High-voltage transmission line fault detection and identification method

Abstract

The invention belongs to the technical field of fault detection, and relates to a high-voltage transmission line fault detection and identification method. The method comprises the following steps of: dividing candidate fault sections according to a topological structure by synchronously acquiring steady-state and transient-state traveling wave signals of key nodes of a power grid; identifying a fault by using a multi-feature fusion criterion and extracting traveling wave features; executing mode recognition-based steady-state section judgment and multi-terminal ranging and polarity verification-based transient positioning in parallel to generate two types of positioning result sets; and finally, carrying out intelligent fusion judgment on the two types of results under topological constraints. According to the method, the problems that a traditional single-point signal analysis method cannot determine a fault section, lacks space positioning capability and is insufficient in adaptability are solved, section-level accurate positioning of the fault is realized, and the positioning accuracy, robustness and fault processing efficiency are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of fault detection technology and relates to a method for fault detection and identification of high-voltage transmission lines. Background Technology

[0002] As a crucial component of the power system, high-voltage transmission lines face increasingly higher demands on fault detection and location technologies due to the continuous expansion of the power system and the deepening of smart grid construction. Traditional methods often rely on analyzing electrical signals, such as current time series, collected from a single monitoring node in the line, and detecting faults by identifying abnormal characteristics in the signals.

[0003] A typical existing solution, such as the fault monitoring system for high-voltage lines in electrical engineering proposed in Chinese Invention Patent Publication No. CN119335299B, improves the histogram-based anomaly detection algorithm by analyzing information such as surge characteristics, harmonic interference, and data mutations in current signals, thereby enhancing the accuracy of identifying abnormal states of high-voltage lines in mines.

[0004] However, this type of method based on single-point signal analysis still has the following main drawbacks: First, the method only evaluates the anomalies in the current time-series data of a single monitoring node, and its judgment is essentially whether a fault exists, rather than where the fault is located. In actual power grids with multi-branch structures, because the grid topology is not integrated, it is impossible to map electrical anomalies to specific physical line sections, resulting in a lack of spatial targeting in fault isolation operations.

[0005] Secondly, the core of this method lies in feature extraction and anomaly scoring of the electrical quantities of a single monitoring node, without involving the synchronous acquisition and joint analysis of signals generated by the transient processes of multiple monitoring nodes. Therefore, it cannot use time-based signal propagation differences to achieve distance estimation, nor can it use the spatial comparison relationship of the initial polarity of the signal to determine the fault direction, and is essentially unable to spatially locate the fault point.

[0006] Although this method introduces features such as harmonics and surges to improve anomaly scoring, it still belongs to the single-stage discrimination mode of feature-scoring. It has the defects of fixed criteria and weak adaptability, making it difficult to adapt to the complex changes of different fault types and power grid operation modes. Summary of the Invention

[0007] In view of this, in order to solve the problems mentioned in the background technology, a fault detection and identification method for high-voltage transmission lines is proposed.

[0008] The objective of this invention can be achieved through the following technical solution: a high-voltage transmission line fault detection and identification method, comprising: S1, acquiring steady-state electrical quantities and transient traveling wave signals synchronously collected by monitoring devices deployed at key nodes of the power grid topology, and recording the absolute arrival time of each signal.

[0009] S2. Based on steady-state electrical quantities, fault identification is performed through multi-feature fusion criteria, and after fault identification, time-frequency distribution features and wavefront polarity features are extracted from transient traveling wave signals.

[0010] S3. Based on the power grid topology, divide the candidate fault sections and perform the following location processing on each section: S31. Extract the steady-state feature quantities of the section based on the current phasors of each node after the fault, and identify the fault state of the section through a pre-trained pattern recognition model to generate the first location result set.

[0011] S32. Based on the absolute time of arrival, time-frequency distribution characteristics, and wavefront polarity characteristics, perform multi-terminal ranging and polarity verification to obtain effective ranging results, and perform cross-validation to generate a second positioning result set.

