Accurate distribution network fault positioning method integrating traveling wave distance measurement and feeder terminal data

Abstract

The invention discloses a distribution network fault accurate positioning method fusing traveling wave distance measurement and feeder terminal data. The method comprises the following steps: collecting traveling wave signals and feeder terminal data when a power distribution network has a fault; extracting traveling wave characteristic parameters from the traveling wave signals, extracting time-frequency domain characteristic parameters from the feeder terminal data, and constructing a comprehensive characteristic vector through the traveling wave characteristic parameters and the time-frequency domain characteristic parameters; a fuzzy C-means clustering algorithm is adopted to classify the comprehensive feature vectors, and fault types are identified; performing weighted fusion on the traveling wave distance measurement result and the feeder terminal distance measurement result according to a fusion weight to obtain a fusion fault distance; and correcting the fused fault distance through a fault type to obtain a final fault distance, and calculating a fault point according to the final fault distance. According to the invention, through constructing a positioning framework based on multi-dimensional feature fusion and fault type identification, improvement of power distribution network fault positioning precision is realized.

Description

Technical Field

[0001] This invention relates to the field of fault location technology, and in particular to a method for accurate fault location in distribution networks that integrates traveling wave ranging and feeder terminal data. Background Technology

[0002] As a crucial component of the power system, the distribution network directly impacts the reliability and quality of power supply for users. With the continuous expansion of distribution network scale and the increasing complexity of its structure, rapid and accurate fault location has become a key technology for ensuring the safe and stable operation of the power grid.

[0003] In existing technologies, single fault location methods are insufficient to adapt to the complex and ever-changing operating environment and diverse fault types of distribution networks. For single-phase ground faults, the traveling wave ranging method suffers from weak traveling wave signals and difficulty in wavefront detection due to the small fault current; while for severe faults such as three-phase short circuits, the feeder terminal ranging method offers relatively accurate impedance calculations due to its smaller transition resistance. The accuracy of the two ranging methods differs under different fault types, but currently there is a lack of an effective fusion mechanism to comprehensively utilize the advantages of both methods. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for accurate fault location in distribution networks that integrates traveling wave ranging and feeder terminal data, solving the problems of poor adaptability of single fault location methods, unstable location accuracy under different fault types, and low location accuracy due to the lack of adjustment mechanism in the fusion strategy in the prior art.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for accurate fault location in distribution networks by fusing traveling wave ranging and feeder terminal data, comprising the following steps: collecting traveling wave signals and feeder terminal data at the time of a distribution network fault; extracting traveling wave feature parameters from the traveling wave signals and time-frequency domain feature parameters from the feeder terminal data, and constructing a comprehensive feature vector using the traveling wave feature parameters and the time-frequency domain feature parameters; classifying the comprehensive feature vector using a fuzzy C-means clustering algorithm to identify the fault type; obtaining traveling wave ranging results by calculating the traveling wave signals and obtaining feeder terminal ranging results by calculating the feeder terminal data; weighting and fusing the traveling wave ranging results and the feeder terminal ranging results according to a fusion weight to obtain a fused fault distance; correcting the fused fault distance according to the fault type to obtain a final fault distance; and calculating the fault point based on the final fault distance.

[0008] As a preferred embodiment of the distribution network fault accurate location method integrating traveling wave ranging and feeder terminal data described in this invention, the step of extracting time-frequency domain feature parameters from the feeder terminal data includes: decomposing the feeder terminal data into multiple intrinsic mode functions through the empirical mode decomposition; performing a Hilbert transform on each intrinsic mode function to obtain instantaneous amplitude, instantaneous phase, and instantaneous frequency; and extracting time-frequency domain feature parameters based on the instantaneous amplitude, instantaneous phase, and instantaneous frequency.

[0009] The beneficial effects of this preferred technical solution are as follows: by extracting the time-frequency domain features of feeder terminal data through empirical mode decomposition and Hilbert transform, it is possible to decompose non-stationary signals and obtain instantaneous amplitude, phase and frequency information. Compared with traditional frequency domain analysis methods, it is more suitable for processing fault transient signals and improves the accuracy and comprehensiveness of feature extraction.

[0010] As a preferred embodiment of the distribution network fault accurate location method integrating traveling wave ranging and feeder terminal data described in this invention, the step of constructing a comprehensive feature vector includes: normalizing the traveling wave feature parameters and time-frequency domain feature parameters respectively, and mapping the traveling wave feature parameters and time-frequency domain feature parameters to the same numerical range; calculating the correlation coefficients between the traveling wave feature parameters and time-frequency domain feature parameters and the fault distance respectively, and filtering feature parameters with correlation higher than a set threshold; aligning the filtered traveling wave feature parameters and time-frequency domain feature parameters according to the time series; and concatenating the aligned traveling wave feature parameters and time-frequency domain feature parameters in a preset order to form a comprehensive feature vector.

[0011] The beneficial effects of this preferred technical solution are as follows: by calculating the correlation coefficient between feature parameters and fault distance and screening highly correlated features, the feature dimensionality can be reduced, redundant information can be removed, and by combining normalization processing and time series alignment, the comparability and temporal consistency of data from different sources can be ensured, thereby improving the representativeness of the comprehensive feature vector and the accuracy of fault identification.

[0012] As a preferred embodiment of the distribution network fault accurate location method integrating traveling wave ranging and feeder terminal data described in this invention, the step of classifying the comprehensive feature vector using a fuzzy C-means clustering algorithm includes: initializing the number and initial position of cluster centers, wherein the number of cluster centers is set to 3-5; calculating the Euclidean distance from each comprehensive feature vector to each cluster center; calculating the membership degree of each comprehensive feature vector to each cluster center based on the reciprocal of the distance, and normalizing the membership degree; updating the position of each cluster center based on the weighted membership degree; determining whether the change in the position of the cluster center is less than the convergence threshold; if not, returning to the distance calculation step; if so, proceeding to the next step; and classifying each comprehensive feature vector to the fault type corresponding to the cluster center with the largest membership degree.

[0013] The advantages of this preferred technical solution are: it uses fuzzy C-means clustering algorithm to identify fault types and achieves soft classification through membership degree calculation, which can better handle fault samples with fuzzy boundaries compared with hard classification methods.

[0014] As a preferred embodiment of the distribution network fault accurate location method integrating traveling wave ranging and feeder terminal data described in this invention, the steps for obtaining the traveling wave ranging result include: extracting the arrival times of the initial wave and reflected wave of the traveling wave, calculating the time difference, and calculating the traveling wave ranging result based on the traveling wave velocity and the time difference; the steps for obtaining the feeder terminal ranging result include: determining the fault current direction based on the feeder terminal data, identifying the fault section, calculating the fault point impedance, and calculating the feeder terminal ranging result based on the impedance per unit length of the line.

[0015] As a preferred embodiment of the distribution network fault accurate location method that integrates traveling wave ranging and feeder terminal data according to the present invention, the steps for obtaining the integrated fault distance include: evaluating the data quality of the traveling wave ranging results and the feeder terminal ranging results respectively, and combining the two sets of data quality to determine a quality level combination; setting an initial integration weight according to the quality level combination; performing cross-validation on the traveling wave ranging results and the feeder terminal ranging results, and adjusting the initial integration weight according to the cross-validation results; and using the adjusted integration weight to perform a weighted calculation on the traveling wave ranging results and the feeder terminal ranging results to obtain the integrated fault distance.

[0016] The beneficial effects of this preferred technical solution are: it establishes a two-dimensional data quality evaluation system based on distance characterization capability and signal quality, which can quantitatively evaluate the reliability of traveling wave ranging and feeder terminal ranging results.

[0017] As a preferred embodiment of the distribution network fault accurate location method integrating traveling wave ranging and feeder terminal data described in this invention, the steps for evaluating the data quality of traveling wave ranging results and feeder terminal ranging results include: obtaining the correlation coefficient between traveling wave characteristic parameters and fault distance, calculating the average value of the correlation coefficients of all traveling wave characteristic parameters as the distance characterization capability index of the traveling wave data; obtaining the correlation coefficient between time-frequency domain characteristic parameters and fault distance, calculating the average value of the correlation coefficients of all time-frequency domain characteristic parameters as the distance characterization capability index of the feeder terminal data; evaluating the signal quality level of the traveling wave data based on the waveform integrity and signal-to-noise ratio of the traveling wave signal; evaluating the signal quality level of the feeder terminal data based on the consistency of fault current direction judgment and the convergence of impedance calculation in the feeder terminal data; combining the distance characterization capability index of the traveling wave data with the signal quality level to determine the comprehensive data quality level of the traveling wave ranging results; combining the distance characterization capability index of the feeder terminal data with the signal quality level to determine the comprehensive data quality level of the feeder terminal ranging results; and combining the comprehensive data quality level of the traveling wave ranging results with the comprehensive data quality level of the feeder terminal ranging results to determine the quality level combination.

