Power distribution network fault positioning method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of intelligent power grids, in particular to a power distribution network fault positioning method and device, computer equipment and a storage medium, and the method comprises the steps: collecting power distribution network data, and carrying out the standardization preprocessing of the power distribution network data; carrying out traveling wave signal feature extraction and enhancement processing on the basis of the power distribution network data subjected to standardization preprocessing; identifying the moment when the fault initial traveling wave arrives at the monitoring point from the enhanced and separated traveling wave signals; the traveling wave signal, the power frequency electrical quantity, the switching state and topological structure multi-mode information are integrated, and intelligent fault diagnosis is carried out by using an AI algorithm; the specific position of a fault point is calculated by using the calibrated time difference of arrival of the multi-terminal traveling waves in combination with the line topology and the corrected wave velocity, a positioning result is output through a multi-source information verification and optimization algorithm, the positioning precision is improved, a reliable basis is provided for rapid first-aid repair, the power failure time is effectively shortened, the power supply reliability of a power distribution network is improved, and the power utilization quality of a user is guaranteed. And the economic loss caused by faults is reduced.

Description

Technical Field

[0001] This invention relates to the field of smart grid technology, specifically to a method, apparatus, computer equipment, and storage medium for locating faults in a distribution network. Background Technology

[0002] As a key link directly connecting the power system and users, the operation of the distribution network directly affects users' electricity experience and the order of social production and life. However, the distribution network is characterized by its wide distribution of lines, many branches, complex topology, and harsh operating environment, resulting in a high frequency of faults. Fault location is the first step in the fault handling of the distribution network, and its efficiency and accuracy directly determine the speed of fault repair and the time of power restoration.

[0003] Existing fault location technologies for distribution networks have several shortcomings: First, they rely on a single data type, such as traveling waves or power frequency electrical quantities, which cannot comprehensively reflect fault characteristics and are susceptible to interference, leading to inaccurate location. Second, their signal processing capabilities are weak; in complex noise environments, they struggle to effectively extract and enhance fault signal features, making it difficult to identify the arrival time of the initial traveling wave and affecting location accuracy. Third, they lack multi-modal information fusion, failing to comprehensively utilize various information sources, making it impossible to accurately determine the fault situation and prone to misjudgment. Fourth, their location algorithms have limited accuracy, failing to fully consider factors such as line topology and wave velocity changes, resulting in inaccurate calculation of the arrival time difference of multi-terminal traveling waves, and lacking multi-source information verification and optimization, leading to large errors in fault location results. This makes it difficult to meet the actual needs of rapid and accurate fault location in distribution networks, affecting fault repair efficiency and power supply reliability.

[0004] In view of this, this application proposes a method, device, computer equipment and storage medium for fault location in power distribution networks. Summary of the Invention

[0005] The purpose of this invention is to provide a method, apparatus, computer equipment, and storage medium for locating faults in power distribution networks, so as to solve the above-mentioned problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for locating faults in a power distribution network includes the following steps:

[0008] S1. Collect power distribution network data and perform standardized preprocessing on the power distribution network data;

[0009] S2. Based on the standardized preprocessed distribution network data, perform traveling wave signal feature extraction and enhancement processing;

[0010] S3. Identify the moment when the initial traveling wave of the fault arrives at the monitoring point from the enhanced and separated traveling wave signal;

[0011] S4. Integrate traveling wave signals, power frequency electrical quantities, switch status and topology multimodal information, and use AI algorithms for intelligent fault diagnosis;

[0012] S5. Using the calibrated multi-terminal traveling wave arrival time difference, combined with the line topology and the corrected wave velocity, calculate the specific location of the fault point, and output the location result through multi-source information verification and optimization algorithm.

[0013] Preferably, step S1 specifically includes:

[0014] S11. Collect the power distribution network data, which includes broadband traveling wave and power frequency signal;

[0015] S12. Perform time alignment and quality verification on the power distribution network data;

[0016] S13. Perform preliminary signal purification processing based on wavelet threshold denoising.

[0017] Preferably, step S2 specifically includes:

[0018] S21. Based on time-frequency analysis, extract the image features of the fault traveling wave;

[0019] S22. Perform homomorphic filtering enhancement on the traveling wave front;

[0020] S23. Based on empirical mode decomposition, perform mode separation and selection.

[0021] Preferably, step S3 specifically includes:

[0022] S31. Based on the wavelet transform modulus maxima, perform preliminary judgment on the arrival time of the wavefront;

[0023] S32. Combine the target recognition algorithm of CNN to perform wave head calibration;

[0024] S33. Based on line parameter correction, perform wavefront time compensation.

[0025] Preferably, step S4 specifically includes:

[0026] S41. Perform early fusion of multimodal features and joint representation learning;

[0027] S42. Based on the attention mechanism, perform mid-term feature fusion and optimization;

[0028] S43. Based on the DS evidence theory, late-stage decision fusion is carried out.

[0029] Preferably, step S5 specifically includes:

[0030] S51. Based on the time difference of the traveling wave at both ends, perform preliminary location calculations for distribution network faults;

[0031] S52. Perform multi-terminal integrated positioning and result optimization;

[0032] S53. Verification of results by combining topology and electrical quantities with spatial constraint optimization.

[0033] To address the aforementioned technical problems, the present invention also provides a power distribution network fault location device, which employs the following technical solution, including:

[0034] The preprocessing module is used to collect power distribution network data and perform standardized preprocessing on the power distribution network data;

[0035] The feature extraction module is used to extract and enhance the features of the traveling wave signal based on the standardized preprocessed distribution network data.

[0036] The identification module is used to identify the moment when the initial traveling wave of the fault arrives at the monitoring point from the enhanced and separated traveling wave signal;

[0037] The diagnostic module integrates multi-modal information such as traveling wave signals, power frequency electrical quantities, switch status, and topology structure, and uses AI algorithms for intelligent fault diagnosis.

[0038] The output module is used to calculate the specific location of the fault point by using the calibrated multi-terminal traveling wave arrival time difference, combined with the line topology and the corrected wave velocity, and output the location result through multi-source information verification and optimization algorithms.

