A power distribution network traveling wave fault location method based on fuzzy interval prediction and wave speed independence

Abstract

This invention proposes a method for fault location in distribution networks based on fuzzy interval prediction and independent of wave velocity, belonging to the technical field of power system automation. The invention includes: deploying a traveling wave detection device for initial time synchronization and continuously collecting historical clock synchronization error data; when a fault occurs, performing initial fault segment location based on the initial traveling wave arrival time and network topology; predicting the current clock error and quantifying the normal distribution characteristics of the prediction error, generating multiple parameter matching pairs to estimate the distances to multiple candidate fault points, and then generating multiple candidate time intervals, determining the optimal time interval based on fuzzy theory; accurately identifying the wavefront of the first reflected wave at the fault point and recording the precise arrival time; constructing a location formula independent of the traveling wave velocity to calculate the fault point location. This invention does not require precise knowledge of the traveling wave velocity and has strong robustness to clock synchronization errors, significantly improving the fault location accuracy and reliability in complex distribution network environments.

Description

Technical Field

[0001] This invention relates to the technical field of power system automation, specifically to a fault location method for distribution networks, and more particularly to a distribution network traveling wave fault location method based on fuzzy interval prediction independent of wave velocity. Background Technology

[0002] With the advancement of new power system construction, the distribution network, as a crucial public infrastructure, is experiencing increasingly complex structures and operating environments, leading to a higher probability of faults. Distribution network fault location technology is a key tool for ensuring a safe and reliable power supply. Traveling wave fault location technology, due to its minimal impact from system structure and high applicability to various fault types, is considered the most promising fault location technology in the power system field to date. However, this technology still faces pressing technical challenges.

[0003] Currently, traveling wave fault location technology, which relies on fault current, is the mainstream approach in both theoretical research and engineering applications. However, existing fault location methods based on the traveling wave principle, especially when applied to complex distribution networks, still face the following key technical challenges:

[0004] High dependence on clock synchronization: Traditional two-end traveling wave ranging methods require high-precision synchronized clocks (such as relying on GPS / BeiDou) at both ends of the line to accurately calculate the time difference of the traveling wave arrival at both ends. Any tiny synchronization error will directly lead to a large positioning error. Although high-precision synchronization devices can mitigate this effect, they will greatly increase engineering costs.

[0005] Traveling wave velocity uncertainty: The traveling wave velocity is affected by line parameters (such as conductor material, ambient temperature, geographical conditions, etc.) and is difficult to know accurately in practical applications. Existing methods usually assume the wave velocity to be a constant value, which introduces uncertainty and causes deviations between theoretical distance measurement results and actual fault location.

[0006] The distribution network has poor structural adaptability: Distribution networks are characterized by a mix of wires and cables, and numerous branches, resulting in a complex structure. Fault traveling waves are easily mixed with reflected waves from non-fault points during transmission. Existing single-ended or double-ended methods used in distribution networks struggle to accurately identify the effective reflected wavefront from the fault point, leading to location failures or decreased accuracy.

[0007] The accuracy of the dual-end traveling wave positioning technology already applied in power distribution networks is significantly affected by the time synchronization accuracy and the uncertainty of the traveling wave velocity. Limited by the inherent principles of the technology, these negative impacts cannot be completely eliminated. Therefore, traveling wave positioning methods that are less dependent on time synchronization accuracy and unaffected by the traveling wave velocity are the future trend of technological development.

[0008] Patent application number 202511415648.X discloses a method and system for locating traveling wave faults at two ends of a multi-control power grid. The method includes: deploying multiple measuring points at the beginning and end of a radial distribution network and at branch terminals to collect topology information and calculate the line-mode wave velocity; extracting the traveling wave line-mode and zero-mode components using post-fault three-phase voltage data through modulus decoupling; selecting the line-mode component with the strongest global abrupt change as a reference, and generating a wavefront time series table through wavelet decomposition and abrupt change detection; combining the topology, time series table, line-mode wave velocity, and corrected zero-mode wave velocity, and using wavefront time difference compensation to correct the fault interval and output the fault location. This invention effectively overcomes the location error caused by clock asynchrony and parameter disturbances through multi-measuring point collaborative wavefront time series analysis and time difference compensation mechanism, achieving accurate ranging unaffected by clock errors and significantly improving ranging accuracy; at the same time, it simplifies the deployment requirements, requiring only a single measuring point on a branch to achieve accurate fault interval identification and location, combining high economy and deployment flexibility. However, the zero-mode wavefront used in the above-mentioned patent for fault location only exists when a ground fault occurs. This invention is only applicable to ground fault situations, and the specific traveling wave velocity needs to be known in advance for fault location. Therefore, the fault location accuracy has an inherent defect that is greatly affected by the uncertainty of the traveling wave velocity. Summary of the Invention

[0009] To address the technical problem of existing technologies being highly dependent on clock synchronization accuracy and traveling wave velocity accuracy, this invention proposes a traveling wave fault location method for distribution networks based on fuzzy interval prediction that is independent of wave velocity, thereby improving the fault location accuracy and reliability under complex distribution network structures.

[0010] To achieve the above objectives, the technical solution of this invention is as follows: a method for locating traveling wave faults in a distribution network based on fuzzy interval prediction independent of wave velocity, comprising the following steps:

[0011] Step 1: Deploy traveling wave detection devices at key nodes of the distribution network, perform preliminary time synchronization on each traveling wave detection device, and continuously collect historical clock synchronization error data of each traveling wave detection device;

[0012] Step 2: When a fault occurs, the initial fault segment location is performed based on the initial arrival time of the traveling wave recorded by each traveling wave detection device and the network topology.

[0013] Step 3: For the initial fault section, the current clock error is predicted using historical clock synchronization error data and the normal distribution characteristics of the prediction error are quantified. Combined with the preset traveling wave velocity range, multiple candidate fault point distances are estimated by generating multiple parameter matching pairs, thereby generating multiple candidate time intervals. Based on fuzzy theory, each candidate time interval is processed to determine the optimal time interval.

[0014] Step 4: Within the optimal time interval, accurately identify the first reflected wavefront of the fault point using signal processing technology, and record the precise arrival time;

[0015] Step 5: Using the initial arrival time of the traveling wave and the arrival time of the first reflected wave from the traveling wave detection devices at both ends of the fault section, construct a positioning formula that is independent of the traveling wave velocity and calculate the location of the fault point.

