A method and apparatus for analyzing partial discharge in cables based on oscillating waves.

Abstract

This application discloses a method and apparatus for partial discharge analysis of cables based on oscillating waves. The method includes applying an oscillating wave to the cable under test and simultaneously acquiring low-frequency envelopes and high-frequency pulses; performing structural normalization based on cable parameters; encoding environmental and operating condition data into conditional vectors and performing conditional normalization; then performing domain adaptive alignment with ideal templates clustered by model, length, and connector / terminal type to obtain difference features; constructing a PRPD joint fingerprint, statistically analyzing pulse phase, amplitude, envelope attenuation rate, and / or equivalent Q; outputting defect location, type, and confidence interval based on difference features and the joint fingerprint; adaptively determining multi-level voltage up-voltage and early shutdown using information gain, and generating maintenance priorities and retest interval suggestions. Through stepwise normalization processing, domain adaptive alignment, and fusion of PRPD joint fingerprint features, the method enhances diagnostic stability and reproducibility, strengthens the detection capability of weak discharges and early defects, and improves adaptability to complex operating conditions and cross-triggering.

Description

Technical Field

[0001] This application relates to the field of power fault detection technology, and in particular to a method and apparatus for analyzing partial discharge in cables based on oscillating waves. Background Technology

[0002] As a crucial component of power transmission and distribution systems, the insulation condition of power cables directly impacts power supply safety and stability. With increasing service life and fluctuations in environmental and load conditions, insulation materials may experience aging, moisture absorption, and mechanical damage, leading to partial discharge under certain operating conditions. Partial discharge is characterized by its early, insidious, and random nature; if not identified and addressed promptly during operation and maintenance, it can escalate into insulation breakdown and power outages. Therefore, in-service monitoring and condition assessment for uninterrupted or short-term power outages have become a practical requirement in cable operation and maintenance.

[0003] The partial discharge detection method based on oscillating waves applies a short-time damped oscillating voltage to the cable under test, repeatedly exciting potential defects through the oscillation process, and acquiring the time-decreasing oscillation waveform and the superimposed partial discharge pulse signal in the sampling channel. Current engineering practices typically involve filtering and detecting the signal after excitation and acquisition to obtain the time, amplitude, and phase information of the partial discharge pulse, and then plotting a phase-related statistical spectrum for qualitative identification of the discharge type. When conditions permit, the location of suspected defects can be estimated by combining information such as arrival time, and suggestions for retesting or maintenance can be given based on existing empirical thresholds or criteria. This type of method has the advantages of relatively simple device structure and good field adaptability, and has been applied in multiple voltage levels and different laying scenarios.

[0004] Under complex field conditions, existing technologies still face several common problems. Differences in line length, laying method, and joint and terminal forms introduce deviations in test results. External factors such as ambient temperature, humidity, and load current also affect the stability of oscillation and discharge characteristics, making judgments based on fixed thresholds or empirical templates insufficient for cross-line applications. Distinguishing between on-site electromagnetic interference, background noise, and repetitive pulses poses challenges to pulse detection; the separability between early and weak discharges in statistical spectra is limited, resulting in insufficient analytical capabilities. The location process is easily affected by factors such as reflection characteristics, equivalent reflection point location, and propagation velocity estimation, leading to uncertainty in distance results under different scenarios. Insufficient data recording and management dimensions make it difficult to reproduce and trace results across batches and operating conditions. Therefore, there is an urgent need to propose a cable partial discharge analysis method based on oscillation waves that improves diagnostic stability, interpretability, and adaptability to complex operating conditions, and provides more operational guidance for defect location and treatment recommendations. Summary of the Invention

[0005] The embodiments of this application provide a cable partial discharge analysis method and apparatus based on oscillating waves to solve the problems of unstable defect diagnosis and location, insufficient interpretability of diagnostic results, and insufficient adaptability to complex working conditions.

[0006] To address the aforementioned technical problems, embodiments of this application disclose the following technical solutions: Firstly, a cable partial discharge analysis method is provided, comprising: exciting the cable under test with an oscillating wave and acquiring an original signal containing a low-frequency envelope and a high-frequency pulse of the oscillating wave; performing structural normalization on the original signal to obtain a standardized oscillating wave feature vector; performing conditional normalization on the standardized oscillating wave feature vector based on a conditional vector to obtain a conditionalized oscillating wave feature vector; performing domain adaptive alignment between the conditionalized oscillating wave feature vector and an ideal template, and performing difference comparison to obtain difference features; and constructing a PRPD connection based on the conditionalized oscillating wave feature vector. The fingerprint includes at least the statistics of pulse phase, amplitude, and low-frequency envelope attenuation rate and / or equivalent quality factor of the oscillating wave; based on the difference features and PRPD joint fingerprint, the defect location, defect type, and confidence interval are output; based on information gain, multi-level overpressure strategies and early shutdown criteria are determined, and maintenance suggestions are output; wherein, the structure normalization is based on the parameters of the cable under test, and the condition vector is determined by environmental and operating condition data collection and encoding; the ideal template includes a template feature vector established according to multiple feature clusters of the cable; the information gain characterizes the reduction in the width of the confidence interval caused by unit overpressure stress energy.

[0007] Furthermore, the oscillating wave excitation of the cable under test includes: charging the energy storage capacitor connected in parallel to the cable under test, and discharging the energy storage capacitor through a fast switch to form a damped oscillating wave; the acquisition of the original signal containing the low-frequency envelope and high-frequency pulse of the oscillating wave includes: performing phase reference alignment and synchronous sampling on the low-frequency envelope and high-frequency pulse of the oscillating wave, and employing at least one of bandpass filtering, noise suppression, anomaly gating and repetitive pulse clustering in the high-frequency pulse detection.

