A power distribution network fault locating method, system, device and storage medium
By digitally processing the three-phase voltage signals and recursively decomposing wavelet packets, an energy feature matrix is constructed and statistical sensitivity analysis is performed. This solves the accuracy and robustness problems of existing distribution network fault location methods under complex operating conditions, and achieves efficient and accurate fault location.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHUNAN COUNTY POWER SUPPLY CO OF STATE GRID ZHEJIANG ELECTRIC POWER CO LTD
- Filing Date
- 2026-04-28
- Publication Date
- 2026-05-29
AI Technical Summary
Existing fault location methods for power distribution networks have low accuracy and poor robustness under complex operating conditions, making it difficult to achieve fast and accurate location of faulty lines, sections, and phases.
By digitally sampling the three-phase voltage signals, a superimposed voltage signal is generated and recursively decomposed into wavelet packets to construct a three-phase energy feature matrix. Statistical sensitivity analysis is then performed to generate a target feature vector, which is then input into a pre-trained fault location model for fault location.
It significantly improves the accuracy and efficiency of fault location, enabling precise identification of fault type, phase, line and section under complex operating conditions, while reducing computational complexity and storage overhead.
Smart Images
Figure CN122109734A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power distribution network fault detection technology, and in particular to a method, system, device and storage medium for locating power distribution network faults. Background Technology
[0002] With the continuous expansion of the scale of power distribution network construction and the widespread access of distributed power sources, the operating conditions of the power distribution system are becoming increasingly complex, and various types of faults occur frequently. If the faulty line, section and phase cannot be located quickly and accurately, it is easy to cause damage to power equipment, decline in power supply reliability, and even cause the scope of the fault to expand, posing a serious threat to the safe and stable operation of the power distribution network.
[0003] Existing fault location methods for distribution networks mostly employ a method based on comparing the amplitude and phase of steady-state power frequency quantities. Specifically, fault identification is achieved by comparing changes in power frequency electrical quantities before and after a fault. However, this method is susceptible to load fluctuations, system impedance changes, and noise interference, resulting in low location accuracy and poor robustness under complex operating conditions, making it difficult to achieve precise fault location in distribution networks. Summary of the Invention
[0004] This invention provides a method, system, device, and storage medium for locating faults in power distribution networks, in order to solve the technical problem of how to improve existing methods for locating faults in power distribution networks and achieve the effect of improving the accuracy of fault location in power distribution networks.
[0005] To address the aforementioned technical problems, this invention provides a method for locating faults in a power distribution network, including... The three-phase voltage signals of the target distribution network are digitally sampled and processed to obtain discrete time series signals; The first voltage sample value and the historical voltage sample value corresponding to the current time and the historical power frequency cycle are extracted from the discrete time series signal, respectively. Based on the second-order historical period compensation term of the first voltage sample value and the historical voltage sample value, a superimposed voltage signal is generated. Perform wavelet packet recursive decomposition of the superimposed voltage signal at a preset scale to obtain the scaling coefficients of each frequency band of the target distribution network; Determine the energy value of each frequency band scale coefficient, and construct the three-phase energy characteristic matrix of the target distribution network based on the energy value; A statistical sensitivity analysis is performed on the three-phase energy feature matrix, and a target feature vector is generated based on the analysis results; wherein, the statistical sensitivity analysis is designed to screen energy features in the three-phase energy feature matrix that are sensitive to changes in the spatial location of the fault. The target feature vector is input into the pre-trained distribution network fault location model to obtain the fault location result of the target distribution network.
[0006] As one preferred embodiment, generating the superimposed voltage signal based on the second-order historical period compensation term of the first voltage sample value and the historical voltage sample value includes: The historical voltage sampling values include the second voltage sampling value corresponding to the two previous power frequency cycles of the current time, and the third voltage sampling value corresponding to the four previous power frequency cycles of the current time. A second-order historical period compensation term is constructed based on the second voltage sample value and the third voltage sample value, wherein the second-order historical period compensation term is the difference between twice the second voltage sample value and the third voltage sample value; The first voltage sample value is combined with the second-order historical period compensation term to perform a differential operation to obtain the superimposed voltage signal.
[0007] As one preferred embodiment, the step of performing wavelet packet recursive decomposition of the superimposed voltage signal at a preset scale to obtain the frequency band scale coefficients of the target distribution network includes: Based on the orthogonal filter corresponding to the wavelet packet basis function, the superimposed voltage signal is decomposed into low-frequency and high-frequency components to obtain low-frequency scaling coefficients and high-frequency scaling coefficients. The low-frequency scaling coefficients and the high-frequency scaling coefficients are recursively decomposed into multiple layers to a preset number of decomposition layers to obtain the scaling coefficients of each frequency band under the preset number of decomposition layers.
[0008] As one preferred embodiment, determining the energy value of each frequency band scale coefficient and constructing the three-phase energy characteristic matrix of the target distribution network based on the energy values includes: Based on the definition of discrete signal energy, the energy value of the frequency band corresponding to each frequency band scaling coefficient is determined, and a set of energy values is obtained. Extract the energy values of each frequency band corresponding to the three-phase voltage signals of the target distribution network from the energy value set, and generate three-phase independent energy feature vectors; The energy feature vectors are combined according to a preset dimension to obtain a three-phase energy feature matrix that reflects the energy distribution of the three-phase voltage frequency band.