[0018] As a preferred embodiment of the distribution network fault accurate location method integrating traveling wave ranging and feeder terminal data described in this invention, the step of setting initial fusion weights based on quality level combinations includes: setting basic fusion weights based on quality level combinations; when the comprehensive data quality level of the traveling wave ranging result is higher than that of the feeder terminal ranging result, the basic weight of the traveling wave ranging is set to 60%-70%; conversely, the basic weight of the feeder terminal ranging is set to 60%-70%; when the quality levels of both are the same, equal weights are set; when the fault type is a single-phase ground fault, the feeder terminal ranging weight is reduced by 5%-15% on the basic weight, and the traveling wave ranging weight is increased accordingly; when the fault type is a two-phase short circuit or a two-phase ground short circuit, the basic fusion weight remains unchanged; when the fault type is a three-phase short circuit, the feeder terminal ranging weight is increased by 5%-15% on the basic weight, and the traveling wave ranging weight is decreased accordingly.

[0019] The beneficial effects of this preferred technical solution are as follows: In view of the difference in accuracy between the two ranging methods under different fault types, a correlation mechanism between fault type and fusion weight is established. In the case of single-phase ground fault, the traveling wave ranging is emphasized, and in the case of three-phase short-circuit fault, the feeder terminal ranging is emphasized, thus realizing the weight allocation according to the fault characteristics.

[0020] As a preferred embodiment of the distribution network fault accurate location method integrating traveling wave ranging and feeder terminal data described in this invention, the step of correcting the integrated fault distance based on the fault type to obtain the final fault distance includes: when the fault type is a single-phase ground fault, the correction coefficient for the integrated fault distance is set to 0.92-0.98; when the fault type is a two-phase short-circuit fault, the correction coefficient for the integrated fault distance is set to 0.98-1.02; when the fault type is a two-phase ground short-circuit fault, the correction coefficient for the integrated fault distance is set to 0.95-1.00; and when the fault type is a three-phase short-circuit fault, the correction coefficient for the integrated fault distance is set to 1.00-1.05.

[0021] Secondly, this invention provides a precise fault location system for distribution networks that integrates traveling wave ranging and feeder terminal data, comprising: a data acquisition module for acquiring traveling wave signals and feeder terminal data during a distribution network fault; a feature extraction and classification module for extracting traveling wave feature parameters from the traveling wave signals and time-frequency domain feature parameters from the feeder terminal data, constructing a comprehensive feature vector, and classifying the comprehensive feature vector using a fuzzy C-means clustering algorithm to identify the fault type; a ranging calculation module for obtaining traveling wave ranging results by calculating the traveling wave signals and obtaining feeder terminal ranging results by calculating the feeder terminal data; and a fusion location module for weighted fusion of the traveling wave ranging results and the feeder terminal ranging results according to fusion weights to obtain a fused fault distance, correcting the fused fault distance according to the fault type to obtain a final fault distance, and calculating the fault point based on the final fault distance.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] By constructing a location framework based on multi-dimensional feature fusion and fault type identification, the accuracy of fault location in distribution networks has been improved. Specifically, empirical mode decomposition and Hilbert transform are used to extract time-frequency domain features from feeder terminal data. A comprehensive feature vector is constructed by combining traveling wave feature parameters, and feature parameters with high representational capabilities are selected through correlation analysis. This solves the problems of insufficient utilization of feature information and difficulties in fusion caused by data heterogeneity in traditional methods. At the same time, fuzzy C-means clustering algorithm is introduced to achieve intelligent identification of fault types, providing a reliable basis for subsequent fusion strategies. This allows the location method to be adjusted according to the actual fault characteristics, exhibiting stronger adaptability and accuracy compared to fixed fusion strategies.

[0024] Secondly, this invention establishes a fusion mechanism driven by both data quality assessment and fault type. Through a dual-dimensional evaluation system of distance characterization capability indicators and signal quality levels, the reliability of traveling wave ranging and feeder terminal ranging results is quantified. Furthermore, cross-validation is used to adjust the fusion weights, overcoming the limitations of fixed or empirically-based weight settings in traditional weighted fusion methods. Further, addressing the significant differences in accuracy between the two ranging methods under different fault types, a correlation mechanism is established between fault type, fusion weights, and correction coefficients. For single-phase ground faults, traveling wave ranging is emphasized; for three-phase short-circuit faults, feeder terminal ranging is emphasized, achieving intelligent fusion tailored to the specific fault. Attached Figure Description

[0025] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. 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.

[0026] Figure 1 This is a schematic diagram of the overall process of a method for accurate location of distribution network faults that integrates traveling wave ranging and feeder terminal data, according to an embodiment of the present invention. Detailed Implementation

[0027] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0028] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for accurate fault location in distribution networks by fusing traveling wave ranging and feeder terminal data is provided, comprising the following steps:

[0029] S1. Collect traveling wave signals and feeder terminal data during power distribution network faults.

[0030] Specifically, when a fault occurs in the distribution network, traveling wave sensors installed at the outgoing lines of the substation collect fault traveling wave signals at a sampling frequency of 1MHz and a sampling duration of 10ms to obtain transient voltage and current waveforms. Simultaneously, feeder terminal units (FTUs) distributed at key nodes along the distribution lines collect three-phase voltage and three-phase current data at the time of the fault, at a sampling frequency of 10kHz and a sampling duration of 200ms. The traveling wave sensors employ a combination of high-frequency current transformers and voltage dividers to ensure accurate capture of the high-frequency transient signals generated by the fault. The feeder terminal units use synchronous sampling technology to ensure time synchronization of data from each measuring point, with a time synchronization accuracy better than 1μs. The collected data is transmitted in real-time to the distribution network fault location master station via an optical fiber communication network for further processing.

[0031] S2. Extract traveling wave feature parameters from the traveling wave signal and extract time-frequency domain feature parameters from the feeder terminal data. Construct a comprehensive feature vector using the traveling wave feature parameters and the time-frequency domain feature parameters.

[0032] For traveling wave signals, after denoising using wavelet transform, the following traveling wave characteristic parameters are extracted: initial traveling wave front arrival time, initial traveling wave amplitude, initial traveling wave steepness, main reflected wave arrival time, reflected wave amplitude, waveform energy, and dominant frequency of the spectrum, totaling 12 characteristic parameters. The initial traveling wave front arrival time is identified using the wavelet modulus maxima method, extracting the time corresponding to the abrupt change point; the waveform steepness is obtained by calculating the rate of change of voltage or current at the leading edge of the wavefront; and the waveform energy is calculated using the square integral of the signal.

[0033] The specific steps for extracting time-frequency domain feature parameters from feeder terminal data include S2.1 to S2.3:

[0034] S2.1 The feeder terminal data is decomposed into multiple intrinsic mode functions through the empirical mode decomposition.

[0035] The acquired fault current signal is used as the raw signal. The empirical mode decomposition algorithm is used for decomposition. First, all extreme points of the signal are identified. Then, the upper and lower envelopes are fitted using cubic spline interpolation, and the mean of the two envelopes is calculated. The first component is obtained by subtracting the mean from the original signal. . judge Does it satisfy the conditions for the intrinsic mode function (IMF), namely, the number of local extrema differs from the number of zero-crossings by at most one, and the mean of the upper and lower envelopes is close to zero? If not, then... Repeat the above process with the original signal as a new one until the IMF condition is met, obtaining the first intrinsic mode function IMF1. Subtract IMF1 from the original signal to obtain the residual signal. ,Will The process of repeatedly decomposing the original signal is repeated to obtain IMF2, IMF3...IMFn sequentially, until the residual signal is a monotonic function or its amplitude is less than a preset threshold. In this embodiment, the fault current signal is decomposed into 8 intrinsic modulo functions.

[0036] S2.2 Perform Hilbert transform on each intrinsic modulus function to obtain the instantaneous amplitude, instantaneous phase and instantaneous frequency.

[0037] Specifically, a Hilbert transform is performed on each eigenmode function IMFi(t) obtained from the decomposition to obtain its analytic signal. The Hilbert transform is defined as: Constructing analytic signals Representing the analytic signal in polar coordinates ,in Instantaneous amplitude, Instantaneous phase, instantaneous frequency Through the above calculations, the instantaneous amplitude, instantaneous phase, and instantaneous frequency information of each intrinsic mode function at each moment on the time axis are obtained, forming the time-frequency distribution characteristics.

