Distribution network overhead line fault identification method, device, equipment, medium and program

Abstract

The invention provides a distribution network overhead line fault identification method, device and equipment, a medium and a program. Relates to the technical field of fault identification. The method comprises the following steps: acquiring a target signal on a distribution network overhead line; determining a target parameter of the VMD in a feasible parameter range of the VMD based on GOA; performing VMD decomposition on the target signal based on the target parameter to obtain a signal feature of the target signal; inputting the signal characteristics into a fault identification model to obtain a fault identification result of the distribution network overhead line; wherein the fault identification model is obtained by training an Xgboost algorithm. According to the invention, the method achieves the effect of improving the fault recognition accuracy of the overhead line of the distribution network.

Description

Technical Field

[0001] The present application relates to the field of fault identification technology, and in particular to a method, device, equipment, medium and program for identifying faults in distribution network overhead lines. Background Art

[0002] Overhead distribution lines are a crucial component of power systems, transmitting energy from substations to end users via overhead conductors. Identifying faults in overhead distribution lines helps ensure safe grid operation, prevent escalating faults, and avoid equipment damage and accidents. It also optimizes grid operation and maintenance strategies and reduces costs.

[0003] In the related art, fault detection of distribution network overhead lines is performed through a fault identification method based on variational mode decomposition (VMD) and support vector machine (SVM), but there are cases of misjudgment or missed judgment, and the accuracy of fault detection is low.

[0004] Based on this, a distribution network overhead line fault identification scheme with high accuracy is proposed. Summary of the Invention

[0005] The present application provides a distribution network overhead line fault identification method, device, equipment, medium and program to achieve the effect of improving the accuracy of distribution network overhead line fault identification.

[0006] In a first aspect, the present application provides a method for identifying a distribution network overhead line fault, comprising:

[0007] Acquire target signals on overhead lines of distribution networks;

[0008] Based on the Gazelle Optimization Algorithm (GOA), the target parameters of VMD are determined within the feasible parameter range of VMD.

[0009] Perform VMD decomposition on the target signal based on the target parameters to obtain the signal characteristics of the target signal;

[0010] The signal features are input into the fault identification model to obtain the fault identification results of the distribution network overhead lines; among them, the fault identification model is trained by the extreme gradient boosting Xgboost algorithm.

[0011] In one possible implementation, determining target parameters of VMD within a feasible parameter range of VMD based on GOA includes:

[0012] Within the feasible parameter range of VMD, multiple individuals that meet the GOA population size requirements are randomly generated to obtain the GOA candidate population; each individual includes a set of VMD parameter combinations;

[0013] According to the feasible parameter range and candidate population, the candidate population is updated through the GOA update strategy to obtain the target population;

[0014] The target parameters of VMD are determined according to the fitness of individuals in the target population.

[0015] In one possible implementation, based on the feasible parameter range and the candidate population, the candidate population is updated using the GOA update strategy to obtain the target population, further comprising:

[0016] For each individual in the candidate population, the VMD parameters corresponding to each individual are used to perform VMD decomposition on the target signal to obtain the modal components of each individual;

[0017] Determine the corresponding fitness through the modal component of each individual;

[0018] According to the fitness of each individual in the candidate population, determine whether the termination condition is met;

[0019] If the termination condition is not met, the candidate population is updated according to the population update strategy of GOA and the fitness of each individual within the feasible parameter range until the termination condition is met;

[0020] If the termination condition is met, the target population is obtained.

[0021] In one possible implementation, for each individual in the candidate population, VMD decomposition is performed on the target signal using the VMD parameters corresponding to each individual to obtain the modal components of each individual, including:

[0022] For each individual in the candidate population, initialize the modal component to obtain the initial component;

[0023] The target signal is decomposed by VMD using the VMD parameters and initial components corresponding to each individual to obtain the modal components of each individual.

[0024] In one possible implementation, the GOA population update strategy includes at least one of the following:

[0025] Update strategies for random perturbations following Brownian motion;

[0026] Update strategy based on Levy flight;

[0027] Update strategies that follow random perturbations and cumulative effects of Brownian motion.

[0028] In one possible implementation, the fault identification model is trained in the following manner:

[0029] Obtain historical signal data of distribution network overhead lines and perform VMD decomposition on the historical signal data to obtain target features;

[0030] Encode the target features through the Transformer model to obtain the encoded feature space representation;

[0031] The Xgboost model is trained based on the feature space representation to obtain the fault recognition model.

[0032] In a second aspect, the present application provides a distribution network overhead line fault device, comprising:

[0033] An acquisition module is used to acquire target signals on overhead lines of the distribution network;

[0034] The processing module is used to determine the target parameters of VMD within the feasible parameter range of VMD based on GOA; perform VMD decomposition on the target signal based on the target parameters to obtain the signal characteristics of the target signal; and input the signal characteristics into the fault identification model to obtain the fault identification results of the distribution network overhead line; wherein the fault identification model is trained based on the Xgboost algorithm.

[0035] In a possible implementation, the processing module is specifically configured to:

[0036] Within the feasible parameter range of VMD, multiple individuals that meet the GOA population size requirements are randomly generated to obtain the GOA candidate population; each individual includes a set of VMD parameter combinations;

[0037] According to the feasible parameter range and candidate population, the candidate population is updated through the GOA update strategy to obtain the target population;

[0038] The target parameters of VMD are determined according to the fitness of individuals in the target population.

[0039] In a possible implementation manner, the processing module is further configured to:

[0040] For each individual in the candidate population, the VMD parameters corresponding to each individual are used to perform VMD decomposition on the target signal to obtain the modal components of each individual;

[0041] Determine the corresponding fitness through the modal component of each individual;

[0042] According to the fitness of each individual in the candidate population, determine whether the termination condition is met;

[0043] If the termination condition is not met, the candidate population is updated according to the population update strategy of GOA and the fitness of each individual within the feasible parameter range until the termination condition is met;

[0044] If the termination condition is met, the target population is obtained.

[0045] In a possible implementation manner, the processing module is further configured to:

[0046] For each individual in the candidate population, initialize the modal component to obtain the initial component;

[0047] The target signal is decomposed by VMD using the VMD parameters and initial components corresponding to each individual to obtain the modal components of each individual.