[0012] S4. Under the constraints of the power grid topology, perform a fusion judgment on the first location result set and the second location result set to output the final fault section.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention divides candidate fault sections according to the power grid topology and integrates the multi-source steady-state and transient electrical quantities synchronously collected by key nodes for analysis, so as to accurately associate electrical signal anomalies with specific line sections. This method solves the defect of the traditional scheme that can only determine whether there is a fault but cannot determine where the fault is, and realizes the section-level positioning of the fault location, which provides a clear operational target for subsequent rapid fault isolation and power restoration, and significantly improves the pertinence and efficiency of fault handling.

[0014] (2) This invention synchronously acquires transient traveling wave signals from multiple key nodes and comprehensively utilizes their absolute time difference of arrival, wavefront polarity relationship, and time-frequency distribution characteristics to perform multi-terminal ranging and cross-validation. This method overcomes the limitations of relying solely on single-point time-series signal analysis. By utilizing the propagation characteristics and direction information of signals in space, it achieves spatial distance estimation and direction determination of fault points, thereby endowing the fault detection system with essentially precise positioning capabilities and greatly reducing the fault tracing range.

[0015] (3) This invention identifies faults by constructing a criterion that integrates multi-dimensional features such as instantaneous rate of change, harmonics, and zero-sequence components, and introduces a pattern recognition model trained on historical data to intelligently identify the steady-state characteristics of the section. This method solves the problems of traditional methods having single features and being unable to flexibly cope with complex fault types and changes in operating modes. Through multi-level and multi-criteria fusion analysis, it significantly improves the accuracy and robustness of fault identification and section discrimination, and reduces the false alarm and false negative rates. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the steps of a high-voltage transmission line fault detection and identification method according to the present invention.

[0018] Figure 2 This is a flowchart illustrating the specific method S2 in this invention.

[0019] Figure 3 This is a flowchart illustrating the specific method of S31 in this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 As shown, the present invention provides a fault detection and identification method for high-voltage transmission lines, including: S1, acquiring steady-state electrical quantities and transient traveling wave signals synchronously collected by monitoring devices deployed at key nodes of the power grid topology, and recording the absolute arrival time of each signal.

[0022] When a fault occurs in a high-voltage transmission line, the symmetry and stability of its electrical system are disrupted. This change is directly reflected in the abrupt changes in the amplitude, phase, harmonics, and sequence components of steady-state electrical quantities such as voltage and current.

[0023] Therefore, by deploying synchronous phasor measurement units at key nodes of the power grid, such as substation busbars, both ends of lines, and T-junctions, three-phase current waveform data and three-phase voltage waveform data are collected synchronously as steady-state electrical quantities.

[0024] Meanwhile, the electromagnetic transient process generated at the moment of the fault will propagate to both ends of the line in the form of a traveling wave. The arrival time of the wavefront and the waveform characteristics of the transient traveling wave signal carry the precise location information of the fault point. Therefore, the transient traveling wave signal can be captured from the above three-phase current waveform data and three-phase voltage waveform data through signal processing technology.

[0025] All monitoring devices are connected to a unified time reference source and undergo time synchronization calibration to eliminate inherent time deviations between devices.

[0026] In one specific embodiment, the specific method of S1 is as follows: the monitoring device deployed at the key nodes of the power grid topology synchronously collects three-phase current waveform data and three-phase voltage waveform data at a uniform sampling rate (e.g., 1MHz), performs time synchronization calibration processing, and uses the processed three-phase current waveform data and three-phase voltage waveform data as steady-state electrical quantities.

[0027] Transient traveling wave signals are captured from steady-state electrical quantities using wavelet transform technology. A wavefront detection algorithm is applied to the transient traveling wave signal of each node, and the corresponding absolute time reference value is read and recorded as the absolute arrival time of the node.

[0028] Please see Figure 2 As shown, S2, based on steady-state electrical quantities, fault identification is performed through multi-feature fusion criteria, and after the fault is identified, time-frequency distribution features and wavefront polarity features are extracted from the transient traveling wave signal.