[0008] Furthermore, the structure normalization includes: performing deconvolution, amplitude-phase equalization, and / or group delay correction on the original signal based on a parameterized line transmission model and / or end-to-end transfer function; estimating the acquisition link gain and / or bandwidth drift through mileage scale points and / or injected calibration pulses to update the normalization parameters; the parameters of the cable under test include at least length, joints and / or terminations, and line laying structure.

[0009] Furthermore, the condition vector includes at least one of temperature, humidity, and load current; the condition normalization includes: generating network output scaling and bias coefficients through environmental and operating parameters, and applying a conditional affine transformation to the standardized oscillating wave feature vector; wherein the standardized oscillating wave feature vector includes the low-frequency envelope attenuation rate of the oscillating wave and / or the equivalent quality factor.

[0010] Furthermore, the domain adaptive alignment includes: selecting a template sub-library that matches the cable under test based on the clustering index of the ideal template; aligning the template feature vector of the template sub-library with the conditional oscillating wave feature vector using at least one of statistical distance minimization, correlation alignment, and adversarial alignment; the clustering index of the ideal template includes the cable model, length range, joint and / or termination type.

[0011] Furthermore, the construction of the PRPD joint fingerprint includes: mapping the high-frequency pulse to the phase axis with the zero-phase point and / or extreme point of the half-cycle of the oscillation wave as a reference; performing exponential and / or damped sine fitting on the low-frequency envelope of the oscillation wave to obtain the low-frequency envelope attenuation rate and / or equivalent quality factor of the oscillation wave; and calculating the joint fingerprint stability, which is obtained by statistical measures of cross-trigger phase stability and / or energy consistency.

[0012] Furthermore, the output defect location includes: in single-end reflection mode, the defect location result is corrected based on the reflection polarity and reflection amplitude ratio of the incident pulse and the first reflected pulse caused by the defect under the same trigger, wherein the reflection amplitude ratio is the ratio of the peak value of the reflected pulse envelope to the peak value of the incident pulse envelope; in dual-end arrival mode, the arrival time difference and propagation velocity acquired synchronously at both ends are used as observations, and the defect location and defect triggering time are jointly estimated using the least squares method and / or maximum likelihood estimation; the confidence interval is obtained through an error propagation model, wherein the input of the error propagation model includes at least the residuals of the joint fingerprint stability and domain adaptive alignment, and in dual-end arrival mode, also includes the variance of the arrival time difference.

[0013] Furthermore, the multi-level press strategy adaptively selects the level step size and maximum number of triggers under the premise of satisfying standard constraints; it terminates the press early when the information gain is lower than the threshold and / or the stability of the joint fingerprint is lower than the preset threshold; and it records the difference features and the PRPD joint fingerprint according to the level.

[0014] Furthermore, the maintenance recommendations include at least maintenance priority and retest interval; the maintenance priority is used to characterize the urgency of defect handling and is expressed in the form of a score; the score is obtained by weighting the width of the confidence interval, the stability of the joint fingerprint, and the intensity of the difference features; the retest interval is determined according to the maintenance priority score and a preset mapping relationship.

[0015] Secondly, a cable partial discharge analysis device based on oscillating waves is provided, comprising: an oscillating wave excitation and acquisition unit, used to apply an oscillating wave and simultaneously acquire a raw signal containing a low-frequency envelope and a high-frequency pulse of the oscillating wave, and also used to acquire parameters of the cable under test and environmental and operating condition data; a structure normalization unit, used to perform structure normalization on the raw signal according to the parameters of the cable under test to obtain a standardized oscillating wave feature vector; and a conditional normalization unit, used to perform conditional vector encoding based on environmental and operating condition data, and to perform conditional normalization on the standardized oscillating wave feature vector according to the conditional vector to obtain a conditionalized oscillating wave feature vector. The system includes: an alignment and comparison unit for adaptively aligning the conditional oscillating wave feature vector with the ideal template and performing a difference comparison to obtain difference features; a joint fingerprint construction unit for generating a PRPD joint fingerprint based on the conditional oscillating wave feature vector; a location and confidence assessment unit for outputting the defect location, defect type, and confidence interval based on the difference features and the PRPD joint fingerprint; an adaptive up-pressure control unit for determining multi-level up-pressure strategies and early shutdown criteria based on information gain; and a display and data management unit for recording and displaying the defect location, defect type, confidence interval, up-pressure level, and maintenance suggestions.

[0016] The above-mentioned technical solutions have at least the following advantages or beneficial effects: Before feature comparison, structural normalization and conditional normalization based on environment and operating conditions are performed sequentially, and the conditional oscillation wave feature vector is adaptively aligned with the ideal template in the domain, which can significantly reduce the distribution offset caused by differences in line length, laying method, connector / terminal model and acquisition link drift; In cross-line and cross-operating condition applications, feature consistency is improved, false alarms and false negatives are reduced, and diagnostic stability and reproducibility are enhanced; By constructing a phase-resolved partial discharge joint fingerprint, the pulse phase, amplitude and low-frequency envelope attenuation rate and equivalent quality factor of the oscillation wave are jointly characterized, and the low-frequency attenuation and high-frequency statistics complement each other, which enhances the detection capability of weak discharge and early defects.

