Single-phase disconnection fault positioning method and device based on BP neural network, and medium

Abstract

The invention discloses a BP neural network-based single-phase disconnection fault positioning method and apparatus, and a medium. The method comprises the steps of collecting a single-side disconnection fault feature set of a power distribution network; mapping the fault feature set to a high-dimensional feature space through a preset kernel function to obtain a mapping result, and screening out a feature subset from the broken line fault feature set based on the mapping result; inputting the feature subset into a preset fault positioning model, searching an optimal model parameter of the fault positioning model by adopting a particle swarm optimization algorithm, and determining a target fault positioning model based on the optimal model parameter; and collecting the electrical characteristic quantity of the single side of the power distribution network, inputting the electrical characteristic quantity into the target fault positioning model to determine an output model output result, and determining the position of the single-phase disconnection fault based on the model output result. According to the invention, the position of the broken line fault can be accurately and rapidly positioned.

Description

Technical Field

[0001] The present application relates to the field of fault location, and in particular to a single-phase line break fault location method, device and medium based on a BP neural network. Background Art

[0002] The distribution network directly serves a large number of users. Its complex branch structure and changing environmental conditions make single-phase disconnection faults frequent. If such faults are not quickly identified and handled, they may cause large-scale power outages and even life-threatening accidents, which seriously deviates from the basic needs of users. Therefore, it is particularly important to accurately locate single-phase disconnection faults in the distribution network.

[0003] Most existing fault location methods rely on limited feature quantities to locate disconnection faults. Since some key feature quantities do not change significantly compared to the normal state before the fault when a disconnection fault occurs in the distribution network, this poses a great challenge to fault judgment. At the same time, it is difficult to accurately determine the threshold of the disconnection fault, which leads to low fault location accuracy and makes it difficult to meet the needs of effectively identifying the disconnection fault of the distribution network.

[0004] Application Contents

[0005] The present application provides a single-phase disconnection fault location method, device and medium based on BP neural network to accurately and quickly locate the position of the disconnection fault.

[0006] In a first aspect, the present application provides a single-phase disconnection fault location method based on a BP neural network, comprising:

[0007] Collecting a line-break fault feature set on one side of the distribution network, wherein the line-break fault feature set includes a phase difference between the fault phase voltage and the negative-sequence current, a negative-sequence voltage and current phase difference, a line voltage amplitude, and a phase voltage amplitude;

[0008] Mapping the fault feature set to a high-dimensional feature space through a preset kernel function to obtain a mapping result, and filtering out a feature subset from the disconnection fault feature set based on the mapping result;

[0009] Inputting the feature subset into a preset fault location model, and using a particle swarm optimization algorithm to search for optimal model parameters of the fault location model, and determining a target fault location model based on the optimal model parameters;

[0010] The electrical characteristic quantity of one side of the distribution network is collected, and the electrical characteristic quantity is input into the target fault location model to determine the output model output result, and the location of the single-phase disconnection fault is determined based on the model output result.

[0011] The embodiment of the present application can comprehensively and accurately obtain the feature set when the fault occurs by collecting the line-break fault feature set on one side of the distribution network, which is convenient for subsequent analysis and fault location; by mapping the fault feature set to a high-dimensional feature space, the nonlinear relationship and hidden features of the data can be better revealed, which is convenient for subsequent extraction of useful feature subsets; by screening out feature subsets from the line-break fault feature set, the most useful information for fault location can be extracted, which is convenient for subsequent mining of the mapping relationship between the line-break features in the feature subset and the line-break fault location, thereby improving the efficiency and accuracy of subsequent fault location; by training the fault location model with feature subsets, the mapping relationship between the line-break features in the feature subset and the line-break fault location can be mined, and by optimizing the fault location model with a particle swarm optimization algorithm, it can be ensured that the model has better generalization ability and robustness when dealing with complex and nonlinear problems, thereby improving the accuracy and reliability of fault location; by determining the location of the single-phase line-break fault through the target fault model, the location of the single-phase line-break fault can be quickly and accurately determined. Compared with the prior art, the present application can accurately and quickly locate the location of the line-break fault.

[0012] Further, the filtering out of a feature subset from the disconnection fault feature set based on the mapping result is specifically:

[0013] Calculating a covariance matrix of the mapping result, and performing eigendecomposition on the covariance matrix to obtain a first eigenvalue and a corresponding eigenvector;

[0014] Perform inner product calculation based on the mapping result, the first eigenvalue and the eigenvector to obtain a projection value, and simplify the calculation process of the projection value by using the kernel function to obtain a simplified equation;

[0015] The simplified equation is solved to obtain a second characteristic value, and a characteristic subset is screened out from the disconnection fault characteristic set based on the second characteristic value.

[0016] In this way, by screening out a feature subset from the disconnection fault feature set, the most useful information for fault location can be extracted, which facilitates subsequent mining of the mapping relationship between the disconnection features in the feature subset and the disconnection fault location, thereby improving the efficiency and accuracy of subsequent fault location.

[0017] Further, the filtering out of a feature subset from the disconnection fault feature set based on the second feature value is specifically:

[0018] Calculating a corresponding contribution rate for each of the second eigenvalues;

[0019] A corresponding cumulative contribution rate is determined based on the contribution rate, a principal component is determined based on the contribution rate and the cumulative contribution rate, and a feature subset is screened out from the line break fault feature set based on the principal component.

[0020] In this way, by screening out a feature subset from the disconnection fault feature set, the most useful information for fault location can be extracted, which facilitates subsequent mining of the mapping relationship between the disconnection features in the feature subset and the disconnection fault location, thereby improving the efficiency and accuracy of subsequent fault location.

[0021] Furthermore, the particle swarm optimization algorithm is used to search for the optimal model parameters of the fault location model, specifically:

[0022] Initialize the initial velocity and initial position of the particle within the preset particle search range;

[0023] Determining a fitness function according to initial model parameters of the fault location model;

[0024] The fitness value is iteratively calculated based on the fitness function, the initial velocity and the initial position until a termination condition is met, and a position matrix of the particle representing the optimal model parameters is output, wherein the position matrix and the velocity matrix of the particle are updated in each iteration.