[0009] As one preferred embodiment, the step of performing statistical sensitivity analysis on the three-phase energy characteristic matrix and generating a target feature vector based on the analysis results includes: Determine the energy features of all wavelet packet decomposition levels in the three-phase energy feature matrix for different fault lines and different fault locations in the target distribution network, and obtain the energy feature vector corresponding to each fault location; Calculate the overall energy level average and spatial dispersion standard deviation of each energy feature vector at different fault locations, and calculate the normalized standard deviation of the energy feature vector based on the overall energy level average and Spassky spatial dispersion standard deviation. All energy feature vectors are filtered based on the normalized standard deviation, and a target feature vector is generated based on the filtering results.
[0010] Another aspect of the present invention provides a power distribution network fault location system, comprising: The sampling module is used to digitally sample and process the three-phase voltage signals of the target distribution network to obtain discrete time series signals; The enhancement module is used to extract the first voltage sample value and the historical voltage sample value corresponding to the current time and the historical power frequency cycle in the discrete time series signal, respectively, and generate a superimposed voltage signal based on the second-order historical period compensation term of the first voltage sample value and the historical voltage sample value. The decomposition module is used to perform wavelet packet recursive decomposition of the superimposed voltage signal at a preset scale to obtain the scale coefficients of each frequency band of the target distribution network. A construction module is used to determine the energy value of each frequency band scale coefficient and construct the three-phase energy characteristic matrix of the target distribution network based on the energy value; The analysis module is used to perform statistical sensitivity analysis on the three-phase energy feature matrix and generate a target feature vector based on the analysis results; wherein, the statistical sensitivity analysis is designed to screen energy features in the three-phase energy feature matrix that are sensitive to changes in the spatial location of the fault. The positioning module is used to input the target feature vector into a pre-trained fault location model to obtain the fault location result of the target distribution network.
[0011] Another embodiment of the present invention provides a power distribution network fault location device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the power distribution network fault location method as described above.
[0012] In another embodiment of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, wherein when the device containing the computer-readable storage medium executes the computer program, the power distribution network fault location method described above is implemented.
[0013] Compared with the prior art, the beneficial effects of the present invention are at least one of the following: 1) This invention effectively counteracts the slow change trend of distribution network load and low-frequency interference components by constructing a superimposed voltage signal with a second-order historical period compensation term, significantly amplifies the transient components of sudden faults, and solves the problem that the traditional single-cycle differential superposition quantity is sensitive to load dynamic changes and the transient characteristic energy is not prominent. At the same time, based on a preset scale, wavelet packet recursive decomposition is carried out on the superimposed voltage signal to achieve a fine characterization of transient high-frequency features, laying an accurate frequency domain foundation for fault feature extraction.
[0014] 2) This invention abandons the empirical frequency band selection method and uses statistical sensitivity analysis to filter energy features that are sensitive to changes in the spatial location of faults to generate target feature vectors. This effectively eliminates redundant features, reduces computational complexity, and greatly improves the pertinence and effectiveness of fault features, making the feature extraction process more suitable for the fault location needs of complex power distribution networks. Attached Figure Description
[0015] Figure 1 This is a flowchart illustrating a power distribution network fault location method in one embodiment of the present invention. Figure 2 This is a structural block diagram of a power distribution network fault location system in one embodiment of the present invention; Figure 3 This is a structural block diagram of a power distribution network fault location device in one embodiment of the present invention; Figure label: Among them, 11. Sampling module; 12. Enhancement module; 13. Decomposition module; 14. Construction module; 15. Analysis module; 16. Positioning module; 21. Processor; 22. Memory. Detailed Implementation
[0016] 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. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.
[0017] In the description of this invention, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.
[0018] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0019] One embodiment of the present invention provides a method for locating faults in a distribution network. For details, please refer to [link to relevant documentation]. Figure 1 , Figure 1 The diagram shown is a flowchart of a power distribution network fault location method according to one embodiment of the present invention, which includes steps S1-S6: S1: The three-phase voltage signals of the target distribution network are digitally sampled and processed to obtain discrete-time series signals; specifically, monitoring devices are deployed at key nodes or line ends of the distribution network to collect three-phase voltage signals in real time, and the analog voltage signals are digitally sampled at a fixed sampling frequency to form discrete-time series signals. in, The sampling point number; The number of sampling points within each power frequency cycle is: These are the three-phase voltages, a, b, and c, respectively.
[0020] S2: Extract the first voltage sample value and the historical voltage sample value corresponding to the current time and the historical power frequency cycle from the discrete time series signal respectively. Generate the superimposed voltage signal based on the second-order historical period compensation term of the first voltage sample value and the historical voltage sample value.
[0021] In the field of distribution network fault location technology, traditional fault feature analysis is mostly based on the superposition of the original voltage signal or a first-order periodic difference. For example, the formula for the superposition of a first-order periodic difference is: Assuming the voltage exhibits a slow-changing trend (load change), let: in, This indicates a slow voltage drift caused by load changes; Let the initial voltage be the voltage during the change. Substitute this into the formula for the superposition of the first-order periodic differences: As can be seen from the equation, even without any faults or disturbances, as long as the load changes, the traditional superimposed quantity will never be zero. The first-order differential has a non-zero response to the first-order change component of the signal, making it unable to effectively distinguish between slow load changes and fault transients, resulting in drift superimposed into the characteristic quantity. Especially in high-resistance grounding and weak transient fault scenarios, the transient characteristic energy is masked by load interference, leading to a significant reduction in fault characteristic identification. To solve the above technical problems, this embodiment constructs a superimposed voltage signal that introduces a second-order historical period compensation term, thereby effectively canceling the slow load change component and simultaneously amplifying the sudden transient characteristics of the fault.
