A power distribution network fault edge cloud cooperative measurement device and method

Abstract

The application discloses a power distribution network fault edge cloud cooperative measurement device and method. The edge side matches wavelet traveling wave characteristics on the high-frequency signal of the feeder terminal, draws a suspected fault time window, establishes a traveling wave arrival time delay table through multi-scale modulus maximum value alignment, and forms a traveling wave arrival disturbance zone. The overlapping degree of the time window of the multiple terminals in the verification section is verified to determine a trigger flag, and the trigger effective disturbance zone is uploaded to the cloud together with the high-frequency signal. The cloud performs signal-to-noise ratio tracking to determine synchronous compression parameters, traces frequency changes along the ridge line to solve dispersion correction coefficients, and converts the frequency-related speed difference into the ranging result. The mutual correlation between the terminals is used to distinguish the reflection dominant component and the background interference component, and the consistency confidence is quantitatively given. According to the confidence evaluation positioning deviation, the deviation node of the traveling wave arrival time delay table of the terminal with one-way narrow standard deviation is corrected and recalculated, and the remaining terminals are output according to the original result, so that the ranging adaptability in the complex operating environment is improved.

Description

Technical Field

[0001] This invention relates to the field of power system fault detection technology, and in particular to a distribution network fault edge-cloud collaborative measurement device and method. Background Technology

[0002] Distribution networks have a wide distribution range and operate in complex environments. In the event of short circuits, open circuits, or other faults, quickly and accurately locating the fault is crucial for minimizing power outage time and reducing economic losses. Traveling wave ranging methods locate faults based on transient traveling wave signals generated by sudden changes in electrical quantities. Compared to traditional impedance ranging methods, their accuracy is less affected by line parameters. In recent years, combined with edge computing and cloud-based collaborative processing architectures, they have gradually become one of the mainstream technologies for fault location in distribution networks.

[0003] However, this approach faces three limitations in practical applications: First, the propagation speed of traveling waves varies with signal frequency, causing dispersion. Most existing methods still calculate distance based on a single fixed wave speed, failing to factor in this frequency-dependent speed difference into the ranging results. Second, the high-frequency signals acquired by the terminal include not only directly propagating traveling waves but also waveforms reflected from line branches, grounding devices, and other structures. These two types of waves are mixed in the same sampled data, and existing methods lack the means to separate them and provide a reliability assessment of the ranging results. Third, once the ranging parameters are set, they are typically used unchanged in each fault, making it impossible to correct the parameters using continuously accumulated operational data. These three limitations collectively restrict the further improvement of the ranging accuracy of the distribution network fault edge-cloud collaborative measurement method. Summary of the Invention

[0004] This invention provides a distribution network fault edge-cloud collaborative measurement device and method, which aims to perform hierarchical processing of traveling wave signals through an edge-cloud collaborative architecture: the edge side completes the initial feature detection and multi-terminal joint trigger judgment, and the cloud side combines dispersion calculation and cross-correlation analysis between terminals to transform dispersion deviation and reflection interference into quantifiable and correctable reference quantities. Furthermore, by leveraging the deviation convergence law accumulated from multiple historical events, the ranging parameters are continuously improved, forming a distribution network fault ranging scheme with self-correcting capabilities.

[0005] The first aspect of this invention proposes a method for collaborative measurement of fault edges and clouds in a distribution network, comprising the following steps:

[0006] Acquire high-frequency signals from the feeder terminals of the distribution network, and perform wavelet traveling wave feature matching on the edge side based on the high-frequency signals to form a suspected fault time window;

[0007] Based on the suspected fault time window, multi-scale modulus maxima alignment is performed to obtain the terminal traveling wave arrival record. The traveling wave arrival delay table is established by statistically analyzing the delay deviation of the terminal traveling wave arrival record. Based on the traveling wave arrival delay table, the deviation and instability section is identified to form the traveling wave arrival disturbance band.

[0008] Based on the arrival disturbance band of the traveling wave, the overlap of the suspected fault time window is verified to determine the trigger flag. The arrival disturbance band of the traveling wave is uploaded to the cloud through the trigger flag for signal-to-noise ratio tracking to determine the synchronization compression parameters. Based on the synchronization compression parameters and the arrival delay table of the traveling wave, the ridge dispersion is calculated to obtain the terminal ranging result.

[0009] Based on the terminal ranging results, the cross-correlation between the ranging of the terminals is analyzed to obtain the terminal correlation intensity spectrum. Based on the terminal correlation intensity spectrum, a subset of refraction and reflection differences is selected to generate a terminal fusion group. Based on the terminal fusion group, the dominant reflection component and the background interference component are identified to determine the consistency confidence.

[0010] Based on the consistency confidence level, a weighted deviation assessment is performed on the terminal ranging results to form a positioning deviation record table. The positioning deviation record table is then screened for deviation convergence trends to establish a compressed feature list. Based on the compressed feature list, the deviation nodes of the traveling wave arrival delay table are oriented and corrected to output the fault ranging results.

[0011] A second aspect of this invention provides a distribution network fault edge-cloud collaborative measurement device, comprising:

[0012] The initial inspection and location module is used to acquire high-frequency signals from the feeder terminals of the distribution network, and to perform wavelet traveling wave feature matching based on the high-frequency signals at the edge side to form a suspected fault time window;

[0013] The time delay modeling module is used to perform multi-scale modulus maxima alignment based on the suspected fault time window to obtain the terminal traveling wave arrival record, to establish a traveling wave arrival time delay table based on the statistical time delay deviation of the terminal traveling wave arrival record, and to identify the deviation instability section based on the traveling wave arrival time delay table to form a traveling wave arrival disturbance band.

[0014] The cloud-based refinement module is used to verify the overlap of the suspected fault time window based on the traveling wave arrival disturbance band to determine the trigger flag, upload the traveling wave arrival disturbance band to the cloud for signal-to-noise ratio tracking to determine the synchronization compression parameters, and perform ridge dispersion calculation based on the synchronization compression parameters and the traveling wave arrival delay table to obtain the terminal ranging result.

[0015] The consistency verification module is used to analyze the inter-terminal ranging correlation based on the terminal ranging results to obtain the terminal correlation intensity spectrum, filter the refraction and reflection difference subsets based on the terminal correlation intensity spectrum to generate a terminal fusion group, and identify the dominant reflection component and background interference component based on the terminal fusion group to determine the consistency confidence.

[0016] The directional output module is used to perform weighted deviation evaluation on the terminal ranging results based on the consistency confidence level to form a positioning deviation record table, screen the positioning deviation record table for deviation convergence trend to establish a compressed feature list, and perform directional correction of the deviation nodes of the traveling wave arrival delay table based on the compressed feature list to output fault ranging results.

[0017] The beneficial effects of this invention are reflected in the following points: First, a sudden change in current captured by a single terminal cannot determine whether it is the same real fault. Non-fault disturbances such as local load switching and temporary grounding can also trigger suspected time windows for individual terminals. This solution introduces overlap verification, expanding the judgment criteria from single terminal readings to the time intersection of suspected time windows of multiple terminals. Only when a disturbance is simultaneously captured by multiple terminals is it designated as a deviation from the instability section and triggered for cloud processing. Local disturbances are thus filtered out, and high-frequency data is no longer fully uploaded to the cloud, reducing the transmission burden of the edge-cloud collaborative link.

[0018] Secondly, this scheme does not follow the conventional practice of setting a single wave velocity for the entire line. Instead, it extracts the dispersion characteristics from the high-frequency signal itself for each event, and uses synchronous compression transform to track the arrival time differences of different frequency bands along the time-frequency ridge, converting them into dispersion correction coefficients for each terminal and each event. Different terminals may be located in different line sections and have different conductor types. A fixed wave velocity naturally has a deviation for some sections. However, the correction coefficients extracted for each event come directly from the actual measured signal of the terminal itself, without relying on empirical wave velocities that may not match the actual line, thus improving the accuracy of the ranging results in adapting to different line structures.

[0019] Finally, the ranging results of this scheme are not output all at once: on the one hand, by analyzing the cross-correlation between the ranging results of the terminals, terminals that conform to the laws of refraction and reflection are selected, and the energy ratio of the dominant reflection component and the background interference component is converted into a consistency confidence level, so that the ranging results have their own credibility label; on the other hand, the convergence trend of the terminal deviation value is tracked across multiple historical events, and only the terminals with the deviation continuously narrowing in one direction are directionally corrected, so that the ranging parameters have the ability to continuously improve with the accumulation of operational data, which makes up for the limitation of existing methods that stop at a single real-time calculation. Attached Figure Description

[0020] Figure 1 This is a flowchart illustrating a method for collaborative measurement of faults in a power distribution network using edge-cloud technology, as described in this invention.

[0021] Figure 2 This is a schematic diagram of the ridge dispersion tracking and measured dispersion value extraction of the present invention.

[0022] Figure 3 This is a structural block diagram of a power distribution network fault edge-cloud collaborative measurement device according to the present invention.

[0023] Wherein: 1-ridge line; 2-frequency band interval; 3-frequency band trajectory; 4-sampling point density; 5-frequency band dispersion offset; 6-stable continuous interval; 7-measured dispersion value; 8-feeder terminal; 9-original high-frequency signal. Detailed Implementation

[0024] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0025] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0026] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0027] The technical solutions of the embodiments of this application will be described below.

[0028] like Figure 1 As shown, this embodiment of the invention provides a method for collaborative measurement of fault edges and clouds in a distribution network, including the following steps S101-S105:

[0029] Step S101: Obtain the high-frequency signal of the distribution network feeder terminal, and perform wavelet traveling wave feature matching on the edge side based on the high-frequency signal to form a suspected fault time window.

[0030] Specifically, high-frequency signals from the distribution network feeder terminals are acquired. Each outgoing switchgear in the feeder terminal is equipped with two sensing channels: a power frequency current transformer and a high-frequency current transformer. The power frequency current transformer calculates the effective current value cycle-by-cycle at a 4kHz sampling rate, continuously monitoring the line's operating status. When the difference between the effective current values ​​of two adjacent power frequency cycles exceeds 20% of the rated current, a disturbance is detected, triggering the high-frequency current transformer in the feeder terminal to start acquisition. Normal fluctuations below this threshold do not trigger the high-frequency acquisition channel. After starting, the high-frequency current transformer acquires high-frequency signals at a 10MHz sampling rate. The acquisition duration is defined by a 1ms threshold before and a 9ms threshold after the trigger time, totaling 10ms, forming a 10ms acquisition window. This window covers the arrival and attenuation process of traveling wave signals generated by typical distribution network faults at the feeder terminal. The frequency response range of the high-frequency signal sensing channel covers tens of kHz to several MHz, matching the typical main frequency band of traveling wave signals on overhead and cable lines in the distribution network. Each feeder terminal independently completes disturbance detection and acquisition triggering. Adjacent terminals determine whether the triggering condition is met based on their local power frequency current RMS value. The triggering status of one terminal does not change the judgment threshold or triggering result of other terminals. The 4kHz power frequency sampling rate corresponds to 80 sampling points per 50Hz cycle, capturing the cycle-to-cycle changes in the current RMS value on a millisecond time scale. The power frequency channel itself does not occupy high-frequency processing resources for extended periods. The 20% disturbance judgment threshold is taken from the historical statistical upper limit of normal load fluctuations on the line. The current change amplitude caused by real faults such as short circuits and open circuits is generally much higher than this level. It is expected that multiple feeder terminals on the same feeder line will trigger high-frequency acquisition successively due to the same real fault. The high-frequency signals recorded by each terminal do not overlap or merge with each other and are all retained as independent records.