[0025] In this way, optimizing the fault location model through the particle swarm optimization algorithm can ensure that the model has better generalization ability and robustness when dealing with complex and nonlinear problems, thereby improving the accuracy and reliability of fault location.

[0026] Furthermore, the calculation formula for updating the position matrix and velocity matrix of the particle is specifically:

[0027]

[0028] In the formula, and are the velocity and position of the ith particle at the k+1th iteration, ω is the inertia weight, and are the velocity and position of the ith particle at the kth iteration, c 1 is the individual particle learning factor, c 2 is the particle group learning factor, P best,i is the historical optimal position of the i-th particle, G best is the historical optimal position of the particle population.

[0029] In a second aspect, the present application provides a single-phase disconnection fault location device based on a BP neural network, comprising: a collection module, a screening module, a determination module and a location module;

[0030] The acquisition module is used to acquire a set of line-break fault characteristics on one side of the distribution network, wherein the set of line-break fault characteristics includes a phase difference between the fault phase voltage and the negative-sequence current, a negative-sequence voltage-current phase difference, a line voltage amplitude, and a phase voltage amplitude;

[0031] The screening module is used to map the fault feature set to a high-dimensional feature space through a preset kernel function to obtain a mapping result, and screen out a feature subset from the disconnection fault feature set based on the mapping result;

[0032] The determination module is used to input the feature subset into a preset fault location model, and use a particle swarm optimization algorithm to search for optimal model parameters of the fault location model, and determine a target fault location model based on the optimal model parameters;

[0033] The positioning module is used to collect electrical characteristic quantities on one side of the distribution network, input the electrical characteristic quantities into the target fault positioning model to determine the output model output result, and determine the location of the single-phase disconnection fault based on the model output result.

[0034] The embodiment of the present application can comprehensively and accurately obtain the feature set when the fault occurs by collecting the line-break fault feature set on one side of the distribution network, which is convenient for subsequent analysis and fault location; by mapping the fault feature set to a high-dimensional feature space, the nonlinear relationship and hidden features of the data can be better revealed, which is convenient for subsequent extraction of useful feature subsets; by screening out feature subsets from the line-break fault feature set, the most useful information for fault location can be extracted, which is convenient for subsequent mining of the mapping relationship between the line-break features in the feature subset and the line-break fault location, thereby improving the efficiency and accuracy of subsequent fault location; by training the fault location model with feature subsets, the mapping relationship between the line-break features in the feature subset and the line-break fault location can be mined, and by optimizing the fault location model with a particle swarm optimization algorithm, it can be ensured that the model has better generalization ability and robustness when dealing with complex and nonlinear problems, thereby improving the accuracy and reliability of fault location; by determining the location of the single-phase line-break fault through the target fault model, the location of the single-phase line-break fault can be quickly and accurately determined. Compared with the prior art, the present application can accurately and quickly locate the location of the line-break fault.

[0035] Further, the screening module includes: a first calculation unit, a second calculation unit and a third calculation unit;

[0036] The first calculation unit is used to calculate the covariance matrix of the mapping result, and perform eigendecomposition on the covariance matrix to obtain a first eigenvalue and a corresponding eigenvector;

[0037] The second calculation unit is used to perform inner product calculation based on the mapping result, the first eigenvalue and the eigenvector to obtain a projection value, and simplify the calculation process of the projection value by using the kernel function to obtain a simplified equation;

[0038] The third calculation unit is used to solve the simplified equation to obtain a second eigenvalue, and filter out a feature subset from the line break fault feature set based on the second eigenvalue.

[0039] Further, the third calculation unit includes: a calculation subunit and a screening subunit;

[0040] The calculation subunit is used to calculate the corresponding contribution rate for each second eigenvalue;

[0041] The screening subunit is used to determine the corresponding cumulative contribution rate based on the contribution rate, determine the principal component based on the contribution rate and the cumulative contribution rate, and screen out a feature subset from the line break fault feature set based on the principal component.

[0042] Further, the determination module includes: an initialization unit, a fourth calculation unit and a fifth calculation unit;

[0043] The initialization unit is used to initialize the initial velocity and initial position of the particle within a preset particle search range;

[0044] The fourth calculation unit is used to determine the fitness function according to the initial model parameters of the fault location model;

[0045] The fifth calculation unit is used to iteratively calculate the fitness value based on the fitness function, the initial velocity and the initial position until a termination condition is met, and output a position matrix of the particle representing the optimal model parameters, wherein the position matrix and velocity matrix of the particle are updated in each iteration.

[0046] In a third aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the single-phase line break fault location method based on a BP neural network as described in the present application is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of a disconnection fault circuit provided by this application;

[0048] Figure 2 It is a flow chart of an embodiment of a single-phase disconnection fault location method based on a BP neural network provided by the present application;

[0049] Figure 3is a schematic diagram of a disconnection fault feature set provided by the present application;

[0050] Figure 4 It is a structural schematic diagram of the fault location model provided by this application;

[0051] Figure 5 It is a circuit diagram for fault location of a 10kV distribution network provided by the present application;

[0052] Figure 6 It is a schematic diagram comparing the actual value and the predicted value of the target fault location model provided by the present application;

[0053] Figure 7 It is a structural schematic diagram of an embodiment of a single-phase line break fault location system based on BP neural network provided by the present application. DETAILED DESCRIPTION

[0054] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0055] It should be understood that the step numbers used in this article are only for the convenience of description and are not intended to limit the order in which the steps are executed.

[0056] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.

[0057] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0058] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0059] The complex branch structure and changing environmental conditions of the distribution network lead to frequent single-phase disconnection faults. If such faults are not quickly identified and handled, they will cause large-scale power outages and even life-threatening accidents. Therefore, it is particularly important to accurately locate single-phase disconnection faults in distribution networks. Most of the existing fault location methods rely on limited feature quantities to locate disconnection faults. However, when a disconnection fault occurs, the change range of some key feature quantities is not significant compared with the normal state before the fault. At the same time, the threshold of the disconnection fault is difficult to determine accurately, resulting in low fault location accuracy.

[0060] Next, the nouns involved in this application are analyzed:

[0061] BP (Back Propagation) neural network is a multi-layer feedforward neural network. Its training process uses the back propagation algorithm. BP neural network minimizes the error between the predicted value and the actual value by continuously adjusting the weights and biases of the network. The key part of BP neural network is back propagation, which propagates the error of the output layer back to the hidden layer and input layer layer by layer through the chain rule, and calculates the gradient of each weight in order to update them.