[0039] Preferably, the identification module includes:

[0040] The initial judgment module is used to make an initial judgment on the arrival time of the wavefront based on the modulus maxima of wavelet transform;

[0041] The calibration module is used to perform wavehead calibration in conjunction with the CNN target recognition algorithm;

[0042] The compensation module is used to perform wavefront time compensation based on line parameter correction.

[0043] To address the aforementioned technical problems, the present invention also provides a computer device that employs the technical solution described below, comprising a memory, a processor, and a network interface. The memory stores computer-readable instructions, and the processor, when executing the computer-readable instructions, implements the steps of the power distribution network fault location method as described in any one of claims 1 to 6.

[0044] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium, which employs the technical solution described below. The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the aforementioned power distribution network fault location method.

[0045] The beneficial effects of this invention are:

[0046] 1. In this invention, by combining time alignment, quality verification and wavelet threshold denoising, not only are the dimensional differences between different types of data such as broadband traveling waves and power frequency signals eliminated, but the data inconsistency caused by different equipment models, transmission delays and environmental interference during the data acquisition process of the power distribution network is also solved. The processed data has a unified time reference, higher integrity and signal-to-noise ratio, providing high-quality data support for subsequent traveling wave signal analysis and multimodal information fusion, avoiding analysis bias caused by data quality defects, and allowing the value of each type of data to be fully explored.

[0047] 2. In this invention, by innovatively integrating multimodal information such as traveling wave signals, power frequency electrical quantities, switch states, and topology, the shortcomings of traditional technologies that rely on only a single data type are overcome. The high-frequency characteristics of traveling wave signals can quickly capture transient fault characteristics, power frequency electrical quantities can reflect the steady-state fault conditions, and switch states and topology provide the background of the power grid's physical connection and operating status. The complementary fusion of multiple types of information comprehensively covers the key characteristics in the fault occurrence and development process, fundamentally improving the comprehensiveness and accuracy of fault diagnosis.

[0048] 3. In this invention, by integrating multimodal information and using AI algorithm diagnosis, information such as traveling waves, power frequency electrical quantities, switch status and topology is comprehensively utilized to give full play to the advantages of each mode, achieve more comprehensive and accurate fault diagnosis, and reduce the misjudgment rate;

[0049] 4. In this invention, the fault location is calculated based on the time difference of arrival of multi-terminal traveling waves, line topology, and corrected wave velocity. The result is then output through multi-source information verification and optimization algorithms, which greatly improves the positioning accuracy, reduces positioning errors, provides a reliable basis for rapid repair, effectively shortens power outage time, improves the reliability of power supply in the distribution network, ensures the power quality for users, and reduces economic losses caused by faults. Attached Figure Description

[0050] Figure 1 This is a flowchart of an embodiment of the power distribution network fault location method of the present invention;

[0051] Figure 2 This is a schematic diagram of a structure of an embodiment of the power distribution network fault location device of the present invention;

[0052] Figure 3This is a schematic diagram of the structure of one embodiment of the computer device of the present invention;

[0053] Figure 4 This is a schematic diagram of the identification module structure in one embodiment of the computer device of the present invention.

[0054] In the diagram: 1. Preprocessing module; 2. Feature extraction module; 3. Recognition module; 4. Diagnosis module; 5. Output module; 6. Memory; 7. Processor; 8. Network interface; 31. Preliminary judgment module; 32. Calibration module; 33. Compensation module. Detailed Implementation

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the specification is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings are used to distinguish different objects and not to describe a particular order.

[0056] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0057] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0058] It should be noted that the power distribution network fault location method provided in the embodiments of the present invention is generally executed by a server / terminal device, and correspondingly, the power distribution network fault location device is generally installed in the server / terminal device.

[0059] It should be understood that the number of terminal devices, networks, and servers is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be used.

[0060] Example 1

[0061] Please refer to Figure 1 and Figure 4The flowchart illustrates an embodiment of the power distribution network fault location method of the present invention. The power distribution network fault location method includes the following steps:

[0062] Step S1: Collect distribution network data and perform standardized preprocessing on the distribution network data;

[0063] In this embodiment, the electronic device (e.g., server / terminal device) on which the power distribution network fault location method runs can receive the power distribution network fault location request through a wired connection or a wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, WiMAXX connection, Zigbee connection, UWB (ultra wideband) connection, and other currently known or future wireless connection methods.

[0064] In this embodiment, step S1 specifically includes the following steps:

[0065] S11. Collect data from the power distribution network, which includes broadband traveling wave and power frequency signals:

[0066] Deploy new broadband current transformers (CTs): These CTs must have high bandwidth (e.g., 0.1Hz-10MHz) to accurately transmit both power frequency current and high-frequency (kHz-MHz level) traveling wave current signals simultaneously, have high dynamic range to avoid saturation during high fault currents, and adopt miniaturized designs such as Rogowski coils or low-power iron cores for easy installation at the distribution network site.

[0067] Record high sampling rate synchronous data: Configure synchronous sampling units at each monitoring point (such as traveling wave series terminals of distribution network lines), and use GPS / BeiDou pulse per second (PPS) or IEEE 1588 precision clock protocol (PTP) to achieve sub-microsecond time synchronization between multiple terminals. The sampling rate must meet the Nyquist sampling theorem. For traveling wave signals, the sampling rate is usually required to be 1-10MHz or higher.

[0068] Standardize and encapsulate the data: In specific implementation, a unified data frame format can be designed according to the data interface standardization requirements adapted to the distribution network autonomous station. The frame structure includes timestamp (accurate to μs), channel identifier (CT phase, position), sampling rate, data length, traveling wave raw waveform array, power frequency effective value array, and equipment status code, etc. High-speed uploading is carried out using protocols such as IEC61850 GOOSE or customized binary protocols.

[0069] The above step S11 can capture complete electrical quantities containing fault transient high-frequency traveling wave and steady-state power frequency information, providing a data basis for the traveling wave method and the traditional impedance method.