[0048] In one possible implementation, the GOA population update strategy includes at least one of the following:

[0049] Update strategies for random perturbations following Brownian motion;

[0050] Update strategy based on Levy flight;

[0051] Update strategies that follow random perturbations and cumulative effects of Brownian motion.

[0052] In one possible implementation, the fault identification model is trained in the following manner:

[0053] Obtain historical signal data of distribution network overhead lines and perform VMD decomposition on the historical signal data to obtain target features;

[0054] Encode the target features through the Transformer model to obtain the encoded feature space representation;

[0055] The Xgboost model is trained based on the feature space representation to obtain the fault recognition model.

[0056] In a third aspect, the present application provides an electronic device, comprising: a memory, a processor;

[0057] Memory stores computer-executable instructions;

[0058] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0059] In a fourth aspect, the present application provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by a processor, they are used to implement the above first aspect and / or various possible implementation methods of the first aspect.

[0060] In a fifth aspect, the present application provides a computer program product, comprising a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementations of the first aspect.

[0061] The distribution network overhead line fault identification method, device, equipment, medium and program provided by the present application obtain the target signal on the distribution network overhead line. Through GOA, the target parameters of VMD are determined within the feasible parameter range of VMD, so that when the target signal is subjected to VMD decomposition, the parameters can be adaptively adjusted to adapt to the signal characteristics in different distribution network overhead lines. The target signal is subjected to VMD decomposition based on the target parameters to obtain the signal characteristics of the target signal; by selecting appropriate target parameters, the problem of VMD decomposition limitation caused by poor VMD parameter selection is avoided, and the characterization capability of the signal characteristics of the target signal is improved, so that the signal characteristics of the target signal can better express the signal characteristics in the distribution network overhead line. Afterwards, the signal characteristics are input into the fault identification model to obtain the fault identification result of the distribution network overhead line, which can improve the accuracy of fault identification and more effectively perform fault classification and fault location. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0063] Figure 1 A schematic diagram of the structure of the Transformer model provided in the embodiment of the present application;

[0064] Figure 2 A schematic diagram of a method for identifying a distribution network overhead line fault according to an embodiment of the present application;

[0065] Figure 3 Schematic diagram of the process of the distribution network overhead line fault identification method provided in the embodiment of the application Figure 1 ;

[0066] Figure 4 Schematic diagram of the process of the distribution network overhead line fault identification method provided in the embodiment of the application Figure 2 ;

[0067] Figure 5 Schematic diagram of the process of the distribution network overhead line fault identification method provided in the embodiment of the application Figure 3 ;

[0068] Figure 6 A schematic diagram of the structure of a distribution network overhead line fault identification device provided in an embodiment of the present application;

[0069] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application.

[0070] The above drawings illustrate specific embodiments of the present application, which will be described in more detail below. These drawings and the textual description are not intended to limit the scope of the present application in any way, but rather to illustrate the concepts of the present application to those skilled in the art by reference to specific embodiments. DETAILED DESCRIPTION

[0071] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0072] In the related art, in the traditional distribution network overhead line fault identification method, the key parameters used in VMD decomposition of signal data are usually determined by fixed values or trial-and-error methods based on experience, resulting in the inability to fully consider the characteristics of different distribution network overhead lines, fault types, and changes in signals under the operating environment when VMD decomposition of signal data, resulting in uneven decomposition effects of VMD decomposition. When using fixed VMD key parameters for VMD decomposition, it is impossible to accurately decompose signal data into modal components with obvious physical meaning. At the same time, there is a lack of deep-level feature mining of signal data, and it is impossible to fully utilize the association between the time series information and multimodal features in the signal data. As a result, in the fault identification process, the feature differentiation between different fault types is not high, and misjudgment or missed judgment is prone to occur, resulting in low fault identification accuracy.

[0073] The distribution network overhead line fault identification method provided in the embodiment of the present application is based on GOA and determines the target parameters of VMD within the feasible parameter range of VMD. By performing VMD decomposition on the target signal through the target parameters, the signal can be accurately decomposed into modal components with obvious physical meanings. Afterwards, the signal characteristics of the target signal are obtained based on the modal components, so that when the target signal is subjected to VMD decomposition, the parameters can be adaptively adjusted to adapt to the signal characteristics in different distribution network overhead lines. The signal characteristics of the target signal are input into the fault identification model trained by the Xgboost algorithm. The fault identification model can effectively use the signal characteristics for fault identification, thereby obtaining the fault identification results of the distribution network overhead lines, thereby reducing the misjudgment or missed judgment phenomenon during fault identification and improving the accuracy of fault identification.

[0074] Figure 1 This is a schematic diagram of the structure of the Transformer model provided in the embodiment of this application. Figure 1As shown in the figure, the Transformer model mainly includes: input layer encoding 11, output layer encoding 12, position encoding 18, encoder 10, decoder 19, linear layer 16 and activation function layer 17, among which: the encoder 10 and decoder 19 also include a multi-head attention mechanism 13, sum normalization 14 and a feedforward network 15 respectively.

[0075] The input layer encoding 11 converts the original input data into a numerical vector representation, laying the foundation for semantic processing; the position encoding 18 embeds temporal information into the numerical vector representation to make up for the lack of sequential perception caused by parallel computing. The output layer encoding 12 converts the original target data into a vector form. The encoder 10 performs high-level feature abstraction on the input vector, captures global dependencies through the multi-head attention mechanism 13, and introduces nonlinear transformations through the feedforward network 15; the decoder 19 gradually predicts the next position of the target sequence based on the encoder output and the generated sequence. Finally, the linear layer 16 maps the decoder output to the target vocabulary dimension, generates word scores, and then converts them into probability distributions by the activation function layer 17, outputting the probability of the predicted result. In related technologies, the activation function used in the activation function layer 17 of the Transformer model is the Softmax classifier. Optionally, the Softmax classifier used in the activation function layer 17 is replaced with an Xgboost classifier.

[0076] Figure 2 This is a schematic diagram of a method for identifying a distribution network overhead line fault provided in an embodiment of the present application. Figure 2 As shown, the specific application scenario of the present application includes a distribution network overhead line 21 and a computing center 22 , wherein a collection device 211 is deployed in the distribution network overhead line 21 .