[0062] A single-phase line break fault in a distribution network refers to a power system fault in which a single-phase line is disconnected due to a break in the conductor or poor contact of a single-phase line in a three-phase AC power system, while the other two-phase lines are not affected.

[0063] Particle Swarm Optimization (PSO) is an optimization algorithm based on swarm intelligence, which uses the cooperation and information sharing of individuals in the swarm to find the optimal solution. In the algorithm, each solution to the optimization problem is regarded as a "particle" in the search space, and these particles have two properties: speed and position. Speed ​​represents the speed of particle movement, and position represents the direction of the particle in the search space. Particles continuously update their positions and speeds by tracking their own historical best positions and the best positions of the group, thereby achieving optimization.

[0064] Based on this, the embodiments of the present application provide a single-phase line break fault location method, device and medium based on a BP neural network, which can accurately and quickly locate the position of the line break fault.

[0065] The embodiments of the present application provide a single-phase line break fault location method, device and medium based on BP neural network, which are specifically explained through the following embodiments. First, the single-phase line break fault location method based on BP neural network in the embodiments of the present application is described.

[0066] The single-phase line break fault location method based on BP neural network provided in the embodiment of the present application relates to the field of fault location. The single-phase line break fault location method based on BP neural network provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, and can also be configured as a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the single-phase line break fault location method based on BP neural network, etc., but is not limited to the above forms.

[0067] The present application can also be used in numerous general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0068] Since most distribution networks are single-power supply systems, the power supply side of the monitoring point is the upstream side, and the load side of the monitoring point is the downstream side. Figure 1 As shown, when a fault occurs at the transformer outlet, the monitoring device T4 detects that the line break fault is located downstream thereof, and T1, T2, and T3 detect that the line break fault is located upstream thereof.

[0069] Embodiment 1

[0070] Please refer to Figure 2 , Figure 2 It is a flow chart of an embodiment of a single-phase disconnection fault location method based on a BP neural network provided by the present application, including steps S101 to S104;

[0071] Step S101, collecting a line break fault feature set on one side of the distribution network, wherein the line break fault feature set includes a phase difference between the fault phase voltage and the negative sequence current, a negative sequence voltage and current phase difference, a line voltage amplitude, and a phase voltage amplitude;

[0072] In some embodiments, voltage transformers or current sensors are installed at key nodes of the distribution network (such as transformer outlets, branch line starting points, etc.) to collect three-phase voltage and three-phase current when a single-phase line break occurs in the distribution network, and then the line voltage amplitude, phase voltage amplitude, negative-sequence voltage and negative-sequence current are calculated based on the three-phase voltage and three-phase current, thereby obtaining the fault phase voltage, negative-sequence current phase difference and negative-sequence voltage and current phase difference, and these features are summarized to obtain a line break fault feature set.

[0073] In some embodiments, the three-phase voltage and three-phase current can be collected by FTU (feeder terminal unit) when a single-phase line break occurs in the distribution network, and then the unbalanced components corresponding to the three-phase voltage and the three-phase current, i.e., the negative-sequence voltage and the negative-sequence current, are calculated respectively; and the line voltage amplitude and the phase voltage amplitude are determined based on the three-phase voltage, and the fault phase voltage value when the fault occurs is monitored and recorded at the same time. After the negative-sequence current and the negative-sequence voltage are obtained, it is necessary to calculate the phase difference between the negative-sequence current and the reference signal (such as the positive-sequence current or a fixed-phase signal), i.e., the negative-sequence current phase difference, and calculate the phase difference between the negative-sequence voltage and the negative-sequence current to obtain the negative-sequence voltage and current phase difference; and summarize these features to obtain a line break fault feature set, wherein the schematic diagram of the line break fault feature set is as shown in FIG. Figure 3 shown.

[0074] Step S102, mapping the fault feature set to a high-dimensional feature space through a preset kernel function to obtain a mapping result, and filtering out a feature subset from the disconnection fault feature set based on the mapping result;

[0075] In some embodiments, before mapping the fault feature set to the high-dimensional feature space through a preset kernel function, the fault feature set is further normalized to convert the data of the fault feature set into a fixed interval range, and the relevant formula for the normalization process is: In the formula, x i ′ is the normalized data, x i is a sample in the ith variable, x min is the minimum value in the sample of the variable, x max is the maximum value in the sample of this variable.

[0076] In some embodiments, the fault feature set is mapped to a high-dimensional feature space through a preset kernel function to obtain a mapping result. Specifically, first, a kernel function is selected to map the fault feature set of the original feature space to a high-dimensional feature space F, so that the data may become linearly separable in this space, thereby better processing the nonlinear data in the fault feature set. The mapping process is: K ij =K(x i ,x j )=φ(x i )φ(x j ), where K ij is the kernel function matrix, φ(x i ) and φ(x j ) is the mapping result of the i-th and j-th variables, x i is a sample in the ith variable, x j A sample in the j-th variable, φ is a nonlinear kernel function, which can be but is not limited to a linear kernel, a polynomial kernel, a radial basis function (RBF) kernel, a Sigmoid kernel, etc.

[0077] It should be noted that the mapping result is actually a kernel matrix calculated by the kernel function, which represents the similarity between the fault feature set samples in the high-dimensional feature space F.