[0016] Preferably, the key nodes of the distribution network include the substation outlet of the main line, the connection point between the cable and the overhead line, or important branch points; all traveling wave detection devices are connected to the central analysis server through a 5G communication network, and the Beidou time synchronization system is used to perform preliminary time synchronization between the central analysis server and each traveling wave detection device, continuously collecting and storing historical clock synchronization error data of each traveling wave detection device; the central analysis server is responsible for aggregating the traveling wave signals and historical clock error data of each traveling wave detection device, and correcting the timestamp through error prediction;

[0017] When a system fault occurs, each traveling wave detection device captures the transient current traveling wave signal generated by the fault and records its arrival time. The central server receives information and timestamps from all traveling wave detection devices that have detected transient current traveling wave signals. Based on the order of the first arrival times of the traveling wave signals and the network topology, the faulty line segment is initially determined. The method is as follows: the transient current traveling wave signal is processed by a bandpass digital filter to separate the high-frequency current traveling wave signal (10 kHz to 1 MHz) and the low-frequency transient current traveling wave signal (50 Hz to 5 kHz). The central server collects the first arrival times of the traveling wave signals reported by all traveling wave detection devices and sorts them from smallest to largest. According to the network topology, the two adjacent traveling wave detection devices corresponding to the earliest two moments of two adjacent detection points are very likely to be the faulty segment. The phase difference between the high-frequency current traveling wave signal and the low-frequency transient current traveling wave signal at different detection points is calculated using Hilbert transform. If the signs of the phase difference are opposite at the detection end, it indicates that a fault has occurred in the corresponding segment, which is taken as the initial faulty segment.

[0018] Preferably, the method for determining the optimal time interval in step three is as follows:

[0019] Step 3.1: Based on historical clock synchronization error data, use time series forecasting methods to predict the clock error at the current moment. Analyze the distribution characteristics of historical prediction errors to determine that the predicted clock error follows a mean of 0 and a standard deviation of . The normal distribution;

[0020] Step 3.2: Based on normal distribution The criteria, combined with the preset range of traveling wave velocities. Determine the range of values ​​for the actual clock error. ;

[0021] Step 3.3: Based on the minimum clock error Predicted clock error Maximum clock error With minimum traveling wave speed Mean traveling wave velocity Maximum traveling wave speed Nine parameter matching pairs are generated by combining them, and the distances to the nine candidate fault points are estimated using the double-ended ranging principle.

[0022] Step 3.4: Based on the estimated distance to the candidate fault point, predict 9 candidate time intervals for the arrival of the first reflected wave at both ends of the traveling wave detection device;

[0023] Step 3.5: Using the nine candidate time intervals as fuzzy input, construct a time membership function, an energy membership function, and a steepness membership function for each candidate time interval, and calculate the comprehensive confidence of each candidate interval through fuzzy inference;

[0024] Step 3.6: Use the defuzzification method to make a decision on the overall confidence of each candidate time interval, and select the candidate time interval with the highest overall confidence as the best time interval.

[0025] Preferably, the nine parameter matching pairs include:

[0026] , , ;

[0027] , , ;

[0028] , , ;

[0029] For each parameter matching pair, the distance from the candidate fault point to the traveling wave detection device A is estimated using the two-end ranging principle. for:

[0030] ;

[0031] in, This represents the total length of the faulty line section. and These represent the times when the initial traveling wave of the fault arrives at traveling wave detection device A and traveling wave detection device B, respectively. , and They represent the first The synchronous clock error and traveling wave velocity in the parameter matching pair;

[0032] The method for generating the nine candidate time intervals is as follows: based on the estimated distances between the nine candidate fault points... The candidate time interval for the first reflected wave from the fault point to arrive at the traveling wave detection device A is calculated. ,in, Corresponding to the first The lower and upper bounds of the candidate time interval for the arrival of the first reflected wave in the parameter matching pair are determined; similarly, nine candidate time intervals for the arrival of the first reflected wave at the fault point in the traveling wave detection device B are predicted.

[0033] Preferably, the method for predicting the clock error at the current moment based on historical clock synchronization error data is as follows: Historical clock synchronization error data of traveling wave detection devices A and B at both ends are obtained within a certain period before the fault occurs; an error time series is constructed; and a time series prediction method is used to predict the clock error at the current moment. Make predictions;

[0034] The historical prediction errors are fitted with probability distribution characteristics using kernel density estimation. Let the prediction error be... Statistical analysis showed that the prediction error was... Follows a mean of 0 and a standard deviation of normal distribution ;in, This is the actual time when the fault occurred, recorded in the historical clock synchronization error data. The time of failure occurrence is predicted using time series forecasting methods;

[0035] Based on the 3σ criterion of normal distribution, actual clock error It falls within the range with a probability of 99.73%. Within; and minimum clock error Average clock error Maximum clock error ;

[0036] The Lower World Upper Realm ;

[0037] The time series prediction method is an autoregressive moving average model, a long short-term memory network, or a Kalman filter.

[0038] Preferably, the time membership function adopts a triangular membership function, with the midpoint of the candidate time interval as the center and the interval length as the support set;

[0039] The energy membership function is constructed based on the wavelet transform modulus maxima energy of the traveling wave signal within the candidate time interval;

[0040] The steepness membership function is constructed based on the rise rate or abrupt slope of the traveling wave front.

[0041] Preferably, the first Temporal membership degree of each candidate time interval:

[0042] ;

[0043] in, and The first The lower and upper bounds of each candidate time interval. Distance from the middle time The absolute deviation is , For the first Peak time of each candidate time interval For the first The duration of each candidate time interval;

[0044] No. Energy membership degree of each candidate time interval Among them, energy value and They represent the first The and the first The wavelet transform of the traveling wave signal in each candidate time interval is performed on the sum of squared moduli.

[0045] No. Steepness of each candidate time interval ,in, and These represent the steepness of the k-th and j-th candidate time intervals, respectively;

[0046] The overall confidence score is obtained by multiplying the membership scores of time, energy, and steepness together: .

[0047] Preferably, the overall confidence level is obtained by comparing the candidate time intervals. The decision-making process employs the principle of maximum membership: the candidate time interval with the highest overall confidence level is selected as the final optimal time interval. ,in, , These represent the upper and lower bounds of the optimal time interval, respectively.

[0048] Signal processing techniques such as wavelet transform or Hilbert-Huang transform are used to detect the modulus maxima of wavelet coefficients in the traveling wave signal within the optimal time interval. Alternatively, Hilbert-Huang transform is used to extract the instantaneous energy spectrum peaks. The times corresponding to the modulus maxima or energy peaks are identified as the fault current traveling wave fronts. The most significant fault current traveling wave fronts within the optimal time interval are searched and identified, marked as the first reflected wave fronts at the fault point, and the arrival time of the first reflected wave for traveling wave detection device A is recorded. and the arrival time of the first reflected wave of the traveling wave detection device B .

