parameters; A dynamic mapping model of ultra-high frequency characteristics and cable defect types is established, real-time characteristic parameters are matched with a dynamically updated defect characteristic reference library, and a defect type recognition result is output; Based on the enhanced characteristic mode components, the arrival time of the signal of each monitoring point is extracted, the signal propagation attenuation coefficient specific to the defect type in the defect type recognition result is combined, and the fault point is accurately positioned through multi-monitoring point space-time coordinate difference calculation.

2. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, The adaptive gain adjustment collection specifically comprises: Real-time acquisition of the instantaneous amplitude A(t) of the ultra-high frequency original signal, calculate the deviation value of the preset effective interval ;​ Based on the deviation value And signal amplitude change rate Dynamic adjustment of gain parameters The adjustment formula is: ； In the formula: represents the first derivative of the signal instantaneous amplitude A(t) with respect to time t, that is, the instantaneous change rate of the signal instantaneous amplitude A(t) with respect to time t; is the gain parameter at the previous moment; is a gain adjustment sensitivity coefficient, which is in the range of (0, 1); is a deviation normalization factor; is a change rate suppression coefficient, which is in the range of (0, 1); , represents the number of signal sampling points, represents the historical deviation sample quantity, and the deviation values in the last 100-500 sampling periods are taken, represents the deviation value of the i-th historical sample, which is calculated by the calculation logic of . When the signal amplitude deviates from the center of the effective interval and needs to be quickly adjusted back to the interval, The greater the value, the closer the signal amplitude is to the center of the interval or the more it needs to avoid excessive adjustment leading to oscillation, The smaller the value, the more intense the signal amplitude fluctuation, and the more it needs to avoid frequent gain parameter oscillation, The greater the value, the more stable the signal amplitude, and the more it needs to improve the gain adjustment response speed, The smaller the value, the more stable A(t) is within [Amin, Amax], Keep the current value.

3. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, The construction of the adaptive threshold function comprises: A time-frequency matrix S(f, t) of the signal is obtained by short-time Fourier transform, and a frequency energy concentration index is calculated wherein respectively denote frequency, time; based on and the noise variance estimate constructing an adaptive threshold function: ； In the formula: is the number of signal sampling points; is a threshold adjustment factor, and the value range is (0, 1); is an energy concentration sensitivity coefficient, ; The stronger the noise interference is and the more dispersed the effective signal frequency energy is The greater the value is, the higher the proportion of the effective signal is and the more concentrated the frequency energy is The smaller the value is, when the energy distribution of different frequency components in the signal is significantly different, that is, part of the frequency energy is highly concentrated and part of the frequency energy is dispersed, The greater the value is, when the energy distribution of each frequency tends to be uniform, The smaller the value is; and the components of the amplitude of the time-frequency matrix are subjected to zero processing to achieve noise reduction.

4. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, In the pulse characteristic mode decomposition, the adaptive mode decomposition algorithm comprises: Initialize the decomposition layer number n = 1, and perform empirical mode decomposition on the denoised data x(t) to obtain the nth order modal component ; Computing Correlation coefficients with each category signal in the preset typical discharge signal template library: , represents the kth discharge template signal; When the termination condition is satisfied and the decomposition is stopped. wherein, represents the nth order modal component a maximum value of correlation coefficients of all category template signals in a preset typical discharge signal template library, is a correlation threshold value, is an energy proportion threshold value, represents the energy of the single order modal component, represents the total energy of the x(t) after noise reduction, .

5. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, The characteristic parameter enhancement processing comprises: An m x m spatial correlation matrix M is constructed, where m is the number of monitoring points, and the matrix element , is the straight-line distance between monitoring points i and j; is the characteristic length of the cable, which is 1 / 5~1 / 3 of the total length of the cable in the monitoring area; is the Pearson correlation coefficient of the characteristic parameters of the two points; based on the matrix M to the modal components performing a weighted fusion ; wherein, denotes the enhanced modal component.

6. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, The dynamic mapping model establishment comprises: Constructing defect feature benchmark library wherein is the frequency band energy proportion feature vector of the kth type of defect. Real-time characteristic parameters The matching operation with the reference library is: ； In the formula: is the eigenvector of the kth type of defect; is the total number of defect types; is the energy proportion of the kth type of defect in the nth frequency band; is the energy proportion of the signal to be identified in the nth frequency band; is the weighted Mahalanobis distance, and the smaller the distance, the higher the similarity; is the diagonal weight matrix; T represents the transposition operation of the vector; wherein the diagonal weight matrix The inverse of the variance of the energy proportion of the kth type of defect in the ith frequency band is taken as the diagonal element, and the non-diagonal element is 0, and the minimum is output based on the above formula The corresponding defect type is the recognition result.

7. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, The acquisition of the signal propagation attenuation coefficient specific to the defect comprises: Based on the defect type recognition result k, the attenuation coefficient is calculated by the following formula : ； wherein: is the cable base attenuation coefficient; is the attenuation correction factor for the kth type of defect; is the distance attenuation exponent, where ∈ (0, 1); is the straight-line distance from the defect point to the monitoring point; is the attenuation saturation distance; wherein, Unit: dB / m, represents the inherent attenuation characteristics of the cable itself to the very high frequency signal in the defect-free state, determined by the cable model, material and working frequency, obtained through factory parameters or early calibration experiments, Dimensionless, the value range is (0, 0.8], obtained by early similar defect experiments, and it is subject to the more serious defects, the larger the value, Dimensionless, the value range is (0, 1), set the buried cable The value is greater than that of the overhead cable , represents the critical distance at which the attenuation rate of the very high frequency signal in the cable changes from rapid growth to slow convergence.

8. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, The accurate positioning of the fault point comprises: Let the coordinates of the monitoring point i be , the signal arrival time be , and the fault point coordinates be (x, y), which are subject to: ； wherein: is the speed of the ultra-high frequency signal propagating in the cable; is the time of the start of the fault in the cable; is the signal propagation attenuation coefficient; is the propagation distance, i.e. the value on the left side of the equation; wherein, ∈ , is a maximum defect attenuation correction factor, is a maximum propagation distance, the more serious the defect type, the closer the propagation distance the greater the value, the more minor the defect type, the farther the propagation distance the smaller the value, the fault point coordinates (x, y) are determined by the least squares solution of the multi-monitoring point equation group.

9. The data processing method based on the ultra-high frequency original signal according to claim 1, characterized in that, The dynamic mapping model further comprises the following real-time optimization steps: calculating a recognition confidence after each completed defect identification , denotes the minimum matching distance; When the reference library feature is updated: ； In the formula: is the updated reference library feature; is the pre-update reference library feature; is the real-time feature parameter; is the historical weight coefficient; wherein, represents a pre-set signal threshold value, when the reference library stability needs to be prioritized to adapt to long-term stable cable defect characteristics, the greater the value, when the dynamic adaptability of the model to new defects or environmental changes needs to be enhanced, the smaller the value.

10. A storage medium, characterized by The storage medium stores a computer program, and when the computer program is executed by the processor, the execution steps of the data processing method based on the ultra-high frequency original signal are realized.

Images