[0017] The aforementioned further technical solutions also have the following beneficial effects: The stability of the joint fingerprint obtained across triggers reflects repeatability and noise effects, improving robustness under complex interference environments; in location and conclusion output, difference features and phase-resolved partial discharge joint fingerprints are used in tandem, with single-end reflection and double-end arrival modes selected according to field conditions; the uncertainty of joint fingerprint stability, domain alignment residuals, and arrival time difference is introduced into error propagation, directly outputting results containing defect location, defect type, and confidence interval, enhancing interpretability and supporting risk-based decision-making; information gain is used as a quantitative criterion, and the reduction in confidence interval width brought about by unit pressure stress energy guides multi-level pressure increases and early shutdown, balancing information efficiency and stress safety while meeting relevant test standard constraints; ideal templates are clustered and matched according to model, length range, joint, and / or terminal type, reducing systematic bias and enhancing adaptability to different cable families and diverse operating conditions; maintenance recommendations are output with a scoring-based maintenance priority and corresponding retest interval, transforming key quantities such as uncertainty, stability, and difference intensity into executable scheduling basis, improving maintenance efficiency and operability. Attached Figure Description

[0018] The specific embodiments of this application will be described in detail below with reference to the accompanying drawings, so that the technical solution and its beneficial effects of this application will be readily apparent.

[0019] Figure 1 The flowchart of the cable partial discharge analysis method based on oscillating waves provided in this application is shown.

[0020] Figure 2 A schematic diagram of the cable partial discharge analysis device 100 based on oscillation waves provided in this application.

[0021] Figure 3 A schematic diagram of the two-end arrival mode provided in this application.

[0022] Explanation of reference numerals in the attached figures: 100, Cable partial discharge analysis device; 101, Cable under test; 102, Terminal; 110, Oscillating wave excitation and acquisition unit; 120, Structure normalization unit; 130, Condition normalization unit; 140, Alignment and comparison unit; 150, Joint fingerprint construction unit; 160, Location and confidence assessment unit; 170, Adaptive up-voltage control unit; 180, Display and data management unit. Detailed Implementation

[0023] To make the objectives, technical solutions, and beneficial effects of this application clearer, the following detailed description, in conjunction with the accompanying drawings and specific embodiments, further illustrates this application. It should be understood that the specific embodiments described in this specification are merely for explaining this application and are not intended to limit it.

[0024] In this application, unless otherwise expressly specified and defined, the term "excitation" refers to the process of applying a voltage or current for testing to the cable under test to trigger a response; the term "oscillating wave / damped AC" refers to a short-duration, time-decreasing AC oscillating voltage formed at the cable end after the energy storage element and the cable under test form a circuit; the term "envelope" refers to a slowly varying curve of the absolute value or peak value of the oscillating waveform over time; the term "envelope decay rate" refers to the exponential decay rate of the low-frequency envelope of the oscillating wave; the term "high-frequency pulse" refers to a high-frequency transient signal generated by partial discharge superimposed on the oscillating wave; the terms "equivalent quality factor" and "equivalent Q" refer to dimensionless indicators characterizing the rate of oscillation decay, with a larger value generally indicating slower decay. The term "upper voltage" refers to the test voltage applied to the cable end under test (e.g., the initial peak value of the oscillating wave or the equivalent value specified by the standard), which is different from the charging voltage of the energy storage capacitor; the term "standard constraint" refers to the limitation of safety and compliance boundaries such as the amplitude, duration, and number of triggers of the test voltage according to relevant cable testing standards.

[0025] In the description of this application, the terms related to signal processing and feature construction are explained as follows: The term "normalization" in a non-conflicting context includes standardizing the amplitude / phase / scale / delay of the original signal or feature to reduce comparability bias caused by differences in channels, devices, or objects; the term "feature vector" refers to the representation of key quantities in a signal or spectrum after numericalization and vectorization; the term "difference feature" refers to the difference measure or residual information obtained by comparing the feature vector of the measured object with the feature vector of a reference or template; the term "phase-resolved partial discharge (PRPD)" refers to a method of phase mapping and statistical representation of the partial discharge pulse using the phase of the oscillating wave as a reference; the term "fingerprint" refers to a set of feature quantities or their combination that can uniquely or stably characterize the state of an object; the term "joint fingerprint stability" refers to the consistency measure of the joint fingerprint in terms of phase distribution and energy distribution under multiple triggering or multi-segment sampling conditions, used to reflect the repeatability and robustness of the results; the term "affine transformation" refers to the linear scaling and bias transformation applied to the feature, often used for condition-dependent scale correction. The term "cross-trigger phase stability" refers to a statistic that measures whether the partial discharge pulses recur in a concentrated manner on the phase axis of the oscillation wave under multiple excitations or triggers; the term "energy consistency" refers to a statistic that measures the stability of the energy or amplitude distribution of the cross-trigger pulses.

[0026] In the description of this application, the terms related to template matching and cross-domain consistency are explained as follows: the term "domain adaptive alignment" refers to the process of making the feature distribution from different operating conditions, lines, or devices more statistically consistent with the reference distribution without changing the semantics of the labels; the term "clustering" refers to grouping templates or samples according to model, length range, connector / terminal type, or other criteria; the term "statistical distance" refers to a statistical measure used to measure the similarity between two distributions or two groups of features, such as a measure based on distribution differences; the term "correlation alignment" refers to an alignment method that maximizes the correlation between features from different sources through projection or transformation; the term "adversarial alignment" refers to an alignment method that uses adversarial learning to reduce the differences in feature distributions from different sources without explicitly labeling domain information.