[0029] Considering that fault identification based on single electrical quantity characteristics may be unable to fully cover different types of line faults such as short circuits and grounding due to noise interference or changes in operating conditions, and that transient traveling wave analysis needs to be started after the fault is confirmed to save computing resources.

[0030] During the propagation of traveling waves, dispersion usually occurs, causing the wave velocity to change with frequency. Therefore, by extracting the time-frequency distribution characteristics that reflect the energy concentration band and time-frequency distribution characteristics of traveling waves, we can use them as the basis for subsequent wave velocity correction.

[0031] Meanwhile, since the directional information of the fault location relative to each monitoring node is the core of distinguishing fault sections and verifying the physical rationality of the ranging results, the wavefront polarity features that can characterize the initial traveling wavefront direction relationship of each monitoring node are extracted for the preliminary identification of fault sections and the physical verification of ranging results.

[0032] In one specific embodiment, the specific method of S2 is as follows: perform real-time analysis on steady-state electrical quantities, calculate the instantaneous rate of change and harmonic content of each phase current and voltage, as well as the amplitude and phase of zero-sequence current and zero-sequence voltage, and combine them into a multi-dimensional feature vector.

[0033] The fault is determined based on preset fault characteristic conditions, which include short-circuit fault characteristic sub-conditions and ground fault characteristic sub-conditions.

[0034] For example, the characteristic sub-conditions for short-circuit faults may include: the instantaneous values ​​of at least two phase currents exceed a set multiple of the rated current, the corresponding phase voltage drops significantly, and the total harmonic content increases suddenly; the characteristic sub-conditions for ground faults may include: the amplitude of the zero-sequence current exceeds the normal operating threshold, the zero-sequence voltage rises simultaneously, and the phase relationship between the zero-sequence current and the zero-sequence voltage conforms to the characteristic angle of ground faults.

[0035] During the discrimination process, if the multidimensional feature vector satisfies the short-circuit fault feature sub-condition or the ground fault feature sub-condition, it is determined that a line fault of the corresponding type has occurred.

[0036] After determining that a line fault has occurred, time-frequency analysis is performed on the transient traveling wave signal, and the energy distribution entropy value within the preset frequency band is extracted as the time-frequency distribution feature.

[0037] Specifically, the method for obtaining the time-frequency distribution characteristics is as follows: the transient traveling wave signal is processed using time-frequency analysis techniques such as short-time Fourier transform to obtain the energy distribution of the signal in the time-frequency domain. The preset frequency band usually covers the main high-frequency band where the energy of the traveling wave signal is concentrated, for example, 5kHz to 200kHz.

[0038] Subsequently, within the preset frequency band, the frequency band is divided into multiple sub-bands, either uniformly or non-uniformly, and the energy value of each sub-band is calculated as a proportion of the total energy of the preset frequency band.

[0039] The energy percentage of each sub-band is considered as a probability value. By calculating the probability entropy of the energy percentage of all sub-bands, a scalar value representing the uniformity of energy distribution is obtained. This value is the energy distribution entropy value and is used as a time-frequency distribution feature.

[0040] The polarity sign of the initial current traveling wavefront at each monitoring node is obtained, and the wavefront polarity feature is constructed based on the comparison relationship of wavefront polarity signs between different nodes.

[0041] Furthermore, the specific method for constructing the wavefront polarity feature is as follows: based on the initial current traveling wave waveform at each monitoring node, the polarity sign of the traveling wave front is determined by detecting the direction of change of the instantaneous current value at the starting point of the traveling wave. For example, the rising edge is denoted as "+", and the falling edge is denoted as "-".

[0042] For any two adjacent monitoring nodes in the power grid topology, compare their wavefront polarity signs: when the wavefront polarity signs of the two nodes are the same, they are marked as having the same polarity; when the wavefront polarity signs of the two nodes are opposite, they are marked as having opposite polarity.