[0022] In this embodiment, the three-phase voltage signal digital sampling processing step is performed first. Specifically, voltage monitoring devices are deployed at key monitoring nodes or line ends of the target distribution network to collect three-phase voltage analog signals during the operation of the distribution network in real time. Then, the collected three-phase voltage analog signals are subjected to analog-to-digital conversion and digital sampling at a fixed sampling frequency to obtain discrete-time series signals of the three-phase voltages. The discrete-time series signal of any phase voltage is denoted as... ,in, The sampling point number is used in this invention. The specific value of the sampling frequency is not limited. Preferably, the number of sampling points N in a single power frequency cycle is 256 points, which is suitable for the characteristic analysis requirements of power frequency signals in power distribution networks.
[0023] It should be noted that when generating the superimposed voltage signal, the corresponding sampled value must first be extracted from the discrete time series signal, specifically the current sampling time. The corresponding first voltage sample value This refers to the voltage sample value at the current moment; simultaneously, it extracts the historical power frequency cycle sample value corresponding to the current moment from the historical voltage sample values, including the second voltage sample value of the sampling point corresponding to the two power frequency cycles before the current moment. And the third voltage sample value of the sampling point corresponding to the previous four power frequency cycles at the current moment. ,in This represents the total number of sampling points corresponding to two power frequency cycles. The total number of sampling points corresponding to four power frequency cycles is given. The above historical voltage sampling values are all extracted from the discrete time sequence signal of the same phase voltage to ensure the consistency of the phase and amplitude of the sampling values.
[0024] In this embodiment, a second-order historical period compensation term is constructed based on the extracted second and third voltage sample values. This second-order historical period compensation term is calculated as the difference between twice the second voltage sample value and the third voltage sample value. That is, the second-order historical period compensation term is... The construction logic of this compensation term is limited only to the above-mentioned numerical operation relationship. The specific operation execution method is not limited. It can be implemented by hardware operation circuit or software programming calculation, both of which are within the protection scope of this invention.
[0025] After constructing the second-order historical period compensation term, the first voltage sample value and the second-order historical period compensation term are combined and differentially analyzed to obtain the superimposed voltage signal. Specifically, the combined differential analysis is the first voltage sample value minus the second-order historical period compensation term, and the superimposed voltage signal is obtained. The calculation formula is: After sorting, it becomes Preferably, the above combined difference operation, sample value extraction, and compensation term construction use the same operational benchmark, all using the sample point number as the time index to ensure the time correspondence of each sample value.
[0026] It should be noted that the superimposed voltage signal generation steps described in this embodiment are applicable to the three-phase voltage signals of the target distribution network (A, B, and C). The same sampling value extraction, compensation term construction, and combined differential operation steps are performed on the discrete-time series signal of each phase voltage, ultimately obtaining the superimposed voltage signals corresponding to each of the three phases. These three-phase superimposed voltage signals are calculated independently and do not interfere with each other, reflecting the fault transient characteristics of each phase voltage separately, thus providing a foundation for the subsequent construction of the three-phase energy characteristic matrix. This invention does not limit the calculation order of the three-phase superimposed voltage signals; parallel or serial calculation methods can be used. As long as the corresponding superimposed voltage signals can be obtained, it falls under the implementation method of this embodiment.
[0027] Furthermore, the method of the present invention has the following advantages: 1) To counteract the dynamic trend of load changes.
[0028] Load changes can be approximated as slow, continuous, approximately linear, or low-order curves. Within a time window of 2N to 4N, due to the very short time, the change in voltage amplitude can be considered a very small disturbance. .
[0029] in, , << .
[0030] Substituting the above equation into the formula for the superposition of second-order periodic differences, we get: At this time, if the load changes steadily: In case of sudden disturbance: Significantly increased As can be seen from the above, when the load changes steadily, the superimposed voltage signal calculation method proposed in this invention is approximately 0, while during sudden disturbances, the superimposed voltage signal increases significantly, which can effectively distinguish between steady-state and transient voltage changes.
[0031] 2) Transient component amplification effect.
[0032] Assuming in Sudden changes occur at all times : Substituting the above equation into the formula for the superposition of second-order periodic differences, we get: And if the mutation occurs : It is evident that the second-order structure amplifies non-stationary abrupt changes through a superposition effect, with the amplitude significantly higher than the background variation. By introducing a second-order difference term, the steady-state component and the slowly changing load component are effectively offset, thereby highlighting the transient characteristics at the moment of fault occurrence.
[0033] S3: Perform wavelet packet recursive decomposition on the superimposed voltage signal at a preset scale to obtain the scale coefficients of each frequency band of the target distribution network; In the field of distribution network fault feature extraction, wavelet packet transform has become a common method for transient signal analysis due to its full-spectrum decomposition capability. However, traditional wavelet packet decomposition often employs empirical fixed-level decomposition and frequency band selection methods, failing to provide a refined multi-scale frequency domain characterization of the transient voltage signal during distribution network faults. This can easily lead to the loss of high-frequency transient features or insufficient feature extraction, especially in scenarios involving high-resistance grounding and weak transient faults, making it difficult to extract identifiable frequency band features from the superimposed voltage signal. To address these issues, this embodiment performs recursive wavelet packet decomposition at a preset scale on the superimposed voltage signal, achieving multi-level fine division of the signal's full spectrum and extracting the scale coefficients corresponding to each frequency band.
[0034] In this embodiment, the input signal for performing wavelet packet recursive decomposition is the superimposed voltage signals corresponding to the three phases A, B, and C of the distribution network. The superimposed voltage signal is a discrete-time voltage sequence signal corrected by a second-order historical period compensation term, which effectively cancels the slow load change component and amplifies the transient characteristics of the fault. Before decomposition, a pair of orthogonal filters corresponding to the wavelet packet basis functions are selected, namely a low-pass filter h(k) and a high-pass filter g(k). The low-pass filter is used to extract the low-frequency approximate features in the superimposed voltage signal, and the high-pass filter is used to extract the high-frequency detail features in the signal. This invention does not limit the specific type of wavelet packet basis function; conventional wavelet packet basis functions such as the db series and sym series can be used. As long as the corresponding orthogonal filter can achieve high- and low-frequency separation of the signal, it is within the protection scope of this invention.
