A Method and System for Diagnosing Interface Defects in Power Cables Based on High-Frequency Oscillating Wave Detection

Abstract

This invention relates to the field of power cable insulation condition detection technology, specifically disclosing a method and system for diagnosing interface defects in power cables based on high-frequency oscillating wave detection. The method involves acquiring a composite time-domain signal at the cable test end; performing variational mode decomposition on the composite time-domain signal to extract the component with the highest center frequency as a candidate discharge pulse group; performing convolutional sparse decomposition on the candidate discharge pulse group and a defect-free reflection waveform template set to extract a pure partial discharge pulse waveform and its reflection path delay value; inputting the delay value into a particle swarm optimization objective function to inversely determine the specific joint or branch node where the defect is located; finally, counting the number of pulse recurrences in each phase interval, and determining the existence and location of the defect after median filtering. This invention can accurately separate aliased signals and precisely locate interface defects in multi-branch, multi-joint topologies.

Description

Technical Field

[0001] This invention relates to the field of power cable insulation condition detection technology, specifically to a method and system for diagnosing interface defects in power cables based on high-frequency oscillation wave detection. Background Technology

[0002] High-frequency oscillating wave detection technology is one of the commonly used methods for partial discharge detection and location in power cables. This technology uses a solid-state switch to rapidly charge and discharge the cable, generating a high-frequency damped oscillating wave voltage, which simultaneously excites partial discharges inside the cable and at the interface. A high-frequency current transformer is then used to collect the discharge pulse signal. In existing implementations, the same high-frequency oscillating wave serves as both the defect excitation source and the detection signal for time-domain reflection location. The defect location is calculated by analyzing the round-trip time of the discharge pulse between the defect and the test end.

[0003] The existing technology has the following shortcomings: In the detection of high-frequency oscillation waves in multi-branch, multi-joint cable lines, existing technologies lack the ability to accurately separate the real discharge pulse from the strongly mixed composite time-domain signal containing the excitation main wave, multiple branch reflection waves, intermediate joint residual waveforms, and weak partial discharge pulses, and to precisely locate the specific joint or branch node to which it belongs. This is to overcome the location failure caused by the inability of the traditional time-domain reflection method to distinguish between reflected interference and discharge pulses. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for diagnosing interface defects in power cables based on high-frequency oscillation wave detection, so as to solve the problems mentioned above.

[0005] The objective of this invention can be achieved through the following technical solutions: A method for diagnosing interface defects in power cables based on high-frequency oscillating wave detection includes the following steps: S1: Acquire the composite time-domain signal at the cable test end under high-frequency oscillation wave excitation. The composite time-domain signal is composed of the superposition of the excitation main wave, the branch node reflected wave, the intermediate joint residual vibration waveform, and the partial discharge pulse generated by the interface defect. S2: Perform variational mode decomposition on the composite time-domain signal to obtain several intrinsic mode components with center frequencies separated from each other, and extract the component with the highest center frequency from all intrinsic mode components as a candidate discharge pulse group; S3: Perform convolutional sparse decomposition on the candidate discharge pulse group and the defect-free reflection waveform template set corresponding to the cable line topology. By solving the sparse coefficients, extract the pure partial discharge pulse waveform and the reflection path delay value associated with the partial discharge pulse waveform from the candidate discharge pulse group. S4: Input the pure partial discharge pulse waveform and its reflection path delay value into the particle swarm optimization objective function with the cable branch length and joint position as prior information. By iteratively searching, the location of the discharge pulse is inverted to determine the specific joint or branch node where the interface defect is located. S5: Count the number of recurrences of pure partial discharge pulses within multiple phase intervals of the same oscillation wave period. When the number of recurrences within a specified phase interval exceeds a preset threshold, output the existence and location information of interface defects.

[0006] As a further aspect of the present invention: S2 specifically includes: Calculate the power spectrum of the composite time-domain signal, and determine the initial value of the number of decomposed modes based on the number of energy peaks in the power spectrum; The composite time-domain signal is decomposed until the center frequency of each intrinsic mode component converges to a stable value; The intrinsic mode component with the highest center frequency and the smallest bandwidth is selected as the candidate discharge pulse group.

[0007] As a further aspect of the present invention: the process for determining the initial value is as follows: After applying a Hanning window to the composite time-domain signal, the power spectral density is calculated piecewise to obtain the piecewise average power spectrum. Perform second derivative calculations on the piecewise average power spectrum, and mark the zero-crossing points where the second derivative changes from positive to negative as the energy peaks; The number of all energy peaks is counted as the initial value for the number of decomposition modes.

[0008] As a further aspect of the present invention: S3 specifically includes: Perform a point-by-point sliding inner product operation between the candidate discharge pulse group and each template in the defect-free reflection waveform template set, and record the peak value of each inner product and its corresponding time offset. The first residual signal is obtained by subtracting the product of the peak value and the corresponding template waveform from the candidate discharge pulse group; Replace the candidate discharge pulse group with the first residual signal, and repeat the inner product operation and subtraction operation until the energy of the residual signal decays to below the preset threshold. The waveforms subtracted each time are summed up to obtain a pure partial discharge pulse waveform, and the corresponding time offset each time is used as the reflection path delay value.