[0049] Preferably, a fault location formula independent of the traveling wave velocity is constructed to calculate the distance from the fault point to the traveling wave detection device A. for:

[0050] ;

[0051] in, The total length of the faulty section. , These represent the times when the initial traveling wave of the fault arrives at traveling wave detection device A and traveling wave detection device B, respectively. and These represent the times when the first reflected wave from the fault point reaches traveling wave detection device A and traveling wave detection device B, respectively.

[0052] Preferably, the energy value ,in, This is the wavelet coefficient matrix obtained by performing wavelet transform on the traveling wave signal within the candidate time interval;

[0053] The steepness ,in, The peak point is at the 1st Indexes in candidate time intervals This represents the amplitude of the traveling wave signal corresponding to the peak point within the k-th candidate interval. This represents the amplitude of the traveling wave signal at the sampling point preceding the peak point within the k-th candidate interval;

[0054] The calculated distance from the fault point to the traveling wave detection device A. Mileage is determined along the corresponding line path in the power distribution network topology. Combined with geographic information system or tower / node coordinate data, the specific latitude and longitude or equipment location is mapped to output the precise geographical location of the fault point.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] Eliminating dependence on wave velocity: In the positioning formula finally adopted in this invention, the traveling wave velocity is completely eliminated in the calculation process, which fundamentally solves the positioning error caused by wave velocity uncertainty and frequency variation characteristics, and greatly improves the accuracy and robustness of ranging.

[0057] The invention reduces the requirements for clock synchronization accuracy by cleverly "absorbing" the initial time synchronization error within the prediction interval through clock error prediction and interval prediction based on historical data and precise search. This error is then corrected by subsequent precise wavefront identification and a wave velocity-independent formula. The final positioning result no longer relies on high-precision absolute time synchronization, allowing the use of lower-cost synchronization schemes (such as standard BeiDou time synchronization), significantly reducing system cost and engineering implementation difficulty.

[0058] Enhancing robustness to clock error prediction uncertainty: This invention introduces a clock error prediction model based on historical data and utilizes the normal distribution characteristics of the prediction error through… The "criteria" are quantified into an error range, enabling the prediction interval to cover the actual clock error with a 99.73% probability, effectively addressing the dynamic changes in clock errors and prediction uncertainties, and further improving the engineering adaptability of the positioning method.

[0059] Enhanced adaptability in complex power distribution networks: This invention uses a prediction and correction mechanism to purposefully search for the first reflected wave of the fault point within the prediction interval, effectively eliminating interference reflected waves generated by other branch points and cable joints in the power distribution network, improving the accuracy and reliability of wavefront identification, and enabling it to better adapt to complex power distribution network environments.

[0060] Enhancing adaptability and robustness in complex power distribution networks: This invention fully retains the theoretical basis of the nine parameter matching pairs, ensuring comprehensive coverage of clock errors and wave velocity uncertainties. Based on this, fuzzy theory is introduced, and through multi-feature fusion and fuzzy inference, the nine candidate time intervals are optimized into a single high-confidence optimal time interval, effectively filtering out interference reflection waves caused by branches, cable joints, etc., significantly improving the accuracy and anti-interference capability of wavefront identification in complex power distribution network environments, and forming a systematic technical architecture of "historical data driven - nine-parameter matching - candidate interval generation - fuzzy fusion - accurate identification - wave velocity independent positioning". Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are 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.

[0062] Figure 1 This is an overall flowchart of the method of the present invention.

[0063] Figure 2 This is a schematic diagram of the multi-terminal traveling wave measurement system in this invention.

[0064] Figure 3 This is a graph showing the historical clock error of Embodiment 2 of the present invention.

[0065] Figure 4 This is a flowchart of the prediction process for the arrival time interval of the first reflected wave at the fault point based on fuzzy theory in Embodiment 2 of the present invention.

[0066] Figure 5 This is a schematic diagram of the generation of 9 parameter matching pairs and candidate time intervals in this invention; where (a) is the correspondence between the time synchronization error and the traveling wave velocity matching pairs, and (b) is the fault distance corresponding to the matching pair.

[0067] Figure 6 The invention embodiment shows the distribution of nine candidate time intervals and their comprehensive confidence scores for devices A and B; wherein, (a) is the distribution of nine candidate time intervals and their comprehensive confidence scores for device A, and (b) is the distribution of nine candidate time intervals and their comprehensive confidence scores for device B.

[0068] Figure 7 The diagram shows the identification of traveling wave signals and reflected waves in an embodiment of the present invention; wherein, (a) is the traveling wave signal of device A, (b) is the traveling wave signal of device B, (c) is the complete signal of device A and the optimal range where the first reflected wave is located, and (d) is the complete signal of device B and the optimal range where the first reflected wave is located.

[0069] Figure 8 The positioning error results at different fault distances in the embodiments of the present invention are shown in the figure. Detailed Implementation

[0070] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0071] Example 1

[0072] like Figure 1As shown, a method for locating traveling wave faults in distribution networks based on fuzzy interval prediction and independent of wave speed is a multi-terminal traveling wave fault location method for distribution networks that does not rely on wave speed and precise clock synchronization. First, the arrival time of the initial traveling wave is obtained through a multi-terminal traveling wave detection device, and the fault section is initially located. Then, the current clock error is predicted based on historical data, and the prediction error range is quantified using the statistical characteristics of normal distribution. Combined with the preset traveling wave speed range, the distance range of the fault point is estimated through matching of 9 parameters, and corresponding 9 candidate time intervals are generated. Next, based on fuzzy theory, each candidate time interval is fuzzified, fuzzy inferenced, and defuzzified to determine the optimal time interval, and the first reflected wave front of the fault point is accurately identified within the optimal time interval. Finally, using the arrival times of the initial traveling wave and the first reflected wave, a location formula independent of the traveling wave speed is constructed to achieve accurate fault location. This invention integrates historical data-driven clock error prediction, normal distribution uncertainty quantification, and fuzzy interval prediction. It does not require precise knowledge of the traveling wave velocity and exhibits strong robustness to clock synchronization errors. It effectively solves the problems of wave velocity uncertainty and high synchronization accuracy requirements in existing traveling wave ranging technologies, significantly improving the fault location accuracy and reliability in complex power distribution network environments. This invention includes the following steps:

[0073] Step 1: Deploy traveling wave detection devices at key nodes of the distribution network, perform preliminary time synchronization on each traveling wave detection device, and continuously collect historical clock synchronization error data of each traveling wave detection device.

