Power distribution network hidden fault detection method, device and equipment and storage medium

Abstract

The application relates to a power distribution network implicit fault detection method, device, equipment and storage medium. The method comprises the following steps: extracting at least two operation state vectors of the power distribution network from the transient recording wave data of the power distribution network; fusing the at least two operation state vectors according to the importance of different operation state vectors to the stable operation of the power distribution network to obtain a fused state vector; and determining the implicit fault detection result of the power distribution network according to the distance between the fused state vector and each reference state vector. Since the reference state vectors contain the state vectors obtained in the case that the power distribution network generates an implicit fault, the distance between the fused state vector and the reference state vector can be used to accurately determine whether there is an implicit fault in the power distribution network, and the implicit fault detection accuracy is improved.

Description

Technical Field

[0001] This application relates to the field of fault detection technology, and in particular to a method, apparatus, equipment and storage medium for detecting latent faults in power distribution networks. Background Technology

[0002] Latent faults in distribution networks refer to faults with weak signals and intermittent occurrences in the early stages. In the early stages, latent faults generally do not cause significant damage to the distribution network. However, as the insulation at the fault location continues to deteriorate, latent faults are very likely to develop into true faults with obvious fault characteristics and great destructive power, thereby affecting the stable operation of the distribution network.

[0003] Currently, fault detection in power distribution networks typically involves setting a fault detection threshold. When the fluctuation of electrical quantities in the power distribution network reaches the fault detection threshold, a fault is identified. However, because latent fault signals are weak and their characteristics are not very obvious, the fluctuation of electrical quantities when a latent fault occurs in the power distribution network is also small enough not to trigger the fault detection threshold, thus failing to detect the latent fault.

[0004] Therefore, in order to detect latent faults, related technologies typically lower the fault detection threshold so that fluctuations in electrical quantities caused by latent faults can trigger the threshold and thus detect the fault. However, system disturbances such as load changes and line switching in the distribution network can also cause fluctuations in electrical quantities, triggering the fault detection threshold and leading to false alarms and low fault detection accuracy. Summary of the Invention

[0005] Therefore, it is necessary to provide a method, device, equipment, and storage medium for detecting latent faults in power distribution networks that can improve the accuracy of fault detection, in order to address the aforementioned technical problems.

[0006] Firstly, this application provides a method for detecting latent faults in a power distribution network, comprising:

[0007] Acquire transient waveform data of the distribution network; wherein the transient waveform data is obtained by collecting electrical quantities of the distribution network when abnormal fluctuations in electrical quantities of the distribution network are detected;

[0008] Extract at least two operating state vectors of the distribution network from the transient waveform data;

[0009] Based on the importance of different operating state vectors to the stable operation of the distribution network, the at least two operating state vectors are fused to obtain a fused state vector;

[0010] The latent fault detection result of the distribution network is determined based on the distance between the fused state vector and each reference state vector; wherein, each reference state vector contains a state vector obtained when a latent fault occurs in the distribution network.

[0011] In one embodiment, determining the latent fault detection result of the distribution network based on the distance between the fused state vector and each reference state vector includes: for each reference state vector, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector; and determining the latent fault detection result of the distribution network based on the sum of the fault scores.

[0012] In one embodiment, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector includes: applying distance weights to weight the distance between the reference state vector and the fused state vector to obtain a first weighted value; applying the vector weights corresponding to the reference state vector to weight the reciprocal of the exponential function value of the first weighted value to obtain a second weighted value; and determining the fault score of the fused state vector under the reference state vector based on the sum of the second weighted value and a preset bias.

[0013] In one embodiment, each reference state vector corresponds to a fault label, the fault label being used to characterize whether the corresponding reference state vector was obtained under the condition of a latent fault in the distribution network; each reference state vector is determined by: obtaining multiple sample state vectors; constructing a classification hyperplane based on the multiple sample state vectors; wherein, sample state vectors corresponding to the same fault label among the multiple sample state vectors are located on the same side of the classification hyperplane; and determining a reference state vector based on the sample state vectors located on both sides of the classification hyperplane whose distance from the classification hyperplane is less than a distance threshold.

[0014] In one embodiment, the step of fusing at least two operating state vectors to obtain a fused state vector based on the importance of different operating state vectors to the stable operation of the distribution network includes: standardizing the at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector; and fusing the standard state vectors according to the importance of different operating state vectors to obtain a fused state vector.

[0015] In one embodiment, each of the operating state vectors includes an operating state value; the standardization process of the at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector includes: determining the mean and standard deviation of the operating state values ​​of the at least two operating state vectors; for each operating state vector, determining the difference between the operating state value and the mean of the operating state vector, and determining the standard state vector of the operating state vector based on the ratio of the difference to the standard deviation.

[0016] Secondly, this application also provides a latent fault detection device for a power distribution network, comprising:

[0017] The data acquisition module is used to acquire transient waveform data of the distribution network; wherein, the transient waveform data is obtained by collecting the electrical quantities of the distribution network when abnormal fluctuations in the electrical quantities of the distribution network are detected.

[0018] The vector extraction module is used to extract at least two operating state vectors of the power distribution network from the transient waveform data.

[0019] The vector fusion module is used to fuse at least two operating state vectors according to the importance of different operating state vectors to the stable operation of the distribution network, so as to obtain a fused state vector.