[0009] As a further aspect of the present invention: the point-by-point sliding inner product operation specifically includes: The candidate discharge pulse groups are normalized in energy to obtain a normalized pulse sequence; The inner product sequence is calculated by performing an inner product calculation on each template in the defect-free reflection waveform template set with a sliding step of one sampling point; Perform parabolic fitting on the inner product sequence, using the vertex position of the fitted curve as the time offset and the inner product value at the vertex as the inner product peak.

[0010] As a further aspect of the present invention: S4 specifically includes: A multi-dimensional location space is constructed based on the cable branch length and joint location, and each particle in the particle swarm is assigned a hypothetical discharge location. The theoretical time delay is calculated based on the propagation path length between each hypothetical discharge location and the test end, and the absolute difference between the theoretical time delay and the reflection path time delay is used as the fitness of the particle. Update the movement speed and spatial position of each particle in the particle swarm in ascending order of fitness; When the minimum fitness value of all particles converges to the preset range, the hypothetical discharge location corresponding to the minimum value is output as the specific joint or branch node where the interface defect is located.

[0011] As a further aspect of the present invention: the calculation of the theoretical time delay value specifically includes: A topology diagram is established based on the cable branch length and joint location, and each edge in the diagram is assigned a propagation delay coefficient per unit length obtained from offline pulse testing. Starting from the node where the test end is located, traverse along the topology graph to the node where the hypothetical discharge occurs, and accumulate the product of the propagation delay coefficient of each edge and the edge length to obtain the basic delay value; Calculate the square root of the number of connectors, use the reciprocal of the square root as a correction factor, multiply the base delay value by the correction factor, and obtain the final theoretical delay value.

[0012] As a further aspect of the present invention: S5 specifically includes: Each oscillation wave period is divided into multiple continuous and non-overlapping phase intervals according to the phase, and the width of each phase interval is a predetermined phase angle. Record the actual phase angle of each pure partial discharge pulse in its respective oscillation wave period, classify it into the corresponding phase interval, and accumulate the count of pulses in each phase interval. The count values ​​obtained by accumulating the counts in each phase interval are arranged in ascending order of phase angle, and are denoted as the accumulating count sequence. Perform sliding window mid-value filtering on the accumulated counting sequence, mark the phase interval where the filtered value is greater than a preset threshold as the specified phase interval, and output the existence judgment of interface defects within the specified phase interval.

[0013] As a further aspect of the present invention: the sliding window midpoint filtering of the accumulated counting sequence specifically includes: The width of the filtering window is determined based on the length of the accumulated counting sequence, and the width is an odd number obtained by taking the square root of the sequence length. For each phase interval in the sequence, take half a window width of adjacent intervals forward and backward with the phase interval as the center, and form an array of unsorted arrays by taking the accumulated count values ​​from the intervals. The values ​​in the array to be sorted are arranged in ascending order, and the value in the middle position after the arrangement is taken as the median filter output of the phase interval. For intervals at both ends of the sequence that are less than half the width of a window, the number of values ​​participating in the sorting is expanded by repeating the endpoint values.

[0014] A power cable interface defect diagnosis system based on high-frequency oscillating wave detection includes: Composite signal acquisition module: Acquires composite time-domain signals at the cable test end under high-frequency oscillation wave excitation. The composite time-domain signal is composed of the superposition of the excitation main wave, the branch node reflected wave, the intermediate joint residual vibration waveform, and the partial discharge pulse generated by the interface defect. Variational Mode Decomposition Module: Performs variational mode decomposition on the composite time-domain signal to obtain several intrinsic mode components with center frequencies separated from each other, and extracts the component with the highest center frequency from all intrinsic mode components as a candidate discharge pulse group; Convolutional Sparse Decomposition Module: Performs convolutional sparse decomposition on the candidate discharge pulse group and the defect-free reflection waveform template set corresponding to the cable line topology. By solving the sparse coefficients, it extracts the pure partial discharge pulse waveform and the reflection path delay value associated with the partial discharge pulse waveform from the candidate discharge pulse group. Particle swarm optimization and localization module: Input the pure partial discharge pulse waveform and its reflection path delay value into the particle swarm optimization objective function with the cable branch length and joint position as prior information. Through iterative search, the occurrence position of the discharge pulse is inverted to determine the specific joint or branch node where the interface defect is located. Phase interval statistics and determination module: Counts the number of recurrences of pure partial discharge pulses in multiple phase intervals within the same oscillation wave period. When the number of recurrences in a specified phase interval exceeds a preset threshold, it outputs the existence determination of interface defects and their location information.

[0015] The beneficial effects of this invention are: (1) For the complex topology of multi-branch and multi-joint cable lines in urban power distribution networks, this invention can separate the pure partial discharge pulse waveform from the strongly mixed composite time domain signal containing the excitation main wave, multiple reflection waves and residual waveforms through the cascaded processing of variational mode decomposition and convolutional sparse decomposition. This effectively avoids the misjudgment of defect location caused by branch reflection and oscillation wave residual interference in the traditional time domain reflection method, and improves the engineering feasibility of defect location.