[0074] like Figure 2 As shown, the constructed multi-terminal traveling wave measurement system for the distribution network involves installing traveling wave detection devices at key nodes of the distribution network (such as substation outlets on main lines, cable-to-overhead line connections, and important branch points). All traveling wave detection devices are connected to the central analysis server via a 5G communication network. The BeiDou time synchronization system is used to perform initial time synchronization between the central analysis server and each traveling wave detection device. Therefore, this invention no longer relies on high-precision absolute time synchronization, allowing for microsecond-level synchronization errors, reducing the stringent requirements on the 5G network and time synchronization signal, and significantly reducing system construction and maintenance costs. The traveling wave detection devices in the multi-terminal traveling wave measurement system continuously collect and store historical clock synchronization error data from each device, serving as the data source for predicting clock synchronization errors in subsequent fault occurrences. The BeiDou time synchronization system provides a coarse but unified time reference, responsible for initial synchronization of the clocks of the central server and each detection device. The central analysis server is responsible for aggregating the traveling wave data and historical clock error data from each device, correcting the timestamps through error prediction, and then running a positioning algorithm that does not rely on high-precision synchronization to achieve accurate fault location in the distribution network.

[0075] Step 2: When a fault occurs, the initial fault segment location is performed based on the initial arrival time of the traveling wave recorded by each traveling wave detection device and the network topology.

[0076] When a system fault occurs, each traveling wave detection device captures the transient current traveling wave signal generated by the fault and records its arrival time. The central server receives information from all traveling wave detection devices that have detected the traveling wave signal, along with their timestamps. Based on the order of the first arrival times of the traveling wave signals and the network topology, it preliminarily determines the line segment where the fault occurred. Specifically: after capturing the fault transient current traveling wave, each traveling wave detection device separates the 10 kHz to 1 MHz high-frequency current traveling wave signal and the 50 Hz to 5 kHz low-frequency transient current traveling wave signal using a bandpass digital filter; the central server collects the first arrival times of the traveling waves reported by all traveling wave detection devices, sorts them in ascending order of time, and, based on the network topology, identifies the adjacent traveling wave detection devices corresponding to the earliest two times of two adjacent detection points as most likely to be the fault segment; using Hilbert transform, the phase difference between the high-frequency signal and the low-frequency signal at different detection points is calculated. If the sign of this phase difference is opposite at the detection end, it indicates that a fault has occurred in the corresponding segment. For example, the phase difference between adjacent devices A and B corresponding to the earliest two moments of the fault traveling wave is calculated by Hilbert transform and the signs of the phase difference between devices A and B are opposite, thus determining that the fault is located on the line segment between devices A and B.

[0077] Step 3: For the initial fault section, the current clock error is predicted using historical clock synchronization error data and the normal distribution characteristics of its prediction error are quantified. Combined with the preset traveling wave velocity range, multiple candidate fault point distances are estimated by generating multiple parameter matching pairs, thereby generating multiple candidate time intervals. Based on fuzzy theory, each candidate time interval is fuzzified, fuzzy inferenced and defuzzified to determine the optimal time interval.

[0078] This step involves interval prediction of the arrival time of the first reflected wave from the fault point based on fuzzy theory.

[0079] like Figure 4 As shown, for the faulty line section determined in step two, the total length of the line in this section is L, for example, the distance between two towers, which is known in advance, meaning the actual length of the line has already been saved as raw data. Let the traveling wave detection devices at both ends of the faulty line section be device A and device B, respectively, and the recorded arrival times of the initial traveling wave of the fault be respectively... and Considering the traveling wave velocity The unknowns and the time synchronization errors of the traveling wave detection devices at both ends The time interval for the first reflected wave from the fault point to reach device A is predicted through the following sub-steps:

[0080] Step 3.1: Based on historical clock synchronization error data, use time series forecasting methods to predict the clock error at the current moment. The distribution characteristics of historical prediction errors were statistically analyzed, and it was determined that the predicted clock error follows a mean of 0 and a standard deviation of 0. It follows a normal distribution.

[0081] The implementation method for clock error prediction based on historical data is as follows: Historical clock synchronization error data of devices A and B within a certain period before the fault occurred are obtained to construct an error time series. Time series prediction methods (such as Autoregressive Moving Average (ARIMA), Long Short-Term Memory (LSTM) network, or Kalman filtering) are then used to predict the clock error at the current moment. To make predictions, the historical prediction error sequence is fitted with its probability distribution characteristics using kernel density estimation. Let the prediction error be... Statistical analysis showed that the prediction error was... Follows a mean of 0 and a standard deviation of The normal distribution, i.e. . This refers to the actual time when the fault occurred, as recorded in historical data. This represents the predicted time of failure using time series forecasting methods. Prediction error. This demonstrates the degree of inconsistency between the fault occurrence time and the predicted occurrence time. By predicting clock errors and statistical characteristics using time series prediction methods, the search range for the arrival time interval of the first reflected wave of the fault traveling wave can be compressed, significantly reducing noise signals and interference from traveling wave signals of adjacent lines, and improving the accuracy of determining the arrival time of the first reflected wave.

[0082] Step 3.2: Based on the normal distribution "Guidelines", combined with a preset range of traveling wave velocities. Determine the range of values ​​for the actual clock error. .

[0083] Set the range of uncertainty parameters:

[0084] Based on the "3σ criterion" of normal distribution, the actual clock error It falls within the range with a probability of 99.73%. Internal. Define the minimum clock error. Average clock error Maximum clock error At the same time, the traveling wave speed is set. The range of values ​​is (e.g., [98%c, 99%c, 100%c], where c is the speed of light). These represent the minimum traveling wave velocity, average traveling wave velocity, and maximum traveling wave velocity, respectively. The propagation speed of a traveling wave in a line typically fluctuates within this range, but cannot be precisely determined.

[0085] Step 3.3: Based on the minimum clock error Clock error Maximum clock error With minimum traveling wave speed Mean traveling wave velocity Maximum traveling wave speed Nine parameter matching pairs are generated by combining them, and the distances to the nine candidate fault points are estimated using the principle of double-ended ranging.

[0086] The method for generating 9 parameter matching pairs and estimating the distance to candidate fault points is as follows: based on time synchronization error , , Three representative values ​​of traveling wave velocity , , Combined, this forms 9 parameter matching pairs:

[0087] , , ;

[0088] , , ;

[0089] , , ;

[0090] For each parameter matching pair, the distance from the candidate fault point to device A is estimated using the two-end ranging principle. for:

[0091] ;

[0092] in, This represents the total length of the faulty line section. and These represent the times when the initial traveling wave of the fault arrives at device A and device B, respectively. , and They represent the first Synchronization clock error and traveling wave velocity in each matching pair.

[0093] Step 3.4: Based on the estimated distance to the candidate fault point, predict nine candidate time intervals for the arrival of the first reflected wave at both ends of the traveling wave detection device.