[0020] The fault detection module is used to determine the latent fault detection result of the distribution network based on the distance between the fused state vector and each reference state vector; wherein, each reference state vector contains a state vector obtained when a latent fault occurs in the distribution network.

[0021] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the various method embodiments provided in the first aspect above.

[0022] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the various method embodiments provided in the first aspect above.

[0023] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the various method embodiments provided in the first aspect above.

[0024] The aforementioned method, device, equipment, and storage medium for detecting latent faults in power distribution networks provide a reliable data foundation for latent fault detection by performing transient waveform recording before and after electrical quantity fluctuations in the power distribution network. Furthermore, the operating state vectors are fused according to the importance of different operating state vectors extracted from the transient waveform data. The resulting fused state vector can accurately express the fault characteristics of the power distribution network. Finally, since the reference state vector includes the state vector obtained under the condition of latent faults, the distance between the fused state vector and the reference state vector can accurately determine whether a latent fault has occurred in the power distribution network, thus improving the fault detection accuracy. Moreover, after detecting a latent fault, it can be repaired in a timely manner, improving the stability of the power distribution network. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is an application environment diagram of a latent fault detection method for a power distribution network in one embodiment;

[0027] Figure 2 This is a flowchart illustrating a method for detecting latent faults in a power distribution network in one embodiment.

[0028] Figure 3 This is a flowchart illustrating the vector fusion steps in one embodiment;

[0029] Figure 4 This is a flowchart illustrating the standardized processing steps in one embodiment;

[0030] Figure 5 This is a flowchart illustrating the steps for determining a reference state vector in one embodiment;

[0031] Figure 6 This is a flowchart illustrating the fault detection steps in one embodiment;

[0032] Figure 7 This is a flowchart illustrating the steps for determining the fault score in one embodiment;

[0033] Figure 8 This is a flowchart illustrating a method for detecting latent faults in a power distribution network in another embodiment.

[0034] Figure 9 This is a structural block diagram of a latent fault detection device for a power distribution network in one embodiment;

[0035] Figure 10 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0037] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0038] The design concept of this application is briefly introduced below:

[0039] Latent faults in distribution networks refer to faults with weak signals and intermittent occurrences in the early stages. In the early stages, latent faults generally do not cause significant damage to the distribution network. However, as the insulation at the fault location continues to deteriorate, latent faults are very likely to develop into true faults with obvious fault characteristics and great destructive power, thereby affecting the stable operation of the distribution network.

[0040] Currently, fault detection in power distribution networks typically involves setting a fault detection threshold. When the fluctuation of electrical quantities in the power distribution network reaches the fault detection threshold, a fault is identified. However, because latent fault signals are weak and their characteristics are not very obvious, the fluctuation of electrical quantities when a latent fault occurs in the power distribution network is also small enough not to trigger the fault detection threshold, thus failing to detect the latent fault.

[0041] Therefore, in order to detect latent faults, related technologies typically lower the fault detection threshold so that fluctuations in electrical quantities caused by latent faults can trigger the threshold and thus detect the fault. However, system disturbances such as load changes and line switching in the distribution network can also cause fluctuations in electrical quantities, triggering the fault detection threshold and leading to false alarms and low fault detection accuracy.

[0042] In view of this, embodiments of this application propose a method, apparatus, device, and storage medium for detecting latent faults in a distribution network. By performing transient waveform recording before and after electrical quantity fluctuations in the distribution network, a reliable data foundation is provided for latent fault detection. Furthermore, the operating state vectors are fused according to the importance of different operating state vectors extracted from the transient waveform data. The resulting fused state vector can accurately express the fault characteristics of the distribution network. Finally, since the reference state vector includes the state vector obtained under the condition of latent faults, the distance between the fused state vector and the reference state vector can accurately determine whether a latent fault has occurred in the distribution network, improving the fault detection accuracy. Moreover, after detecting a latent fault, it can be repaired in a timely manner, improving the stability of the distribution network.

[0043] The latent fault detection method for distribution networks provided in this application can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store data that server 104 needs to process, such as transient waveform data and reference state vectors. The data storage system can be integrated onto server 104 or placed in the cloud or on other network servers. Terminal 102 can send transient waveform data of the distribution network to server 104. Server 104 extracts at least two operating state vectors of the distribution network from the transient waveform data. Based on the importance of different operating state vectors to the stable operation of the distribution network, it fuses the at least two operating state vectors to obtain a fused state vector. Based on the distance between the fused state vector and each reference state vector, it determines the latent fault detection result of the distribution network.

[0044] The terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, etc. The server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0045] In one exemplary embodiment, such as Figure 2 As shown, a method for detecting latent faults in a distribution network is provided, which can be applied to... Figure 1 Taking server 104 as an example, the explanation includes:

[0046] S201, acquire transient waveform data of the distribution network.

[0047] Under normal operation, the electrical quantities (such as voltage and current) of the distribution network are usually in a stable power frequency state. However, when a fault or disturbance occurs in the distribution network, a short-term (usually millisecond to second-level) "transient process" will occur, and the electrical quantities of the distribution network will also fluctuate abnormally. The transient waveform data in this application embodiment is obtained by collecting the electrical quantities of the distribution network when abnormal fluctuations are detected.