[0078] In some embodiments, a feature subset is selected from the disconnection fault feature set based on the mapping result, including: calculating the covariance matrix of the mapping result, and performing eigendecomposition on the covariance matrix to obtain a first eigenvalue and a corresponding eigenvector; performing inner product calculation based on the mapping result, the first eigenvalue and the eigenvector to obtain a projection value, and simplifying the calculation process of the projection value through the kernel function to obtain a simplified equation; solving the simplified equation to obtain a second eigenvalue, and selecting a feature subset from the disconnection fault feature set based on the second eigenvalue. Specifically: first, when the mapping result K is obtained, ij After that, we need to calculate a variable mapping result φ(x i ) In the formula, C F is the covariance matrix, n is the number of samples in the mapping result, φ is the nonlinear kernel function, φ(x i ) is the mapping result of the i-th variable, T represents the transposition, and then the covariance matrix C F Perform eigendecomposition to obtain the first eigenvalue λ and the corresponding eigenvector v. The formula for eigendecomposition is: λv=C F v. Secondly, based on the mapping result φ(x i ), the first eigenvalue λ and the eigenvector v are inner-producted to obtain the projection value λ[φ(xi )·v], the formula for calculating the inner product is: λ[φ(x i )·v]=φ(x i )·C F v, where λ[φ(x i )·v] is the projection value, φ(x i ) is the mapping result of the i-th variable, C F is the covariance matrix, v is the eigenvector; if the first eigenvalue λ is not 0, then in the high-dimensional space F, the eigenvector v can be mapped to the data set φ(x i ) is linearly represented as: At this time, in the high-dimensional space F, the calculation process of the projection value is: The calculation process of the projection value is simplified by using a kernel function, and a simplified equation nλα=Kα is obtained, where λ is the second eigenvalue, α is the eigenvector (in a high-dimensional space) of the kernel function matrix K, and K is the kernel function matrix. Finally, by solving nλα=Kα, the second eigenvalue λ of the kernel function matrix K and its corresponding eigenvector α can be obtained, and a feature subset can be screened out from the disconnection fault feature set based on the second eigenvalue λ.

[0079] In some embodiments, selecting a feature subset from the disconnection fault feature set based on the second eigenvalue includes: calculating a corresponding contribution rate for each second eigenvalue; determining a corresponding cumulative contribution rate based on the contribution rate, determining a principal component based on the contribution rate and the cumulative contribution rate, and selecting a feature subset from the disconnection fault feature set based on the principal component. Specifically, first, after obtaining the second eigenvalue λ, it is necessary to calculate a corresponding contribution rate Con(λ) for each second eigenvalue λ. i ), the contribution rate calculation formula is: In the formula, Con(λ i ) is the contribution rate of the i-th second eigenvalue, then the contribution rates of the second eigenvalues ​​are accumulated to obtain the cumulative contribution rate, and finally, the contribution rate Con(λ i ) is the largest and the cumulative contribution rate reaches or exceeds a preset threshold (such as 90%) as the characteristic vector corresponding to the second eigenvalue as the principal component, and based on these principal components, a feature subset is screened out from the line break fault feature set for subsequent fault analysis or model training.

[0080] It should be noted that the mapping result φ(x i ) is projected in the direction of each element to transform the data from the original feature space into a new space composed of the most important features (i.e., principal components). The calculation formula is:

[0081] In this way, by screening out a feature subset from the disconnection fault feature set, the most useful information for fault location can be extracted, which facilitates subsequent mining of the mapping relationship between the disconnection features in the feature subset and the disconnection fault location, thereby improving the efficiency and accuracy of subsequent fault location.

[0082] Step S103, inputting the feature subset into a preset fault location model, and using a particle swarm optimization algorithm to search for optimal model parameters of the fault location model, and determining a target fault location model based on the optimal model parameters;

[0083] In some embodiments, the preset fault location model is constructed based on the BP neural network. When constructing the fault location model, it is necessary to define the model output as a disconnection upstream and downstream judgment logic quantity, wherein the disconnection upstream and downstream judgment logic quantity is the position of the disconnection fault point relative to the monitoring point. On the premise that a disconnection fault occurs in a known line, when the neural network output is 1, it is marked as the disconnection is located upstream of the monitoring point, and when the neural network output is -1, it is marked as the disconnection is located downstream of the monitoring point.

[0084] In some embodiments, the fault location model includes an input layer, a hidden layer and an output layer. Figure 4 As shown, the hidden layer is activated by the logsig function, the output layer is activated by the tansig function, the extracted features and their labels are sent to the input layer, the fault location prediction label is output from the output layer, and the real label is compared with the predicted label of the model. A large amount of data is used for training, the prediction error of the model is calculated, and the model with the smallest prediction error is saved. This model is used as the distribution network fault location model.

[0085] In some embodiments, the particle swarm optimization algorithm is used to search for the optimal model parameters of the fault location model, specifically: initializing the initial velocity and initial position of the particle within a preset particle search range; determining the fitness function according to the initial model parameters of the fault location model; iteratively calculating the fitness value based on the fitness function, the initial velocity and the initial position until the termination condition is met, and outputting the position matrix of the particle representing the optimal model parameters, wherein the position matrix and velocity matrix of the particle are updated in each iteration. Specifically, first, the size of the particle swarm (i.e., the number of particles) is set, and each particle P i Each is a separate neural network with independent parameters, randomly determining each particle P i The initial velocity v i and the initial position x i , where position represents a combination of model parameters, and speed represents the direction and magnitude of parameter update; secondly, the error norm of the BP neural network is used to measure the fitness function: min||T sim -Ttrain ||, where T sim Represents the discrimination result of the neural network under the current weight and threshold, T train represents the actual discrimination result of the neural network training set to determine the fitness function; then, for each particle P i , use its current position (i.e., model parameters) to calculate the fitness value. If the fitness value of the current particle is better than its historical best position (i.e., the best fitness value in the previous iteration), then update the P value of the particle. best,i , and find the particle with the best fitness value in the entire particle group, and update the G of the particle best Finally, based on the current position, current speed, P best,i and G best Iteratively update the speed and position of each particle. After each iteration, check whether the termination condition is met. If the termination condition is met, output the particle position matrix representing the optimal model parameters. This position matrix contains the optimal parameter settings of the fault localization model. The optimal model parameters include the optimal weights and the optimal thresholds. The optimal weight network is assigned to the target fault localization model for fault localization.

[0086] In some embodiments, the termination condition may be setting a preset error accuracy or a preset number of iterations, and when the error accuracy reaches the requirement or reaches a maximum number of iterations, it is determined that the termination condition is satisfied.

[0087] In this way, optimizing the fault location model through the particle swarm optimization algorithm can ensure that the model has better generalization ability and robustness when dealing with complex and nonlinear problems, thereby improving the accuracy and reliability of fault location.

[0088] In some embodiments, the calculation formula for updating the position matrix and velocity matrix of the particle is specifically:

[0089]

[0090] In the formula, and are the velocity and position of the ith particle at the k+1th iteration, ω is the inertia weight, and are the velocity and position of the ith particle at the kth iteration, c 1 is the individual particle learning factor, c 2 is the particle group learning factor, P best,i is the historical optimal position of the i-th particle, G best is the historical optimal position of the particle population.