[0027] In this application, the terms related to field objects and calibration are explained as follows: the term "joint and / or termination" refers to the intermediate accessory used to connect two ends of a cable and the termination accessory used to connect the cable end to the equipment or to insulate it from the outside, respectively, and the two usually have different reflection characteristics; the term "mileage scale point" refers to a reference point or event at a known location along the line that can be used for time / distance calibration; the term "calibration pulse" refers to a reference pulse intentionally injected to calibrate the acquisition link or time scale; the term "acquisition link gain" refers to the amplification / attenuation coefficient of the acquisition system for the signal amplitude; the term "bandwidth drift" refers to the change in passband width or amplitude-frequency characteristics of the acquisition link frequency response caused by changes in time or environmental conditions.

[0028] In the description of this application, the terms related to location and uncertainty assessment are explained as follows: the term "single-end reflection mode" refers to a method of estimating the defect distance by completing excitation and acquisition at the same port and using the time relationship, polarity, and amplitude ratio of the incident and reflected signals; the term "double-end reflection mode / double-end arrival mode" refers to a method of simultaneously or separately acquiring signals at both ends of a cable and estimating the distance based on the arrival information at both ends; the term "time difference of arrival" refers to the difference in arrival times measured at both ends of the signal; the term "error propagation model" refers to a model that propagates error sources such as time difference of arrival uncertainty, uncertainty components corresponding to joint fingerprint stability, and alignment residuals to the variance of distance, type, or other output quantities through sensitivity or Jacobian approximation and uses them for interval estimation. These explanations of terms are intended to facilitate understanding of the content of this application. Unless the context otherwise requires, terms such as "including," "comprising," and "having" are non-exclusive terms; "and / or" indicates that the related objects can exist individually or in combination; and "first," "second," etc., are used only for distinction and do not limit the order, quantity, or priority.

[0029] Partial discharge is a common and representative insulation defect in power cables during operation. Due to defects such as microcracks, inclusions, moisture, or aging within or on the surface of the insulation material, electrical breakdown occurs in localized areas under operating stress or overvoltage, generating high-frequency currents or electromagnetic pulses at the nanosecond to microsecond level. This phenomenon usually does not immediately lead to overall insulation breakdown, but it repeatedly accumulates energy at the defect site, causing electrical densification, carbonization channel expansion, and increased dielectric loss, accompanied by localized temperature rise and insulation performance degradation. With the passage of operating time and the superposition of stress cycles, partial discharge activity may gradually intensify, eventually inducing overall insulation breakdown, resulting in power outages and equipment damage.

[0030] To this end, this application proposes an exemplary cable partial discharge analysis method and apparatus based on oscillating waves. By briefly discharging an energy storage capacitor to the cable under test, an equivalent oscillating circuit is formed, generating a short-term damped oscillating voltage on the measured resistor, so that the potential defect is repeatedly excited and responded to in a continuous half-cycle.

[0031] Figure 1 The flowchart of the cable partial discharge analysis method based on oscillating waves provided in this application is shown. Figure 1 As shown, the exemplary analysis method provided in this embodiment includes the following steps.

[0032] S1 involves exciting the tested cable 101 with an oscillating wave and acquiring the original signal containing the low-frequency envelope and high-frequency pulse of the oscillating wave; S2 involves structural normalization of the original signal to obtain a standardized oscillating wave feature vector; S3 involves conditionally normalizing the standardized oscillating wave feature vector from step S2 based on the conditional vector to obtain a conditionalized oscillating wave feature vector; S4 involves adaptively aligning the conditionalized oscillating wave feature vector from step S3 with the ideal template and performing a difference comparison to obtain difference features; S5 involves constructing a PRPD joint fingerprint based on the conditionalized oscillating wave feature vector from step S3, which includes at least the pulse phase, amplitude, and statistics of the low-frequency envelope attenuation rate and / or equivalent quality factor of the oscillating wave; S6 involves outputting the defect location, defect type, and confidence interval based on the difference features from step S4 and the PRPD joint fingerprint from step S5; S7 involves determining the multi-level up-voltage strategy and early shutdown criteria based on the information gain and outputting maintenance suggestions. The structural normalization in step S2 is based on the parameters of the cable under test 101. The condition vector in step S3 is determined by collecting and encoding environmental and operating condition data. The ideal template in step S4 includes template feature vectors established based on multiple feature clusters of the cable. The information gain in step S7 characterizes the reduction in the width of the confidence interval caused by unit compressive stress energy. In some specific embodiments, the following optional settings are also provided: for example, the parameters of the cable under test 101 include at least the length of the cable under test 101, the joints and / or terminals 102, and the line laying structure; the ideal template in step S4 is a reference template established by clustering according to cable type, length range, joint and / or terminal 102 type, and the template elements are template feature vectors matching the cable category; the maintenance suggestions output in step S7 include at least maintenance priority and retest interval.