[0048] Abnormal fluctuations in electrical quantities in the distribution network can be abnormal fluctuations in a single electrical quantity, such as voltage fluctuations exceeding the voltage fluctuation threshold or current fluctuations exceeding the current fluctuation threshold. Abnormal fluctuations in electrical quantities in the distribution network can also be abnormal fluctuations in multiple electrical quantities, such as voltage fluctuations exceeding the voltage fluctuation threshold and current fluctuations exceeding the current fluctuation threshold. These are not listed one by one here.

[0049] Transient waveform data includes various electrical quantities of the distribution network collected before and after abnormal fluctuations, such as three-phase voltage, three-phase current, zero-sequence voltage, zero-sequence current, active power, reactive power, frequency, and circuit breaker opening and closing indicator lights, without specific limitations here.

[0050] Optionally, when abnormal fluctuations in the electrical quantities of the distribution network are detected, transient waveform data of the distribution network is obtained by using a feeder terminal unit (FTU) to perform transient waveform recording for a preset duration at a preset sampling frequency and sampling accuracy. The sampling frequency can be 12500 Hz, the sampling accuracy can be 16-bit analog-to-digital conversion, and the preset duration can be 240 milliseconds.

[0051] S202, extract at least two operating state vectors of the distribution network from transient waveform data.

[0052] The operating state vector is extracted from transient waveform data and is used to characterize the operating state of the distribution network. The operating state vector can be the true effective value, the coefficient of variation, the skewness, the harmonic content, etc., without specific limitations.

[0053] Optionally, if at least two operating state vectors include the true RMS value of the current, the true RMS value of the current, Irms, can be determined by the following formula:

[0054]

[0055] Where N is the number of sampling points, i k Let be the instantaneous current value at the k-th sampling point.

[0056] Optionally, if at least two operating state vectors include current discrepancy coefficients, the current discrepancy coefficients Cv can be determined by the following formula:

[0057]

[0058] Where σ is the standard deviation of the instantaneous current value at each sampling point, and μ is the mean of the instantaneous current value at each sampling point.

[0059] Optionally, if at least two operating state vectors include current skewness, the current skewness skew(X) can be determined by the following formula:

[0060]

[0061] Where k2 is the second-order central moment of the instantaneous current value at each sampling point, and k3 is the third-order central moment of the instantaneous current value at each sampling point.

[0062] Optionally, if at least two operating state vectors include harmonic content, the harmonic content can be at least one of the 3rd harmonic content, the 5th harmonic content, and the 7th harmonic content, and the single harmonic content S can be determined by the following formula:

[0063]

[0064] Among them, S x For the Xth harmonic content, I x I is the effective value of the Xth harmonic current. r This represents the effective value of the fundamental current. By analyzing these harmonics, we can reflect the characteristics of the signal from a frequency domain perspective, and combine them with other time domain characteristics to achieve multi-domain feature fusion.

[0065] S203, based on the importance of different operating state vectors to the stable operation of the distribution network, at least two operating state vectors are fused to obtain a fused state vector.

[0066] Optionally, based on the importance of different operating state vectors to the stable operation of the distribution network, the weight of each operating state vector is determined, and at least two operating state vectors are weighted and fused to obtain a fused state vector. If the operating state vector is of high importance to the stable operation of the distribution network, a higher weight can be assigned to this operating state vector; if the operating state vector is of low importance to the stable operation of the distribution network, a higher weight can be assigned to this operating state vector. For example, if operating state vector 1 is [3] with a weight of 0.4 and operating state vector 5 is [5] with a weight of 0.6, then the fused state vector can be [1.2,3].

[0067] S204. Based on the distance between the fused state vector and each reference state vector, determine the latent fault detection result of the distribution network.

[0068] Each reference state vector includes a state vector acquired under conditions of latent faults in the distribution network. Optionally, historical waveform data of distribution networks in multiple different regions (such as urban, rural, and industrial areas) can be acquired. This historical waveform data can be collected under conditions of latent faults in the distribution network or under conditions of disturbances in the distribution network. Then, reference state vectors are extracted from these two types of historical waveform data. Each extracted reference state vector contains a state vector acquired under conditions of latent faults in the distribution network, which can characterize the features of latent faults in the distribution network. Each extracted reference state vector also contains a state vector acquired under conditions of disturbances in the distribution network, which can characterize the features of disturbances in the distribution network. Thus, based on the distance between the fused state vector and the reference state vectors, it is possible to determine whether the fused state vector was acquired under conditions of latent faults or disturbances in the distribution network, thereby obtaining the latent fault detection result.

[0069] The distance between the fused state vector and the reference state vector can be determined by Euclidean distance, Manhattan distance, cosine similarity, etc., and no specific limitation is made here.

[0070] In the above embodiments, by performing transient waveform recording before and after the electrical quantity fluctuations of the distribution network, a reliable data foundation is provided for the detection of latent faults. Furthermore, the operating state vectors are fused according to the importance of different operating state vectors extracted from the transient waveform data. The resulting fused state vector can accurately express the fault characteristics of the distribution network. Finally, since the reference state vector includes the state vector obtained under the condition of latent faults, the distance between the fused state vector and the reference state vector can accurately determine whether a latent fault has occurred in the distribution network, thereby improving the fault detection accuracy. Moreover, after a latent fault is detected, it can be repaired in a timely manner, improving the stability of the distribution network.