[0031] On the edge side, a fault-probable time window is formed by initial detection using wavelet traveling wave feature matching based on high-frequency signals. The high-frequency signal is decomposed into four levels using a db4 wavelet basis, and the modulus maxima points and their corresponding times and signs are extracted at each scale. The abrupt change front of the traveling wave is characterized by the simultaneous appearance of modulus maxima with consistent signs at multiple decomposition scales. This characteristic is significantly different from the random noise, where the maximum signs are chaotic and the amplitude decays irregularly at each scale. When the ratio of the modulus maxima amplitudes of two adjacent decomposition scales falls within a preset interval, and at least three of the four decomposition levels show modulus maxima with consistent signs, that moment is determined to be a traveling wave feature matching point. The fault-probable time window is defined centered on the traveling wave feature matching point, extending forward by 0.5 ms and backward by 1.5 ms, covering the time range during which the traveling wave may continue to propagate and undergo multiple reflections at the feeder terminal. When multiple traveling wave feature matching points appear within the same 10 ms acquisition window, a corresponding fault-probable time window is defined for each matching point. Overlapping time windows are merged, while non-overlapping windows are retained separately. The wavelet decomposition performed on the edge side only involves four layers, and the number of decomposition layers will not be increased. More refined dispersion feature extraction is performed in the cloud, while the edge side only handles the feature matching judgment itself. When no matching combination of modulus maxima is detected in the high-frequency signal, no fault-probability time window is formed in this acquisition. The feeder terminal continues to monitor the line status according to the effective value of the power frequency current, waiting for the next disturbance to trigger a new round of high-frequency acquisition. If there are high-frequency oscillations caused by the switching action of power electronic equipment near the traveling wave feature matching point, their modulus maxima usually appear only on one or two adjacent decomposition scales, failing to meet the matching condition of at least three layers with consistent signs. Therefore, the high-frequency signal waveform generated by such oscillations will not be misjudged as the leading edge of the traveling wave abrupt change.

[0032] Step S102: Perform multi-scale modulus maxima alignment based on the suspected fault time window to obtain the terminal traveling wave arrival record; establish a traveling wave arrival delay table based on the statistical delay deviation of the terminal traveling wave arrival record; and identify the deviation instability section based on the traveling wave arrival delay table to form a traveling wave arrival disturbance band.

[0033] Specifically, multi-scale modulus maxima alignment is performed based on the suspected fault time window to obtain the arrival record of the traveling wave. Within the suspected fault time window, modulus maxima points with consistent signs are selected from each decomposition scale, and their corresponding precise sampling times are recorded for each scale. The temporal resolution of wavelet decomposition decreases as the decomposition scale increases, and the positioning of coarse-scale modulus maxima points on the time axis is correspondingly widened, while the positioning of fine-scale modulus maxima points is closer to the actual arrival time of the traveling wave leading edge. Multi-scale alignment is calculated according to the formula t=Σ(t_i / Δt_i) / Σ(1 / Δt_i), where t_i is the sampling time corresponding to the modulus maxima point on the i-th decomposition scale, Δt_i is the temporal resolution corresponding to that scale, and t is the arrival time of the aligned traveling wave. Each scale is weighted according to the reciprocal of its temporal resolution, and the higher the resolution of the scale, the greater its contribution to the final alignment time. The sign of the modulus maxima point itself represents the polarity of the traveling wave. During the alignment process, the signs of each scale have been checked and matched in the S101 determination stage, so there is no need to repeat the determination in this step. The sign is directly used as the traveling wave polarity record. The arrival time and polarity of each aligned ripple event together constitute a terminal ripple arrival record, which is the complete record of this event for the terminal. When multiple independent modulus maxima combinations (originating from multiple actual arrival events covered by the same suspected fault time window) occur successively within the same fault-prone time window, each set of modulus maxima is aligned separately, generating its own independent terminal ripple arrival record, which is not merged into a single record. The accuracy of the alignment time t depends on the time resolution of the finest decomposition scale. Even if the coarse-scale modulus maxima points are blurred due to widening, the arrival time retained in the weighted average terminal ripple arrival record is still dominated by the fine-scale positioning accuracy.

[0034] In some embodiments, establishing a traveling wave arrival delay table based on the statistical delay deviation of the terminal traveling wave arrival record includes: extracting the arrival time and estimate from the terminal traveling wave arrival record to obtain an arrival estimate comparison table; performing forward and reverse path drop normalization correction on the arrival estimate comparison table to generate a forward and reverse path drop map; extracting the cumulative drop deviation using the forward and reverse path drop map to obtain a cumulative drop set; and establishing a traveling wave arrival delay table based on the terminal summary delay deviation of the cumulative drop set.

[0035] An arrival estimation comparison table is obtained by extracting arrival times and estimates from the terminal traveling wave arrival records. The actual arrival time corresponding to each terminal traveling wave arrival record is added to the theoretical propagation time obtained by dividing the line length from the terminal to the reference point (the outlet of the substation switchgear) by the nominal wave velocity. The nominal wave velocity is taken as 270 m / s, a commonly used empirical value for the propagation speed of traveling waves in overhead lines. The length of each line in the actual line parameter database is entered according to the design drawings and remains fixed for each terminal. The arrival estimation comparison table records both the actual arrival time and the estimated arrival time for each terminal. The difference between the two is recorded simultaneously. A positive difference indicates that the actual arrival of the traveling wave is later than the estimate, while a negative difference indicates that it arrives earlier. When the line includes sections with mixed cable and overhead laying, the theoretical propagation time is calculated by dividing the length of each section by the corresponding wave velocity (the wave velocity of cable sections is usually lower than that of overhead sections) and then summing the results. The nominal wave velocity of the overhead line is not uniformly applied. When multiple records exist for the same terminal in the terminal traveling wave arrival record (corresponding to multiple independent arrival events), the estimated time and difference are extracted separately for each record in the arrival estimation comparison table, without merging multiple records into a single estimation result. The difference in the arrival estimation comparison table itself has not undergone any normalization or path correction processing; it is merely the raw deviation between the terminal traveling wave arrival record and the theoretical estimate. Records in the arrival estimation comparison table where the absolute value of the difference exceeds the theoretical propagation time are also retained as raw difference records without numerical truncation or labeling removal, and the reasonableness of the difference is not judged at this stage. The estimated arrival time of the same terminal in different batches of events is recalculated for each batch based on the actual trigger time and fixed line length and wave velocity, without reusing the estimation results calculated in previous batches, even if the line parameters of the terminal have not changed between the two events.

[0036] The arrival valuation comparison table is normalized and corrected for forward and reverse path elevation differences to generate a forward and reverse path elevation difference map. The difference for each terminal in the arrival valuation comparison table is calculated for both forward and reverse paths: the forward path is the line segment from the terminal along the route towards the next adjacent terminal, and the reverse path is the line segment from the terminal against the route towards the previous adjacent terminal. The line length and nominal wave speed for each direction are substituted into the theoretical propagation time to calculate the difference for the terminal in each direction in the arrival valuation comparison table. The forward and reverse path elevation difference is the difference between the forward and reverse differences for the terminal. The elevation difference reflects whether the positional deviation of the terminal relative to its two adjacent terminals is symmetrical; a high degree of symmetry should result in an elevation difference close to zero. The elevation difference is normalized using the shorter segment of the corresponding line length. The normalized elevation difference value is dimensionless, allowing direct comparison of elevation differences between terminals on different length segments. The forward and reverse path elevation difference map records the normalized elevation difference value for each terminal, arranged according to the actual order of the terminals on the line. When a terminal is located at the end of a line and has only one adjacent terminal on one side, the difference in the arrival estimation comparison table for this terminal is only in one direction. In the forward and reverse path drop diagrams, the drop for this type of terminal is directly taken as the difference in its only existing direction, without subtracting between forward and reverse directions. In the forward and reverse path drop diagrams, the drops for the same terminal in different event batches are calculated independently, without averaging or comparing trends with the drops of previous batches for that terminal. The normalization process uses a shorter segment length rather than a longer segment or average length because the physical meaning of the drop itself is limited to the segment with the shorter propagation time and higher measurement sensitivity between two segments. Normalizing with a longer segment would dilute the discernibility of the drop.

[0037] The cumulative deviation of elevation difference is extracted using the forward and reverse path elevation difference map to obtain the cumulative deviation set. Terminals in the forward and reverse path elevation difference map are arranged into a sequence according to their actual arrangement on the line. Starting from the terminal at the beginning of the line, the normalized elevation difference value of each terminal is accumulated sequentially. The cumulative deviation of the k-th terminal is the sum of the elevation difference values ​​of the first k terminals in the sequence. The accumulation direction is consistent with the physical direction from the beginning to the end of the line; no reverse accumulation is performed. The cumulative deviation of each terminal is summarized into the cumulative deviation set, and its value is expanded sequentially along the line direction according to the accumulation position. When there is no systematic deviation on the line, the values ​​of the forward and reverse path elevation difference map for each terminal fluctuate around zero, and the cumulative deviation tends to be stable overall, without continuously increasing or decreasing unidirectionally with the accumulation position. When the recorded length of a certain segment on the line does not match the actual length, the cumulative deviation will show a continuous unidirectional drift starting from that segment. In the same event, when multiple branches share the same substation outgoing switchgear as the starting point of the line, the cumulative height difference of each branch is calculated independently, without sharing the same accumulation benchmark, and the terminals of different branches are not included in the same accumulation sequence. The cumulative deviation in the cumulative height difference set itself has no upper limit; even if the cumulative value of a terminal far exceeds that of other terminals, it is still recorded according to the original accumulation result without truncation. The starting point for the accumulation of the cumulative height difference set is fixed at the first terminal of each branch at the substation outgoing switchgear. Even if this terminal itself has only one adjacent terminal at the starting point of the line, or its height difference value is different from that of intermediate terminals, it is still used as the starting point of the accumulation sequence and is not skipped or processed separately. If a terminal in the forward and reverse path height difference diagram fails to collect valid data or has a missing height difference value in this event, the cumulative deviation of the cumulative height difference set at that terminal uses the cumulative value of the previous terminal, and normal accumulation resumes from the next terminal with a valid height difference value.

[0038] A traveling wave arrival delay table is created by summarizing the cumulative drop set for each terminal and aggregating the delay deviation. During aggregation, the cumulative drop deviation of the terminal in the cumulative drop set is used as the deduction amount: this cumulative drop deviation is subtracted from the original difference in the arrival estimation comparison table for that terminal to obtain the terminal's delay deviation. The systematic component that accumulates continuously along the line is thus deducted, and the retained portion reflects the local delay anomaly of that terminal relative to its neighboring terminals. The traveling wave arrival delay table records the delay deviation for each terminal, along with the original arrival time of that terminal in the terminal's traveling wave arrival record; both values ​​correspond to the same event. Multiple independent arrival records for the same terminal (corresponding to multiple arrivals within the same event) are calculated separately for their respective delay deviations. The same terminal can correspond to multiple delay deviation records in the traveling wave arrival delay table, and they are not merged into a single value. If the cumulative deviation in the cumulative drop set is used because the drop value of a certain terminal is missing, and the cumulative value of the previous terminal is used, the time delay deviation calculation of that terminal is still based on its original difference in the arrival estimation comparison table. The calculation of the time delay deviation of that terminal is not skipped due to the use of the cumulative value in the cumulative drop set. The time delay deviation values ​​in the traveling wave arrival time delay table are not normalized; the unit is consistent with the original difference value, and both are presented in the time dimension, without being converted to a dimensionless ratio. Multiple branches under the same substation outgoing switchgear each establish independent traveling wave arrival time delay tables. Terminals of different branches do not appear in the same time delay table. An abnormal time delay deviation in one branch does not change the values ​​in the time delay tables of other branches. The range of time delay deviation values ​​after deducting the systematic cumulative component is usually significantly smaller than the original difference in the arrival estimation comparison table. This narrowing range can serve as an indirect basis for judging whether there is an overall parameter error (such as a deviation in line length entry) in the line section where the terminal is located.