[0091] Step S104, collecting electrical characteristic quantities on one side of the distribution network, and inputting the electrical characteristic quantities into the target fault location model to determine an output model output result, and determining the location of the single-phase disconnection fault based on the model output result.

[0092] In some embodiments, electrical characteristic quantities on one side of the distribution network are acquired through sensors or data acquisition units, wherein the electrical characteristic quantities may be, but are not limited to, the phase difference between the fault phase voltage and the negative-sequence current, the negative-sequence voltage and current phase difference, the line voltage amplitude, and the phase voltage amplitude; thereafter, the electrical characteristic quantities are preprocessed, and the preprocessed electrical characteristic quantities are input into the target fault location model to obtain the model output result, and the location of the single-phase line break fault is determined based on the model output result. For example, when the model output result is in the interval [0.5, 1], the single-phase line break fault is considered to be located upstream of the monitoring point; when the model output result is in the interval [-1, -0.5], the single-phase line break fault is considered to be located downstream of the monitoring point.

[0093] The following is a verification and explanation of the technical effects used in this method. This embodiment uses a traditional technical solution to conduct a comparative test with the method of the present invention, and compares the test results by means of scientific demonstration to verify the real effect of this method. Figure 5 As shown in the circuit diagram for fault location of a 10kV distribution network, 14 groups of single-phase line break faults are simulated at different locations, and electrical feature quantities are obtained from 11 monitoring points. After KPCA dimensionality reduction, the electrical feature quantities are input into the model for training. In the data set, the ratio of the training set to the test set is 7:3, and the ratio of the upstream fault data to the downstream fault data contained in the feature data is approximately 5:3. Figure 6 It is a schematic diagram for comparing the true value and the predicted value of the target fault location model. It can be seen that the predicted value tested by the method provided by the present invention is very close to the true value of the fault location, and the purpose of section location can be achieved.

[0094] Table 1 is an evaluation of the results predicted by the target fault location model and the traditional single criterion model. From Table 1, it can be seen that the target fault location model proposed in the embodiment of the present application has excellent performance in prediction accuracy, indicating that the prediction results of the target fault location model are more accurate and can effectively extract feature information, so that the neural network learns the feature quantity that has a greater impact on fault location, thereby improving the accuracy of fault location.

[0095] Table 1: Comparison of the accuracy of the target fault location model and the traditional single criterion

[0096] method Accuracy Negative sequence voltage and current phase angle difference 88.14％ Phase difference between fault phase voltage and negative sequence current 86.44％ Line voltage amplitude 93.22％ Phase voltage amplitude 89.83％ BP neural network training set 100％ BP neural network test set 100％

[0097] The embodiment of the present application can comprehensively and accurately obtain the feature set when the fault occurs by collecting the line-break fault feature set on one side of the distribution network, which is convenient for subsequent analysis and fault location; by mapping the fault feature set to a high-dimensional feature space, the nonlinear relationship and hidden features of the data can be better revealed, which is convenient for subsequent extraction of useful feature subsets; by screening out feature subsets from the line-break fault feature set, the most useful information for fault location can be extracted, which is convenient for subsequent mining of the mapping relationship between the line-break features in the feature subset and the line-break fault location, thereby improving the efficiency and accuracy of subsequent fault location; by training the fault location model with feature subsets, the mapping relationship between the line-break features in the feature subset and the line-break fault location can be mined, and by optimizing the fault location model with a particle swarm optimization algorithm, it can be ensured that the model has better generalization ability and robustness when dealing with complex and nonlinear problems, thereby improving the accuracy and reliability of fault location; by determining the location of the single-phase line-break fault through the target fault model, the location of the single-phase line-break fault can be quickly and accurately determined. Compared with the prior art, the present application can accurately and quickly locate the location of the line-break fault.

[0098] Embodiment 2

[0099] Please refer to Figure 7 , Figure 7 It is a structural schematic diagram of an embodiment of a single-phase disconnection fault location system based on a BP neural network provided by the present application, comprising a collection module 100, a screening module 200, a determination module 300 and a location module 400;

[0100] The acquisition module 100 is used to acquire a line-break fault feature set on one side of the distribution network, wherein the line-break fault feature set includes a phase difference between the fault phase voltage and the negative-sequence current, a negative-sequence voltage and current phase difference, a line voltage amplitude, and a phase voltage amplitude;

[0101] In some embodiments, voltage transformers or current sensors are installed at key nodes of the distribution network (such as transformer outlets, branch line starting points, etc.) to collect three-phase voltage and three-phase current when a single-phase line break occurs in the distribution network, and then the line voltage amplitude, phase voltage amplitude, negative-sequence voltage and negative-sequence current are calculated based on the three-phase voltage and three-phase current, thereby obtaining the fault phase voltage, negative-sequence current phase difference and negative-sequence voltage and current phase difference, and these features are summarized to obtain a line break fault feature set.

[0102] In some embodiments, the three-phase voltage and three-phase current can be collected by FTU (feeder terminal unit) when a single-phase line break occurs in the distribution network, and then the unbalanced components corresponding to the three-phase voltage and the three-phase current, i.e., the negative-sequence voltage and the negative-sequence current, are calculated respectively; and the line voltage amplitude and the phase voltage amplitude are determined based on the three-phase voltage, and the fault phase voltage value when the fault occurs is monitored and recorded at the same time. After the negative-sequence current and the negative-sequence voltage are obtained, it is necessary to calculate the phase difference between the negative-sequence current and the reference signal (such as the positive-sequence current or a fixed-phase signal), i.e., the negative-sequence current phase difference, and calculate the phase difference between the negative-sequence voltage and the negative-sequence current to obtain the negative-sequence voltage and current phase difference; and summarize these features to obtain a line break fault feature set, wherein the schematic diagram of the line break fault feature set is as shown in FIG. Figure 3 shown.