[0016] (2) This invention does not require additional hardware beyond the high-frequency current transformer. It can determine the specific intermediate joint or branch node where the interface defect is located by relying solely on particle swarm optimization and phase interval statistical filtering, and simultaneously output the defect existence determination. On-site maintenance personnel can directly perform targeted maintenance based on the output joint number or branch name, reducing the blindness of cable line fault diagnosis and the workload of disassembly inspection. Attached Figure Description

[0017] The invention will now be further described with reference to the accompanying drawings.

[0018] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system block diagram of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Please see Figure 1 As shown, this invention is a method for diagnosing interface defects in power cables based on high-frequency oscillation wave detection, comprising the following steps: S1: Acquire the composite time-domain signal at the cable test end under high-frequency oscillation wave excitation. The composite time-domain signal is composed of the superposition of the excitation main wave, the branch node reflected wave, the intermediate joint residual vibration waveform, and the partial discharge pulse generated by the interface defect. S2: Perform variational mode decomposition on the composite time-domain signal to obtain several intrinsic mode components with center frequencies separated from each other, and extract the component with the highest center frequency from all intrinsic mode components as a candidate discharge pulse group; S3: Perform convolutional sparse decomposition on the candidate discharge pulse group and the defect-free reflection waveform template set corresponding to the cable line topology. By solving the sparse coefficients, extract the pure partial discharge pulse waveform and the reflection path delay value associated with the partial discharge pulse waveform from the candidate discharge pulse group. S4: Input the pure partial discharge pulse waveform and its reflection path delay value into the particle swarm optimization objective function with the cable branch length and joint position as prior information. By iteratively searching, the location of the discharge pulse is inverted to determine the specific joint or branch node where the interface defect is located. S5: Count the number of recurrences of pure partial discharge pulses within multiple phase intervals of the same oscillation wave period. When the number of recurrences within a specified phase interval exceeds a preset threshold, output the existence and location information of interface defects.

[0021] In S1, a composite time-domain signal is acquired at the cable test end under high-frequency oscillation wave excitation. The composite time-domain signal is composed of the superposition of the excitation main wave, the branch node reflected wave, the intermediate joint residual vibration waveform, and the partial discharge pulse generated by the interface defect. Specifically, it includes: When performing high-frequency oscillating wave detection, a high-frequency current transformer is installed at the grounding lead of the cable under test. This transformer has a sampling bandwidth of 100kHz to 50MHz and is used to couple transient current signals flowing through the grounding wire. The high-voltage output terminal of the high-frequency oscillating wave tester is connected to the cable conductor, and the high-voltage divider inside the tester is simultaneously connected to the cable test terminal to record the applied oscillating wave voltage waveform. The tester is started, and a high-frequency damped oscillating wave voltage with a frequency of 100kHz to 1MHz is applied to the cable. The voltage amplitude gradually increases from 0kV to a preset peak value, typically 1.5 times the rated phase voltage of the cable. During each oscillating wave application, the high-frequency current transformer continuously acquires the time-domain signal at a sampling rate of no less than 100 mega-sampling points per second, while the high-voltage divider simultaneously acquires the voltage time-domain waveform. The acquired composite time-domain signal is composed of the following four components superimposed: the excitation main wave directly applied to the cable by the tester; the reflected waves generated by impedance changes at each branch node and intermediate joint of the cable line; the residual vibration waveform formed by dielectric loss and impedance mismatch as the oscillating wave propagates inside the cable; and the partial discharge pulse generated under the excitation of the oscillating wave electric field when defects such as air gaps, stripping, or semiconductive layer tips exist at the cable interface. All acquired composite time-domain signals, along with the synchronized voltage waveform, are stored in the control computer as the basis data for subsequent signal decomposition and defect location. Throughout the acquisition process, the output of the high-frequency current transformer and the output of the high-voltage divider are strictly synchronized to ensure that the occurrence time of each signal component relative to the voltage phase of the oscillating wave can be accurately obtained during subsequent analysis.

[0022] In S2, variational mode decomposition is performed on the composite time-domain signal to obtain several intrinsic mode components with mutually separated center frequencies. The component with the highest center frequency is then extracted from all intrinsic mode components as a candidate discharge pulse group, specifically including: Parameter configuration before performing variational mode decomposition on the composite time-domain signal. The power spectrum of the composite time-domain signal is calculated, and the initial number of decomposition modes is determined based on the number of energy peaks in the power spectrum. The specific calculation process is as follows: The composite time-domain signal is divided into multiple data segments in chronological order, each segment having a length of 256 sampling points, with 128 sampling points overlapping between adjacent segments. A Hanning window is applied to each data segment, i.e., each sampling point is multiplied by the Hanning window function value, where the window function value is equal to 0.5 multiplied by (1 minus the cosine value), and the angle of the cosine value is the product of twice pi and the sampling point index divided by the data segment length minus 1. Then, a Fast Fourier Transform is performed on each windowed data segment. The square of the magnitude of each frequency component after the transform is divided by the data segment length to obtain the power spectral density value of that data segment. The power spectral density values ​​of all data segments are averaged at the same frequency point to obtain the segmented average power spectrum.