[0070] S12. Perform time alignment and quality verification on the distribution network data;

[0071] First, data alignment is performed based on precise timestamps: using the high-precision timestamps reported by each terminal as a benchmark, a unified timeline is established at the data access layer (main station or edge computing unit). For minor packet arrival delays caused by network jitter, timestamps are used for reordering, rather than the order in which the packets arrive.

[0072] Next, perform data integrity verification: check whether each data packet is complete (such as checksum and CRC check), whether there are missing points or duplicates, mark incomplete data packets, and perform linear interpolation repair based on previous and subsequent points (applicable to non-drastic change segments).

[0073] Then, abnormal data detection and labeling are performed: thresholds (such as current surges, waveform distortion rates) and rules (such as whether it is within the equipment's range) are set to automatically detect obvious abnormal data. For example, if the current value at a certain point exceeds the rated peak value of the CT, the data in that time period is marked as "suspicious" for subsequent algorithms to choose to ignore or handle specially.

[0074] Step S12 above can eliminate data time stamp deviations caused by communication delays and differences in device response, and identify and label abnormal data to ensure spatiotemporal consistency of subsequent fusion and computation.

[0075] S13. Perform preliminary signal purification processing based on wavelet threshold denoising;

[0076] First, select the wavelet basis function and the number of decomposition levels: choose a wavelet basis with waveform characteristics similar to the traveling wave signal, such as the Daubechies (db) series or the Symlets (sym) series. The number of decomposition levels J is determined based on the signal sampling frequency fs and the highest effective frequency of interest fmax, usually satisfying fs / 2. J+1 > fmax.

[0077] Then, wavelet thresholding denoising is performed: Discrete wavelet transform (DWT) is applied to the acquired discrete traveling wave signal sequence s(n) to obtain the detail coefficients of each layer. and approximation coefficients ;

[0078] Next, select a threshold: use an adaptive threshold, such as Stein's unbiased risk estimation (SURE) threshold or a fixed threshold. ;

[0079] A soft thresholding function is used to obtain a smoother reconstructed signal:

[0080] ;

[0081] in: This represents the k-th detail coefficient of the j-th layer after thresholding. This represents the original k-th detail coefficient of the j-th level obtained by discrete wavelet transform. The sign function takes the sign of the input value (1 for positive, -1 for negative, and 0 for zero). This indicates taking the absolute value. The noise reduction threshold is represented by the following formula: , The noise standard deviation is represented by the estimated value of the absolute median of the first level detail coefficients (MAD / 0.6745), and N represents the signal length (number of sampling points).

[0082] Detail factor after thresholding and the final approximation coefficient Perform inverse wavelet transform (IDWT) to obtain the denoised signal. This step can effectively filter out white noise and some periodic interference.

[0083] Step S13 above can suppress noise interference in step S11, initially improve the signal-to-noise ratio (SNR) of the signal, and prepare for fine operations such as wavefront detection.

[0084] Step S2: Based on the standardized preprocessed distribution network data, perform traveling wave signal feature extraction and enhancement processing;

[0085] In this embodiment, step S2 specifically includes the following steps:

[0086] S21. Based on time-frequency analysis, extract the image features of the fault traveling wave;

[0087] For the denoised signal Performing continuous wavelet transform (CWT) includes the following:

[0088] ;

[0089] Where: CWT(a,b) represents the wavelet coefficients at scale a and translation b, a represents the scale factor, which is inversely proportional to the frequency (the smaller a is, the higher the frequency), and b represents the translation factor, corresponding to time. This represents the input time-domain signal (in this case, the denoised signal). Represents the mother wavelet function. Represents the complex conjugate of the mother wavelet function. This represents the normalization factor, ensuring that the wavelet energy remains constant across different scales;

[0090] Calculate the modulus of CWT coefficients It is mapped to the grayscale range of [0, 255], with the horizontal axis representing time b and the vertical axis representing scale a (or converted to frequency f). The energy distribution patterns of traveling waves generated by different faults (such as single-phase grounding and phase-to-phase short circuit) in the time-frequency plane are different, forming distinguishable "texture" features.

[0091] All generated time-frequency graphs are standardized to a uniform size (e.g., 256x256 pixels) and normalized to eliminate the influence of amplitude dimensions, making them easier to input into subsequent deep learning models.

[0092] Step S21 can convert a one-dimensional time-domain traveling wave signal into a two-dimensional time-frequency image (such as a grayscale image or a pseudo-color image), which intuitively shows the distribution of signal energy in the time and frequency dimensions, making it easier for visual algorithms such as CNN to identify the image features of various types of fault traveling waves.

[0093] S22. Perform homomorphic filtering enhancement on the traveling wave front;

[0094] For time domain signals Take the absolute value (emphasizing the magnitude) and then take the natural logarithm to obtain... (add To prevent zero values, a Discrete Fourier Transform (DFT) is then performed to obtain the spectrum. ;

[0095] Design a high-pass or band-pass filter in the frequency domain, for example, an ideal high-pass filter:

[0096] ;

[0097] in, This represents the filter gain at frequency index k, where k represents the discrete frequency index. This indicates the cutoff frequency index, which is usually set to the index of the main frequency band of the traveling wave (e.g., above 10kHz). Indicates high-frequency gain, typically >1 (e.g., 1.5-2.5), used to enhance the high-frequency components of traveling waves. This indicates low-frequency gain, typically <1 (e.g., 0.5), used to suppress power frequency and low-frequency noise;

[0098] Filtering in the frequency domain Then to Perform inverse DFT to obtain Finally, perform exponentiation. The enhanced time-domain signal is obtained This process can effectively improve the contrast of the traveling wave head relative to the background.

[0099] Through the above step S22, specific frequency components related to the fault traveling wave front can be selectively enhanced in the time-frequency domain, while suppressing background noise and power frequency components, improving the salience of the wave front, and dealing with waveform distortion and noise interference.

[0100] S23. Based on empirical mode decomposition, perform mode separation and selection;

[0101] For the enhanced signal Empirical Mode Decomposition (EMD) is performed to obtain a series of IMF components from high frequency to low frequency. and a residual term ,Right now ,in: This represents the input signal (the enhanced signal). This represents the i-th intrinsic mode function (IMF), arranged in descending order of frequency. The term represents the residual, which indicates the average trend or DC component of the signal, and n represents the total number of IMFs obtained from the decomposition.

