Power distribution network line single-phase earth fault section selection method and device based on multiple characteristic quantities, computer equipment, storage medium and computer program product

Abstract

The invention relates to a power distribution network line single-phase earth fault section selection method and device based on multiple characteristic quantities, computer equipment, a storage medium and a computer program product. The method comprises the steps of collecting zero-sequence current waveform data of each detection node in a first half cycle under the condition that a single-phase earth fault occurs in a power distribution network line; obtaining a target fault feature set and a fault section selection model for the power distribution network line; according to the target fault feature set, performing feature extraction on the zero-sequence current waveform data corresponding to the detection nodes at the two ends of any section to obtain a fault feature quantity of any section under each target fault feature; inputting the fault characteristic quantity of any section under each target fault characteristic into a fault section selection model to obtain a section type of any section; and determining a fault section in each section according to the section type of each section. By adopting the method, accurate positioning can be performed when the power distribution network has a fault.

Description

Technical Field

[0001] The present application relates to the technical field of power systems, and in particular, to a method, device, computer device, storage medium, and computer program product for single-phase grounding fault section selection of a distribution network line based on multiple characteristic quantities. Background Art

[0002] The distribution network is an important link in the power system that directly faces users, and its safe and stable operation is crucial for ensuring energy security. The long-term operation of the power system shows that most power outages are caused by distribution network faults, and most of these faults are single-phase grounding faults.

[0003] Currently, resonant grounding is widely used in medium and low voltage distribution networks, and when a single-phase grounding fault occurs, it can continue to operate for 1 to 2 hours. However, since the input of the arc suppression coil suppresses the fault current, the fault characteristics are weak, making it difficult to perform accurate fault location.

[0004] Therefore, there is a problem in the traditional technology that the fault location of the distribution network cannot be accurately performed. Summary of the Invention

[0005] Based on this, in view of the above technical problems, it is necessary to provide a method, device, computer device, computer-readable storage medium, and computer program product for single-phase grounding fault section selection of a distribution network line based on multiple characteristic quantities that can perform accurate location when a fault occurs in the distribution network.

[0006] A method for single-phase grounding fault section selection of a distribution network line based on multiple characteristic quantities is applied to a distribution network line. The distribution network line includes multiple sections, and detection nodes are provided at both ends of any section. The method includes:

[0007] In the case of a single-phase grounding fault occurring in the distribution network line, collecting zero-sequence current waveform data of each detection node in the first half cycle;

[0008] Obtaining a target fault feature set and a fault section selection model for the distribution network line; the target fault features in the target fault feature set are selected from each candidate fault feature; the fault section selection model is obtained by training a support vector machine model with a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under different fault conditions under each candidate fault feature;

[0009] According to the target fault feature set, extracting features from the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section to obtain the fault feature quantities of any section under each target fault feature;

[0010] Inputting the fault feature quantities of any section under each target fault feature into the fault section selection model to obtain the section type of any section;

[0011] Determine the faulty section in each section according to the section type of each section.

[0012] In one embodiment, obtaining a target fault feature set and a fault section selection model for a distribution network line includes:

[0013] Obtain a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under different fault conditions for each candidate fault feature.

[0014] Use a genetic algorithm to train a support vector machine model based on the training sample set to obtain an optimal fault feature set and an optimal hyperparameter value set.

[0015] Based on the optimal fault feature set, determine the target fault feature set, and based on the optimal hyperparameter value set, construct a fault section selection model.

[0016] In one embodiment, using a genetic algorithm to train a support vector machine model based on the training sample set to obtain an optimal fault feature set and an optimal hyperparameter value set includes:

[0017] Based on the value ranges of the hyperparameters of the support vector machine model, randomly generate multiple chromosomes as the initial population; any chromosome corresponds to a set of hyperparameter values; any set of hyperparameter values includes a regularization parameter and a kernel width parameter.

[0018] Obtain the individual fitness values and fault feature screening results corresponding to each chromosome in the initial population, and perform population evolution based on the individual fitness values corresponding to each chromosome to generate multiple new chromosomes as the new population; the individual fitness value corresponding to any chromosome represents the quality of the chromosome in the initial population; the fault feature screening result corresponding to any chromosome is determined after feature screening of each candidate fault feature during the process of training the model using the hyperparameter value set corresponding to the chromosome.

[0019] When the current iteration number is less than the maximum iteration number, use the new population as the initial population, update the current iteration number, and return to obtain the individual fitness values and fault feature screening results corresponding to each chromosome in the initial population until the current iteration number is equal to the maximum iteration number.

[0020] When the current iteration number is equal to the maximum iteration number, use the chromosome with the highest individual fitness value in the new population as the optimal chromosome, determine the hyperparameter value set corresponding to the optimal chromosome as the optimal hyperparameter value set, and determine the fault feature screening result corresponding to the optimal chromosome as the optimal fault feature set.

[0021] In one embodiment, obtaining the individual fitness values and fault feature screening results corresponding to each chromosome in the initial population includes:

[0022] Decode each chromosome in the initial population to obtain the set of hyperparameter values corresponding to each chromosome;

[0023] Based on the sets of hyperparameter values corresponding to different chromosomes, construct support vector machine models corresponding to different sets of hyperparameter values;

[0024] For the set of hyperparameter values corresponding to any chromosome, use the training sample set to train the support vector machine model corresponding to the set of hyperparameter values, and based on the model training results, determine the fault feature screening result and the individual fitness value corresponding to the chromosome.

[0025] In one embodiment, for the set of hyperparameter values corresponding to any chromosome, use the training sample set to train the support vector machine model corresponding to the set of hyperparameter values, and based on the model training results, determine the fault feature screening result and the individual fitness value corresponding to the chromosome, including:

[0026] For the set of hyperparameter values corresponding to any chromosome, use the training sample set to train the support vector machine model corresponding to the set of hyperparameter values to obtain the feature importance of each candidate fault feature;

[0027] Adopt the sequential forward floating search method to perform feature screening based on the feature importance of each candidate fault feature to obtain the fault feature screening result corresponding to the chromosome;

[0028] Based on the fault feature screening result corresponding to the chromosome, determine the individual fitness value corresponding to the chromosome.

[0029] In one embodiment, the candidate fault features include the cosine similarity, Pearson correlation coefficient, Spearman correlation coefficient, dynamic time warping distance, grey relational degree, relative entropy of energy, absolute distance, Euclidean distance, Hausdorff distance, Bhattacharyya coefficient, peak difference degree, and peak-to-peak difference degree between the zero-sequence current waveforms at both ends of the section in the first half cycle after a single-phase grounding fault occurs.