[0077] The acquisition device 211 can collect multi-dimensional operating data such as current, voltage, temperature, and vibration of the distribution network overhead lines in real time to obtain signal data. The acquisition device 211 can communicate with the computing center 22 via wireless or wired communication technology and transmit the collected signal data to the computing center 22. As the core processing unit, the computing center 22 can analyze the signal data, accurately identify the fault status of the distribution network overhead lines, and obtain the fault status of the distribution network overhead lines. At the same time, the computing center 22 can also generate a diagnostic report containing the fault type, location, and severity.

[0078] The following specific embodiments describe in detail the technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0079] Figure 3Schematic diagram of the process of the distribution network overhead line fault identification method provided in the embodiment of the application Figure 1 .like Figure 3 As shown, the method includes:

[0080] S301: Acquire a target signal on a distribution network overhead line.

[0081] Overhead distribution lines are used to distribute electrical energy within a power system, typically from substations to consumers. Target signals are signals collected from these lines for fault detection and analysis. Optionally, the target signals are the voltage and / or current on these lines.

[0082] Optionally, multiple signal injection devices and multiple acquisition devices are deployed in the distribution network overhead lines. The signal injection devices use a signal generator to inject an original injection signal into the distribution network overhead lines. The original injection signal is a pulse signal with a specific frequency and amplitude. The original injection signal can propagate in the distribution network overhead lines. When the original injection signal encounters a fault point or a point where the line impedance changes in the distribution network overhead lines, a reflected signal is generated. The acquisition device can collect the reflected signal and the original injection signal on the distribution network overhead lines. The reflected signal collected by the acquisition device is a sequence of voltage or current values that change over time.

[0083] For example, a signal injection device and a collection device are installed at the starting point of the distribution network overhead line, and a collection device is installed at the midpoint and end point of the distribution network overhead line. The signal injection device at the starting point injects an original injection signal with a frequency of 50Hz and an amplitude of 1V. The three collection devices collect the signals and record the time-varying sequence of voltage values to obtain the target signal. Optionally, the above process can be repeated in sunny and rainy environments, as well as in load-variable environments, to collect target signals under different conditions.

[0084] 302. Based on the GOA, determine target parameters of the VMD within a feasible parameter range of the VMD.

[0085] GOA is an optimization algorithm based on gazelle behavior. It simulates the gazelle's swift running and agile mobility to find the optimal solution to complex optimization problems. VMD is an adaptive signal decomposition method used to decompose a complex signal into a set of K intrinsic mode functions (IMFs) with different frequencies and center frequencies. This decomposes the complex signal into K IMF components. IMF components are also called modal components. The number of K is determined by the number of modal components. The target parameters are a set of VMD parameters determined through GOA optimization.

[0086] First, parameters such as the population size, number of iterations, and search space of the GOA are set. Next, the parameters are encoded, converting the VMD parameters into individuals within the GOA. VMD parameters can include center frequency, penalty factor, modal components, and / or bandwidth. Each individual in the GOA represents a set of VDM parameters. The fitness of each individual is then evaluated using a fitness function. This fitness function can be a performance metric based on the decomposition results, such as signal reconstruction error or the physical significance of the modal components. Based on their fitness, high-performing individuals are selected for reproduction, simulating the rapid running and agile movement of gazelles. In the crossover and mutation phase, these selected individuals undergo crossover and mutation to generate new individuals, simulating the evolution of a gazelle population. In the population update phase, individuals with lower fitness are replaced with newly generated individuals, updating the population. In the iteration phase, the fitness of each individual is repeatedly calculated and the population is updated based on the fitness until a termination criterion is reached, thereby determining the target parameters for VMD.

[0087] Optionally, determining the target parameters of VMD within the feasible parameter range of VMD by GOA may include a penalty factor and a modal component.

[0088] S303 : Perform VMD decomposition on the target signal based on the target parameters to obtain signal characteristics of the target signal.

[0089] The target signal on the distribution network overhead line is used as the input of the VMD algorithm, and the parameters of VMDE are set according to the target parameters. VMD decomposition is performed to obtain the modal components of the target signal.

[0090] For example, assuming that the target parameters include the penalty factor α and the modal component K, the target signal on the distribution network overhead line includes the injection reflection signal f(t) composed of the injection signal and the reflection signal. VMD can decompose the injection reflection signal f(t) into K components with a center frequency ω k The modal component u k (t), and make K modal components u k The sum of the bandwidths of (t) is minimized, thereby achieving effective signal decomposition. Optionally, the injected reflection signal f(t) is decomposed into K components with a center frequency ω by using the VMD determination formula. k The modal component u k (t). Furthermore, the VMD determination formula may be a variational constraint determination formula for VMD. The variational constraint determination formula for VMD is:

[0091]

[0092] in, represents the partial derivative with respect to t; δ(t) is the unit impulse function; * represents the convolution operation; Represents the square of the L2 norm, used to measure error; constraints To ensure that all modal components u k The sum of the injected reflected signal f(t) is equal to the injected reflected signal f(t).

[0093] In order to simplify the solution process, the variational constraint of VMD can be converted into an unconstrained variational problem by introducing the penalty factor α and the Lagrange multiplier λ, and then the unconstrained variational determination formula of VMD is obtained as follows:

[0094]

[0095] The modal component u is iteratively updated using the alternating direction multiplier method k (t), center frequency ω k and Lagrange multiplier λ until the convergence condition is met. Modal component u k The update formula of (t) is:

[0096]

[0097] Where n is the number of iterations.

[0098] Center frequency ω k The update formula is:

[0099]

[0100] in, represents the kth modal component u in the nth iteration k The spectrum of (t).

[0101] The convergence condition is that the change of the modal component between two adjacent iterations is less than the preset threshold ε. The convergence condition is expressed as:

[0102]

[0103] in, represents the modal component at round n+1; represents the modal component at n rounds.

[0104] Through the iterative calculation of the above VMD, K modal components u are obtained based on the criterion of minimum envelope entropy k (t).

[0105] Optionally, the K modal components u of the extracted target signal can be k (t) Signal characteristics as target signal.

[0106] Optionally, the K modal components u of the target signal can also be k(t) and use the extracted target features as the signal features of the target signal. k The time domain indicators and frequency domain indicators of (t) are analyzed to obtain the target characteristics of each modal component.

[0107] The target characteristics of each modal component may include mean, variance, peak, kurtosis, effective value, crest factor, impulse factor, shape factor, and margin factor.