[0103] The screening module 200 is used to map the fault feature set to a high-dimensional feature space through a preset kernel function to obtain a mapping result, and screen out a feature subset from the disconnection fault feature set based on the mapping result;

[0104] In some embodiments, before mapping the fault feature set to the high-dimensional feature space through a preset kernel function, the fault feature set is further normalized to convert the data of the fault feature set into a fixed interval range, and the relevant formula for the normalization process is: In the formula, x i ′ is the normalized data, x i is a sample in the ith variable, x min is the minimum value in the sample of the variable, x max is the maximum value in the sample of this variable.

[0105] In some embodiments, the fault feature set is mapped to a high-dimensional feature space through a preset kernel function to obtain a mapping result. Specifically, first, a kernel function is selected to map the fault feature set of the original feature space to a high-dimensional feature space F, so that the data may become linearly separable in this space, thereby better processing the nonlinear data in the fault feature set. The mapping process is: K ij =K(x i ,x j )=φ(x i )φ(x j ), where K ij is the kernel function matrix, φ(x i ) and φ(x j ) is the mapping result of the i-th and j-th variables, x i is a sample in the ith variable, x jA sample in the j-th variable, φ is a nonlinear kernel function, which can be but is not limited to a linear kernel, a polynomial kernel, a radial basis function (RBF) kernel, a Sigmoid kernel, etc.

[0106] It should be noted that the mapping result is actually a kernel matrix calculated by the kernel function, which represents the similarity between the fault feature set samples in the high-dimensional feature space F.

[0107] In some embodiments, the screening module 200 includes: a first calculation unit, a second calculation unit and a third calculation unit; the first calculation unit is used to calculate the covariance matrix of the mapping result, and perform eigendecomposition on the covariance matrix to obtain a first eigenvalue and a corresponding eigenvector; the second calculation unit is used to perform inner product calculation based on the mapping result, the first eigenvalue and the eigenvector to obtain a projection value, and simplify the calculation process of the projection value through the kernel function to obtain a simplified equation; the third calculation unit is used to solve the simplified equation to obtain a second eigenvalue, and filter out a feature subset from the disconnection fault feature set based on the second eigenvalue. Specifically: First, when the mapping result K is obtained ij After that, we need to calculate a variable mapping result φ(x i ) In the formula, C F is the covariance matrix, n is the number of samples in the mapping result, φ is the nonlinear kernel function, φ(x i ) is the mapping result of the i-th variable, T represents the transposition, and then the covariance matrix C F Perform eigendecomposition to obtain the first eigenvalue λ and the corresponding eigenvector v. The formula for eigendecomposition is: λv=C F v. Secondly, based on the mapping result φ(x i ), the first eigenvalue λ and the eigenvector v are inner-producted to obtain the projection value λ[φ(x i )·v], the formula for calculating the inner product is: λ[φ(x i )·v]=φ(x i )·C F v, where λ[φ(x i )·v] is the projection value, φ(x i ) is the mapping result of the i-th variable, C F is the covariance matrix, v is the eigenvector; if the first eigenvalue λ is not 0, then in the high-dimensional space F, the eigenvector v can be mapped to the data set φ(x i ) is linearly represented as: At this time, in the high-dimensional space F, the calculation process of the projection value is: The calculation process of the projection value is simplified by using a kernel function, and a simplified equation nλα=Kα is obtained, where λ is the second eigenvalue, α is the eigenvector (in a high-dimensional space) of the kernel function matrix K, and K is the kernel function matrix. Finally, by solving nλα=Kα, the second eigenvalue λ of the kernel function matrix K and its corresponding eigenvector α can be obtained, and a feature subset can be screened out from the disconnection fault feature set based on the second eigenvalue λ.

[0108] In some embodiments, the third calculation unit includes: a calculation subunit and a screening subunit; the calculation subunit is used to calculate the corresponding contribution rate for each second eigenvalue; the screening subunit is used to determine the corresponding cumulative contribution rate based on the contribution rate, determine the principal component based on the contribution rate and the cumulative contribution rate, and screen out the feature subset from the disconnection fault feature set based on the principal component. Specifically, first, after obtaining the second eigenvalue λ, it is necessary to calculate the corresponding contribution rate Con(λ) for each second eigenvalue λ. i ), the contribution rate calculation formula is: In the formula, Con(λ i ) is the contribution rate of the i-th second eigenvalue, then the contribution rates of the second eigenvalues ​​are accumulated to obtain the cumulative contribution rate, and finally, the contribution rate Con(λ i ) is the largest and the cumulative contribution rate reaches or exceeds a preset threshold (such as 90%) as the characteristic vector corresponding to the second eigenvalue as the principal component, and based on these principal components, a feature subset is screened out from the line break fault feature set for subsequent fault analysis or model training.

[0109] It should be noted that the mapping result φ(x i ) is projected in the direction of each element to transform the data from the original feature space into a new space composed of the most important features (i.e., principal components). The calculation formula is:

[0110] In this way, by screening out a feature subset from the disconnection fault feature set, the most useful information for fault location can be extracted, which facilitates subsequent mining of the mapping relationship between the disconnection features in the feature subset and the disconnection fault location, thereby improving the efficiency and accuracy of subsequent fault location.

[0111] The determination module 300 is used to input the feature subset into a preset fault location model, and use a particle swarm optimization algorithm to search for optimal model parameters of the fault location model, and determine a target fault location model based on the optimal model parameters;

[0112] In some embodiments, the preset fault location model is constructed based on the BP neural network. When constructing the fault location model, it is necessary to define the model output as a disconnection upstream and downstream judgment logic quantity, wherein the disconnection upstream and downstream judgment logic quantity is the position of the disconnection fault point relative to the monitoring point. On the premise that a disconnection fault occurs in a known line, when the neural network output is 1, it is marked as the disconnection is located upstream of the monitoring point, and when the neural network output is -1, it is marked as the disconnection is located downstream of the monitoring point.

[0113] In some embodiments, the fault location model includes an input layer, a hidden layer and an output layer. Figure 4 As shown, the hidden layer is activated by the logsig function, the output layer is activated by the tansig function, the extracted features and their labels are sent to the input layer, the fault location prediction label is output from the output layer, and the real label is compared with the predicted label of the model. A large amount of data is used for training, the prediction error of the model is calculated, and the model with the smallest prediction error is saved. This model is used as the distribution network fault location model.