[0102] Calculate the instantaneous frequency and energy of each IMF component. Traveling wavefronts typically exhibit high-frequency, short-duration components with abrupt energy changes; therefore, select the first few (e.g., , The IMF with the highest instantaneous frequency and concentrated energy at the suspected fault moment is taken as the dominant mode of the traveling wave;

[0103] The selected key IMF components are summed to obtain a reconstructed signal focused on the traveling wave component. S is the index set of key IMFs. This signal is purer and facilitates accurate wavefront detection later.

[0104] Through the above step S23, the signal can be adaptively decomposed into intrinsic mode functions (IMFs) of different frequency scales, separating the IMFs mainly contributed by fault traveling waves, removing noise and low-frequency oscillation components, and dealing with problems with complex frequency components.

[0105] Through the above step S2, traveling wave image features that can characterize the fault type and location can be extracted and enhanced from the initially purified signal, serving subsequent AI recognition and multimodal fusion.

[0106] Step S3: Identify the moment when the initial traveling wave of the fault arrives at the monitoring point from the enhanced and separated traveling wave signal;

[0107] In this embodiment, step S3 specifically includes the following steps:

[0108] S31. Based on the wavelet transform modulus maxima, perform preliminary judgment on the arrival time of the wavefront;

[0109] Selecting wavelets with tight support and symmetry, such as db4 or sym4, whose waveforms are similar to the rising edge of the traveling wave front, can produce significant transformation coefficients.

[0110] For reconstructed signals Discrete wavelet transform is performed at a specific scale (usually an intermediate scale that balances frequency and time resolution, such as the 3rd or 4th layer), and its magnitude is calculated. Where s is the scale, It is a translation;

[0111] On scale s, search about The local maxima points, which correspond to singular points or abrupt changes in the signal, are defined by the time interval of the first modulus maxima exceeding a set threshold. It was initially determined to be the arrival time of the wavefront.

[0112] By using the above step S31, the modulus maxima characteristic of wavelet transform at signal singularities can be utilized to preliminarily determine the possible arrival time of the wavefront.

[0113] S32, combined with the CNN target recognition algorithm, performs wave head calibration;

[0114] Based on the initial judgment time Centered on the fault, a small segment (e.g., from one power frequency cycle before the fault to two power frequency cycles after the fault) of the original or enhanced time-domain waveform is extracted and converted into a fixed-size (e.g., 64x64) grayscale image. Based on historical fault waveform data, the accurate starting pixel of the wavefront (corresponding to the time) is manually marked. ), forming a training set;

[0115] Construct a lightweight CNN, such as one containing 2-3 convolutional layers (for extracting local features like waveform edges and peaks), pooling layers, and fully connected layers. The output layer can be designed as a regression problem (directly predicting wavefront time). For example, in classification problems (discretizing the time axis into multiple segments and predicting the segment where the wavefront is located), labeled data can be used for training, and the loss function can be the mean squared error (MSE) or cross-entropy.

[0116] The waveform segment to be analyzed is input into the trained CNN model to obtain its predicted precise wavefront timing. This technology can learn the essential visual features of wavefronts from complex waveform distortions, making it more robust than traditional thresholding methods.

[0117] Through the above step S32, the powerful image pattern recognition capability of the convolutional neural network (CNN) can be utilized to perform fine recognition and classification of the local waveform of the wave head, accurately identify the true fault traveling wave starting point from the candidate points, and overcome the interference of non-fault signals and the influence of waveform distortion.

[0118] S33, based on line parameter correction, performs wavefront time compensation;

[0119] The distribution network line is considered as a unit-length model with distributed resistance R, inductance L, capacitance C, and conductance G. Traveling wave propagation constant. Where ω is the angular frequency and phase velocity is... It is a function of frequency (dispersion effect), where: The propagation constant is represented by R, L, C, and G, which represent the resistance, inductance, capacitance, and conductance per unit length of the line, respectively. j represents the imaginary unit, and ω represents the angular frequency. , This indicates taking the imaginary part of a complex number;

[0120] By utilizing operating traveling wave signals such as reclosing or historical line switching, a known operating signal (such as switch closing) is injected when the system is fault-free, and the arrival time difference of the traveling waves at both ends is recorded. Given the line length Then the equivalent wave velocity can be inferred or corrected based on the model. Or line unit length parameter;

[0121] The correction formula can be simplified to ,in, This represents the corrected equivalent traveling wave propagation velocity. Indicates the total length of the line. This represents the time difference of arrival of the operating signal measured at both ends. It comprehensively reflects the actual propagation characteristics of the line and is more accurate than the theoretical wave speed;

[0122] For a fault traveling wave, the wave velocity corresponding to its dominant frequency component may differ from the corrected wave velocity. Therefore, a frequency-wave velocity lookup table or empirical formula should be established based on the fault traveling wave dominant frequency extracted in step S21. Find the corresponding more precise wave velocity The propagation time from the monitoring point to the fault point is... The compensation formula is: ,in It is an estimated length of the section from the monitoring point to the fault point. This compensation can reduce the system error introduced by inaccurate wave velocity.

[0123] Step S33 above can take into account the dispersion, attenuation, and wave velocity changes caused by inaccurate line parameters during the propagation of the traveling wave, and perform micro-compensation on the detected wavefront time to improve the absolute time calibration accuracy.

[0124] Step S4: Integrate traveling wave signals, power frequency electrical quantities, switch status and topology multimodal information, and use AI algorithms for intelligent fault diagnosis;

[0125] In this embodiment, step S4 may specifically include the following steps:

[0126] S41. Perform early fusion of multimodal features and joint representation learning;

[0127] The different modal features extracted from the same time window are time-aligned (already completed in step S12). For example, the time-frequency image features (matrix), the power frequency current / voltage effective value sequence (vector), and the switch change state (one-hot encoded vector) are flattened and then spliced ​​in the feature dimension to form an ultra-long joint feature vector.