[0035] Based on the selected orthogonal filters, the superimposed voltage signal of a single corresponding signal is simultaneously decomposed into low-frequency and high-frequency values to obtain the low-frequency scaling coefficients and high-frequency scaling coefficients of the first layer.
[0036] Specifically, the low-frequency (approximate) scaling coefficients are expressed as: The high-frequency (approximate) scaling factor is expressed as: Where n is the time index after signal downsampling, this layer of decomposition achieves the first frequency band bisection of the superimposed voltage signal, initially separating the signal into two independent feature components: a low-frequency band and a high-frequency band. It should be noted that the wavelet packet decomposition of the three-phase superimposed voltage signal is an independent execution process, that is, the superimposed voltage signal of each phase A, B, and C is decomposed into high and low frequencies in the above manner to obtain the corresponding first-level low-frequency and high-frequency scaling coefficients for each phase, ensuring the independence and specificity of fault feature extraction for each phase. This invention does not limit the execution order of the three-phase decomposition; serial or parallel computation is acceptable.
[0037] After completing the first-level wavelet packet decomposition, the obtained low-frequency scaling coefficients are... and high-frequency scaling coefficients The process involves recursive multi-level decomposition until a preset decomposition level is reached. In this embodiment, the preset decomposition level is denoted as . Preferred By selecting 11 layers, this decomposition layer number can completely cover the spectral range of transient voltage signals during power distribution network faults, enabling a refined characterization of high-frequency transient features.
[0038] The specific method of recursive decomposition is as follows: for the th Any scale coefficient obtained from the layer The low-frequency and high-frequency sub-node scaling coefficients of the j-th layer are obtained by performing high- and low-frequency orthogonal decomposition using the low-pass filter h(k) and the high-pass filter g(k).
[0039] Specifically, the low-frequency child node scaling factor is expressed as: The high-frequency child node scaling factor is expressed as: in, , , For the first The index number of the layer scale coefficients. It should be noted that during the recursive decomposition process, all scale coefficients of each layer need to undergo high-low frequency decomposition to ensure the integrity of the frequency band division of each layer. This invention does not limit the operation priority of the recursive decomposition and can be executed in the order of low frequency priority or high frequency priority. As long as all scale coefficients under the preset number of layers can be obtained, it is a implementation method of this embodiment.
[0040] Through the above recursive multi-level decomposition, the preset number of decomposition levels is reached. Then, the scaling factors of all frequency bands under this layer can be obtained, the first... The layer contains Group scaling factor, denoted as Each set of scaling coefficients uniquely corresponds to a feature component of the superimposed voltage signal within a frequency band, and the frequency bands corresponding to all scaling coefficients are non-overlapping and non-exclusive, completely covering the entire spectrum of the superimposed voltage signal. In this embodiment, the time index of each frequency band scaling coefficient is obtained after downsampling, ensuring the consistency of the time dimension of scaling coefficients of different levels and different frequency bands. The bandwidth is determined by the preset number of decomposition levels. The higher the number of decomposition levels, the finer the frequency band division, and the more accurate the characterization of the high-frequency characteristics of fault transients.
[0041] S4: Determine the energy value of the scaling coefficient for each frequency band, and construct the three-phase energy characteristic matrix of the target distribution network based on the energy value.
[0042] In distribution network fault characteristic analysis, frequency band energy characteristics are core data reflecting the transient characteristics of faults. Traditional methods often directly use the scaling coefficients after wavelet packet decomposition for fault identification, without specifically extracting and integrating the energy distribution patterns of each frequency band. This results in insufficient fault feature identification and makes it difficult to accurately determine the faulty phase and line based on the energy correlation of three-phase voltages. To address these issues, this embodiment, based on the definition of discrete signal energy, constructs a feature matrix that comprehensively reflects the frequency band energy distribution of three-phase voltages by accurately calculating the energy values of the scaling coefficients for each frequency band.
[0043] In this embodiment, the prerequisite for constructing the three-phase energy feature matrix is that the wavelet packet recursive decomposition of the superimposed voltage signal has been completed, that is, the scaling coefficients of each frequency band corresponding to the three phases A, B, and C of the target distribution network have been obtained through the steps described above. Let the first... The set of frequency band scaling factors of the layer is ,in For frequency band indexing, For time indexing, each set of scaling coefficients corresponds to an independent frequency band feature component. It should be noted that the frequency band scaling coefficients of the three-phase voltages are all obtained through wavelet packet decomposition with the same parameters, ensuring the consistency of frequency band division and the comparability of energy characteristics of each phase. This invention does not impose a unique limitation on the specific parameters of wavelet packet decomposition (such as wavelet basis functions and the number of decomposition levels). As long as the frequency band scaling coefficients covering the entire spectrum of the signal can be obtained, they are all within the protection scope of this invention.
[0044] Based on the definition of discrete signal energy, the energy value of the frequency band corresponding to each frequency band scaling factor is calculated. In this embodiment, the energy of a discrete signal is defined as the sum of the squares of the amplitudes of each sampling point of the signal. For any frequency band scaling factor... Its corresponding frequency band energy value The calculation formula is: in, Let be the absolute value of the amplitude at the nth sampling point. The energy values of the scaling coefficients for each frequency band of the three-phase voltage are calculated sequentially using the above formula, resulting in the energy value sets for each of the three phases. These are denoted as . , , ,in, This represents the energy value of the 0th frequency band of phase A, and the rest are calculated similarly. This invention does not limit the calculation precision of the energy value; an appropriate calculation precision can be selected according to actual engineering needs, as long as it accurately reflects the energy distribution characteristics of each frequency band.