[0094] The method for generating nine candidate time intervals is as follows: based on the estimated distances between the nine candidate fault points... The candidate time interval for the first reflected wave from the fault point to arrive at device A is calculated. Among them, the lower bound of time Upper Boundary of Time ,in, , Corresponding to the first The lower and upper bounds of the candidate time intervals for the arrival of the first reflected wave in each matching pair are determined. Similarly, nine candidate time intervals for the arrival of the first reflected wave at the fault point in device B can be predicted. These nine candidate intervals together constitute the input space for fuzzy inference.

[0095] Step 3.5: Using the nine candidate time intervals as fuzzy inputs, construct a time membership function, an energy membership function, and a steepness membership function for each candidate time interval, and calculate the comprehensive confidence of each candidate interval through fuzzy inference.

[0096] The method for fuzzification and membership function construction is as follows: extract three fuzzy features for each candidate time interval: time membership, energy membership, and steepness membership.

[0097] Time membership The time membership function uses a triangular membership function, centered at the midpoint of the candidate time interval, with the interval length as the support set. It characterizes the degree to which each moment within the candidate time interval belongs to a valid reflected wavefront. , Indicates the first Let the temporal membership degree of each candidate time interval be denoted as . The lower bound of the time intervals for each candidate time interval is: The upper bound of time is ,but:

[0098] Mid-time , No. Duration of each candidate time interval .

[0099] Order No. The peak time of each candidate time interval is The peak value is obtained by performing peak detection on the traveling wave signal waveform within the candidate time interval and taking the time corresponding to the maximum amplitude. The absolute deviation from the midpoint time is... Then, the time membership degree is:

[0100] .

[0101] In other words, the closer the peak value is to the center of the interval, the higher the time membership degree. This index ensures that the candidate fault location whose actual arrival time of the traveling wave front is closer to the theoretical center point of the interval obtains a higher membership degree, thereby effectively suppressing peak offset interference caused by noise and improving the reliability of identifying the arrival time of the first reflected wave of the fault traveling wave.

[0102] Energy membership Based on the wavelet transform modulus maxima energy of the traveling wave signal within the candidate time interval, an energy membership function is constructed to reflect the significance of the wavefront. , Indicates the first The energy membership of each candidate time interval is used to perform CWT transform (using Morlet wavelet, scale range 1~64) on the traveling wave signal within the candidate time interval to obtain the wavelet coefficient matrix. Take its absolute value to obtain the modulus. For the first Given a time interval, select all time points within that candidate time interval. and all scales Sum of squares of moduli: Then, the energy values ​​of all intervals are normalized to [0,1]. The higher the energy value, the higher the energy membership degree. and They represent the first The and the first The candidate time interval traveling wave signals are subjected to CWT transform with sum of squared magnitudes. Energy membership is used to quantize the wavelet energy of the traveling wave signal, giving higher weight to candidate intervals with significant wavefronts and concentrated energy, thereby effectively enhancing the identification capability of real fault reflection wavefronts and suppressing noise interference.

[0103] Steepness membership Based on the rise rate or abrupt slope (absolute value of forward difference) of the traveling wave front, a steepness membership function is constructed to characterize the sharpness of the wave front. Let the peak point be at the [missing information]. The index in the candidate time interval is The steepness is obtained by performing peak detection on the traveling wave signal waveform within the candidate time interval and taking the moment corresponding to the maximum amplitude; then: Then normalization is performed: The greater the steepness, the higher the steepness membership degree. This represents the original traveling wave signal amplitude corresponding to the peak point within the k-th candidate interval. This indicates the signal amplitude at the sampling point preceding the peak point. and They represent the first The and the first The steepness of each candidate time interval.

[0104] The overall confidence score is obtained by multiplying the membership scores of time, energy, and steepness together: This method integrates the temporal position, energy concentration, and wavefront steepness of the waveform through a product form, with these three factors jointly determining the reliability of an interval as a valid reflecting wavefront. By implementing the "AND" operation in fuzzy logic through the product form, it integrates the three fuzzy features of temporal position, energy concentration, and wavefront steepness, thereby improving the robustness and accuracy of reflecting wavefront identification.

[0105] Step 3.6: Use the defuzzification method to make a decision on the overall confidence of each candidate time interval, and select the candidate time interval with the highest overall confidence as the best time interval.

[0106] The method for determining the optimal time interval for the first reflected wavefront based on the comprehensive confidence score is as follows: The defuzzification method directly compares the comprehensive confidence scores calculated from each candidate time interval. The principle of maximum membership (i.e., selecting the overall confidence level) is adopted. The decision is made by using the interval corresponding to the maximum value, thereby transforming the fuzzy confidence distribution into a clear optimal time interval, providing a unique and reliable time interval input for subsequent accurate fault location.

[0107] The candidate time interval with the highest overall confidence level was selected as the final optimal time interval. The same principle applies to device A, and the optimal time interval for device B can be obtained similarly. , These represent the upper and lower bounds of the optimal time interval, respectively.

[0108] Step 4: Within the optimal time interval, accurately identify the first reflected wavefront of the fault point using signal processing technology, and record the precise arrival time.

[0109] The method for accurately identifying the first reflected wavefront within the optimal time interval is as follows:

[0110] The optimal time interval determined in step three Within this process, wavelet transform or Hilbert-Huang transform is used to perform wavelet transform on the original traveling wave signal within the optimal time interval. The modulus maxima of the wavelet coefficients are detected, or the instantaneous energy spectrum peak is extracted using Hilbert-Huang transform. The time corresponding to the modulus maxima or energy peak is identified as the most significant fault current traveling wave front. The most significant fault current traveling wave front within the optimal time interval is searched and identified, marked as the first reflected wave front of the fault point, and its precise arrival time is recorded. (For device A) and (For device B).

[0111] The signal processing techniques include, but are not limited to, wavelet transform, used to identify traveling wavefronts within the optimal time interval.

[0112] Step 5: Using the initial arrival time of the traveling wave and the arrival time of the first reflected wave from the traveling wave detection devices at both ends of the fault section, construct a positioning formula that is independent of the traveling wave velocity and calculate the location of the fault point.

[0113] Two-ended traveling wave fault location that does not depend on wave velocity

[0114] Using the initial traveling wave arrival time of the fault obtained in step two and The arrival time of the first reflected wave obtained in step four and A fault location formula independent of the traveling wave velocity is constructed to calculate the distance from the fault point to device A. for:

[0115] ;

[0116] in, The distance from the fault point to the traveling wave detection device A at one end is denoted as . The total length of the faulty section. , These represent the times when the initial traveling wave of the fault arrives at device A and device B, respectively. and These represent the times when the first reflected wave from the fault point reaches device A and device B, respectively.

[0117] Step 6: Output the location of the fault point.

[0118] Calculated distance Combined with the distribution network topology, the calculated fault distance Mileage is determined along the corresponding line path in the power distribution network topology. Combined with geographic information system (GIS) or tower / node coordinate data, the specific latitude and longitude or equipment location is mapped to obtain the precise geographical location of the fault point and the final fault location result.