[0043] According to the power grid topology connection order, the polarity relationship of all adjacent node pairs is arranged sequentially to form a polarity comparison sequence. Based on the distribution pattern of same polarity and opposite polarity relationships in the polarity comparison sequence, a matrix reflecting the polarity characteristics of the wavefront is constructed.

[0044] S3. Divide the candidate fault sections according to the power grid topology and perform location processing on each section.

[0045] Considering the complex topology of high-voltage transmission lines, direct global positioning is inefficient and inaccurate, and the reliability of positioning results from a single dimension is insufficient.

[0046] Therefore, candidate fault sections are divided using power grid topology information, and steady-state characteristic localization and transient traveling wave localization are performed on each section to generate a first localization result set and a second localization result set, respectively.

[0047] The division of candidate fault sections is based on the power grid topology, with key nodes where monitoring devices are deployed as section boundary nodes, and the line sections between adjacent key nodes as units, thus dividing the area into multiple candidate fault sections.

[0048] Please see Figure 3 As shown, S31, based on the current phasors of each node after the fault, the steady-state feature quantities of the section are extracted, and the section fault state is identified through a pre-trained pattern recognition model to generate the first localization result set.

[0049] In one specific embodiment, the specific method of S31 is as follows: after determining that a line fault has occurred, considering that the current phasor relationship between the inside and outside of the fault section follows Kirchhoff's current law and there is an essential difference, the synchronously collected current phasor corresponding to the boundary nodes at both ends of each candidate fault section is obtained, the vector sum of the current phasors at both ends of the section is calculated as the differential current phasor, and the vector difference is calculated as the braking current phasor.

[0050] The differential current phasor reflects the total fault current flowing into the candidate fault section, while the braking current phasor reflects the degree of current imbalance at both ends of the section, and its value is larger when there is a fault outside the section.

[0051] Polar coordinate transformations were performed on the differential current phasor and the braking current phasor respectively, and their amplitude ratio and phase difference were extracted as steady-state characteristic quantities of the section. These characteristic quantities quantitatively describe the electrical response mode of the section under fault conditions.

[0052] The steady-state characteristics of the section are input into a pattern recognition model trained based on historical fault data for discrimination.

[0053] The pattern recognition model divides the feature space into internal fault clustering regions and external fault clustering regions. Specifically, it collects historical line fault data, extracts historical steady-state feature quantities, and, based on actual fault records and on-site verification information, labels each historical sample to determine whether its corresponding candidate fault segment is a real fault segment, thus forming a sample label.

[0054] For example, if a historical fault is confirmed to have occurred in section A through line inspection, then all steady-state characteristic quantity samples corresponding to section A extracted from the waveform data of this fault are marked as internal faults (positive samples), while steady-state characteristic quantity samples corresponding to other non-faulted sections (such as sections B and C) at the same time are marked as external faults (negative samples).

[0055] An internal fault sample set is constructed using samples marked as actual fault segments, and an external fault sample set is constructed using samples marked as non-fault segments.

[0056] Supervised machine learning algorithms, such as support vector machines, random forests, or gradient boosting decision trees, are used to train on internal and external fault sample sets, with historical steady-state features as input features and sample labels as training targets.

[0057] After training, the model will form a classification boundary in the feature space. The region consisting of the coordinates of the points that are judged to be faults located inside the segment is defined as the internal fault cluster region; otherwise, it is defined as the external fault cluster region.

[0058] If the projection point of the steady-state characteristic quantity of a section in the characteristic space falls into the internal fault clustering region, it indicates that the electrical response mode of the section is highly similar to the historical internal fault mode, and the fault is determined to be located inside the candidate section; otherwise, it indicates that its mode is closer to external faults or normal operation, and the fault is determined not to be within the candidate section.

[0059] Finally, the judgment results of all candidate fault segments are summarized, and the segments that are determined to be faulty inside are included in the first location result set.

[0060] S32. Based on the absolute time of arrival, time-frequency distribution characteristics, and wavefront polarity characteristics, perform multi-terminal ranging and polarity verification to obtain effective ranging results, and perform cross-validation to generate a second positioning result set.