[0108] For example, the K modal components u can be obtained by determining the formula corresponding to the target feature. k The corresponding target features are obtained in (t).

[0109] Specifically, the modal component u k (t) mean μ k The formula for determining is:

[0110]

[0111] Where N represents the signal u k The total number of sample points of (t); μ k (i) represents the value of the kth modal component at the i-th sample point.

[0112] Modal component u k (t) Variance The formula for determining is:

[0113]

[0114] in, represents the variance of the kth modal component.

[0115] Modal component u k (t) Peak value P k The formula for determining is:

[0116] P k =max(u k (i))

[0117] Among them, max means taking u k The maximum value operation in (i).

[0118] Modal component u k (t) Kurtosis K k The formula for determining is:

[0119]

[0120] Modal component u k (t)RMS k The formula for determining is:

[0121]

[0122] Modal component u k (t) Peak factor CF k The formula for determining is:

[0123]

[0124] Modal component u k (t) Pulse factor IF k The formula for determining is:

[0125]

[0126] Modal component u k (t) Form factor FF k The formula for determining is:

[0127]

[0128] Modal component u k (t) Margin factor MF k The formula for determining is:

[0129]

[0130] By the above determination formula, each modal component u k (t) corresponds to 9 features. Each modal component u can be obtained by integrating the 9 features of mean, variance, peak, kurtosis, effective value, peak factor, impulse factor, shape factor and margin factor. k (t) corresponds to the target feature. Afterwards, the signal feature space formed by the signal features of all modal components is used as the signal feature of the target signal, where the dimension of the signal feature space is K×9, where K is the number of modal components.

[0131] S304: Input the signal features into a fault identification model to obtain a fault identification result of the distribution network overhead line; wherein the fault identification model is obtained by training the Xgboost algorithm.

[0132] Xgboost is an efficient machine learning algorithm based on the gradient boosting tree. It improves prediction accuracy by building multiple weak prediction models and combining them into a strong prediction model. Using the Xgboost algorithm to train a fault identification model enables accurate identification and diagnosis of distribution network overhead line faults. Signal features are input into the trained fault identification model, which then outputs a fault identification result, including both normal and faulty conditions. This fault identification result provides a reliable basis for maintenance and inspection of distribution network overhead lines.

[0133] The distribution network overhead line fault identification method provided in the embodiment of the present application obtains the target signal on the distribution network overhead line; uses GOA to determine the target parameters of VMD, improves the accuracy of VMD signal decomposition, adapts to the characteristics of the target signal in different distribution network overhead lines, and enables the comprehensive extraction of the characteristics of the target signal during decomposition. The target signal is subjected to VMD decomposition based on the determined target parameters, so that the signal characteristics of the target signal have stronger characterization capabilities. Afterwards, the signal characteristics are input into the fault identification model to obtain the fault identification results of the distribution network overhead line; the fault identification model trained by the Xgboost algorithm has high accuracy and generalization ability, can reduce fault identification misjudgments and missed judgments, and improve the accuracy of fault identification.

[0134] Figure 4 Schematic diagram of the process of the distribution network overhead line fault identification method provided in the embodiment of the application Figure 2 .like Figure 4 As shown, this embodiment Figure 3 Based on the embodiment, a distribution network overhead line fault identification method is described in detail. In this method, the above step S302 further includes:

[0135] S401. Within the feasible parameter range of VMD, a plurality of individuals that meet the GOA population size requirement are randomly generated to obtain a candidate population of GOA; each individual includes a set of VMD parameter combinations.

[0136] Obtain a feasible parameter range for the VMD. Optionally, the feasible parameter range for the VMD includes a feasible parameter range for the center frequency, a feasible parameter range for the bandwidth, a feasible parameter range for the penalty factor, and / or a feasible parameter range for the modal component. Exemplarily, the feasible parameter range for the VMD may include: a feasible parameter range for the center frequency and a feasible parameter range for the bandwidth; or, the feasible parameter range for the VMD may also include: a feasible parameter range for the bandwidth and a feasible parameter range for the penalty factor; or, the feasible parameter range for the VMD may also include: a feasible parameter range for the penalty factor and a feasible parameter range for the modal component. Optionally, the feasible parameter range for the VMD may be obtained by reading the maximum and minimum values of the feasible parameters for the VMD stored in a database; or, the feasible parameter range for the VMD may be obtained by obtaining the maximum and minimum values of the feasible parameters for the VMD set by an operator; or, the feasible parameter range for the VMD may be obtained by obtaining a set of feasible parameters for the VMD in a database.

[0137] The GOA population requirement refers to the number of candidate solutions that must be defined before GOA begins searching for the optimal solution. The number of candidate solutions can be thought of as the size of a gazelle population or the number of individual gazelles. The GOA population requirement determines the size and diversity of the GOA algorithm's search space. Optionally, the GOA population requirement can be any positive integer, such as 5, 10, 17, or 44.

[0138] Within the feasible parameter range of VMD, a plurality of individuals that meet the GOA population size requirement are randomly generated. Each individual includes a set of VMD parameter combinations. The VMD parameter combinations should include parameters within the feasible parameter range of VMD. Exemplarily, when the feasible parameter range of VMD includes the feasible parameter range of the center frequency and the feasible parameter range of the bandwidth, the VMD parameter combination should include the center frequency and the bandwidth, i.e., each individual includes a set of combinations of the center frequency and the bandwidth; when the feasible parameter range of VMD includes the feasible parameter range of the penalty factor and the feasible parameter range of the modal component, the VMD parameter combination should include the penalty factor and the modal component, i.e., each individual includes a set of combinations of the penalty factor and the modal component.

[0139] For example, if the number of GOA populations is N, the feasible parameter range of VMD includes the penalty factor α. The feasible parameter range [α min ,α max ] and the feasible parameter range of modal component k [k min ,k max ]. Where α min Indicates the minimum value of the penalty factor α; α max represents the maximum value of penalty factor α; k min Indicates the minimum value of modal component k; k maxIndicates the maximum value of modal component k. min ,α max ] Randomly generate N penalty factor values ɑ N , in [k min ,k max ] randomly generates N modal components with values k N . N The N penalty factor values and k N The N modal components in are combined in pairs to obtain N individuals. This set of N individuals is used as the candidate population of GOA.