[0039] In some embodiments, identifying deviation instability segments and forming a traveling wave arrival disturbance band based on the traveling wave arrival delay table includes: rolling the delay deviation amount in the traveling wave arrival delay table according to sampling batches to obtain a deviation rolling sequence; generating a terminal wave intensity spectrum by comparing the positive and negative deviation amplitude differences through the deviation rolling sequence; removing periodic attributable fluctuations from the terminal wave intensity spectrum to determine deviation instability segments; and forming a traveling wave arrival disturbance band based on the statistical fluctuation range and distribution location of the deviation instability segments.

[0040] The travel wave arrival delay table is arranged in a rolling sequence by sampling batches to obtain the delay deviation. The delay deviations recorded by the same terminal in different sampling batches (corresponding to different suspected fault events) are arranged in chronological order. The delay deviation of each batch is taken from the terminal's record in the travel wave arrival delay table for that particular event. Batches with an interval exceeding 30 days between two events are assigned to different rolling windows. An interval exceeding 30 days is considered to have crossed different operating condition intervals, and the window is recalculated based on the interval. The width of the rolling window is the 10 most recent batches. The window moves forward with each new batch, and the earliest batch's value moves out of the window, always retaining the delay deviation values ​​of 10 batches within the window. The deviation rolling sequence records the delay deviation values ​​of each batch within the window for each terminal, arranged chronologically by batch time. The arrangement order itself does not change the value of the delay deviation for each batch. If a terminal has fewer than 10 historical batches, the deviation rolling sequence records the number of existing batches. If the window width is less than 10, no fixed placeholder value is used to make it 10 batches. The deviation rolling sequences of different terminals on the same branch line are maintained independently. The number of batches within a terminal window does not affect the number of batches in other terminal windows on the same branch line. When multiple records are corresponding to the same terminal in the ripple arrival delay table due to multiple arrivals, the rolling arrangement selects the record with the largest absolute value of delay deviation among all records in that batch to participate in the deviation rolling sequence. The remaining records do not change the value of that batch in the rolling sequence, but are only retained separately as supplementary records for that batch in the ripple arrival delay table. Each time a new batch is added to the ripple arrival delay table, a value is synchronously added to the deviation rolling sequence of the corresponding terminal. The addition operation is triggered independently for each terminal, and the addition of other terminal sequences is not suspended because a certain terminal has not experienced a ripple event in its current batch.

[0041] The terminal fluctuation intensity spectrum is generated by comparing the difference between positive and negative deviation amplitudes in the deviation from the rolling sequence. The positive deviation amplitude of the terminal is obtained by summing the absolute values ​​of the time delay deviations that are positive in the rolling sequence and dividing by the number of positive values; the negative deviation amplitude is obtained by summing the absolute values ​​of the time delay deviations that are negative and dividing by the number of negative values. When all values ​​in the window have the same sign (only positive or only negative), the amplitude in the other direction is recorded as 0, and the fluctuation intensity calculation of the terminal is not abandoned due to the lack of a value with the opposite sign. The fluctuation intensity is calculated according to the formula I=|A_p-A_n|, where A_p is the positive deviation amplitude of the terminal, A_n is the negative deviation amplitude, and I is the fluctuation intensity. When the terminal's line segment parameters have long-term input deviations, the time delay deviation stably shifts to one side, increasing the I value; when the I value is close to 0, the positive and negative deviations are symmetrical, more closely resembling the behavior of random fluctuations. Each terminal occupies one record in the terminal fluctuation intensity spectrum, storing its fluctuation intensity value. The values ​​are expanded according to the actual arrangement order of the terminals on the line, maintaining the same arrangement standard as the deviation from the rolling sequence and previous stages. For terminals with fewer than 10 batches within the deviation from the rolling sequence window, the fluctuation intensity is calculated based on the difference between the positive and negative amplitudes under the existing batch count, and the generation of fluctuation intensity values ​​is not delayed due to the window not being full. After the batch combination of the same terminal within the deviation from the rolling sequence window changes, its corresponding fluctuation intensity value in the terminal fluctuation intensity spectrum is recalculated accordingly, and the fluctuation intensity values ​​obtained under the previous batch combination are no longer retained. Excessive differences in the fluctuation intensity values ​​between two adjacent terminals in the terminal fluctuation intensity spectrum do not trigger additional processing. The spectrum presents the actual calculation results of each terminal as is, without smoothing or interpolation to smooth out the differences between adjacent terminals. Positive and negative deviation amplitudes are averaged according to the number of values ​​in their respective directions, rather than averaging according to the total number of batches in the window. When the number of positive and negative values ​​is unbalanced, the two averages still reflect the typical amplitude of the deviation in their respective directions, and are not diluted by the number of values ​​in the other direction due to the smaller number of values.

[0042] The deviation from instability is determined by removing periodic attributable fluctuations from the terminal fluctuation intensity spectrum. For each terminal in the spectrum, the fluctuation intensity is subtracted from the historical periodic baseline value corresponding to the current season. The periodic baseline value is calculated as the average fluctuation intensity of the same batch of terminals in the same season for each of the four seasons (spring, summer, autumn, and winter), reflecting the normal contribution of temperature changes causing periodic fluctuations in line impedance to the fluctuation intensity. If the residual fluctuation intensity after subtraction exceeds a fixed threshold corresponding to the voltage level of the line where the terminal is located, the terminal is determined to be in an unstable state; terminals whose residual fluctuation intensity does not exceed the threshold are considered to have fluctuations mainly explained by periodic factors and are not determined to be unstable. Terminals determined to be unstable in the fluctuation intensity spectrum are merged according to their adjacency on the line: if there is no non-instable terminal interval between two unstable terminals, they are merged into the same segment; if there is a non-instable terminal interval, they are each treated as an independent segment. The deviation from instability segment records the starting and ending terminals of each segment and the number of unstable terminals contained within the segment. A single isolated unstable terminal also constitutes a separate deviation from instability segment, containing only that single terminal. If the historical data of a terminal is insufficient to cover a particular season, the periodic benchmark value for that season shall be replaced by the median of the average fluctuation intensity of other terminals at the same line voltage level for that season. The instability assessment of a single terminal in the terminal fluctuation intensity spectrum shall not be skipped due to incomplete historical data for that terminal. Within the same segment deviating from the instability zone, the residual fluctuation intensity values ​​of adjacent unstable terminals are not required to be similar; they only need to be spatially adjacent. A segment may contain multiple terminals with significantly different residual fluctuation intensities, or it may consist of only a single terminal with residual fluctuation intensities close to the threshold.

[0043] A traveling wave arrival disturbance zone is formed based on the statistical fluctuation range and distribution location of the deviation instability sections. The fluctuation range of each segment within the deviation instability section is obtained by summing the line lengths between the starting and ending terminals of that segment, according to the segment lengths between adjacent terminals in the line parameter database. For a deviation instability section consisting of a single terminal, the fluctuation range is half the shorter segment length between that terminal and its adjacent terminal, which is taken as the equivalent influence range of the isolated instability point. The distribution location is taken as the midpoint of the line coordinates between the starting and ending terminals of that segment, with the substation outgoing switchgear as the zero point, and the segment lengths accumulated along the line direction. Both the fluctuation range and distribution location values ​​of each deviation instability section are recorded in the traveling wave arrival disturbance zone, arranged in chronological order along the line direction. When multiple deviation instability sections exist on the same branch line, each section is recorded separately in the traveling wave arrival disturbance zone, and they are not merged into the same zone due to proximity in distribution location. Once the fluctuation range and distribution location of the deviation from the instability zone are determined, they only change with the intensity spectrum of the terminal fluctuations in this round and the calculation results of the deviation from the instability zone, and are not affected by other data outside this stage. Strips with larger fluctuation range values ​​in the traveling wave arrival disturbance zone usually correspond to areas with a wide and continuous distribution of unstable terminals on the line. The time delay deviations of each terminal within these strips exhibit similar directional offsets, which physically point more likely to a systematic error in the line parameter input than to a single-point measurement anomaly. The distribution location coordinates of the traveling wave arrival disturbance zone and the terminal installation location coordinates recorded in the line parameter database share the same set of zero points and direction conventions. No additional coordinate transformation is required for direct comparison between the two coordinate systems. The distribution locations of the traveling wave arrival disturbance zones generated by two adjacent branches are not compared because although the coordinate zero points of different branches share the substation outgoing switchgear, they follow the direction of their respective branches, and the values ​​are not comparable across branches.

[0044] Step S103: Based on the overlap of the suspected fault time window of the traveling wave arrival disturbance zone, determine the trigger flag. Upload the traveling wave arrival disturbance zone to the cloud through the trigger flag to determine the synchronization compression parameters by signal-to-noise ratio tracking. Based on the synchronization compression parameters and the traveling wave arrival delay table, perform ridge dispersion solution to obtain the terminal ranging result.

[0045] Specifically, the trigger flag is determined based on the overlap of the suspected fault time windows when the traveling wave arrives at the disturbance zone. For each terminal covered by the unstable section in the traveling wave arrival zone, the start and end times of the suspected fault time window corresponding to this event are checked for each terminal. The time window data is taken from the independently defined window of each terminal. When the time windows of any two terminals intersect on the time axis, it is recorded as an overlap. The overlap degree is calculated according to the formula o=m / C(n,2), where n is the total number of terminals in the unstable section, C(n,2) is the total number of terminal pairs that can be obtained by combining n terminals pairwise, m is the number of terminal pairs that actually overlap, and o is the overlap degree. When the overlap degree exceeds 0.6, it is determined that the unstable section corresponds to a real line disturbance event, and the trigger flag is set to valid; for sections where the overlap degree does not exceed 0.6, it is considered that the suspected fault time windows of each terminal in the section are independent and do not point to the same event, and the trigger flag remains invalid. The trigger flag is determined individually for each deviation from the unstable section. Multiple deviations from the unstable section in a single event may each receive a valid or invalid determination, without affecting each other. When the total number of terminals, n, is 1 (a section consisting of isolated unstable terminals), there is no terminal pair available for comparison overlap. The trigger flag is then determined directly based on whether the absolute value of the terminal's own time delay deviation exceeds a fixed threshold for the voltage level of its line. For a deviation from the unstable section determined by the trigger flag to be valid, its coverage area uses the recorded start and end terminals within the deviation from the unstable section, and the fluctuation range value uses the recorded value of the traveling wave arriving at the disturbance band; neither is recalculated in this step. The overlap threshold of 0.6 is taken from the lower quartile of the overlap distribution of suspected fault time windows for each terminal in historical real fault events on the same line. Overlap below this level is relatively rare in historical real fault samples.

[0046] The traveling wave is triggered by a flag, and the data is uploaded to the cloud for signal-to-noise ratio tracking to determine synchronization and compression parameters. For example... Figure 2As shown, the step of uploading the traveling wave to the disturbance zone, specifically for the deviation and instability section determined by the trigger flag, involves uploading the original high-frequency signal 9 collected by the feeder terminal 8 within the suspected fault time window, along with the recorded fluctuation range and distribution location within the disturbance zone, to the cloud. Sections with invalid trigger flags are not uploaded, and the original high-frequency signal 9 remains at the edge. The cloud calculates the signal energy of the uploaded high-frequency signal segment by segment using a 0.1 millisecond short-time window and compares it with the energy of the background noise segment outside the suspected fault time window before the traveling wave arrives, obtaining a curve showing the signal-to-noise ratio (SNR) changing over time. The window function bandwidth required for synchronous compression transformation is taken as the empirical bandwidth corresponding to the lowest value of the SNR curve within the entire time window; the lower the SNR, the wider the corresponding window function bandwidth, exchanging stronger smoothness for noise suppression; the higher the SNR, the narrower the window function bandwidth, preserving finer temporal resolution. The window function bandwidth and the corresponding minimum value of the signal-to-noise ratio (SNR) curve are written into the synchronization compression parameters, which are determined separately for each terminal. Even if different terminals belong to the same traveling wave arriving at the disturbance zone and deviating from the unstable section, their synchronization compression parameters may differ. When there is no obvious trough in the SNR curve, the window function bandwidth is taken as the empirical bandwidth corresponding to the overall average of the curve, and the minimum point is no longer identified separately. The uploaded data received by the cloud is processed into processing tasks according to the sections where the trigger flag is valid. Each task only processes terminal data within the same deviation from the unstable section. The calculation process of the synchronization compression parameters for different sections is independent of each other, and intermediate results are not shared. Sections in the traveling wave arriving at the disturbance zone with larger recorded fluctuation range values ​​usually cover more terminals. When processing such sections, the cloud needs to calculate the SNR curve for each terminal separately, and the processing time is roughly proportional to the number of terminals covered.