[0061] In one specific embodiment, the effective ranging result is obtained by: calculating the propagation time difference between any pair of nodes based on the absolute arrival time of the transient traveling wave signal recorded by each monitoring node, and calculating the estimated distance from the fault location to each node in combination with the traveling wave propagation speed.

[0062] Considering the dispersion effect of traveling waves in transmission lines, i.e., different frequency components have different propagation speeds, using a fixed wave speed would lead to ranging errors. Therefore, it is necessary to use time-frequency distribution characteristics for wave speed correction.

[0063] The specific method is as follows: by analyzing the peak value of the energy distribution entropy, the dominant frequency band with the most concentrated energy in the transient traveling wave signal of this fault is identified. Based on the frequency-wave velocity correspondence established in advance through simulation, i.e., the preset dispersion curve, the optimal propagation speed corresponding to the dominant frequency band is queried and determined. The above preliminary distance estimate is recalculated and corrected using this optimal propagation speed to obtain the corrected distance estimate.

[0064] Using the corrected distance estimate, geometric positioning can be performed using two- or three-end ranging formulas to infer the possible positional relationship of the fault location relative to each monitoring node. For example, the fault location may be on the line segment between node A and node B.

[0065] Based on the traveling wave propagation theory, when the fault point is located between two monitoring nodes, the initial current traveling wave generated by the fault will propagate towards these two nodes, resulting in opposite polarities of the initial wavefronts detected by the two nodes. Therefore, a polarity criterion is established based on this theory to verify the physical rationality of the ranging results.

[0066] The specific verification method is as follows: compare the fault location relationship inferred from the geometric positioning with the extracted wavefront polarity features. For all node pairs inferred to be where the fault point is located, check whether the polarity relationship recorded in the wavefront polarity features of the node pair is an antipolarity relationship.

[0067] If the inferred location matches the polarity relationship (i.e., it conforms to anti-polarity), then the ranging result is physically consistent and should be retained. If it does not match, for example, if the inferred fault is between two points but the two points have the same polarity, then the preliminary ranging result may be unreliable due to wavefront identification errors, reflected wave interference, or other reasons, and should be discarded. The ranging results retained through this verification are considered valid ranging results.

[0068] Furthermore, since multiple effective ranging results are obtained from multiple pairs of different nodes, they are independent of each other and may have slight deviations, so they need to be fused to obtain the optimal positioning estimate.

[0069] Specifically, the cross-validation method is as follows: for all valid distance measurement results, i.e. the sum or difference of the distances from the fault point to the two nodes calculated, a topological association graph is constructed with the monitoring node as the vertex and the distance estimate as the edge weight. Graph theory optimization algorithms, such as least squares estimation-based optimization algorithms, are applied to iteratively solve the problem with the objective of minimizing the overall variance between all edge weights (observed values) in the graph and the theoretical distance value calculated from a hypothetical single fault point.

[0070] This process can select the optimal combination of node pairs that best makes all observations self-consistent, and simultaneously estimate the distance from the fault point to each node.

[0071] During the optimization process, the confidence level of the distance estimate for each node is calculated. The specific method for calculating the confidence level is as follows: based on the optimal location of the fault point obtained from the optimization model, the theoretical value of each effective distance measurement result is calculated backward. The residual of each effective distance measurement result is defined as the absolute value of the difference between its measured value and the theoretical value.

[0072] The confidence level is set to be inversely proportional to the square of the residuals, that is, the smaller the residuals, the higher the confidence level of the ranging result. The effective ranging results are weighted and fused according to the confidence level to generate the final fault location estimate in each candidate fault segment.

[0073] Specifically, the confidence scores of each effective ranging result are normalized, and the normalized confidence scores are directly used as the weight coefficients for each result in the weighted fusion. Subsequently, multiple effective ranging results located in the same candidate segment are weighted and averaged according to their weights, and the result is the final fault location estimate for that segment.