[0071] Different operating state vectors may have different dimensions, therefore, before vector fusion in S203, the different operating state vectors can be standardized. In an exemplary embodiment, such as... Figure 3 As shown, the vector fusion steps in S203 are further refined, including:

[0072] S301, standardize at least two running state vectors to obtain the standard state vector corresponding to each running state vector.

[0073] Optionally, standardizing the running state vector can map different running state vectors to the same interval, for example, scaling the running state vector to the [0,1] interval; standardizing the running state vector can also be done by dividing the running state vector by the vector length to normalize the running state vector; standardizing the running state vector can also be done based on the mean and standard deviation of different running state vectors.

[0074] S302, based on the importance of different operating state vectors, the standard state vectors are fused to obtain the fused state vector.

[0075] Optionally, based on the importance of different operating state vectors, the weight of each standard state vector is determined, and the standard state vectors are weighted and fused to obtain a fused state vector. For example, if the weight of operating state vector 1 is 0.4 and the weight of operating state vector 2 is 0.6, then the weight of standard state vector 1 corresponding to operating state vector 1 is 0.4, and the weight of standard state vector 2 corresponding to operating state vector 2 is 0.6.

[0076] In the above embodiments, different operating state vectors are standardized to maintain a unified dimension, avoiding the inability of the fused state vector to correctly express fault characteristics due to different dimensions. This can improve the accuracy of the fused state vector and thus enhance the accuracy of fault detection results.

[0077] In one exemplary embodiment, each running state vector includes running state values, such as... Figure 4 As shown, the standardization process in S301 is further refined, including:

[0078] S401, determine the mean and standard deviation of the running state values ​​of at least two running state vectors.

[0079] S402, for each operating state vector, determine the difference between the operating state value and the mean of the operating state vector, and determine the standard state vector of the operating state vector based on the ratio of the difference to the standard deviation.

[0080] Specifically, when the running state vector includes running state values, the running state vector is a 1-dimensional vector, and the mean and standard deviation of different running state values ​​can be determined. For example, the mean of different running state values ​​is 8 and the standard deviation is 0.5. For running state vector 1[5], its running state value is 5, the difference between the running state value and the mean is 3, and the ratio of the difference to the standard deviation is 6. Then the standard state vector of running state vector 1 is[6].

[0081] In the above embodiments, standardization based on the mean and standard deviation of the operating state values ​​can eliminate the influence caused by the different dimensions of different operating state values ​​and improve the accuracy of the standard state vector.

[0082] Optionally, each reference state vector corresponds to a fault label, which is used to characterize whether the corresponding reference state vector was obtained under the condition of a latent fault in the distribution network. For example, if the reference state vector was obtained under the condition of a latent fault in the distribution network, the fault label of the reference state vector is 1; if the reference state vector was not obtained under the condition of a latent fault in the distribution network, the fault label of the reference state vector is -1.

[0083] In one exemplary embodiment, such as Figure 5 As shown, each reference state vector is determined in the following way:

[0084] S501, obtain multiple sample state vectors.

[0085] The multiple sample state vectors are extracted from historical waveform data collected from distribution networks in multiple different regions. Optionally, for each historical waveform data, the true RMS value of the current, the current dispersion coefficient, the current skewness, the third harmonic content of the current, the fifth harmonic content of the current, and the sixth harmonic content of the current are extracted to form a sample state vector. Furthermore, a fault label is set for each sample state vector. If the historical waveform data was collected under conditions of latent faults in the distribution network, the fault label of the sample state vector collected from the historical waveform data can be 1. If the historical waveform data was collected under conditions of disturbances in the distribution network, the fault label of the sample state vector collected from the historical waveform data can be -1.

[0086] S502, construct a classification hyperplane based on multiple sample state vectors.

[0087] The classification hyperplane is a hyperplane in space that can divide multiple sample state vectors into two parts. It can also be called the classification boundary. Sample state vectors with the same fault label are located on the same side of the classification hyperplane.

[0088] S503. Based on the sample state vectors located on both sides of the classification hyperplane, the sample state vectors whose distance from the classification hyperplane is less than a distance threshold are determined to be reference state vectors.

[0089] Specifically, after obtaining multiple sample state vectors, a classification hyperplane can be found among the multiple sample state vectors such that sample state vectors with the same fault label are located on the same side of the classification hyperplane. In this way, after obtaining the fused state vector, the fault label of the sample state vector on the side where the fused state vector is located can be used as the latent fault detection result of the fused state vector by determining which side of the classification hyperplane the fused state vector is on.

[0090] It is understandable that for multiple sample state vectors, their positions in space are already fixed. Therefore, the direction and position of the classification hyperplane are actually only affected by the few sample state vectors closest to the classification hyperplane. These few sample state vectors can be called support vectors (i.e., reference state vectors). In this embodiment, the sample state vectors whose distance from the classification hyperplane is less than a distance threshold are used as reference state vectors. The distance threshold can be set according to actual needs. For example, if the distance threshold is 0, the reference state vector is located on the classification hyperplane.

[0091] In the above embodiments, the sample state vector that is close to the classification hyperplane is used as the reference state vector. Then, it is possible to determine whether there is a hidden fault in the distribution network based on the reference state vector, thereby improving the accuracy of hidden fault detection.

[0092] In one exemplary embodiment, such as Figure 6 As shown, the fault detection steps in S204 are further detailed, including:

[0093] S601, for each reference state vector, determine the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector.