[0140] S402: Based on the feasible parameter range and the candidate population, the candidate population is updated through the GOA update strategy to obtain the target population.

[0141] Based on the candidate population, the fitness of each individual in the candidate population is determined. Within the feasible parameter range, the candidate population is updated according to the fitness of each individual through the GOA update strategy to obtain an updated candidate population. When the updated candidate population meets the termination condition of GOA, the candidate population obtained after the last update of the candidate population through the GOA update strategy is used as the target population. When the updated candidate population does not meet the termination condition of GOA, the candidate population is continued to be updated according to the feasible parameter range and the candidate population through the GOA update strategy until the updated candidate population meets the termination condition of GOA. Optionally, the termination condition of GOA can be that the number of current iterations is greater than or equal to the maximum number of iterations, and / or that the fitness of the updated candidate population meets the fitness requirement.

[0142] S403: Determine the target parameters of VMD according to the fitness of individuals in the target population.

[0143] The fitness value of each individual in the target population is determined, and based on the fitness values of the individuals in the target population and the meaning of fitness, the VMD parameter combination corresponding to the individual with the best fitness is selected as the target parameter for VMD. Exemplarily, when the fitness is a performance score, the VMD parameter combination corresponding to the individual with the largest fitness value in the target population is selected as the target parameter for VMD. Exemplarily, when the fitness is an error, the VMD parameter combination corresponding to the individual with the smallest fitness value in the target population is selected as the target parameter for VMD.

[0144] Figure 5 Schematic diagram of the process of the distribution network overhead line fault identification method provided in the embodiment of the application Figure 3 .like Figure 5 As shown, this embodiment Figure 4Based on the embodiment, a method for identifying a distribution network overhead line fault is described in detail. In this method, step S402 can be implemented by the following steps:

[0145] S4021. For each individual in the candidate population, perform VMD decomposition on the target signal using the VMD parameters corresponding to each individual to obtain the modal components of each individual.

[0146] For each individual in the candidate population, the target signal is subjected to VMD decomposition using the VMD parameters corresponding to each individual. The determination process of VMD decomposition is consistent with the determination process of VMD iterative calculation.

[0147] Optionally, in step S4021, when performing VMD decomposition on the target signal, only one round of VMD iterative calculation is performed to reduce computational overhead and improve computational efficiency. In practical applications, multiple rounds of VMD iterative calculation may be performed based on actual needs. This embodiment does not specifically limit the number of rounds of iterative calculation.

[0148] S4022. Determine the corresponding fitness through the modal component of each individual.

[0149] First, a fitness function is defined that can characterize the modal components. For example, the fitness function is one of the modal component's envelope entropy, reconstruction error, energy distribution, or frequency. The fitness function is used to determine the fitness of each individual modal component.

[0150] For example, assuming that the defined fitness function is envelope entropy, the envelope entropy of the modal component can be calculated using the envelope entropy determination formula, where the envelope entropy determination formula is:

[0151]

[0152] Where M is the number of sampling points of the target signal; p j is the normalized envelope value of the jth sampling point; the modal component u is obtained by the Hilbert transform method k The envelope e of (t) k (t), for the envelope e k (t) is normalized to obtain p j , p j The formula for determining is:

[0153]

[0154] The modal component u of each individual is obtained by determining the formula using the envelope entropy: k The envelope entropy F(p j ), the envelope entropy F(p j ) is taken as the modal component u of each individualk (t)’s fitness.

[0155] S4023. Determine whether the termination condition is met based on the fitness of each individual in the candidate population.

[0156] The termination condition is the condition used in the GOA algorithm to stop population updates. When the termination condition is that the fitness value reaches the fitness value requirement, the optimal fitness of each individual in the candidate population is determined based on the fitness of each individual in the candidate population, and whether the optimal fitness meets the fitness value requirement is determined. If the optimal fitness meets the fitness value requirement, the termination condition is determined to be met; otherwise, the termination condition is determined to be not met.

[0157] In one possible implementation, the termination condition may be that the number of iterations satisfies a maximum iteration condition. When the termination condition may be that the number of iterations satisfies the maximum iteration condition, whether the termination condition has been met is determined based on the current number of iterations of the candidate population. For example, if the current number of iterations is less than the maximum iteration condition, the termination condition is determined to have been met; otherwise, the termination condition is determined not to have been met.

[0158] S4024. If the termination condition is not met, the candidate population is updated according to the population update strategy of GOA and the fitness of each individual within the feasible parameter range until the termination condition is met.

[0159] If the termination condition is not met, the candidate population is updated according to the population update strategy of GOA and the fitness of each individual within the feasible parameter range, and steps S4021 to S4023 are repeated until the termination condition is met.

[0160] S4025. If the termination condition is reached, the target population is obtained.

[0161] If the termination condition is met, the current candidate population is used as the target population to obtain the target population.

[0162] In a possible implementation, step S4021 includes:

[0163] Step 1: For each individual in the candidate population, initialize the modal component to obtain the initial component.

[0164] For each individual in the candidate population, the modal components and center frequencies of each individual during VMD decomposition are initialized to obtain the initial modal components and initial center frequencies of each individual, thereby obtaining the initial components. Those skilled in the art will appreciate that there are multiple ways to initialize the modal components and center frequencies corresponding to each individual, and any one of these initialization methods may be selected, and this should not be construed as a limitation on the technical solutions of the present application.

[0165] Optionally, the modal component of each individual is set to 0 to obtain the initial modal component of each individual; the center frequency of each individual is set to the frequency mean of the target signal to obtain the initial center frequency, and the initial component is obtained according to the initial modal component and the initial center frequency.

[0166] Optionally, the amplitude and phase of the modal component of each individual in the candidate population are randomly generated to obtain the initial modal component. For example, the amplitude ranges from [0, 1], and a random value is selected within the amplitude range to obtain the amplitude of the modal component; the phase ranges from [0, 2π], and a random value is selected within the phase range to obtain the phase of the modal component. The initial modal component is obtained based on the amplitude and phase of the modal component. Optionally, a target spectrum of the target signal is obtained through Fourier transform. The main frequency component in the target spectrum is selected as the initial center frequency of the modal component.

[0167] Step 2: Perform VMD decomposition on the target signal using the VMD parameters and initial components corresponding to each individual to obtain the modal components of each individual.