[0074] By traversing all candidate segments, segments whose final fault location estimates fall within their physical length range and whose estimates are generated by polarity-verified data fusion are identified as potential fault segments. All these potential fault segments and their estimated internal locations are then aggregated to form the second location result set. This result set provides physically verified and data-fused location information based on transient traveling waves.

[0075] It should be noted that if the signal-to-noise ratio of the transient traveling wave signal is lower than the preset threshold (e.g., 10dB), then S32 is skipped and the first positioning result set is used directly as the final output.

[0076] S4. Under the constraints of the power grid topology, perform a fusion judgment on the first location result set and the second location result set to output the final fault section.

[0077] In a specific embodiment, the specific method of S4 is as follows: under the constraints of the power grid topology, based on the first positioning result set and the second positioning result set, the following fusion judgment logic is executed: (i) if there is a candidate fault section that is included in both the first positioning result set and the second positioning result set, it means that the section is determined to be a possible fault section under both steady-state feature identification and transient traveling wave positioning, and then the section is directly determined and output as a fault section.

[0078] (ii) If there is no common section as described above, the candidate fault sections with the highest confidence are selected from the first location result set and the second location result set and denoted as section A and section B, respectively. Based on the power grid topology such as the node connection diagram, the distance between these two sections is calculated, for example, by the number of line sections in the middle or the physical distance.

[0079] If the distance is less than the preset topology proximity threshold, it means that although the specific segments located by the two methods are not exactly the same, the possible fault locations indicated are adjacent in the electrical topology, and the fault is likely to actually occur in that adjacent area or near the connection point of the two segments.

[0080] The topological proximity threshold is set according to the actual situation, and is usually set to be adjacent or with an interval of no more than two segments.

[0081] In this case, segment A and segment B, as well as all segments in between, are merged into a continuous extended fault segment, which is then output as the final fault segment.

[0082] Otherwise, this indicates that the most likely fault locations of the two methods are electrically far apart, and the automatic fusion is deemed to have failed. An instruction to locate the fault is output and request manual intervention is given, leaving the final decision to the operators.

[0083] It should be noted that if both the first and second location result sets are empty sets, the region with the highest similarity is output as the final fault segment based on the similarity between the feature vector of fault identification in S2 and the historical fault mode; if there is no region that meets the conditions, an instruction to fail the location and request manual intervention is output.

[0084] In summary, this invention first acquires steady-state electrical quantities and transient traveling wave signals through synchronous monitoring; then, it reliably identifies faults and triggers subsequent analysis by using multi-feature fusion criteria of steady-state electrical quantities, while extracting time-frequency features and polarity features from the transient traveling wave for precise positioning.

[0085] Next, candidate segments are divided according to the power grid topology, and segment state discrimination based on steady-state differential principle and pattern recognition and transient positioning based on multi-terminal traveling wave ranging, polarity verification and data fusion are performed in parallel to generate preliminary positioning result sets.

[0086] Finally, under the physical constraints of the power grid topology, the two result sets are intelligently fused and judged. The unique and reliable final fault section is output through the intersection priority and proximity merging rules, or manual intervention is requested when the conclusions conflict.

[0087] This method achieves full-process automation and high reliability from fault detection, feature extraction, parallel localization to intelligent decision-making by integrating multi-dimensional information such as steady-state and transient states, electrical quantities and topological relationships.

[0088] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0089] Those skilled in the art will recognize that the algorithmic steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0090] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0091] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0092] Finally, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for detecting and identifying faults in a high voltage transmission line, characterized in that, The method comprises the following steps: S1, acquiring steady-state electrical quantities and transient traveling wave signals synchronously collected by monitoring devices arranged at key nodes of a power grid topology, and recording absolute arrival times of each signal; S2, based on the steady-state electrical quantities, identifying a fault through a multi-feature fusion criterion, and extracting time-frequency distribution features and wave head polarity features from the transient traveling wave signals after identifying the fault; S3, dividing candidate fault sections according to the power grid topology, and performing the following positioning processing on each section: S31, extracting section steady-state feature quantities based on current phasors of each node after the fault, and identifying a section fault state through a pre-trained pattern recognition model to generate a first positioning result set; S32, based on the absolute arrival times, the time-frequency distribution features and the wave head polarity features, performing multi-terminal ranging and polarity verification to obtain effective ranging results, and cross-verifying the effective ranging results to generate a second positioning result set; S4, under the constraint of the power grid topology, performing fusion judgment on the first positioning result set and the second positioning result set to output a final fault section.