[0030] A single-phase grounding fault section selection device for a distribution network line based on multiple feature quantities, which is applied to a distribution network line. The distribution network line includes multiple sections, and detection nodes are arranged at both ends of any section. The device includes:

[0031] An acquisition module, configured to collect the zero-sequence current waveform data of each detection node in the first half cycle when a single-phase grounding fault occurs in the distribution network line;

[0032] An acquisition module, configured to acquire a target fault feature set and a fault section selection model for a distribution network line; the target fault features in the target fault feature set are screened from each candidate fault feature; the fault section selection model is obtained by training a support vector machine model with a training sample set; the training sample set includes fault feature quantities of each section of the distribution network line under each candidate fault feature under different fault conditions.

[0033] An extraction module, configured to extract features from the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section according to the target fault feature set, and obtain the fault feature quantities of any section under each target fault feature.

[0034] An input module, configured to input the fault feature quantities of any section under each target fault feature into the fault section selection model to obtain the section type of any section.

[0035] A determination module, configured to determine the fault section among each section according to the section types of each section.

[0036] A computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0037] A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0038] A computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0039] The above-mentioned single-phase grounding fault section selection method, device, computer equipment, storage medium and computer program product for distribution network lines based on multiple feature quantities collect the zero-sequence current waveform data of each detection node in the first half cycle when a single-phase grounding fault occurs in the distribution network line; obtain the target fault feature set and fault section selection model for the distribution network line; the target fault features in the target fault feature set are screened from each candidate fault feature; the fault section selection model is obtained by training a support vector machine model with a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under each candidate fault feature under different fault conditions; according to the target fault feature set, feature extraction is performed on the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section to obtain the fault feature quantities of any section under each target fault feature; input the fault feature quantities of any section under each target fault feature into the fault section selection model to obtain the section type of any section; according to the section types of each section, determine the fault section in each section; in this way, the fault features of each section under each target fault feature can be quickly extracted, and the fault section can be quickly and accurately selected in the distribution network line based on the fault features of each section under each target fault feature by the fault section selection model. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is an application environment diagram of a single-phase grounding fault section selection method for distribution network lines based on multiple feature quantities in an embodiment;

[0042] Figure 2 It is a flowchart of a single-phase grounding fault section selection method for distribution network lines based on multiple feature quantities in an embodiment;

[0043] Figure 3 It is a schematic diagram of a 10kV resonant grounding distribution network simulation model in an embodiment;

[0044] Figure 4 It is a schematic diagram of the implementation framework of a feature subset optimization method in an embodiment;

[0045] Figure 5 It is a flowchart of a single-phase grounding fault section selection in an embodiment;

[0046] Figure 6Schematic flowchart of a method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities in another embodiment;

[0047] Figure 7 Structural block diagram of a device for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities in an embodiment;

[0048] Figure 8 Internal structure diagram of a computer device in an embodiment. Detailed implementation manners

[0049] In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0050] The method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities provided by the embodiments of the present application can be applied to an application environment as shown in Figure 1 In the figure. Among them, the terminal 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed in the cloud or other network servers. When a single-phase grounding fault occurs in the distribution network line, the server 104 collects the zero-sequence current waveform data of each detection node in the first half cycle; the server 104 obtains a target fault feature set and a fault section selection model for the distribution network line; the target fault features in the target fault feature set are screened from each candidate fault feature; the fault section selection model is obtained by training a support vector machine model with a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under different fault conditions under each candidate fault feature; the server 104 extracts features from the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section according to the target fault feature set, and obtains the fault feature quantities of any section under each target fault feature; the server 104 inputs the fault feature quantities of any section under each target fault feature into the fault section selection model to obtain the section type of any section; the server 104 determines the fault section among each section according to the section types of each section. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart vehicle-mounted devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented by an independent server or a server cluster composed of multiple servers.

[0051] In an exemplary embodiment, as Figure 2 shown, a single-phase grounding fault section selection method for a distribution network line based on multiple feature quantities is provided. Taking the method applied to the Figure 1 server 104 in

[0052] as an example, the following steps S202 to S210 are included. Among them:

[0053] The distribution network line includes multiple sections, and detection nodes are arranged at both ends of each section.

[0054] The zero-sequence current waveform data in the first half cycle can refer to the zero-sequence current waveform signal in the first half cycle when a single-phase grounding fault occurs.

[0055] Optionally, when a single-phase grounding fault occurs in the distribution network line, the server collects the zero-sequence current waveform data of each detection node in the first half cycle.

[0056] Step S204, obtain a target fault feature set and a fault section selection model for the distribution network line; the target fault features in the target fault feature set are screened from each candidate fault feature; the fault section selection model is obtained by training a support vector machine model with a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under each candidate fault feature in different fault conditions.

[0057] Among them, the target fault feature set can be a subset of the candidate fault feature set. The candidate fault features can characterize the difference in the zero-sequence current waveforms at both ends of the section. By using a feature subset optimization method to screen the candidate fault feature set, the target fault feature set can be obtained, and the feature importance of the target fault features meets the preset requirements.

[0058] Among them, the fault section selection model can refer to a model used to screen out the fault section. In practical applications, the fault section selection model can be a pre-trained support vector machine model (Support Vector Machine, SVM).

[0059] Optionally, the server obtains a target fault feature set and a fault section selection model for the distribution network line.

[0060] Step S206, according to the target fault feature set, extract features from the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section, and obtain the fault feature quantities of any section under each target fault feature.

[0061] Among them, the fault feature quantity of any section under each target fault feature may refer to the feature value of any section under each target fault feature. When the target fault feature is the cosine similarity, the fault feature quantity of section A under the target fault feature of cosine similarity is the cosine similarity value of the zero-sequence current waveforms of the detection nodes at both ends of section A.

[0062] Optionally, the server extracts features from the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section according to the target fault feature set, and obtains the fault feature quantity of any section under each target fault feature.

[0063] Step S208: Input the fault feature quantity of any section under each target fault feature into the fault section selection model to obtain the section type of any section.

[0064] Among them, the section type may include two types: a sound section and an abnormal section.

[0065] Optionally, the server inputs the fault feature quantity of any section under each target fault feature into the fault section selection model to obtain the section type of any section.

[0066] Step S210: Determine the fault section among each section according to the section type of each section.

[0067] Among them, the fault section may refer to the section where a single-phase grounding fault occurs.

