Distribution network overhead line injection type fault positioning method, device, equipment and medium

Abstract

The invention provides a distribution network overhead line injection type fault positioning method and device, equipment and a medium, and relates to the technical field of distribution network fault positioning. The method comprises the following steps: acquiring a reflection signal of a target signal under the condition that the target signal is injected into a distribution network overhead line; and inputting the reflected signal to a pre-trained target fault positioning model to obtain position information of the distribution network overhead line fault point, the target fault positioning model being constructed after optimizing hyper-parameters of a target deep learning model based on LEA, and the target deep learning model including an LSTM model, a GRU or a BiLSTM model. According to the method, the hyper-parameters of the target deep learning model are optimized through LEA, the learning ability of the model for fault features in the reflected signals is remarkably improved, the optimized model can more accurately capture subtle changes of the signals in the time domain and the frequency domain, and therefore the accuracy of fault positioning is remarkably improved.

Description

Technical Field

[0001] The present application relates to the technical field of distribution network fault location, and particularly to a method, device, equipment and medium for locating injection-type faults in overhead distribution lines. Background Technique

[0002] Overhead distribution lines are an important part of the distribution network and play a key role in delivering electrical energy from the substation to the user end. Compared with cable lines, overhead lines have the advantages of low cost, convenient construction and easy maintenance. However, due to their exposed nature, overhead lines are vulnerable to the influence of the external environment (such as lightning strikes, wind and rain, or trees) and equipment aging, resulting in frequent faults, which in turn affect the safe and stable operation of the distribution network. Therefore, quickly and accurately locating the fault point is of great practical significance for improving the reliability and operation and maintenance efficiency of the power system.

[0003] In the related art, signal processing means such as Fast Fourier Transform (FFT) and Wavelet Transform (WT) are usually used to extract features such as amplitude, frequency and phase of the signal actively injected into the line and its reflected signal, and these features are input into machine learning models such as Support Vector Machine (SVM) or decision tree for fault location, but there is a problem of low fault location accuracy. Summary of the Invention

[0004] The present application provides a method, device, equipment and medium for locating injection-type faults in overhead distribution lines to improve the problem of low fault location accuracy of injection-type faults in overhead distribution lines.

[0005] In a first aspect, the present application provides a method for locating injection-type faults in overhead distribution lines, including:

[0006] When the target signal is injected into the overhead distribution line, obtain the reflected signal of the target signal;

[0007] The reflected signal is input into a pre-trained target fault location model to obtain the location information of the fault point on the overhead distribution line. The target fault location model is constructed by optimizing the hyperparameters of the target deep learning model based on the Lotus Effect Algorithm (LEA). The target deep learning model includes a Long Short-Term Memory Network (LSTM) model, a Gated Recurrent Unit (GRU), or a Bidirectional Long Short-Term Memory Network (BiLSTM) model.

[0008] In a possible implementation, the target signal includes at least one of a high-frequency pulse signal, a sine wave signal with a specific frequency, a multi-frequency modulation signal, and a chirp signal.

[0009] In a possible implementation, the hyperparameters include the number of hidden layers, the number of neurons, and the learning rate. The hyperparameters are optimized in the following way: the initial population of LEA is randomly generated based on the search range corresponding to the hyperparameters, and each individual in the initial population represents a set of hyperparameter combinations; a fitness function is constructed based on the mean square error of fault location of the target deep learning model on the training set, and the fitness function is used to evaluate the performance of each set of hyperparameter combinations; for each individual in the initial population, based on the fitness function and according to the exploration phase principle of LEA, the hyperparameter combination of the individual is updated according to the target factors; according to the exploitation phase principle of LEA, local search is performed with the best solution obtained in the exploration phase as the core to update the hyperparameter combination of the individual; according to the exploitation phase reinforcement principle of LEA, local search is performed by simulating the movement of water droplets in the pits of lotus leaves with the best solution obtained in the exploitation phase as the core to update the hyperparameter combination of the individual; the current iteration number is determined, and if the current iteration number reaches the target iteration number, the hyperparameters corresponding to the best solution obtained in the exploitation phase reinforcement are determined as the optimized hyperparameters.

[0010] In a possible implementation, the method for locating an injection-type fault on an overhead distribution line further includes: if the current iteration number does not reach the target iteration number, return to execute the step of updating the hyperparameter combination of each individual in the initial population based on the fitness function and according to the exploration phase principle of LEA according to the target factors until the target iteration number is reached.

[0011] In a possible implementation, the target factors include separation behavior, alignment behavior, cohesion behavior, food attraction, and enemy distraction. According to the exploration phase principle of LEA, the hyperparameter combination of an individual is updated according to the target factors, including: for each factor in the target factors, determining the product of the factor value corresponding to the factor and the weight coefficient corresponding to the factor to obtain the weighted value corresponding to the factor; based on the sum of the weighted values corresponding to each factor, determining the velocity vector of the individual; and adjusting the position of the individual according to the velocity vector and the current hyperparameter combination corresponding to the individual to obtain the updated hyperparameter combination.

[0012] In a possible implementation, inputting the reflection signal into a pre-trained target fault location model includes: inputting the reflection signal into a target filter for filtering to obtain filtered data, where the target filter includes a Butterworth filter or a Chebyshev filter; performing normalization processing on the filtered data to obtain normalized data; formatting the normalized data according to the input format requirements of the target fault location model to obtain formatted data; and inputting the formatted data into the target fault location model.

[0013] In a possible implementation, the target fault location model is trained in the following manner: obtaining a training set, where the training set includes reflection signal samples and the actual fault locations corresponding to the reflection signal samples; inputting the reflection signal samples into the target fault location model to obtain the fault prediction locations corresponding to the reflection signal samples; calculating the mean square error as the loss function based on the fault prediction locations and the actual fault locations; and using an optimization algorithm based on stochastic gradient descent to adjust the model parameters through backpropagation algorithm so that the loss function value of the target fault location model on the training set gradually decreases.

[0014] In a second aspect, the present application provides a positioning device for an injection-type fault in a distribution network overhead line, including:

[0015] An acquisition module, configured to acquire the reflection signal of the target signal when the target signal is injected into the distribution network overhead line;

[0016] A fault location module, configured to input the reflection signal into a pre-trained target fault location model to obtain the position information of the fault point of the distribution network overhead line, where the target fault location model is constructed based on optimizing the hyperparameters of the target deep learning model by LEA, and the target deep learning model includes an LSTM model, a GRU, or a BiLSTM model.

