Cable fault abnormal data self-checking method and system based on multi-source traveling wave comparison

Abstract

The application discloses a cable fault abnormal data self-checking method and system based on multi-source traveling wave comparison, relates to the field of power system protection and fault diagnosis, and comprises the following steps: taking a cross-connection grounding box as a multi-point observation node, collecting and buffering overlapping data segments which can be cross-correlated by same-window triggering under the condition that sampling clocks are inconsistent; extracting wave head transient characteristics of multi-phase sheath traveling waves, obtaining relative time differences of each channel by cross-correlation, and constructing a relative time difference matrix and an energy ratio matrix in combination with wave head neighborhood energy ratios; generating a normalized reliability weight according to a signal-to-noise ratio and energy consistency, jointly estimating wave head arrival times of each channel in a weighted iterative time estimation model, and identifying and correcting outlier abnormal data to obtain a self-checked arrival time sequence; and then mapping and outputting a fault point position in combination with a grounding box position / segment length and a traveling wave propagation speed. Thus, reliable positioning is achieved without high-precision synchronization.

Description

Technical Field

[0001] This application relates to the field of power system protection and fault diagnosis technology, and in particular to a method and system for self-verification of cable fault anomaly data based on multi-source traveling wave comparison. Background Technology

[0002] High-voltage cables are the lifeblood of urban power grids, and their operational reliability is paramount. Traditional cable fault location techniques mainly include impedance methods and traveling wave methods. While the impedance method is simple in principle, it is easily affected by factors such as transition resistance and load current fluctuations, resulting in large ranging errors and failing to meet the precise location requirements of modern power grids. In contrast, the traveling wave method utilizes the propagation characteristics of transient traveling waves generated by faults in the cable for ranging. Because it is unaffected by fault type and load, it theoretically possesses higher location accuracy and has therefore become the mainstream technology for cable fault monitoring.

[0003] However, current traveling wave positioning technology still has significant limitations in practical applications. The single-ended traveling wave method mainly relies on analyzing waveform reflection characteristics, but in complex cable networks, the reflected wave from a fault point is easily confused with the reflected wave from points of impedance discontinuity, leading to identification difficulties. While the widely used double-ended traveling wave method solves some of the identification problems, it heavily relies on high-precision clock synchronization (such as GPS or BeiDou timing) and high-speed communication links at both ends. If the timing signal is unstable or there is a system time difference, the ranging error will increase sharply. Furthermore, when dealing with multi-circuit lines or complex branch structures, the double-ended method also faces challenges in terms of principle adaptability and equipment cost.

[0004] Furthermore, existing monitoring methods for the cross-interconnection grounding systems commonly used in high-voltage cables are often limited in function. Current testing devices typically require the removal of the interconnecting conductors and sheath protectors before testing can be conducted. Such invasive operations are not only cumbersome to maintain, but also prone to creating safety hazards due to operational errors during reconnection. Summary of the Invention

[0005] This application provides a method, system, storage medium, computer program product, and electronic device for self-verification of cable fault anomaly data based on multi-source traveling wave comparison, which at least solves the problems in current related technologies where cable fault location relies excessively on high-precision clock synchronization and single-point waveform criteria, and where the location reliability and engineering applicability are insufficient in complex line environments.

[0006] In a first aspect, embodiments of this application provide a self-verification method for cable fault anomaly data based on multi-source traveling wave comparison. The method includes: in response to the same fault event occurring in a target cable line, acquiring multi-phase sheath traveling wave signals collected by multi-source sensor arrays at at least two cross-interconnected grounding boxes along the target cable line; wherein the sampling clocks of each sensor in the multi-source sensor array are independent and time asynchrony is allowed, and for the same fault event, the at least two cross-interconnected grounding boxes complete the acquisition and buffering of the multi-phase sheath traveling wave signals within the same fault trigger acquisition window, so that the multi-phase sheath traveling wave signals acquired at different cross-interconnected grounding boxes have an overlapping interval on the time axis that can be used for cross-correlation analysis; extracting transient features from the multi-phase sheath traveling wave signals to obtain a preprocessed signal sequence containing traveling wave front features; performing cross-correlation analysis on any two signals in the preprocessed signal sequence, calculating the relative time difference between sensor pairs at different locations or different phases, and calculating the relative time difference between each sensor pair in the wave front neighborhood based on the preprocessed signal sequence. The energy ratio between each sensor pair is calculated based on the energy characteristics, thereby constructing a relative time difference matrix and an energy ratio matrix for the multi-source signals. Based on the signal-to-noise ratio index of each sensor and the energy ratio matrix, normalized weights characterizing the reliability of signal correlation between each sensor pair are calculated, and a reliability weight matrix is ​​generated. A time estimation model based on weighted iteration is constructed using the relative time difference matrix and the reliability weight matrix. The estimated arrival time of each sensor is updated through iterative calculation, and abnormal sensor data deviating from the statistical distribution are identified during the iteration process. The abnormal sensor data is corrected using weighted constraints of normal sensor data to obtain a target wavefront arrival time sequence after self-verification and calibration. Based on the target wavefront arrival time sequence, cable line length parameters, and traveling wave propagation speed, the fault location is calculated and output. The cable line length parameters include the position parameters of the at least two cross-interconnected grounding boxes on the target cable line and / or the segment length parameters of adjacent cable segments, which are used to map the target wavefront arrival time sequence to the fault location.

[0007] Secondly, embodiments of this application provide a cable fault anomaly data self-verification system based on multi-source traveling wave comparison. The system includes: an acquisition and buffering unit, used to acquire multi-phase sheath traveling wave signals collected by multi-source sensor arrays at at least two cross-interconnected grounding boxes along the target cable line in response to the same fault event; wherein the sampling clocks of each sensor in the multi-source sensor array are independent and time asynchrony is allowed, and for the same fault event, the at least two cross-interconnected grounding boxes complete the acquisition and buffering of the multi-phase sheath traveling wave signals within the same fault trigger acquisition window, so that the multi-phase sheath traveling wave signals collected at different cross-interconnected grounding boxes have an overlapping interval on the time axis that can be used for cross-correlation analysis; a transient feature extraction unit, used to extract transient features from the multi-phase sheath traveling wave signals to obtain a preprocessed signal sequence containing traveling wave front features; and a time difference and energy consistency construction unit, used to perform cross-correlation analysis on any two signals in the preprocessed signal sequence, calculate the relative time difference between sensor pairs at different locations or different phases, and construct a time difference and energy consistency model based on the preprocessed signal sequence within the wave front neighborhood. The system calculates the energy ratio between each sensor pair based on its energy characteristics, thereby constructing a relative time difference matrix and an energy ratio matrix for the multi-source signals. A reliability weight generation unit calculates normalized weights characterizing the reliability of signal correlation between each sensor pair based on the signal-to-noise ratio index of each sensor and the energy ratio matrix, and generates a reliability weight matrix. A time estimation self-verification unit constructs a weighted iterative time estimation model using the relative time difference matrix and the reliability weight matrix. It iteratively updates the estimated wavefront arrival time of each sensor, identifies abnormal sensor data deviating from the statistical distribution during the iteration process, and corrects the abnormal sensor data using weighted constraints on normal sensor data to obtain a self-verified target wavefront arrival time sequence. A fault location output unit calculates and outputs the fault location based on the target wavefront arrival time sequence, cable length parameters, and traveling wave propagation speed. The cable length parameters include the position parameters of the at least two cross-interconnected grounding boxes on the target cable line and / or the segment length parameters of adjacent cable segments, used to map the target wavefront arrival time sequence to the fault location.

[0008] Thirdly, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of the cable fault anomaly data self-verification method based on multi-source traveling wave comparison according to any embodiment of the present application.

[0009] Fourthly, embodiments of this application provide a storage medium storing a computer program thereon, characterized in that, when the program is executed by a processor, it implements the steps of the cable fault anomaly data self-verification method based on multi-source traveling wave comparison according to any embodiment of this application.

[0010] Fifthly, embodiments of this application provide a computer program product, including a computer program / instructions, which, when executed by a processor, implement the steps of the cable fault anomaly data self-verification method based on multi-source traveling wave comparison according to any embodiment of this application.

[0011] The cable fault anomaly data self-verification method and system based on multi-source traveling wave comparison provided in this application can achieve at least the following technical effects: (1) The multi-point simultaneous acquisition and relative timing alignment mechanism for non-ideal sampling conditions in engineering enables traveling wave data across cross-interconnected grounding boxes and multi-phase sheaths to remain comparable and fusionable without relying on a unified absolute timing reference. Since different acquisition points complete acquisition and buffering within the same fault triggering window, a time overlap interval that can be used for cross-correlation analysis is naturally formed, and the timing correlation between multi-source signals can be established through relative time difference. Thus, the estimation of the arrival time of the traveling wave is no longer blocked by the engineering reality of "independent sampling clocks and asynchronous existence", but is transformed into a calculable and calibrable relative constraint, so that subsequent location inference has a stable time reference basis and reproducibility.

[0012] (2) On the one hand, by constructing a relative time difference matrix and a wavefront neighborhood energy ratio matrix, the signal correlation is not only dependent on a single waveform shape, but is simultaneously constrained by both time consistency and energy statistical characteristics, thereby improving the stability of wavefront correlation. On the other hand, by introducing normalized weights based on signal-to-noise ratio and energy consistency calculations, data quality is explicitly projected into the solution process, making high-confidence channels contribute more to the estimation results and suppressing the influence of low-confidence channels. Based on this, an abnormal channel deviating from the statistical distribution is identified through a weighted iterative time estimation model, and the abnormal data is corrected using the weighted constraints formed by normal channels, ultimately obtaining a wavefront arrival time sequence that has been self-calibrated. Thus, automatic detection and automatic correction of factors such as occasional noise, local channel degradation, and sensor anomalies can be achieved, thereby significantly reducing the transmission and amplification of abnormal data on the positioning results.