[0068] Optionally, the server determines the fault section among each section according to the section type of each section.

[0069] In the above method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities, when a single-phase grounding fault occurs in the distribution network line, the zero-sequence current waveform data of each detection node in the first half cycle is collected; the target fault feature set and the fault section selection model for the distribution network line are obtained; the target fault features in the target fault feature set are screened from each candidate fault feature; the feature importance of the target fault feature meets the preset requirements; according to the target fault feature set, features are extracted from the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section to obtain the fault feature quantity of any section under each target fault feature; the fault feature quantity of any section under each target fault feature is input into the fault section selection model to obtain the section type of any section; according to the section type of each section, the fault section is determined among each section; in this way, the fault features of each section under each target fault feature can be quickly extracted, and based on the fault features of each section under each target fault feature by the fault section selection model, the fault section can be quickly and accurately selected in the distribution network line.

[0070] In an exemplary embodiment, obtaining a target fault feature set and a fault section selection model for a distribution network line includes: obtaining a training sample set; the training sample set includes fault feature quantities of each section of the distribution network line under each candidate fault feature in different fault conditions; using a genetic algorithm to train a support vector machine model based on the training sample set to obtain an optimal fault feature set and an optimal hyperparameter value set; determining the target fault feature set based on the optimal fault feature set, and constructing a fault section selection model based on the optimal hyperparameter value set.

[0071] Among them, the training sample set includes fault feature quantities of each section of the distribution network line under each candidate fault feature in different fault conditions. The training sample set is obtained by extracting features from a simulation data set, and the simulation data set includes zero-sequence current waveform data at both ends of the section under various fault conditions.

[0072] In practical applications, based on Figure 3 the shown simulation model, various single-phase grounding faults are simulated by modifying fault parameters such as the fault occurrence location, fault initial phase angle, and fault grounding resistance, and the zero-sequence current waveform data in the first half cycle after the fault at both ends of each section are collected, thereby forming a simulation data set. After actual simulation in this application, 720 groups of samples are selected from the simulation data of 80 single-phase grounding faults in the distribution network, including 360 groups of training set samples and 360 groups of test set samples.

[0073] Among them, the genetic algorithm may refer to the GA (Genetic Algorithm) algorithm.

[0074] Among them, the support vector machine model may correspond to the fault section selection model.

[0075] Among them, the optimal fault feature set may be the fault feature set corresponding to the minimum recognition error.

[0076] Among them, the optimal hyperparameter value set may be the hyperparameter value set corresponding to the minimum recognition error.

[0077] Among them, the target fault feature set may be the fault feature set adopted by the final fault section selection model.

[0078] Optionally, the server obtains the training sample set and uses a genetic algorithm to train the support vector machine model based on the training sample set to obtain an optimal fault feature set and an optimal hyperparameter value set. The server determines the target fault feature set based on the optimal fault feature set, and the server constructs a fault section selection model based on the optimal hyperparameter value set.

[0079] In this embodiment, a training sample set is obtained. The training sample set includes the fault feature quantities of each section of the distribution network line under different fault conditions for each candidate fault feature. The genetic algorithm is used to train the support vector machine model based on the training sample set to obtain the optimal fault feature set and the optimal hyperparameter value set. Based on the optimal fault feature set, the target fault feature set is determined, and based on the optimal hyperparameter value set, a fault section selection model is constructed. In this way, the genetic algorithm can be used to optimize the hyperparameters of the support vector machine, and at the same time, the optimal fault feature set and the optimal hyperparameter value set are determined during this process, avoiding the cumulative errors in multiple links such as feature selection, parameter optimization, and model training, which is beneficial to improving the section selection accuracy of the fault section selection model.

[0080] In an exemplary embodiment, using the genetic algorithm to train the support vector machine model based on the training sample set to obtain the optimal fault feature set and the optimal hyperparameter value set includes: randomly generating a plurality of chromosomes as the initial population based on the value ranges of the hyperparameters of the support vector machine model; any chromosome corresponds to a set of hyperparameter values; any set of hyperparameter values includes a regularization parameter and a kernel width parameter; obtaining the individual fitness values and the fault feature screening results corresponding to each chromosome in the initial population, and evolving the population based on the individual fitness values corresponding to each chromosome to generate a plurality of new chromosomes as the new population; the individual fitness value corresponding to any chromosome represents the quality of the chromosome in the initial population; the fault feature screening result corresponding to any chromosome is determined after feature screening of each candidate fault feature during the process of model training using the hyperparameter value set corresponding to the chromosome. When the current iteration number is less than the maximum iteration number, the new population is used as the initial population, the current iteration number is updated, and the individual fitness values and the fault feature screening results corresponding to each chromosome in the initial population are obtained and returned until the current iteration number is equal to the maximum iteration number. When the current iteration number is equal to the maximum iteration number, the chromosome with the highest individual fitness value in the new population is used as the optimal chromosome, the hyperparameter value set corresponding to the optimal chromosome is determined as the optimal hyperparameter value set, and the fault feature screening result corresponding to the optimal chromosome is determined as the optimal fault feature set.

[0081] Among them, the hyperparameters of the support vector machine model may include the regularization parameter c and the kernel width parameter γ of the SVM model.

[0082] Among them, the value ranges of the hyperparameters of the support vector machine model may refer to the search spaces of the regularization parameter c and the kernel width parameter γ of the SVM model.

[0083] Among them, the set of hyperparameter values may refer to the parameter value set formed by a regularization parameter c and a kernel width parameter γ.

[0084] Among them, the individual fitness values corresponding to each chromosome can characterize the performance of each chromosome in the initial population.

[0085] Among them, the fault feature screening result corresponding to any chromosome can be determined by constructing an SVM model using the hyperparameter value set corresponding to the chromosome and screening each candidate fault feature during the process of training the SVM model.

[0086] Optionally, the server randomly generates multiple chromosomes as the initial population based on the value ranges of the hyperparameters of the support vector machine model; then, the server obtains the individual fitness values and fault feature screening results corresponding to each chromosome in the initial population, and evolves the population based on the individual fitness values corresponding to each chromosome to generate multiple new chromosomes as the new population; then, the server determines whether the current iteration number is equal to the maximum iteration number: if the current iteration number is less than the maximum iteration number, the server uses the new population as the initial population, increments the current iteration number by one, and returns to obtain the individual fitness values and fault feature screening results corresponding to each chromosome in the initial population until the current iteration number is equal to the maximum iteration number; while if the current iteration number is equal to the maximum iteration number, the server uses the chromosome with the highest individual fitness value in the new population as the optimal chromosome, determines the hyperparameter value set corresponding to the optimal chromosome as the optimal hyperparameter value set, and determines the fault feature screening result corresponding to the optimal chromosome as the optimal fault feature set.