[0128] The concatenated high-dimensional features are input into a deep autoencoder. The encoder learns a low-dimensional joint representation vector z that can condense the key information of all modes through multi-layer nonlinear transformations. This z captures the deep correlations across modes, such as the correspondence between "abrupt change in the traveling wave of a certain phase", "increase in the power frequency current of that phase", and "start of the corresponding switch protection".

[0129] The low-dimensional joint representation vector z is used as the output of the early fusion and is fed into the subsequent diagnostic classification network. This method requires high quality and accurate alignment of data from each modality, is sensitive to outliers, and has a deep degree of fusion.

[0130] Through the above step S41, data or low-level feature fusion can be performed, allowing the model to learn the intrinsic relationship between different modalities from the original stage and make full use of their complementarity.

[0131] S42, based on the attention mechanism, performs mid-term feature fusion and optimization;

[0132] Design an independent feature extraction sub-network for each modality (e.g., CNN for image processing, RNN for sequence processing, MLP for state variables) to obtain high-level feature vectors respectively. (i represents the mode);

[0133] The weights for each modality feature are calculated using an attention network (usually a small fully connected network). ;

[0134] ; ;in, Let v, W, b represent the attention score for the i-th modality, and let v, W, b represent the learnable weight vector, weight matrix, and bias vector, respectively. Hyperbolic tangent activation function, This represents the high-level feature vector of the i-th mode. Let M represent the normalized attention weight of the i-th modality, and M represent the total number of modalities. Represents an exponential function;

[0135] This reflects the importance of the i-th modality in the current context;

[0136] Calculate the weighted sum to obtain the fusion features. When a high-resistance ground fault occurs, the traveling wave characteristics may be weak, and the model may automatically assign higher weights to the power frequency zero-sequence current characteristics. However, during a metallic short circuit, the weight of the traveling wave characteristic increases.

[0137] The purpose of step S42 is to dynamically weight and fuse the high-level features extracted from different modalities in the intermediate layer of the deep neural network, so that the model can autonomously focus on the modal features most relevant to the current fault diagnosis, thus solving the problems of difficulty in determining a suitable fusion strategy and different contributions of data from different modalities to the final decision.

[0138] S43, based on DS evidence theory, performs late-stage decision fusion;

[0139] Multiple independent fault diagnosis models are trained. For example, Model A classifies fault types based on traveling wave time-frequency diagrams and outputs the confidence (probability) distribution for each type of fault; Model B diagnoses based on power frequency variation characteristics; and Model C performs fault interval inference based on topology and switch information. Each model outputs a Basic Probability Assignment (BPA) function. This indicates that model i considers the fault to belong to type [type]. Reliability;

[0140] For the two models and The synthesized BPA is:

[0141] , ,in, This represents the basic probability assignment to proposition A after fusion; Let A represent the basic probability assignments of Model 1 to proposition B and Model 2 to proposition C; B and C are propositions (such as subsets of the fault type set) under the two model identification frameworks, respectively; A represents the proposition considered after synthesis, and... K represents the normalization factor (conflict coefficient). , used to handle conflicting evidence; Represents the empty set;

[0142] in, This is a normalization factor used to handle conflicting evidence. B and C are propositions (such as subsets of the fault type set) under the two model identification frameworks, respectively. This rule can effectively fuse even conflicting evidence;

[0143] The BPAs of all models are synthesized pairwise to obtain the comprehensive BPA. The proposition with the highest reliability is selected as the final diagnostic result (such as fault type, suspected fault segment). This method does not require low-level alignment of modal data, has strong fault tolerance, and belongs to late-stage fusion.

[0144] Step S43 above can integrate the results of multiple independent diagnostic models (each based on a different modality or different algorithm) at the decision level, improving the reliability and robustness of the final diagnostic results, especially when some modality data is missing or of poor quality.

[0145] Through the above step S4, multimodal information such as traveling wave signals, power frequency electrical quantities, switch status, and topology can be integrated, and comprehensive diagnosis can be performed using AI algorithms to achieve fault type identification, preliminary screening of fault ranges, and reliability assessment, providing more reliable input and constraints for precise location.

[0146] Step S5: Using the calibrated multi-terminal traveling wave arrival time difference, combined with the line topology and the corrected wave velocity, calculate the specific location of the fault point, and output the location result through multi-source information verification and optimization algorithm;

[0147] In this embodiment, step S5 may specifically include the following steps:

[0148] S51. Based on the time difference of the traveling wave at both ends, perform preliminary location calculations for distribution network faults;

[0149] Obtain the absolute arrival time of the traveling wavefront at both ends of the faulty line (designated as M and N ends) after the precise calibration in step three. and And the equivalent wave velocity v and the total length L of the line after correction by step 3.3;

[0150] The double-ended positioning formula is as follows Where: x represents the distance from the fault point to end M, L represents the total length of the line, and v represents the equivalent propagation speed of the traveling wave on the line (already corrected). This represents the absolute time it takes for the initial traveling wave of the fault to reach the M terminal. This indicates the absolute time it takes for the initial traveling wave of the fault to reach the N terminal;

[0151] molecular Equivalent to That is, the total path length of the wave propagating from the fault point to the M and N ends (2x or the corrected form of 2L−x), which is obtained by dividing by 2;

[0152] Calculate x and ensure that 0 ≤ x ≤ L. If x exceeds this range, it indicates that there may be a large error in the time calibration or that the wave velocity selection is inappropriate, and an alarm should be triggered.

[0153] By using the time difference of arrival of the initial traveling wave of the fault detected at both ends (or multiple ends) of the line through the above step S51, the approximate distance from the fault point to one end can be calculated.

[0154] S52. Perform multi-terminal integrated positioning and result optimization;

[0155] For a line with k terminals, it can be formed Each pair of dual-end positioning pairs calculates a fault location. (i, j are terminal numbers). At the same time, more hypotheses can be formed by using the single-end positioning principle (which requires knowing the arrival time of the reflected wave at the other end).