[0013] This technical solution reconstructs multi-point, multi-phase traveling wave observation from a "ranging problem dependent on absolute synchronization" to a "multi-source consistency constraint problem based on relative time difference," and automatically suppresses and corrects anomalous data through reliability weighting and iterative self-verification. Therefore, the output fault location not only originates from the mapping calculation of propagation speed and line parameters, but is also based on the arrival time sequence that has undergone consistency verification and self-calibration, thus enabling the location results to have higher stability, reliability, and repeatability under complex engineering conditions. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 A flowchart is shown as an example of a cable fault anomaly data self-verification method based on multi-source traveling wave comparison according to an embodiment of this application; Figure 2 A schematic diagram illustrating the operational mechanism of an example of a cable fault anomaly data self-verification method based on multi-source traveling wave comparison according to an embodiment of this application is shown. Figure 3 A schematic diagram showing the simulation results comparing the fault location accuracy of different methods under dynamic signal-to-noise ratio interference environment is presented; Figure 4 The figure shows a simulation comparison of the impact of clock synchronization error on fault location accuracy in different methods; Figure 5 This is a schematic diagram of the fault location convergence process in the presence of abnormal sensor data, provided in an embodiment of this application. Figure 6 This is a visualized heatmap of the sensor reliability weight matrix provided in the embodiments of this application; Figure 7 A structural block diagram of an example of a cable fault anomaly data self-verification system based on multi-source traveling wave comparison according to an embodiment of this application is shown. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] It should be noted that, regarding the difficulty of identifying reflected waves in single-ended traveling wave ranging, some improved solutions have emerged in related technologies. For example, some studies have proposed injecting high-amplitude narrow pulses into the line and detecting the returned waveform, using adaptive filtering to compare the differences between normal and fault waveforms for location; other scholars have used mathematical morphology or wavelet packet processing methods to extract the maxima of the traveling wave front mode through multi-resolution morphological gradient transformation to determine the fault section; still other solutions utilize the difference in propagation speed between zero-mode and line-mode traveling wave components for ranging. However, when facing complex cable networks, the fundamental problem of distinguishing between reflected waves from fault points and reflected waves from points of impedance discontinuity remains unsolved.

[0018] For two- and multi-terminal traveling wave ranging, in addition to the aforementioned reliance on clock synchronization, its accuracy is also highly sensitive to inherent differences in line parameters and hardware systems. Measurement errors in line length and time differences in the hardware of primary and secondary equipment directly introduce ranging deviations. Furthermore, for multi-terminal network positioning in distribution networks or urban cables, although some improved methods based on multi-terminal information or specific waveform types have emerged, they still have limitations in solving reflected wave interference and eliminating false fault points in practical applications.

[0019] In the monitoring of cross-connected grounding systems, utilizing sheath circulating current information for auxiliary location is one of the current research directions. Current monitoring methods typically use the amplitude ratio of the sheath circulating current to determine the faulty phase; however, this indicator may show the same value at multiple points within a specific range, making it impossible to pinpoint the specific fault location. Some studies have proposed comparing the difference in the rate of change of power frequency current at different grounding boxes, or using the parameter differences between ideal and actual sheath currents for judgment; however, these methods usually focus on power frequency quantity analysis, only achieving fault section location, and are difficult to achieve precise location. Furthermore, current intelligent grounding box devices mostly operate independently, lacking comprehensive diagnostic capabilities based on multi-source data fusion.

[0020] It should be understood that the above description of the relevant technologies is intended only to help the public better understand the inventive spirit and motivation of this application, and is not intended to limit this application. Furthermore, the technical solutions described in the above-mentioned relevant technologies are not prior art, and may also be undisclosed technical solutions, such as those under research or in the laboratory stage.

[0021] The technical solutions in this application, including the collection, storage, use, processing, transmission, provision, and disclosure of users' personal information, comply with relevant laws and regulations and do not violate public order and good morals.

[0022] Figure 1A flowchart illustrating an example of a cable fault anomaly data self-verification method based on multi-source traveling wave comparison according to an embodiment of this application is shown.

[0023] Regarding the execution subject of the method in the embodiments of this application, it can be any controller or processor with computing or processing capabilities, such as a cable integrated online monitoring platform controller, an edge computing gateway, or a grounding box-side intelligent acquisition terminal controller. It performs preprocessing, cross-correlation comparison, reliability weighted fusion, and anomaly self-checking iteration on the multi-phase sheath traveling wave signals collected by the multi-source sensor array, and outputs the fault location result by combining the cable line parameters and the traveling wave propagation speed.

[0024] In some examples, the execution entity can be integrated and configured in electronic devices or terminals through software, hardware, or a combination of both, and can be deployed in a distributed manner: for example, the acquisition terminal at each cross-interconnected grounding box completes the acquisition of traveling wave signals and local caching / preliminary processing, while the host computer or cloud platform completes multi-source comparison and fusion, abnormal data self-verification, and location solution; and the types of terminals or electronic devices can include, but are not limited to, intelligent grounding boxes, fault location devices, substation monitoring hosts, distribution network automation terminals, industrial computers, or servers.

[0025] like Figure 1 As shown, in step S110, in response to the same fault event occurring in the target cable line, the multiphase sheath traveling wave signal collected by the multi-source sensor array at at least two cross-interconnected grounding boxes laid along the target cable line is acquired.

[0026] Here, the sampling clocks of each sensor in the multi-source sensor array are independent and time asynchrony is allowed. For the same fault event, at least two cross-connected grounding boxes complete the acquisition and buffering of the multiphase sheath traveling wave signal within the same fault trigger acquisition window, so that the multiphase sheath traveling wave signals acquired at different cross-connected grounding boxes have an overlapping interval on the time axis that can be used for cross-correlation analysis.

[0027] It should be understood that the types of fault events can be diverse, including but not limited to cable insulation breakdown, transient discharge or short circuit caused by joint / terminal defects, single-phase grounding faults, phase-to-phase short circuit faults, abnormal discharge of sheath / metal shielding layer to ground, and transient faults caused by external force damage. All of these can excite observable transient traveling wave signals in the cable sheath circuit. "Same fault event" refers to the transient traveling wave process caused by the same fault source that starts at the same time, and the signals collected at each cross-interconnected grounding box meet the temporal adjacency and wavefront shape consistency within the preset trigger window (such as wavefront polarity, main frequency bandwidth, or energy distribution characteristics meeting preset consistency conditions), thereby merging the multi-point observation data of the same fault source into the data set of the same fault event.

[0028] In some implementations, after the same fault event is triggered, the system simultaneously initiates the acquisition process of a multi-source sensor array at at least two cross-connected grounding boxes along the target cable line. The multi-source sensor array may include a high-frequency current sensor (e.g., a Rogowski coil) for acquiring the sheath traveling wave transients, a sheath voltage sampling unit (e.g., a high-frequency voltage divider / coupling unit), and necessary anti-aliasing filtering and front-end isolation circuitry; for the three-phase sheath, the traveling wave signals of the A / B / C phase sheaths, such as the traveling wave current / voltage, are acquired separately.

[0029] In addition, to ensure the consistency of the "same fault event", the acquisition trigger can be triggered by the local sheath transient change criterion (such as sheath current change threshold, short-term energy surge, differential peak exceeding limit), and an additional "event fingerprint" consistency check (such as the polarity of the first wavefront, the main frequency energy bandwidth, and the ratio of the shape of the first peak to the second peak) can be added, thereby avoiding different grounding boxes from mistaking different disturbances as the same event and participating in subsequent comparisons.

[0030] It should be noted that in engineering implementation, the sampling clocks of each grounding box are independent of each other, allowing for unknown deviations and drifts; therefore, this embodiment does not pursue absolute time synchronization across points, but instead adopts a "pre-trigger + post-trigger" window caching mechanism at each sampling point.

[0031] For example, a local circular buffer continuously rolls through the cache. When a trigger occurs, a signal segment containing a preset duration before and after the trigger is frozen and written to local storage (or edge computing unit memory). By appropriately setting the acquisition window (such as covering the critical periods of wavefront arrival and early reflection), sufficient overlap in the time axis between signals acquired from different grounding boxes can be ensured. Thus, even in the absence of high-precision time synchronization and high-speed synchronization links, multi-point data can still enter a comparable and fused processing channel under the constraint of "the same event, the same window".

[0032] In step S120, transient features are extracted from the multiphase sheath traveling wave signal to obtain a preprocessed signal sequence containing traveling wave front features.

[0033] It should be noted that the acquired multiphase sheath traveling wave signal usually contains power frequency components, switching noise, partial discharge spikes, ringing introduced by interconnect structures, and amplitude-frequency distortion caused by differences in sensor bandwidth. Therefore, transient feature extraction is required before entering cross-correlation to form a preprocessed signal sequence containing wavefront features.

[0034] In some implementations, power frequency removal and bandpass filtering are first performed (e.g., preserving energy within the effective frequency band of the traveling wave, suppressing low-frequency power frequency and high-frequency random noise), combined with amplitude normalization or gain correction to reduce the impact of differences in sensor sensitivity on correlation. Subsequently, the wavefront is enhanced, for example by employing short-time energy / envelope extraction, continuous wavelet transform, or differential operators to highlight abrupt changes, resulting in a sharper, alignable feature in the time domain. Furthermore, to avoid over-processing causing wavefront timing shifts, wavefront enhancement should employ a processing chain with minimal phase distortion (such as linear-phase FIR filtering or symmetrical window wavelets), and the processing parameters (filter bandwidth, window length, threshold) should be fixed to the device configuration or adaptively adjusted by the event intensity.

[0035] More specifically, the preprocessing output can contain two types of information: first, a wavefront enhancement sequence for cross-correlation (e.g., a normalized transient edge sequence or envelope sequence); and second, a wavefront neighborhood energy description for subsequent energy calculations (e.g., an energy integral sequence near candidate wavefronts). This significantly improves the wavefront's identifiability against a noisy background, making cross-correlation peaks more concentrated and relative time difference estimation more stable. Furthermore, it makes the outputs from different grounding boxes, different phases, and different sensors more statistically comparable.

[0036] In step S130, cross-correlation analysis is performed on any two signals in the preprocessed signal sequence to calculate the relative time difference between each sensor pair at different positions or phases, and the energy ratio between each sensor pair is calculated based on the energy characteristics of the preprocessed signal sequence in the wavefront neighborhood, thereby constructing the relative time difference matrix and energy ratio matrix of the multi-source signals.

[0037] In some implementations, the cross-correlation search range can be defined first (e.g., a reasonable delay interval around the trigger) to avoid spurious peaks introduced by long-distance reflections or non-fault disturbances; the delay corresponding to the cross-correlation peak is the relative alignment of the two signals in the wavefront characteristics. If there are differences in sampling rates at different acquisition points or insufficient time resolution, subsampling interpolation (such as parabolic fitting of the peak or refined estimation based on frequency domain phase) can be performed in the neighborhood of the cross-correlation peak to obtain a more refined relative time difference estimate. By traversing sensor pairs at different locations and with different phases, a relative time difference matrix can be formed, whose elements express the relative arrival time difference constraints between any two channels.

[0038] In addition, energy features are calculated in the neighborhood of the wavefront and an energy ratio matrix is ​​formed. Specifically, a neighborhood window of fixed or adaptive length can be extracted near the candidate wavefront of each signal (e.g., centered on the wavefront time or the start of the wavefront). The energy metric within the window (such as the sum of squares integral, envelope energy integral, or multi-scale energy aggregation) is calculated. Then, the ratio of the energy metrics of any two channels is obtained to obtain the energy ratio matrix.