[0087] In this embodiment, multiple chromosomes are randomly generated as the initial population based on the value ranges of the hyperparameters of the support vector machine model; any chromosome corresponds to a set of hyperparameter values; any set of hyperparameter values includes a regularization parameter and a kernel width parameter; the individual fitness values and the fault feature screening results corresponding to each chromosome in the initial population are obtained, and population evolution is performed based on the individual fitness values corresponding to each chromosome to generate multiple new chromosomes as the new population; the individual fitness value corresponding to any chromosome represents the quality of the chromosome in the initial population; the fault feature screening result corresponding to any chromosome is determined after feature screening of each candidate fault feature during the process of model training using the hyperparameter value set corresponding to the chromosome; when the current iteration number is less than the maximum iteration number, the new population is used as the initial population, the current iteration number is updated, and the process returns to obtaining the individual fitness values and the fault feature screening results corresponding to each chromosome in the initial population until the current iteration number is equal to the maximum iteration number; when the current iteration number is equal to the maximum iteration number, the chromosome with the highest individual fitness value in the new population is used as the optimal chromosome, the hyperparameter value set corresponding to the optimal chromosome is determined as the optimal hyperparameter value set, and the fault feature screening result corresponding to the optimal chromosome is determined as the optimal fault feature set; in this way, based on the interaction mechanism between the genetic algorithm and the feature selection process, the screening of fault features and the optimization of hyperparameter values can be achieved simultaneously.

[0088] In an exemplary embodiment, obtaining the individual fitness values and the fault feature screening results corresponding to each chromosome in the initial population includes: decoding each chromosome in the initial population to obtain the hyperparameter value set corresponding to each chromosome; constructing support vector machine models corresponding to different hyperparameter value sets based on the hyperparameter value sets corresponding to different chromosomes; for the hyperparameter value set corresponding to any chromosome, training the support vector machine model corresponding to the hyperparameter value set using the training sample set, and determining the fault feature screening result and the individual fitness value corresponding to the chromosome based on the model training result.

[0089] Optionally, the server decodes each chromosome in the initial population to obtain the hyperparameter value set corresponding to each chromosome, the server constructs support vector machine models corresponding to different hyperparameter value sets based on the hyperparameter value sets corresponding to different chromosomes, and for the hyperparameter value set corresponding to any chromosome, the server trains the support vector machine model corresponding to the hyperparameter value set using the training sample set, and determines the fault feature screening result and the individual fitness value corresponding to the chromosome based on the model training result.

[0090] In this embodiment, by decoding each chromosome in the initial population, a set of hyperparameter values corresponding to each chromosome is obtained; based on the sets of hyperparameter values corresponding to different chromosomes, support vector machine models corresponding to different sets of hyperparameter values are constructed; for the set of hyperparameter values corresponding to any chromosome, a training sample set is used to train the support vector machine model corresponding to the set of hyperparameter values, and based on the model training result, the fault feature screening result and the individual fitness value corresponding to the chromosome are determined; in this way, each chromosome can be decoded into each set of hyperparameter values, which is beneficial to subsequent genetic algorithm operations, so as to further train different fault selection models based on different sets of hyperparameter values and evaluate the performance of different fault selection models.

[0091] In an exemplary embodiment, for the set of hyperparameter values corresponding to any chromosome, a training sample set is used to train the support vector machine model corresponding to the set of hyperparameter values, and based on the model training result, the fault feature screening result and the individual fitness value corresponding to the chromosome are determined, including: for the set of hyperparameter values corresponding to any chromosome, a training sample set is used to train the support vector machine model corresponding to the set of hyperparameter values to obtain the feature importance of each candidate fault feature; the sequential forward floating search method is used to perform feature screening based on the feature importance of each candidate fault feature to obtain the fault feature screening result corresponding to the chromosome; based on the fault feature screening result corresponding to the chromosome, the individual fitness value corresponding to the chromosome is determined.

[0092] Among them, the feature importance of a candidate fault feature may be the degree of influence of the value candidate fault feature on fault recognition.

[0093] Optionally, for the set of hyperparameter values corresponding to any chromosome, the server uses a training sample set to train the support vector machine model corresponding to the set of hyperparameter values, determines the feature importance of each candidate fault feature by training the support vector machine model, then the server deletes the feature with the smallest importance from the candidate fault feature set and adds it to the feature ranking table, recalculates the importance of each feature in the candidate fault feature set, and deletes the next feature until the candidate fault feature set is an empty set. After feature ranking, the server uses the sequential forward floating search method to screen out the feature subset to obtain the fault feature screening result corresponding to the chromosome, and the server determines the individual fitness value corresponding to the chromosome based on the fault feature screening result corresponding to the chromosome.

[0094] In practical applications, the population chromosomes are decoded to obtain the corresponding SVM hyperparameter values. Then, combined with the above hyperparameter values, the importance R of each candidate fault feature in the candidate fault feature set is calculated by training the SVM p :

[0095] (1)

[0096] In the formula, , is the Lagrange multiplier in the SVM decision function; and , y i and y j are the training sample labels, k(x i , x j ) is the SVM kernel function, and the symbol indicates that feature p is deleted.

[0097] Then, delete the candidate fault feature with the minimum importance from the candidate fault feature set and add it to the feature ranking table. Then, recalculate the importance of each candidate fault feature in the candidate fault feature set and delete the next candidate fault feature until the candidate fault feature set becomes an empty set.

[0098] Then, after feature ranking, use the Sequential Floating Forward Selection (SFFS) method to screen the feature subset. This method starts from an empty set, adds L features in each round to make the performance of the feature subset more excellent, and then tries to remove R features to make the performance of the feature subset more stable and excellent. During the search process, evaluate the performance of the feature subset according to formula (2):

[0099] (2)

[0100] In the formula, E is the recognition error of the SVM; M is the dimension of the feature subset; N is the total number of candidate features; A and B represent the weights of the recognition error and the subset dimension respectively. The smaller the fitness, the more excellent the performance of the feature subset. Among them, the calculation formula of the recognition error E is:

[0101] (3)

[0102] In the formula, P i is the probability output value of the SVM on the validation sample; t i =(1 + y i ) / 2, y i is the validation sample label; m is the total number of validation samples.