[0023] Perform second-order derivative operations on the above segmented average power spectrum. Calculate the first-order derivative at each frequency point in the power spectrum, which is the power spectral density value at that frequency point minus the power spectral density value at the previous frequency point, and then divided by the frequency interval between the two adjacent frequency points. Similarly, perform a difference operation on the first-order derivative to obtain the second-order derivative at each frequency point, which is the first-order derivative at that frequency point minus the first-order derivative at the previous frequency point, and then divided by the frequency interval. Traverse along the frequency axis from smallest to largest, marking the frequency points where the second-order derivative changes from positive to negative as the positions of energy peaks. Count the number of all marked energy peaks, and use this number as the initial value for the number of decomposed modes in variational mode decomposition.

[0024] The penalty factor for variational mode decomposition is set to 2000, and the convergence tolerance is 1 x 10^-7. The center frequency and bandwidth of each modal component are initialized. Using a composite time-domain signal as input, decomposition is performed according to the iterative framework of variational mode decomposition: in each iteration, the spectrum of each modal component is updated first, then the center frequency of each modal component is updated, and finally the Lagrange multipliers are updated. After each iteration, the sum of squares of the differences between the spectra of the modal components obtained in the current iteration and the previous iteration is calculated, and then divided by the sum of squares of the spectra of the modal components in the previous iteration. If this ratio is less than the convergence tolerance, the iteration stops. At this point, the center frequencies of all modal components have converged to stable values. After decomposition, the same number of intrinsic mode components as the initial number of modes are obtained, each component having its own center frequency and bandwidth. The intrinsic mode component with the largest center frequency value is selected, and its bandwidth is confirmed to be the smallest among all components. This component is used as a candidate discharge pulse group for subsequent convolutional sparse decomposition steps.

[0025] In the above process, the center frequency is defined as the frequency value corresponding to the maximum amplitude in the spectrum of the modal component. The bandwidth is defined as the frequency value at which the amplitude of the modal component drops to its maximum value in the spectrum. The frequency range width corresponding to the multiple. By filtering out the intrinsic mode components with the highest center frequency and the smallest bandwidth, the high-frequency, narrow-band characteristics of the partial discharge pulse can be effectively preserved, while suppressing interference from low-frequency oscillations and broadband noise.

[0026] In S3, the candidate discharge pulse group and the defect-free reflection waveform template set corresponding to the cable line topology are convolved and sparsely decomposed. By solving the sparse coefficients, the pure partial discharge pulse waveform and the reflection path delay value associated with the partial discharge pulse waveform are extracted from the candidate discharge pulse group. Specifically, this includes: First, a template set of defect-free reflected waveforms corresponding to the cable line topology is constructed. This template set consists of multiple time-series waveforms, each corresponding to the reflected waveform generated by a specific branch node or intermediate joint in the cable line under a defect-free state and excited by a unit pulse. The templates are obtained as follows: before the cable is put into operation or when the defect-free status is confirmed, a narrow pulse with a width of 10 nanoseconds and an amplitude of 1 volt is applied to the cable. The reflected waveform is collected at the test end using a high-frequency current transformer, and the waveform is normalized by energy to serve as the template for that node. This operation is repeated for all branch nodes and intermediate joints to obtain the template set.

[0027] The candidate discharge pulse group is used as the input signal, denoted as the signal sequence. Energy normalization is performed on this candidate discharge pulse group: first, the sum of squares of all sample points in the sequence is calculated; then, the value of each sample point in the sequence is divided by the square root of the sum of squares to obtain the normalized pulse sequence. The total energy of the normalized pulse sequence is 1.

[0028] The normalized pulse sequence is subjected to a point-by-point sliding inner product operation with each template in the defect-free reflection waveform template set. Specifically, for a normalized pulse sequence with L sampling points and a template with M sampling points (where L is greater than M), the starting point of the template is aligned with the first sampling point of the pulse sequence. The sum of the products of the corresponding sampling point values ​​in the overlapping portion of the two sequences is calculated to obtain the first inner product value. Then, the template is slid one sampling point to the right, and the sum of the products of the overlapping portions is calculated again to obtain the second inner product value. This operation is repeated until the last sampling point of the template is aligned with the last sampling point of the pulse sequence to obtain the last inner product value. All inner product values ​​are arranged in sliding order to form an inner product value sequence, the length of which is L minus M plus 1.

[0029] Parabolic fitting is performed on the inner product sequence to determine the precise peak position and time offset. Let k be the index corresponding to the maximum value in the inner product sequence. The inner product values ​​at indices k-1, k, and k+1 are denoted as follows: , , Calculate the vertex offset of the fitted parabola using the following formula. The calculation expression is: ; in, The inner product value at index k-1 The inner product value at index k. It is the inner product value at index k+1. The value of is between -0.5 and +0.5. The peak value of the inner product at the vertex is calculated by the following formula: ; in, This represents the exact inner product peak obtained from the fit. The time offset equals the index k plus... Then multiply by the sampling interval, which is the reciprocal of the sampling rate. Record the peak value of the inner product corresponding to this template and its time offset.