[0033] Figure 2 A schematic diagram of the cable partial discharge analysis device 100 based on oscillation waves provided in this application. Figure 2As shown, the device includes: an oscillating wave excitation and acquisition unit 110, which applies an oscillating wave to the cable under test 101 and simultaneously acquires the raw signal containing the low-frequency envelope and high-frequency pulse of the oscillating wave, and also acquires the parameters of the cable under test 101 and environmental and operating condition data; a structure normalization unit 120, which performs structure normalization on the raw signal output by the oscillating wave excitation and acquisition unit 110 according to the parameters of the cable under test 101 to obtain a standardized oscillating wave feature vector; a conditional normalization unit 130, which performs conditional vector encoding on the environmental and operating condition data acquired by the oscillating wave excitation and acquisition unit 110, and performs conditional normalization on the standardized oscillating wave feature vector output by the structure normalization unit 120 to obtain a conditionalized oscillating wave feature vector; and an alignment and comparison unit 140, which performs alignment and comparison on the cable under test 101. The system comprises: a conditional normalization unit 130, a joint fingerprinting unit 150, and a display and data management unit 180. The latter is used to perform domain adaptive alignment of the conditional oscillation feature vector output by the conditional normalization unit 130 with the ideal template and to perform difference comparison to obtain difference features; a joint fingerprinting unit 150, which generates a PRPD joint fingerprint based on the conditional oscillation feature vector output by the conditional normalization unit 130; a location and confidence assessment unit 160, which outputs the defect location, defect type, and confidence interval based on the difference features output by the alignment and comparison unit 140 and the PRPD joint fingerprint output by the joint fingerprinting unit 150; an adaptive up-pressure control unit 170, which determines the multi-level up-pressure strategy and early shutdown criteria based on information gain; and an adaptive up-pressure control unit 170, which records and displays the defect location, defect type, confidence interval, up-pressure level, and maintenance suggestions.

[0034] The following is combined Figure 1 and Figure 2 The exemplary cable partial discharge analysis method and apparatus based on oscillating waves described above will be described in detail.

[0035] Figure 3 This is a schematic diagram of the two-end arrival mode provided in this application. Figure 3 As shown, in this embodiment, the oscillating wave excitation and acquisition unit 110 includes an energy storage capacitor C and a fast switch K. This unit performs oscillating wave excitation and acquisition on the two joints and / or terminals 102 of the cable under test 101 to form a defect detection in a dual-end arrival mode, thereby distinguishing it from... Figure 2The single-ended reflection mode is shown. Based on the above exemplary embodiment, step S1 further includes the following: charging the energy storage capacitor C connected in parallel to the cable under test 101, discharging the energy storage capacitor C through a fast switch K to form a damped oscillation wave in the cable under test 101, and simultaneously acquiring the original signal containing the low-frequency envelope and high-frequency pulse of the oscillation wave. In this embodiment, the oscillation wave excitation and acquisition unit 110 is equipped with a low-frequency envelope and high-frequency pulse synchronous sampling channel and a phase reference channel, performs phase reference alignment and synchronous sampling on the low-frequency envelope and high-frequency pulse of the oscillation wave, and sets a signal processing module with at least one of the functions of bandpass filtering, noise suppression, anomaly gating and repetitive pulse clustering in the sampling channel of the high-frequency pulse, and applies corresponding signal processing to the high-frequency pulse in the high-frequency pulse detection. By setting up low-frequency envelope and high-frequency pulse synchronous sampling channels and phase reference channels during the oscillating wave excitation and acquisition stages, and introducing bandpass filtering, noise suppression, anomaly gating, and repetitive pulse clustering in the high-frequency channel, the signal-to-noise ratio and detectability of partial discharge pulses can be significantly improved, and unrelated interferences such as radio waves and switching actions can be suppressed. Phase reference alignment is conducive to the stable reproduction of subsequent phase mapping and reduces statistical diffusion caused by trigger jitter. The short-time damped oscillation formed by the parallel discharge of the energy storage capacitor C can repeatedly excite potential defects within multiple half-cycles, improve the probability of weak discharge detection, and avoid the additional stress exposure caused by long-term power frequency pressure, thus taking into account both field feasibility and the safety boundary of the object under test.

[0036] In the above exemplary analysis method, step S2 can also be implemented by the following: The structure normalization unit 120 establishes a parameterized line transmission model and / or end-to-end transfer function based on the length of the cable 101 under test, the line laying method, the type of joint and / or terminal 102, and the parameters of the acquisition link. This is to perform deconvolution, amplitude and phase equalization, and / or group delay correction on the oscillating wave excitation and the original signal output by the acquisition unit 110. Then, the acquisition link gain and / or bandwidth drift are estimated through mileage scale points and / or injected calibration pulses to update the normalization parameters. For example, in the equivalent transmission model... Response to data acquisition link Below, the original signal Performing deconvolution and amplitude-phase equalization yields the following result in the frequency domain: ;in, To obtain the signal spectrum, A regularization term is used to suppress high-frequency noise amplification; subsequently, group delay correction is performed to eliminate peak position shift caused by dispersion. To eliminate field equipment drift, the structure normalization unit 120 can estimate the acquisition link gain and bandwidth drift based on mileage scale points (such as interface reflections at known locations) and / or calibration pulses. The gain and passband are updated online; when a complete model is unavailable, calibration curves or empirical compensation factors can be used to achieve equivalent normalization. By deconvolution, amplitude and phase equalization, and group delay correction, combined with online compensation for gain and bandwidth drift of the acquisition link using mileage scale points / calibration pulses, the system errors introduced by equipment differences and time-varying drift can be significantly reduced, providing a fundamental support for the comparability of results across lines and batches.

[0037] In the above exemplary analysis method, step S3 specifically includes the following: The conditional normalization unit 130 encodes environmental and operating conditions such as temperature, humidity, and load current into a conditional vector z, and generates scaling factors through the parameter generation network. With bias coefficient The standardized oscillating wave eigenvector obtained in step S2 Performing an affine transformation yields: ;in, This represents element-wise multiplication by channel. The above standardized feature vectors... It should at least include the low-frequency envelope attenuation rate of the oscillating wave and / or the equivalent quality factor; under network-free conditions, the above transformation can also be achieved by looking up tables or linear regression coefficients to reduce the influence of temperature, humidity and load fluctuations on the feature scale and bias. Conditional vector-driven affine normalization further ensures that the above features maintain statistical consistency under multiple operating conditions, reducing misjudgments caused by environmental and operating condition deviations.