[0103] Among them, the optimization objective of the hyperparameters is the same as that of the feature subset. Therefore, according to the fault feature screening result, the fitness value of the chromosome is calculated in combination with formula (2).

[0104] In this embodiment, for the set of hyperparameter values corresponding to any chromosome, a training sample set is used to train the support vector machine model corresponding to the set of hyperparameter values to obtain the feature importance of each candidate fault feature; the sequential forward floating search method is used to perform feature screening based on the feature importance of each candidate fault feature to obtain the fault feature screening result corresponding to the chromosome; based on the fault feature screening result corresponding to the chromosome, the individual fitness value corresponding to the chromosome is determined; in this way, the support vector machine models constructed with different sets of hyperparameter values can be trained, and during the training of each support vector machine model, feature screening and evaluation of the quality of the model are realized, improving the model training efficiency.

[0105] In an exemplary embodiment, the candidate fault features include the cosine similarity, Pearson correlation coefficient, Spearman correlation coefficient, dynamic time warping distance, grey relational degree, relative entropy of energy, absolute distance, Euclidean distance, Hausdorff distance, Bhattacharyya coefficient, peak difference degree, and peak-to-peak difference degree between the zero-sequence current waveforms in the first half cycle at both ends of the section after a single-phase ground fault occurs.

[0106] To obtain the fault feature quantities of each candidate fault feature, it is necessary to normalize the zero-sequence current waveforms i1 and i2 in the first half cycle after the fault at both ends of the section. The formula is as follows:

[0107] (4)

[0108] In the formula, is the normalized value of the nth data point of the zero-sequence current i p ; is the value of the nth data point of the zero-sequence current i p ; K is the total number of data points of the zero-sequence current. In this example, the data sampling frequency is 10 kHz, so K is 100. In practical applications, each candidate fault feature can be represented by the symbols in Table 1 below.

[0109] Table 1

[0110]

[0111] Among them, the cosine similarity f1 refers to the cosine similarity between the zero-sequence current waveforms in the first half cycle at both ends of the section after the fault occurs. The extraction method of the cosine similarity f1 is expressed as:

[0112] (5)

[0113] In the formula, the smaller the absolute value of f1, the more obvious the difference between the zero-sequence current waveforms in the first half cycle at both ends of the section after the fault occurs.

[0114] Among them, the Pearson correlation coefficient f2 can be regarded as a special cosine similarity, and the extraction method of the Pearson correlation coefficient f2 is expressed as:

[0115] (6)

[0116] In the formula, and are the means of the zero-sequence current waveforms i1 and i2 respectively.

[0117] Among them, the Spearman correlation coefficient f3 is defined as the Pearson correlation coefficient between rank variables, that is, first change the original data of the zero-sequence current waveform i1 and the zero-sequence current waveform i2 according to the position size of the descending order, and then substitute it into formula (6) to obtain the Spearman correlation coefficient value.

[0118] Among them, when calculating the dynamic time warping distance f4, take the zero-sequence current data i1 and i2 at both ends of the section and calculate the dynamic time warping distance one by one. The calculation process is as follows:

[0119] (7)

[0120] Then, calculate its per-unit value to obtain the dynamic time warping distance between i1 and i2, which can be expressed as:

[0121] (8)

[0122] Among them, the smaller f4 is, the more similar the two zero-sequence current waveforms are, and the larger it is, the greater the difference between the two waveforms.

[0123] Among them, the grey relational grade f5 refers to the grey relational grade between the zero-sequence current waveforms at both ends of the section. The larger its value, the more similar the two waveforms are. Its extraction method is expressed as:

[0124] (9)

[0125] In the formula, f5 is the grey relational grade between the zero-sequence current waveforms at both ends of the section, ξ is the grey distinguishing coefficient, and the value range can be (0, 1). Here, 0.5 is taken.

[0126] Among them, the energy relative entropy f6 refers to the energy relative entropy between the zero-sequence current waveforms at both ends of the section. The larger its value, the greater the possibility that the section type is a fault section. Its calculation process is expressed as:

[0127] First, calculate the total energy of the zero-sequence current at both ends of the section, which can be expressed as:

[0128] (10)

[0129] (11)

[0130] Then, calculate the ratio of the energy of each sampling point in the zero-sequence current waveform to the total energy. The formula is as follows:

[0131] (12)

[0132] (13)

[0133] Finally, calculate its relative entropy to obtain the relative energy entropy between i1 and i2, which can be expressed as:

[0134] (14)

[0135] Among them, the absolute distance f7 refers to the difference between waveforms. The larger the absolute distance f7, the greater the difference between waveforms. When calculating, first normalize the zero-sequence current data at both ends of the section according to Equation (4) to obtain the normalized zero-sequence current and , and then calculate the absolute distance f7. The extraction method of the absolute distance f7 is as follows:

[0136] (15)

[0137] Among them, the larger the Euclidean distance measure f8, the greater the difference between waveforms. The extraction method of the Euclidean distance is as follows:

[0138] (16)

[0139] Among them, the larger the Hausdorff distance f9, the greater the difference between waveforms. The Hausdorff distance can be calculated using the normalized zero-sequence current and and can be expressed as:

[0140] (17)

[0141] Among them, the Bhattacharyya coefficient f 10 characterizes the overlapping amount of the two waveforms. The larger its value, the larger the overlapping area of the two waveforms, indicating a higher degree of similarity between the two waveforms. The extraction method is as follows:

[0142] (18)

[0143] In the formula, S is the number of blocks, which represents that the maximum value after waveform normalization is evenly divided into S intervals; and are respectively and the ratio of the number of sampling points distributed in the kth equal division interval to the total number of data.

[0144] Among them, the peak difference degree f11 is the peak difference degree between zero-sequence current waveforms. The larger the value, the greater the difference between the two waveforms. The calculation process is as follows:

[0145] First, calculate the peaks of the zero-sequence currents at both ends of the section. The formula is as follows:

[0146] (19)

[0147] (20)

[0148] Then, calculate the peak difference degree between zero-sequence currents, which can be expressed as:

[0149] (21)

[0150] where the peak-to-peak difference degree f 12 is the peak-to-peak difference degree between zero-sequence current waveforms. The larger the value, the greater the difference between the two waveforms. The calculation process can be expressed as:

[0151] First, calculate PP1 and PP2:

[0152] (22)

[0153] (23)

[0154] Then, calculate the peak-to-peak difference degree between zero-sequence currents, which can be expressed as:

[0155] (24)

[0156] In this embodiment, by constructing multiple candidate fault features, it is beneficial to construct a comprehensive fault assessment system.