[0045] Extract the energy values of each frequency band of the corresponding three-phase voltage signal from each energy value set to generate three-phase independent energy feature vectors. Specifically, extract the energy value set of phase A... Arranged in frequency band index order to form the A-phase energy characteristic vector. Similarly, the set of energy values for phase B... Arranged as B-phase energy eigenvectors Set the energy values of phase C Arranged as C-phase energy eigenvectors It should be noted that the three-phase energy feature vectors all have a dimension of 2J (J is the wavelet packet decomposition level), and the frequency band correspondence of the elements in each vector remains consistent, i.e. , , The m-th element corresponds to the energy value of the m-th frequency band, ensuring the effectiveness of subsequent three-phase energy comparison analysis.
[0046] The energy feature vectors are combined according to a preset dimension to obtain a three-phase energy feature matrix. In this embodiment, the preset dimension is "frequency band dimension - phase dimension", that is, the three-phase energy feature matrix is a 3×2 J-dimensional matrix, and its construction form is as follows: Preferably, the element order of each energy feature vector remains unchanged during the combination process, ensuring that each row in the matrix corresponds to the full-band energy characteristics of one phase voltage, and each column corresponds to the energy characteristics of different phase voltages in the same frequency band. This invention does not limit the storage format of the matrix; various data organization methods such as arrays and lists can be used, as long as the frequency band energy distribution information of the three-phase voltages can be completely preserved, all of which fall under the implementation methods of this embodiment.
[0047] S5: Perform statistical sensitivity analysis on the three-phase energy feature matrix and generate target feature vectors based on the analysis results; among them, the statistical sensitivity analysis is designed to screen energy features in the three-phase energy feature matrix that are sensitive to changes in the spatial location of the fault.
[0048] In the feature selection stage of distribution network fault location, traditional methods often rely on experience or fixed rules to select frequency band energy features after wavelet packet decomposition, failing to fully consider the sensitivity of different frequency bands to changes in fault spatial location. This leads to problems such as redundancy and insufficient discriminative ability in the selected features. Especially in multi-line and complex fault scenarios, the lack of feature specificity can easily affect the accuracy and real-time performance of fault location. To solve the above technical problems, this embodiment uses a statistical sensitivity analysis mechanism to adaptively select energy features sensitive to changes in fault spatial location, generating highly discriminative and low-redundancy target feature vectors, providing accurate input for subsequent fault location models.
[0049] In this embodiment, the prerequisite for performing statistical sensitivity analysis is that the three-phase energy characteristic matrix of the target distribution network has been obtained. This matrix is... dimension( The wavelet packet is pre-decomposed into a certain number of layers (preferably J=11 layers). Each row corresponds to the full-band energy characteristics of a phase voltage, and each column corresponds to the energy characteristics of different phase voltages in the same frequency band.
[0050] First, it is necessary to determine the energy features of all wavelet packet decomposition levels in the three-phase energy feature matrix for different faulted lines and fault locations in the target distribution network, thereby obtaining the energy feature vector corresponding to each fault location. Specifically, assuming the target distribution network contains L lines, M typical fault locations are selected on each line (e.g., at 10%, 30%, 50%, 70%, and 90% of the line length). For each fault type (e.g., single-phase ground fault, transient overvoltage), the energy values of all frequency bands in the three-phase energy feature matrix corresponding to each faulted line and fault location are extracted and combined according to the frequency band index order to form an energy feature vector, denoted as: in, This invention provides a line index. It does not limit the number of faulty lines, the number of fault locations selected, or their specific locations; these can be flexibly set according to the actual scale and operating conditions of the target distribution network, as long as the main fault scenarios are covered.
[0051] Next, the overall energy level average and spatial dispersion standard deviation of each energy characteristic (i.e., the energy value sequence corresponding to each frequency band) are calculated at different fault locations. In this embodiment, for the first... Energy characteristics of each frequency band, overall energy level The calculation formula is: This mean reflects the overall distribution level of the energy characteristics of this frequency band; Spatial dispersion standard deviation The calculation formula is: The standard deviation reflects the dispersion of the energy characteristics of this frequency band under different fault lines and different fault locations. The greater the dispersion, the higher the sensitivity of the energy in this frequency band to changes in the spatial location of the fault. It should be noted that this invention does not strictly limit the calculation accuracy of the mean and standard deviation. An appropriate calculation accuracy can be selected according to the actual engineering needs, as long as it can accurately reflect the statistical distribution characteristics of the energy characteristics.
[0052] The overall energy level is calculated based on the above average value. and the standard deviation of spatial dispersion The normalized standard deviation (NDS) of each energy characteristic is calculated to eliminate the influence of differences in energy amplitude across different frequency bands. The formula for calculating the normalized standard deviation is: in, For the first A small wave packet transforms into a sub-band; This represents the standard deviation of the energy characteristic at different fault locations. The mean of this sub-band.
[0053] By normalizing, the sensitivity of frequency band energy characteristics at different levels can be uniformly quantified, making the sensitivity comparison between frequency bands more reasonable and avoiding the problem of overestimation or underestimation of sensitivity due to the overall high energy amplitude of some frequency bands. This invention does not limit the calculation method of the normalization standard deviation; if other normalization methods exist that can achieve uniform sensitivity quantification, they can also be applied to this step, and all are within the scope of protection of this invention.