[0017] In a third aspect, the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0018] The memory is used to store computer execution instructions;

[0019] A processor for executing computer-executable instructions stored in a memory to implement the method according to any one of the first aspects.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium storing computer-executable instructions that, when executed, are used to implement the method according to any one of the first aspects.

[0021] In a fifth aspect, the present application provides a computer program product including a computer program that, when executed, implements the method according to any one of the first aspects.

[0022] The method, device, equipment, and medium for locating an injection-type fault in a distribution network overhead line provided by the present application, in the case where a target signal is injected into the distribution network overhead line, obtain a reflection signal of the target signal; input the reflection signal into a pre-trained target fault location model to obtain location information of a fault point on the distribution network overhead line, where the target fault location model is constructed by optimizing hyperparameters of a target deep learning model based on LEA, and the target deep learning model includes an LSTM model, a GRU, or a BiLSTM model. In this process, the hyperparameters of the target deep learning model are optimized by LEA, significantly improving the model's learning ability for fault features in the reflection signal, enabling the optimized model to more accurately capture subtle changes in the signal in the time domain and frequency domain, and thus significantly improving the accuracy of fault location. Description of the Drawings

[0023] The drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0024] Figure 1 A schematic diagram of an application scenario of the method for locating an injection-type fault in a distribution network overhead line provided by an exemplary embodiment of the present application;

[0025] Figure 2 A flowchart of the method for locating an injection-type fault in a distribution network overhead line provided by an exemplary embodiment of the present application;

[0026] Figure 3 A schematic diagram of the structure of an LSTM unit in an LEA-LSTM model provided by an exemplary embodiment of the present application;

[0027] Figure 4 Another flowchart of the method for locating an injection-type fault in a distribution network overhead line provided by an exemplary embodiment of the present application;

[0028] Figure 5A structural schematic diagram of a positioning device for injection-type faults in a distribution network overhead line provided by an exemplary embodiment of the present application;

[0029] Figure 6 A structural schematic diagram of an electronic device provided by an exemplary embodiment of the present application.

[0030] Through the above-mentioned drawings, specific embodiments of the present application have been shown, and there will be a more detailed description hereinafter. These drawings and written descriptions are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Description of the Embodiment

[0031] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0032] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, products, or devices.

[0033] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties. The collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards, and corresponding operation entrances are provided for users to choose to authorize or refuse.

[0034] During the research process, the inventors found that in order to trigger the reflected signal of the line, it is usually necessary to actively inject high-frequency signals into the overhead distribution line. These signals are usually electrical signals with specific frequencies, amplitudes, and waveforms, which can propagate in the line and be reflected at the fault point (such as line break or short circuit, etc.). The fault point will change the signal propagation characteristics, resulting in a significant difference between the reflected signal and the normal signal. By installing sensors (such as current, voltage, or temperature sensors, etc.) on the distribution line, the propagation and reflection process of the signal on the line are collected. Among them, the collected signal data contains information such as the amplitude, time delay, and frequency change of the reflected signal. In the signal analysis stage, mathematical tools such as FFT and WT are usually used. FFT is used to convert the signal from the time domain to the frequency domain, extract the spectrum information of the signal, and identify the frequency components and their amplitude distributions. When a fault occurs in the line, the spectrum characteristics of the signal will change, and the amplitudes of some frequency components will fluctuate significantly. Through FFT analysis, these spectrum changes can be captured, providing a basis for fault judgment. WT, as a time-frequency analysis method, can provide both the time information and frequency information of the signal. Compared with FFT, WT is more sensitive in processing transient signals and can decompose the signal at different time and frequency resolutions. Therefore, it is suitable for analyzing the characteristics of the frequency changing with time in the signal. Through WT, the multi-scale information of the signal can be extracted, which helps to capture the instantaneous change characteristics of the signal. In fault location, due to the different waveforms and spectrum characteristics of the fault signal and the normal signal, WT can effectively distinguish the normal signal from the fault signal, especially when faults such as short circuit and line break occur. In the signal preprocessing stage, through FFT and WT, etc., information such as the frequency, amplitude, and time delay of the signal is obtained, and further features with discriminative power for fault location are extracted from them. Among them, typical features include amplitude change, frequency offset, phase change, and time delay difference of the signal, which can reflect the state change of the line. Especially when a fault occurs, the amplitude and frequency characteristics of the signal will change significantly. After the extracted features are processed, a feature vector containing fault information is formed, providing input data for the training and prediction of the machine learning model. Then, the extracted feature vector is input into the machine learning model for analysis and classification. Commonly used machine learning models include SVM and decision tree. These models can find the boundary between the fault signal and the normal signal in the multi-dimensional feature space. In the model training stage, a large number of historical data sets are required, which contain signal characteristics under different types of faults (such as line break, short circuit, poor contact, etc.) and normal states. Through training, the machine learning model can learn how to accurately judge the fault location according to the signal characteristics.

[0035] Considering that the signal is affected by various factors such as changes in line parameters, noise interference, and multipath effects during propagation, its characteristics are extremely complex. In the related art, FFT can only analyze the frequency-domain characteristics of the signal and cannot effectively capture the time-varying characteristics of the signal; although WT can take into account the time-frequency characteristics to a certain extent, it is still not accurate and comprehensive enough for signal feature extraction in complex fault situations. Specifically, there are deficiencies in aspects such as weak signal changes caused by faults, the precise position and feature description of signal mutation points, resulting in the extracted features being difficult to accurately reflect the essential information of the fault and affecting the accuracy of fault location.

[0036] To solve the above problems, the embodiments of the present application provide a positioning scheme for injection-type faults in distribution network overhead lines. By using LEA to optimize the hyperparameters of the target deep learning model, the learning ability of the model for fault features in the reflected signal is improved, so as to obtain a target fault positioning model with better performance; an offline active injection signal device is used to inject a target signal into the distribution network overhead line, and the reflected signal of the line to the injected signal is collected at a suitable position, and the collected reflected signal is input into the optimized target fault positioning model. Based on in-depth analysis and learning of the signal data, the model outputs the position information of the fault point of the distribution network overhead line to achieve high-precision fault positioning; and through high-accuracy fault positioning, the workload of manual inspection and fault troubleshooting is greatly reduced, the operation and maintenance cost is reduced, and the power outage time caused by faults is effectively reduced, significantly improving the power supply reliability and user satisfaction.