[0094] Specifically, the distance between the reference state vector and the fused state vector can be Euclidean distance, Manhattan distance, cosine similarity, etc. Taking Euclidean distance as an example, the Euclidean distance between the reference state vector and the fused state vector can be directly used as the fault score of the fused state vector under the reference state vector. Alternatively, weights can be set for the reference state vectors according to their importance, and the Euclidean distance can be weighted, with the weighted value used as the fault score of the fused state vector under the reference state vector. It is understandable that the fault score is used to characterize the fault situation in the distribution network and is not limited to the form of a score; it can also be a fault degree, fault value, etc., which will not be listed here.

[0095] S602 determines the latent fault detection results of the distribution network based on the sum of the fault scores.

[0096] Specifically, a score threshold can be set. If the sum of the scores of all faults is greater than the score threshold, the result of the latent fault detection in the distribution network is determined to be that a latent fault exists. If the sum of the scores of all faults is not greater than the score threshold, the result of the latent fault detection in the distribution network is determined to be that no latent fault exists.

[0097] In the above embodiments, the latent fault detection result is determined based on the fault score of the fused state vector under each reference state vector, which can fully combine different reference state vectors for fault detection and improve the accuracy of fault detection.

[0098] In one exemplary embodiment, such as Figure 7 As shown, the steps for determining the fault score in S601 are further refined, including:

[0099] S701 uses distance weighting to weight the distance between the reference state vector and the fused state vector to obtain the first weighted value.

[0100] S702, using the vector weights corresponding to the reference state vector, weights the reciprocal of the exponential function value of the first weighted value to obtain the second weighted value.

[0101] S703, based on the sum of the second weighting value and the preset bias, determine the fault score of the fused state vector under the reference state vector.

[0102] Specifically, the distance weight and preset bias can be set based on experience, multiple trials, and actual needs, and are not specifically limited here. The vector weight of the reference state vector can be set according to the importance of the reference state vector. For example, the closer the reference state vector is to the classification hyperplane, the higher its importance can be considered, and the higher its weight. The fault score can be obtained by directly using the sum of the second weighted value and the preset bias, or by weighting the sum of the second weighted value and the preset bias.

[0103] In the above embodiments, by setting distance weights, vector weights, and preset biases, more flexibility is added to the process of determining fault scores. This allows for the determination of fault scores from fused state vectors based on actual needs, thereby improving the accuracy of fault scores.

[0104] In low-dimensional space, it may be impossible to find a classification hyperplane capable of classifying multiple sample state vectors into two classes based on fault labels. Therefore, in an exemplary embodiment, based on the idea of ​​Support Vector Machines (SVMs), a radial basis function (RBF) is used to map multiple sample state vectors to a high-dimensional space. Based on the RBF and SVM decision functions, the SVM boundary (i.e., the classification hyperplane) is established through iterative training, yielding the hyperparameters in the RBF and decision functions, as well as the support vectors constituting the SVM boundary. Optionally, the weights of different running state vectors can also be optimized and updated during iterative training to maximize the classification margin.

[0105] Optionally, the expression for the radial basis function kernel is:

[0106]

[0107] Where γ is the kernel function width parameter (controlling the range of influence of the sample), x i and x j For two different sample state vectors It is the square of the Euclidean distance between the two sample state vectors.

[0108] The decision function of a support vector machine is:

[0109]

[0110] in, For support vector x i Lagrange multipliers, y i x ∈{−1,1} is a support vector i The fault label is (1 or -1), b is the bias term, and x is the state vector that needs to be determined by the decision function to determine whether it belongs to a latent fault.

[0111] Optionally, during iterative training, a hybrid algorithm of gray wolf and simulated annealing is used to optimize the hyperparameters. The parameters of the gray wolf and simulated annealing algorithm are set as follows: gray wolf population size is 20, maximum number of iterations is 100, initial annealing temperature is 100, cooling coefficient is 0.95, and the termination condition is that after 20 consecutive iterations, the optimal fitness improvement is <0.001.

[0112] In an exemplary embodiment, a branch line experiences intermittent discharge due to insulation aging. After detecting abnormal fluctuations in the electrical quantities of the distribution network, transient waveform recording is performed through the line FTU to obtain transient waveform data. The sampling frequency is 12500 Hz, 16-bit analog-to-digital conversion is used, and the recording duration is 240 ms. Six operational state vectors were extracted from the transient waveform data: true effective value Irms = 36.2A; coefficient of variation Cv = 5.2 / 38.2 = 0.136; skewness S = 1.8; 3rd harmonic content H3 = 8.2%; 5th harmonic content H5 = 3.1%; 7th harmonic content H7 = 1.5%. After standardizing the six operational state vectors, they were weighted according to their respective weights and then concatenated to obtain a fused state vector [36.2, 0.136, 1.8, 8.2%, 3.1%, 1.5%]. The fused state vector was input into the decision function, and the decision function output f(x) = 2.3, which is greater than the score threshold of 0.8. Therefore, a latent fault was identified (confidence level 99.2%).

[0113] Optionally, after each latent fault detection result is obtained, relevant fault information can be stored, including:

[0114] 1) Original waveform (transient waveform recording data): Compressed by algorithm (compression rate 58%) and stored as a binary file;

[0115] 2) Fusion state vector: [36.2, 0.136, 1.8, 8.2%, 3.1%, 1.5%];

[0116] 3) Classification confidence: d=2.3.