[0156] All calculated This forms a set, which is then analyzed using density clustering algorithms (such as DBSCAN) or statistical methods. DBSCAN can identify dense regions (most consistent results) and discrete points (outliers), grouping those points that fall into the highest density cluster (core point set). Considered a reliable result;

[0157] For the selected reliable result set The median (which is more resistant to outliers than the mean) was used as the initial localization result. Meanwhile, the standard deviation of this set is calculated as a measure of the uncertainty of the positioning results.

[0158] Through the above step S52, when there are multiple (>2) monitoring terminals on the line, multiple sets of time difference information can be used for comprehensive positioning, and the most reliable result can be selected through consistency verification, effectively suppressing the impact of multipath effect and individual terminal data anomalies.

[0159] S53. Result verification and spatial constraint optimization combining topology and electrical quantities;

[0160] Preliminary positioning results Map it onto the actual line geographic information system (GIS) or topology model to determine the specific line section where it is located (between which two switches or towers).

[0161] During the fault inspection, check whether the direction of the fault current flowing through the switches (if measured) at both ends of the section where the initial location point is located matches the location of the fault point predicted by the location results (the fault point should be near the node where the current direction changes). At the same time, verify whether there is indeed a protection action or fault indicator triggered in this section.

[0162] The location result is optimized to its local section and can be described as "XX line, located between tower A and tower B, approximately Y meters away from tower A". Combining the fault type diagnosis result from step S4 and the uncertainty measurement in this step, a complete fault diagnosis report containing the precise location, fault type, confidence level, and suggested investigation scope is generated. This step ensures the operability of the location result in engineering.

[0163] Through the above step S53, information such as the distribution network topology, sectionalizing switch status, and electrical quantities after a fault (such as current distribution) can be used to verify the physical rationality and spatial constraints of the preliminary location results, correct the location results to the vicinity of the nearest topological node (such as a tower or sectionalizing switch), and exclude obviously impossible intervals.

[0164] Through the above step S5, the precise arrival time difference of the multi-terminal traveling wave can be used, combined with the line topology and the corrected wave velocity, to calculate the specific location of the fault point, and through multi-source information verification and optimization algorithms, the final high-precision positioning result can be output.

[0165] In some optional implementations of this embodiment, this embodiment can also be software-based, modularized, and productized to form a set of distribution network fault traveling wave location function modules that can be embedded in existing distribution network automation systems or run independently, and its effectiveness and reliability can be verified through engineering pilot projects.

[0166] First, we standardize the data interface and design the module integration, defining a stable and efficient data interaction interface between the module and external systems (such as distribution network master station, traveling wave terminal, SCADA, GIS) to ensure that the module can be seamlessly integrated into the existing technology system.

[0167] The input interface design includes: real-time stream interface, service call interface, and static data interface, etc.

[0168] The real-time streaming interface can use a message queue (such as Kafka, MQTT) to subscribe to the fault recording data stream of the traveling wave terminal (conforming to the standardized format of step S11).

[0169] The service call interface can provide a RESTful API or WebService, which allows the main station system to proactively push fault-related information (fault time, trip switch, related terminal ID, etc.) to trigger localization analysis after detecting a fault event.

[0170] The static data interface can synchronize line topology, tower coordinates, and length parameters from the GIS system; and synchronize equipment parameters (CT ratio, line type) from the production management system.

[0171] The output interface design includes a result push interface, which refers to pushing the complete fault diagnosis report generated in step five to the distribution network dispatch system and mobile inspection APP in real time through message queues or API callbacks.

[0172] The raw data, intermediate results, and final report of each location are stored in a structured manner in a relational database (such as MySQL) or a time-series database (such as InfluxDB) for historical querying, statistical analysis, and algorithm iteration.

[0173] It can also be designed as a microservice within the module, encapsulating the algorithms of steps S1 to S5 into independent microservices (such as "data preprocessing service", "wave head calibration service", "AI diagnostic service", "positioning calculation service"), and communicating through an internal service bus (such as gRPC) to improve the maintainability, scalability and reliability of the system.

[0174] In practice, a workflow engine can be developed to schedule the execution of each microservice according to the logical order of steps S1-S5, and an exception handling mechanism can be set up (such as degradation or retry strategies when a service times out or data is missing).

[0175] Real-time monitoring is possible through a human-machine interface (HMI) developed via a web front-end or desktop client: displaying the status of each terminal and network communication status; fault display: highlighting the fault location and fault range on the map background; report viewing: providing detailed waveform graphs, time-frequency graphs, AI diagnostic process, location calculation details, and final reports for each fault; parameter configuration: allowing maintenance personnel to configure algorithm parameters (such as thresholds, model versions), line parameters, communication parameters, etc.; statistical analysis: providing historical statistical reports on faults, such as location accuracy, algorithm time consumption, and distribution of different fault types.

[0176] The positioning function module was deployed and tested in a real power distribution network environment. Operational data was collected to evaluate whether it met the economic indicators of "fault range positioning accuracy ≥ 99% and fault positioning accuracy ≤ 200m". Based on the feedback, the algorithm and system were optimized and iterated.

[0177] Select several or more representative distribution network lines (with different network structures and load characteristics), install new broadband CTs and smart terminals, and deploy positioning function modules (which can be deployed on the city-level company's cloud platform or edge computing station).

[0178] During the pilot phase, detailed data on every line fault (including actual faults and experimental artificial grounding faults) were recorded. Experts were organized to review and confirm the true location and type of each fault afterward, forming a gold standard test set.

[0179] Accuracy calculation: Accuracy = (Number of times the module correctly identifies the fault range) / (Total number of faults) * 100%.

[0180] Positioning accuracy calculation: For precise positioning results, calculate the absolute distance error between them and the golden standard position. For example, the error of all faults should be ≤200 meters, and the average error should be calculated.

[0181] Iterative optimization: In-depth analysis of cases that fail to meet the targets (whether it is a wavefront calibration error, fusion error, or inaccurate parameters, etc.), targeted adjustments to the algorithm model, parameters, or data preprocessing process, and continuous iteration to ensure that the module performance stably reaches and exceeds the predetermined goals, ultimately forming a mature, reliable, and scalable distribution network traveling wave fault location module.