[0038] What needs to be known is that It is the Hilbert transform of the eigenmode functions. It is an analytic signal composed of intrinsic modulus functions and their Hilbert transform. It is the imaginary unit.

[0039] S2.3 Extract time-frequency domain feature parameters based on the instantaneous amplitude, instantaneous phase, and instantaneous frequency.

[0040] The time-frequency information obtained through Hilbert transform yields the following time-frequency domain feature parameters: average instantaneous frequency, instantaneous frequency standard deviation, peak instantaneous amplitude, mean instantaneous amplitude, instantaneous energy (square integral of instantaneous amplitude), number of phase abrupt change points, and energy proportion of the dominant frequency component. The top 5 IMF components with the highest energy proportions are selected for feature extraction, with 7 feature parameters extracted for each component, totaling 35 time-frequency domain feature parameters. In addition, 8 steady-state feature parameters are extracted, including changes in phase current amplitude before and after the fault, zero-sequence current amplitude, and negative-sequence current amplitude. These, along with the time-frequency domain feature parameters, form the feeder terminal data feature set.

[0041] The steps for constructing the comprehensive feature vector include A1 to A4:

[0042] A1. Normalize the traveling wave characteristic parameters and the time-frequency domain characteristic parameters respectively, and map the traveling wave characteristic parameters and the time-frequency domain characteristic parameters to the same numerical range.

[0043] Because the dimensions and numerical ranges of the traveling wave characteristic parameters and the feeder terminal time-frequency domain characteristic parameters differ significantly, normalization is required. A minimum-maximum normalization method is used to map all characteristic parameters to the interval [0, 1]. and These are the minimum and maximum values ​​of the characteristic parameter in the historical fault sample database, respectively. In this embodiment, a normalized parameter table is established based on the minimum and maximum values ​​of each characteristic parameter, which are statistically analyzed using a sample database containing 500 sets of historical fault data. After normalization, the 12 traveling wave characteristic parameters and the 43 feeder terminal characteristic parameters are all within the same numerical range, eliminating the influence of dimensions.

[0044] A2. Calculate the correlation coefficients between the traveling wave characteristic parameters and the time-frequency domain characteristic parameters and the fault distance, and filter the characteristic parameters whose correlation is higher than the set threshold.

[0045] The Pearson correlation coefficient is used to evaluate the correlation between each characteristic parameter and the actual fault distance. For the characteristic parameter xi and the fault distance d, the correlation coefficient is calculated using the following formula: ,in, For covariance, and The standard deviations are used as the reference values. The correlation coefficient for each feature parameter is calculated based on the historical fault sample database. A threshold of 0.3 is set, and feature parameters with an absolute correlation coefficient greater than 0.3 are selected. After screening, 9 out of 12 traveling wave feature parameters and 28 out of 43 feeder terminal feature parameters meet the criteria, retaining a total of 37 highly correlated feature parameters. This step effectively removes redundant features with weak correlation to fault distance, reduces feature dimensionality, and improves the efficiency and accuracy of subsequent processing.

[0046] A3. Align the filtered traveling wave characteristic parameters and time-frequency domain characteristic parameters according to the time series.

[0047] Because the sampling frequencies of the traveling wave signal and the feeder terminal data are different, and there may be slight deviations in the start time of acquisition, time series alignment is required. The fault occurrence time is used as the reference time point t0, which is determined by retroactively calculating the arrival time of the initial wavefront of the traveling wave. The timestamps corresponding to all characteristic parameters are uniformly converted to relative times with respect to t0, ensuring that characteristic parameters from different sources correspond to the same time period of the same fault event on the time axis. For characteristic parameters requiring time-series characteristics (such as instantaneous frequency change trends), linear interpolation is used to interpolate low-sampling-rate data onto the high-sampling-rate time axis, ensuring consistency in time resolution.

[0048] A4. The aligned traveling wave feature parameters and time-frequency domain feature parameters are concatenated in a preset order to form a comprehensive feature vector.

[0049] Following the order of "traveling wave characteristics - time-frequency domain characteristics - steady-state characteristics", the 37 aligned feature parameters are arranged sequentially to form a 37-dimensional comprehensive feature vector. The first nine elements are traveling wave characteristic parameters, and the middle 28 elements are feeder terminal time-frequency domain characteristic parameters. The comprehensive feature vector fully characterizes the transient and steady-state characteristics of the fault, integrating the information advantages of both traveling wave ranging and feeder terminal ranging methods, and providing a comprehensive data foundation for subsequent fault type identification and accurate location.

[0050] S3. The fuzzy C-means clustering algorithm is used to classify the comprehensive feature vector and identify the fault type.

[0051] The Fuzzy C-Means (FCM) clustering algorithm identifies fault types through soft classification, allowing each sample to belong to multiple categories with different membership degrees. Compared to hard classification methods, it is more suitable for handling fault samples with ambiguous boundaries. The specific steps include S3.1~S3.6:

[0052] S3.1 Initialize the number and initial position of cluster centers, wherein the number of cluster centers is set to 3-5.

[0053] Based on common fault types in distribution networks, this embodiment sets the number of cluster centers to four, corresponding to single-phase grounding faults, two-phase short-circuit faults, two-phase ground-to-short-circuit faults, and three-phase short-circuit faults, respectively. The initial location of the cluster centers is selected using an initialization strategy: first, a sample is randomly selected from the historical fault sample database as the first cluster center; then, the distance from all samples to the selected cluster center is calculated, and the sample with the farthest distance is selected as the next cluster center; this process is repeated until four initial cluster centers are selected. Each cluster center is a 37-dimensional feature vector, representing the typical characteristics of a certain type of fault. The fuzzy weighting index is set to 2.0; this parameter controls the degree of fuzziness in the classification, with a larger value resulting in a smoother classification boundary. The convergence threshold is set to 0.001 to determine whether the algorithm has reached a stable state.

[0054] S3.2 Calculate the Euclidean distance from each integrated feature vector to each cluster center.

[0055] For a fault sample to be classified, the similarity between its comprehensive feature vector and the four cluster centers is calculated. Euclidean distance is used as the similarity metric; a smaller distance indicates greater similarity. The Euclidean distance is calculated by subtracting the corresponding dimensions of two 37-dimensional feature vectors, squared the difference, summing the 37 squared values, and taking the square root. In this embodiment, the calculated distances from the current fault sample to the four cluster centers are: 2.35 to center 1, 4.67 to center 2, 3.89 to center 3, and 5.12 to center 4. These distance values ​​indicate that the fault sample is closest to the first cluster center (single-phase ground fault type). To improve computational efficiency, the system uses matrix operations to simultaneously calculate the distances from all samples to all cluster centers.

[0056] S3.3 Calculate the membership degree of each comprehensive feature vector to each cluster center based on the reciprocal of the distance, and normalize the membership degree.

[0057] Membership degree represents the probability that a fault sample belongs to a specific fault type, ranging from 0 to 1, with a higher value indicating a higher probability of belonging. The calculation of membership degree is based on the inverse relationship of distance: the closer the distance, the higher the membership degree; the farther the distance, the lower the membership degree. Specifically, for each fault type, the ratio of the current sample's distance to that type to its distance to all types is first calculated. Then, these ratios are squared, and finally, the inverse is calculated and normalized to ensure that the sum of the sample's membership degrees to all types equals 1.

[0058] In this embodiment, the membership degrees of the current fault sample to the four cluster centers are calculated as follows: 0.547 for center 1, 0.138 for center 2, 0.200 for center 3, and 0.115 for center 4. The sum of the four membership degrees is 1.000, which meets the normalization requirement. The maximum membership degree of 0.547 corresponds to a single-phase ground fault of type 1, indicating that there is a 54.7% probability that the fault belongs to the single-phase ground fault type. This soft classification method is more flexible than the traditional hard classification (which can only be classified into one type) and can reflect the uncertainty and ambiguity of fault characteristics.

[0059] S3.4 Update the position of each cluster center according to the membership degree weighting.

[0060] The cluster center positions are updated using a weighted average method, where samples with higher membership degrees have a greater impact on the cluster center positions. Specifically, for each cluster center, the membership degrees of all samples to that center are collected, the membership degrees are squared to obtain weights, and then these weights are used to calculate a weighted average of the feature vectors of all samples to obtain the new cluster center positions. This embodiment updates the cluster centers based on the current 500 historical fault samples and newly occurring fault samples.