[0039] Thus, the relative time difference matrix provides evidence of "time consistency", while the energy ratio matrix provides evidence of "amplitude consistency / propagation attenuation consistency". The combination of these two pieces of evidence constitutes the dual-constraint basis for the correlation of multi-source signals, enabling subsequent fusion solutions to not depend on a single waveform shape.

[0040] In step S140, based on the signal-to-noise ratio index and energy ratio matrix of each sensor, the normalized weights characterizing the reliability of the signal correlation between each sensor pair are calculated, and a reliability weight matrix is ​​generated.

[0041] It should be noted that since the coupling strength, background noise, and sensor status of the sheath traveling wave may differ at different grounding boxes, differentiated confidence levels can be assigned to the correlation results of each sensor pair. Here, based on the signal-to-noise ratio (SNR) index and energy ratio matrix of each sensor, normalized weights characterizing "signal correlation confidence" are calculated, and a reliability weight matrix is ​​generated. The SNR index can be defined as the ratio or difference between the signal energy in the wavefront neighborhood and the noise neighborhood before the fault, or it can be defined by combining peak amplitude and noise standard deviation; the energy ratio matrix is ​​used to measure the degree of energy consistency between two channels. For example, the closer the energy ratio is to the expected range (or the closer it is to the group statistical center), the higher the corresponding confidence level.

[0042] In constructing the weights, an unnormalized confidence score can be generated for each pair of channels (e.g., a quality score obtained from a monotonic mapping of SNR, a consistency score obtained from energy ratio deviation, and fused in a multiplicative or additive manner). Then, normalization is performed to ensure that the weights in the same row / column or the entire matrix meet a controllable scale (avoiding numerical instability caused by excessively large individual weights). Simultaneously, to avoid systematic bias caused by overall degradation of a particular sensor, lower and upper bound constraints can be added to the weights, and the weight distribution can be smoothed. In this way, any subsequent time estimation based on matrix constraints will naturally favor high-quality, highly consistent channel pairs, suppressing disturbances to results from low SNR or energy-abnormal channel pairs, thereby improving the robustness and repeatability of the fusion estimation.

[0043] In step S150, a time estimation model based on weighted iteration is constructed using the relative time difference matrix and the reliability weight matrix. The estimated arrival time of each sensor is updated by iterative calculation. During the iteration process, abnormal sensor data that deviates from the statistical distribution is identified. The abnormal sensor data is corrected by the weighted constraints of the normal sensor data, and the target wavefront arrival time sequence after self-verification and calibration is obtained.

[0044] Here, a weighted iterative time estimation model is constructed using the relative time difference matrix and the reliability weight matrix. Its core is to use the unknown "arrival time of each sensor wavefront" as the variable to be estimated, so that the arrival time difference between any two channels conforms as much as possible to the calculated relative time difference, and the contribution of each constraint term is controlled by the weight matrix.

[0045] In some implementations, the problem can be formulated as a weighted consistency optimization: initialize arrival times (e.g., using a reference channel as time zero, and coarsely aligning the rest according to relative time differences), and then iteratively minimize the "weighted residual" (the residual being the deviation between the current estimated time difference and the relative time difference). The iteration can employ robust update strategies (such as weighted least squares iterative updates or damped gradient updates) and set convergence criteria (such as the magnitude of residual decrease or an upper limit on the number of iterations).

[0046] Furthermore, after each iteration, the distribution characteristics of the residuals related to each channel are statistically analyzed to identify channels that continuously deviate from the statistical center of the population (e.g., residuals significantly greater than the median absolute deviation threshold, or inconsistent performance across multiple high-weight constraints). These channels are marked as anomalous sensor data. Instead of directly "deleting" anomalous channels, they are corrected using weighted consistency constraints of normal channels. Specifically, the arrival times of anomalous channels are projected and calibrated using a consensus time base formed by a majority of high-confidence channels, bringing them back to an interpretable range consistent with the population. The final output is a self-calibrated target wavefront arrival time sequence. This transforms multi-source observations from raw time series potentially containing anomalies and noise into time series that have undergone consistency verification and automatic correction, significantly reducing the amplification and transmission of localization results from individual sensor degradation, sporadic interference, or local coupling anomalies.

[0047] In step S160, the location of the fault point is calculated and output based on the target wavefront arrival time sequence, cable line length parameters, and traveling wave propagation speed.

[0048] Here, the cable line length parameter includes the location parameters of at least two cross-connected grounding boxes on the target cable line (such as mileage coordinates along the line) and / or the segment length parameters of adjacent cable segments (applicable to segmented calibration, and line representation with joints / branches), to map the target wavefront arrival time sequence to the fault location. The traveling wave propagation speed can be obtained from the cable type, dielectric parameters, and historical calibration, or it can be calibrated during system commissioning through known events or artificially injected pulses.

[0049] Furthermore, to ensure consistency between the output results and the maintenance line log, the location mapping can be explicitly mapped into the line coordinate system / segmentation system. When only the location parameter is used, the output can be the mileage value along the line; when the segment length parameter is introduced, the output can be "the cable segment number + the distance within the segment", ensuring that the mapping result meets physical boundary constraints (such as being located between two grounding boxes or within a certain segment length). Thus, the cable location result is calculated based on the self-verified arrival time sequence, and the time dimension is robustly mapped to the location dimension usable by the line parameters, making the output both numerically stable and facilitating rapid location and verification during on-site maintenance.

[0050] Regarding the implementation details of obtaining the multiphase sheath traveling wave signal in step S110, in some examples of the embodiments of this application, a ring buffer queue is constructed in the acquisition terminal of each cross-interconnected grounding box to perform real-time cyclic overwriting and storage of the original sampling data output by the multi-source sensor array, and the buffer depth of the ring buffer queue is configured to at least cover the sum of the preset pre-trigger length and the preset post-trigger length.

[0051] In some implementations, to capture fleeting high-frequency transient traveling wave signals, the acquisition terminals (e.g., embedded systems based on FPGAs or high-performance DSPs) within each cross-connected grounding box enable a ring buffer queue based on the first-in-first-out (FIFO) principle in the underlying storage space. This queue does not wait for a fault trigger signal but operates normally, continuously receiving analog-to-digital conversion (ADC) data output from the multi-source sensor array at a set high sampling rate (e.g., above 10MHz), and overwriting the earliest written historical data in real time.

[0052] Regarding the physical storage depth of the circular buffer queue, its capacity strictly corresponds to the sum of the preset pre-trigger length and post-trigger length (and can also reserve a certain amount of redundancy margin), thus constructing a time sliding window at the hardware level that can always "go back to the past", solving the problem that traditional trigger-based acquisition may lose key wavefront information due to startup delay.

[0053] Then, the transient change detection logic is run using a local independent clock. When a signal amplitude change or rate of change exceeds a preset threshold is detected, a local trigger anchor point is generated in response to the same fault event. The local trigger anchor point is used to characterize the sampling time corresponding to the trigger determination.

[0054] While data is continuously cached, the acquisition terminal uses its local independent clock (without the need for external GPS timing signals) to run transient change detection logic in parallel. This logic monitors the characteristic changes of the input signal in real time and uses amplitude exceeding the limit criterion or a rate of change criterion based on morphological gradients to identify fault characteristics. Once the change in the current or voltage signal of any phase sheath exceeds the preset trigger threshold, the system determines that a fault event has occurred at the current moment and immediately locks the write pointer position of the current circular buffer queue, marking it as a local trigger anchor point. This anchor point logically represents the moment when the acquisition terminal "subjectively" perceives the fault occurrence. It is entirely dependent on the local time base, thereby physically severing the dependence on the unified absolute clock of the entire network and reducing the system's requirements for hardware synchronization accuracy.

[0055] Then, using the local trigger anchor point as the timing reference, the historical data segment containing the preset pre-trigger length is backtracked from the circular buffer queue, and the data segment continues to be collected until the subsequent data segment with the preset post-trigger length is covered. The combined data segment is used as the multiphase sheath traveling wave signal and local buffering is performed.

[0056] Here, using the generated local trigger anchor point as the timing zero point, the acquisition terminal performs specific memory read operations to assemble a complete traveling wave data packet. Specifically, the system first controls the read pointer to trace back from the anchor point position, reading a historical data segment with a length equal to the "preset pre-trigger length". This data not only contains the traveling wave front at the moment the fault occurs, but also the background noise data before the wave front arrives, providing a necessary benchmark for subsequent signal-to-noise ratio calculation and weight matrix construction. Subsequently, the system continues to acquire and record subsequent data segments with a length equal to the preset post-trigger length to capture the reflected wave and attenuation process of the traveling wave. The traced historical segment and the acquired subsequent segment are seamlessly spliced ​​in the time domain, thus forming the "multiphase sheath traveling wave signal" of this node. By adopting a "store first, retrieve later" mechanism, it is ensured that the integrity of the signal waveform is not affected by the difference in the sensitivity of the trigger criterion under asynchronous triggering conditions.

[0057] In some implementations, the pre-trigger length is set to be no less than the sum of the maximum traveling wave propagation delay between any two cross-connected grounding boxes in at least two cross-connected grounding boxes and the maximum allowable sampling clock deviation of the system, in order to tolerate the trigger timing deviation caused by different wavefront arrival times and asynchronous sampling clocks of each acquisition terminal, and to ensure that the signal segments acquired by each acquisition terminal contain the same faulty traveling wavefront and complete wavefront background noise on the physical time axis.

[0058] It should be noted that the pre-trigger length, a core parameter in this embodiment, is not arbitrarily selected, but configured to cover the sum of two physical constraints: "maximum traveling wave transmission delay" and "maximum sampling clock deviation." The maximum traveling wave transmission delay is determined by the furthest cable distance between any two cross-connected grounding boxes and the traveling wave velocity. This ensures that even if a fault occurs at the far end, the near end triggers first due to the wavefront arriving earlier, and the later-triggered far-end acquisition terminal can still retrieve the wavefront data that arrived earlier but was not yet covered from the buffer queue through a sufficiently long "backtracking" operation. The maximum sampling clock deviation is determined by the temperature drift characteristics and calibration period of the system crystal oscillator, used to tolerate time slippage generated by nodes after long-term operation. This configuration supports a "wide window fallback," meaning that as long as the fault traveling wave falls within this wide window, regardless of the absolute clock dispersion of each node, the signal segments acquired by each terminal will inevitably have an overlapping interval containing the same fault wavefront on the physical time axis. This lays a solid data foundation for subsequent clock error elimination through cross-correlation analysis.