[0037] Figure 1 It is a schematic diagram of the application scenario of the positioning method for injection-type faults in distribution network overhead lines provided by an exemplary embodiment of the present application. As Figure 1 shown, the application scenario includes a client 11 and a server 12, where the number of clients 11 can be at least one. In practical applications, when the server 12 detects a positioning instruction for injection-type faults in the distribution network overhead line submitted by the client 11, it executes the positioning method for injection-type faults in the distribution network overhead line provided by the present application to obtain the position information of the fault point of the distribution network overhead line.

[0038] It should be noted that the server 12 can also be replaced by a server cluster or other computing devices with a certain computing power. The first client and the second client can both be a mobile phone, a computer, a laptop, or a personal digital assistant (Personal Digital Assistant, abbreviated as PDA), etc.

[0039] Next, in combination with Figure 1 the application scenario, refer to Figure 2A method for locating injection-type faults in distribution network overhead lines according to an exemplary embodiment of the present application is described. It should be noted that the above application scenarios are only shown for the convenience of understanding the spirit and principle of the present application, and the embodiments of the present application are not limited by Figure 1 the application scenarios shown. It should be noted that 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 with reference to the accompanying drawings.

[0040] Figure 2 FIG. is a schematic flow chart of a method for locating injection-type faults in distribution network overhead lines provided by an exemplary embodiment of the present application. As Figure 2 shown, the method for locating injection-type faults in distribution network overhead lines includes the following steps:

[0041] S201. When a target signal is injected into a distribution network overhead line, obtain the reflected signal of the target signal.

[0042] Exemplarily, a target signal is injected into the distribution network overhead line through a pre-built offline active injection signal system. Among them, the offline active injection signal system includes, but is not limited to, devices such as a signal generator, a power amplifier, and a coupler; determine parameters such as the amplitude and frequency of the target signal to be injected, and inject the signal into the distribution network overhead line through a coupler; through a high-precision data acquisition card or sensor pre-deployed at the monitoring point of the distribution network overhead line, etc., collect the reflected signal of the line in real time. Among them, the selection of the monitoring point needs to ensure that the collected reflected signal can accurately reflect the response of the line to the injected signal. Correspondingly, obtain the reflected signal collected by the high-precision data acquisition card or sensor, etc.

[0043] S202. Input the reflected signal into a pre-trained target fault location model to obtain the location information of the fault point of the distribution network overhead line. Among them, the target fault location model is constructed by optimizing the hyperparameters of the target deep learning model based on LEA, and the target deep learning model includes an LSTM model, a GRU or a BiLSTM model.

[0044] Among them, LEA is a new evolutionary algorithm that combines the efficient operators in the dragonfly algorithm with the superhydrophobic and self-cleaning characteristics (lotus effect) of lotus leaves for extraction and local search operations; the superhydrophobic property of lotus leaves enables water droplets to quickly gather and slide on its surface, while taking away dust and dirt at the same time. This phenomenon is called the lotus effect; LEA simulates this property and performs local search through the movement behavior of water droplets on the lotus leaf, so as to achieve efficient exploration and optimization of the solution space.

[0045] Exemplarily, the target deep learning model is an LSTM model. Based on LEA, the hyperparameters of the LSTM model are optimized to obtain the optimized hyperparameters. Based on the optimized hyperparameters, an LSTM model is constructed to obtain the target fault location model, that is, the LEA-LSTM model. Among them, the LEA-LSTM model includes an input layer, a hidden layer, a fully connected layer, and an output layer. Correspondingly, if the input reflected signal is time series data, features such as the amplitude and phase of the reflected signal can be used as inputs; the hidden layer processes the time series data through LSTM units, which can effectively handle the long-term dependence relationships in the time series data and can be used to extract the fault feature information in the reflected signal; the fully connected layer is used to integrate the feature information output by the hidden layer and further extract high-dimensional features; the output layer is used to output the location information of the fault point on the overhead distribution line. Among them, the location information includes the distance of the fault point from the injection end or the number of the line section where the fault point is located, etc. Exemplarily, Figure 3 FIG. is a schematic structural diagram of an LSTM unit in the LEA-LSTM model provided by an exemplary embodiment of the present application. As Figure 3 shown, h t-1 represents the output of the LSTM unit at time t-1 and serves as the long-term memory input of the LSTM unit at time t; X t represents another input of the LSTM unit at time t; h t represents the output of the LSTM unit at time t; C t-1 represents another output of the LSTM unit at time t-1 and also serves as an input of the LSTM unit at time t. Correspondingly, C t corresponds to its output in the LSTM unit at time t.

[0046] Correspondingly, if the target deep learning model is a GRU, based on LEA, the hyperparameters of the GRU are optimized to obtain the optimized hyperparameters. Based on the optimized hyperparameters, a GRU is constructed to obtain the target fault location model, that is, the LEA-GRU. Further, the reflected signal is input into the pre-trained LEA-GRU to obtain the location information of the fault point on the overhead distribution line. Among them, the location information includes the distance of the fault point from the injection end or the number of the line section where the fault point is located, etc.

[0047] Correspondingly, if the target deep learning model is a BiLSTM model, based on LEA, the hyperparameters of the BiLSTM model are optimized to obtain the optimized hyperparameters. Based on the optimized hyperparameters, a BiLSTM model is constructed to obtain the target fault location model, that is, the LEA-BiLSTM model. Further, the reflected signal is input into the pre-trained LEA-BiLSTM model to obtain the location information of the fault point on the overhead distribution line. Among them, the location information includes the distance of the fault point from the injection end or the number of the line section where the fault point is located, etc.

[0048] The fault location method for distribution network overhead lines provided by the embodiments of this application optimizes the hyperparameters of the target deep learning model through LEA, significantly improving the model's learning ability for fault features in the reflected signal, enabling the optimized model to more accurately capture the subtle changes of the signal in the time domain and frequency domain, thus significantly improving the accuracy of fault location and fully meeting the high-precision location requirements in complex distribution network environments; in addition, through highly accurate fault location, the workload of manual inspection and fault troubleshooting is greatly reduced, the operation and maintenance costs are reduced, and the power outage time caused by faults is effectively reduced, significantly improving the power supply reliability and user satisfaction.