[0061] Taking the first cluster center as an example, the membership degree of all samples to this center was calculated. It was found that 156 samples had a membership degree exceeding 0.3, and these samples mainly represented single-phase grounding faults. The squared membership degree of each sample was used as a weight, and the feature vectors of these 156 samples were weighted and summed. This sum was then divided by the total weight to obtain the new cluster center positions. The first dimension feature value of the new positions was updated from 0.68 to 0.71, the second dimension from 0.52 to 0.49, and the remaining 35 dimensions were adjusted accordingly. Through iterative updates, the four cluster centers gradually moved towards the true distribution centers of their respective fault types, making the classification results more accurate.

[0062] S3.5 Determine whether the change in the cluster center position is less than the convergence threshold. If not, return to the distance calculation step. If yes, proceed to the next step.

[0063] After each iteration, it is necessary to check whether the cluster center positions have stabilized. This is done by calculating the change in the position of each cluster center before and after the current iteration, using Euclidean distance to measure the magnitude of the change. The changes in the positions of the four cluster centers are calculated, and the maximum value is used as the criterion. When the maximum change is less than the preset convergence threshold of 0.001, the cluster centers are considered stable, the algorithm converges, and the iteration terminates; otherwise, the updated cluster center positions are used, and the process returns to step S3.2 to recalculate the distance and membership, starting a new round of iteration.

[0064] In this embodiment, the algorithm converges on the 8th iteration. The convergence trajectory of the iteration process is as follows: the maximum position change is 1.253 in the 1st iteration, indicating a significant adjustment in the cluster centers; the change decreases to 0.687 in the 2nd iteration; 0.342 in the 3rd; 0.178 in the 4th; 0.089 in the 5th; 0.043 in the 6th; 0.018 in the 7th; and finally, it decreases to 0.00087 in the 8th, which is less than the convergence threshold of 0.001, at which point the algorithm terminates. The entire iteration process exhibits rapid convergence characteristics, with large changes in the first few iterations, gradually stabilizing thereafter. To prevent the algorithm from failing to converge under special circumstances, the system sets a maximum iteration limit of 100 times; exceeding this limit forces termination and outputs the current result.

[0065] S3.6. Classify each comprehensive feature vector into the fault type corresponding to the cluster center with the highest membership degree.

[0066] After the algorithm converges, the fault type is determined based on the final membership matrix. For each fault sample, the sample with the highest membership degree among all cluster centers is identified, and the sample is classified into the fault type corresponding to that cluster center. In this embodiment, the final membership degree distribution of the current fault sample is as follows: 0.563 for center 1, 0.125 for center 2, 0.192 for center 3, and 0.120 for center 4. The maximum membership degree is 0.563, corresponding to the first cluster center; therefore, the fault is identified as a single-phase ground fault.

[0067] Meanwhile, the system records the maximum membership value of 0.563 as the classification confidence index. Confidence reflects the reliability of the classification result; the closer the value is to 1, the more certain the classification; the closer the value is to 0.5, the more ambiguous the classification. When the confidence is below 0.4, it indicates that the fault characteristics are in the boundary area of ​​multiple types, and the classification result is not certain enough. The system will issue a classification uncertainty warning, prompting maintenance personnel to conduct manual review or make a comprehensive judgment based on other information.

[0068] A correspondence was established between fault types and cluster centers: Cluster center 1 represents single-phase ground faults, characterized by a large zero-sequence current, a significant drop in fault phase voltage, and a relatively weak traveling wave signal; Cluster center 2 represents two-phase short-circuit faults, characterized by a large negative-sequence current, a decrease in phase-to-phase voltage, and a significant traveling wave signal; Cluster center 3 represents two-phase ground faults, exhibiting both zero-sequence and negative-sequence components; Cluster center 4 represents three-phase short-circuit faults, characterized by a simultaneous and significant increase in all three-phase currents, a severe voltage drop, and the strongest traveling wave signal. The final output of the fault type identification result is "single-phase ground fault, confidence level 56.3%", providing important fault type basis for subsequent adaptive fusion localization.

[0069] S4. By calculating the traveling wave signal, the traveling wave ranging result is obtained. By calculating the feeder terminal data, the feeder terminal ranging result is obtained.

[0070] The steps for obtaining the traveling wave ranging result include: extracting the arrival times of the initial wave and the reflected wave of the traveling wave, calculating the time difference, and calculating the traveling wave ranging result based on the traveling wave velocity and the time difference.

[0071] In this embodiment, the arrival times of the initial and reflected waves of the traveling wave are extracted to calculate the time difference, and the traveling wave ranging result is calculated based on the traveling wave velocity and the time difference. The principle of traveling wave ranging is to locate the fault by utilizing the propagation characteristics of high-frequency transient waves generated during a fault in the line. At the instant a fault occurs, the fault point generates traveling waves that propagate towards both ends. These traveling waves are reflected after reaching the end or branch point of the line, forming reflected waves that return. By measuring the time difference between the initial and reflected waves arriving at the monitoring point, and combining this with the propagation velocity of the traveling wave in the line, the fault distance can be calculated.

[0072] The specific ranging process is as follows: First, wavelet transform technology is used to process the acquired traveling wave voltage signal to accurately identify the arrival time of the initial wavefront. Wavelet transform can analyze the signal simultaneously in the time and frequency domains, effectively capturing the abrupt change characteristics of the traveling wavefront. In this embodiment, the arrival time of the initial wavefront at the traveling wave sensor installed at the substation outgoing line is detected to be 1.235 milliseconds after the fault occurs. The initial wave is the first wavefront of the traveling wave generated at the fault point that propagates directly to the measurement point, and its propagation path is the line from the fault point to the measurement point.

[0073] Next, the traveling wave signal is analyzed to identify the main reflected wavefronts. A reflected wave is a wave that reflects back after the initial wave reaches the end of the line, a branch point, or another point of impedance discontinuity. In this embodiment, the distribution line end is an open circuit, exhibiting strong reflection characteristics. Through wavelet modulus maxima detection, the arrival time of the first significant reflected wave at the measurement point is identified as 3.847 milliseconds after the fault occurs.

[0074] Calculate the round-trip time difference of the traveling wave: Subtract the arrival time of the initial wave from the arrival time of the reflected wave to obtain a time difference of 3.847 - 1.235 = 2.612 milliseconds. This time difference represents the total time it takes for the traveling wave to propagate from the measurement point to the end of the line and then reflect back. The corresponding propagation distance is twice the distance from the measurement point to the end of the line.

[0075] The propagation speed of traveling waves in the transmission line is determined. The propagation speed of traveling waves in transmission lines is close to the speed of light, but it is affected by the line structure and dielectric properties. For overhead lines, the propagation speed is approximately 290-295 meters per microsecond; for cable lines, due to the higher dielectric constant, the propagation speed decreases to 170-180 meters per microsecond. In this embodiment, the distribution line is a hybrid structure of overhead and cable lines. Based on the length ratio of the two types of lines, the equivalent traveling wave propagation speed is calculated to be 280 meters per microsecond.

[0076] The fault distance is calculated based on the principle of two-end ranging. A time difference of 2.612 milliseconds (2612 microseconds) corresponds to a propagation distance of 280 meters / microsecond × 2612 microseconds = 731.36 meters, which is the total round-trip distance of the traveling wave. The fault distance is equal to half of the round-trip distance, i.e., 731.36 meters ÷ 2 = 365.68 meters. This distance is from the installation location of the traveling wave sensor (the outgoing line end of the substation) to the fault point.

[0077] Considering the existence of multiple branches in the actual line, which may generate multiple reflected waves, wavefront matching verification is required based on the line topology. The system confirms that the identified reflected waves indeed originate from the end of the line by matching the arrival times of each reflected wave with the known branch point locations, thus verifying the validity of the ranging results. Considering the potential error of 1-2 sampling points in wavefront identification and the uncertainty in wave velocity calculation, the traveling wave ranging result is recorded as 365.68 meters, with an estimated ranging uncertainty of ±15 meters.

[0078] The steps to obtain the distance measurement results of the feeder terminal include: determining the direction of the fault current based on the feeder terminal data, identifying the section where the fault is located, calculating the impedance of the fault point, and calculating the distance measurement results of the feeder terminal based on the impedance per unit length of the line.

[0079] The fault current direction is determined based on feeder terminal data, the fault section is identified, the fault point impedance is calculated, and the feeder terminal ranging result is calculated based on the impedance per unit length of the line. The feeder terminal ranging method is based on multiple feeder terminal units deployed in the distribution network automation system, and fault location is achieved by analyzing the electrical quantity information of each monitoring point.

[0080] The first step is fault current direction determination and section location. In this embodiment, five feeder terminal units (FTUs) are installed on the distribution line, numbered FTU1 to FTU5, dividing the entire line into four sections. Each feeder terminal unit can monitor the three-phase voltage and current at its location in real time. By comparing the fault current direction measured by each feeder terminal and applying Kirchhoff's current law, the section in which the fault occurred can be determined.