[0168] First, perform Hilbert transform on the target signal f(t) to obtain the analytical signal of the target signal f(t) Afterwards, the analytical signal is determined by Fourier transform One-sided spectrum For the modal component u in the initial component k (t), perform Hilbert transform and get the modal component u k Modal components of (t) Afterwards, the modal components u are determined by Fourier transform k One-sided spectrum of (t)

[0169] The initial center frequency ω in the initial component k The variational constraint determination formula of VMD is introduced to construct the variational constraint of VMD. Then, Lagrangian uses the Lagrangian multiplier method to construct the augmented Lagrangian function and obtain the non-constrained variational determination formula of VMD. Then, the alternating direction multiplier method is used to iteratively update the modal component u k (t), center frequency ω k and the Lagrange multiplier λ until the convergence condition is met.

[0170] In one possible implementation, the GOA population update strategy includes at least any one of the following: an update strategy following random perturbations of Brownian motion; an update strategy based on Lévy flight; an update strategy following random perturbations and cumulative effects of Brownian motion.

[0171] Among them, Brownian motion can simulate the irregular movement of tiny particles in a fluid. The random perturbations of Brownian motion can increase the diversity of the candidate population, helping the GOA algorithm to escape the local optimum when updating the candidate population. Lévy flight is a random walk model that follows the Lévy distribution and has a long tail characteristic, that is, large step sizes occasionally appear. When updating through Lévy flight, it helps the GOA algorithm to conduct a global search and increase the possibility of finding the global optimal solution. The cumulative effect refers to the fact that during the iteration process of the GOA algorithm, when updating the individual positions in the candidate population, not only the random perturbations caused by the current Brownian motion are considered, but also the cumulative impact of the random perturbations caused by previous Brownian motion.

[0172] The GOA population update strategy can update the individual positions in the candidate population within the VMD feasible parameter range. The individual positions represent the parameter combination of VMD. The update strategy following the random perturbation of Brownian motion can simulate the free grazing of gazelles in the absence of predators or predator tracking. The update formula of the update strategy following the random perturbation of Brownian motion is:

[0173]

[0174] in, is the position of individual i at the tth iteration; S represents the movement speed of individual gazelle i; R is a random number between [0,1] generated at the tth iteration; Indicates that at the tth iteration, the top gazelle individual with the best fitness in the candidate population, and the rest of the individuals need to move towards the top gazelle individual; R B represents a random vector based on Brownian motion, R B Satisfies the normal probability distribution function:

[0175]

[0176] Where x represents a random variable, μ represents the expected value, and σ represents the standard deviation.

[0177] The update strategy based on Lévy flight can simulate the escape behavior of gazelles using Wright flight in the early stage of discovering predators. The update formula of the update strategy based on Lévy flight is:

[0178]

[0179] Among them, θ is -1 or 1, indicating two movement directions, which can be randomly selected to determine whether to move in the forward or reverse direction; R L represents a random vector based on Lévy flight, R L Determined by the probability distribution function, R L The probability distribution function of is:

[0180]

[0181] Among them, f L (x; ɑ, γ) represents the probability distribution function of the Lévy flight; x represents the input value of the probability distribution function; ɑ represents the stability parameter; γ represents the scale parameter, which is related to the step size of the Lévy flight.

[0182] The update strategy that follows the random perturbation and cumulative effect of Brownian motion can simulate the escape behavior of gazelles after discovering the predator. The update formula of the update strategy that follows the random perturbation and cumulative effect of Brownian motion is:

[0183]

[0184]

[0185] Where CF represents the cumulative effect of the predator. As t increases, the value of CF will change, reflecting the dynamic changes of the predator's influence. t represents the tth iteration, and T represents the maximum number of iterations.

[0186] For ease of understanding, the GOA algorithm is further explained in detail. The GOA algorithm includes:

[0187] Step 1: Initialization.

[0188] Randomly generate gazelle individuals within the solution space of the GOA algorithm, each representing a potential solution. Optionally, the population size can be any positive integer, such as 5 gazelles, 10 gazelles, 18 gazelles, or 40 gazelles. For example, if the population size is 10, 10 gazelle individuals are randomly generated within the solution space of the GOA algorithm.

[0189] Step 2: Population iteration.

[0190] Generate a random number r, and use the random number r to determine the current GOA population update strategy. When r is less than 0.5, the gazelles graze freely without the threat of predators, and the GOA population update strategy is to adopt an update strategy that follows the random perturbation of Brownian motion. Further, the iterative process is divided into an early stage and a late stage. In the early stage, when r is greater than or equal to 0.5, the GOA population update strategy is an update strategy based on Lévy flight; in the late stage, when r is greater than or equal to 0.5, the GOA population update strategy is an update strategy that follows the random perturbation and cumulative effect of Brownian motion. Optionally, the iterative process is divided into an early stage and a late stage based on the maximum number of iterations. The iterative process in which the current number of iterations is less than or equal to half of the maximum number of iterations is divided into the early stage; the iterative process in which the current number of iterations is greater than half of the maximum number of iterations is divided into the late stage. For example, if the maximum number of iterations is 50 and the current number of iterations is 20, since 20 is less than half of 50, it is determined to be the early stage, and the GOA population update strategy is the update strategy based on Lévy flight.

[0191] Step 3: Escape phase.

[0192] Whether the gazelle successfully escapes is determined by the predator's hunting success rate and a random number.

[0193] Step 4: The iteration ends.

[0194] If the current number of iterations is greater than or equal to the maximum number of iterations, the iteration is terminated and the target population is obtained.

[0195] If the current number of iterations is less than the maximum number of iterations, continue to perform step 2 until the number of iterations is greater than or equal to the maximum number of iterations.

[0196] In one possible implementation, the fault identification model is trained in the following manner:

[0197] Step 1: Obtain the historical signal data of the distribution network overhead lines and perform VMD decomposition on the historical signal data to obtain the target features.

[0198] Collect historical signal data from distribution network overhead lines. This data contains rich characterization information about the distribution network overhead lines. Optionally, preprocess the collected historical signal data to obtain processed historical signal data. This data is then subjected to VMD decomposition to obtain target features. Data preprocessing includes denoising and normalization, which can improve data quality. The VMD algorithm is used to decompose the historical signal data into multiple modal components. Features corresponding to each modal component are extracted from each modal component to obtain the target features.