[0049] In some embodiments, the target signal includes at least one of a high-frequency pulse signal, a sine wave signal with a specific frequency, a multi-frequency modulation signal, and a chirp signal.

[0050] Among them, the high-frequency pulse signal is a short-time signal with fast rise and fall times, and its frequency range is usually higher than 1 MHz. This signal can propagate quickly in the line and produce obvious reflections at the fault point. Because it can quickly respond and capture the reflection characteristics of the fault point, it is suitable for detecting faults such as line breaks and short circuits in the line; the sine wave signal is a periodic signal with a fixed frequency, and its waveform is a sine curve. The sine wave signal with a specific frequency refers to a single-frequency signal with a precisely selected frequency, which is suitable for analyzing the impedance characteristics of the line and can effectively detect high-resistance faults or line parameter changes; the multi-frequency modulation signal is a composite signal composed of multiple frequency components, usually generated by modulation techniques (such as frequency division multiplexing), and its spectrum contains multiple discrete frequency points, which is suitable for fault detection in complex line environments, can analyze the reflection characteristics of multiple frequency points simultaneously, and improve the accuracy of fault location; the chirp signal is a signal whose frequency changes with time, usually showing a linear increase or decrease in frequency (i.e., a frequency-swept signal). Its characteristic is a wide frequency coverage range, which can provide rich frequency-domain information, is suitable for detecting frequency-sensitive faults in the line, can capture the reflection characteristics of the signal at different frequencies, and is suitable for analyzing the frequency response characteristics of the line.

[0051] Correspondingly, according to the electrical characteristics of the distribution network overhead line and the fault location requirements, select the appropriate injection signal type. For example, in one implementation, the selected target signal is at least one of a high-frequency pulse signal, a sine wave signal with a specific frequency, a multi-frequency modulation signal, and a chirp signal. For example, for rapid fault detection, a high-frequency pulse signal can be selected; for high-resistance fault detection, a sine wave signal with a specific frequency can be selected; for complex line environments, a multi-frequency modulation signal or a chirp signal can be selected, etc.

[0052] In another implementation manner, the selected target signal is any combination of high-frequency pulse signals, sine wave signals with specific frequencies, multi-frequency modulation signals, chirp signals, etc. For example, the target signal is a combination of a high-frequency pulse signal and a sine wave signal with a specific frequency.

[0053] Exemplarily, in actual distribution network line testing, from signal injection to the output of the fault location result, the entire process takes about 2-3 seconds. Compared with traditional manual inspection and methods based on simple electrical quantity analysis (with an average time consumption of 15-30 minutes), the efficiency is increased by dozens of times, enabling power operation and maintenance personnel to quickly respond to faults and repair them in a timely manner, significantly reducing the power outage time. Specifically, taking the distribution network of a medium-sized city as an example, the annual power outage time due to faults can be reduced by about 30%, which is of positive significance for enhancing the social benefits of power enterprises and user satisfaction.

[0054] In the embodiments of the present application, by flexibly selecting a suitable injection signal type according to the electrical characteristics and fault location requirements of the distribution network overhead line, it helps to further improve the accuracy of fault location and effectively improve the speed of fault location, which is of positive significance for reducing the power outage time and enhancing user satisfaction, etc.

[0055] Considering that in the related art, when determining the parameters of the fault location model, it mostly relies on empirical judgment or simple trial-and-error methods, which are not only inefficient but also difficult to find the optimal parameter combination of the model. For example, for the kernel function parameters or penalty factors in the support vector machine (SVM), or the tree depth or branch node threshold and other parameters in the decision tree model, there is a lack of scientific and effective optimization strategies. Due to improper parameter selection, the performance of the model cannot be fully exerted, further reducing the accuracy and reliability of fault location. Therefore, in some embodiments, the hyperparameters include the number of hidden layers, the number of neurons, and the learning rate, and the hyperparameters are optimized in the following manner: randomly generate the initial population of the Lévy flight for exploration (LEA) based on the search range corresponding to the hyperparameters, and each individual in the initial population represents a set of hyperparameter combinations; construct a fitness function based on the mean squared error of fault location of the target deep learning model on the training set, and the fitness function is used to evaluate the performance of each set of hyperparameter combinations; for each individual in the initial population, based on the fitness function, according to the exploration phase principle of LEA, update the hyperparameter combination of the individual according to the target factor; according to the exploitation phase principle of LEA, perform local search with the best solution obtained in the exploration phase as the core to update the hyperparameter combination of the individual; according to the enhanced exploitation phase principle of LEA, perform local search with the best solution obtained in the exploitation phase as the core, simulating the movement of water droplets in the concave pits of lotus leaves, to update the hyperparameter combination of the individual; determine the current iteration number, and if the current iteration number reaches the target iteration number, determine the hyperparameters corresponding to the best solution obtained in the enhanced exploitation phase as the optimized hyperparameters.

[0056] Exemplarily, taking the LEA-LSTM model as an example, set the population size of LEA to N, and determine the search range of the number of hidden layers h as [h min , h max , the search range of the number of neurons n is [n min , n max , the search range of the learning rate r is [r min , r max , etc.; randomly generate the initial population of LEA based on the search ranges corresponding to the hyperparameters. Each individual i in the initial population represents a set of hyperparameter combinations of the LEA-LSTM model [h i , n i , r i , ……], where i = 1, 2, …, N; the maximum number of algorithm iterations is T. Correspondingly, take the mean squared error of fault location of LEA-LSTM on the training set as the objective function, and construct the fitness function based on this. Among them, the mean squared error of fault location MSE and the fitness fitness respectively satisfy the following formulas:

[0057]

[0058] fitness = 1 / (1 + MSE)

[0059] Among them, m is the number of training samples, y j is the actual fault location of the j-th sample, is the fault location of the j-th sample predicted by the LEA-LSTM model under the parameter combination represented by each individual; further evaluate the fault location performance of the model by calculating the difference between the predicted location and the actual location. According to the fitness function, the smaller the mean squared error MSE, the higher the fitness fitness.