[0119] Example 2

[0120] A method for locating traveling wave faults in distribution networks based on fuzzy interval prediction and independent of wave velocity is proposed in this embodiment. The method uses a wave with a length of... Taking an overhead power line as an example, traveling wave detection devices A and B are installed at both ends of the line, with a sampling frequency of 10 MHz. The system has continuously collected and stored historical clock synchronization error data of devices A and B over the past 30 days. The actual location of the fault point is at a distance of [distance missing] from traveling wave detection device A. The fault time is defined as .

[0121] Step 1: Constructing a multi-terminal traveling wave measurement system

[0122] Traveling wave detection devices A and B are deployed at the substation outlet of the main power distribution line. All traveling wave detection devices are connected to the central analysis server via a 5G communication network and use the BeiDou time synchronization system for initial time synchronization, allowing for synchronization errors at the microsecond level. The system continuously collects historical clock synchronization error data from each traveling wave detection device.

[0123] Step 2: Fault Section Location

[0124] When a single-phase ground fault occurs on the line, traveling wave detection devices A and B respectively capture the transient current traveling wave generated by the fault and record the arrival time of the initial traveling wave of the fault. , Based on the order of arrival of the traveling waves and the network topology, the central server initially determined that the fault was located on the line segment between traveling wave detection device A and device B.

[0125] Step 3: Predict the arrival time interval of the first reflected wave at the fault point based on fuzzy theory. The specific implementation includes the following steps:

[0126] Step 3.1: Clock error prediction based on historical clock synchronization error data

[0127] Historical clock synchronization error data for devices A and B over the past 30 days are obtained. Preferably, an ARIMA (Autoregressive Moving Average) model is used to predict the clock error at the current moment, yielding the predicted value. The distribution characteristics of historical prediction errors were analyzed, and the prediction errors were found to follow a mean of 0 and a standard deviation of [missing value]. It follows a normal distribution. Figure 3 The historical clock synchronization error data is obtained by collecting historical clock synchronization error data from devices A and B over the past 30 days and plotting it as a time series curve or scatter plot in chronological order. This plot can visually show the trend of error changes over time, whether there is periodic drift, outliers, and the overall fluctuation range, thereby verifying the stationarity of the error series and providing a basis for ARIMA modeling.

[0128] Step 3.2: Set the range of uncertainty parameters

[0129] Based on the "3σ criterion" of normal distribution, the actual clock error falls within the interval with a probability of 99.73%. =[0.3-0.6,0.3+0.6]=[−0.3,0.9]μs. At this time: , , Meanwhile, based on the overhead line parameters, the range of values ​​for the traveling wave velocity is set as follows: , , .

[0130] Step 3.3: Generate 9 parameter matching pairs and estimate the distance to candidate fault points.

[0131] Will , , and , , Nine parameter matching pairs are formed. For each parameter matching pair, the distance from the candidate fault point to device A is estimated using the two-end ranging formula. for:

[0132] ;

[0133] in, , The calculation results are shown in Table 1:

[0134] Table 1. Estimated distance to candidate fault points based on 9 parameter matching pairs

[0135]

[0136] The distances of all candidate fault points are within the range of 0 to 15,000 m, which is consistent with physical reality. Figure 5 This is a schematic diagram of the generation of 9 parameter matching pairs and candidate time intervals in this invention. It combines the predicted clock error value with the traveling wave velocity into 9 sets of parameter matching pairs, and calculates the fault distance corresponding to each set of parameters using the double-ended traveling wave ranging formula, thereby drawing (a) the matching correspondence and (b) the fault distance distribution map. Figure 5 It can be intuitively seen how different combinations of clock errors and wave speeds affect the fault location results, thereby generating multiple candidate time intervals and providing a basis for subsequent fuzzy inference.

[0137] Step 3.4: Generate 9 candidate time intervals (taking device A as an example)

[0138] Based on the distance of candidate fault points Calculate the candidate time interval for the first reflected wave from the fault point to arrive at device A: , To ensure that each interval contains at least one sampling point (sampling interval 0.1 μs), the interval width is set to be no less than 0.2 μs. Nine candidate time intervals are calculated as shown in Table 2 (unit: μs).

[0139] Table 2 Nine candidate time intervals for device A

[0140]

[0141] Similarly, for device B, the distance from the fault point to device B is... Nine candidate time intervals for device B can be obtained as shown in Table 3 (unit: μs):

[0142] Table 3. Nine candidate time intervals for device B

[0143]

[0144] As can be seen from Tables 2 and 3, the width of the nine candidate time intervals for device A is relatively narrow (approximately 0.68~0.71 μs), while the width of the intervals for device B is significantly wider (approximately 1.34~1.38 μs). This reflects that because device B is farther from the fault point, the traveling wave propagation time is longer, which amplifies the clock error and wave velocity uncertainty, resulting in a wider range of candidate time intervals. At the same time, the center values ​​of each interval in the two tables show a regular shift as the sequence number changes, reflecting the influence of different parameter matching on the propagation time.

[0145] Step 3.5: Fuzzification and Membership Function Construction

[0146] Extract three fuzzy features for each candidate time interval:

[0147] Time membership The triangular membership function is used, with the midpoint of the candidate time interval as the center and the interval length as the support set, to characterize the degree to which a time within the interval belongs to an effective reflected wavefront. For example, for the fifth interval [50.93, 51.63] μs in Table 2, the midpoint is 51.28 μs and the interval length is 0.69 μs. If the peak value appears near the midpoint, the time membership degree is close to 1.

[0148] Energy membership The wavelet transform modulus maximum energy within each interval reflects the saliency of the wavefront. The sum of squares of the wavelet transform modulus within each interval is calculated; the interval with the highest energy corresponds to an energy membership degree of 1, and the rest are normalized proportionally.

[0149] Steepness membership The sharpness of the wavefront is characterized by the rise rate (maximum signal differential value) of the traveling wavefront. Normalization is also applied.

[0150] Step 3.6: Determine the optimal time interval for the first reflected wavefront based on the overall confidence level.

[0151] The comprehensive confidence score is calculated based on three fuzzy inference factors: time membership, energy membership, and steepness membership. For each candidate time interval... (Total 9), calculate the values ​​of these three membership degrees separately, and then multiply them together to obtain the overall confidence level of the interval: .

[0152] Figure 6 The confidence level is obtained by calculating the time membership degree, energy membership degree, and steepness membership degree for each of the nine candidate time intervals for both device A and device B, and then multiplying these three values ​​together. Then, a bar chart was drawn with the candidate interval number as the horizontal axis and the overall confidence level as the vertical axis; through Figure 6 It can be seen that the distribution characteristics of the comprehensive confidence of devices A and B (e.g., which number interval has the highest confidence) and the difference in the peak position of the confidence of the two devices are obvious, so the candidate interval with the highest confidence is selected as the best time interval for the first reflected wavefront.