[0054] After calculating the normalized standard deviation of all frequency band energy features, the features are filtered based on the normalized standard deviation to generate a target feature vector. In this embodiment, the filtering rule is to sort all frequency band energy features in descending order of normalized standard deviation. The larger the normalized standard deviation, the higher the sensitivity of the frequency band energy feature to changes in fault spatial location, and the stronger its discrimination ability. Preferably, the top 10 energy features with the highest normalized standard deviation are selected as effective features, and combined according to their original frequency band index order to form the target feature vector. ,in The frequency band energy characteristic corresponding to the largest normalized standard deviation. The frequency band energy feature that ranks 10th in terms of normalized standard deviation. It should be noted that the present invention does not impose a unique limit on the number of features to be selected. It can be flexibly adjusted according to the complexity of the fault types of the target distribution network and the requirements for positioning accuracy, such as selecting the first 8 or the first 12 features, as long as the discriminative ability and computational efficiency of the target feature vector can be guaranteed.
[0055] S6: Input the target feature vector into the pre-trained distribution network fault location model to obtain the fault location result of the target distribution network.
[0056] In the field of distribution network fault location, traditional fault location methods often require the construction of multiple intelligent models for different fault types and different lines, resulting in high consumption of model storage and computing resources. Furthermore, the lack of a collaborative processing mechanism between fault type identification, fault phase determination, and fault line and section location makes the overall process lengthy and difficult to meet the needs of rapid and accurate fault location in large-scale distribution networks. To address these issues, this embodiment employs a pre-trained distribution network fault location model, combining the energy distribution patterns of the target feature vector and the three-phase energy feature matrix to achieve integrated determination of fault type, fault phase, and fault line and section location, significantly improving the efficiency and accuracy of fault location. The following provides a detailed explanation of this step.
[0057] In this embodiment, the pre-trained distribution network fault location model is preferably a hierarchical general regression neural network (GRNN). This model adopts a four-layer structure, including an input layer, a pattern layer (hidden layer), a summation layer, and an output layer. Its training process has been completed in the preceding stage: historical voltage disturbance data under typical operating conditions (including single-phase grounding faults and transient overvoltage samples) are selected. After signal acquisition, voltage superposition construction, wavelet packet decomposition, energy feature calculation, and statistical sensitivity screening as described above, a training sample set is formed. The feature vectors of the training samples and their corresponding disturbance type labels are stored in the pattern layer. No backpropagation iterative training is required, resulting in high model training efficiency and strong generalization ability. It should be noted that this invention does not limit the specific type of fault location model; other intelligent models with the same fault discrimination capabilities (such as support vector machines, BP neural networks, etc.) can also be applied and are all within the scope of protection of this invention.
[0058] First, the target feature vector is input into the pre-trained fault location model for disturbance type identification. In this embodiment, the target feature vector consists of 10 energy features most sensitive to changes in the spatial location of the fault, selected through statistical sensitivity analysis. The number of input layer nodes is consistent with the dimension of the target feature vector, both being 10. The mode layer uses radial basis functions as activation functions to measure the similarity between the input target feature vector and the training sample feature vectors. The summation layer performs weighted accumulation and normalization processing on the mode layer output through weighted summation units and unweighted summation units, respectively. The output layer outputs continuous values or category identifiers, where an output of 0 indicates a transient overvoltage and an output of 1 indicates a single-phase ground fault.
[0059] The specific discrimination process is as follows: For the target feature vector E, the model output is: in, For the first Energy feature vectors of training samples; Labels for the corresponding disturbance types; This represents the number of training samples. The smoothing factor for the radial basis function (set via cross-validation) is used to compare the output with a preset threshold (preferably 0.5). If the voltage is ≥0.5, it is determined to be a single-phase ground fault; otherwise, it is a transient overvoltage.
[0060] If the fault type is determined to be a single-phase ground fault, the faulty phase must first be determined based on the three-phase energy characteristic matrix. Specifically, the total energy value corresponding to the voltages of phases A, B, and C in the three-phase energy characteristic matrix is calculated. The faulty phase is determined based on the relationship between the total energy values of each phase. The criterion is that the total energy value corresponding to the voltage of any phase is greater than that of the other two phases, that is, the phase with the largest total energy value is the faulty phase. For example, if the total energy value corresponding to the voltage of phase A is significantly greater than that of phases B and C (usually the difference ratio is set to ≥3:1), then phase A is determined to be the faulty phase. This invention does not limit the specific value of this difference ratio and can be flexibly adjusted according to the actual operating conditions of the distribution network.
[0061] After determining the fault phase, the energy characteristic data of that fault phase is extracted from each line of the target distribution network. The energy characteristic distribution of the fault phase of each line is compared to identify the faulty line. Assuming the target distribution network contains L lines, and the faulty phase is denoted as p, the energy characteristic vector corresponding to the faulty phase p for each line l is extracted. The distribution differences of the energy characteristic vectors of the faulty phases of all lines are compared, and the line with the most significant faulty phase energy characteristics (e.g., the highest total energy or the highest peak energy in the critical sensitive frequency band) is identified as the faulty line. It should be noted that this invention does not limit the number of lines or the dimensions of comparison for the faulty phase energy characteristics, as long as the faulty line can be distinguished from the non-faulty line through differences in energy distribution.