[0030] After performing the above operation on all templates, find the maximum value among all inner product peaks, and take the template corresponding to the maximum value as the best template for this matching. Record the inner product peak corresponding to the maximum value as the current best peak value, and record its corresponding time offset as the current time delay. Then, subtract the product of the current best peak value and the best template waveform from the candidate discharge pulse group: that is, for each sampling point, subtract the current best peak value multiplied by the value of the best template waveform at the sampling point from the value of the candidate discharge pulse group at that sampling point (the template waveform needs to be translated and aligned according to the time offset), to obtain the first residual signal.

[0031] The original candidate discharge pulse group is replaced with the first residual signal, and the above-mentioned energy normalization, point-by-point sliding inner product, parabolic fitting, peak comparison, and signal subtraction operations are repeated. Each repetition yields a new optimal peak value and its corresponding time offset, and generates a new residual signal. The repetition process continues until the energy of the current residual signal (i.e., the sum of squares of all sampling points) decays below a preset threshold. This preset threshold is set to one percent of the energy of the original candidate discharge pulse group.

[0032] The waveforms subtracted each time (i.e., the optimal peak value multiplied by the corresponding translated optimal template waveform) are aligned in time and then accumulated to obtain a pure partial discharge pulse waveform. At the same time, the corresponding time offsets are arranged in sequence as a reflection path delay value sequence, where each delay value represents the time it takes for the partial discharge pulse to propagate from the defect point to the test end.

[0033] In S4, the pure partial discharge pulse waveform and its reflection path delay value are input into the particle swarm optimization objective function, which uses the cable branch length and joint position as prior information. Through iterative search, the location of the discharge pulse is retrieved, determining the specific joint or branch node where the interface defect is located. Specifically, this includes: A multi-dimensional location space is established based on the actual topology of the cable under test. The cable line includes one test end node, multiple branch nodes, and multiple intermediate joint nodes. The length of each cable segment is measured in meters, and the position coordinates of each branch node and each intermediate joint are recorded. An undirected topology graph is constructed using these nodes as vertices and the cable segments as edges. Each particle in the particle swarm optimization algorithm is assigned a hypothetical discharge location, represented by the distance coordinates of a specific node or edge in the topology graph. Initially, 50 particles are randomly generated in the particle swarm, with each particle's initial position uniformly distributed throughout the topology graph, and its initial movement speed set to zero.

[0034] For each particle representing a hypothetical discharge location, the theoretical time delay between it and the test end is calculated. The specific calculation process consists of three steps. Step 1: Obtain the propagation time delay coefficient per unit length of the cable line through offline pulse testing. Specifically, when the cable is in a defect-free state, a narrow pulse with a width of 10 nanoseconds is applied to the test end. Simultaneously, the round-trip time of the pulse is measured at a reflection point at a known distance from either end. The round-trip time is divided by twice the distance to obtain the propagation time delay coefficient per unit length, in nanoseconds per meter. For different branches and joints, if the cable type is the same, the same coefficient is used. Step 2: Starting from the node where the test end is located, traverse along the shortest path in the topology graph to the node where the hypothetical discharge location is located. If the hypothetical location is located in the middle of an edge, traverse to the starting point of that edge and then proceed along that edge to the target coordinates. Accumulate the product of the propagation time delay coefficient of each traversed edge and the length of that edge (in meters), i.e., accumulate the time delay contribution of each cable segment to obtain the basic time delay value, in nanoseconds. Step 3: Calculate the square root reciprocal correction factor for the number of joints. Count the total number of all intermediate joints (excluding branch nodes) in the entire topology diagram, denoted as the total number of joints. Calculate the square root of the total number of joints, and then divide 1 by this square root to obtain the correction factor. Multiply the base time delay value by this correction factor to obtain the final theoretical time delay value. This correction factor is used to compensate for the additional propagation delay caused by multiple reflections of the oscillating wave in the case of multiple joints.

[0035] Calculate the fitness of each particle. Record the reflection path delay value output in step S3 as the measured delay value. For each particle, subtract the measured delay value from its theoretical delay value, and take the absolute value. Use this absolute difference as the particle's fitness. The smaller the fitness, the closer the particle's hypothetical position is to the actual defect position.

[0036] The movement speed and spatial position of each particle in the particle swarm are updated in ascending order of fitness. The update rule is as follows: for each particle, its historical best position (i.e., the position with the lowest fitness in all historical iterations) is recorded, along with the global best position of the entire particle swarm (i.e., the position with the lowest fitness among all particles). The new speed of each particle consists of three parts: the first part is the original speed multiplied by an inertia weight, which is initially set to 0.8 and decays by a factor of 0.99 after each iteration; the second part is the individual cognition part, which is a random number multiplied by an individual learning factor and then multiplied by the difference between the historical best position and the current position, where the individual learning factor is 1.5; the third part is the social cognition part, which is another random number multiplied by a social learning factor and then multiplied by the difference between the global best position and the current position, where the social learning factor is 1.5. The new speed is added to the current position to obtain the particle's new position. If the new position exceeds the spatial boundary of the topology graph, it is restricted to the boundary.