[0038] In the above exemplary analysis method, step S4 specifically includes the following: The alignment and comparison unit 140 first clusters and builds a library of ideal templates according to cable type, length range, joint and / or terminal type. Then, based on the parameters of the cable 101 under test, it selects a template sub-library through indexing. Subsequently, it compares the template feature vector of the selected template sub-library with... Adaptive domain alignment can be performed between the two domains, employing at least one of the following methods: statistical distance minimization (e.g., distribution difference measurement), correlation alignment (maximizing the correlation between the two domains), and / or adversarial alignment (making the two domains difficult to distinguish using a discriminator), to make the distributions of the two domains more statistically consistent. After alignment, the difference features are calculated, which can be defined as residual vectors. ,in To match template vectors and their derived norms, relevance, etc., for subsequent defect localization and output, clustered library selection and feature distribution level alignment are used to converge the statistical differences between "object and template" from the source. Then, the "residual inconsistency" is quantified by the difference features, which not only improves the accuracy of template matching, but also provides a measurable source of residuals for subsequent uncertainty modeling.

[0039] In the above exemplary analysis method, step S5 specifically includes the following: The joint fingerprint construction unit 150 uses the zero-phase point and / or extreme point of the oscillation half-cycle as a reference, maps each high-frequency pulse to the phase axis, and statistically analyzes the phase-amplitude distribution to obtain a PRPD (phase-resolved partial discharge) spectrum; and performs exponential and / or damped sine fitting on the low-frequency envelope to obtain the low-frequency envelope attenuation rate and / or equivalent quality factor. Across multiple triggers, the joint fingerprint stability is calculated, which can be obtained by weighting phase concentration (e.g., circular statistics) and energy consistency (e.g., normalized complement of the coefficient of variation), used to quantify repeatability and noise impact. Finally, the above statistics are aggregated into a numerical vector representation of the PRPD joint fingerprint. This process jointly characterizes the phase-amplitude statistics and envelope attenuation rate, and uses cross-trigger stability to measure repeatability and noise impact, significantly improving the separability and detection probability of early or weak discharges, and maintaining higher robustness in complex interference environments.

[0040] To facilitate the implementation of step S5 above, the joint fingerprint stability is defined. A weighted combination of phase concentration and energy consistency: ;in, The length of the composite vector in circular statistics, i.e., the phase concentration. The coefficient of variation of the pulse energy. and These are weighting coefficients. The following is a calculation example defined in this embodiment. Several discharge pulse phases under triggering conditions are selected. With pulse energy Statistical results: Phase concentration Energy variation coefficient Take weight It is 0.6. If the value is 0.4, then: Therefore, it can be seen that the phase distribution is relatively concentrated and the energy consistency is good, forming a joint fingerprint with a stability of 0.872. This stability can be directly used for uncertainty mapping and pressure decision-making.

[0041] In the above exemplary analysis method, step S6 specifically includes the following: In single-end reflection mode, the positioning and confidence assessment unit 160 corrects the ranging based on the arrival times, reflection polarity, and reflection amplitude ratio of the incident and reflected pulses triggered in the same instance. The reflection amplitude ratio is the ratio of the peak value of the reflected pulse envelope to the incident pulse including the peak value. In dual-end arrival mode, using the arrival time difference and propagation velocity obtained synchronously from both ends as observations, the least squares method and / or maximum likelihood estimation are used to jointly estimate the defect location and triggering time. To provide a confidence interval, an error propagation model is constructed: the arrival time difference variance, the uncertainty component obtained by monotonically mapping the joint fingerprint stability, and the uncertainty component obtained by monotonically mapping the domain adaptive alignment residual are used as error sources. These are propagated to the distance variance through a sensitivity coefficient, and the interval result is output using a statistical interval estimation method.

[0042] Combination Figure 3 The diagram shown illustrates the dual-arrival mode. An example is given below to illustrate the above process. Assume that when locating a defect in the dual-arrival mode, the pulse propagation speed is... The time difference can be measured Then the estimated distance to the defect point is: Consider the following three sources of uncertainty: time of arrival variance. The uncertainty is obtained by combining the joint fingerprint stability component and the domain alignment residual component, where sampling and time calibration are used to obtain the overall uncertainty. The joint fingerprint stability component is composed of Mapped to distance standard deviation ,Pick ,get Alignment residual measure , mapped to ,Pick ,get Treating the above three as independent error sources, the distance variance obtained by first-order Taylor expansion is: Substitute the values Then there is Under the normal approximation, the half-width of the 95% confidence interval is approximately 8.17m, therefore... Based on the above calculation process, when using the single-end reflection mode, only the arrival time difference term needs to be removed or the reflection time difference variance needs to be used instead.

[0043] In the above exemplary analysis method, step S7 specifically includes the following: The adaptive uppressure control unit 170 calculates the information gain: ;in This represents the width of the confidence interval after applying pressure at the k-th level. The unit pressure stress energy of the k-th level is characterized. Under the premise of meeting standard constraints, such as test voltage amplitude, duration, and number of triggers, the information gain IG and target CI are used as criteria to adaptively select the next level, adjust the level step size, or trigger early shutdown. When the diagnosis converges, for example, CI reaches the target, or the CI change slows down, or IG is lower than the preset threshold, the pressure is terminated, and maintenance suggestions such as maintenance priority and retest interval are output. Among them, the retest interval can be determined by converting the priority score into a time interval through a preset monotonically decreasing mapping.