[0059] Regarding the implementation details of signal preprocessing in step S120, in some examples of embodiments of this application, baseline drift correction is performed on the multiphase sheath traveling wave signal, and hybrid denoising processing is performed by combining adaptive median filtering and wavelet threshold transformation to retain the high-frequency transient components of the signal and generate a denoised traveling wave signal with a high signal-to-noise ratio; then, amplitude normalization processing (e.g., normalization to the [0, 1] interval or unit energy) is performed on the denoised traveling wave signal to eliminate the energy attenuation difference of the fault traveling wave at different propagation distances, so that the signals output by each sensor are comparable in the energy dimension.

[0060] In some implementations, after acquiring the original signal, the system first performs baseline drift correction by subtracting the long-term mean of the signal or using a high-pass filter to eliminate the DC component caused by sensor zero drift or low-frequency induced voltage. Subsequently, considering the complex electromagnetic environment within the cable trench, this embodiment innovatively employs a hybrid denoising strategy combining adaptive median filtering and wavelet threshold transformation. Specifically, an adaptive median filter can first be used to filter out impulse noise (salt-and-pepper noise) generated by high-voltage switching operations, and then wavelet transform (such as using a db4 wavelet basis for multi-level decomposition) can be used to perform multi-scale threshold processing on the signal to remove Gaussian white noise.

[0061] Then, the traveling wave signals of each phase in the denoised traveling wave signal are respectively denoised as follows: Using pre-defined structural elements For each Perform morphological decomposition and calculate morphological gradient waveforms To highlight the abrupt change edge features of the signal: Equation (1) In the formula, Represents morphological dilation operations. Represents morphological erosion operation. This indicates the phase index, used to distinguish different phase signals in a multiphase sheath traveling wave; For time variables, Indicates the first The time-domain amplitude sequence of the denoised traveling wave signal is used to characterize the transient response of the fault traveling wave in that phase. Indicates the first Morphological gradient waveform of a phase traveling wave signal.

[0062] Here, in order to extract the denoised traveling wave signal To accurately extract waveheads, this embodiment introduces a mathematical morphology processing method, utilizing preset structural elements. (For example, a flat structuring element with a width of 5-10 sampling points) performs morphological decomposition on the signal.

[0063] In equation (1), the dilation operation takes the local maximum value within the structuring element coverage window, which smooths the signal troughs and expands the peaks; while the erosion operation takes the local minimum value within the window, which smooths the peaks and expands the troughs. The morphological gradient obtained by subtracting the two is... Physically, this effectively eliminates the stable DC or slowly varying portion of the signal, transforming the edges of sharp abrupt changes in the signal (i.e., the instant the traveling wave front arrives) into significant unipolar pulse peaks. This significantly sharpens the wave front characteristics, making wave front timing identification no longer dependent on absolute amplitude thresholds, but rather on the signal's rate of change, thus adapting to fault signals with varying degrees of attenuation.

[0064] Then, in each morphological gradient waveform The modulus maxima are detected, and the sampling time corresponding to the modulus maxima is determined as the arrival time of the single-phase candidate wavefront of the corresponding phase.

[0065] After obtaining the morphological gradient waveform, the system locks the arrival time of the candidate wavefronts for each phase by searching for local maxima. However, single-phase signals are highly susceptible to partial discharge or random interference, leading to misjudgments. Therefore, this embodiment further utilizes the physical characteristic of "simultaneous propagation of multiphase signals": theoretically, for traveling waves generated by the same fault source, the induced currents on the three-phase sheaths should arrive at the same monitoring point almost simultaneously. The system calculates the deviation between the candidate times of phases A, B, and C. If the time difference between the triggering time of a phase and other phases exceeds a preset simultaneity threshold (e.g., 2 microseconds, which depends on the inter-phase coupling delay), the point is determined to be isolated interference and eliminated; otherwise, the time that passes the verification is retained. This significantly reduces the single-phase false triggering rate and ensures that the wavefront times participating in subsequent positioning calculations have a clear physical correlation with the fault.

[0066] Furthermore, at the same cross-connection grounding box, utilizing the simultaneity characteristics of multiphase signal propagation, the time deviation between the arrival times of the single-phase candidate wavefronts of each phase is calculated. Isolated abrupt changes exceeding a preset simultaneity threshold are eliminated, and the time that passes the consistency check is retained as the final candidate wavefront arrival time. Centered on the final candidate wavefront arrival time, a wavefront neighborhood window of a preset length is extracted from the denoised traveling wave signal, and the denoised traveling wave signal, morphological gradient waveform, final candidate wavefront arrival time, and wavefront neighborhood window together constitute a preprocessed signal sequence.

[0067] More specifically, to reduce the amount of data required for subsequent cross-correlation calculations and eliminate interference from noise in non-faulty sections, the system uses the arrival time of the final candidate wavefront that has passed consistency verification as the center, and extracts a fixed-length time window (e.g., from 200 sampling points before the wavefront to 800 sampling points after the wavefront) from the original high signal-to-noise ratio denoised traveling wave signal, defining it as the wavefront neighborhood window. Data segments within this window, their corresponding morphological gradient waveforms, and the determined arrival time labels are collectively packaged into a preprocessed signal sequence. This structured data processing method not only provides a standardized energy statistics interval for subsequent energy ratio-based weight calculations but also provides precisely aligned regions of interest for multi-source cross-correlation analysis, thereby significantly improving the algorithm's efficiency and the convergence speed of the localization model.

[0068] Regarding the implementation details of constructing the relative time difference matrix and energy ratio matrix of the multi-source signals in step S130, in some examples of the embodiments of this application, for any two sensors in the sensor set... and Denoising traveling wave signal segments corresponding to the wavefront neighborhood window are extracted from the preprocessed signal sequence and denoised as follows: and .

[0069] Here, based on the wavefront neighborhood window determined in the previous steps, the corresponding denoised traveling wave signal segments are precisely extracted from the preprocessed signal sequence and labeled as follows: and It is not a simple copy, but rather it strictly locks the analysis range of the signal to the core section containing the transient energy of the fault, and discards the steady-state components or redundant background noise outside the window.

[0070] Then, in the time span corresponding to the wavefront neighborhood window Internally, based on and Construct cross-correlation function from overlapping intervals : Equation (2) In the formula, This represents a time-lag variable.

[0071] In order to accurately calculate the time difference between the two signals, this embodiment constructs a cross-correlation function. In equation (2), the upper limit of integration is set to... Instead of a fixed window length This means that integration is performed only within the physically overlapping region of the two signal slides. This effectively avoids issues arising from the lag time. When the value is large, human error is introduced due to zero padding when the window slides out of the boundary.

[0072] Then, within the preset search range, search for the lag time that maximizes the cross-correlation function. and will Determined to be a sensor With sensors The fine relative time difference between them is used to construct a relative time difference matrix by traversing all sensor combinations. .

[0073] By searching within the preset search range Reaching the global maximum value The system can obtain the relative time difference of sub-sampling point accuracy, thereby eliminating the influence of the absolute clock asynchrony of each sensor and realizing "relative synchronization" by utilizing waveform similarity, providing core data support for asynchronous positioning.

[0074] Then, the transient energy of each denoised traveling wave signal segment within the wavefront neighborhood window was calculated using the square integral method. and build sensors With sensors Energy ratio : Equation (3) In the formula, For sensor index placeholders, take or ; Indicates sensor Transient energy within the wavefront neighborhood window, Indicates the corresponding sensor The denoised traveling wave signal segment within the wavefront neighborhood window.

[0075] It should be noted that real traveling wave signals inevitably carry significant energy, while simple noise interference, although it may have a high amplitude (such as pulse interference), has a short duration and usually low total energy. Equation (3) describes the calculation logic of transient energy, which is to integrate and accumulate the square of the signal amplitude within the wavefront neighborhood window. Compared to a single amplitude peak, the energy... It comprehensively reflects the overall impact intensity of the fault traveling wave over a period of time and has better noise immunity.

[0076] Furthermore, when When the energy exceeds the preset lower limit, the sensor... With sensors Energy ratio and energy ratio Assembled into an energy ratio matrix; when When the energy level is not greater than the preset lower limit, the corresponding sensor pair will be marked as low confidence and will not participate in the construction of the energy ratio matrix; the energy ratio matrix is ​​used to quantitatively characterize the consistency of energy attenuation of the same fault traveling wave on different propagation paths and phases.

[0077] Here, based on the calculated energy values, the system constructs an energy ratio matrix to characterize the consistency of signal attenuation. To prevent logical errors and numerical instability, this embodiment introduces a strict "divide-by-zero" mechanism: only when the sensor energy used as the denominator is zero... The energy ratio is only calculated when the energy level exceeds the preset lower limit (i.e., the effective signal threshold). And fill in the matrix; conversely, if If the value is too small, it indicates that the sensor may be in a blind zone or malfunctioning. The system directly marks this sensor pair as low confidence and excludes it from matrix construction. This prevents irrelevant data from entering the subsequent weight calculation process at the data source, ensuring that the energy ratio matrix accurately reflects the physical attenuation of the faulty traveling wave along different propagation paths (i.e.,...). (It fluctuates around the theoretical decay value), thus providing a high-confidence reference feature for subsequent self-verification.

[0078] Regarding the implementation details of generating the reliability weight matrix in step S140, in some examples of embodiments of this application, for each sensor in the sensor set... Based on the background noise data segment located before the wavefront neighborhood window in the preprocessed signal sequence, the standard deviation of the background noise of the sensor is calculated. A preset lower limit for the standard deviation of background noise is set. To obtain the equivalent noise amount used for weight calculation. .

[0079] Here, before calculating the weights, the system first performs a benchmark assessment of the signal-to-noise ratio (SNR) of the sensor data. Specifically, the system backtracks the pure background noise data segment in the preprocessed signal sequence that lies before the wavefront neighborhood window and calculates its standard deviation. To prevent issues caused by a sensor being in a silent state or data acquisition being stuck. The noise level approaches zero, which can lead to a numerical explosion (divide-by-zero anomaly) during weight calculation. This embodiment introduces a preset noise lower limit. (For example, the value is taken as twice the ADC quantization noise), and then the equivalent noise amount is calculated, which gives the algorithm strong engineering robustness and ensures that the weight calculation model still has mathematical stability even in an extremely ideal silent environment.

[0080] Then, the transient energy of each denoised traveling wave signal segment is retrieved. , and the energy ratio of each sensor pair For any two different sensors and In order to satisfy and All are greater than the preset lower energy limit Under the condition of constructing the original reliability weights Computational model: Equation (4) In the formula, Indicates unnormalized sensor pairs The original reliability weights, and This is the preset adjustment coefficient.