[0157] For the convenience of understanding by those skilled in the art, the following introduces the full process of the fault section selection model from the training stage to the prediction stage, corresponding to as Figure 5 , which mainly includes two stages: offline training and online application. The steps in the model training stage include:

[0158] Step 1: Obtain the historical data set. In the PSCAD / EMTDC simulation software, a 10 kV resonant grounding distribution network simulation model with a radial network structure is built, as Figure 3 shown. CL and OL represent cable lines and overhead lines respectively. This simulation model has a total of 5 feeders with a mixture of overhead lines and cable lines, and 18 detection nodes are installed on the feeders to collect zero-sequence current data after a single-phase grounding fault occurs. Among them, the sampling frequency of the detection nodes is set to 10 kHz. According to Figure 3The shown simulation model simulates various single-phase grounding faults by modifying fault parameters such as the fault occurrence location, fault initial phase angle, and fault grounding resistance, and collects the zero-sequence current waveform data of the first half cycle after the fault at both ends of each section. The samples selected in this application are from 80 sets of simulation data of single-phase grounding faults in the distribution network, with a total of 720 sets of data, including 360 sets of training set data and 360 sets of test set data.

[0159] Step 2: Construct a candidate fault feature set. Construct 12 candidate faults as shown in Table 1, extract features from the obtained 720 sets of data, and obtain a training sample set.

[0160] Step 3: Select the Support Vector Machine (SVM) as the fault section selection model and embed it in the process of optimizing the feature subset, that is, the fault feature screening process. In this process, use the Genetic Algorithm (GA) to optimize the hyperparameters of SVM, including the penalty factor c and the kernel function coefficient γ. Figure 4 It is the implementation framework of the feature subset optimization method, mainly including three parts: population initialization, fitness value calculation, and population evolution.

[0161] Step 3.1: In the process of population initialization, set the search space of the regularization parameter c and the kernel width parameter γ of the SVM model. Subsequently, encode the hyperparameters of the SVM model using binary coding and randomly generate multiple chromosomes as the initial population.

[0162] Step 3.2: Calculate the fitness value. First, perform feature ranking. Specifically, decode the population chromosomes to obtain the corresponding SVM hyperparameter values. Combine the hyperparameter values and calculate the importance R of each feature in the candidate feature set by training SVM. p . The importance R p is calculated as shown in the above formula (1). Delete the feature with the smallest importance from the candidate feature set and add it to the feature ranking table. Then, recalculate the importance of each candidate fault feature in the remaining candidate feature set and delete the next candidate fault feature until the candidate fault feature set is an empty set. Then, perform feature subset search. Specifically, after feature ranking, use the Sequential Floating Forward Selection (SFFS) method to screen the feature subset. This method starts from an empty set, selects and adds L features in each round to make the performance of the feature subset better, and then tries to remove R features to make the performance of the feature subset more stable and excellent. During the search process, evaluate the performance of the feature subset according to the above formula (2). The optimization goal of the hyperparameters is the same as that of the feature subset. Therefore, according to the subset search results, calculate the fitness value of the chromosome in combination with formula (2).

[0163] Step 3.3: Population evolution. Sort the individuals according to their fitness values, and then generate the next generation of individuals through three genetic evolution operations: selection, crossover, and mutation, thus forming a new population.

[0164] Step 4: After the genetic algorithm optimization is completed, output the optimal fault feature set and its corresponding optimized SVM hyperparameter values simultaneously. Then, construct an SVM segment selection model based on the optimal fault feature set and the optimized SVM hyperparameter values.

[0165] The above steps 1 - 4 can be summarized as follows: According to the candidate fault feature set shown in Table 1 above, extract 12 feature quantities of the zero-sequence current at both ends of the segment, obtaining a total of 360 groups of 1×12 - dimensional feature vectors representing the segment fault features, and splice them with the corresponding sample labels to construct a training sample set of size 360×13. Set the preset parameters of the feature subset optimization method according to Table 2 below, input the training sample set into the feature subset optimization method to obtain the optimal fault feature set and the optimal hyperparameter value group. The fault features included in the optimal fault feature set are Hausdorff distance, Euclidean distance, Bhattacharyya coefficient, and peak difference degree. The optimal hyperparameter value group includes the optimal regularization parameter c and kernel width parameter γ of the SVM model. Finally, train the SVM segment selection model based on the optimal fault feature set.

[0166] Table 2

[0167]

[0168] The steps in the model inference stage include:

[0169] Step 5: After a single - phase ground fault occurs in the distribution network line, collect the zero - sequence current waveforms in the first half - cycle after the fault at the detection nodes on both sides of the distribution network segment. Then, extract the fault feature values of the zero - sequence current waveforms on both sides of the segment under the target fault feature quantities according to the optimal fault feature set, and input them into the trained SVM segment selection model to obtain the type of this segment, that is, the fault segment or the healthy segment.

[0170] Advantages of the technical solution of this application: (1) Aiming at the problem that the multi - feature fusion method in the current single - phase ground fault location of the distribution network depends on experience to select feature quantities, this application embeds the SVM model into the feature subset optimization process, which can adapt to the model to select key feature quantities without human experience; (2) To avoid the cumulative error of multiple links, this application uses the interaction mechanism between the genetic algorithm and the feature selection algorithm to simultaneously optimize the feature quantities and the SVM hyperparameters, enabling the SVM to maximize the use of the information of the optimized feature quantities and improving the engineering practical value of the method.

[0171] In another embodiment, such as Figure 6As shown, a method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities is provided. Taking the server 104 in Figure 1 as an example for illustration, it includes the following steps:

[0172] Step S602, in the case of a single-phase grounding fault occurring in the distribution network line, collect the zero-sequence current waveform data of each detection node in the first half cycle.

[0173] Step S604, obtain a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under different fault conditions under each candidate fault feature.

[0174] Step S606, use the genetic algorithm to train the support vector machine model based on the training sample set to obtain the optimal fault feature set and the optimal hyperparameter value set.

[0175] Step S608, based on the optimal fault feature set, determine the target fault feature set, and, based on the optimal hyperparameter value set, construct a fault section selection model; the target fault features in the target fault feature set are selected from each candidate fault feature.

[0176] Step S610, according to the target fault feature set, extract the features of the zero-sequence current waveform data corresponding to the detection nodes at both ends of any section to obtain the fault feature quantities of any section under each target fault feature.