[0044] In the specific step S7 described above, the maintenance priority can also be displayed in the form of a score, which is obtained by weighting the width of the confidence interval output from step S6, the stability of the joint fingerprint, and the intensity of the differential features. For example, the score S of one scoring model is modeled as follows: ;in, and The sum of is 1. This indicates linear normalization to [0, 1] based on the upper and lower limits of the engineering parameters; The strength of the difference feature. Taking the example of a 95% confidence interval half-width of approximately 8.17m as the reference example for this model, the CI width is taken as 16.3m, and the upper and lower limits for engineering are taken as 5m-60m, resulting in... Approximately 0.205, taken as the stability of the joint fingerprint. Normalized value of difference feature intensity (Obtained from in-database statistics or upper limit calibration), weight and If we take the values ​​0.4, 0.6, and 0.6 respectively, then we have The value is approximately 31. According to an exemplary score-retest interval mapping table, in the above example, a retest interval of 180 days is required, meaning routine retesting is sufficient. The exemplary mapping table is as follows: Table 1:

[0045]

[0046] Table 1 shows the exemplary mapping relationship between the scoring and retest intervals mentioned above. In other embodiments, if continuous mapping is required, the parameter configuration can be retained according to the preset mapping relationship, such as a monotonically decreasing function, to adapt to different operation and maintenance strategies.

[0047] In some embodiments, the display and data management unit 180 can record key data such as difference features, joint fingerprints, condition vectors, template clustering indexes and calibration parameters according to the pressure level and / or trigger dimension, which facilitates cross-operating condition retesting, comparative analysis and result traceability.

[0048] In summary, the specific implementation details of each embodiment and step provided in this application all revolve around an ordered chain of "structural normalization—conditional normalization—domain adaptive alignment—joint fingerprint—location and confidence assessment—information gain-driven upsampling": the two-stage normalization and domain alignment at the front end significantly reduce the distribution offset across lines and operating conditions, improving the consistency and transferability of features; the joint fingerprint, composed of complementary "phase-amplitude" statistics and "low-frequency attenuation / quality factor," strengthens the separability of early and weak discharges, and characterizes repeatability and noise impact with stability; the location stage incorporates the stability of the joint fingerprint, alignment residuals, and arrival time difference into error propagation, achieving result output with confidence intervals, improving interpretability and the executability of operation and maintenance decisions; information gain quantifies the convergence of the confidence interval brought by each unit of stress, reducing unnecessary electrical stress exposure while meeting standard constraints, balancing test efficiency and equipment safety boundaries, thereby achieving stable, reproducible, and quantifiable identification and location of defective insulation under complex field conditions.

[0049] It should be noted that the specific embodiments described above are only used to illustrate the technical solutions of this application, and are not intended to limit this application. For those skilled in the art, various modifications and changes can be made without departing from the spirit and scope of the technical solutions of this application, and all of these should fall within the protection scope of this application.

Claims

1. An oscillating wave-based cable partial discharge analysis method, characterized by, The method comprises: exciting the measured cable with an oscillatory wave and collecting a raw signal containing an oscillatory wave low-frequency envelope and a high-frequency pulse; performing structural normalization on the raw signal to obtain a normalized oscillatory wave feature vector; performing conditional normalization on the normalized oscillatory wave feature vector based on a conditional vector to obtain a conditional oscillatory wave feature vector; performing domain adaptive alignment on the conditional oscillatory wave feature vector and an ideal template, and performing difference comparison to obtain a difference feature; constructing a PRPD joint fingerprint according to the conditional oscillatory wave feature vector, including at least the statistical quantity of the pulse phase, amplitude, and oscillatory wave low-frequency envelope decay rate and / or equivalent quality factor; outputting a defect position, defect type, and confidence interval based on the difference feature and the PRPD joint fingerprint; determining a multi-grade overvoltage strategy and early shutdown criterion according to the information gain, and outputting an operation and maintenance suggestion; wherein the structural normalization is based on parameters of the measured cable, and the conditional vector is determined by environmental and working condition data collection and coding; the structural normalization includes: performing deconvolution, amplitude and phase equalization, and / or group delay correction on the raw signal based on a parameterized line transmission model and / or an end-to-end transfer function; estimating the acquisition link gain and / or bandwidth drift by the milepost point and / or injecting a calibration pulse to update the normalization parameters; the conditional normalization includes: generating network output scaling coefficients and bias coefficients through environmental and working condition parameters, and applying conditional affine transformation to the normalized oscillatory wave feature vector; The constructing PRPD joint fingerprint comprises: mapping the high-frequency pulse to the phase axis with the zero-phase point and / or the extreme point of the half cycle of the oscillation wave as the reference; performing exponential and / or damped sinusoidal fitting on the low-frequency envelope of the oscillation wave to obtain the low-frequency envelope decay rate and / or the equivalent quality factor of the oscillation wave, and calculating the joint fingerprint stability by weighting the circular statistics (Q ) representing the phase concentration and the coefficient of variation normalized complementary quantity (C ) representing the energy consistency. The calculation formula is as follows: ​ in, and These are circular statistics ( ) and the normalized complement of the coefficient of variation ( The weighting coefficients; the ideal template includes a template feature vector established according to the clustering of multiple features of the cable; the information gain represents the width reduction of the confidence interval caused by unit overvoltage stress energy.