[0081] The direction of fault current is determined using a power direction criterion. For each feeder terminal, the product of its measured voltage and current is calculated to obtain the complex power. When the real part of the complex power is positive, it indicates that the power flows towards the line, i.e., the current flows into the line; when the real part is negative, it indicates that the power flows out of the line, i.e., the current flows out of the line. For normal sections, the directions of current inflow and outflow should be consistent; for fault sections, the current upstream of the fault point flows towards the line, and the current downstream flows out of the line.

[0082] The calculation results of this embodiment show that FTU1 measured the current flow direction along the line, FTU2 measured the current flow direction along the line, FTU3 measured the current outflow line, FTU4 measured the current outflow line, and FTU5 measured the current outflow line. Based on the point of change in current direction, the fault is determined to be located in section II between FTU2 and FTU3. A query of the line topology database shows that the starting mileage of section II is 200 meters, the ending mileage is 800 meters, and the total length of the section is 600 meters. Section location narrows the fault search range, providing a basis for accurate distance measurement.

[0083] The second step is to calculate the impedance at the fault point. Based on the voltage and current data measured at the upstream feeder terminal FTU2, the line impedance from FTU2 to the fault point is calculated using the impedance method. The impedance calculation formula differs for different types of faults. In this embodiment, the identified fault type is a single-phase ground fault, so the influence of the zero-sequence component needs to be considered.

[0084] The calculation method for fault circuit impedance is as follows: Extract the phase voltage and phase current of the faulted phase, and simultaneously extract the zero-sequence components of the three-phase voltage and current. The zero-sequence component reflects the degree of three-phase imbalance and is particularly important in single-phase ground faults. A zero-sequence compensation coefficient is introduced to correct the calculation; this coefficient is related to the zero-sequence impedance and positive-sequence impedance of the line. For overhead lines, the zero-sequence impedance is typically 3-4 times the positive-sequence impedance, and the zero-sequence compensation coefficient is generally between 0.8 and 1.2.

[0085] In this embodiment, the data measured by FTU2 are as follows: fault phase (phase A) voltage amplitude 2.35 kV, phase angle -68 degrees; fault phase current amplitude 1.25 kA, phase angle -85 degrees; zero-sequence voltage amplitude 1.87 kV, phase angle -72 degrees; zero-sequence current amplitude 0.58 kA, phase angle -88 degrees. According to the line parameter database, the positive-sequence impedance of this section of the line is 0.25 + j0.35 ohms per kilometer, and the zero-sequence impedance is 0.68 + j1.15 ​​ohms per kilometer. The zero-sequence compensation coefficient is calculated using the following formula:

[0086] Where k is the zero-order compensation coefficient, It is the zero-sequence impedance per unit length of the line. It is the positive sequence impedance per unit length of the line.

[0087] Substituting the numerical values, the zero-sequence compensation coefficient is approximately 0.86 + j0.64. Then, the fault loop impedance calculation formula is used:

[0088] Where: Zf is the fault point impedance in ohms, Ua is the fault phase voltage, and Ia is the fault phase current. It is zero-sequence voltage. It is the zero-sequence current, and k is the zero-sequence compensation coefficient.

[0089] Substituting the measured data into the formula and performing complex number operations, the impedance at the fault point is calculated to be 0.042 + j0.059 ohms. The real part of the impedance represents the resistive component, and the imaginary part represents the reactive component. Calculating the magnitude of the impedance, which is the square root of the sum of the squares of the real and imaginary parts, yields 0.0725 ohms.

[0090] The third step is to calculate the fault distance based on the impedance per unit length of the line. The fault distance equals the impedance at the fault point divided by the impedance per unit length of the line. The positive sequence impedance per unit length for this section is retrieved from the line parameter database; the modulus is 0.431 ohms per kilometer, or 0.000431 ohms per meter. The fault distance is calculated as 0.0725 ohms ÷ 0.000431 ohms / meter = 168.2 meters. This distance is from FTU2 to the fault point.

[0091] Converting relative distance to absolute distance: The cumulative distance from the FTU2 installation location to the line start point is 200 meters. Adding the calculated 168.2 meters, the absolute distance from the fault point to the line start point is 200 + 168.2 = 368.2 meters. Considering potential manufacturing and installation errors in the line parameters, the possibility that transition resistance might affect the accuracy of impedance calculation, and instrument errors in voltage and current measurements, the distance measurement result recorded at the feeder terminal is 368.2 meters, with an estimated distance measurement uncertainty of ±25 meters.

[0092] Thus, two independent ranging results were obtained: the traveling wave ranging result was 365.68 meters (uncertainty ±15 meters), and the feeder terminal ranging result was 368.2 meters (uncertainty ±25 meters). The difference between the two sets of results was 2.52 meters, with a relative deviation of 0.69%, indicating that the ranging results of the two methods are basically consistent and mutually verify the effectiveness of the ranging method.

[0093] S5. The traveling wave ranging result and the feeder terminal ranging result are weighted and fused according to the fusion weight to obtain the fused fault distance.

[0094] The steps for obtaining the fusion fault distance include S5.1~S5.4:

[0095] S5.1 Evaluate the data quality of the traveling wave ranging results and the feeder terminal ranging results separately, and combine the two sets of data quality to determine the quality level combination.

[0096] The steps for evaluating the data quality of traveling wave ranging results and feeder terminal ranging results include B1 to B7:

[0097] B1. Obtain the correlation coefficient between traveling wave characteristic parameters and fault distance, and calculate the average value of the correlation coefficients of all traveling wave characteristic parameters as an indicator of the distance representation capability of traveling wave data.

[0098] Distance characterization capability reflects the predictive ability of characteristic parameters for fault distance. The correlation coefficient measures the strength of the linear relationship between two variables, ranging from -1 to +1; a larger absolute value indicates a stronger correlation. In this embodiment, the nine extracted traveling wave characteristic parameters are: initial wavefront arrival time, initial wave amplitude, waveform steepness, reflected wave arrival time, reflected wave amplitude, waveform energy, dominant frequency, wavefront sharpness, and signal duration.

[0099] Based on 500 sets of data from a historical fault sample database, the correlation coefficient between each traveling wave characteristic parameter and the actual fault distance was calculated. The results show that the correlation coefficients are as follows: initial wavefront arrival time is 0.89, reflected wave arrival time is 0.86, waveform energy is 0.72, initial wave amplitude is 0.68, waveform steepness is 0.64, reflected wave amplitude is 0.61, dominant frequency is 0.58, wavefront sharpness is 0.71, and signal duration is 0.55.

[0100] The arithmetic mean of the correlation coefficients of the nine traveling wave characteristic parameters yields the distance characterization index of the traveling wave data: 0.89 + 0.86 + 0.72 + 0.68 + 0.64 + 0.61 + 0.58 + 0.71 + 0.55 divided by 9, resulting in 0.693. This high index value indicates that the traveling wave characteristic parameters generally have a strong correlation with the fault distance, and the traveling wave ranging method has good distance characterization capability.

[0101] B2. Obtain the correlation coefficient between time-frequency domain characteristic parameters and fault distance, and calculate the average value of the correlation coefficients of all time-frequency domain characteristic parameters as an indicator of the distance characterization capability of feeder terminal data.

[0102] For feeder terminal data, this embodiment retains 28 highly correlated time-frequency domain characteristic parameters, including the instantaneous frequency, instantaneous amplitude, instantaneous energy, and steady-state characteristics such as zero-sequence current and negative-sequence current of each intrinsic mode function. Based on the same historical fault sample library, the correlation coefficient between each time-frequency domain characteristic parameter and the fault distance is calculated.

[0103] The correlation coefficients of the main characteristic parameters are as follows: instantaneous frequency of the first intrinsic mode function is 0.76, instantaneous frequency of the second intrinsic mode function is 0.68, instantaneous amplitude of the first intrinsic mode function is 0.72, instantaneous energy is 0.69, zero-sequence current amplitude is 0.58, negative-sequence current amplitude is 0.62, current change is 0.71, real part of impedance is 0.81, imaginary part of impedance is 0.77, etc. Summing the correlation coefficients of the 28 time-frequency domain characteristic parameters and dividing by 28 yields a distance characterization index of 0.652 for the feeder terminal data.

[0104] This index is slightly lower than the traveling wave data's 0.693, indicating that the traveling wave data has a certain advantage in distance characterization, but the feeder terminal data also has good characterization ability, and the two can complement each other.