[0047] In some embodiments, the step of obtaining terminal ranging results by ridge dispersion calculation based on the synchronization compression parameters and the traveling wave arrival delay table includes: performing ridge fidelity weighted sampling based on the synchronization compression parameters to form a compressed sampling sequence; using the compressed sampling sequence to track the ridge frequency change trajectory to determine the measured dispersion value; performing wave velocity dispersion analysis on the measured dispersion value to generate a dispersion correction coefficient; and performing wave velocity conversion on the traveling wave arrival delay table based on the dispersion correction coefficient to obtain the terminal ranging results.

[0048] Based on the synchronous compression parameters, ridge-based fidelity weighted sampling is performed to form a compressed sampling sequence. The original high-frequency signal 9 undergoes synchronous compression transformation according to the window function bandwidth determined in the synchronous compression parameters. The transformation result generates a concentrated energy ridge trajectory 1 on the time-frequency plane. Each point on ridge 1 corresponds to a time and the dominant instantaneous frequency at that time. Fidelity weighted sampling along ridge 1 extracts the time, frequency, and energy values ​​point by point along the ridge trajectory 1. The sampling point density 4 is not fixed: the sampling interval for sections with larger curvature of ridge 1 is half of the original sampling interval, while the sampling interval for sections with smaller curvature and gentler frequency changes is the original sampling interval itself. Under both sampling intervals, sections with larger curvature retain more sampling points describing frequency changes. All sampling points collected along the ridge are arranged into the compressed sampling sequence, recording the time, frequency, and energy point by point, and arranged in chronological order. Sampling points in the encrypted sections and the original interval sections are not distinguished in the compressed sampling sequence and are recorded uniformly according to the same set of three-value formats. When the window function bandwidth in the synchronization compression parameters is relatively wide (corresponding to terminals with low signal-to-noise ratio), the ridge trajectory itself is relatively smooth, and the curvature is generally small. The proportion of segments using encrypted sampling in the compressed sampling sequence is also correspondingly low. When the ridge trajectory experiences a brief interruption in the time-frequency plane, the compressed sampling sequence does not generate sampling points in that interruption interval; sampling points before and after the interruption are recorded independently, without interpolation across interruptions. When the overall ridge trajectory shows a monotonically decreasing frequency over time, the correspondence between the sampling density distribution and the ridge curvature of the compressed sampling sequence still follows the same rule, without changing to a fixed sampling interval due to the overall monotonic trend. If the synchronization compression parameters of the same terminal change due to recalculation of the signal-to-noise ratio curve, the compressed sampling sequence is regenerated according to the changed window function bandwidth, without retaining the sampling points corresponding to the previous values.

[0049] For example, the step of using the compressed sampling sequence to trace the ridge frequency change trajectory to determine the measured dispersion value includes: extracting the instantaneous frequency of each frequency band ridge to form a frequency band frequency trajectory group based on the compressed sampling sequence; calculating the arrival time offset of adjacent frequency bands for the frequency band frequency trajectory group to obtain a frequency band dispersion offset set; tracing the dispersion stability characteristics based on the frequency band dispersion offset set to output a dispersion stability record; and selecting the measured dispersion value from the dispersion stability record by frequency band.

[0050] The instantaneous frequencies of the ridges in each frequency band are extracted from the compressed sampling sequence to form frequency band trajectory groups. Sampling points in the compressed sampling sequence are assigned to predefined frequency band intervals 2 according to their frequency values. Frequency band intervals 2 are divided into 50kHz increments, covering the frequency response range of the sensing channel from tens of kHz to several MHz. A subset of sampling points arranged chronologically within the same frequency band interval 2 constitutes a frequency band trajectory 3 corresponding to that frequency band. Each frequency band to which a sampling point falls generates a frequency trajectory, which are then combined into a frequency band trajectory group. Frequency bands to which no sampling point falls do not generate a corresponding trajectory. If the same sampling point falls exactly on the boundary between two frequency bands due to the ridge crossing the frequency band boundary, it is assigned to the frequency band with the higher value. In the compressed sampling sequence, if the time gaps formed by the interruption of the ridge line still belong to the same frequency band interval before and after the gap, the sampling points before and after the gap are still arranged continuously in chronological order within the frequency trajectory of that frequency band; the interruption does not change the arrangement order within the trajectory. In a frequency trajectory group, the number of sampling points for the same frequency band is directly related to the length of time the ridge line traverses that band. The longer the ridge line stays in a frequency band, the more sampling points are generated for that band's corresponding trajectory. When the compressed sampling sequence covers only a narrow frequency range, the number of trajectories generated in the frequency trajectory group is correspondingly smaller. This situation is common in acquisition events where narrowband noise interference is stronger than the actual traveling wave signal, which differs from the typical traveling wave signal's behavior of spanning multiple frequency bands and exhibiting a significant frequency decrease over time. No comparisons are made between the trajectories in the frequency trajectory group based on energy or quantity. Each trajectory only reflects the temporal direction of the ridge line within its corresponding frequency band; the relative importance between frequency bands is not determined in this step.

[0051] For each frequency band frequency trajectory group, the arrival time offset of adjacent frequency bands is calculated to obtain the frequency band dispersion offset set. Adjacent frequency bands in the frequency band frequency trajectory group are arranged from high to low frequency values. The difference between the start time of the high-frequency band frequency trajectory 3 and the start time of the low-frequency band frequency trajectory 3 is taken as the frequency band dispersion offset 5 for this pair of adjacent frequency bands. A positive difference indicates that the high-frequency component arrives before the low-frequency component, which conforms to the physical law that the propagation speed of high-frequency components on the transmission line is usually faster than that of low-frequency components. Each entry in the frequency band dispersion offset set is a pair of adjacent frequency bands and their corresponding arrival time offset, arranged in descending frequency order. If one of the two adjacent frequency bands in the frequency band frequency trajectory group does not generate a corresponding trajectory, the arrival time offset of this pair of frequency bands is not calculated, and the record of this pair of frequency bands does not appear in the frequency band dispersion offset set. It is skipped, and the system directly compares the adjacent frequency bands with existing trajectories in the next pair of frequency band frequency trajectory groups. Frequency band pairs with negative arrival time offsets are also retained according to the original difference record without sign correction or marking as abnormal. Such anomalous offsets may indeed occur under specific line structures. When the same frequency band in the frequency band dispersion offset set is simultaneously the high-frequency band of two adjacent frequency band pairs and the low-frequency band of another frequency band pair, the starting time of the trajectory of that frequency band is taken independently in the two offset calculations and is not adjusted in any way due to its participation in the two calculations. When the number of generated trajectories in the frequency band trajectory group is less than two, there are no adjacent frequency band pairs available for calculation, and this step does not generate a frequency band dispersion offset set for that terminal. The magnitude of the offset in the frequency band dispersion offset set does not have a fixed proportional relationship with the difference in center frequencies between two adjacent frequency bands. The offset corresponding to the same 50kHz frequency band interval may be significantly different in different frequency intervals. This nonlinearity itself is a direct manifestation of the dispersion phenomenon.

[0052] Based on the frequency band dispersion offset set, the dispersion stability record is output by tracking the dispersion stability characteristics. Frequency band pairs in the frequency band dispersion offset set are arranged from high to low frequency. The offset signs of adjacent frequency band pairs are checked for consistency. Frequency band pairs with consecutively consistent signs are recorded as a stable and persistent interval (6), with the interval length determined by the number of frequency band pairs contained within it. When the stable and persistent interval (6) has a length of 3 or more frequency band pairs, it is determined that the interval reflects the true dispersion trend; intervals with a length of less than 3 frequency band pairs are considered local fluctuations and are not included in the stable and persistent characteristics. Each terminal identifies a stable and persistent interval and records it segment by segment in the dispersion stability record, recording the average offset between the start and end frequency bands and the interval's value. When multiple stable and persistent intervals satisfying the length condition exist for the same terminal, the dispersion stability record is recorded separately for each interval, without merging them into a single interval or retaining only the longest interval. When the offset symbols in a frequency band dispersion offset cluster frequently alternate between adjacent frequency band pairs, and no consecutive intervals of the same sign with a length of 3 frequency band pairs can be found, the longest segment among all alternating symbol intervals is directly recorded in the dispersion stability record. The requirement of meeting the 3-frequency-band pair threshold is no longer applied. Only the actual interval length is additionally marked next to the average offset value corresponding to this interval to distinguish its reliability from that of a regular stable interval. The purpose of marking the actual interval length is to allow the average offset value itself to carry a direct clue to its reliability: the closer the interval length is to the 3-frequency-band pair threshold, the fewer data points support the average value. Therefore, the reliability of the same average offset value supported by 5 frequency band pairs is not the same as that supported by 2 frequency band pairs.

[0053] The measured dispersion values ​​are determined by filtering the dispersion stability records by frequency band. If multiple stable intervals exist in the dispersion stability records, the one with the longest interval is selected, and the average offset of that interval is taken as the measured dispersion value for this event at that terminal. When multiple intervals with the same length exist simultaneously, the one with the larger absolute value of the average offset is selected. The offset with the larger absolute value corresponds to a more obvious frequency-time offset feature and carries more identifiable information. The measured dispersion value is recorded for each terminal, along with the start and end frequency bands of its source interval. Records in the dispersion stability records that indicate an actual interval length of less than 3 frequency band pairs can still be used as sources of measured dispersion values, but they must meet the condition that the record is the longest interval among all intervals in the dispersion stability records of that terminal. No separate selection criteria are set for intervals that are too short. Records with a negative average offset in the measured dispersion values ​​are also retained as original values ​​without absolute value conversion. The negative value itself still has interpretable physical meaning in the wave velocity dispersion analysis stage. If the longest interval of a terminal changes due to the regeneration of the dispersion stabilization record, the measured dispersion value will be changed to the average offset corresponding to the new longest interval, and the values ​​from previous batches will not be retained. The selection of measured dispersion values ​​is based solely on the absolute values ​​of the interval length and the average offset in the dispersion stabilization record, without considering the type of deviation and instability segment or the trigger flag determination results of the terminal. The selection logic applies uniformly to all terminals. When there is exactly one interval in the dispersion stabilization record, no comparison of length or absolute value is involved, and the average offset corresponding to this unique interval is directly used as the measured dispersion value.

[0054] Wave velocity dispersion analysis is performed on the measured dispersion values ​​to generate dispersion correction coefficients. The dispersion correction coefficients are calculated using the formula k = a × Δτ + b, where Δτ is the measured dispersion value (dimensions are time, units are μs), and a is the regression slope (units are μs) calibrated for that voltage level. -1), b is the dimensionless regression intercept, and k is the dimensionless dispersion correction coefficient; a and b are taken from the linear regression results of the ratio of measured dispersion values ​​to known wave velocity deviations in historical calibration tests of lines of the same voltage level. Lines of different voltage levels each have their own independent set of a and b values, and parameters are not shared across voltage levels. When the measured dispersion value is 0, the dispersion correction coefficient k is equal to b itself, and the regression intercept term directly gives the baseline correction amount in the absence of dispersion. A dispersion correction coefficient k value greater than 1 indicates that the equivalent wave velocity corresponding to the event at this terminal is lower than the nominal wave velocity, and the actual propagation time is longer than estimated based on the nominal wave velocity; a k value less than 1 indicates the opposite. The dispersion correction coefficient is generated separately for each terminal. The k value of different terminals within the same deviation instability section may be different, and it is not mandatory for all terminals within the section to share the same k value. When historical calibration test data for a line voltage level is lacking, a and b are replaced by calibration parameters from adjacent lines of the same conductor type, and uncalibrated default values ​​are not used. The measured dispersion values ​​corresponding to multiple historical events of the same terminal are different, and the dispersion correction coefficients calculated by substituting them into the formula for each event are also independent. The k value calculated for previous events is not reused, even if the values ​​of a and b do not change between two events. The range of the k value in historical calibration test data usually falls between 0.9 and 1.1. The calculation results outside this range are still recorded as the original output of the formula without numerical limiting. The fact that the value exceeds the range can itself be used as an observation clue that there is an anomaly in this calibration or this event of the terminal.