2. The oscillating wave based cable partial discharge analysis method of claim 1, wherein, The excitation of the measured cable with an oscillatory wave includes: charging an energy storage capacitor connected in parallel to the measured cable, and discharging the energy storage capacitor through a fast switch to form a decaying oscillatory wave; the collection of the raw signal containing the oscillatory wave low-frequency envelope and the high-frequency pulse includes: phase reference alignment and synchronous sampling of the oscillatory wave low-frequency envelope and the high-frequency pulse, and at least one of band-pass filtering, noise suppression, abnormal gating, and repeated pulse clustering is used in high-frequency pulse detection.

3. The oscillatory wave-based cable partial discharge analysis method of claim 1, wherein: the parameters of the measured cable include at least length, joints and / or terminals, and line laying structure.

4. The oscillating wave based cable partial discharge analysis method of claim 1, wherein, The conditional vector includes at least one of temperature, humidity, and load current; the normalized oscillatory wave feature vector includes the oscillatory wave low-frequency envelope decay rate and / or the equivalent quality factor.

5. The oscillating wave based cable partial discharge analysis method of claim 1, wherein, The domain adaptive alignment includes: selecting a template sub-library matching the measured cable based on the clustering index of the ideal template; and aligning the template feature vectors of the template sub-library and the conditional oscillatory wave feature vector using at least one of statistical distance minimization, correlation alignment, and adversarial alignment; the clustering index of the ideal template includes the model, length interval, joint and / or terminal type of the cable.

6. The oscillating wave based cable partial discharge analysis method of claim 1, wherein, the output of the defect position includes: In the single-end reflection mode, the defect position is corrected according to the reflection polarity of the incident pulse and the first reflection pulse caused by the defect under the same trigger, and the reflection amplitude ratio of the reflection pulse envelope peak value to the incident pulse envelope peak value; In the double-end arrival mode, the arrival time difference and the propagation speed synchronously collected at both ends are observed, and the defect position and the defect trigger time are jointly estimated by using the least square method and / or the maximum likelihood estimation; The confidence interval is obtained by an error propagation model, and the input of the error propagation model at least includes the joint fingerprint stability and the residual error of the domain adaptive alignment, and in the double-end arrival mode, the variance of the arrival time difference is further included.

7. The oscillating wave based cable partial discharge analysis method of claim 6, wherein, The multi-grade up-pressing strategy adaptively selects the gear step and the maximum trigger number under the premise of meeting the standard constraints; the up-pressing is terminated in advance when the information gain is lower than a threshold and / or the joint fingerprint stability is lower than a preset threshold; and the difference feature and the PRPD joint fingerprint are recorded according to the gear.

8. The oscillating wave based cable partial discharge analysis method of claim 7, wherein, The operation and maintenance suggestion at least includes an operation and maintenance priority and a re-measurement interval; The operation and maintenance priority is used to represent the urgency of defect disposal and is expressed in the form of a score; the score is obtained by weighted calculation on the width of the confidence interval, the joint fingerprint stability and the intensity of the difference feature; and the re-measurement interval is determined according to the score of the operation and maintenance priority and a preset mapping relationship.

9. An oscillating wave-based cable partial discharge analysis apparatus, characterized by, Comprise: An oscillating wave excitation and collection unit for applying an oscillating wave and synchronously collecting a raw signal containing an oscillating wave low-frequency envelope and a high-frequency pulse, and for collecting parameters of a measured cable and environmental and working condition data; A structure normalization unit for performing structure normalization on the raw signal according to the parameters of the measured cable to obtain a standardized oscillating wave feature vector; A condition normalization unit for performing condition vector coding according to the environmental and working condition data, and performing condition normalization on the standardized oscillating wave feature vector according to the condition vector to obtain a conditioned oscillating wave feature vector; An alignment and comparison unit for performing domain adaptive alignment of the conditioned oscillating wave feature vector with an ideal template and performing difference comparison to obtain a difference feature; A joint fingerprint construction unit for generating a PRPD joint fingerprint according to the conditioned oscillating wave feature vector; A positioning and confidence evaluation unit for outputting a defect position, a defect type and a confidence interval based on the difference feature and the PRPD joint fingerprint; An adaptive up-pressing control unit for determining a multi-grade up-pressing strategy and an early stop criterion according to an information gain; A display and data management unit for recording and displaying the defect position, the defect type, the confidence interval, the up-pressing gear and the operation and maintenance suggestion; The structure normalization unit is configured to perform deconvolution, amplitude and phase equalization and / or group delay correction on the raw signal based on a parameterized line transmission model and / or an end-to-end transfer function; and estimate the collection link gain and / or bandwidth drift by a milepost point and / or an injection calibration pulse to update the normalization parameters. The condition normalization is configured to generate a network output scaling coefficient and a biasing coefficient through the environment and working condition parameters, and to impose a conditional affine transformation on the normalized oscillatory wave feature vector; The joint fingerprint construction unit is configured to map the high-frequency pulse to a phase axis with reference to a zero-phase point and / or an extreme point of a half cycle of the oscillation wave; perform exponential and / or damped sinusoidal fitting on a low-frequency envelope of the oscillation wave to obtain a low-frequency envelope decay rate and / or an equivalent quality factor of the oscillation wave, and calculate a joint fingerprint stability (S) by weighting a circular statistic (C) representing phase concentration and a coefficient of variation normalized complementary quantity (CV) representing energy consistency, ), ), ), ; in, and These are circular statistics ( ) and the normalized complement of the coefficient of variation ( The weighting coefficients.

Images