[0037] When the change in the minimum fitness value of all particles over five consecutive iterations is less than one-thousandth of the minimum fitness value itself, the minimum fitness value is considered to have converged to a preset range. At this point, the hypothetical discharge location corresponding to the minimum value is output as the specific joint or branch node where the interface defect is located. If the location is at a node when converged, the node name is directly output; if it is in the middle of an edge, the actual length of the edge from the test end is output.

[0038] In S5, the number of recurrences of pure partial discharge pulses within multiple phase intervals of the same oscillation wave period is counted. When the number of recurrences within a specified phase interval exceeds a preset threshold, the existence and location information of the interface defect are output, specifically including: The clean partial discharge pulse waveform is phase-aligned with the synchronously acquired high-voltage divider voltage waveform. The start and end points of each oscillation cycle are extracted from the voltage waveform; one oscillation cycle is defined as the time interval between two adjacent zero-crossing points of the voltage waveform. Each oscillation cycle is divided into multiple continuous and non-overlapping phase intervals according to phase angle. Specifically, the width of each phase interval is set to 10 degrees, thus a complete 360-degree cycle is divided into 36 phase intervals. The first phase interval corresponds to 0 to 10 degrees, the second to 10 to 20 degrees, and so on, with the 36th corresponding to 350 to 360 degrees.

[0039] For each pulse in a pure partial discharge pulse waveform, determine its phase angle value by comparing its occurrence time on the time axis with the phase angle at the same moment in the synchronization voltage waveform. Place the pulse into the phase interval corresponding to its phase angle. Within the same oscillation wave period, accumulate the number of pulses appearing in each phase interval. Traverse all oscillation wave periods (typically applying 20 to 30 oscillation waves) to obtain the total pulse count value for each phase interval. Arrange the count values ​​of all phase intervals in phase order to form an accumulated count sequence of length 36.

[0040] Perform sliding window mid-range filtering on the above accumulated count sequence. First, determine the width of the filtering window: the length of the accumulated count sequence is 36. Calculate the square root of 36, which is 6. Round 6 to the nearest integer, which is still 6. To ensure the window width is odd, add 1 to 6, resulting in 7. Therefore, the filtering window width is set to 7 phase intervals. For each phase interval in the sequence (index from 1 to 36), take the 3 adjacent intervals forward and 3 adjacent intervals backward, along with the center interval itself, for a total of 7 intervals. Extract the accumulated count values ​​corresponding to these 7 intervals to form an array to be sorted. If the center interval is close to the end of the sequence, for example, the interval with index 1, and there are fewer than 3 intervals forward, repeat the end value: copy the value of the interval with index 1 multiple times to make up the 3 values ​​needed to move forward. Specifically, for index 1, 3 values ​​are needed to move forward, so take all the values ​​of index 1; take the values ​​of indices 2, 3, and 4 backward, along with the center, for a total of 7 values. For index 2, the value of index 1 is taken twice and the value of index 1 is taken once (the value of index 1 can actually be repeated) to ensure that each central interval obtains 7 values. Then, the 7 values ​​in the unsorted array are arranged in ascending order, and the value in the 4th position after arrangement (i.e., the median) is taken as the filtered output value of that central interval. The above operation is performed on all 36 intervals in sequence to obtain a new filtered count sequence of length 36.

[0041] A preset threshold is set. This threshold is determined based on background noise statistics from multiple defect-free cable tests: the same testing process is performed on a known defect-free cable, the average count value for each phase interval is calculated, and five times this average value is used as the preset threshold. In this embodiment, the preset threshold is set to 3. Each value in the filtered count sequence is compared with the preset threshold. If the value of a certain phase interval is greater than 3, the phase interval is marked as a specified phase interval. If one or more specified phase intervals exist, it is determined that there is an interface defect in the cable; if no phase interval has a value greater than the threshold, it is determined that there is no obvious interface defect in the cable. When a defect is determined to exist, the location information of the joint or branch node where the defect is located, output in step S4, is combined to finally output the determination of the existence of the interface defect and its specific location information. This location information includes: the intermediate joint number where the defect is located, the branch node name, or the actual length of the defect from the test end.