[0062] Based on the fault phase determination results and fault line identification results, and combined with the differences in the energy spatial distribution of the three-phase energy feature matrix, the location of the fault section is determined. In this embodiment, the energy feature vector of the fault phase p of the faulty line y is... According to the line segment division rules (such as dividing the line into 10 segments evenly by length), extract the energy feature sub-vectors corresponding to each segment, and calculate the energy difference between adjacent segments. ( (for the segment index), when When the maximum value is reached, the corresponding segment location is the faulty segment. This is because when a fault occurs, the energy distribution between the faulty segment and the adjacent non-faulty segment undergoes a significant abrupt change, and the location with the largest absolute value of the energy difference precisely corresponds to the fault boundary. This invention does not impose a unique limitation on the number of line segments; it can be adjusted to 5, 8, etc., depending on the positioning accuracy requirements, all of which are implementation methods of this embodiment.
[0063] If the fault type is determined to be transient overvoltage, the location of the fault section is directly determined based on the energy distribution characteristics of the three-phase energy characteristic matrix. Since the transient overvoltage experiences energy attenuation during propagation along the line, its energy distribution satisfies… or This refers to the section where the energy peak corresponds to or is adjacent to the transient overvoltage occurrence. Specifically, energy characteristic data of each line and section are extracted, and the peak point with the largest energy value is selected. The section where this peak point is located is the fault section of the transient overvoltage.
[0064] Finally, the fault type identification results, fault phase identification results (transient overvoltage without a fault phase can be marked as "none"), fault line identification results, and fault section location are integrated to form a complete fault location result for the target distribution network. For example, the location result can be represented as "single-phase ground fault, phase A, line 3, section 5" or "transient overvoltage, line 1, section 2". The fault location process in this embodiment significantly improves the location efficiency and accuracy through a step-by-step convergence mechanism of "type identification - phase identification - line identification - section location". Moreover, it can cover multiple fault types and multiple line scenarios with only one pre-trained model, significantly reducing computational complexity and storage overhead. It does not require precise line parameters or high-precision time synchronization equipment, making it highly practical for engineering applications and adaptable to complex operating conditions such as high-resistance grounding and weak transient faults.
[0065] Another embodiment of the present invention provides a power distribution network fault location system. For details, please refer to [link to relevant documentation]. Figure 2 , Figure 2 The diagram shown is a structural block diagram of a power distribution network fault location system according to one embodiment of the present invention, which includes: The sampling module 11 is used to perform digital sampling processing on the three-phase voltage signal of the target distribution network to obtain a discrete time series signal; Enhancement module 12 is used to extract the first voltage sample value and the historical voltage sample value corresponding to the current time and the historical power frequency cycle in the discrete time series signal, respectively, and generate a superimposed voltage signal based on the second-order historical cycle compensation term of the first voltage sample value and the historical voltage sample value; The decomposition module 13 is used to perform wavelet packet recursive decomposition of the superimposed voltage signal at a preset scale to obtain the scale coefficients of each frequency band of the target distribution network. The construction module 14 is used to determine the energy value of each frequency band scale coefficient and construct the three-phase energy characteristic matrix of the target distribution network based on the energy value; Analysis module 15 is used to perform statistical sensitivity analysis on the three-phase energy feature matrix and generate a target feature vector based on the analysis results; wherein, the statistical sensitivity analysis is designed to screen energy features in the three-phase energy feature matrix that are sensitive to changes in the spatial location of the fault. The positioning module 16 is used to input the target feature vector into a pre-trained distribution network fault location model to obtain the fault location result of the target distribution network.
[0066] Another embodiment of the present invention provides a power distribution network fault location device, specifically, see [link to details]. Figure 3 This is a structural block diagram of a power distribution network fault location device provided in an embodiment of the present invention. The power distribution network fault location device provided in this embodiment includes a processor 21, a memory 22, and a computer program stored in the memory 22 and configured to be executed by the processor 21. When the processor 21 executes the computer program, it implements the steps described in the above-described power distribution network fault location method embodiment, for example... Figure 1 The steps S1 to S6 described above; or, when the processor 21 executes the computer program, it implements the functions of each module in the above-described device embodiments, such as the sampling module 11.
[0067] For example, the computer program may be divided into one or more modules, which are stored in the memory 22 and executed by the processor 21 to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the power distribution network fault location device.
[0068] The power distribution network fault location device may include, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will understand that the schematic diagram is merely an example of a power distribution network fault location device and does not constitute a limitation on the device. It may include more or fewer components than shown in the diagram, or combine certain components, or use different components. For example, the power distribution network fault location device may also include input / output devices, network access devices, buses, etc.
[0069] The processor 21 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. The processor 21 is the control center of the power distribution network fault location equipment, connecting various parts of the equipment via various interfaces and lines.
[0070] The memory 22 can be used to store the computer programs and / or modules. The processor 21 implements various functions of the power distribution network fault location device by running or executing the computer programs and / or modules stored in the memory 22 and calling the data stored in the memory 22. The memory 22 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 22 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.
[0071] If the integrated modules of the power distribution network fault location equipment are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.
[0072] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.
[0073] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform steps in the power distribution network fault location method of the above embodiments, for example... Figure 1 Steps S1 to S6 as described above.
[0074] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.
Claims
1. A method for locating faults in a power distribution network, characterized in that, include: The three-phase voltage signals of the target distribution network are digitally sampled and processed to obtain discrete time series signals; The first voltage sample value and the historical voltage sample value corresponding to the current time and the historical power frequency cycle are extracted from the discrete time series signal, respectively. Based on the second-order historical period compensation term of the first voltage sample value and the historical voltage sample value, a superimposed voltage signal is generated. Perform wavelet packet recursive decomposition of the superimposed voltage signal at a preset scale to obtain the scaling coefficients of each frequency band of the target distribution network; Determine the energy value of each frequency band scale coefficient, and construct the three-phase energy characteristic matrix of the target distribution network based on the energy value; A statistical sensitivity analysis is performed on the three-phase energy feature matrix, and a target feature vector is generated based on the analysis results; wherein, the statistical sensitivity analysis is designed to screen energy features in the three-phase energy feature matrix that are sensitive to changes in the spatial location of the fault. The target feature vector is input into the pre-trained distribution network fault location model to obtain the fault location result of the target distribution network.