[0199] Optionally, the historical signal data of the distribution network overhead lines can be obtained by manual input, collection by a collection device, system interface call and / or file upload. Exemplarily, when the historical signal data collection device of the distribution network overhead lines is collected, the collection device should be arranged in the distribution network overhead lines, or, arranged at a position where the corresponding historical signal data can be collected. During the collection process, the collection device collects different operating states of the lines, and the operating states include normal operating states, short-circuit fault states, and / or high-resistance fault states. At the same time, the collection device also collects data under environmental conditions such as different weather and / or different loads to ensure that the historical signal data collected by the collection device is comprehensive and representative, and to provide sufficient samples for fault identification model training.

[0200] Step 2: Encode the target features through the Transformer model to obtain the encoded feature space representation.

[0201] The target features are input into the Transformer model for deep feature encoding. The Transformer model's multi-head attention mechanism encodes the target features according to the set number of heads and dimensions, generating an encoded feature space representation. The multi-head attention mechanism dynamically focuses on the associations between different target features, capturing potential dependencies between them.

[0202] Optionally, the injected reflection signal f(t) is divided into S function segments using a sliding window. Each segment is subjected to VMD decomposition to extract the modal components of the S segments. Next, the nine features corresponding to the modal components of each S segment are extracted to obtain the target feature for each segment. In this case, the dimension of the target feature is S × the total number of modal components × 9. The target feature is then input into the Transformer model, where its multi-head attention mechanism encodes the target feature according to the set number of heads and dimensions, resulting in the encoded feature space representation Z.

[0203] Step 3: Train the Xgboost model based on the feature space representation to obtain a fault recognition model.

[0204] First, the feature space representation is divided into training and test sets in a 7:3 ratio. The activation function layer of the Transformer model is replaced with an XgBoost classifier to obtain the XgBoost model. The XgBoost model is then trained using the training set data. During training, an appropriate kernel function must be selected based on the data characteristics.

[0205] During the training process, cross-validation was used to carefully tune key Xgboost parameters, such as the number of decision trees and maximum tree depth. After model training, a fault recognition model was obtained. The trained model was then used to classify and identify faults in the test set data, yielding the fault recognition results for the test set. The performance of the model was then evaluated based on the fault recognition results from the test set.

[0206] Optionally, based on the fault identification results of the test set, the accuracy is determined using an accuracy determination formula; the recall is determined using a recall determination formula; the precision is determined using a precision determination formula; and the F1 value is determined using an F1 determination formula. The performance of the fault identification model is evaluated based on the accuracy, recall, precision, and F1 value. The fault identification model parameters are further optimized based on the evaluation results to improve the performance of the fault identification model in identifying distribution network overhead line faults.

[0207] For example, the accuracy determination formula is:

[0208]

[0209] Among them, TP is the number of true positives, TN is the number of true negatives, FP is the number of false positives, and FN is the number of false negatives.

[0210] The recall rate is determined by:

[0211]

[0212] The formula for determining the accuracy is:

[0213]

[0214] The formula for determining F1 is:

[0215]

[0216] In one possible implementation, an Xgboost model is trained based on feature space representation to obtain a fault recognition model. The Xgboost model can be replaced with other applicable models or algorithms.

[0217] The method for identifying faults in overhead distribution lines provided in the embodiments of the present application has at least the following advantages:

[0218] 1. Accurate VMD decomposition of the target signal: By optimizing the VMD target parameters through GOA, the VMD decomposition of the target signal can be adaptively adjusted to suit the signal characteristics of different distribution network overhead lines. This also avoids the VMD decomposition limitations caused by poor VMD parameter selection. The modal components obtained during VMD decomposition using GOA-optimized VMD target parameters have strong characterization capabilities and can better represent the signal characteristics of distribution network overhead lines. Compared with traditional methods with fixed VMD parameters, the reconstruction error with the original signal is reduced by approximately 30%-50%.

[0219] 2. Through the deep encoding of the Transformer, the characterization capability of signal features is further improved. By combining the feature extraction methods of VMD and Transformer, it is possible to distinguish highly differentiated features, further improving the accuracy of fault classification between different fault types. Compared with the fault classification methods in the related art, the fault classification accuracy of the distribution network overhead line fault identification method proposed in the embodiment of this application is improved by approximately 20%-30%.

[0220] 3. By combining the Transformer model and the Xgboost model, the advantages of the Transformer model and the Xgboost model can be fully utilized. Transformer's deep encoding of target features provides Xgboost with a more discriminative feature space representation, enabling Xgboost to more effectively classify and locate faults. In actual distribution network overhead line fault detection experiments, the fault location accuracy of the distribution network overhead line fault identification method provided by the embodiment of the present application is improved by about 15%-25% compared with traditional methods, and the fault classification accuracy rate reaches 94%. It can detect the fault type and location more quickly and accurately, effectively reducing fault troubleshooting time and power outage losses.

[0221] Figure 6 This is a schematic diagram of the structure of the distribution network overhead line fault identification device provided in the embodiment of the present application. Figure 6 As shown, the distribution network overhead line fault identification device 60 provided in this embodiment includes:

[0222] An acquisition module 601 is used to acquire a target signal on an overhead line of a distribution network;

[0223] Processing module 602 is used to obtain the target signal on the distribution network overhead line; based on the GOA, determine the target parameters of VMD within the feasible parameter range of VMD; perform VMD decomposition on the target signal based on the target parameters to obtain the signal characteristics of the target signal; input the signal characteristics into the fault identification model to obtain the fault identification result of the distribution network overhead line; wherein the fault identification model is trained based on the Xgboost algorithm.

[0224] In one possible implementation, processing module 602 is specifically configured to: randomly generate multiple individuals that meet the GOA population size requirement within a feasible parameter range of VMD to obtain a candidate population of GOA; each individual includes a set of VMD parameter combinations; based on the feasible parameter range and the candidate population, update the candidate population using the GOA update strategy to obtain a target population; and determine the target parameters of VMD based on the fitness of the individuals in the target population.

[0225] In one possible implementation, the processing module 602 is further configured to: perform VMD decomposition on the target signal for each individual in the candidate population using the VMD parameters corresponding to each individual to obtain the modal components of each individual; determine the corresponding fitness of each individual through the modal components of each individual; determine whether a termination condition is met based on the fitness of each individual in the candidate population; if the termination condition is not met, update the candidate population according to the population update strategy of GOA and the fitness of each individual within a feasible parameter range until the termination condition is met; and if the termination condition is met, obtain the target population.