[0060] Correspondingly, for each individual in the initial population, based on the fitness function, according to the exploration stage principle of LEA, simulate the global search behavior of the dragonfly algorithm, and update the position of the individual in combination with the target factors. The position is also the hyperparameter combination of LEA-LSTM. In some embodiments, the target factors include separation behavior, alignment behavior, cohesion behavior, food attraction, and enemy distraction. According to the exploration stage principle of LEA, update the hyperparameter combination of the individual according to the target factors, including: for each factor in the target factors, determine the product of the factor value corresponding to the factor and the weight coefficient corresponding to the factor to obtain the weighted value corresponding to the factor; based on the sum of the weighted values corresponding to each factor, determine the velocity vector of the individual; according to the velocity vector and the current hyperparameter combination corresponding to the individual, adjust the position of the individual to obtain the updated hyperparameter combination.

[0061] Exemplarily, the separation behavior refers to the calculation of the separation degree between individual i and neighboring individuals. The separation degree S iSatisfy the following formula:

[0062]

[0063] where N i is the neighborhood set of individual i; r n is the neighborhood radius; t is the current iteration number; N′ is the number of individuals in N i neighborhood; X i t and X j t respectively represent the positions of individuals with current indices i and j in the t-th evolutionary iteration.

[0064] The alignment behavior means that individual i tends to align with the flight directions of neighborhood individuals, and the alignment degree A i satisfies the following formula:

[0065]

[0066] where V j t is the velocity vector of individual j in the t-th iteration optimization; |N i | is the number of neighborhood individuals; A i t represents the alignment degree of the individual with current index i in the t-th evolutionary iteration.

[0067] The cohesion behavior means that individual i tends to approach the central position of neighborhood individuals, and the cohesion degree C i satisfies the following formula:

[0068]

[0069] Assume there is a food source position F (which can be understood as the position of the optimal solution), and the food attraction degree F i t in the t-th evolutionary iteration satisfies the following formula:

[0070]

[0071] Assume there is an enemy position E (which can be understood as the position of a poor solution), and the enemy distraction degree E i t in the t-th evolutionary iteration satisfies the following formula:

[0072]

[0073] Combining the above five factors, the update formula for the velocity vector v i of individual i is:

[0074]

[0075] where ω is the inertia weight; a 1 、a 2 、a 3 、a 4 and a 5 correspond to the weight coefficients of the five factors of separation behavior, alignment behavior, cohesion behavior, food attraction, and enemy distraction respectively; the position vector X i of individual i is updated by the formula:

[0076]

[0077] Correspondingly, when an individual cannot find a better solution in the neighborhood, it will fly with a random movement step size, where the random movement step size L (similar to the velocity vector in the particle swarm algorithm) satisfies the following formula:

[0078]

[0079] where Levy(β) is a random vector following the Levy distribution, and β is the parameter of the Levy distribution. Correspondingly, the position update formula becomes:

[0080]

[0081] Furthermore, according to the principle of the exploitation stage (i.e., the local pollination stage) of LEA, local search is carried out with the best solution obtained in the exploration stage as the core to update the hyperparameter combination of the individual. For example, assuming that the position of the best solution currently found is X t best , in the exploitation stage of LEA, the position update formula of individual i is:

[0082]

[0083] where r l is a random coefficient in the interval [0,1]. As the number of iterations increases, the value range of r l gradually shrinks. In fact, at the beginning of the algorithm, the movement step size is long, and at the end of the algorithm, the movement step size is short until it converges to the optimum.

[0084]

[0085] In this algorithm, the values of the neighborhood radius r n (in the exploration stage) and r l (in the exploitation stage) of individual i are used to balance exploration and exploitation. r n is an incremental value, and the search range is gradually focused through the continuous increase of r n . The variable r lAdjust according to the number of repetitions of the algorithm to control the movement steps from long to short.

[0086] Furthermore, according to the principle of strengthening in the development stage of LEA, with the best solution obtained in the development stage as the core, simulate the movement of water droplets in the pits of lotus leaves for local search, and update the hyperparameter combination of individuals. Among them, by moving the water droplet to the first pit on the leaf, the pit will eventually be filled, and the water will overflow from the leaf into the next pit. Assume that the water droplet i (i.e., individual i) moves in the j-th pit (local optimal region) among k pits on the lotus leaf (solution space). Assume that the position of the water droplet i is X i , and the speed is V i , the deeper the pit and the larger its capacity, the higher the quality of the local optimal solution; in the t-th iteration optimization, the capacity of the pit j satisfies the following formula:

[0087]

[0088] Among them, C t j is the capacity of the pit j in the t-th evolutionary iteration, and f t ij is the fitness of the water droplet i in the pit j in the t-th evolutionary iteration, and f max , f min are the maximum and minimum fitness values in the pit j respectively; τ is a constant representing the maximum capacity of the target function MSE pit.

[0089] In the t-th iteration, the probability P i (transferring from the current pit j to another pit m) of the water droplet i to select different pits is calculated based on factors such as pit capacity:

[0090]

[0091] Among them, k is the number of all pits in the t-th iteration; after selecting the pit m, the overflowing water moves towards the pit m; when encountering a pit g with a larger capacity than the source pit j along the way, the overflowing water will be injected into the pit g with a larger capacity and the movement stops; otherwise, the overflowing water will continue to reach the selected pit m and be injected; the source pit j is removed at the end of the movement; the larger the capacity of the pit, the greater the probability of reusing it during the execution of the algorithm and the more likely it is to be selected to capture the overflowing water flowing towards it; during the local search process in the pits on the lotus leaf, consider a velocity increment q for each water droplet, then the water droplet velocity update and movement position satisfy the following formula:

[0092]

[0093] Correspondingly, the velocity update and movement position of the water droplet after overflowing from the previous pit satisfy the following formula:

[0094]

[0095] After being strengthened in the development stage of LEA, determine the current iteration number. If the current iteration number reaches the target iteration number, obtain the best solution obtained by the strengthening in the development stage, that is, the optimal parameter combination [h, n, r,...] of the LEA-LSTM model best , [h, n, r,...] best The corresponding hyperparameters are the optimized hyperparameters

[0096] Based on the above embodiments, in some embodiments, the method for locating the injection-type fault of the distribution network overhead line further includes: if the current iteration number does not reach the target iteration number, return to execute the step of updating the hyperparameter combination of each individual in the initial population according to the fitness function and the exploration stage principle of LEA according to the target factors until the target iteration number is reached