[0153] The overall confidence level calculation results are shown in Table 4 (taking device A as an example):

[0154] Table 4 Overall Confidence Level of Device A

[0155]

[0156] It can be seen that the third interval has the highest overall confidence level (0.125), and [50.17, 50.85] μs is determined to be the optimal time interval for device A.

[0157] Similarly, fuzzy reasoning was performed on device B, and the comprehensive confidence calculation results are shown in Table 5. It can be seen that the comprehensive confidence of the 7th interval is the highest (0.867), and its optimal time interval is [99.42, 100.76] μs.

[0158] Table 5 Overall Confidence Level of Device A

[0159]

[0160] Figure 7 Sub-figures (a) and (b) are obtained by separately acquiring the original fault traveling wave signals of devices A and B, plotting their complete time-domain waveforms; then, the optimal time intervals determined in step 3.6 are superimposed on the complete waveforms to obtain sub-figures (c) and (d). Figure 7 It can be seen that the initial fault wavefront and the first reflected wavefront can be clearly identified in the traveling wave waveforms of both device A and device B; the optimal time interval accurately covers the significant positions of their respective wavefronts, and the interval widths are different (the interval for device A is narrower, and the interval for device B is wider), which verifies the effectiveness of the optimal interval selection method based on fuzzy reasoning.

[0161] Step 4: Accurately identify the first reflected wave head within the optimal time interval.

[0162] Within the optimal time interval [50.17, 50.85] μs of device A, wavelet transform was used to search for the traveling wave signal and identify the most significant reflected wavefront. The signal peak appeared at 50.50 μs, therefore it was recorded as follows: Similarly, within the optimal time interval [99.42, 100.76] μs of device B, the arrival time of the reflected wavefront is identified: .

[0163] Step 5: Two-End Traveling Wave Fault Location Independent of Wave Velocity

[0164] Using the initial traveling wave arrival time , and the arrival time of the first reflected wave , Substituting into the positioning formula that is independent of wave speed:

[0165]

[0166] Calculate the time difference: ,

[0167] Then the distance: .

[0168] The absolute positioning error is .

[0169] Step Six: Output Fault Location

[0170] Calculated distance Combined with the distribution network topology, the fault point was determined to be 5035.1m downstream of device A, and the location result was output. To verify the effectiveness of the proposed invention, Table 6 shows the location results at fault distances of 5500m-10000m from end A.

[0171] Table 6 Location Results

[0172]

[0173] Figure 8 It is a line graph or scatter plot drawn based on the 11 sets of different actual fault distances (5000m~10000m) and their corresponding absolute error data in Table 6, with the actual fault location as the horizontal axis and the absolute error as the vertical axis; through Figure 8 It can be seen that within the range of 5000m to 10000m, the absolute positioning error is generally small (maximum approximately 35.54m), and the error distribution is relatively uniform, without a significant increasing trend with increasing distance. This verifies that the method of the present invention has good positioning accuracy and stability under different fault distances.

[0174] In this embodiment, the maximum actual fault error was 35.54 m, verifying the accuracy of the method of the present invention. Even in the presence of clock error (0.5 μs) and wave velocity uncertainty, the reflected wavefront can still be accurately identified and wave velocity-independent precise positioning can be achieved through matching of 9 parameters and fuzzy inference.

[0175] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for power distribution network fault location based on fuzzy interval prediction and wave speed independent, characterized in that, The steps are as follows: Step 1: Deploy traveling wave detection devices at key nodes of the distribution network, perform preliminary time synchronization on each traveling wave detection device, and continuously collect historical clock synchronization error data of each traveling wave detection device; Step 2: When a fault occurs, the initial fault segment location is performed based on the initial arrival time of the traveling wave recorded by each traveling wave detection device and the network topology. Step 3: For the initial fault section, the current clock error is predicted using historical clock synchronization error data and the normal distribution characteristics of the prediction error are quantified. Combined with the preset traveling wave velocity range, multiple candidate fault point distances are estimated by generating multiple parameter matching pairs, thereby generating multiple candidate time intervals. Based on fuzzy theory, each candidate time interval is processed to determine the optimal time interval. Step 4: Within the optimal time interval, accurately identify the first reflected wavefront of the fault point using signal processing technology, and record the precise arrival time; Step 5: Using the initial arrival time of the traveling wave and the arrival time of the first reflected wave from the traveling wave detection devices at both ends of the fault section, construct a positioning formula that is independent of the traveling wave velocity and calculate the location of the fault point.

2. The method of claim 1, wherein the method is a traveling wave fault location method for power distribution networks based on fuzzy interval prediction and independent of wave velocity. The key nodes of the power distribution network include the substation outlet of the main line, the connection point of the cable and the overhead line, or important branch points; all traveling wave detection devices are connected to the central analysis server through the 5G communication network, and the Beidou time synchronization system is used to perform preliminary time synchronization between the central analysis server and each traveling wave detection device, and to continuously collect and store the historical clock synchronization error data of each traveling wave detection device. The central analysis server is responsible for aggregating traveling wave signals and historical clock error data from various traveling wave detection devices, and correcting timestamps through error prediction. When a system fault occurs, each traveling wave detection device captures the transient current traveling wave signal generated by the fault and records its arrival time. The central server receives information and timestamps from all traveling wave detection devices that have detected transient current traveling wave signals. Based on the order of the first arrival times of the traveling wave signals and the network topology, the faulty line segment is initially determined. The method is as follows: the transient current traveling wave signal is processed by a bandpass digital filter to separate the high-frequency current traveling wave signal (10 kHz to 1 MHz) and the low-frequency transient current traveling wave signal (50 Hz to 5 kHz). The central server collects the first arrival times of the traveling wave signals reported by all traveling wave detection devices and sorts them from smallest to largest. According to the network topology, the two adjacent traveling wave detection devices corresponding to the earliest two moments of two adjacent detection points are very likely to be the faulty segment. The phase difference between the high-frequency current traveling wave signal and the low-frequency transient current traveling wave signal at different detection points is calculated using Hilbert transform. If the signs of the phase difference are opposite at the detection end, it indicates that a fault has occurred in the corresponding segment, which is taken as the initial faulty segment.