[0042] Please see Figure 2 As shown, the power cable interface defect diagnosis system based on high-frequency oscillation wave detection includes: Composite signal acquisition module: Acquires composite time-domain signals at the cable test end under high-frequency oscillation wave excitation. The composite time-domain signal is composed of the superposition of the excitation main wave, the branch node reflected wave, the intermediate joint residual vibration waveform, and the partial discharge pulse generated by the interface defect. Variational Mode Decomposition Module: Performs variational mode decomposition on the composite time-domain signal to obtain several intrinsic mode components with center frequencies separated from each other, and extracts the component with the highest center frequency from all intrinsic mode components as a candidate discharge pulse group; Convolutional Sparse Decomposition Module: Performs convolutional sparse decomposition on the candidate discharge pulse group and the defect-free reflection waveform template set corresponding to the cable line topology. By solving the sparse coefficients, it extracts the pure partial discharge pulse waveform and the reflection path delay value associated with the partial discharge pulse waveform from the candidate discharge pulse group. Particle swarm optimization and localization module: Input the pure partial discharge pulse waveform and its reflection path delay value into the particle swarm optimization objective function with the cable branch length and joint position as prior information. Through iterative search, the occurrence position of the discharge pulse is inverted to determine the specific joint or branch node where the interface defect is located. Phase interval statistics and determination module: Counts the number of recurrences of pure partial discharge pulses in multiple phase intervals within the same oscillation wave period. When the number of recurrences in a specified phase interval exceeds a preset threshold, it outputs the existence determination of interface defects and their location information.

[0043] The working principle of this invention is as follows: A composite time-domain signal is acquired at the test end of a cable under high-frequency oscillation wave excitation. This signal is composed of the superposition of the excitation main wave, the reflected wave from the branch node, the residual waveform of the intermediate joint, and the partial discharge pulse generated by interface defects. Variational mode decomposition is performed on the composite time-domain signal to obtain several intrinsic mode components with separated center frequencies. The component with the highest center frequency and the smallest bandwidth is extracted as a candidate discharge pulse group. The candidate discharge pulse group is then convolved with a defect-free reflection waveform template set corresponding to the cable line topology for sparse decomposition. Through point-by-point sliding inner product, parabolic fitting, and iterative subtraction, the signal is obtained from... The pure partial discharge pulse waveform and its reflection path delay value are extracted from the candidate discharge pulse group. Then, the pure partial discharge pulse waveform and its reflection path delay value are input into the particle swarm optimization objective function with the cable branch length and joint position as prior information. The occurrence position of the discharge pulse is inverted through iterative search to determine the specific joint or branch node where the interface defect is located. Finally, the recurrence number of pure partial discharge pulses is counted in multiple phase intervals of the same oscillation wave period. After sliding window mid-value filtering, when the recurrence number in the specified phase interval exceeds the preset threshold, the existence determination of the interface defect and its location information are output.

[0044] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for diagnosing interface defects in power cables based on high-frequency oscillating wave detection, characterized in that, Includes the following steps: S1: Acquire the composite time-domain signal at the cable test end under high-frequency oscillation wave excitation. The composite time-domain signal is composed of the superposition of the excitation main wave, the branch node reflected wave, the intermediate joint residual vibration waveform, and the partial discharge pulse generated by the interface defect. S2: Perform variational mode decomposition on the composite time-domain signal to obtain several intrinsic mode components with center frequencies separated from each other, and extract the component with the highest center frequency from all intrinsic mode components as a candidate discharge pulse group; S3: Perform convolutional sparse decomposition on the candidate discharge pulse group and the defect-free reflection waveform template set corresponding to the cable line topology. By solving the sparse coefficients, extract the pure partial discharge pulse waveform and the reflection path delay value associated with the partial discharge pulse waveform from the candidate discharge pulse group. Specifically, this includes: performing point-by-point sliding inner product operation on the candidate discharge pulse group and each template in the defect-free reflection waveform template set, and recording the peak value of each inner product and its corresponding time offset. The first residual signal is obtained by subtracting the product of the peak value and the corresponding template waveform from the candidate discharge pulse group; Replace the candidate discharge pulse group with the first residual signal, and repeat the inner product operation and subtraction operation until the energy of the residual signal decays to below the preset threshold. The waveforms subtracted each time are summed up to obtain a pure partial discharge pulse waveform, and the corresponding time offset each time is used as the reflection path delay value. S4: Input the pure partial discharge pulse waveform and its reflection path delay value into the particle swarm optimization objective function with the cable branch length and joint position as prior information. By iteratively searching, the location of the discharge pulse is inverted to determine the specific joint or branch node where the interface defect is located. S5: Count the number of recurrences of pure partial discharge pulses within multiple phase intervals of the same oscillation wave period. When the number of recurrences within a specified phase interval exceeds a preset threshold, output the existence and location information of interface defects.

2. The method for diagnosing interface defects of power cables based on high-frequency oscillation wave detection according to claim 1, characterized in that, S2 specifically includes: Calculate the power spectrum of the composite time-domain signal, and determine the initial value of the number of decomposed modes based on the number of energy peaks in the power spectrum; The composite time-domain signal is decomposed until the center frequency of each intrinsic mode component converges to a stable value; The intrinsic mode component with the highest center frequency and the smallest bandwidth is selected as the candidate discharge pulse group.

3. The method for diagnosing interface defects of power cables based on high-frequency oscillation wave detection according to claim 2, characterized in that, The process for determining the initial value is as follows: After applying a Hanning window to the composite time-domain signal, the power spectral density is calculated piecewise to obtain the piecewise average power spectrum. Perform second derivative calculations on the piecewise average power spectrum, and mark the zero-crossing points where the second derivative changes from positive to negative as the energy peaks; The number of all energy peaks is counted as the initial value for the number of decomposition modes.