[0097] Exemplarily, if the current iteration number is 100 and the target iteration number is 200, in the case that the current iteration number does not reach the target iteration number, return to execute the exploration stage, development stage and development stage strengthening of LEA in sequence until the target iteration number 200 is reached

[0098] In the embodiments of the present application, LEA can accurately optimize the hyperparameters of the LEA-LSTM model through its unique exploration stage, development stage and development stage strengthening, so that the optimized model can more accurately capture the subtle changes of the signal in the time domain and frequency domain, thereby effectively improving the accuracy and reliability of fault location

[0099] In some embodiments, inputting the reflection signal into the pre-trained target fault location model includes: inputting the reflection signal into the target filter for filtering to obtain the filtered data, where the target filter includes a Butterworth filter or a Chebyshev filter; performing normalization processing on the filtered data to obtain the normalized data; formatting the normalized data according to the input format requirements of the target fault location model to obtain the formatted data; inputting the formatted data into the target fault location model

[0100] Exemplarily, the reflected signal is input into a target filter for filtering processing to remove clutter signals such as environmental noise and electromagnetic interference, and the data after filtering processing is obtained. The target filter includes but is not limited to Butterworth filters or Chebyshev filters, etc. Among them, when selecting a filter, it is necessary to select appropriate filter types and parameters such as the order and cut-off frequency of the filter according to the frequency characteristics of the signal and the noise distribution situation, so as to retain the useful information in the signal to the greatest extent and improve the signal-to-noise ratio of the signal. Correspondingly, the Min-Max (Minimum-Maximum) normalization method is used to map the amplitude range of the signal to an interval, and the calculation formula is as follows:

[0101]

[0102] where x is the amplitude of the original signal, x min and x max are the minimum and maximum values of the signal amplitude respectively, and x norm is the amplitude of the normalized signal; through the normalization process, the signals collected in different batches have a unified dimension, which is convenient for subsequent model processing, and also helps to improve the training efficiency and stability of the model; and according to the input format of the LEA-LSTM model, the data after normalization processing is converted into a three-dimensional tensor structure, and the dimensions are (number of samples, time step, number of features) in sequence; further, the formatted data is input into the LEA-LSTM model, and the position information of the fault point of the distribution network overhead line is output through the LEA-LSTM model.

[0103] In addition, in the model training stage, it is also necessary to divide the formatted data set. For example, 70%-80% of the data is used as the training set for model training; 10%-15% of the data is used as the validation set for adjusting the hyperparameters of the model; 10%-15% of the data is used as the test set for evaluating the performance of the model. By dividing the formatted data set, while ensuring sufficient training, the problem of model overfitting can be effectively avoided.

[0104] In the embodiment of the present application, by using a Butterworth filter or a Chebyshev filter, etc. to filter the reflected signal, noise and interference can be effectively removed, and the consistency of the data is ensured through normalization processing, so that signals of different amplitudes and scales can be uniformly processed, reducing the errors caused by data amplitude differences, thereby effectively improving the accuracy of fault location; in addition, the formatting process makes the data meet the input format requirements of the target fault location model, can better adapt to different types and sources of reflected signals, reduces the fault location errors caused by signal quality problems, improves the applicability of the model in various application scenarios and the performance of fault location, and enables it to provide accurate and reliable fault detection and location services in complex environments.

[0105] In some embodiments, the target fault location model is trained as follows: Obtain a training set, where the training set includes reflection signal samples and the actual fault locations corresponding to the reflection signal samples; Input the reflection signal samples into the target fault location model to obtain the predicted fault locations corresponding to the reflection signal samples; Based on the predicted fault locations and the actual fault locations, calculate the mean square error as the loss function; Use an optimization algorithm based on stochastic gradient descent to adjust the model parameters through backpropagation algorithm so that the value of the loss function of the target fault location model gradually decreases on the training set.

[0106] Exemplarily, according to the hyperparameter combination [h, n, r,...] optimized by LEA best Construct an LSTM neural network structure, use the pre-divided training set data to train the LSTM neural network structure, and use the mean square error as the loss function, and the loss function satisfies the following formula:

[0107]

[0108] where P is the number of training samples, and y p is the actual fault location of the p-th sample, is the fault location predicted by the model; Use the Adam optimizer (an optimization algorithm based on stochastic gradient descent) to update the parameters, calculate the gradient through the backpropagation algorithm, and make the value of the loss function of the model gradually decrease on the training set; During the training process, use the validation set data to evaluate and optimize the model, monitor performance indicators such as the loss function value and location accuracy of the model, and when the indicators no longer improve or show signs of overfitting, adjust the model hyperparameters or take regularization measures such as L1 / L2 regularization; After the training is completed, use the test set data to comprehensively evaluate the performance indicators of the model and obtain the prediction result of the final fault location; The relevant formulas involved in the model training process are shown as follows:

[0109]

[0110] Among them, Accuracy is the accuracy rate, that is, the proportion of samples correctly predicted by the model in the total samples; Recall is the recall rate, also known as the complete retrieval rate, that is, the proportion of positive class samples correctly identified by the model in the actual positive class samples; Precision is the precision rate, also known as the accurate retrieval rate, that is, among the samples predicted as true positives by the model, the proportion of those that are truly true positives; F1 refers to the harmonic mean, comprehensively balancing the precision rate and the recall rate; TP is the true positive, that is, the number of samples where the error between the actual fault location and the predicted fault location is within the allowable range; TN is the true negative, that is, the number of fault-free samples that are correctly judged as fault-free; FP is the false positive, that is, the number of fault-free samples that are wrongly judged as having faults; FN is the false negative, that is, the number of faulty samples that are wrongly judged as fault-free; MSE is the mean square error; MAE is the mean absolute error; R 2 is the coefficient of determination, and its value range is [0, 1]. The closer it is to 1, the better the fitting effect of the model; is the average value of the true values.

[0111] For example, in the experimental tests for various distribution network overhead line fault scenarios, compared with the traditional fault location method based on electrical quantity analysis, the average location error of the above LEA-LSTM model is reduced by about 40%; for short-circuit faults, the average location error of the traditional method is about 500 meters, while after using the LEA-LSTM model, the average location error can be controlled within 300 meters; for high-resistance faults, the error of the traditional method often exceeds 800 meters, and after using the LEA-LSTM model, the error can be reduced to about 500 meters; compared with the unoptimized LSTM model, the optimized model also has a significant improvement in location accuracy, and the average location error is reduced by about 25%; and in complex multi-fault scenarios, it can accurately identify and locate multiple fault points, and the accuracy rate is increased by more than 30% compared with the traditional method, greatly reducing the time and cost of fault troubleshooting and improving the operation and maintenance efficiency of the distribution network system.