[0105] B3. Evaluate the signal quality level of traveling wave data based on the waveform integrity and signal-to-noise ratio of the traveling wave signal.

[0106] Signal quality level reflects the reliability and availability of the acquired data. For traveling wave data, evaluation is mainly based on two aspects: waveform integrity and signal-to-noise ratio.

[0107] Waveform integrity assessment: This involves checking whether the initial wavefront and main reflected wavefront in the traveling wave signal are clearly distinguishable, and whether there are any missing waveforms or severe distortions. In this embodiment, the initial wavefront identified by wavelet transform exhibits clear characteristics, with a wavefront steepness of 2.3 kV per microsecond and a rise time of 4 microseconds, consistent with typical traveling wave characteristics. The main reflected wave is also clearly distinguishable, with an amplitude of 68% of the initial wave, meeting the criteria for reflected wave identification. Therefore, the waveform integrity is assessed as excellent.

[0108] Signal-to-noise ratio (SNR) assessment: Calculate the ratio of the effective signal power of the traveling wave to the background noise power. In this embodiment, the SNR is estimated using the ratio of signal energy before and after wavelet denoising. The original signal energy is 15.8 units, and the noise energy is 0.6 units. The SNR is calculated as 15.8 divided by 0.6, which equals 26.3, equivalent to approximately 14.2 dB. According to the evaluation criteria, an SNR higher than 12 dB is excellent, 8-12 dB is good, 4-8 dB is acceptable, and below 4 dB is unacceptable. In this embodiment, the SNR is 14.2 dB, which falls into the excellent category.

[0109] Based on both waveform integrity and signal-to-noise ratio, the signal quality level of the traveling wave data is determined to be excellent. Excellent indicates very high data quality, making it a reliable basis for ranging.

[0110] B4. Evaluate the signal quality level of the feeder terminal data based on the consistency of fault current direction judgment and the convergence of impedance calculation in the feeder terminal data.

[0111] Consistency assessment of fault current direction judgment: The direction of the fault current measured by each feeder terminal unit is checked to ensure it conforms to Kirchhoff's Current Law. In this embodiment, the current direction judgment results of the five feeder terminals are clear: FTU1 and FTU2 show current flowing towards the line, while FTU3, FTU4, and FTU5 show current flowing out of the line. The point of change in direction is clearly located between FTU2 and FTU3, with no contradictory or ambiguous direction judgments. The confidence level of the power direction criterion calculation results for each feeder terminal is above 0.85, indicating that the direction judgment is highly reliable. The consistency of fault current direction judgment is rated as excellent.

[0112] Convergence evaluation of impedance calculation: This involves checking whether the impedance calculation process at the fault point is stable and convergent, and whether the calculation results from sampling data at different times are consistent. In this example, 50 sampling windows within 10 milliseconds after the fault occur are used to calculate the impedance at the fault point, resulting in 50 calculation results. The standard deviation of these 50 results is calculated; the standard deviation of the impedance magnitude is 0.0038 ohms, and the relative standard deviation is 5.2%. According to the evaluation criteria, a relative standard deviation of less than 8% is excellent, 8%-15% is good, 15%-25% is acceptable, and greater than 25% is unacceptable. In this example, the relative standard deviation is 5.2%, which falls into the excellent category, indicating that the impedance calculation is very stable.

[0113] Based on the consistency of fault current direction judgment and the convergence of impedance calculation, the signal quality level of the feeder terminal data is determined to be excellent.

[0114] B5. Combine the distance representation capability index of traveling wave data with the signal quality level to determine the comprehensive data quality level of traveling wave ranging results.

[0115] The comprehensive data quality rating considers both the data's representational ability and signal quality. This embodiment uses a two-dimensional evaluation matrix for rating. The distance representational ability index is divided into three levels: above 0.65 for strong representation, 0.50-0.65 for medium representation, and below 0.50 for weak representation. The signal quality rating is divided into four levels: excellent, good, acceptable, and unacceptable.

[0116] The evaluation parameters for traveling wave ranging results are as follows: a distance characterization capability index of 0.693 indicates a strong characterization capability, and the signal quality level is excellent. Based on the evaluation matrix, the strong characterization capability combined with excellent signal quality results in a comprehensive data quality level of A (the highest level). Level A indicates excellent data quality and highly reliable ranging results, which can be used as the primary basis for fusion calculations.

[0117] B6. Combine the distance characterization capability index of the feeder terminal data with the signal quality level to determine the comprehensive data quality level of the feeder terminal ranging results.

[0118] The evaluation parameters for the feeder terminal ranging results are as follows: a distance characterization capability index of 0.652 falls into the medium characterization range (close to the strong characterization threshold of 0.65), and the signal quality level is excellent. Based on the evaluation matrix, the combination of medium characterization capability and excellent signal quality results in a comprehensive data quality level of B (second-highest). Level B indicates good data quality and reliable ranging results, which can serve as an important reference for fusion calculations.

[0119] It should be noted that although the overall quality level of the feeder terminal ranging result is B, its signal quality is also excellent, only slightly inferior to traveling wave data in terms of distance characterization capability.

[0120] B7. Combine the overall data quality level of the traveling wave ranging results with the overall data quality level of the feeder terminal ranging results to determine the quality level combination.

[0121] The quality level combination is used for subsequent fusion weight setting. In this embodiment, the traveling wave ranging result is grade A, and the feeder terminal ranging result is grade B, forming an "AB" quality level combination. This combination indicates that the traveling wave ranging data quality is slightly better than the feeder terminal ranging data, and the traveling wave ranging result should be given a slightly higher weight when allocating fusion weights.

[0122] A database of correspondences between quality level combinations and fusion strategies has been established, including various scenarios such as AA combination (both have comparable and high quality), AB combination (traveling wave quality is better), BA combination (feeder terminal quality is better), and BB combination (both have comparable and good quality).

[0123] S5.2. Set the initial fusion weights according to the quality level combination.

[0124] Specifically, it includes:

[0125] The basic fusion weight is set according to the combination of quality levels. When the comprehensive data quality level of the traveling wave ranging result is higher than that of the feeder terminal ranging result, the basic weight of the traveling wave ranging is set to 60%-70%. Conversely, the basic weight of the feeder terminal ranging is set to 60%-70%. When the quality levels of both are the same, equal weights are set.

[0126] When the fault type is a single-phase ground fault, the feeder terminal ranging weight is reduced by 5%-15% on the basic weight, and the traveling wave ranging weight is increased accordingly.

[0127] When the fault type is a two-phase short circuit or a two-phase-to-ground short circuit, the basic fusion weight remains unchanged;

[0128] When the fault type is a three-phase short circuit, the feeder terminal ranging weight is increased by 5%-15% on the basis of the basic weight, and the traveling wave ranging weight is reduced accordingly.

[0129] For example, the initial fusion weights are set according to the principle of "quality first, type adjustment". First, the basic weights are determined based on the combination of data quality levels, and then fine-tuned according to the fault type.

[0130] The basic fusion weight is set based on the quality level combination: In this embodiment, the quality level combination is AB, indicating that the overall data quality level of the traveling wave ranging result is higher than that of the feeder terminal ranging result. According to the setting rules, when the traveling wave ranging quality is higher than that of the feeder terminal ranging, the basic weight of the traveling wave ranging is set within the range of 60%-70%. Considering that the quality difference between the two is not large in this embodiment (level A and level B are adjacent), the middle value of this range, 65%, is used as the basic weight of the traveling wave ranging, and the corresponding basic weight of the feeder terminal ranging is 35%. If the quality level combination is AC or a larger difference, the upper limit of the range of 70% will be selected; if the quality levels of both are the same (such as AA or BB combination), then equal weights of 50% each are set.

[0131] Adjusting the base weights based on the fault type: In this embodiment, the fault type identified in step S3 is a single-phase ground fault. According to the characteristics of power system faults, the fault current is relatively small during a single-phase ground fault, and although the traveling wave signal is relatively weak, it still has certain characteristics. However, the feeder terminal ranging method is significantly affected by zero-sequence current measurement errors, which reduces the accuracy of impedance calculation. Therefore, for single-phase ground faults, the feeder terminal ranging weight should be reduced, and the traveling wave ranging weight should be increased accordingly.

[0132] According to the adjustment rules, the feeder terminal ranging weight is reduced by 5%-15% from the base weight during a single-phase ground fault. This embodiment, considering the high signal quality levels, uses the lower limit of this range, 10%, for adjustment. Specifically, the base weight for feeder terminal ranging is reduced by 10 percentage points from 35% to 25%, and correspondingly, the base weight for traveling wave ranging is increased by 10 percentage points from 65% to 75%. The adjusted initial fusion weights are: traveling wave ranging weight 75%, and feeder terminal ranging weight 25%.