[0182] The beneficial effects of implementing this embodiment are:

[0183] (1) By standardizing and preprocessing the distribution network data, the dimensional differences between different data sources can be effectively eliminated, the data quality can be improved, and a solid foundation can be laid for subsequent analysis.

[0184] (2) By extracting and enhancing the features of traveling wave signals, fault features can be accurately captured and the signal identification can be enhanced. Even in complex noise environments, fault signals can be accurately identified.

[0185] (3) By integrating multimodal information and using AI algorithms for diagnosis, we can make full use of information such as traveling waves, power frequency electrical quantities, switch status and topology, give full play to the advantages of each mode, achieve more comprehensive and accurate fault diagnosis, and reduce the misjudgment rate.

[0186] (4) By calculating the fault location based on the time difference of arrival of multi-terminal traveling waves, line topology and corrected wave velocity, and outputting the results through multi-source information verification and optimization algorithms, the positioning accuracy is greatly improved, the positioning error is reduced, a reliable basis for rapid repair is provided, the power outage time is effectively shortened, the power supply reliability of the distribution network is improved, the power quality of users is guaranteed, and the economic losses caused by the fault are reduced.

[0187] This invention can be used in numerous general-purpose or special-purpose computer system environments or configurations, such as: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. This invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0188] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by instructing related hardware through computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM).

[0189] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0190] Example 2

[0191] Further reference Figure 2 and Figure 4 As a response to the above Figure 1 The present invention provides an embodiment of a power distribution network fault location device, which is implemented in accordance with the method shown. Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0192] like Figure 2 As shown, the power distribution network fault location device in this embodiment includes: a preprocessing module 1, a feature extraction module 2, an identification module 3, a diagnosis module 4, and an output module 5, wherein:

[0193] Preprocessing module 1 is used to collect distribution network data and perform standardized preprocessing on the distribution network data;

[0194] Feature extraction module 2 is used to extract and enhance the features of traveling wave signals based on the standardized preprocessed distribution network data;

[0195] Identification module 3 is used to identify the moment when the initial traveling wave of the fault arrives at the monitoring point from the enhanced and separated traveling wave signal;

[0196] Diagnostic module 4 is used to integrate traveling wave signals, power frequency electrical quantities, switch status and topology multimodal information, and use AI algorithms for intelligent fault diagnosis.

[0197] Output module 5 is used to calculate the specific location of the fault point by using the calibrated multi-terminal traveling wave arrival time difference, combined with the line topology and the corrected wave velocity, and output the location result through multi-source information verification and optimization algorithm.

[0198] In some optional implementations of this embodiment, the identification module 3 includes:

[0199] The preliminary judgment module 31 is used to make a preliminary judgment on the arrival time of the wavefront based on the modulus maxima of wavelet transform;

[0200] The calibration module 32 is used to perform wavehead calibration in conjunction with the CNN target recognition algorithm;

[0201] The compensation module 33 is used to perform wavefront time compensation based on line parameter correction.

[0202] The beneficial effects of implementing this embodiment are:

[0203] (1) By standardizing and preprocessing the distribution network data, the dimensional differences between different data sources can be effectively eliminated, the data quality can be improved, and a solid foundation can be laid for subsequent analysis.

[0204] (2) By extracting and enhancing the features of traveling wave signals, fault features can be accurately captured and the signal identification can be enhanced. Even in complex noise environments, fault signals can be accurately identified.

[0205] (3) By integrating multimodal information and using AI algorithms for diagnosis, we can make full use of information such as traveling waves, power frequency electrical quantities, switch status and topology, give full play to the advantages of each mode, achieve more comprehensive and accurate fault diagnosis, and reduce the misjudgment rate.

[0206] (4) By calculating the fault location based on the time difference of arrival of multi-terminal traveling waves, line topology and corrected wave velocity, and outputting the results through multi-source information verification and optimization algorithms, the positioning accuracy is greatly improved, the positioning error is reduced, a reliable basis for rapid repair is provided, the power outage time is effectively shortened, the power supply reliability of the distribution network is improved, the power quality of users is guaranteed, and the economic losses caused by the fault are reduced.

[0207] Example 3

[0208] To address the aforementioned technical problems, embodiments of the present invention also provide a computer device, for details please refer to... Figure 3 , Figure 3 This is a basic structural block diagram of the computer device in this embodiment.

[0209] The aforementioned computer device includes a memory 6, a processor 7, and a network interface 8 that are interconnected via a system bus. It should be noted that only a computer device with components 6, 7, and 8 is shown in the figure. However, it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions. Its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0210] The aforementioned computer equipment can be desktop computers, laptops, handheld computers, cloud servers, and other computing devices. These computer devices can interact with users through keyboards, mice, remote controls, touchpads, or voice control devices.

[0211] The aforementioned memory 6 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the aforementioned memory 6 may be an internal storage unit of the aforementioned computer device, such as the hard disk or memory of the computer device. In other embodiments, the aforementioned memory 6 may also be an external storage device of the aforementioned computer device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. of the computer device. Of course, the aforementioned memory 6 may also include both the internal storage unit and the external storage device of the aforementioned computer device. In this embodiment, the aforementioned memory 6 is typically used to store the operating system and various application software installed on the aforementioned computer device, such as computer-readable instructions for power distribution network fault location methods. In addition, the aforementioned memory 6 may also be used to temporarily store various types of data that have been output or will be output.

[0212] In some embodiments, the processor 7 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 7 is typically used to control the overall operation of the computer device. In this embodiment, the processor 7 is used to run computer-readable instructions stored in the memory 6 or to process data, such as running computer-readable instructions for the power distribution network fault location method.

[0213] The aforementioned network interface 8 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the aforementioned computer equipment and other electronic devices.

[0214] The beneficial effects of implementing this embodiment are:

[0215] (1) By standardizing and preprocessing the distribution network data, the dimensional differences between different data sources can be effectively eliminated, the data quality can be improved, and a solid foundation can be laid for subsequent analysis.