[0112] Accordingly, in the simulation experiment of line parameter drift, when the line resistance changes within the range of ±20%, the fault location accuracy of the LEA-LSTM model can still remain above 90%, while the accuracy of the traditional machine learning model drops to about 70%; in the experiment of suffering from different degrees of electromagnetic interference, in the case of strong interference (such as signal-to-noise ratio of 5 dB) for the LEA-LSTM model, the false alarm rate is only 5% and the missed alarm rate is 3%; in the case of medium interference (signal-to-noise ratio of 15 dB), both the false alarm rate and the missed alarm rate are less than 1%; while for the traditional method under the same interference conditions, the false alarm rate and the missed alarm rate are as high as 15% and 10% respectively in the case of strong interference, and the false alarm rate and the missed alarm rate are as high as 8% and 5% respectively in the case of medium interference; when facing new fault types such as composite faults, after a small number of sample learning, the LEA-LSTM model can effectively locate new faults, and the location accuracy can reach more than 85%, fully reflecting its adaptability and robustness to new faults, and effectively ensuring the reliability and stability of the power system operation.

[0113] In the embodiment of the present application, the target fault location model optimized by LEA can adapt to complex scenarios such as the simultaneous existence of multiple fault points, complex and diverse fault types, and changes in line topology. Compared with the traditional machine learning model, the optimized model has stronger generalization ability when processing large-scale and high-dimensional signal feature data, avoids the overfitting problem, and can be accurately extended to the fault location tasks under different line conditions and fault conditions, with strong applicability; in addition, the global search and local optimization mechanisms of LEA endow the LEA-LSTM model with good adaptability and robustness, enabling it to automatically adjust parameters to maintain high performance when facing various changes and interferences during the operation of the distribution network overhead line, and for new fault types or complex mixed fault situations, after a small number of sample learning, it can quickly adapt and achieve relatively accurate fault location, demonstrating a powerful ability to cope with complex and changeable fault scenarios, and effectively ensuring the stable operation of the power system.

[0114] It should be noted that the above embodiments are mainly elaborated in detail taking the LEA-LSTM model as an example. For the LEA-GRU and BiLSTM models, the corresponding processes such as hyperparameter optimization, model training, and model application are similar to those of the LEA-LSTM model, and will not be elaborated here.

[0115] Figure 4 Another flowchart of the method for locating the injection-type fault of the distribution network overhead line provided by the exemplary embodiment of the present application. As Figure 4 shown, the method for locating the injection-type fault of the distribution network overhead line includes:

[0116] S401. Obtain a reflection signal dataset.

[0117] S402. Preprocess the reflection signal dataset, where the preprocessing includes at least one of filtering, normalization, dataset partitioning, and formatting.

[0118] Exemplarily, input the reflection signal into a target filter for filtering to remove clutter signals such as environmental noise and electromagnetic interference, and obtain the data after filtering. The target filter includes, but is not limited to, Butterworth filters or Chebyshev filters, etc. Among them, when selecting a filter, the appropriate filter type and parameters such as the order and cut-off frequency of the filter need to be selected according to the frequency characteristics of the signal and the noise distribution to maximize the retention of useful information in the signal and improve the signal-to-noise ratio; and format the data after normalization according to the input format of the model, for example, convert it into a three-dimensional tensor structure with dimensions (number of samples, time steps, number of features) in sequence; further, partition the formatted dataset. For example, 70%-80% of the data is used as the training set for model training; 10%-15% of the data is used as the validation set for adjusting the hyperparameter tuning of the model; 10%-15% of the data is used as the test set for evaluating the performance of the model.

[0119] S403. Optimize the hyperparameters of the target deep learning model based on LEA to obtain the optimized hyperparameters.

[0120] Among them, the target deep learning model includes LSTM models, GRU, or BiLSTM models, etc.; the hyperparameters include the number of hidden layers, the number of neurons, and the learning rate, etc. Exemplarily, randomly generate the initial population of LEA based on the search range corresponding to the hyperparameters. Each individual in the initial population represents a set of hyperparameter combinations; construct a fitness function based on the mean squared error of fault location on the training set of the target deep learning model. The fitness function is used to evaluate the performance of each set of hyperparameter combinations; for each individual in the initial population, based on the fitness function, according to the exploration stage principle of LEA, update the hyperparameter combination of the individual according to the target factors; according to the exploitation stage principle of LEA, perform local search with the best solution obtained in the exploration stage as the core to update the hyperparameter combination of the individual; according to the exploitation stage reinforcement principle of LEA, with the best solution obtained in the exploitation stage as the core, simulate the movement of water droplets in the pits of lotus leaves to perform local search and update the hyperparameter combination of the individual; determine the current iteration number. If the current iteration number reaches the target iteration number, then determine the hyperparameters corresponding to the best solution obtained in the exploitation stage reinforcement as the optimized hyperparameters.

[0121] S404. Construct a target deep learning model based on the optimized hyperparameters to obtain a target fault location model.

[0122] Exemplarily, according to the hyperparameter combination [h, n, r,...] optimized by LEA bestConstruct an LSTM neural network structure.

[0123] S405. Train the target fault location model to obtain a trained target fault location model.

[0124] Exemplarily, obtain a training set, where the training set includes reflection signal samples and the actual fault locations corresponding to the reflection signal samples; input the reflection signal samples into the target fault location model to obtain the predicted fault locations corresponding to the reflection signal samples; based on the predicted fault locations and the actual fault locations, calculate the mean square error as the loss function; adopt an optimization algorithm based on stochastic gradient descent, and adjust the model parameters through the backpropagation algorithm so that the value of the loss function of the target fault location model gradually decreases on the training set.

[0125] S406. When the target signal is injected into the distribution network overhead line, obtain the reflection signal of the target signal.

[0126] S407. Input the reflection signal into the trained target fault location model to obtain the location information of the fault point on the distribution network overhead line.