[0114] In some embodiments, the determination module 300 includes: an initialization unit, a fourth calculation unit, and a fifth calculation unit; the initialization unit is used to initialize the initial velocity and initial position of the particle within a preset particle search range; the fourth calculation unit is used to determine the fitness function according to the initial model parameters of the fault location model; the fifth calculation unit is used to iteratively calculate the fitness value based on the fitness function, the initial velocity, and the initial position until the termination condition is met, and output the position matrix of the particle representing the optimal model parameters, wherein the position matrix and velocity matrix of the particle are updated in each iteration. Specifically, first, the size of the particle group (i.e., the number of particles) is set, and each particle P i Each is a separate neural network with independent parameters, randomly determining each particle P i The initial velocity v i and the initial position x i , where position represents a combination of model parameters, and speed represents the direction and magnitude of parameter update; secondly, the error norm of the BP neural network is used to measure the fitness function: min||T sim -T train ||, where T sim Represents the discrimination result of the neural network under the current weight and threshold, T train represents the actual discrimination result of the neural network training set to determine the fitness function; then, for each particle P i , use its current position (i.e., model parameters) to calculate the fitness value. If the fitness value of the current particle is better than its historical best position (i.e., the best fitness value in the previous iteration), then update the P value of the particle.best,i , and find the particle with the best fitness value in the entire particle group, and update the G of the particle best Finally, based on the current position, current speed, P best,i and G best Iteratively update the speed and position of each particle. After each iteration, check whether the termination condition is met. If the termination condition is met, output the particle position matrix representing the optimal model parameters. This position matrix contains the optimal parameter settings of the fault localization model. The optimal model parameters include the optimal weights and the optimal thresholds. The optimal weight network is assigned to the target fault localization model for fault localization.

[0115] In some embodiments, the termination condition may be setting a preset error accuracy or a preset number of iterations, and when the error accuracy reaches the requirement or reaches a maximum number of iterations, it is determined that the termination condition is satisfied.

[0116] In this way, optimizing the fault location model through the particle swarm optimization algorithm can ensure that the model has better generalization ability and robustness when dealing with complex and nonlinear problems, thereby improving the accuracy and reliability of fault location.

[0117] In some embodiments, the calculation formula for updating the position matrix and velocity matrix of the particle is specifically: In the formula, and are the velocity and position of the ith particle at the k+1th iteration, ω is the inertia weight, and are the velocity and position of the ith particle at the kth iteration, c 1 is the individual particle learning factor, c 2 is the particle group learning factor, P best,i is the historical optimal position of the i-th particle, G best is the historical optimal position of the particle population.

[0118] The positioning module 400 is used to collect electrical characteristic quantities on one side of the distribution network, input the electrical characteristic quantities into the target fault positioning model to determine the output model output result, and determine the location of the single-phase disconnection fault based on the model output result.

[0119] In some embodiments, electrical characteristic quantities on one side of the distribution network are acquired through sensors or data acquisition units, wherein the electrical characteristic quantities may be, but are not limited to, the phase difference between the fault phase voltage and the negative-sequence current, the negative-sequence voltage and current phase difference, the line voltage amplitude, and the phase voltage amplitude; thereafter, the electrical characteristic quantities are preprocessed, and the preprocessed electrical characteristic quantities are input into the target fault location model to obtain the model output result, and the location of the single-phase line break fault is determined based on the model output result. For example, when the model output result is in the interval [0.5, 1], the single-phase line break fault is considered to be located upstream of the monitoring point; when the model output result is in the interval [-1, -0.5], the single-phase line break fault is considered to be located downstream of the monitoring point.

[0120] The following is a verification and explanation of the technical effects used in this method. This embodiment uses a traditional technical solution to conduct a comparative test with the method of the present invention, and compares the test results by means of scientific demonstration to verify the real effect of this method. Figure 5 As shown in the circuit diagram for fault location of a 10kV distribution network, 14 groups of single-phase line break faults are simulated at different locations, and electrical feature quantities are obtained from 11 monitoring points. After KPCA dimensionality reduction, the electrical feature quantities are input into the model for training. In the data set, the ratio of the training set to the test set is 7:3, and the ratio of the upstream fault data to the downstream fault data contained in the feature data is approximately 5:3. Figure 6 It is a schematic diagram for comparing the true value and the predicted value of the target fault location model. It can be seen that the predicted value tested by the method provided by the present invention is very close to the true value of the fault location, and the purpose of section location can be achieved.

[0121] Table 1 is an evaluation of the results predicted by the target fault location model and the traditional single criterion model. From Table 1, it can be seen that the target fault location model proposed in the embodiment of the present application has excellent performance in prediction accuracy, indicating that the prediction results of the target fault location model are more accurate and can effectively extract feature information, so that the neural network learns the feature quantity that has a greater impact on fault location, thereby improving the accuracy of fault location.

[0122] Table 1: Comparison of the accuracy of the target fault location model and the traditional single criterion

[0123]

[0124]

[0125] The embodiment of the present application can comprehensively and accurately obtain the feature set when the fault occurs by collecting the line-break fault feature set on one side of the distribution network, which is convenient for subsequent analysis and fault location; by mapping the fault feature set to a high-dimensional feature space, the nonlinear relationship and hidden features of the data can be better revealed, which is convenient for subsequent extraction of useful feature subsets; by screening out feature subsets from the line-break fault feature set, the most useful information for fault location can be extracted, which is convenient for subsequent mining of the mapping relationship between the line-break features in the feature subset and the line-break fault location, thereby improving the efficiency and accuracy of subsequent fault location; by training the fault location model with feature subsets, the mapping relationship between the line-break features in the feature subset and the line-break fault location can be mined, and by optimizing the fault location model with a particle swarm optimization algorithm, it can be ensured that the model has better generalization ability and robustness when dealing with complex and nonlinear problems, thereby improving the accuracy and reliability of fault location; by determining the location of the single-phase line-break fault through the target fault model, the location of the single-phase line-break fault can be quickly and accurately determined. Compared with the prior art, the present application can accurately and quickly locate the location of the line-break fault.

[0126] The information interaction, execution process, etc. between the modules in the above-mentioned single-phase line break fault location system based on BP neural network are based on the same concept as the embodiment of the single-phase line break fault location method based on BP neural network in the first aspect of the present invention, and the technical effects achieved are basically the same. For specific contents, please refer to the description in the first embodiment of the method of the present invention, and will not be repeated here.