[0216] (2) By extracting and enhancing the features of traveling wave signals, fault features can be accurately captured and the signal identification can be enhanced. Even in complex noise environments, fault signals can be accurately identified.

[0217] (3) By integrating multimodal information and using AI algorithms for diagnosis, we can make full use of information such as traveling waves, power frequency electrical quantities, switch status and topology, give full play to the advantages of each mode, achieve more comprehensive and accurate fault diagnosis, and reduce the misjudgment rate.

[0218] (4) By calculating the fault location based on the time difference of arrival of multi-terminal traveling waves, line topology and corrected wave velocity, and outputting the results through multi-source information verification and optimization algorithms, the positioning accuracy is greatly improved, the positioning error is reduced, a reliable basis for rapid repair is provided, the power outage time is effectively shortened, the power supply reliability of the distribution network is improved, the power quality of users is guaranteed, and the economic losses caused by the fault are reduced.

[0219] Example 4

[0220] The present invention also provides another embodiment, namely, a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the power distribution network fault location method described above.

[0221] The beneficial effects of implementing this embodiment are as follows:

[0222] (1) By standardizing and preprocessing the distribution network data, the dimensional differences between different data sources can be effectively eliminated, the data quality can be improved, and a solid foundation can be laid for subsequent analysis.

[0223] (2) By extracting and enhancing the features of traveling wave signals, fault features can be accurately captured and the signal identification can be enhanced. Even in complex noise environments, fault signals can be accurately identified.

[0224] (3) By integrating multimodal information and using AI algorithms for diagnosis, we can make full use of information such as traveling waves, power frequency electrical quantities, switch status and topology, give full play to the advantages of each mode, achieve more comprehensive and accurate fault diagnosis, and reduce the misjudgment rate.

[0225] (4) By calculating the fault location based on the time difference of arrival of multi-terminal traveling waves, line topology and corrected wave velocity, and outputting the results through multi-source information verification and optimization algorithms, the positioning accuracy is greatly improved, the positioning error is reduced, a reliable basis for rapid repair is provided, the power outage time is effectively shortened, the power supply reliability of the distribution network is improved, the power quality of users is guaranteed, and the economic losses caused by the fault are reduced.

[0226] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0227] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A method for locating faults in a power distribution network, characterized in that, Includes the following steps: S1. Collect power distribution network data and perform standardized preprocessing on the power distribution network data; S2. Based on the standardized preprocessed distribution network data, perform traveling wave signal feature extraction and enhancement processing; S3. Identify the moment when the initial traveling wave of the fault arrives at the monitoring point from the enhanced and separated traveling wave signal; S4. Integrate traveling wave signals, power frequency electrical quantities, switch status and topology multimodal information, and use AI algorithms for intelligent fault diagnosis; S5. Using the calibrated multi-terminal traveling wave arrival time difference, combined with the line topology and the corrected wave velocity, calculate the specific location of the fault point, and output the location result through multi-source information verification and optimization algorithm.

2. The method for locating faults in a power distribution network according to claim 1, characterized in that, Step S1 specifically includes: S11. Collect the power distribution network data, which includes broadband traveling wave and power frequency signal; S12. Perform time alignment and quality verification on the power distribution network data; S13. Perform preliminary signal purification processing based on wavelet threshold denoising.

3. The method for locating faults in a power distribution network according to claim 1, characterized in that, Step S2 specifically includes: S21. Based on time-frequency analysis, extract the image features of the fault traveling wave; S22. Perform homomorphic filtering enhancement on the traveling wave front; S23. Based on empirical mode decomposition, perform mode separation and selection.

4. The method for locating faults in a power distribution network according to claim 1, characterized in that, Step S3 specifically includes: S31. Based on the wavelet transform modulus maxima, perform preliminary judgment on the arrival time of the wavefront; S32. Combine the target recognition algorithm of CNN to perform wave head calibration; S33. Based on line parameter correction, perform wavefront time compensation.

5. The method for locating faults in a power distribution network according to claim 1, characterized in that, Step S4 specifically includes: S41. Perform early fusion of multimodal features and joint representation learning; S42. Based on the attention mechanism, perform mid-term feature fusion and optimization; S43. Based on the DS evidence theory, late-stage decision fusion is carried out.

6. The method for locating faults in a power distribution network according to claim 1, characterized in that, Step S5 specifically includes: S51. Based on the time difference of the traveling wave at both ends, perform preliminary location calculations for distribution network faults; S52. Perform multi-terminal integrated positioning and result optimization; S53. Verification of results by combining topology and electrical quantities with spatial constraint optimization.

7. A power distribution network fault location device, characterized in that, include: The preprocessing module (1) is used to collect power distribution network data and perform standardized preprocessing on the power distribution network data; The feature extraction module (2) is used to extract and enhance the features of the traveling wave signal based on the standardized preprocessed distribution network data. The identification module (3) is used to identify the moment when the initial traveling wave of the fault arrives at the monitoring point from the enhanced and separated traveling wave signal; The diagnostic module (4) is used to integrate traveling wave signals, power frequency electrical quantities, switch status and topology multimodal information, and use AI algorithms for intelligent fault diagnosis. The output module (5) is used to calculate the specific location of the fault point by using the calibrated multi-terminal traveling wave arrival time difference, combined with the line topology and the corrected wave velocity, and output the location result through multi-source information verification and optimization algorithm.

8. The power distribution network fault location device according to claim 7, characterized in that, The identification module (3) includes: The initial judgment module (31) is used to make an initial judgment on the arrival time of the wavefront based on the maximum value of the wavelet transform modulus; The calibration module (32) is used to perform wavefront calibration in conjunction with the CNN target recognition algorithm; The compensation module (33) is used to perform wavefront time compensation based on line parameter correction.

9. A computer device, characterized in that, It includes a memory (6), a processor (7) and a network interface (8), wherein the memory (6) stores computer-readable instructions, and the processor (7) executes the computer-readable instructions to implement the steps of the power distribution network fault location method as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the power distribution network fault location method as described in any one of claims 1 to 6.

Images