[0081] Here, a weight calculation model integrating multi-dimensional features is constructed. Specifically, in equation (4), it is not a simple linear superposition, but rather a product of three physical constraint terms:

[0082] (1) Signal-to-noise ratio coupling term By utilizing the inverse relationship of noise, low-noise (high signal-to-noise ratio) sensors can be directly assigned higher base weights. (2) Energy symmetry term This term is essentially the ratio of the geometric mean to the arithmetic mean, if and only if The maximum value of 1 is obtained at a time to punish sensor pairs with significant energy differences (physically, the energy of the same fault traveling wave should be similar at adjacent nodes), and to exclude the case of abnormal amplification of single-ended signals. (3) Ratio Consistency Item Using the exponential decay function, the deviation from the theoretical energy ratio ( The signal will be severely punished.

[0083] Through the combined effect of the above three factors, only when two sensors simultaneously meet the criteria of "low noise", "equivalent energy" and "conforming to the attenuation law" can they obtain high weight, thereby achieving accurate screening of high-quality data sources.

[0084] It should be noted that when or Not greater than the preset lower energy limit At that time, the corresponding sensor pair is marked as low confidence and then... Specifically, the system has strict admission criteria, requiring that the two sensors involved in the calculation have sufficient energy. and All must be greater than the preset lower energy limit If the energy of any sensor falls below this threshold, it indicates that it has failed to effectively capture the traveling wave signal (possibly due to a dead zone upstream of the fault or a sensor malfunction). The system then directly applies the corresponding original weight. The value is set to 0, thus constructing a "hard threshold cutoff" mechanism to prevent weak noise disturbances from being erroneously amplified by Equation (4) and to ensure that the final weight matrix only reflects the correlation between valid fault signals.

[0085] Then, normalization is performed on the original reliability weights to make them suitable for any target sensor. The sum of its association weights with the other sensors satisfies The normalization calculation formula is: Equation (5) In the formula, For the sensor traversal variables.

[0086] Furthermore, the normalized reliability weights are... Combine them to generate a reliability weight matrix. .

[0087] Here, after obtaining the original weights, the system further performs row normalization processing on them, targeting the sensor. All associated weights are transformed into "trust assignments" in a probabilistic sense, so that the sum of all associated weights is strictly equal to 1. This ensures that the subsequent iterative estimates remain stable in the numerical space and do not diverge or drift due to the accumulation of weights, thus ensuring that the self-verification algorithm can quickly converge to the true value.

[0088] Regarding the implementation details of iteratively updating the estimated wavefront arrival time in step S150, in some examples of embodiments of this application, the relative time difference is called. With reliability weight For any sensor pair in a sensor network, a physical constraint relationship is established based on the absolute time difference equaling the relative observation value, and a residual variable is defined to characterize the degree of violation of this constraint relationship. : Equation (6) In the formula, and The sensors to be estimated are respectively With sensors The absolute wavefront arrival time relative to the virtual reference time.

[0089] Here, the system targets each pair of sensors with a valid connection in the sensor network (i.e., weights). A physical consistency constraint based on "the absolute time difference should be equal to the relative observation value" is constructed, and the degree of violation of this constraint is quantified by equation (6), where This is the absolute time relative to the virtual reference frame that needs to be solved. Specifically, in an ideal, noise-free system, it is the estimated time difference between two points. It should be strictly equal to the time delay of the cross-correlation measurement. At this point, the residual The result is zero. Therefore, the complex distributed asynchronous measurement problem is transformed into a mathematical residual minimization problem, providing a clear quantitative indicator for subsequent optimal estimation.

[0090] Construct a global objective function based on minimizing weighted residuals, and utilize a robust loss function. To suppress the impact of large residual abnormalities on the overall estimate, a nonlinear mapping is applied to the residual variables. Equation (7) In the formula, express The robust loss function value is used as the residual penalty term.

[0091] Here, in order to solve for the optimal solution from noisy observation data... The system constructs a globally weighted optimization model. Equation (7) describes the solution objective of this model; specifically, it utilizes the weights calculated in the preceding steps. Weighting the residuals ensures that data with high signal-to-noise ratio and good energy matching dominates the estimation; simultaneously, a nonlinear robust loss function can be introduced. (For example, using the Huber loss function or the Bisquare function) instead of the traditional L2 squared loss. Thus, It can perform "soft truncation" or weight reduction mapping on large residual terms, thereby significantly suppressing the pulling effect of anomalous sensor data (outliers) on the overall solution, so that the algorithm can still obtain a robust global optimal solution even when some sensors are severely disturbed.

[0092] A reference constraint is introduced to fix the virtual reference time, ensuring that the arrival time of the absolute wavefront satisfies the zero-mean benchmark: Equation (8) By solving for the minimum value of the global objective function under the condition of satisfying the reference constraint, the estimation results of the absolute wavefront arrival time of each sensor relative to the virtual reference time are obtained.

[0093] It should be noted that, since the constraint relationship in equation (6) only involves the time difference, the system of equations mathematically suffers from "translation uncertainty" (i.e., all...). Simultaneously add a constant The residual value remains unchanged, leading to a non-unique solution. To address this rank deficiency problem, this embodiment introduces equation (8). The reference constraint shown forces that the sum of the estimated times of all sensors must be zero, which mathematically anchors the virtual reference time to the statistical centroid of the arrival times of all sensors. This provides a stable absolute reference benchmark for the iterative algorithm, eliminates numerical drift, and ensures that the iterative process can quickly converge to a unique definite solution, thus achieving a precise mapping from "relative time" to "absolute time".

[0094] Regarding the implementation details of correcting and obtaining the target wavefront arrival time sequence in step S150, in some examples of embodiments of this application, for any target sensor Determine its associated sensor set ;in, Includes sensors There is an effective relative time difference And corresponding reliability weight Sensors .

[0095] Perform an iterative update step based on a weighted average, utilizing the current estimated arrival times of each associated sensor. relative time difference Estimate the target sensor The next round of absolute arrival time : Equation (9) exist Not greater than the preset minimum weight and threshold At that time, keep Or the sensor Marked as low credibility and skipped in this round of updates.

[0096] Here, after establishing the global optimization objective, the system enters the core iterative solution phase. For any target sensor in the network... The system first dynamically selects its associated sensor set. It only accepts "reliable neighbors" with valid observations and non-zero weights.

[0097] Equation (9) reveals the physical meaning of single-step iteration, which describes the "weighted multi-source voting" process. Specifically, Represents neighbor sensors Based on its current state and relative observations, the target sensor The "inference" is made based on the arrival time; and by using a weighted average, the inferences from all neighbors are fused into the target sensor's inference. The next round of estimates Furthermore, to prevent numerical instability caused by excessively small neighbor weights (i.e., isolated nodes), the denominator of equation (9) is less than a threshold. This triggers protection logic (skipping updates). Thus, by utilizing the redundancy of information across the entire network, random measurement noise at a single point is suppressed, allowing the estimated value to gradually approach the true physical value.

[0098] Then, after each iteration update, zero-mean processing is performed on the estimated arrival times of all participating sensors to satisfy the reference constraints; the iterative update step is started with the initial assumption values ​​until the change in the estimated arrival times of all participating sensors is less than the preset convergence threshold or the preset upper limit of the number of iterations is reached.

[0099] Here, after each iteration update is completed, the system must immediately update all... Perform zero-mean processing (i.e.) ), to force the satisfaction of the reference constraint in equation (8); This represents the arithmetic mean of the estimated wavefront arrival times of all participating sensors in the current iteration. Mathematically, this eliminates the rank deficiency of the equation system, preventing the solution from drifting across the entire time axis; physically, it establishes a stable virtual reference frame. The system continuously monitors the norm of change of the estimates between consecutive iterations, and determines iterative convergence when the changes at all nodes are below a preset convergence threshold (e.g., 1 μs). This ensures that the algorithm can stably converge to the global optimum within a finite number of steps (usually less than 20).

[0100] Then, after iterative convergence, based on the relative time difference... Calculate the weighted residual scale for each sensor to identify anomalous sensor data: Equation (10) In the formula, Indicates sensor The weighted residual scale.

[0101] when Greater than the preset anomaly detection threshold At that time, the corresponding sensor The sensor was determined to be faulty.

[0102] It should be noted that iterative convergence does not necessarily mean the result is correct, as there may be anomalous sensors with systematic biases (such as crystal oscillators with severe frequency deviations). Therefore, the system calculates the weighted residual scale for each sensor. In equation (10), if a sensor It's normal, so the absolute moment of its eventual convergence... with neighbors The moment The difference should be compared with the measured value. A high degree of agreement. Therefore, This actually quantifies the incompatibility between the sensor and the entire network, when Exceeding the preset threshold When this happens, the system determines that the sensor is an abnormal node. This enables a deep diagnostic process, moving from "statistical outliers" to "physical violations," accurately identifying faulty data sources that, while participating in iterations, consistently fail to integrate into the overall consistency.

[0103] Then, a self-checking and correction mechanism is activated: for abnormal sensors. This is done by temporarily removing the sensor from the set of sensors participating in the iteration, resulting in the normal set of sensors. and using only a normal sensor set Abnormal sensors The arrival time is reconstructed: Equation (11) In the formula, Represents a normal sensor set Any normal sensor number in the system, Indicating abnormal sensor The corrected estimated wavefront arrival time. Indicates sensor With sensors The relative time difference between the observations, Indicates sensor pair Reliability weight.

[0104] when Not greater than the preset minimum weight and threshold At that time, adopt The weighted median or median as .

[0105] Furthermore, using the revised Replace the original anomaly estimate and write it back to the wavefront arrival time sequence, then re-trigger iterative updates and zero-mean processing until all anomaly sensors are eliminated or the preset self-checking iteration limit is reached.

[0106] Specifically, once the abnormal sensor is locked The system immediately initiates self-checking and correction logic, removing it from the trusted set and utilizing only the remaining set of normal sensors. Reconstruct it. In equation (11), since If your own readings are unreliable, then let all your reliable neighbors do the work. Utilize their reliable moments and relative observations To reverse the process The system should have the theoretical arrival time. Furthermore, if the effective neighbor weights are insufficient, the system uses the more robust median algorithm as a fallback. Finally, the original outliers are replaced with corrected values, and the iteration is restarted until no more outliers are detected in the network. This gives the system self-healing capabilities; even if some sensors fail, a high-precision arrival time sequence can be recovered through data reconstruction, ensuring the reliability of the final fault location.

[0107] Regarding the implementation details of step S160, in some examples of embodiments of this application, environmental monitoring parameters associated with the target cable line are obtained, and based on the propagation characteristic model of the cable insulation material, the nominal traveling wave propagation speed is compensated and corrected using the environmental monitoring parameters to obtain the actual traveling wave propagation speed under the current operating conditions. Environmental monitoring parameters include at least one of the following: temperature, humidity, soil resistivity, and sheath grounding status parameters.

[0108] It should be noted that, because the dielectric constant of high-voltage cable insulation is affected by environmental factors such as temperature and humidity, using the nominal wave velocity will lead to significant positioning errors. Therefore, this embodiment uses an environmental monitoring module deployed at the cross-connection grounding box to collect real-time data on temperature, humidity, soil resistivity, and sheath grounding status parameters within the cable trench or tunnel.