[0117] In terms of storage strategy, the aforementioned fault information can be retained at the edge for a period of time and then permanently saved to the cloud.

[0118] Based on the above embodiments, in an exemplary embodiment, such as Figure 8 As shown in the embodiments of this application, the method for detecting latent faults in a power distribution network includes:

[0119] S801, obtain multiple sample state vectors; construct a classification hyperplane based on the multiple sample state vectors; determine each reference state vector based on the sample state vectors located on both sides of the classification hyperplane whose distance from the classification hyperplane is less than a distance threshold.

[0120] S802, acquire transient waveform data of the distribution network.

[0121] S803 extracts at least two operating state vectors of the distribution network from transient waveform data.

[0122] S804, determine the mean and standard deviation of the running state values ​​of at least two running state vectors.

[0123] S805: For each operating state vector, determine the difference between the operating state value and the mean of the operating state vector, and determine the standard state vector of the operating state vector based on the ratio of the difference to the standard deviation.

[0124] S806 fuses the standard state vectors according to the importance of different operating state vectors to obtain a fused state vector.

[0125] S807: For each reference state vector, a distance weight is used to weight the distance between the reference state vector and the fused state vector to obtain a first weighted value; the vector weight corresponding to the reference state vector is used to weight the reciprocal of the exponential function value of the first weighted value to obtain a second weighted value; the fault score of the fused state vector under the reference state vector is determined based on the sum of the second weighted value and the preset bias.

[0126] S808 determines the results of latent fault detection in the distribution network based on the sum of the fault scores.

[0127] For specific implementation methods of S801-S808, please refer to the above embodiments, which will not be repeated here.

[0128] It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to 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 above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0129] Based on the same inventive concept, this application also provides a latent fault detection device for a distribution network to implement the aforementioned method for detecting latent faults in a distribution network. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the latent fault detection device for a distribution network provided below can be found in the limitations of the latent fault detection method for a distribution network described above, and will not be repeated here.

[0130] In one exemplary embodiment, such as Figure 9 As shown, a latent fault detection device for a power distribution network is provided, comprising: a data acquisition module 901, a vector extraction module 902, a vector fusion module 903, and a fault detection module 904, wherein:

[0131] The data acquisition module 901 is used to acquire transient waveform data of the distribution network; wherein, the transient waveform data is obtained by collecting the electrical quantities of the distribution network when abnormal fluctuations are detected in the electrical quantities of the distribution network.

[0132] The vector extraction module 902 is used to extract at least two operating state vectors of the distribution network from the transient waveform data;

[0133] The vector fusion module 903 is used to fuse at least two operating state vectors according to the importance of different operating state vectors to the stable operation of the distribution network, so as to obtain a fused state vector;

[0134] The fault detection module 904 is used to determine the latent fault detection result of the distribution network based on the distance between the fused state vector and each reference state vector; wherein each reference state vector contains the state vector obtained when a latent fault occurs in the distribution network.

[0135] In one exemplary embodiment, the fault detection module 904 includes:

[0136] The distance determination unit is used to determine the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector for each reference state vector.

[0137] The result determination unit is used to determine the latent fault detection results of the distribution network based on the sum of the fault scores.

[0138] In one exemplary embodiment, the distance determination unit is specifically used for:

[0139] Distance weights are used to weight the distance between the reference state vector and the fused state vector to obtain a first weighted value. The vector weights corresponding to the reference state vector are used to weight the reciprocal of the exponential function value of the first weighted value to obtain a second weighted value. The fault score of the fused state vector under the reference state vector is determined based on the sum of the second weighted value and the preset bias.

[0140] In an exemplary embodiment, each reference state vector corresponds to a fault label, which is used to characterize whether the corresponding reference state vector was obtained under the condition of a latent fault in the distribution network; the fault detection module 904 is specifically used for:

[0141] Obtain multiple sample state vectors; construct a classification hyperplane based on the multiple sample state vectors; wherein, sample state vectors corresponding to the same fault label among the multiple sample state vectors are located on the same side of the classification hyperplane; determine the reference state vector based on the sample state vectors located on both sides of the classification hyperplane whose distance from the classification hyperplane is less than a distance threshold.

[0142] In one exemplary embodiment, the vector fusion module 903 includes:

[0143] The processing unit is used to standardize at least two running state vectors to obtain a standard state vector corresponding to each running state vector.

[0144] The fusion unit is used to fuse the standard state vectors according to the importance of different operating state vectors to obtain a fused state vector.

[0145] In an exemplary embodiment, each running state vector includes a running state value; the processing unit is specifically used for:

[0146] Determine the mean and standard deviation of the operating state values ​​for at least two operating state vectors; for each operating state vector, determine the difference between the operating state value and the mean of the operating state vector, and determine the standard state vector of the operating state vector based on the ratio of the difference to the standard deviation.

[0147] Each module in the aforementioned latent fault detection device for the power distribution network can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0148] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores transient waveform data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for detecting latent faults in a power distribution network.