2. A method of high voltage transmission line fault detection and identification as claimed in claim 1, wherein, The specific method of S1 is as follows: Synchronously collecting three-phase current waveform data and three-phase voltage waveform data through monitoring devices arranged at key nodes of a power grid topology, and performing time synchronization calibration processing; Taking the processed three-phase current waveform data and three-phase voltage waveform data as steady-state electrical quantities; Capturing transient traveling wave signals from the steady-state electrical quantities through wavelet transform technology; Recording times of the transient traveling wave signals arriving at each monitoring node as absolute arrival times.

3. A method of high voltage transmission line fault detection and identification as claimed in claim 1, wherein, The specific method of S2 is as follows: Performing real-time analysis on the steady-state electrical quantities, calculating instantaneous change rates, harmonic contents of each phase current and voltage, and amplitudes and phases of zero-sequence current and zero-sequence voltage; Combining the instantaneous change rates, the harmonic contents, the zero-sequence current amplitudes, the zero-sequence current phases, the zero-sequence voltage amplitudes and the zero-sequence voltage phases into a multi-dimensional feature vector; According to a pre-set fault feature condition, when the multi-dimensional feature vector meets a short-circuit fault feature sub-condition or a ground fault feature sub-condition, it is determined that a line fault occurs; After determining that a line fault occurs, performing time-frequency analysis on the transient traveling wave signals to extract energy distribution entropy values of the transient traveling wave signals in a pre-set frequency band range as time-frequency distribution features; Acquiring polarity signs of initial current traveling wave wave heads at each monitoring node, and constructing a wave head polarity feature based on a comparison relationship of the polarity signs between different nodes.

4. A method of high voltage transmission line fault detection and identification as claimed in claim 3, wherein, The specific construction method of the wave head polarity feature is as follows: According to initial current traveling wave waveforms at each monitoring node, determining polarity signs of the traveling wave wave heads by detecting change directions of current instantaneous values at starting points of the traveling waves; For any two adjacent monitoring nodes in the power grid topology, comparing the polarity signs of the wave heads of the two nodes: when the polarity signs of the wave heads of the two nodes are the same, marking as a same polarity relationship; when the polarity signs of the wave heads of the two nodes are opposite, marking as an opposite polarity relationship; According to a connection order of the power grid topology, arranging polarity relationships of all adjacent node pairs in sequence to form a polarity comparison sequence; Based on distribution modes of the same polarity relationships and the opposite polarity relationships in the polarity comparison sequence, constructing a wave head polarity feature matrix.

5. A method of high voltage transmission line fault detection and identification as claimed in claim 1, wherein, The specific method of dividing candidate fault sections according to the power grid topology is as follows: The key nodes of the monitoring device are taken as section boundary nodes according to the power grid topology structure, and the line sections between adjacent key nodes are taken as units to divide into a plurality of candidate fault sections.

6. A method of high voltage transmission line fault detection and identification as claimed in claim 1, wherein, The specific method of S31 is: After determining that a line fault occurs, the current phasors synchronously collected by the boundary nodes at both ends of each candidate fault section are obtained, the vector sum of the current phasors at both ends of the section is calculated as a differential current phasor, and the vector difference thereof is calculated as a braking current phasor; The polar coordinate transformation is performed on the differential current phasor and the braking current phasor respectively, and the amplitude ratio and the phase difference thereof are extracted as section steady-state characteristic quantities; The section steady-state characteristic quantities are input into a pattern recognition model trained based on historical fault data for discrimination, wherein the pattern recognition model divides a feature space into an internal fault clustering area and an external fault clustering area; If the projection point of the section steady-state characteristic quantities in the feature space falls into the internal fault clustering area, it is determined that the fault is located inside the candidate section; otherwise, it is determined that the fault is not inside the candidate section; The discrimination results of all candidate fault sections are summarized to generate a first positioning result set.