[0055] The terminal ranging result is obtained by converting the wave velocity in the traveling wave arrival delay table based on the dispersion correction coefficient. The corrected wave velocity is calculated by dividing the nominal wave velocity by the dispersion correction coefficient k. The corrected wave velocity is inversely proportional to the dispersion correction coefficient; when k is greater than 1, the corrected wave velocity is lower than the nominal wave velocity, and when k is less than 1, the corrected wave velocity is higher than the nominal wave velocity. The terminal ranging result is the known line length from the terminal to the reference point, plus the product of the time delay deviation recorded by the terminal in the traveling wave arrival delay table and the corrected wave velocity. A positive time delay deviation indicates that the actual arrival is later than the theoretical estimate, and the corresponding distance correction is positive. The ranging result is correspondingly greater than the original line length, indicating that there is an additional equivalent propagation distance between the terminal and the actual disturbance source. The terminal ranging result is generated independently for each terminal. The ranging result of a terminal does not refer to the calculation process or intermediate values ​​of other terminals in the same deviation and instability section, even if multiple terminals in the section share the same triggering event. When multiple delay deviation records exist for the same terminal in the traveling wave arrival delay table, the terminal ranging results are calculated separately for each record. Each delay deviation corresponds to an independent ranging result, without weighted merging of multiple records. When the dispersion correction coefficient is 0, the divisor for correcting the wave velocity is 0. In this step, the nominal wave velocity of the voltage level of the terminal is directly substituted into the ranging formula, and the wave velocity correction step is not performed. The unit of the terminal ranging result is consistent with the line length, presented in meters. The ranging results of different terminals are output together with the original delay deviation in their respective traveling wave arrival delay tables. The two values ​​are presented together, allowing for tracing back the original delay source corresponding to each ranging result.

[0056] Step S104: Based on the terminal ranging results, analyze the cross-correlation between the ranging of the terminals to obtain the terminal correlation intensity spectrum. Based on the terminal correlation intensity spectrum, select the refraction and reflection difference subset to generate the terminal fusion group. Based on the terminal fusion group, identify the dominant reflection component and the background interference component to determine the consistency confidence.

[0057] In some embodiments, the step of analyzing the cross-correlation between terminals based on the terminal ranging results to obtain the terminal association strength spectrum includes: performing time-series trajectory parsing on the terminal ranging results to generate a terminal trajectory group; analyzing the correlation between each terminal trajectory based on the terminal trajectory group to obtain a terminal association coefficient; eliminating common mode trends and extracting discrete clustering features based on the terminal association coefficient to form an association strength distribution; and delineating the boundaries between strong and weak associated terminals from the association strength distribution to obtain the terminal association strength spectrum.

[0058] Terminal trajectory groups are generated by analyzing the time-series trajectory of the terminal ranging results. Multiple independent records corresponding to the same terminal in the terminal ranging results are sorted according to their original arrival times in the traveling wave arrival delay table. The sorted distance value sequence constitutes a time-series trajectory for that terminal. The first value in the trajectory usually comes from the record where the traveling wave directly reaches the terminal, and its distance value is closest to the actual distance from the terminal to the actual disturbance point. The second and subsequent values ​​mostly come from traveling waves reflected back from other structures on the line (such as branch access points, the terminal's own grounding device), and the corresponding distance values ​​are systematically larger. The terminal trajectory group consists of the time-series trajectories of all terminals. The trajectory length is consistent with the number of records for that terminal in the terminal ranging results; the more records a terminal has, the longer its trajectory. Terminals with only a single record in the terminal ranging results have a trajectory length of 1, containing only one value, and are also included in the terminal trajectory group, without being individually removed due to their short length. The trajectory lengths of different terminals on the same branch line can vary. The length differences are not normalized. Shorter trajectories maintain the same recording format as longer trajectories, but contain fewer value entries. The order of values ​​for each trajectory within a terminal trajectory group is fixed and remains unchanged. Even if the arrival time of a record changes due to recalculation of terminal ranging results, the relative order of values ​​within the trajectory is preserved. While the arrival times of multiple records from the same terminal in the terminal ranging results may differ, time-series trajectory analysis prioritizes arrival times, disregarding the magnitude of delay deviations. Even if the absolute value of a delay deviation is significantly greater than other records, the trajectory still occupies its corresponding position in the terminal trajectory group based on its arrival time.

[0059] Terminal correlation coefficients are obtained by analyzing the correlation between terminal trajectories based on terminal trajectory groups. The value sequences of any two terminals within the same deviation and instability zone are matched one-to-one according to their positions (first value to first value, second value to second value). The comparison range is limited to the value pairs covered by the shorter of the two trajectories; values ​​from the longer trajectory that exceed this range are not included in the comparison. The terminal correlation coefficient is calculated using the formula c = 1 - Σ|d_i - d_j| / (n × L), where d_i and d_j are the values ​​of the two trajectories at the same position, n is the number of value pairs participating in the comparison, L is the total length of the branch line containing the terminal pair, used to normalize the accumulated difference to a dimensionless interval, and c is the terminal correlation coefficient. A c value closer to 1 indicates that the values ​​of the two trajectories at the corresponding positions are closer, and the two terminals are likely to have captured the same set of physical events; a c value closer to 0 indicates that the two trajectories are significantly different. The terminal correlation coefficient is calculated for each terminal pair individually. The number of terminal pairs that can be formed from n terminals on a branch line is determined by pairwise combinations. All terminal pairs must be calculated, without skipping any pairs due to physical distance. When the trajectory lengths of two terminals in a terminal trajectory group are both 1, the number of value pairs n participating in the comparison is 1, and the terminal correlation coefficient is still calculated using the same formula without using any other simplified algorithms. If the trajectory length of the same terminal pair in a terminal trajectory group changes due to the regeneration of the terminal ranging result of one terminal, the terminal correlation coefficient is recalculated based on the two changed trajectories, and the comparison range is changed to the new shorter trajectory length. The previously calculated coefficient is no longer retained.

[0060] Based on the terminal correlation coefficient, common-mode trends are eliminated, and discrete clustering characteristics are extracted to form a correlation strength distribution. The terminal correlation coefficients of all terminal pairs within the same unstable deviation section are first calculated as their arithmetic mean as the common-mode baseline for that section. This baseline reflects the overall background correlation level of the section, which may originate from the inherent electromagnetic coupling of the line itself and is unrelated to whether a specific pair of terminals captured the same event. The common-mode baseline is subtracted from the terminal correlation coefficient of each terminal pair to obtain the net correlation value after deducting the common-mode component. Terminal pairs with net correlation values ​​significantly deviating from 0 are more likely to reflect the true consistency of event capture than the overall background coupling level of the section. The net correlation values ​​of all terminal pairs after deducting the common-mode baseline are arranged in ascending order of physical distance between the terminal pairs on the branch line to form a correlation strength distribution. The calculation of the common-mode baseline covers all terminal pairs within the section and does not remove or recalculate the mean for some terminal pairs with abnormally high or low terminal correlation coefficient values. When the number of terminal pairs within the same segment is less than 3, the statistical significance of the common-mode baseline is limited. The association strength distribution is then changed to directly use the original values ​​of the terminal association coefficients without common-mode deduction. The sign of the net association value in the association strength distribution itself has meaning: a positive value indicates that the association coefficient of the terminal pair is higher than the overall background level of the segment, and a negative value indicates that it is lower than the background level. After arranging the distribution according to physical spacing, intervals with concentrated positive values ​​often correspond to terminal groups that actually captured the same set of reflection events, while intervals with concentrated negative values ​​are more likely to be terminals that are independent of each other but happen to fall within the same deviation instability segment. Branches in the association strength distribution where the net association value is symmetrically distributed around 0 and there are no obvious clusters of positive or negative values ​​are considered to indicate that the overall association relationship between terminals within that branch is weak. This distribution pattern itself is retained as a record along with the association strength distribution.

[0061] The terminal association strength spectrum is obtained by defining the boundaries between strong and weak associations in the association strength distribution. After sorting the net association values ​​in the distribution, the upper quartile is taken as the strong association boundary; terminal pairs with net association values ​​exceeding this boundary are considered strongly associated. The lower quartile is taken as the weak association boundary; terminal pairs with net association values ​​below this boundary are considered weakly associated. Terminal pairs between the upper and lower quartiles are considered moderately associated and are not classified into either the strong or weak association category. The terminal association strength spectrum records the net association value and corresponding strong / weak association category label for each terminal pair. The terminal association strength spectrum uses strong and weak association labels to mark the terminal pairs most likely to jointly capture the same set of reflections and the independent terminal pairs, respectively. Terminal pairs in between are labeled as moderately associated. The same terminal may appear in multiple terminal pairs simultaneously, with some terminal pairs being classified as strongly associated and others as weakly associated. In such cases, the independent judgment results for each terminal pair are recorded, without uniform adjustment due to inconsistencies in the category of the same terminal in different terminal pairs. When the net association value in the association strength distribution is changed to the original value of the terminal association coefficient, the strong and weak association boundaries are still determined according to the upper and lower quartiles of that batch of values. The judgment process does not change to other threshold algorithms due to different data sources. For terminal pairs in the intermediate association category in the terminal association strength spectrum, their net association values ​​are still fully preserved in the spectrum, but they do not carry strong or weak association category labels. This step does not perform further category subdivision on these terminal pairs. When the upper and lower quartiles of the association strength distribution are exactly equal, the strong and weak association boundaries degenerate into the same value. At this time, terminal pairs with a net association value equal to this value are uniformly judged as having intermediate association and are not classified into the strong association side.

[0062] Based on the terminal correlation intensity spectrum, a subset of refraction and reflection differences is selected to generate a terminal fusion group. The terminal fusion group is not a newly added physical device, but rather a set of feeder terminals (numbered 8) within the same deviation instability section that are determined to have jointly captured the same set of reflected traveling waves. It serves as a terminal affiliation list for subsequent energy proportion analysis, and its membership depends on whether the reflection distance sequences recorded by each terminal in the terminal trajectory group match. For terminal pairs determined to be strongly correlated in the terminal correlation intensity spectrum, the distance values ​​at the same location of both terminals are retrieved from the terminal trajectory group for each pair and the difference is calculated. The comparison range is the shorter of the two terminal trajectory group sequences. If the difference monotonically increases or decreases with the trajectory position, consistent with the physical behavior of the traveling wave sequentially passing through branch access points, grounding devices, and other structures, and the corresponding distances for each reflection increasing or decreasing step by step, then the terminal pair is determined to have captured the same series of reflections and is included in the refraction and reflection difference subset. Strongly correlated terminal pairs with fluctuating but not monotonically increasing differences are excluded from this subset. All terminals appearing in the refraction and reflection difference subset are included in the terminal fusion group. A terminal is included as long as it falls within any strongly correlated terminal pair that satisfies the monotonicity condition, regardless of whether it satisfies monotonicity in all strongly correlated terminal pairs in its branch. If no strongly correlated terminal pair satisfying the monotonicity law is found in the same deviation and instability segment, then the terminals involved in all strongly correlated terminal pairs in that segment directly constitute the terminal fusion group. For terminal pairs determined to be weakly or moderately correlated in the terminal correlation intensity spectrum, the difference monotonicity check is skipped; whether their terminals can enter the terminal fusion group depends solely on whether they simultaneously fall within a strongly correlated terminal pair. When there are many strongly correlated terminal pairs in the same segment, the refraction and reflection difference subsets are determined independently pair by pair, and the terminal fusion group is the union of all terminals involved in terminal pairs that satisfy the condition.