[0149] Those skilled in the art will understand that Figure 10 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0150] In one exemplary embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to: acquire transient waveform data of a distribution network; wherein the transient waveform data is obtained by collecting electrical quantities of the distribution network when abnormal fluctuations in electrical quantities of the distribution network are detected; extract at least two operating state vectors of the distribution network from the transient waveform data; fuse the at least two operating state vectors according to the importance of different operating state vectors to the stable operation of the distribution network to obtain a fused state vector; determine the latent fault detection result of the distribution network according to the distance between the fused state vector and each reference state vector; wherein each reference state vector contains a state vector acquired when a latent fault occurs in the distribution network.

[0151] In one embodiment, the latent fault detection result of the distribution network is determined based on the distance between the fused state vector and each reference state vector, including: for each reference state vector, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector; and determining the latent fault detection result of the distribution network based on the sum of the fault scores.

[0152] In one embodiment, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector includes: using distance weights to weight the distance between the reference state vector and the fused state vector to obtain a first weighted value; using the vector weights corresponding to the reference state vector to weight the reciprocal of the exponential function value of the first weighted value to obtain a second weighted value; and determining the fault score of the fused state vector under the reference state vector based on the sum of the second weighted value and a preset bias.

[0153] In one embodiment, each reference state vector corresponds to a fault label, which is used to characterize whether the corresponding reference state vector was obtained under the condition of a latent fault in the distribution network. Each reference state vector is determined as follows: multiple sample state vectors are obtained; a classification hyperplane is constructed based on the multiple sample state vectors; wherein the sample state vectors corresponding to the same fault label among the multiple sample state vectors are located on the same side of the classification hyperplane; and the reference state vector is determined based on the sample state vectors located on both sides of the classification hyperplane whose distance from the classification hyperplane is less than a distance threshold.

[0154] In one embodiment, at least two operating state vectors are fused to obtain a fused state vector based on the importance of different operating state vectors to the stable operation of the distribution network. This includes: standardizing at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector; and fusing the standard state vectors according to the importance of different operating state vectors to obtain a fused state vector.

[0155] In one embodiment, each operating state vector includes an operating state value; standardizing at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector includes: determining the mean and standard deviation of the operating state values ​​of at least two operating state vectors; for each operating state vector, determining the difference between the operating state value and the mean of the operating state vector, and determining the standard state vector of the operating state vector based on the ratio of the difference to the standard deviation.

[0156] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When executed by a processor, the computer program performs the following: acquiring transient waveform data of a distribution network; wherein the transient waveform data is obtained by collecting electrical quantities of the distribution network when abnormal fluctuations in electrical quantities of the distribution network are detected; extracting at least two operating state vectors of the distribution network from the transient waveform data; fusing the at least two operating state vectors according to the importance of different operating state vectors to the stable operation of the distribution network to obtain a fused state vector; and determining the latent fault detection result of the distribution network according to the distance between the fused state vector and each reference state vector; wherein each reference state vector contains a state vector acquired when a latent fault occurs in the distribution network.

[0157] In one embodiment, the latent fault detection result of the distribution network is determined based on the distance between the fused state vector and each reference state vector, including: for each reference state vector, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector; and determining the latent fault detection result of the distribution network based on the sum of the fault scores.

[0158] In one embodiment, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector includes: using distance weights to weight the distance between the reference state vector and the fused state vector to obtain a first weighted value; using the vector weights corresponding to the reference state vector to weight the reciprocal of the exponential function value of the first weighted value to obtain a second weighted value; and determining the fault score of the fused state vector under the reference state vector based on the sum of the second weighted value and a preset bias.

[0159] In one embodiment, each reference state vector corresponds to a fault label, which is used to characterize whether the corresponding reference state vector was obtained under the condition of a latent fault in the distribution network. Each reference state vector is determined as follows: multiple sample state vectors are obtained; a classification hyperplane is constructed based on the multiple sample state vectors; wherein the sample state vectors corresponding to the same fault label among the multiple sample state vectors are located on the same side of the classification hyperplane; and the reference state vector is determined based on the sample state vectors located on both sides of the classification hyperplane whose distance from the classification hyperplane is less than a distance threshold.

[0160] In one embodiment, at least two operating state vectors are fused to obtain a fused state vector based on the importance of different operating state vectors to the stable operation of the distribution network. This includes: standardizing at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector; and fusing the standard state vectors according to the importance of different operating state vectors to obtain a fused state vector.

[0161] In one embodiment, each operating state vector includes an operating state value; standardizing at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector includes: determining the mean and standard deviation of the operating state values ​​of at least two operating state vectors; for each operating state vector, determining the difference between the operating state value and the mean of the operating state vector, and determining the standard state vector of the operating state vector based on the ratio of the difference to the standard deviation.

[0162] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following: acquiring transient waveform data of a distribution network; wherein the transient waveform data is obtained by collecting electrical quantities of the distribution network when abnormal fluctuations in electrical quantities of the distribution network are detected; extracting at least two operating state vectors of the distribution network from the transient waveform data; fusing the at least two operating state vectors according to the importance of different operating state vectors to the stable operation of the distribution network to obtain a fused state vector; and determining the latent fault detection result of the distribution network based on the distance between the fused state vector and each reference state vector; wherein each reference state vector contains a state vector acquired when a latent fault occurs in the distribution network.

[0163] In one embodiment, the latent fault detection result of the distribution network is determined based on the distance between the fused state vector and each reference state vector, including: for each reference state vector, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector; and determining the latent fault detection result of the distribution network based on the sum of the fault scores.