[0133] If the fault type is a two-phase short circuit or a two-phase ground short circuit, the basic fusion weight remains unchanged at 65% and 35%, respectively. If the fault type is a three-phase short circuit, the feeder terminal ranging weight is increased by 5%-15% on the basic weight, and the traveling wave ranging weight is reduced accordingly, because the fault current is large during a three-phase short circuit, and the ranging accuracy of the feeder terminal impedance method is higher.

[0134] S5.3 Cross-validate the traveling wave ranging results and the feeder terminal ranging results, and adjust the initial fusion weights based on the cross-validation results.

[0135] Specifically, the relative deviation between the two sets of ranging results is calculated. In this embodiment, the traveling wave ranging result is 365.68 meters, and the feeder terminal ranging result is 368.2 meters, with a difference of 2.52 meters. The relative deviation is calculated by dividing the difference by the average of the two: 2.52 divided by the average of 367 meters, resulting in 0.69%. A relative deviation of less than 2% is considered highly consistent, 2%-5% is basically consistent, 5%-10% indicates a difference, and greater than 10% indicates a significant difference requiring further analysis. In this embodiment, the relative deviation of 0.69% falls within the highly consistent range, indicating that the ranging results from the two methods mutually verify each other and have high reliability.

[0136] Secondly, the rationality was verified by combining the line topology information. A query of the distribution network geographic information system showed that the fault occurred in section II, with a section range of 200-800 meters. The two distance measurement results, 365.68 meters and 368.2 meters, were both within this section and relatively close to FTU2, consistent with the fault section location determined in step S4, thus verifying the rationality of the distance measurement results.

[0137] Next, the overlap of ranging uncertainties was analyzed. The uncertainty range of the traveling wave ranging results is 350.68-380.68 meters (365.68±15 meters), and the uncertainty range of the feeder terminal ranging results is 343.2-393.2 meters (368.2±25 meters). There is a large overlap between the two uncertainty ranges, with the overlap range being 350.68-380.68 meters, and the overlap rate reaching 87%, further confirming the consistency of the two sets of results.

[0138] Based on cross-validation analysis, the initial fusion weights were fine-tuned. Since the two sets of ranging results were highly consistent, with a relative deviation of only 0.69%, it indicates that both methods performed well in this fault, and the weight allocation can be appropriately balanced. Furthermore, considering that the uncertainty of traveling wave ranging (±15 meters) is less than that of feeder terminal ranging (±25 meters), the accuracy advantage of traveling wave ranging is significant, and its higher weight is maintained.

[0139] Taking all factors into consideration, the traveling wave ranging weight in the initial fusion weighting was slightly adjusted from 75% to 72%, and the feeder terminal ranging weight was slightly adjusted from 25% to 28%, an adjustment of 3 percentage points. The final fusion weights are: traveling wave ranging weight 72%, and feeder terminal ranging weight 28%.

[0140] S5.4. The adjusted fusion weights are used to perform weighted calculations on the traveling wave ranging results and the feeder terminal ranging results to obtain the fused fault distance.

[0141] The weighted fusion calculation uses a linear weighted average method, which sums the two sets of ranging results according to determined weights. The calculation process is as follows:

[0142] The contribution of the traveling wave ranging result is 365.68 meters multiplied by 72%, which equals 263.29 meters; the contribution of the feeder terminal ranging result is 368.2 meters multiplied by 28%, which equals 103.10 meters. Adding the two together, the fused fault distance is 263.29 meters plus 103.10 meters, which equals 366.39 meters.

[0143] The fused fault distance of 366.39 meters combines the advantages of both ranging methods, making it more reliable than the results of a single method. Compared to the traveling wave ranging result of 365.68 meters, the fused result shows a slight increase with a deviation of 0.71 meters; compared to the feeder terminal ranging result of 368.2 meters, the fused result shows a decrease with a deviation of 1.81 meters. The fused result falls between the two sets of original results and is closer to the more accurate traveling wave ranging result, meeting the expected effect of weighted fusion.

[0144] The uncertainty of the fused fault distance is estimated. Since the fusion process integrates two sets of independent ranging results, the uncertainty is reduced. The fusion uncertainty is estimated using a weighted square method: the square of the traveling wave ranging uncertainty of 15 meters is multiplied by the square of the weight 72%, plus the square of the feeder terminal ranging uncertainty of 25 meters multiplied by the square of the weight 28%, and then the square root is taken. The calculated fusion uncertainty is approximately 12 meters. The fused fault distance is recorded as 366.39 meters, with an uncertainty of ±12 meters, i.e., a confidence interval of 354.39–378.39 meters.

[0145] S6. Correct the fused fault distance according to the fault type to obtain the final fault distance, and calculate the fault point based on the final fault distance.

[0146] The steps to obtain the final fault distance include:

[0147] When the fault type is a single-phase ground fault, the correction factor for the fused fault distance is set to 0.92-0.98;

[0148] The fault type identified in this embodiment is a single-phase ground fault. Statistical analysis of historical fault data shows that single-phase ground faults, due to their smaller fault current, unstable arc, and larger fluctuations in grounding resistance, tend to result in overestimating the distance measurement. Regression analysis based on 500 historical single-phase ground fault samples revealed that the average distance measurement deviation for single-phase ground faults is systematically overestimated by 3%-6%.

[0149] According to the correction rules, the correction factor for the distance of a fused fault during a single-phase ground fault is set within the range of 0.92-0.98. In this embodiment, considering good signal quality, clear waveforms, and that the ranging deviation should be at the lower end of the statistical distribution, the upper limit of this range, 0.96, is selected as the correction factor. If the signal quality is poor or the fault characteristics are atypical, the lower limit of the range, 0.92, will be selected for a larger correction.

[0150] Calculating the final fault distance: Multiplying the fused fault distance of 366.39 meters by a correction factor of 0.96 yields a final fault distance of 351.7 meters. The corrected distance is reduced by 14.69 meters, a correction of 4%, which is within the expected range of 3%-6%. The uncertainty of the final fault distance remains at ±12 meters, i.e., the confidence interval is 339.7-363.7 meters.

[0151] When the fault type is a two-phase short-circuit fault, the correction factor for the fused fault distance is set to 0.98-1.02;

[0152] When the fault type is a two-phase-to-ground short-circuit fault, the correction factor for the fused fault distance is set to 0.95-1.00;

[0153] When the fault type is a three-phase short-circuit fault, the correction factor for the fused fault distance is set to 1.00-1.05.

[0154] If the fault type is a two-phase short circuit fault, the correction factor is set to 0.98-1.02, which is close to 1, indicating that the ranging system error for this type of fault is small and basically no correction is needed. If the fault type is a two-phase-to-ground short circuit fault, the correction factor is set to 0.95-1.00, which requires a small correction. If the fault type is a three-phase short circuit fault, the correction factor is set to 1.00-1.05, which may require an upward correction because the fault impedance is very small during a three-phase short circuit, which can easily lead to an underestimation of the ranging result.

[0155] The fault point was calculated based on the final fault distance: the final fault distance of 351.7 meters is the distance measured from the starting point of the line (the outgoing end of the substation). The distribution network geographic information system was consulted to obtain the topology and tower location information of the distribution line. The location 351.7 meters from the starting point corresponds to the vicinity of tower number 12 on the line, which is located in an industrial park at coordinates of 116.385 degrees east longitude and 39.912 degrees north latitude.

[0156] Further investigation of the line equipment information at this location revealed a pole-mounted transformer, numbered T-0312, with a capacity of 315 kVA, located 15 meters behind tower number 12. It connects to three low-voltage branches downstream. Based on the fault type (single-phase grounding) and the calculated fault distance, the system initially determined that the fault point is located on the line section between tower number 12 and transformer T-0312. Possible causes of the fault include: flashover of insulators in this section, contact between conductors and tree branches, or damage to the transformer's down conductor insulation.

[0157] A fault location report is generated, displaying information such as the fault location, fault type, and location reliability. This report is then pushed to the dispatch center and repair personnel via the distribution automation system. Repair personnel, carrying equipment, rushed to the site of tower No. 12. Upon on-site investigation, a single-phase grounding fault caused by contact between phase A conductor and tree branches was found 10 meters behind tower No. 12. The actual fault location was 348 meters from the starting point, with a location error of only 3.7 meters compared to the system's calculated final fault distance of 351.7 meters.