[0109] For example, the system utilizes a pre-defined propagation characteristic model of the insulating material (e.g., a dielectric constant fitting curve based on the temperature change characteristics of XLPE material) to dynamically compensate and correct the nominal traveling wave velocity. This eliminates systematic errors introduced by changes in environmental conditions, ensuring the accuracy of the wave velocity parameters subsequently substituted into the positioning equation. It is a real-time truth value that is highly consistent with the current physical environment.

[0110] Then, the cross-connection grounding boxes included in the cable line length parameters are... The location parameters are represented as line mileage coordinates. And represent the location of the fault point as The reference point at the time of the fault occurrence is represented as The target wavefront arrival time sequence is compared with the mileage coordinates. The corresponding calibration wavefront arrival time is denoted as ,based on Construct a system of positioning equations: Equation (12) Here, after determining the wave speed Then, the system maps the cable topology to a one-dimensional coordinate system for each sensor node or cross-connection grounding box. Based on its arrival time after self-calibration Construct the positioning equation (12). Specifically, the node Capture the moment of wave head , equal to the time when the fault occurred Add traveling wave from the fault point propagation to sensor The required time (distance divided by velocity) is thus established. This establishes an analytical relationship connecting the time observation domain and the spatial location domain, providing a foundational model for subsequent mathematical solutions.

[0111] Then, based on the reliability weight matrix And the set of abnormal sensors, deriving the overall confidence level of the nodes corresponding to each cross-interconnected grounding box. Among them, for the node corresponding to the abnormal sensor, For non-abnormal nodes, Rather than The sum of the incident weights from non-abnormal nodes is positively correlated and normalized.

[0112] here, The description focuses on the relationship between "sensor pairs" (two-dimensional), while the localization equation is for "individual sensor nodes" (one-dimensional). To achieve this dimensionality transformation, this embodiment designs a node comprehensive confidence level. The derived algorithm: First, for sensors judged as "abnormal", directly set their confidence level. A "one-vote veto" is used to completely remove it from the location calculation; for "non-abnormal" nodes, their confidence level is... It is calculated as its value in the matrix. The sum of the incident weights from all other non-abnormal nodes is normalized. Therefore, the more reliable neighbors a node is "trusted" by (i.e., the better the cross-correlation and the more reasonable the energy ratio with its neighbors), the greater its voice in the final location vote, thus realizing the mapping from "relative reliability" to "absolute confidence".

[0113] Then, a weighted least squares objective function with comprehensive confidence constraints is constructed. : Equation (13) By solving the objective function Minimize parameters as well as This is to determine the final location of the fault and output it.

[0114] To solve for the optimal fault location from a set of overdetermined equations that may contain measurement noise, the system constructs a weighted least squares (WLS) objective function. Here, equation (13) defines the objective of this optimization: minimizing the "weighted sum of squared time difference residuals" of all sensor nodes involved in the calculation. Represents the first The deviation between the observation time and the theoretical time of each node, and This is then used as a weighting coefficient to adjust the contribution of this deviation to the overall objective. The system solves the problem using numerical optimization algorithms (such as the Gauss-Newton method or grid search method) to achieve this. Minimized and Therefore, through weighted fitting, the positioning results are automatically aligned with sensor nodes possessing "high confidence," while simultaneously... The joint solution eliminates the dependence on the absolute wave emission time, achieving high-precision positioning without the need for time synchronization.

[0115] Furthermore, based on the optimal residual value obtained from the solution and the proportion of abnormal sensors marked during the self-verification process, a reliability index for the positioning result is quantified and generated, and the location of the fault point is determined. Output in sync with reliability metrics.

[0116] Specifically, reliability indicators can be generated based on two dimensions: first, the minimum value of the objective function (optimal residual value), where a smaller residual indicates more physically consistent observation data from each sensor; and second, the proportion of abnormal sensors marked during the self-verification process, where a lower proportion indicates higher overall network health. For example, the system merges these two parameters into a score or confidence interval of 0-100, which is output synchronously with the fault location. This provides maintenance personnel with an intuitive basis for judgment, enabling them to distinguish between "high-deterministic location results" and "suspected results requiring further investigation," significantly improving the efficiency of maintenance decision-making.

[0117] Figure 2 The diagram illustrates an example of the operation mechanism of a self-verification method for cable fault anomaly data based on multi-source traveling wave comparison according to an embodiment of this application.

[0118] like Figure 2 As shown, this operational mechanism encompasses a closed-loop control process from data acquisition to result output. First, in the input phase, the system acquires high-frequency transient traveling wave signals in parallel using a multi-source sensor array, while simultaneously utilizing environmental sensors to obtain environmental parameters such as temperature and humidity. The data stream then enters the preprocessing phase, sequentially performing filtering and denoising, and morphological decomposition to extract clear wavefront features from complex background noise. The core processing phase is the logical center of this solution. It first calculates the relative time difference between nodes through multi-source cross-correlation, then uses a reliability weighting and anomaly detection module to identify "untrusted" nodes in the network, triggering a self-calibration mechanism to iteratively correct and reconstruct abnormal data using redundant information from normal nodes. Finally, after parameter correction and threshold adaptation (including wave velocity correction based on environmental parameters and dynamic matching of the algorithm's internal convergence threshold), the system synchronously outputs the corrected and accurate fault location results and their corresponding reliability indicators in the output phase, providing maintenance personnel with diagnostic data that combines accuracy and reliability.

[0119] To verify the effectiveness of the multi-source traveling wave comparison self-calibration method proposed in this application, a distributed parameter model of a 100 km long single-core high-voltage cable was constructed based on a simulation platform, and the nominal traveling wave propagation velocity was set to 190 m / µs. Six monitoring nodes were uniformly deployed along the cable line to simulate the non-contact multi-source sensor array layout at the cross-connection grounding box. Using the Monte Carlo method, 1000 fault events were randomly generated within a 10 km to 90 km range of the cable, covering low-resistance grounding, high-resistance grounding, and flashover faults to ensure the diversity and representativeness of the test samples.

[0120] To focus on examining the algorithm's robustness and desynchronization capability under harsh conditions, the experiment introduced multidimensional interference variables: on the one hand, Gaussian white noise was superimposed on the acquired signal, and the signal-to-noise ratio (SNR) was set in the range of -5 dB to 30 dB to simulate a strong noise environment; on the other hand, to address the asynchronous problem that is the core issue of this application, random time jitter was introduced into the sampling clock of each monitoring node. The standard deviation The time interval is varied between 0 and 10 µs to simulate extreme clock deviation scenarios caused by the absence of GPS timing or loss of satellite signal lock. Comparative experiments are conducted to verify the accuracy advantage of this method over the traditional two-end traveling wave method.

[0121] Figure 3 This diagram illustrates the simulation results comparing the fault location accuracy of different methods under dynamic signal-to-noise ratio (SNR) interference environments. The horizontal axis represents the SNR after adding random noise in the simulation environment, expressed in decibels (dB), with a range of -5dB to 30dB to simulate different operating conditions from strong to weak interference. The vertical axis represents the root mean square error (RMSE) of fault location obtained from multiple Monte Carlo simulations, expressed in meters (m). A smaller RMSE value indicates higher location accuracy.

[0122] exist Figure 3 In the text, "traditional dual-end method" is used as the baseline method, which adopts the error change trend of dual-end traveling wave positioning method based on single wavefront time identification; "the method in this paper" represents the error change trend of positioning method based on multi-source cross-correlation relative time difference and reliability weight iteration proposed in this application.

[0123] like Figure 3 As shown, with the decrease in signal-to-noise ratio (i.e., increased noise interference, shifting towards the negative horizontal axis), the positioning errors of both methods increase. However, under the same signal-to-noise ratio conditions, the root mean square error of the proposed method is significantly lower than that of the traditional two-end method. Especially in strong noise environments below 0 dB, the proposed method can still maintain a low error level, while the error of the traditional method increases sharply. This verifies that the proposed method has stronger anti-noise interference capability and higher positioning accuracy compared to the prior art.

[0124] Figure 4 The figure shows a simulation comparison of the impact of clock synchronization error on fault location accuracy in different methods. The horizontal axis in the figure represents the standard deviation of random jitter superimposed on the sampling clocks of each sensor. The unit is microseconds (μs), used to simulate asynchronous sampling scenarios in environments without a unified clock source (such as without GPS) or with unstable timing signals; the vertical axis represents the root mean square error (RMSE) of the fault location result, in meters (m).

[0125] like Figure 4 As shown, the positioning error of the baseline dual-end method increases linearly and sharply with the increase of clock jitter (for example, the error exceeds 600m with a 5μs jitter), indicating that the baseline method is highly dependent on absolute time synchronization and fails once it loses synchronization. In contrast, the method in this paper maintains near-perfect accuracy throughout the entire test range, with the error consistently kept at an extremely low level. This result intuitively confirms that the core relative time difference comparison and iterative self-verification algorithm of this application has a significant "desynchronization" capability, and can automatically correct errors through geometric constraints even when there are large random time differences between sensors, thereby achieving high-precision positioning without the need for external high-precision time synchronization.

[0126] Figure 5 This is a schematic diagram of the fault location convergence process under the condition of abnormal sensor data provided in the embodiments of this application. In the figure, the horizontal axis represents the number of iterations of the self-verification algorithm, and the vertical axis represents the estimated deviation of the fault location result relative to the actual location (unit: km). The simulation scenario is set such that one of the sensors (e.g., Sensor 3) is artificially injected with a time deviation of 20μs to simulate the scenario of the sensor being falsely triggered by strong interference or hardware failure.

[0127] like Figure 5 As shown, in the initial calculation stage (when the number of iterations is 0), the directly weighted positioning result has a large deviation (approximately 1.8 km) due to the influence of this abnormal data. However, as the iteration process progresses, the method of this application uses a multi-source cross-correlation matrix to detect that the consistency between this sensor and other nodes is extremely low, and then dynamically reduces the reliability weight of this sensor through algorithm logic. The curve shows that the positioning deviation decreases rapidly with the number of iterations, and finally converges and stabilizes near the actual fault point (error less than 50 m) after about 5 iterations. This result intuitively verifies that the abnormal data self-verification mechanism proposed in this application has excellent "self-immunity" capability, can automatically identify and isolate interference data, and can still ensure that the system outputs high-precision positioning results even when some sensors fail.