[0063] In some embodiments, determining the consistency confidence level based on the identification of the dominant reflection component and the background interference component using the terminal fusion group includes: splitting the terminal fusion group by branch to obtain the branch association intensity distribution; performing valley segmentation on the branch association intensity distribution to obtain the dominant reflection component; dividing the dominant reflection component into low-intensity categories based on secondary clustering of reflection intensity to obtain the background interference component; and comparing the energy proportions of the dominant reflection component and the background interference component to determine the consistency confidence level.

[0064] The branch association strength distribution is obtained by splitting the terminal fusion group according to its branches. Terminals in the terminal fusion group are grouped according to their respective branches, with each branch corresponding to a subset. Terminals within a subset retain their original net association values ​​in the association strength distribution; these values ​​are not recalculated due to the grouping operation. The branch association strength distribution records the net association values ​​of all terminals within each branch, unfolding according to the physical arrangement of terminals along the branch, maintaining consistency with previous stages. In cases where the same terminal in the terminal fusion group is counted multiple times due to participation in multiple strongly associated terminal pairs, only one record is retained in the branch association strength distribution after grouping. The maximum value among the net association values ​​of the multiple terminal pairs in which the terminal participates is taken. The maximum value is used here instead of the mean: the maximum value corresponds to the relationship between the terminal and its most closely associated terminal, best representing whether it truly captured the same set of reflection events; the mean would mix strong and weak associations, weakening discernibility. The net correlation value recorded in the branch correlation strength distribution maintains the same unit as the source correlation strength distribution, and the value range is not rescaled due to grouping or maximum value taking. For multiple branches under the same substation outgoing switchgear, their respective branch correlation strength distributions are established separately and are not merged into a single overall distribution record simply because these branches share the same substation. When the number of terminal fusion group members within a branch is less than two, the branch correlation strength distribution is still recorded according to the existing members, and the generation of the branch's distribution record is not delayed due to the insufficient number of members. The absolute magnitude of the net correlation values ​​in the branch correlation strength distributions of different branches is not compared; the distribution of each branch only reflects the relative correlation strength between terminals within that branch.

[0065] For example, the step of segmenting the branch correlation intensity distribution to obtain the dominant reflection component includes: calculating the intensity gradient using the branch correlation intensity distribution to generate an intensity gradient spectrum; locating the zero-crossing points where the gradient changes from negative to positive based on the intensity gradient spectrum to form a set of valley points; performing segmentation on the branch correlation intensity distribution according to the set of valley points to generate a set of candidate reflection components; and performing proximity continuity verification on the set of candidate reflection components to obtain the dominant reflection component.

[0066] Intensity gradient spectrum is generated by calculating intensity gradients using the branch-linkage intensity distribution. For adjacent terminals arranged along the branch in the branch-linkage intensity distribution, the difference between their net linkage values ​​is divided by the length of the line segment between the two terminals to obtain the intensity gradient at that location. A positive gradient sign indicates that the net linkage value increases along the branch direction, while a negative sign indicates a decrease. The intensity gradient spectrum records the intensity gradient value at each location, corresponding to the midpoint of the segment between two adjacent terminals. The gradient values ​​are arranged according to the branch direction. The intensity gradient spectrum represents the spatial rate of change of the net linkage value along the branch: near-zero gradients indicate similar net linkage values ​​between adjacent terminals, suggesting they belong to roughly the same reflection group; large absolute gradient values ​​indicate a steep drop between the two terminals, suggesting they belong to different reflection groups. When there are two terminals in the branch-linkage intensity distribution, the intensity gradient spectrum contains only one gradient value, corresponding to the segment between this unique pair of terminals. When adjacent terminals have equal net linkage values, the intensity gradient at that location is set to 0. Locations with a gradient of 0 are also included in the intensity gradient spectrum without special marking or skipping. The alternation of positive and negative gradient values ​​in the intensity gradient spectrum directly reflects the undulation of the branch correlation intensity distribution along the branch. More frequent alternation indicates a less smooth change in the net correlation value along the branch. The intensity gradient is normalized by segment length rather than directly taking the difference in net correlation values ​​because the terminal spacing is uneven: even small differences are converted to lower gradients for larger spacings, and vice versa for smaller spacings. Normalization allows for comparison of gradients at different positions. In the intensity gradient spectrum, the positions where two adjacent gradients have opposite signs and their absolute values ​​both exceed the median of all absolute gradients for that branch correspond to segments on the branch correlation intensity distribution curve with sharp inflections. These segments often exhibit local peaks or troughs in the branch correlation intensity distribution.

[0067] Valley point sets are formed by locating zero-crossing points where the gradient changes from negative to positive based on the intensity gradient spectrum. When the gradient values ​​of two adjacent positions along a branch in the intensity gradient spectrum are negative and positive respectively, a zero-crossing point exists between them. The position of the zero-crossing point is determined by linear interpolation of the gradient values ​​on both sides, with the interpolation result falling between the two original positions. Zero-crossing points where the gradient changes from positive to negative are not included in the judgment range of this step; only zero-crossing points where the gradient changes from negative to positive are retained. All zero-crossing points that meet the conditions, along with their interpolated position coordinates, constitute the valley point set. Branches with consistent gradient signs throughout the intensity gradient spectrum and no zero-crossing points where the gradient changes from negative to positive are typically short branches with few terminals, and the intensity distribution of the branch itself is insufficient to form a complete peak-valley undulation pattern. These branches do not generate valley point set records, and segmentation is treated as a single segment for the entire branch. When the distance between two adjacent zero-crossing points is less than the shortest terminal distance of the branch, the two zero-crossing points are merged into one, and the midpoint of their position coordinates is taken as the merged valley point. Duplicate valley points that are too close to each other are not retained in the valley point set. When determining the zero-crossing point position through interpolation, the ratio of the distance from the gradient value of the previous position to 0 to the sum of the absolute values ​​of the gradients on both sides is taken. The physical distance between the two original positions is linearly divided according to this ratio. The closer the ratio is to the previous side, the closer the zero-crossing point position is to the previous terminal. The number of valley points in the valley point set is consistent with the actual number of zero-crossing points that change from negative to positive in the intensity gradient spectrum. Before merging, the two numbers are completely equal. Merging only reduces the number of records in the valley point set when the distance between zero-crossing points is too close.

[0068] Based on the valley point set, the branch correlation intensity distribution is segmented to generate candidate reflection component sets. The valley points in the valley point set are arranged along the branch direction according to their location coordinates. The interval of branch correlation intensity distribution between two adjacent valley points is divided into a candidate segment. Each candidate segment is assumed to correspond to the contribution of an arrival path (including direct wave arrival or a subsequent reflection) on that branch, hence the name candidate reflection component. The distance from the branch start point to the first valley point and from the last valley point to the branch end point are also each divided into a candidate segment, covering the entire length of the branch and leaving no intervals not included in any candidate segment. All candidate segments are listed as candidate reflection component sets, and the net correlation value sequence of the start and end positions and the branch correlation intensity distribution within each segment is recorded. Branches without any valley points in the valley point set are counted as unique candidate segments in the candidate reflection component set without further subdivision. The candidate segments from the starting point to the first valley point and from the last valley point to the end point are recorded in the same format as the candidate segments divided between adjacent valley points. Both record the start and end positions and the net correlation value sequence within the segment, without using different recording methods depending on whether the segment is located at either end of the branch. The number of segments in the candidate reflection component set is one more than the number of valley points in the valley point set. This quantitative relationship stems from the geometric rule that n valley points divide the branch into n+1 segments, and each generated candidate reflection component set satisfies this quantitative relationship. When a candidate segment contains 0 terminals, the segment is still included in the candidate reflection component set, marked as a spatial segment without terminals, and recorded separately from candidate segments containing terminals.

[0069] The dominant reflection component is obtained by performing a proximity continuity check on the candidate reflection component set. For each candidate segment in the candidate reflection component set, the net correlation value sequence of the branch correlation strength distribution within the segment is checked to see if it shows a unidirectional change (continuously increasing or continuously decreasing, without direction reversal). Candidate segments showing a unidirectional change are determined to pass the continuity check; candidate segments with direction reversal within the net correlation value sequence of the branch correlation strength distribution are determined to fail. Spatial segments without terminals are considered to automatically meet the continuity condition and are also determined to pass the check. Among the candidate segments that pass the continuity check, the segment with the longest length is determined to be the dominant reflection component. The remaining candidate segments that pass the check are not included in the dominant reflection component and are separately labeled as segments that pass the check but are not dominant. Physically, a series of real reflections originates from a traveling wave striking branch access points, grounding devices, and other reflection points sequentially along the same path. The ranging values ​​corresponding to each reflection accumulate step by step along the propagation direction, appearing as a monotonic trend without reversal on the net correlation value curve. Sudden increases and decreases in the sequence are more likely due to several independent reflections being mistakenly merged by valley values; therefore, only the longest monotonic segment is identified as the dominant reflection component. If all candidate segments in the candidate reflection component set fail the continuity check, the longest segment is directly taken as the dominant reflection component, and the decision in this step is not abandoned due to failure to pass the check. When multiple segments with equal lengths that pass the check coexist, the one with the larger absolute value at the end of the net correlation value sequence is taken as the dominant reflection component. Once the dominant reflection component is determined, its corresponding start and end positions and the net correlation value sequence of the branch correlation strength distribution are fixed as the original records of the corresponding candidate segments in the candidate reflection component set; this step does not perform any smoothing or correction on these values.

[0070] The background interference component is obtained by secondary clustering based on reflection intensity to divide the dominant reflection component into low-intensity categories. The absolute value of the net association value of each terminal within the dominant reflection component is taken as the reflection intensity. Secondary clustering is then performed based on the reflection intensity, resulting in two categories: one with generally higher reflection intensity values ​​and the other with generally lower reflection intensity values. The clustering uses a one-dimensional K-means method, with the initial cluster centers being the maximum and minimum values ​​of the reflection intensity sequence, iterating until the boundaries between the two categories no longer change. The category with lower reflection intensity is identified as the background interference component, and its corresponding terminals and net association values ​​are recorded together. The category with higher reflection intensity remains within the dominant reflection component, with terminals from each category appearing only in the records of one of the categories. When the number of terminals within the dominant reflection component is 1, the minimum sample size required for secondary clustering is insufficient, and no corresponding background interference component is generated in this step; the terminal remains entirely within the dominant reflection component. When the category boundaries obtained from the two clusters precisely classify the two terminals with the closest reflection intensity values ​​into different categories, the boundaries are not adjusted, and the two terminals are still assigned to the dominant reflection component and the background interference component respectively according to their respective clustering results. During the iterative process of secondary clustering, if after a certain iteration all terminals are clustered into the same category while the other category contains no terminals, the iteration result is considered invalid. The category division obtained from the previous iteration is then used as the final result to ensure that both categories contain terminals after clustering. If the reflection intensity sequence itself has highly concentrated values ​​and the difference between the maximum and minimum values ​​is very small, secondary clustering is still performed according to the same process. The clustering result may simply mechanically divide terminals with extremely similar values ​​into two groups. In this case, the background interference components obtained are also recorded as is according to the clustering result without any additional rationality judgment.