[0127] The device embodiments described above are merely illustrative, wherein the modules described as separate components may or may not be physically separated, i.e., may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the method of this embodiment.

[0128] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the single-phase line break fault location method based on the BP neural network as described in the first embodiment above is implemented.

[0129] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-monitorable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0130] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application.

[0131] It is particularly pointed out that for those skilled in the art, any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A single-phase line break fault location method based on BP neural network, characterized in that: include: Collecting a line-break fault feature set on one side of the distribution network, wherein the line-break fault feature set includes a phase difference between the fault phase voltage and the negative-sequence current, a negative-sequence voltage and current phase difference, a line voltage amplitude, and a phase voltage amplitude; Mapping the fault feature set to a high-dimensional feature space through a preset kernel function to obtain a mapping result, and filtering out a feature subset from the disconnection fault feature set based on the mapping result; Inputting the feature subset into a preset fault location model, and using a particle swarm optimization algorithm to search for optimal model parameters of the fault location model, and determining a target fault location model based on the optimal model parameters; The electrical characteristic quantity of one side of the distribution network is collected, and the electrical characteristic quantity is input into the target fault location model to determine the output model output result, and the location of the single-phase disconnection fault is determined based on the model output result.

2. The single-phase disconnection fault location method based on BP neural network according to claim 1 is characterized in that: The filtering out of a feature subset from the disconnection fault feature set based on the mapping result is specifically: Calculating a covariance matrix of the mapping result, and performing eigendecomposition on the covariance matrix to obtain a first eigenvalue and a corresponding eigenvector; Perform inner product calculation based on the mapping result, the first eigenvalue and the eigenvector to obtain a projection value, and simplify the calculation process of the projection value by using the kernel function to obtain a simplified equation; The simplified equation is solved to obtain a second characteristic value, and a characteristic subset is screened out from the disconnection fault characteristic set based on the second characteristic value.

3. The single-phase disconnection fault location method based on BP neural network according to claim 2 is characterized in that: The filtering out of a feature subset from the disconnection fault feature set based on the second feature value is specifically: Calculating a corresponding contribution rate for each of the second eigenvalues; A corresponding cumulative contribution rate is determined based on the contribution rate, a principal component is determined based on the contribution rate and the cumulative contribution rate, and a feature subset is screened out from the line break fault feature set based on the principal component.

4. The single-phase disconnection fault location method based on BP neural network according to claim 1 is characterized in that: The particle swarm optimization algorithm is used to search for the optimal model parameters of the fault location model, specifically: Initialize the initial velocity and initial position of the particle within the preset particle search range; Determining a fitness function according to initial model parameters of the fault location model; The fitness value is iteratively calculated based on the fitness function, the initial velocity and the initial position until a termination condition is met, and a position matrix of the particle representing the optimal model parameters is output, wherein the position matrix and the velocity matrix of the particle are updated in each iteration.

5. The single-phase disconnection fault location method based on BP neural network according to claim 4 is characterized in that: The calculation formula for updating the position matrix and velocity matrix of the particle is specifically: In the formula, and are the velocity and position of the ith particle at the k+1th iteration, ω is the inertia weight, and are the speed and position of the ith particle at the kth iteration, c1 is the individual learning factor of the particle, c2 is the group learning factor of the particle, P best,i is the historical optimal position of the i-th particle, G best is the historical optimal position of the particle population.

6. A single-phase line break fault location device based on BP neural network, characterized in that: include: Acquisition module, screening module, determination module and positioning module; The acquisition module is used to acquire a set of line-break fault characteristics on one side of the distribution network, wherein the set of line-break fault characteristics includes a phase difference between the fault phase voltage and the negative-sequence current, a negative-sequence voltage-current phase difference, a line voltage amplitude, and a phase voltage amplitude; The screening module is used to map the fault feature set to a high-dimensional feature space through a preset kernel function to obtain a mapping result, and screen out a feature subset from the disconnection fault feature set based on the mapping result; The determination module is used to input the feature subset into a preset fault location model, and use a particle swarm optimization algorithm to search for optimal model parameters of the fault location model, and determine a target fault location model based on the optimal model parameters; The positioning module is used to collect electrical characteristic quantities on one side of the distribution network, input the electrical characteristic quantities into the target fault positioning model to determine the output model output result, and determine the location of the single-phase disconnection fault based on the model output result.

7. The single-phase disconnection fault location device based on BP neural network according to claim 6 is characterized in that: The screening module includes: a first calculation unit, a second calculation unit and a third calculation unit; The first calculation unit is used to calculate the covariance matrix of the mapping result, and perform eigendecomposition on the covariance matrix to obtain a first eigenvalue and a corresponding eigenvector; The second calculation unit is used to perform inner product calculation based on the mapping result, the first eigenvalue and the eigenvector to obtain a projection value, and simplify the calculation process of the projection value by using the kernel function to obtain a simplified equation; The third calculation unit is used to solve the simplified equation to obtain a second eigenvalue, and filter out a feature subset from the line break fault feature set based on the second eigenvalue.

8. The single-phase disconnection fault location device based on BP neural network according to claim 7 is characterized in that: The third calculation unit includes: a calculation subunit and a screening subunit; The calculation subunit is used to calculate the corresponding contribution rate for each second eigenvalue; The screening subunit is used to determine the corresponding cumulative contribution rate based on the contribution rate, determine the principal component based on the contribution rate and the cumulative contribution rate, and screen out a feature subset from the line break fault feature set based on the principal component.

9. The single-phase disconnection fault location device based on BP neural network according to claim 6, characterized in that: The determination module includes: an initialization unit, a fourth calculation unit and a fifth calculation unit; The initialization unit is used to initialize the initial velocity and initial position of the particle within a preset particle search range; The fourth calculation unit is used to determine the fitness function according to the initial model parameters of the fault location model; The fifth calculation unit is used to iteratively calculate the fitness value based on the fitness function, the initial velocity and the initial position until a termination condition is met, and output a position matrix of the particle representing the optimal model parameters, wherein the position matrix and velocity matrix of the particle are updated in each iteration.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the single-phase line break fault location method based on BP neural network as described in any one of claims 1 to 5 is implemented.

Images