[0128] Figure 6 This is the sensor reliability weight matrix provided in the embodiments of this application ( This is a visualization heatmap showing the distribution of cross-correlation reliability weights among sensor nodes under a simulated high-resistivity fault scenario. The horizontal and vertical axes represent sensor numbers (index 0-5), and the brightness of the color represents the magnitude of the weight value. Brighter colors (approaching yellow / 1.0) indicate a more reasonable signal correlation and energy ratio between the two sensors, resulting in higher confidence; darker colors (approaching deep purple / 0.0) indicate lower confidence.

[0129] like Figure 6 As shown, the heatmap exhibits two significant characteristics: First, the cross-weighted region between the sensors closest to the fault point (such as Sensor 2 and Sensor 3) is the brightest, indicating that the algorithm correctly identified the high-quality signal source based on the energy decay law and assigned it a high computational weight. Second, for Sensor 4, which was simulated to have a fault, its corresponding 4th row and 4th column appear as a dark area in the graph (weight close to 0), forming a clear "cross" dark zone. This intuitively verifies that the self-verification mechanism based on multi-source comparison proposed in this application can sensitively detect abnormal nodes and effectively "isolate" them from the positioning calculation by automatically reducing their weight, thereby preventing the fault or interference of a single sensor from contaminating the overall positioning result and ensuring the robustness of the system.

[0130] Simulation results strongly validate the significant advantages of the proposed multi-source traveling wave ratio self-calibration method under complex working conditions, particularly its complete resolution of the dependence of traditional technologies on high-precision time synchronization and signal quality. Firstly, this method exhibits strong environmental adaptability: in environments with low signal-to-noise ratios (e.g., 0dB) and strong interference, random noise is effectively suppressed through multi-source cross-correlation fusion, resulting in a significantly lower positioning error compared to the traditional dual-end method. More importantly, experiments demonstrate that this method possesses a "desynchronization" characteristic. Even in extreme scenarios where the sensor sampling clock experiences significant random jitter (e.g., 10μs) and there is no external GPS time synchronization, the algorithm can still maintain stable positioning accuracy due to the geometric constraints of the relative time difference matrix. This overcomes the industry pain point of accurate distance measurement for high-voltage cables in underground tunnels and other areas without satellite signals.

[0131] Furthermore, the experiments visually demonstrate the algorithm's unique intrinsic self-verification and self-healing mechanisms. By monitoring the iteration process and the weight matrix distribution, it can be seen that when sensor malfunctions or data anomalies occur in the system, the algorithm can intelligently identify "dirty data" based on the laws of physical energy propagation and effectively isolate faulty nodes from the computing network by dynamically reducing reliability weights. This automatic correction capability based on iterative convergence not only ensures the system has extremely high fault tolerance when operating without changing the wiring method of the cross-connection cables, but also avoids the risk of overall positioning failure due to single-point sensor failure, demonstrating significant engineering application value.

[0132] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of combined actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application. In the above embodiments, the descriptions of each embodiment have their own emphasis; for parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0133] Figure 7 A structural block diagram of an example of a cable fault anomaly data self-verification system based on multi-source traveling wave comparison according to an embodiment of this application is shown.

[0134] like Figure 7 As shown, the cable fault anomaly data self-verification system 700 based on multi-source traveling wave comparison includes an acquisition and buffer unit 710, a transient feature extraction unit 720, a time difference and energy consistency construction unit 730, a credibility weight generation unit 740, a time estimation self-verification unit 750, and a fault location output unit 760.

[0135] The acquisition and buffering unit 710 is used to acquire multiphase sheath traveling wave signals collected by multi-source sensor arrays at at least two cross-interconnected grounding boxes along the target cable line in response to the same fault event. The sampling clocks of each sensor in the multi-source sensor array are independent and time asynchrony is allowed. For the same fault event, the at least two cross-interconnected grounding boxes complete the acquisition and buffering of the multiphase sheath traveling wave signals within the same fault trigger acquisition window, so that the multiphase sheath traveling wave signals acquired at different cross-interconnected grounding boxes have an overlapping interval on the time axis that can be used for cross-correlation analysis.

[0136] The transient feature extraction unit 720 is used to extract transient features from the multiphase sheath traveling wave signal to obtain a preprocessed signal sequence containing traveling wave front features.

[0137] The time difference and energy consistency construction unit 730 is used to perform cross-correlation analysis on any two signals in the preprocessed signal sequence, calculate the relative time difference between each sensor pair at different positions or different phases, and calculate the energy ratio between each sensor pair based on the energy characteristics of the preprocessed signal sequence in the wavefront neighborhood, thereby constructing the relative time difference matrix and energy ratio matrix of the multi-source signal.

[0138] The credibility weight generation unit 740 is used to calculate the normalized weights characterizing the credibility of the signal correlation between each sensor pair based on the signal-to-noise ratio index of each sensor and the energy ratio matrix, and to generate a reliability weight matrix.

[0139] The time estimation self-verification unit 750 is used to construct a time estimation model based on weighted iteration using the relative time difference matrix and the reliability weight matrix. It updates the estimated wavefront arrival time of each sensor through iterative calculation, identifies abnormal sensor data that deviates from the statistical distribution during the iteration process, and corrects the abnormal sensor data using the weighted constraints of normal sensor data to obtain the target wavefront arrival time sequence after self-verification and calibration.

[0140] The fault location output unit 760 is used to calculate and output the fault location based on the target wavefront arrival time sequence, cable line length parameters, and traveling wave propagation speed; wherein, the cable line length parameters include the position parameters of the at least two cross-interconnected grounding boxes on the target cable line and / or the segment length parameters of adjacent cable segments, so as to map the target wavefront arrival time sequence to the fault location.

[0141] In some embodiments, this application provides a non-volatile computer-readable storage medium storing one or more programs including execution instructions. The execution instructions can be read and executed by electronic devices (including but not limited to computers, servers, or network devices) to perform the steps of any of the above-described cable fault anomaly data self-verification methods based on multi-source traveling wave comparison.

[0142] In some embodiments, this application also provides a computer program product, the computer program product including a computer program stored on a non-volatile computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the steps of any of the above-described cable fault anomaly data self-verification methods based on multi-source traveling wave comparison.

[0143] In some embodiments, this application also provides an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the steps of a cable fault anomaly data self-verification method based on multi-source traveling wave comparison.

[0144] The above-described product can perform the methods provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects for performing the methods. Technical details not described in detail in this embodiment can be found in the methods provided in the embodiments of this application.

[0145] The electronic devices in this application can exist in various forms, including but not limited to: mobile communication devices, ultra-mobile personal computer devices, portable entertainment devices, or other airborne electronic devices with data interaction functions.

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A self-verification method for cable fault anomaly data based on multi-source traveling wave comparison, characterized in that, The method includes: In response to the same fault event occurring in the target cable line, multiphase sheath traveling wave signals are acquired by multi-source sensor arrays at at least two cross-interconnected grounding boxes along the target cable line; wherein the sampling clocks of each sensor in the multi-source sensor array are independent and time asynchrony is allowed, and for the same fault event, the at least two cross-interconnected grounding boxes complete the acquisition and buffering of the multiphase sheath traveling wave signals within the same fault trigger acquisition window, so that the multiphase sheath traveling wave signals acquired at different cross-interconnected grounding boxes have overlapping intervals on the time axis that can be used for cross-correlation analysis; Transient features are extracted from the multiphase sheath traveling wave signal to obtain a preprocessed signal sequence containing traveling wave front features; Perform cross-correlation analysis on any two signals in the preprocessed signal sequence to calculate the relative time difference between each sensor pair at different positions or phases, and calculate the energy ratio between each sensor pair based on the energy characteristics of the preprocessed signal sequence in the wavefront neighborhood, thereby constructing the relative time difference matrix and energy ratio matrix of the multi-source signal. Based on the signal-to-noise ratio index of each sensor and the energy ratio matrix, the normalized weights characterizing the reliability of the signal correlation between each sensor pair are calculated, and a reliability weight matrix is ​​generated. A time estimation model based on weighted iteration is constructed using the relative time difference matrix and the reliability weight matrix. The estimated arrival time of each sensor is updated by iterative calculation. During the iteration process, abnormal sensor data that deviates from the statistical distribution is identified. The abnormal sensor data is corrected by the weighted constraints of normal sensor data to obtain the target wavefront arrival time sequence after self-verification and calibration. Based on the target wavefront arrival time sequence, cable line length parameters, and traveling wave propagation speed, the fault location is calculated and output; wherein, the cable line length parameters include the position parameters of the at least two cross-interconnected grounding boxes on the target cable line and / or the segment length parameters of adjacent cable segments, so as to map the target wavefront arrival time sequence to the fault location.

2. The method according to claim 1, characterized in that, The acquisition of multiphase sheath traveling wave signals collected by multi-source sensor arrays at at least two cross-interconnected grounding boxes along the target cable line includes: A circular buffer queue is constructed in the acquisition terminal of each cross-interconnected grounding box to perform real-time cyclic overwriting and storage of the raw sampling data output by the multi-source sensor array, and the buffer depth of the circular buffer queue is configured to at least cover the sum of the preset pre-trigger length and the preset post-trigger length. The transient change detection logic is run using a local independent clock. When a signal amplitude change or rate of change exceeds a preset threshold is detected, a local trigger anchor point is generated in response to the same fault event. The local trigger anchor point is used to characterize the sampling time corresponding to the trigger determination. Using the local trigger anchor point as a timing reference, the historical data segment containing the preset pre-trigger length is backtracked from the circular buffer queue, and the data segment continues to be collected until the subsequent data segment with the preset post-trigger length is covered. The combined data segment is used as the multiphase sheath traveling wave signal and local buffering is performed. The pre-trigger length is configured to be no less than the sum of the maximum traveling wave transmission delay between any two cross-interconnected grounding boxes and the maximum allowable sampling clock deviation of the system, so as to tolerate the trigger timing deviation caused by different wavefront arrival times and asynchronous sampling clocks of each acquisition terminal, and to ensure that the signal segments acquired by each acquisition terminal contain the same fault traveling wavefront and complete wavefront background noise on the physical time axis.