[0071] The consistency confidence level is determined by comparing the energy proportions of the dominant reflection component and the background interference component. The energy of the dominant reflection component is obtained by summing the squares of the reflection intensities of each terminal within the dominant reflection component; similarly, the energy of the background interference component is obtained by summing the squares of the reflection intensities of each terminal within the background interference component. The consistency confidence level is calculated using the formula q = E_r / (E_r + E_n), where E_r is the energy of the dominant reflection component, E_n is the energy of the background interference component, and q is the consistency confidence level, ranging from 0 to 1. A consistency confidence level q closer to 1 indicates a higher proportion of the dominant reflection component's energy in the total energy, and a higher reliability of the current terminal ranging result; a q value closer to 0.5 indicates that the energy of the dominant reflection component and the background interference component are roughly equivalent, and the reliability of the ranging result decreases accordingly. In the case where the background interference component contains no terminals (the number of terminals within the dominant reflection component is 1, and no corresponding background interference component is generated), E_n is 0, and the consistency confidence level q is directly equal to 1. Consistency confidence is calculated separately for each dominant reflection component. If a branch is divided into multiple candidate reflection components, each component corresponds to an independent combination of the dominant reflection component and the background interference component, and each also obtains an independent consistency confidence value. There is no unified confidence merging for the entire branch. The consistency confidence is determined solely by the energy of the sum of the squares of the absolute values ​​of the net correlation values ​​(reflection intensity), without incorporating the correlation determination results from the terminal correlation intensity spectrum or the terminal fusion group stage. The processing volumes for the two stages differ, even if the same terminal participates in the calculations in both stages. Consistency confidence is calculated independently for each event and is not accumulated across batches. A high q-value for a particular terminal in a given event does not guarantee the reliability of all its ranging measurements. During maintenance, the terminal must be verified sequentially using the positioning deviation record table.

[0072] Step S105: Based on the consistency confidence level, the terminal ranging results are weighted and the deviation is evaluated to form a positioning deviation record table. The positioning deviation record table is screened for deviation convergence trend to establish a compressed feature list. Based on the compressed feature list, the deviation nodes of the traveling wave arrival delay table are oriented and corrected to output the fault ranging results.

[0073] Specifically, a positioning deviation record table is generated by weighting the terminal ranging results based on the consistency confidence level. The terminal ranging results of all terminals within the same deviation and instability section are weighted and summed according to their respective consistency confidence levels, then divided by the total confidence level to obtain the confidence-weighted average distance for that section. The weighted average is calculated using the formula D_w=Σ(q_i×D_i) / Σq_i, where D_i is the terminal ranging result of the i-th terminal, q_i is the consistency confidence level corresponding to that terminal, and D_w is the confidence-weighted average distance. The positioning deviation of each terminal is the difference between its own terminal ranging result and D_w. Terminals with higher consistency confidence levels have a greater weight in the influence of their ranging results on D_w, and their deviation values ​​are closer to the original readings of that terminal. The positioning deviation record table records the positioning deviation and the corresponding consistency confidence level for each terminal, and these two values ​​are retained together to distinguish whether the confidence levels behind deviations of the same magnitude are reliable. For terminals with a consistency confidence level of 0 (q_i), this terminal is not included in the numerator and denominator of the weighted average calculation. Its positioning deviation is instead calculated as the difference between its own distance measurement result and D_w, using the same method as other terminals, except that this terminal itself does not contribute to the value of D_w. The sign of the deviation value in the positioning deviation record table corresponds to the direction of the terminal's distance measurement result relative to D_w; a positive value indicates that the terminal's distance measurement result is greater than the confidence-weighted average distance, and a negative value indicates the opposite. When the overall consistency confidence level (q_i) is low, the weighted average calculation process remains unchanged, still using the formula. D_w is more affected by the balancing influence of the distance measurement results from low-confidence terminals. The positioning deviation record table truthfully records the calculation results under this low-confidence environment, without additionally indicating the overall low confidence level of the segment.

[0074] In some embodiments, the step of screening the positioning deviation record table for deviation convergence trend to establish a compressed feature list includes: calculating the recurrence frequency of deviation values ​​through the positioning deviation record table to form a deviation frequency distribution; delineating the range of terminals with concentrated deviation values ​​based on the deviation frequency distribution to obtain a deviation-dense terminal band; tracking the narrowing magnitude of deviation values ​​for the deviation-dense terminal band to form a narrowing trend table; and identifying terminals with unidirectional continuous narrowing based on the narrowing trend table to establish a compressed feature list.

[0075] The frequency distribution of deviation values ​​is formed by calculating the recurrence frequency of deviation values ​​through the positioning deviation record table. The deviation values ​​recorded in the positioning deviation record tables of the same terminal in different historical batches are divided into intervals of 0.5 meters. The number of times each historical batch of the terminal falls into each interval is counted, obtaining the recurrence frequency of each interval for that terminal. The deviation value corresponding to the interval with the highest recurrence frequency in the positioning deviation record table can be regarded as the typical deviation level that repeatedly occurs in multiple events for that terminal, distinguishing it from occasional one-off deviations. The deviation frequency distribution records the recurrence frequency of each interval for each terminal, with intervals arranged from smallest to largest deviation value. When the number of historical batches for a terminal is less than 5, the statistical significance of the recurrence frequency is limited; the deviation frequency distribution still records the frequency of each interval according to the existing batches, without pausing the generation of distribution records for that terminal due to the small number of batches. When all historical deviation values ​​of the same terminal fall into the same 0.5-meter interval, the recurrence frequency of that interval equals the total number of historical batches for that terminal; the frequency of other intervals is 0. The deviation frequency distribution accurately records this highly concentrated distribution pattern. In the positioning deviation record table, records from a particular batch of the same terminal that did not contribute to the confidence-weighted average due to a consistency confidence level of 0 are still included in the interval statistics of the deviation frequency distribution, and are not excluded simply because the confidence level is 0. The 0.5-meter interval width is taken from the typical measurement error level of the terminal ranging results themselves. If the width is too narrow, similar deviations caused by the same physical cause will be split into different intervals, diluting the statistical significance of the recurrence frequency. If the width is too wide, deviations from different causes will be mixed into the same interval. 0.5 meters is an empirical balance between these two risks.

[0076] Based on the frequency distribution of deviations, the range of terminals with concentrated deviation values ​​is delineated to obtain a deviation-dense terminal zone. When the recurrence frequency of a single interval in the deviation frequency distribution exceeds 0.6% of the total number of historical batches for that terminal, the deviation value of that terminal is considered highly concentrated and belongs to a concentrated terminal. Terminals with a proportion not exceeding 0.6% have deviation values ​​dispersed across multiple intervals and do not belong to concentrated terminals. Concentrated terminals are merged according to their adjacency on the line. If there are no non-concentrated terminal intervals between adjacent concentrated terminals, they are merged into the same segment; if there are intervals, they are each formed into independent segments. Concentrated terminal zones are stored as segments. Deviation-dense terminal zones correspond to sections on the line where deviations repeatedly and stably recur. This concentration is mostly caused by systematic factors such as errors in the recording of the line length for that segment, rather than occasional single measurement fluctuations. The criteria for determining that terminals within the same segment are concentrated based on the deviation frequency distribution may not fall within the same interval, but this step only registers the terminal affiliation and does not retain specific deviation intervals for each terminal. For terminals with fewer than 5 historical batches in the deviation frequency distribution, the denominator in the ratio calculation still uses the actual number of historical batches for that terminal. No other criteria are used due to the small number of batches. Terminals with a ratio exceeding 0.6 are still considered concentrated terminals. The 0.6 ratio threshold is taken from the distribution boundary between concentrated and non-concentrated terminals in historical real data. Terminals exceeding this ratio exhibit a significantly stronger regularity in their deviations falling within the same interval in most batches than should show random fluctuations. There is no upper limit to the number of concentrated terminals within the same segment of a densely distributed deviation terminal zone. A segment can cover all terminals identified as concentrated terminals on that branch, provided they are physically adjacent to each other and do not contain any non-concentrated terminals in between.

[0077] A narrowing trend table is generated for terminals with dense deviations, showing the narrowing of tracking deviation values. The densely biased terminal bands are used here only to identify the tracking targets. The narrowing magnitude of each terminal is measured by the batch-to-batch standard deviation of the positioning deviation in its positioning deviation record table. A sliding window is formed using the most recent five batches, and the standard deviation of the positioning deviations stored in the positioning deviation record table is calculated window by window. The smaller the standard deviation, the more concentrated the deviation values ​​within the window and the narrower the fluctuation. If a terminal's line segment has fixed causes such as line length input errors, its calculated positioning deviation will steadily converge to a certain non-zero level as historical events accumulate, and the standard deviation of the sliding window will decrease window by window. For terminals only affected by random measurement noise, the positioning deviation is irregularly distributed among batches, and the standard deviation will not continuously decrease. Using five batches as the sliding window strikes a balance between timely reflection of recent narrowing and resistance to disturbances from single abnormal batches. The narrowing trend table records the standard deviation sequence of consecutive sliding windows for each terminal, and also records the mean of the positioning deviation within each window, with the sequence arranged chronologically according to the latest batch within the window. A negative difference in the standard deviations of two adjacent windows indicates narrowing, while a positive difference indicates expansion. When there are fewer than 5 historical batches for a terminal, making it impossible to form a complete sliding window, the narrowing trend table is changed to calculate a single standard deviation value based on all existing batches, without further sequence expansion. In the positioning deviation record table, if the standard deviation sequence of a certain terminal is 0 throughout, it is recorded according to the actual result and considered a special case of extreme narrowing. Within the same segment of a deviation-dense terminal zone, the standard deviations of the sliding windows for different terminals are calculated independently and do not affect each other. The sliding window length of 5 batches is consistent with the historical batch threshold used to determine the deviation-dense terminal zone, and the judgment basis for each stage of the same terminal does not contradict each other due to different window lengths.

[0078] A compressed feature list is established to identify terminals exhibiting unidirectional continuous narrowing based on the narrowing trend table. In the standard deviation sequence of the narrowing trend table, if the difference between adjacent windows is consecutively negative for three or more consecutive windows, the terminal's deviation is determined to be exhibiting unidirectional continuous narrowing. If the sign changes from negative to positive in the sequence, interrupting the narrowing continuity, it is not considered unidirectional continuous narrowing. All terminals determined to exhibit unidirectional continuous narrowing are added to the compressed feature list. For each terminal, the standard deviation value corresponding to its latest window, the mean of the positioning deviation within that latest window (as the convergence deviation), and the number of consecutively narrowed windows are recorded. Terminals with a standard deviation sequence of 0 throughout the narrowing trend table have no adjacent window differences for judgment. These terminals are considered to have converged to the extreme based on their constant standard deviation of 0, and are also included in the compressed feature list. The number of consecutively narrowed windows is recorded as the total number of all sliding windows for that terminal. When a terminal's historical batches are insufficient to form three consecutive sliding windows, the minimum number of windows required to determine unidirectional continuous narrowing is not met. Such terminals will not appear in the compressed feature list, but their narrowing trend table records will still be retained separately based on the existing number of windows. There is no upper limit to the number of consecutive narrowing windows in the compressed feature list. If a terminal has sufficient historical batches and its standard deviation consistently decreases, its number of consecutive narrowing windows can continue to grow, and the list will accurately record this cumulative result. For terminals whose standard deviation in the narrowing trend table first decreases continuously, then rebounds, and then begins to decrease again, the number of consecutive narrowing windows accumulated before the rebound is discarded, and the count is restarted from the first decreasing window after the rebound. The number of consecutive windows corresponding to this terminal in the compressed feature list will be recorded based on the restarted result.