[0177] Step S612, input the fault feature quantities of any section under each target fault feature into the fault section selection model to obtain the section type of any section.

[0178] Step S614, according to the section types of each section, determine the fault section in each section.

[0179] It should be noted that the specific limitations of the above steps can refer to the specific limitations of a method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities above.

[0180] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed 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 executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0181] Based on the same inventive concept, an embodiment of the present application further provides a multi-feature-based single-phase grounding fault section selection device for a distribution network line for implementing the above-mentioned multi-feature-based single-phase grounding fault section selection method for a distribution network line. The implementation solutions provided by this device to solve problems are similar to the implementation solutions described in the above method. Therefore, the specific limitations in one or more embodiments of the multi-feature-based single-phase grounding fault section selection device for a distribution network line provided below can refer to the limitations on the multi-feature-based single-phase grounding fault section selection method for a distribution network line in the foregoing, and will not be elaborated herein.

[0182] In an exemplary embodiment, as Figure 7 shown, a multi-feature-based single-phase grounding fault section selection device for a distribution network line is provided, including: a collection module 702, an acquisition module 704, an extraction module 706, an input module 708, and a determination module 710, where:

[0183] The collection module 702 is configured to collect zero-sequence current waveform data of each detection node in the first half cycle when a single-phase grounding fault occurs in the distribution network line;

[0184] The acquisition module 704 is configured to acquire a target fault feature set and a fault section selection model for the distribution network line; the target fault features in the target fault feature set are screened from each candidate fault feature; the fault section selection model is obtained by training a support vector machine model with a training sample set; the training sample set includes fault feature quantities of each section of the distribution network line under each candidate fault feature in different fault conditions;

[0185] The extraction module 706 is configured to perform feature extraction on the zero-sequence current waveform data corresponding to the two ends of any section according to the target fault feature set to obtain the fault feature quantities of any section under each target fault feature;

[0186] The input module 708 is configured to input the fault feature quantities of any section under each target fault feature into the fault section selection model to obtain the section type of any section;

[0187] The determination module 710 is configured to determine the fault section among each section according to the section types of each section.

[0188] In one embodiment, the obtaining module 704 is specifically configured to obtain a training sample set; the training sample set includes fault feature quantities of each section of the distribution network line under each candidate fault feature in different fault conditions; use a genetic algorithm to train the support vector machine model based on the training sample set to obtain an optimal fault feature set and an optimal hyperparameter value set; determine a target fault feature set based on the optimal fault feature set, and construct a fault section selection model based on the optimal hyperparameter value set.

[0189] In one embodiment, the obtaining module 704 is specifically configured to randomly generate a plurality of chromosomes as an initial population based on the value ranges of the hyperparameters of the support vector machine model; any chromosome corresponds to a set of hyperparameter values; any set of hyperparameter values includes a regularization parameter and a kernel width parameter; obtain the individual fitness values and fault feature screening results corresponding to each chromosome in the initial population, and perform population evolution based on the individual fitness values corresponding to each chromosome to generate a plurality of new chromosomes as a new population; the individual fitness value corresponding to any chromosome represents the quality of the chromosome in the initial population; the fault feature screening result corresponding to any chromosome is determined after feature screening of each candidate fault feature during the process of training the model using the hyperparameter value set corresponding to the chromosome; in the case where the current iteration number is less than the maximum iteration number, use the new population as the initial population, update the current iteration number, and return to obtain the individual fitness values and fault feature screening results corresponding to each chromosome in the initial population until the current iteration number is equal to the maximum iteration number; in the case where the current iteration number is equal to the maximum iteration number, use the chromosome with the highest individual fitness value in the new population as the optimal chromosome, determine the hyperparameter value set corresponding to the optimal chromosome as the optimal hyperparameter value set, and determine the fault feature screening result corresponding to the optimal chromosome as the optimal fault feature set.

[0190] In one embodiment, the obtaining module 704 is specifically configured to decode each chromosome in the initial population to obtain the hyperparameter value set corresponding to each chromosome; construct a support vector machine model corresponding to different hyperparameter value sets based on the hyperparameter value sets corresponding to different chromosomes; for the hyperparameter value set corresponding to any chromosome, use the training sample set to train the support vector machine model corresponding to the hyperparameter value set, and determine the fault feature screening result and the individual fitness value corresponding to the chromosome based on the model training result.

[0191] In one embodiment, the obtaining module 704 is specifically configured to, for the hyperparameter value set corresponding to any chromosome, train the support vector machine model corresponding to the hyperparameter value set by using the training sample set to obtain the feature importance degrees of each candidate fault feature; use the sequential forward floating search method to perform feature screening based on the feature importance degrees of each candidate fault feature to obtain the fault feature screening result corresponding to the chromosome; and determine the individual fitness value corresponding to the chromosome based on the fault feature screening result corresponding to the chromosome.

[0192] In one embodiment, the candidate fault features include the cosine similarity, Pearson correlation coefficient, Spearman correlation coefficient, dynamic time warping distance, grey relational grade, relative entropy of energy, absolute distance, Euclidean distance, Hausdorff distance, Bhattacharyya coefficient, peak difference degree, and peak-to-peak difference degree between the zero-sequence current waveforms at the two ends of the section in the first half cycle after a single-phase grounding fault occurs.

[0193] Each module in the above single-phase grounding fault section selection device for a distribution network line based on multiple feature quantities can be implemented in whole or in part by software, hardware, and their combination. The above modules can be embedded in the processor in the computer device in the form of hardware or be independent of the processor, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above respective modules.

[0194] In an exemplary embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 8 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the single-phase grounding fault section selection data for a distribution network line based on multiple feature quantities. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a single-phase grounding fault section selection method for a distribution network line based on multiple feature quantities.

[0195] Those skilled in the art can understand that Figure 8The structure shown is only a block diagram of some structures related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have a different component layout.

[0196] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor is caused to execute the steps of the above-mentioned method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities. The steps of the method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities here may be the steps in the method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities in each of the above embodiments.

[0197] In one embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by the processor, the processor is caused to execute the steps of the above-mentioned method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities. The steps of the method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities here may be the steps in the method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities in each of the above embodiments.

[0198] In one embodiment, a computer program product is provided, including a computer program. When the computer program is executed by the processor, the processor is caused to execute the steps of the above-mentioned method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities. The steps of the method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities here may be the steps in the method for selecting a single-phase grounding fault section of a distribution network line based on multiple feature quantities in each of the above embodiments.