3. The method according to claim 1, characterized in that, The step of extracting transient features from the multiphase sheath traveling wave signal to obtain a preprocessed signal sequence containing traveling wavefront features includes: Baseline drift correction is performed on the multiphase sheath traveling wave signal, and hybrid denoising processing is performed by combining adaptive median filtering and wavelet threshold transformation to retain the high-frequency transient components of the signal and generate a denoised traveling wave signal with a high signal-to-noise ratio. The traveling wave signals of each phase in the denoised traveling wave signal are respectively denoted as follows: Using pre-defined structural elements For each Perform morphological decomposition and calculate morphological gradient waveforms To highlight the abrupt change edge features of the signal: ， In the formula, Represents morphological dilation operations. Represents morphological erosion operation. This indicates the phase index, used to distinguish different phase signals in a multiphase sheath traveling wave; For time variables, Indicates the first The time-domain amplitude sequence of the denoised traveling wave signal is used to characterize the transient response of the fault traveling wave in that phase. Indicates the first Morphological gradient waveform of a phase-traveling wave signal; In various morphological gradient waveforms The modulus maxima are detected in the middle, and the sampling time corresponding to the modulus maxima is determined as the arrival time of the single-phase candidate wavefront of the corresponding phase; At the same cross-connection grounding box, the time deviation between the arrival times of the single-phase candidate wavefronts of each phase is calculated by utilizing the propagation simultaneity characteristics of multiphase signals. Isolated abrupt changes that exceed the preset simultaneity threshold are eliminated, and the time that passes the consistency check is retained as the final candidate wavefront arrival time. Centered on the arrival time of the final candidate wavefront, a wavefront neighborhood window of a preset length is extracted from the denoised traveling wave signal, and the denoised traveling wave signal, the morphological gradient waveform, the arrival time of the final candidate wavefront, and the wavefront neighborhood window together constitute a preprocessed signal sequence.

4. The method according to claim 3, characterized in that, The process involves performing cross-correlation analysis on any two signals in the preprocessed signal sequence to calculate the relative time difference between sensor pairs at different positions or phases, and calculating the energy ratio between sensor pairs based on the energy characteristics of the preprocessed signal sequence in the wavefront neighborhood, thereby constructing a relative time difference matrix and an energy ratio matrix for the multi-source signals. This includes: For any two sensors in the sensor set and Denoising traveling wave signal segments corresponding to the wavefront neighborhood window are extracted from the preprocessed signal sequence and denoised as follows: and ; The time span corresponding to the wavefront neighborhood window Internally, based on and Construct cross-correlation function from overlapping intervals : ， In the formula, Indicates a time-lag variable; Search within a preset search range for the lag time that maximizes the cross-correlation function. and will Determined to be a sensor With sensors The fine relative time difference between them is used to construct a relative time difference matrix by traversing all sensor combinations. ; The transient energy of each denoised traveling wave signal segment within the wavefront neighborhood window is calculated using the square integral method. and build sensors With sensors Energy ratio : ， In the formula, For sensor index placeholders, take or ; Indicates sensor Transient energy within the wavefront neighborhood window, Indicates the corresponding sensor Denoising traveling wave signal segments within the wavefront neighborhood window; when When the energy exceeds the preset lower limit, the sensor... With sensors Energy ratio and energy ratio Assembled into an energy ratio matrix; when When the energy level is not greater than the preset lower limit, the corresponding sensor pair is marked as low confidence and does not participate in the construction of the energy ratio matrix; wherein, the energy ratio matrix is ​​used to quantitatively characterize the consistency of energy attenuation of the same fault traveling wave on different propagation paths and phases.

5. The method according to claim 4, characterized in that, Based on the signal-to-noise ratio (SNR) index of each sensor and the energy ratio matrix, normalized weights characterizing the reliability of signal correlation between each sensor pair are calculated, and a reliability weight matrix is ​​generated, including: For each sensor in the sensor set Based on the background noise data segment in the preprocessed signal sequence that is located before the wavefront neighborhood window, the standard deviation of the background noise of the sensor is calculated. A preset noise lower limit is set for the standard deviation of the background noise. To obtain the equivalent noise amount used for weight calculation. ; Calling the transient energy of each denoised traveling wave signal segment , and the energy ratio of each sensor pair For any two different sensors and In order to satisfy and All are greater than the preset lower energy limit Under the condition of constructing the original reliability weights Computational model: ， In the formula, Indicates unnormalized sensor pairs The original reliability weights, and The preset adjustment coefficient; when or Not greater than the preset lower energy limit At that time, the corresponding sensor pair is marked as low confidence and then... ; The original reliability weights are normalized so that for any target sensor The sum of its association weights with the other sensors satisfies The normalization calculation formula is: ， In the formula, For the sensor traversal variables; Apply each normalized reliability weight Combine them to generate a reliability weight matrix. .

6. The method according to claim 5, characterized in that, The step of constructing a weighted iterative time estimation model using the relative time difference matrix and the reliability weight matrix, and updating the estimated wavefront arrival time of each sensor through iterative calculation, includes: Call relative time difference With reliability weight For any sensor pair in a sensor network, a physical constraint relationship is established based on the absolute time difference equaling the relative observation value, and a residual variable is defined to characterize the degree of violation of this constraint relationship. : ， In the formula, and The sensors to be estimated are respectively With sensors The absolute wavefront arrival time relative to the virtual reference time; Construct a global objective function based on minimizing weighted residuals, and utilize a robust loss function. To suppress the impact of large residual abnormalities on the overall estimate, a nonlinear mapping is applied to the residual variables. ， In the formula, express The robust loss function value is used as the residual penalty term; A reference constraint is introduced to fix the virtual reference time, such that the arrival time of the absolute wavefront satisfies a zero-mean benchmark: ， By solving for the minimum value of the global objective function under the condition of satisfying the reference constraints, the absolute wavefront arrival time estimation results of each sensor relative to the virtual reference time are obtained.

7. The method according to claim 6, characterized in that, The process of identifying anomalous sensor data that deviates from the statistical distribution during the iteration process, correcting the anomalous sensor data using weighted constraints of normal sensor data, and obtaining a target wavefront arrival time sequence after self-calibration includes: For any target sensor Determine its associated sensor set ;in, Includes sensors There is an effective relative time difference And corresponding reliability weight Sensors ; Perform an iterative update step based on a weighted average, utilizing the current estimated arrival times of each associated sensor. relative time difference Estimate the target sensor The next round of absolute arrival time : ， exist Not greater than the preset minimum weight and threshold At that time, keep Or the sensor Marked as low confidence and skipped in this round of updates; After each iteration update, zero-mean processing is performed on the estimated arrival times of all participating sensors to satisfy the reference constraint; the iterative update step is started with the initial assumption value until the change in the estimated arrival times of all participating sensors is less than the preset convergence threshold or the preset iteration number limit is reached. After iterative convergence, based on the relative time difference Calculate the weighted residual scale for each sensor to identify anomalous sensor data: ， In the formula, Indicates sensor The weighted residual scale; when Greater than the preset anomaly detection threshold At that time, the corresponding sensor The sensor was determined to be faulty. Activate self-checking and correction mechanism: for abnormal sensors This is done by temporarily removing the sensor from the set of sensors participating in the iteration, resulting in the normal set of sensors. and using only the normal sensor set Abnormal sensors The arrival time is reconstructed: ， In the formula, Represents a normal sensor set Any normal sensor number in the system, Indicating abnormal sensor The corrected estimated wavefront arrival time. Indicates sensor With sensors The relative time difference between the observations, Indicates sensor pair Reliability weights; when Not greater than the preset minimum weight and threshold At that time, adopt The weighted median or median as ; Using the revised Replace the original anomaly estimate and write it back to the wavefront arrival time sequence, then re-trigger iterative updates and zero-mean processing until all anomaly sensors are eliminated or the preset self-checking iteration limit is reached.

8. The method according to claim 7, characterized in that, The calculation and output of the fault location based on the target wavefront arrival time sequence, cable line length parameters, and traveling wave propagation speed includes: The environmental monitoring parameters associated with the target cable line are acquired, and based on the propagation characteristic model of the cable insulation material, the nominal traveling wave propagation velocity is compensated and corrected using the environmental monitoring parameters to obtain the actual traveling wave propagation velocity under the current operating conditions. The environmental monitoring parameters include at least one of the following: temperature, humidity, soil resistivity, and sheath grounding status parameters. The cross-interconnection grounding boxes included in the cable line length parameters. The location parameters are represented as line mileage coordinates. And represent the location of the fault point as The reference point at the time of the fault occurrence is represented as The target wavefront arrival time sequence is compared with the mileage coordinates. The corresponding calibration wavefront arrival time is denoted as ,based on Construct a system of positioning equations: ， Based on reliability weight matrix And the set of abnormal sensors, deriving the overall confidence level of the nodes corresponding to each cross-interconnected grounding box. Among them, for the node corresponding to the abnormal sensor, For non-abnormal nodes, Rather than The sum of the incident weights from non-abnormal nodes is positively correlated and normalized; Construct a weighted least squares objective function with comprehensive confidence constraints. : ， By solving the objective function Minimize parameters as well as To determine and output the final location of the fault; Based on the optimal residual value obtained from the solution and the proportion of abnormal sensors marked during the self-verification process, a reliability index for the positioning result is quantified and generated, and the location of the fault point is determined. Output synchronously with the aforementioned reliability indicators.

9. A self-verification system for cable fault anomaly data based on multi-source traveling wave comparison, characterized in that, The system includes: A data acquisition and buffering unit is used to acquire multiphase sheath traveling wave signals collected by multi-source sensor arrays at at least two cross-interconnected grounding boxes along the target cable line in response to the same fault event. The sampling clocks of each sensor in the multi-source sensor array are independent and time asynchrony is allowed. For the same fault event, the at least two cross-interconnected grounding boxes complete the acquisition and buffering of the multiphase sheath traveling wave signals within the same fault-triggered acquisition window, so that the multiphase sheath traveling wave signals acquired at different cross-interconnected grounding boxes have overlapping intervals on the time axis that can be used for cross-correlation analysis. The transient feature extraction unit is used to extract transient features from the multiphase sheath traveling wave signal to obtain a preprocessed signal sequence containing traveling wave front features. The time difference and energy consistency construction unit is used to perform cross-correlation analysis on any two signals in the preprocessed signal sequence, calculate the relative time difference between each sensor pair at different positions or different phases, and calculate the energy ratio between each sensor pair based on the energy characteristics of the preprocessed signal sequence in the wavefront neighborhood, thereby constructing the relative time difference matrix and energy ratio matrix of the multi-source signal. The credibility weight generation unit is used to calculate the normalized weights characterizing the credibility of the signal correlation between each sensor pair based on the signal-to-noise ratio index of each sensor and the energy ratio matrix, and to generate a reliability weight matrix. The time estimation self-verification unit is used to construct a time estimation model based on weighted iteration using the relative time difference matrix and the reliability weight matrix. It updates the estimated wavefront arrival time of each sensor through iterative calculation, identifies abnormal sensor data that deviates from the statistical distribution during the iteration process, and corrects the abnormal sensor data using the weighted constraints of normal sensor data to obtain the target wavefront arrival time sequence after self-verification and calibration. The fault location output unit is used to calculate and output the fault location based on the target wavefront arrival time sequence, cable line length parameters, and traveling wave propagation speed; wherein, the cable line length parameters include the position parameters of the at least two cross-interconnected grounding boxes on the target cable line and / or the segment length parameters of adjacent cable segments, so as to map the target wavefront arrival time sequence to the fault location.

Images