7. A method of high voltage transmission line fault detection and identification as claimed in claim 6, wherein, The pattern recognition model divides the feature space into an internal fault clustering area and an external fault clustering area, specifically: The historical steady-state characteristic quantities are extracted from the collected historical line fault data, and according to the actual fault records and the field verification information, each historical sample is marked whether the corresponding candidate fault section is a real fault section to form a sample label; The samples marked as real fault sections are used to construct an internal fault sample set, and the samples marked as non-fault sections are used to construct an external fault sample set; A supervised machine learning algorithm is used to train the internal fault sample set and the external fault sample set with the historical steady-state characteristic quantities as input features and the sample labels as training targets; After the training is completed, the region composed of the coordinate points judged as the fault located inside the section is defined as the internal fault clustering area, otherwise, it is defined as the external fault clustering area.

8. A method of high voltage transmission line fault detection and identification as claimed in claim 1, wherein, The method for obtaining the effective ranging result is: Based on the absolute arrival time of the transient traveling wave signals recorded by each monitoring node, the propagation time difference between any node pair is calculated, and the distance estimation value of the fault occurrence point to each node is calculated in combination with the traveling wave propagation speed; The peak value of the energy distribution entropy is analyzed by using the time-frequency distribution characteristics to identify the dominant frequency band of the transient traveling wave signal energy, the best propagation speed corresponding to the dominant frequency band is determined according to the preset dispersion curve, and the distance estimation value is corrected by using the best propagation speed; The position relationship of the fault occurrence point relative to each monitoring node is inferred by geometric positioning by using the corrected distance estimation value, and a polarity criterion is established based on the traveling wave propagation theory as follows: If the fault occurrence point is inferred to be located on the line segment between two monitoring nodes according to the corrected distance estimation value, the initial current traveling wave head polarities detected by the nodes should theoretically have an inverse polarity relationship; The inferred fault position relationship and the extracted wave head polarity characteristics are compared, and for all node pairs whose fault points are inferred to be located therebetween, it is checked whether the wave head polarity characteristics conform to the inverse polarity relationship. If not, the corresponding ranging result is rejected, and the ranging result passing the verification is kept as a valid ranging result.

9. A method of high voltage transmission line fault detection and identification as claimed in claim 1, wherein, The specific method of the cross-validation is: A topological correlation graph is constructed with the monitoring nodes as the vertices and the distance estimates as the edge weights, a graph theory algorithm is applied, the optimal node pair combination is screened out with the minimum overall variance of all edge weights in the graph as the optimization objective, and the confidence of each node distance estimate is calculated; The valid ranging results are weighted and fused according to the confidence to generate the final fault location estimate in each candidate fault section; The final fault location estimates of all candidate sections are summarized, the section whose estimate is located in its physical range and passes the polarity verification is determined as a potential fault section, and a second positioning result set is formed.

10. A method of high voltage transmission line fault detection and identification as claimed in claim 1, wherein, The specific method of the S4 is: According to the constraints of the power grid topology, the following fusion judgment logic is performed on the first positioning result set and the second positioning result set: (i) If there is a candidate fault section included in both the first positioning result set and the second positioning result set, the section is directly determined and output as the fault section; (ii) If there is no such common section, the candidate fault sections with the highest confidence are selected from the first positioning result set and the second positioning result set respectively, the distance between the two sections is calculated based on the power grid topology; If the distance is less than a preset topological proximity threshold, the two sections are merged into a continuous fault section, and the final fault section is output; Otherwise, an indication of positioning failure and a request for manual intervention are output.

Images