3. The method of claim 1 or 2, wherein the method is a traveling wave fault location method for power distribution networks that is independent of wave velocity based on fuzzy interval prediction. The method for determining the optimal time interval in step three is as follows: Step 3.1: Based on historical clock synchronization error data, predict the clock error at the current time using a time series prediction method ; The distribution characteristics of the statistical history prediction error are predicted, and it is determined that the predicted clock error is subject to a normal distribution with a mean of 0 and a standard deviation of ​ Step 3.2: based on normal distribution criteria, in combination with a preset range of wave velocity values of the actual clock error ; Step 3.3: Based on the minimum clock error Predicted clock error Maximum clock error With minimum traveling wave speed Mean traveling wave velocity Maximum traveling wave speed Nine parameter matching pairs are generated by combining them, and the distances to the nine candidate fault points are estimated using the double-ended ranging principle. Step 3.4: Based on the estimated distance to the candidate fault point, predict 9 candidate time intervals for the arrival of the first reflected wave at both ends of the traveling wave detection device; Step 3.5: Using the nine candidate time intervals as fuzzy input, construct a time membership function, an energy membership function, and a steepness membership function for each candidate time interval, and calculate the comprehensive confidence of each candidate interval through fuzzy inference; Step 3.6: Use the defuzzification method to make a decision on the overall confidence of each candidate time interval, and select the candidate time interval with the highest overall confidence as the best time interval.

4. The method for locating traveling wave faults in distribution networks based on fuzzy interval prediction independent of wave velocity, as described in claim 3, is characterized in that... The nine parameter matching pairs include: 、 、 ； 、 、 ； 、 、 ； For each parameter matching pair, the distance from the candidate fault point to the traveling wave detection device A is estimated using the two-end ranging principle. for: ； in, This represents the total length of the faulty line section. and These represent the times when the initial traveling wave of the fault arrives at traveling wave detection device A and traveling wave detection device B, respectively. , and They represent the first The synchronous clock error and traveling wave velocity in the parameter matching pair; The method for generating the nine candidate time intervals is as follows: based on the estimated distances between the nine candidate fault points... The candidate time interval for the first reflected wave from the fault point to arrive at the traveling wave detection device A is calculated. ,in, Corresponding to the first The lower and upper bounds of the candidate time interval for the arrival of the first reflected wave in the parameter matching pair are determined; similarly, nine candidate time intervals for the arrival of the first reflected wave at the fault point in the traveling wave detection device B are predicted.

5. The method for locating traveling wave faults in distribution networks based on fuzzy interval prediction independent of wave velocity, as described in claim 4, is characterized in that... The method for predicting the clock error at the current moment based on historical clock synchronization error data is as follows: Historical clock synchronization error data from traveling wave detection devices A and B at both ends are acquired within a certain period before the fault occurs. An error time series is constructed, and a time series prediction method is used to predict the clock error at the current moment. Make predictions; The historical prediction errors are fitted with probability distribution characteristics using kernel density estimation. Let the prediction error be... Statistical analysis showed that the prediction error was... Follows a mean of 0 and a standard deviation of normal distribution ;in, This is the actual time when the fault occurred, recorded in the historical clock synchronization error data. The time of failure occurrence is predicted using time series forecasting methods; Based on the 3σ criterion of normal distribution, actual clock error It falls within the range with a 99.73% probability. Within; and minimum clock error Average clock error Maximum clock error ; The Lower World Upper Realm ; The time series prediction method is an autoregressive moving average model, a long short-term memory network, or a Kalman filter.

6. The method for locating traveling wave faults in distribution networks based on fuzzy interval prediction independent of wave velocity, as described in claim 4 or 5, is characterized in that... The time membership function adopts a triangular membership function, with the midpoint of the candidate time interval as the center and the interval length as the support set; The energy membership function is constructed based on the wavelet transform modulus maxima energy of the traveling wave signal within the candidate time interval; The steepness membership function is constructed based on the rise rate or abrupt slope of the traveling wave front.

7. The method for locating traveling wave faults in distribution networks based on fuzzy interval prediction independent of wave velocity as described in claim 6, characterized in that, No. Temporal membership degree of each candidate time interval: ； in, and The first The lower and upper bounds of each candidate time interval. Distance from the middle time The absolute deviation is , For the first Peak time of each candidate time interval For the first The duration of each candidate time interval; No. Energy membership degree of each candidate time interval Among them, energy value and They represent the first The and the first The wavelet transform of the traveling wave signal in each candidate time interval is performed on the sum of squared moduli. No. Steepness of each candidate time interval ,in, and These represent the steepness of the k-th and j-th candidate time intervals, respectively; The overall confidence score is obtained by multiplying the membership scores of time, energy, and steepness together: .

8. The method for locating traveling wave faults in distribution networks based on fuzzy interval prediction independent of wave velocity, as described in claim 7, is characterized in that... The overall confidence level is obtained by comparing the time intervals of each candidate. The decision-making process employs the principle of maximum membership: the candidate time interval with the highest overall confidence level is selected as the final optimal time interval. ,in, , These represent the upper and lower bounds of the optimal time interval, respectively. Signal processing techniques such as wavelet transform or Hilbert-Huang transform are used to detect the modulus maxima of wavelet coefficients in the traveling wave signal within the optimal time interval. Alternatively, Hilbert-Huang transform is used to extract the instantaneous energy spectrum peaks. The times corresponding to the modulus maxima or energy peaks are identified as the fault current traveling wave fronts. The most significant fault current traveling wave fronts within the optimal time interval are searched and identified, marked as the first reflected wave fronts at the fault point, and the arrival time of the first reflected wave for traveling wave detection device A is recorded. and the arrival time of the first reflected wave of the traveling wave detection device B .

9. The method for locating traveling wave faults in distribution networks based on fuzzy interval prediction independent of wave velocity, as described in claim 8, is characterized in that... Construct a fault location formula independent of the traveling wave velocity, and calculate the distance from the fault point to the traveling wave detection device A. for: ； in, The total length of the faulty section. , These represent the times when the initial traveling wave of the fault arrives at traveling wave detection device A and traveling wave detection device B, respectively. and These represent the times when the first reflected wave from the fault point reaches traveling wave detection device A and traveling wave detection device B, respectively.

10. The method for locating traveling wave faults in distribution networks based on fuzzy interval prediction independent of wave velocity, as described in any one of claims 7-9, is characterized in that... The energy value ,in, This is the wavelet coefficient matrix obtained by performing wavelet transform on the traveling wave signal within the candidate time interval; steepness ,in, The peak point is at the 1st Indexes in candidate time intervals This represents the amplitude of the traveling wave signal corresponding to the peak point within the k-th candidate interval. This represents the amplitude of the traveling wave signal at the sampling point preceding the peak point within the k-th candidate interval; The calculated distance from the fault point to the traveling wave detection device A. Mileage is determined along the corresponding line path in the power distribution network topology. Combined with geographic information system or tower / node coordinate data, the specific latitude and longitude or equipment location is mapped to output the precise geographical location of the fault point.

Images