2. The method for locating faults in a distribution network as described in claim 1, characterized in that, The step of generating a superimposed voltage signal based on the second-order historical period compensation term of the first voltage sample value and the historical voltage sample value includes: The historical voltage sampling values include the second voltage sampling value corresponding to the two previous power frequency cycles of the current time, and the third voltage sampling value corresponding to the four previous power frequency cycles of the current time. A second-order historical period compensation term is constructed based on the second voltage sample value and the third voltage sample value, wherein the second-order historical period compensation term is the difference between twice the second voltage sample value and the third voltage sample value; The first voltage sample value is combined with the second-order historical period compensation term to perform a differential operation to obtain the superimposed voltage signal.
3. The method for locating faults in a distribution network as described in claim 1, characterized in that, The step of performing wavelet packet recursive decomposition of the superimposed voltage signal at a preset scale to obtain the frequency band scale coefficients of the target distribution network includes: Based on the orthogonal filter corresponding to the wavelet packet basis function, the superimposed voltage signal is decomposed into low-frequency and high-frequency components to obtain low-frequency scaling coefficients and high-frequency scaling coefficients. The low-frequency scaling coefficients and the high-frequency scaling coefficients are recursively decomposed into multiple layers to a preset number of decomposition layers to obtain the scaling coefficients of each frequency band under the preset number of decomposition layers.
4. The method for locating faults in a distribution network as described in claim 1, characterized in that, The step of determining the energy value of each frequency band scale coefficient and constructing the three-phase energy characteristic matrix of the target distribution network based on the energy value includes: Based on the definition of discrete signal energy, the energy value of the frequency band corresponding to each frequency band scaling coefficient is determined, and a set of energy values is obtained. Extract the energy values of each frequency band corresponding to the three-phase voltage signals of the target distribution network from the energy value set, and generate three-phase independent energy feature vectors; The energy feature vectors are combined according to a preset dimension to obtain a three-phase energy feature matrix that reflects the energy distribution of the three-phase voltage frequency band.
5. The distribution network fault location method as described in claim 1, characterized in that, The statistical sensitivity analysis of the three-phase energy characteristic matrix, and the generation of a target feature vector based on the analysis results, includes: Determine the energy features of all wavelet packet decomposition levels in the three-phase energy feature matrix for different fault lines and different fault locations in the target distribution network, and obtain the energy feature vector corresponding to each fault location; Calculate the overall energy level average and spatial dispersion standard deviation of each energy feature vector at different fault locations, and calculate the normalized standard deviation of the energy feature vector based on the overall energy level average and the spatial dispersion standard deviation. All energy feature vectors are filtered based on the normalized standard deviation, and a target feature vector is generated based on the filtering results.
6. The method for locating faults in a distribution network as described in claim 1, characterized in that, The step of inputting the target feature vector into a pre-trained distribution network fault location model to obtain the fault location result of the target distribution network includes: The target feature vector is input into the fault location model to identify the disturbance type and obtain the fault type discrimination result. The fault type includes single-phase ground fault and transient overvoltage. Based on the fault type discrimination result and the energy distribution characteristics of the three-phase energy characteristic matrix, the target distribution network is subjected to fault phase discrimination and fault line identification respectively, and the fault phase discrimination result and fault line identification result are obtained. Based on the fault phase determination results and the fault line identification results, as well as the energy spatial distribution differences of the three-phase energy characteristic matrix, the location of the fault section of the target distribution network is determined. Based on the fault type identification result, the fault phase identification result, the fault line identification result, and the fault section location, the fault location result of the target distribution network is determined.
7. The distribution network fault location method as described in claim 6, characterized in that, The step of determining the fault location result of the target distribution network based on the fault type discrimination result, the fault phase discrimination result, the fault line identification result, and the fault section location includes: If the fault type determination result is a single-phase ground fault, calculate the total energy value corresponding to the phase voltage in the three-phase energy characteristic matrix, and determine the fault phase based on the total energy. Extract the energy characteristic data of the faulty phase in each line of the target distribution network, compare the energy characteristic distribution of the faulty phase in each line, and determine the faulty line based on the comparison results.
8. A power distribution network fault location system, characterized in that, include: The sampling module is used to digitally sample and process the three-phase voltage signals of the target distribution network to obtain discrete time series signals; The enhancement module is used to extract the first voltage sample value and the historical voltage sample value corresponding to the current time and the historical power frequency cycle in the discrete time series signal, respectively, and generate a superimposed voltage signal based on the second-order historical period compensation term of the first voltage sample value and the historical voltage sample value. The decomposition module is used to perform wavelet packet recursive decomposition of the superimposed voltage signal at a preset scale to obtain the scale coefficients of each frequency band of the target distribution network. A construction module is used to determine the energy value of each frequency band scale coefficient and construct the three-phase energy characteristic matrix of the target distribution network based on the energy value; The analysis module is used to perform statistical sensitivity analysis on the three-phase energy feature matrix and generate a target feature vector based on the analysis results; wherein, the statistical sensitivity analysis is designed to screen energy features in the three-phase energy feature matrix that are sensitive to changes in the spatial location of the fault. The localization module is used to input the target feature vector into a pre-trained distribution network fault localization model to obtain the fault localization result of the target distribution network.
9. A power distribution network fault location device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the power distribution network fault location method as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the device containing the computer-readable storage medium executes the computer program, it implements the power distribution network fault location method as described in any one of claims 1 to 7.