[0079] Based on the compressed feature list, the fault location results are output after directional correction of the deviation nodes in the traveling wave arrival delay table. If the latest window standard deviation recorded for each terminal in the compressed feature list is lower than the fixed accuracy threshold corresponding to its line voltage level, the terminal is determined to have reached a stable level suitable for correction, and the convergence deviation of that terminal in the compressed feature list is used for correction. The delay deviation recorded for that terminal in the traveling wave arrival delay table is the deviation node that needs correction. The convergence deviation is converted into an equivalent delay correction amount according to the terminal's corrected wave velocity. This correction amount is subtracted from the delay deviation in the traveling wave arrival delay table to obtain the corrected delay deviation. The corrected delay deviation from the traveling wave arrival delay table is then substituted back into the terminal location result calculation process to obtain the corrected fault location result for that terminal. For terminals outside the compressed feature list that have not reached the threshold, their traveling wave arrival delay table records remain unchanged, and the original terminal location results are recorded. The fault location results consist of all calibrated terminals and their corrected distances, with both types of terminals recorded in the same format. Regardless of whether calibration is performed, the original time delay deviation of the terminal in the traveling wave arrival delay table is not overwritten or deleted; the corrected value is stored separately as a new value. The proportion of calibrated terminals within the same deviation and instability section increases with the number of historical batches, without a fixed target value. The final distance of each terminal in the fault location results is given as the distance along the line from "a certain number of meters from the substation outgoing switchgear," along with whether directional calibration has been performed and the original distance before calibration, for maintenance and patrol personnel to verify the formation process of the distance location results at the corresponding tower or cable section.

[0080] To implement the above-described method embodiment, a distribution network fault edge-cloud collaborative measurement method is proposed to achieve the corresponding functions and technical effects. See also... Figure 3 , Figure 3 This diagram illustrates a structural block diagram of a distribution network fault edge-cloud collaborative measurement device 300 according to an embodiment of this application. For ease of explanation, only the parts relevant to this embodiment are shown. The distribution network fault edge-cloud collaborative measurement device 300 provided in this embodiment includes:

[0081] The initial inspection and positioning module 301 is used to acquire high-frequency signals from the feeder terminal of the distribution network and perform wavelet traveling wave feature matching on the edge side based on the high-frequency signals to form a suspected fault time window.

[0082] The time delay modeling module 302 is used to perform multi-scale modulus maxima alignment based on the suspected fault time window to obtain the terminal traveling wave arrival record, to establish a traveling wave arrival time delay table based on the statistical time delay deviation of the terminal traveling wave arrival record, and to identify the deviation instability section based on the traveling wave arrival time delay table to form a traveling wave arrival disturbance band.

[0083] The cloud-based refinement module 303 is used to verify the overlap of the suspected fault time window based on the traveling wave arrival disturbance band to determine the trigger flag, upload the traveling wave arrival disturbance band to the cloud for signal-to-noise ratio tracking to determine the synchronization compression parameters, and perform ridge dispersion solution based on the synchronization compression parameters and the traveling wave arrival delay table to obtain the terminal ranging result.

[0084] The consistency verification module 304 is used to analyze the inter-terminal ranging correlation based on the terminal ranging results to obtain the terminal correlation intensity spectrum, filter the refraction and reflection difference subsets according to the terminal correlation intensity spectrum to generate a terminal fusion group, and identify the dominant reflection component and background interference component based on the terminal fusion group to determine the consistency confidence.

[0085] The directional output module 305 is used to perform weighted deviation evaluation on the terminal ranging results based on the consistency confidence level to form a positioning deviation record table, screen the positioning deviation record table for deviation convergence trend to establish a compressed feature list, and perform directional correction of the deviation nodes of the traveling wave arrival delay table based on the compressed feature list to output fault ranging results.

[0086] The aforementioned distribution network fault edge-cloud collaborative measurement device 300 can implement a distribution network fault edge-cloud collaborative measurement method according to the above method embodiments. The options in the above method embodiments are also applicable to this embodiment, and will not be detailed here. The remaining contents of this application embodiment can be referred to the contents of the above method embodiments, and will not be repeated in this embodiment.

[0087] The above embodiments are not an exhaustive list based on the present invention, and there may be many other embodiments not listed. Any substitutions and improvements made without departing from the concept of the present invention are within the protection scope of the present invention.

Claims

1. A method for collaborative measurement of fault edges and clouds in a distribution network, characterized in that, include: Acquire high-frequency signals from the feeder terminals of the distribution network, and perform wavelet traveling wave feature matching on the edge side based on the high-frequency signals to form a suspected fault time window; Based on the suspected fault time window, multi-scale modulus maxima alignment is performed to obtain the terminal traveling wave arrival record. The traveling wave arrival delay table is established by statistically analyzing the delay deviation of the terminal traveling wave arrival record. Based on the traveling wave arrival delay table, the deviation and instability section is identified to form the traveling wave arrival disturbance band. Based on the arrival disturbance band of the traveling wave, the overlap of the suspected fault time window is verified to determine the trigger flag. The arrival disturbance band of the traveling wave is uploaded to the cloud through the trigger flag for signal-to-noise ratio tracking to determine the synchronization compression parameters. Based on the synchronization compression parameters and the arrival delay table of the traveling wave, the ridge dispersion is calculated to obtain the terminal ranging result. Based on the terminal ranging results, the cross-correlation between the ranging of the terminals is analyzed to obtain the terminal correlation intensity spectrum. Based on the terminal correlation intensity spectrum, a subset of refraction and reflection differences is selected to generate a terminal fusion group. Based on the terminal fusion group, the dominant reflection component and the background interference component are identified to determine the consistency confidence. Based on the consistency confidence level, a weighted deviation assessment is performed on the terminal ranging results to form a positioning deviation record table. The positioning deviation record table is then screened for deviation convergence trends to establish a compressed feature list. Based on the compressed feature list, the deviation nodes of the traveling wave arrival delay table are oriented and corrected to output the fault ranging results.

2. The method according to claim 1, characterized in that, The step of identifying the deviation from the instability section and forming the traveling wave arrival disturbance band based on the traveling wave arrival delay table includes: The travel wave arrival delay table is sorted by sampling batches and the delay deviation is arranged in a rolling sequence to obtain the deviation rolling sequence; The terminal fluctuation intensity spectrum is generated by comparing the positive and negative deviation amplitude differences of the deviation rolling sequence. The deviation from the instability zone is determined by removing periodic attributable fluctuations from the terminal fluctuation intensity spectrum; Based on the statistical fluctuation range and distribution location of the aforementioned deviation and instability section, a traveling wave arrival disturbance zone is formed.

3. The method according to claim 1, characterized in that, The step of obtaining the terminal ranging result by performing ridge dispersion calculation based on the synchronous compression parameters and the traveling wave arrival delay table includes: Based on the aforementioned synchronous compression parameters, ridge-fidelity weighted sampling is performed to form a compressed sampling sequence; The measured dispersion value is determined by tracing the ridge frequency change trajectory using the compressed sampling sequence. Wave velocity dispersion analysis is performed on the measured dispersion values ​​to generate dispersion correction coefficients; Based on the dispersion correction coefficient, the wave velocity is converted into the arrival delay table of the traveling wave to obtain the terminal ranging result.

4. The method according to claim 1, characterized in that, The step of analyzing the inter-terminal ranging cross-correlation based on the terminal ranging results to obtain the terminal association strength spectrum includes: The terminal ranging results are used to perform time-series trajectory parsing to generate a terminal trajectory group; Based on the analysis of the terminal trajectory group, the correlation coefficient between each terminal trajectory is obtained. Based on the terminal correlation coefficient, common mode trends are eliminated, discrete clustering features are extracted, and a correlation strength distribution is formed. The terminal association strength spectrum is obtained by delineating the strong and weak association terminal boundaries from the association strength distribution.

5. The method according to claim 1, characterized in that, The determination of consistency confidence based on the identification of the dominant reflection component and the background interference component by the terminal fusion group includes: The branch association strength distribution is obtained by splitting the terminal fusion group according to the branch; Valley segmentation is performed on the branch correlation intensity distribution to obtain the dominant reflection component; The background interference component is obtained by secondary clustering of the dominant reflection component according to the reflection intensity to divide it into low-intensity categories. The consistency confidence level is determined by comparing the energy proportions of the dominant reflection component and the background interference component.

6. The method according to claim 1, characterized in that, The step of screening the positioning deviation record table for deviation convergence trend and establishing a compressed feature list includes: The frequency of deviation value recurrence is calculated using the positioning deviation record table to form a deviation frequency distribution; Based on the frequency distribution of the deviation, the range of terminals where the deviation values ​​are concentrated is delineated to obtain the deviation-dense terminal zone; A narrowing trend table is generated for the narrowing of tracking deviation values ​​in the terminal band with dense deviations; Based on the narrowing trend table, a compression feature list is established to identify terminals that continuously narrow in one direction.

7. The method according to claim 1, characterized in that, The step of establishing a traveling wave arrival delay table based on the statistical delay deviation of the terminal traveling wave arrival record includes: An arrival estimation comparison table is obtained by extracting the arrival time and estimation based on the terminal traveling wave arrival record; The arrival valuation comparison table is normalized and corrected for forward and reverse path drop differences to generate a forward and reverse path drop difference map; The cumulative elevation difference set is obtained by extracting the cumulative elevation difference deviation using the aforementioned forward and reverse path elevation difference maps. The accumulated drop set is used to establish a traveling wave arrival delay table based on the terminal summary delay deviation.

8. The method according to claim 3, characterized in that, The step of determining the measured dispersion value by tracing the ridge frequency change trajectory using the compressed sampling sequence includes: Based on the compressed sampling sequence, the instantaneous frequencies of the ridge lines of each frequency band are extracted to form a frequency band trajectory group; For the frequency trajectory group of the frequency band, the arrival time offset of adjacent frequency bands is calculated to obtain the frequency band dispersion offset set; Based on the frequency band dispersion offset set, track the dispersion stability persistence characteristics and output dispersion stability records; The measured dispersion values ​​are determined by filtering the dispersion stability records according to frequency bands.

9. The method according to claim 5, characterized in that, The step of segmenting the branch correlation intensity distribution to obtain the dominant reflection component includes: Intensity gradient spectrum is generated by calculating the intensity gradient using the associated intensity distribution of the branch lines. Based on the intensity gradient spectrum, the zero-crossing points where the gradient changes from negative to positive are located, forming a set of valley points; Based on the valley point set, the branch correlation intensity distribution is segmented to generate a candidate reflection component set; The dominant reflection component is obtained by performing a proximity continuity check on the candidate reflection component set.

10. A distribution network fault edge-cloud collaborative measurement device, characterized in that, include: The initial inspection and location module is used to acquire high-frequency signals from the feeder terminals of the distribution network, and to perform wavelet traveling wave feature matching based on the high-frequency signals at the edge side to form a suspected fault time window; The time delay modeling module is used to perform multi-scale modulus maxima alignment based on the suspected fault time window to obtain the terminal traveling wave arrival record, to establish a traveling wave arrival time delay table based on the statistical time delay deviation of the terminal traveling wave arrival record, and to identify the deviation instability section based on the traveling wave arrival time delay table to form a traveling wave arrival disturbance band. The cloud-based refinement module is used to verify the overlap of the suspected fault time window based on the traveling wave arrival disturbance band to determine the trigger flag, upload the traveling wave arrival disturbance band to the cloud for signal-to-noise ratio tracking to determine the synchronization compression parameters, and perform ridge dispersion calculation based on the synchronization compression parameters and the traveling wave arrival delay table to obtain the terminal ranging result. The consistency verification module is used to analyze the inter-terminal ranging correlation based on the terminal ranging results to obtain the terminal correlation intensity spectrum, filter the refraction and reflection difference subsets based on the terminal correlation intensity spectrum to generate a terminal fusion group, and identify the dominant reflection component and background interference component based on the terminal fusion group to determine the consistency confidence. The directional output module is used to perform weighted deviation evaluation on the terminal ranging results based on the consistency confidence level to form a positioning deviation record table, screen the positioning deviation record table for deviation convergence trend to establish a compressed feature list, and perform directional correction of the deviation nodes of the traveling wave arrival delay table based on the compressed feature list to output fault ranging results.

Images