[0164] In one embodiment, determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector includes: using distance weights to weight the distance between the reference state vector and the fused state vector to obtain a first weighted value; using the vector weights corresponding to the reference state vector to weight the reciprocal of the exponential function value of the first weighted value to obtain a second weighted value; and determining the fault score of the fused state vector under the reference state vector based on the sum of the second weighted value and a preset bias.

[0165] In one embodiment, each reference state vector corresponds to a fault label, which is used to characterize whether the corresponding reference state vector was obtained under the condition of a latent fault in the distribution network. Each reference state vector is determined as follows: multiple sample state vectors are obtained; a classification hyperplane is constructed based on the multiple sample state vectors; wherein the sample state vectors corresponding to the same fault label among the multiple sample state vectors are located on the same side of the classification hyperplane; and the reference state vector is determined based on the sample state vectors located on both sides of the classification hyperplane whose distance from the classification hyperplane is less than a distance threshold.

[0166] In one embodiment, at least two operating state vectors are fused to obtain a fused state vector based on the importance of different operating state vectors to the stable operation of the distribution network. This includes: standardizing at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector; and fusing the standard state vectors according to the importance of different operating state vectors to obtain a fused state vector.

[0167] In one embodiment, each operating state vector includes an operating state value; standardizing at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector includes: determining the mean and standard deviation of the operating state values ​​of at least two operating state vectors; for each operating state vector, determining the difference between the operating state value and the mean of the operating state vector, and determining the standard state vector of the operating state vector based on the ratio of the difference to the standard deviation.

[0168] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0169] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0170] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting latent faults in a power distribution network, characterized in that, The method includes: Acquire transient waveform data of the distribution network; wherein the transient waveform data is obtained by collecting electrical quantities of the distribution network when abnormal fluctuations in electrical quantities of the distribution network are detected; Extract at least two operating state vectors of the distribution network from the transient waveform data; Based on the importance of different operating state vectors to the stable operation of the distribution network, the at least two operating state vectors are fused to obtain a fused state vector; For each reference state vector, the fault score of the fused state vector under the reference state vector is determined based on the distance between the reference state vector and the fused state vector; the latent fault detection result of the distribution network is determined based on the sum of the fault scores; wherein each reference state vector contains a state vector obtained when a latent fault occurs in the distribution network.

2. The method according to claim 1, characterized in that, The step of determining the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector includes: The distance between the reference state vector and the fused state vector is weighted using distance weighting to obtain a first weighted value; Using the vector weights corresponding to the reference state vector, the reciprocal of the exponential function value of the first weighted value is weighted to obtain the second weighted value; The fault score of the fused state vector under the reference state vector is determined based on the sum of the second weighted value and the preset bias.

3. The method according to any one of claims 1 to 2, characterized in that, Each reference state vector corresponds to a fault label, which is used to characterize whether the corresponding reference state vector was obtained under the condition that a latent fault occurred in the distribution network; each reference state vector is determined in the following manner: Obtain multiple sample state vectors; A classification hyperplane is constructed based on the plurality of sample state vectors; wherein, the sample state vectors corresponding to the same fault label among the plurality of sample state vectors are located on the same side of the classification hyperplane; A reference state vector is determined based on the sample state vectors located on both sides of the classification hyperplane that have a distance less than a distance threshold from the classification hyperplane.

4. The method according to any one of claims 1 to 2, characterized in that, The step of fusing at least two operating state vectors based on their relative importance to the stable operation of the distribution network to obtain a fused state vector includes: The at least two running state vectors are standardized to obtain a standard state vector corresponding to each running state vector. Based on the importance of the different operating state vectors, the standard state vectors are fused to obtain a fused state vector.

5. The method according to claim 4, characterized in that, Each of the aforementioned operating state vectors includes an operating state value; the standardization process for the at least two operating state vectors to obtain a standard state vector corresponding to each operating state vector includes: Determine the mean and standard deviation of the operating state values ​​of the at least two operating state vectors; For each of the aforementioned operating state vectors, the difference between the operating state value of the operating state vector and the mean is determined, and the standard state vector of the operating state vector is determined based on the ratio of the difference to the standard deviation.

6. The method according to claim 4, characterized in that, The step of fusing the standard state vectors according to their importance to obtain a fused state vector includes: The weight of each standard state vector is determined based on the importance of the different operating state vectors. Based on the weight of each of the standard state vectors, the standard state vectors are weighted and fused to obtain a fused state vector.

7. A latent fault detection device for a power distribution network, characterized in that, The device includes: The data acquisition module is used to acquire transient waveform data of the distribution network; wherein, the transient waveform data is obtained by collecting the electrical quantities of the distribution network when abnormal fluctuations in the electrical quantities of the distribution network are detected. The vector extraction module is used to extract at least two operating state vectors of the power distribution network from the transient waveform data. The vector fusion module is used to fuse at least two operating state vectors according to the importance of different operating state vectors to the stable operation of the distribution network, so as to obtain a fused state vector. The fault detection module is used to determine the fault score of the fused state vector under the reference state vector based on the distance between the reference state vector and the fused state vector for each reference state vector; and to determine the latent fault detection result of the distribution network based on the sum of the fault scores; wherein each reference state vector contains a state vector obtained when a latent fault occurs in the distribution network.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images