[0226] In one possible implementation, the processing module 602 is further used to: initialize the modal components of each individual in the candidate population to obtain initial components; perform VMD decomposition on the target signal using the VMD parameters and initial components corresponding to each individual to obtain the modal components of each individual.

[0227] In one possible implementation, the GOA population update strategy includes at least any one of the following: an update strategy following random perturbations of Brownian motion; an update strategy based on Lévy flight; an update strategy following random perturbations and cumulative effects of Brownian motion.

[0228] In one possible implementation, the fault identification model is trained in the following manner: historical signal data of the distribution network overhead lines is obtained, and the historical signal data is decomposed by VMD to obtain target features; the target features are encoded through a Transformer model to obtain an encoded feature space representation; and an Xgboost model is trained based on the feature space representation to obtain a fault identification model.

[0229] The distribution network overhead line fault identification device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effects are similar, and are not described in detail in this embodiment.

[0230] Figure 7 This is a schematic diagram of the structure of the electronic device provided in the embodiment of the present application. Figure 7As shown, the electronic device 70 provided in this embodiment includes: at least one processor 701 and a memory 702. Optionally, the device 70 further includes a communication component 703. The processor 701, the memory 702 and the communication component 703 are connected via a bus 704.

[0231] During the specific implementation process, at least one processor 701 executes the computer-executable instructions stored in the memory 702, so that the at least one processor 701 performs the above method.

[0232] The specific implementation process of the processor 701 can be found in the above method embodiment. Its implementation principle and technical effects are similar and will not be repeated here in this embodiment.

[0233] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), etc. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the present invention may be directly implemented by a hardware processor or implemented by a combination of hardware and software modules in the processor.

[0234] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (NVM), such as at least one disk memory.

[0235] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be divided into address buses, data buses, control buses, etc. For ease of presentation, the buses in the drawings of the embodiments of the present application are not limited to just one bus or just one type of bus.

[0236] An embodiment of the present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0237] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed, any of the above methods is implemented.

[0238] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile memory device or a combination thereof, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0239] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be an integral part of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist in a device as discrete components.

[0240] The division of units is merely a logical functional division; actual implementations may employ alternative divisions, such as combining or integrating multiple units or components into another system, or omitting or disabling certain features. Furthermore, any direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between devices or units, either through an interface, electrical, mechanical, or other means.

[0241] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0242] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0243] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and other media that can store program code.

[0244] Those skilled in the art will appreciate that all or part of the steps in the above-described method embodiments can be implemented using hardware associated with program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0245] Finally, it should be noted that those skilled in the art will readily identify other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art not disclosed herein. The present invention is not limited to the precise structure described above and illustrated in the accompanying drawings, and various modifications and variations may be made without departing from the scope thereof. The scope of the present invention is limited solely by the appended claims.

Claims

1. A method for identifying faults in overhead lines of a distribution network, characterized in that: include: Acquiring a target signal on the distribution network overhead line; Determining target parameters of the variational mode decomposition (VMD) within a feasible parameter range of the VMD based on the Gazelle Optimization Algorithm (GOA); Performing VMD decomposition on the target signal based on the target parameter to obtain a signal feature of the target signal; The signal features are input into a fault identification model to obtain a fault identification result of the distribution network overhead line; wherein the fault identification model is trained by the extreme gradient boosting Xgboost algorithm.

2. The method according to claim 1, characterized in that The method of determining target parameters of the variational mode decomposition (VMD) based on the Gazelle Optimization Algorithm (GOA) within a feasible parameter range of the VMD includes: Within the feasible parameter range of the VMD, a plurality of individuals that meet the GOA population quantity requirement are randomly generated to obtain a candidate population of the GOA; each of the individuals includes a set of VMD parameter combinations; According to the feasible parameter range and the candidate population, the candidate population is updated by the update strategy of the GOA to obtain a target population; The target parameters of the VMD are determined according to the fitness of individuals in the target population.

3. The method according to claim 2, characterized in that The updating of the candidate population by the GOA update strategy according to the feasible parameter range and the candidate population to obtain the target population further includes: For each individual in the candidate population, performing VMD decomposition on the target signal using the VMD parameters corresponding to each individual to obtain the modal component of each individual; Determining the corresponding fitness through the modal component of each individual; Determining whether a termination condition is met based on the fitness of each individual in the candidate population; If the termination condition is not met, the candidate population is updated according to the population update strategy of the GOA and the fitness of each individual within the feasible parameter range until the termination condition is met; If the termination condition is met, the target population is obtained.

4. The method according to claim 3, characterized in that For each individual in the candidate population, performing VMD decomposition on the target signal using the VMD parameters corresponding to each individual to obtain the modal component of each individual, including: For each individual in the candidate population, initializing the modal component to obtain an initial component; The target signal is subjected to VMD decomposition using the VMD parameters corresponding to each individual and the initial components to obtain the modal components of each individual.

5. The method according to claim 2, characterized in that The population update strategy of the GOA includes at least one of the following: Update strategies for random perturbations following Brownian motion; Update strategy based on Levy flight; Update strategies that follow random perturbations and cumulative effects of Brownian motion.

6. The method according to claim 1, characterized in that The fault identification model is trained in the following way: Obtain historical signal data of the distribution network overhead line, and perform VMD decomposition on the historical signal data to obtain target features; Encoding the target features through a Transformer model to obtain an encoded feature space representation; The Xgboost model is trained based on the feature space representation to obtain the fault recognition model.

7. A distribution network overhead line fault identification device, characterized in that: include: An acquisition module, configured to acquire a target signal on the overhead line of the distribution network; A processing module is configured to determine target parameters of a variational mode decomposition (VMD) within a feasible parameter range of the VMD based on a gazelle optimization algorithm (GOA); perform VMD decomposition on the target signal based on the target parameters to obtain signal characteristics of the target signal; and input the signal characteristics into a fault identification model to obtain a fault identification result for the distribution network overhead line; wherein the fault identification model is trained based on the Xgboost algorithm.

8. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The method comprises a computer program, which implements the method according to any one of claims 1 to 6 when the computer program is executed.

Images