4. The method for diagnosing interface defects in power cables based on high-frequency oscillation wave detection according to claim 1, characterized in that, The point-by-point sliding inner product operation specifically includes: The candidate discharge pulse groups are normalized in energy to obtain a normalized pulse sequence; The inner product sequence is calculated by performing an inner product calculation on each template in the defect-free reflection waveform template set with a sliding step of one sampling point; Perform parabolic fitting on the inner product sequence, using the vertex position of the fitted curve as the time offset and the inner product value at the vertex as the inner product peak.

5. The method for diagnosing interface defects in power cables based on high-frequency oscillating wave detection according to claim 1, characterized in that, S4 specifically includes: A multi-dimensional location space is constructed based on the cable branch length and joint location, and each particle in the particle swarm is assigned a hypothetical discharge location. The theoretical time delay is calculated based on the propagation path length between each hypothetical discharge location and the test end, and the absolute difference between the theoretical time delay and the reflection path time delay is used as the fitness of the particle. Update the movement speed and spatial position of each particle in the particle swarm in ascending order of fitness; When the minimum fitness value of all particles converges to the preset range, the hypothetical discharge location corresponding to the minimum value is output as the specific joint or branch node where the interface defect is located.

6. The method for diagnosing interface defects of power cables based on high-frequency oscillating wave detection according to claim 5, characterized in that, The calculation of the theoretical time delay value specifically includes: A topology diagram is established based on the cable branch length and joint location, and each edge in the diagram is assigned a propagation delay coefficient per unit length obtained from offline pulse testing. Starting from the node where the test end is located, traverse along the topology graph to the node where the hypothetical discharge occurs, and accumulate the product of the propagation delay coefficient of each edge and the edge length to obtain the basic delay value; Calculate the square root of the number of connectors, use the reciprocal of the square root as a correction factor, multiply the base delay value by the correction factor, and obtain the final theoretical delay value.

7. The method for diagnosing interface defects of power cables based on high-frequency oscillation wave detection according to claim 1, characterized in that, S5 specifically includes: Each oscillation wave period is divided into multiple continuous and non-overlapping phase intervals according to the phase, and the width of each phase interval is a predetermined phase angle. Record the actual phase angle of each pure partial discharge pulse in its respective oscillation wave period, classify it into the corresponding phase interval, and accumulate the count of pulses in each phase interval. The count values ​​obtained by accumulating the counts in each phase interval are arranged in ascending order of phase angle, and are denoted as the accumulating count sequence. Perform sliding window mid-value filtering on the accumulated counting sequence, mark the phase interval where the filtered value is greater than a preset threshold as the specified phase interval, and output the existence judgment of interface defects within the specified phase interval.

8. The method for diagnosing interface defects of power cables based on high-frequency oscillating wave detection according to claim 7, characterized in that, The process of performing sliding window value filtering on the accumulated count sequence specifically includes: The width of the filtering window is determined based on the length of the accumulated counting sequence, and the width is an odd number obtained by taking the square root of the sequence length. For each phase interval in the sequence, take half a window width of adjacent intervals forward and backward with the phase interval as the center, and form an array of unsorted arrays by taking the accumulated count values ​​from the intervals. The values ​​in the array to be sorted are arranged in ascending order, and the value in the middle position after the arrangement is taken as the median filter output of the phase interval. For intervals at both ends of the sequence that are less than half the width of a window, the number of values ​​participating in the sorting is expanded by repeating the endpoint values.

9. A power cable interface defect diagnosis system based on high-frequency oscillating wave detection, characterized in that, The method for diagnosing interface defects of power cables based on high-frequency oscillating wave detection as described in any one of claims 1-8 includes: Composite signal acquisition module: Acquires composite time-domain signals at the cable test end under high-frequency oscillation wave excitation. The composite time-domain signal is composed of the superposition of the excitation main wave, the branch node reflected wave, the intermediate joint residual vibration waveform, and the partial discharge pulse generated by the interface defect. Variational Mode Decomposition Module: Performs variational mode decomposition on the composite time-domain signal to obtain several intrinsic mode components with center frequencies separated from each other, and extracts the component with the highest center frequency from all intrinsic mode components as a candidate discharge pulse group; Convolutional Sparse Decomposition Module: Performs convolutional sparse decomposition on the candidate discharge pulse group and the defect-free reflection waveform template set corresponding to the cable line topology. By solving the sparse coefficients, it extracts the pure partial discharge pulse waveform and the reflection path delay value associated with the partial discharge pulse waveform from the candidate discharge pulse group. Particle swarm optimization and localization module: Input the pure partial discharge pulse waveform and its reflection path delay value into the particle swarm optimization objective function with the cable branch length and joint position as prior information. Through iterative search, the occurrence position of the discharge pulse is inverted to determine the specific joint or branch node where the interface defect is located. Phase interval statistics and determination module: Counts the number of recurrences of pure partial discharge pulses in multiple phase intervals within the same oscillation wave period. When the number of recurrences in a specified phase interval exceeds a preset threshold, it outputs the existence determination of interface defects and their location information.

Images