[0199] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memories can include read-only memory (ROM), magnetic tapes, floppy disks, flash memories, optical memories, high-density embedded non-volatile memories, resistive random access memories (ReRAM), magnetoresistive random access memories (MRAM), ferroelectric random access memories (FRAM), phase change memories (PCM), graphene memories, etc. Volatile memories can include random access memory (RAM) or external cache memories, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0200] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, 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, it should be considered as the scope described in this specification.

[0201] The above-described embodiments merely represent several implementation manners of the present application. Their descriptions are relatively specific and detailed, but they should not be construed as limiting the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the appended claims.

Claims

1. A method for selecting a section for a single-phase ground fault in a distribution network based on multiple characteristic quantities, characterized in that: Applied to a power distribution network line, the power distribution network line includes a plurality of sections, and detection nodes are arranged at both ends of any of the sections. The method includes: In the event of a single-phase grounding fault in the distribution network line, collecting zero-sequence current waveform data of each detection node in the first half cycle; Obtain a target fault feature set and a fault section selection model for the distribution network line; the target fault features in the target fault feature set are screened from the candidate fault features; the fault section selection model is obtained by training a support vector machine model using a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under each candidate fault feature under different fault conditions; According to the target fault feature set, feature extraction is performed on zero-sequence current waveform data corresponding to the detection nodes at both ends of any of the sections to obtain the fault feature quantity of any of the sections under each target fault feature; Inputting the fault feature quantity of any of the sections under each of the target fault features into the fault section selection model to obtain the section type of any of the sections; A faulty section is determined in each of the sections according to the section type of each of the sections.

2. The method according to claim 1, characterized in that The obtaining of a target fault feature set and a fault section selection model for the distribution network line includes: Obtaining the training sample set; Using a genetic algorithm to train a support vector machine model based on the training sample set to obtain an optimal fault feature set and an optimal hyperparameter value set; Based on the optimal fault feature set, the target fault feature set is determined, and based on the optimal hyperparameter value set, the fault segment selection model is constructed.

3. The method according to claim 2, characterized in that The adopting of a genetic algorithm to train a support vector machine model based on the training sample set to obtain an optimal fault feature set and an optimal hyperparameter value set includes: Based on the value range of each hyperparameter of the support vector machine model, a plurality of chromosomes are randomly generated as an initial population; any of the chromosomes corresponds to a set of hyperparameter value sets; any of the hyperparameter value sets includes a regularization parameter and a kernel width parameter; The individual fitness value and fault feature screening result corresponding to each of the chromosomes in the initial population are obtained, and the population is evolved based on the individual fitness value corresponding to each of the chromosomes to generate multiple new chromosomes as a new population; the individual fitness value corresponding to any of the chromosomes represents the quality of the chromosome in the initial population; the fault feature screening result corresponding to any of the chromosomes is determined after feature screening of each of the candidate fault features in the process of model training using the hyperparameter value set corresponding to the chromosome; In the case where the current number of iterations is less than the maximum number of iterations, the new population is used as the initial population, the current number of iterations is updated, and the individual fitness values ​​and fault feature screening results corresponding to each chromosome in the initial population are obtained and returned until the current number of iterations is equal to the maximum number of iterations; When the current number of iterations is equal to the maximum number of iterations, the chromosome with the highest individual fitness value in the new population is taken as the optimal chromosome, the hyperparameter value set corresponding to the optimal chromosome is determined as the optimal hyperparameter value set, and the fault feature screening result corresponding to the optimal chromosome is determined as the optimal fault feature set.

4. The method according to claim 3, characterized in that The obtaining of the individual fitness value and the fault feature screening result corresponding to each chromosome in the initial population includes: Decoding each of the chromosomes in the initial population to obtain a hyperparameter value set corresponding to each of the chromosomes; Based on the hyperparameter value sets corresponding to the different chromosomes, construct support vector machine models corresponding to the different hyperparameter value sets; For any hyperparameter value set corresponding to the chromosome, the support vector machine model corresponding to the hyperparameter value set is trained using the training sample set, and based on the model training result, the fault feature screening result and the individual fitness value corresponding to the chromosome are determined.

5. The method according to claim 4, characterized in that For any of the hyperparameter value sets corresponding to the chromosomes, the support vector machine model corresponding to the hyperparameter value set is trained using the training sample set, and based on the model training result, the fault feature screening result and the individual fitness value corresponding to the chromosome are determined, including: For a hyperparameter value set corresponding to any of the chromosomes, the support vector machine model corresponding to the hyperparameter value set is trained using the training sample set to obtain the feature importance of each of the candidate fault features; Using a sequence forward floating search method, feature screening is performed based on the feature importance of each candidate fault feature to obtain a fault feature screening result corresponding to the chromosome; Based on the fault feature screening result corresponding to the chromosome, the individual fitness value corresponding to the chromosome is determined.

6. The method according to claim 1, characterized in that The candidate fault features include cosine similarity, Pearson correlation coefficient, Spearman correlation coefficient, dynamic time warping distance, grey correlation, energy relative entropy, absolute distance, Euclidean distance, Hausdorff distance, Bhattacharyya coefficient, peak difference and peak-to-peak difference between the zero-sequence current waveforms of the first half cycle at both ends of the section after the single-phase grounding fault occurs.

7. A single-phase ground fault section selection device for a distribution network line based on multiple characteristic quantities, characterized in that: Applied to a power distribution network line, the power distribution network line includes multiple sections, and detection nodes are arranged at both ends of any of the sections. The device includes: A collection module, used for collecting zero-sequence current waveform data of each detection node in the first half cycle when a single-phase grounding fault occurs in the distribution network line; An acquisition module is used to acquire a target fault feature set and a fault section selection model for the distribution network line; the target fault features in the target fault feature set are screened from the candidate fault features; the fault section selection model is obtained by training a support vector machine model using a training sample set; the training sample set includes the fault feature quantities of each section of the distribution network line under each candidate fault feature under different fault conditions; An extraction module, used to extract features from zero-sequence current waveform data corresponding to the detection nodes at both ends of any of the sections according to the target fault feature set, to obtain fault feature quantities of any of the sections under each target fault feature; An input module, used for inputting the fault feature quantity of any of the sections under each of the target fault features into the fault section selection model to obtain the section type of any of the sections; The determination module is used to determine the faulty section in each of the sections according to the section type of each of the sections.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

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

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

Images