[0158] Example 2 illustrates a schematic scheme for a method of accurately locating distribution network faults by integrating traveling wave ranging and feeder terminal data. It should be noted that the technical solution of this system for accurately locating distribution network faults by integrating traveling wave ranging and feeder terminal data belongs to the same concept as the technical solution of the aforementioned method for accurately locating distribution network faults by integrating traveling wave ranging and feeder terminal data. Details not described in detail in this embodiment can be found in the description of the technical solution of the aforementioned method for accurately locating distribution network faults by integrating traveling wave ranging and feeder terminal data.

[0159] This embodiment also provides a distribution network fault accurate location system that integrates traveling wave ranging and feeder terminal data, including:

[0160] The data acquisition module is used to collect traveling wave signals and feeder terminal data during power distribution network faults.

[0161] The feature extraction and classification module is used to extract traveling wave feature parameters from traveling wave signals, extract time-frequency domain feature parameters from feeder terminal data, construct a comprehensive feature vector, and use a fuzzy C-means clustering algorithm to classify the comprehensive feature vector and identify the fault type.

[0162] The ranging calculation module is used to obtain the traveling wave ranging result by calculating the traveling wave signal and to obtain the feeder terminal ranging result by calculating the feeder terminal data.

[0163] The fusion positioning module is used to perform weighted fusion of the traveling wave ranging result and the feeder terminal ranging result according to the fusion weight to obtain the fusion fault distance, and to correct the fusion fault distance according to the fault type to obtain the final fault distance, and to calculate the fault point according to the final fault distance.

[0164] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for accurate fault location in distribution networks that integrates traveling wave ranging and feeder terminal data, characterized in that, Includes the following steps: Collect traveling wave signals and feeder terminal data during power distribution network faults; Traveling wave characteristic parameters are extracted from the traveling wave signal, and time-frequency domain characteristic parameters are extracted from the feeder terminal data. A comprehensive feature vector is constructed by combining the traveling wave characteristic parameters and the time-frequency domain characteristic parameters. The comprehensive feature vector is classified using a fuzzy C-means clustering algorithm to identify the fault type; The traveling wave ranging result is obtained by calculating the traveling wave signal, and the feeder terminal ranging result is obtained by calculating the feeder terminal data. The traveling wave ranging result and the feeder terminal ranging result are weighted and fused according to the fusion weight to obtain the fused fault distance; The fused fault distance is corrected by the fault type to obtain the final fault distance, and the fault point is calculated based on the final fault distance.

2. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 1, characterized in that, The steps for extracting time-frequency domain feature parameters from feeder terminal data include: The feeder terminal data is decomposed into multiple intrinsic mode functions through the empirical mode decomposition. Perform a Hilbert transform on each intrinsic mode function to obtain the instantaneous amplitude, instantaneous phase, and instantaneous frequency; Time-frequency domain feature parameters are extracted based on the instantaneous amplitude, instantaneous phase, and instantaneous frequency.

3. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 2, characterized in that, The steps to construct a comprehensive feature vector include: The traveling wave characteristic parameters and the time-frequency domain characteristic parameters are normalized respectively, and the traveling wave characteristic parameters and the time-frequency domain characteristic parameters are mapped to the same numerical range. Calculate the correlation coefficients between the traveling wave characteristic parameters and the time-frequency domain characteristic parameters and the fault distance, and filter the characteristic parameters whose correlation is higher than the set threshold. The filtered traveling wave characteristic parameters and time-frequency domain characteristic parameters are aligned according to the time series. The aligned traveling wave feature parameters and time-frequency domain feature parameters are concatenated in a preset order to form a comprehensive feature vector.

4. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 3, characterized in that, The steps for classifying the comprehensive feature vector using the fuzzy C-means clustering algorithm include: Initialize the number and initial positions of cluster centers, with the number of cluster centers set to 3-5; Calculate the Euclidean distance from each composite feature vector to each cluster center; The membership degree of each comprehensive feature vector to each cluster center is calculated based on the reciprocal of the distance, and the membership degree is normalized. Update the position of each cluster center based on its membership degree; Determine if the change in the cluster center position is less than the convergence threshold. If not, return to the distance calculation step; if yes, proceed to the next step. Each comprehensive feature vector is classified into the fault type corresponding to the cluster center with the highest membership degree.

5. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 4, characterized in that, The steps for obtaining the traveling wave ranging result include: The arrival times of the initial wave and the reflected wave of the traveling wave are extracted and the time difference is calculated. The traveling wave ranging result is calculated based on the wave velocity and the time difference. The steps to obtain the distance measurement results of the feeder terminal include: determining the direction of the fault current based on the feeder terminal data, identifying the section where the fault is located, calculating the impedance of the fault point, and calculating the distance measurement results of the feeder terminal based on the impedance per unit length of the line.

6. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 5, characterized in that, The steps for obtaining the fusion fault distance include: The data quality of traveling wave ranging results and feeder terminal ranging results are evaluated separately, and the two sets of data quality are combined to determine the quality level combination. The initial fusion weights are set according to the combination of quality levels; Cross-validate the traveling wave ranging results and the feeder terminal ranging results, and adjust the initial fusion weights based on the cross-validation results; The adjusted fusion weights are used to weight the traveling wave ranging results and the feeder terminal ranging results to obtain the fused fault distance.

7. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 6, characterized in that, The steps for evaluating the data quality of traveling wave ranging results and feeder terminal ranging results include: Obtain the correlation coefficient between traveling wave characteristic parameters and fault distance, and calculate the average value of the correlation coefficients of all traveling wave characteristic parameters as an indicator of the distance representation capability of traveling wave data. Obtain the correlation coefficient between time-frequency domain characteristic parameters and fault distance, and calculate the average value of the correlation coefficients of all time-frequency domain characteristic parameters as an indicator of the distance characterization capability of feeder terminal data; The signal quality level of traveling wave data is evaluated based on the waveform integrity and signal-to-noise ratio of the traveling wave signal. The signal quality level of the feeder terminal data is evaluated based on the consistency of fault current direction determination and the convergence of impedance calculation in the feeder terminal data. By combining the distance representation capability index of traveling wave data with the signal quality level, the comprehensive data quality level of traveling wave ranging results is determined. By combining the distance characterization capability index of feeder terminal data with the signal quality level, the comprehensive data quality level of the feeder terminal ranging results is determined. The overall data quality level of the traveling wave ranging results is combined with the overall data quality level of the feeder terminal ranging results to determine the quality level combination.

8. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 7, characterized in that, The steps for setting initial fusion weights based on quality level combinations include: The basic fusion weight is set according to the combination of quality levels. When the comprehensive data quality level of the traveling wave ranging result is higher than that of the feeder terminal ranging result, the basic weight of the traveling wave ranging is set to 60%-70%. Conversely, the basic weight of the feeder terminal ranging is set to 60%-70%. When the quality levels of both are the same, equal weights are set. When the fault type is a single-phase ground fault, the feeder terminal ranging weight is reduced by 5%-15% on the basic weight, and the traveling wave ranging weight is increased accordingly. When the fault type is a two-phase short circuit or a two-phase-to-ground short circuit, the basic fusion weight remains unchanged; When the fault type is a three-phase short circuit, the feeder terminal ranging weight is increased by 5%-15% on the basis of the basic weight, and the traveling wave ranging weight is reduced accordingly.

9. The method for accurate fault location in distribution networks by integrating traveling wave ranging and feeder terminal data as described in claim 8, characterized in that, The steps for correcting the fused fault distance based on the fault type to obtain the final fault distance include: When the fault type is a single-phase ground fault, the correction factor for the fused fault distance is set to 0.92-0.98; When the fault type is a two-phase short-circuit fault, the correction factor for the fused fault distance is set to 0.98-1.02; When the fault type is a two-phase-to-ground short-circuit fault, the correction factor for the fused fault distance is set to 0.95-1.00; When the fault type is a three-phase short-circuit fault, the correction factor for the fused fault distance is set to 1.00-1.

05.

10. A precise fault location system for distribution networks integrating traveling wave ranging and feeder terminal data, employing the method described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect traveling wave signals and feeder terminal data during power distribution network faults. The feature extraction and classification module is used to extract traveling wave feature parameters from traveling wave signals, extract time-frequency domain feature parameters from feeder terminal data, construct a comprehensive feature vector, and use a fuzzy C-means clustering algorithm to classify the comprehensive feature vector and identify the fault type. The ranging calculation module is used to obtain the traveling wave ranging result by calculating the traveling wave signal and to obtain the feeder terminal ranging result by calculating the feeder terminal data. The fusion positioning module is used to perform weighted fusion of the traveling wave ranging result and the feeder terminal ranging result according to the fusion weight to obtain the fusion fault distance, and to correct the fusion fault distance according to the fault type to obtain the final fault distance, and to calculate the fault point according to the final fault distance.

Images