[0127] Exemplarily, input the reflection signal into the target filter for filtering to obtain the filtered data. The target filter includes a Butterworth filter or a Chebyshev filter; perform normalization processing on the filtered data to obtain the normalized data; format the normalized data according to the input format requirements of the target fault location model to obtain the formatted data; input the formatted data into the target fault location model.

[0128] Furthermore, based on the target fault location model such as the LEA-LSTM model, capture the subtle changes of the signal in the time domain and frequency domain, and perform analysis, so as to output the location information of the fault point on the distribution network overhead line.

[0129] In summary, the present application has at least the following advantages:

[0130] First, by optimizing the hyperparameters of the target deep learning model through LEA, the learning ability of the model for fault features in the reflection signal is significantly improved, so that the optimized model can more accurately capture the subtle changes of the signal in the time domain and frequency domain, thereby significantly improving the accuracy of fault location and fully meeting the high-precision location requirements in the complex distribution network environment; in addition, through high-accuracy fault location, the workload of manual inspection and fault troubleshooting is greatly reduced, the operation and maintenance cost is reduced, and the power outage time caused by faults is effectively reduced, significantly improving the power supply reliability and user satisfaction.

[0131] Second, by flexibly selecting a suitable injection signal type according to the electrical characteristics and fault location requirements of the distribution network overhead line, it helps to further improve the accuracy of fault location and effectively enhance the speed of fault location, which is of positive significance for reducing power outage time and improving user satisfaction, etc.

[0132] Third, through its unique exploration stage, development stage, and development stage enhancement, LEA can accurately optimize the hyperparameters of the LEA-LSTM model, enabling the optimized model to more precisely capture the subtle changes of signals in the time domain and frequency domain, thereby effectively improving the accuracy and reliability of fault location.

[0133] Fourth, the target fault location model optimized by LEA can adapt to complex scenarios such as the simultaneous presence of multiple fault points, diverse fault types, and changes in line topology. Compared with traditional machine learning models, the optimized model has stronger generalization ability when dealing with large-scale and high-dimensional signal feature data, avoiding overfitting problems, and can be accurately generalized to fault location tasks under different line conditions and fault conditions, with strong applicability; in addition, the global search and local optimization mechanisms of LEA endow the LEA-LSTM model with good adaptability and robustness, enabling it to automatically adjust parameters to maintain high performance when facing various changes and interferences during the operation of the distribution network overhead line. Moreover, for new fault types or complex mixed fault situations, after learning a small number of samples, it can quickly adapt and achieve relatively accurate fault location, demonstrating a powerful ability to handle complex and variable fault scenarios, effectively ensuring the stable operation of the power system.

[0134] The following is an embodiment of the device of the present application, which can be used to execute the method embodiment of the present application. For details not disclosed in the device embodiment of the present application, please refer to the method embodiment of the present application.

[0135] Figure 5 It is a schematic structural diagram of a positioning device for injection-type faults in a distribution network overhead line provided by an exemplary embodiment of the present application. As Figure 5 shown, the positioning device 50 for injection-type faults in the distribution network overhead line includes an acquisition module 51 and a fault location module 52, where:

[0136] The acquisition module 51 is used to acquire the reflected signal of the target signal when the target signal is injected into the distribution network overhead line;

[0137] The fault location module 52 is used to input the reflected signal into a pre-trained target fault location model to obtain the position information of the fault point on the distribution network overhead line. The target fault location model is constructed by optimizing the hyperparameters of the target deep learning model based on LEA, and the target deep learning model includes an LSTM model, a GRU, or a BiLSTM model.

[0138] In a possible implementation, the target signal includes at least one of a high-frequency pulse signal, a sine wave signal of a specific frequency, a multi-frequency modulation signal, and a chirp signal.

[0139] In a possible implementation, the hyperparameters include the number of hidden layers, the number of neurons, and the learning rate. The fault location module 52 can specifically be used to: randomly generate an initial population of LEA based on the search range corresponding to the hyperparameters, where each individual in the initial population represents a set of hyperparameter combinations; construct a fitness function based on the mean square error of fault location of the target deep learning model on the training set, and the fitness function is used to evaluate the performance of each set of hyperparameter combinations; for each individual in the initial population, based on the fitness function, according to the exploration phase principle of LEA, update the hyperparameter combination of the individual according to the target factors; according to the exploitation phase principle of LEA, perform local search with the best solution obtained in the exploration phase as the core to update the hyperparameter combination of the individual; according to the exploitation phase reinforcement principle of LEA, with the best solution obtained in the exploitation phase as the core, simulate the movement of water droplets in the pits of a lotus leaf to perform local search and update the hyperparameter combination of the individual; determine the current iteration number, and if the current iteration number reaches the target iteration number, determine the hyperparameters corresponding to the best solution obtained in the exploitation phase reinforcement as the optimized hyperparameters.

[0140] In a possible implementation, the fault location module 52 can also be used to: if the current iteration number does not reach the target iteration number, return to execute the step of updating the hyperparameter combination of each individual in the initial population based on the fitness function and according to the exploration phase principle of LEA according to the target factors until the target iteration number is reached.

[0141] In a possible implementation, the target factors include separation behavior, alignment behavior, cohesion behavior, food attraction, and enemy distraction. The fault location module 52 can also be used to: for each factor in the target factors, determine the product of the factor value corresponding to the factor and the weight coefficient corresponding to the factor to obtain the weighted value corresponding to the factor; based on the sum of the weighted values corresponding to each factor, determine the velocity vector of the individual; according to the velocity vector and the current hyperparameter combination corresponding to the individual, adjust the position of the individual to obtain the updated hyperparameter combination.

[0142] In a possible implementation, the fault location module 52 can also be used to: input the reflected signal into a target filter for filtering to obtain filtered data, where the target filter includes a Butterworth filter or a Chebyshev filter; perform normalization processing on the filtered data to obtain normalized data; format the normalized data according to the input format requirements of the target fault location model to obtain formatted data; input the formatted data into the target fault location model.

[0143] In a possible implementation manner, the target fault location model is trained as follows: Obtain a training set, where the training set includes reflection signal samples and the actual fault locations corresponding to the reflection signal samples; input the reflection signal samples into the target fault location model to obtain the predicted fault locations corresponding to the reflection signal samples; based on the predicted fault locations and the actual fault locations, calculate the mean square error as the loss function; adopt an optimization algorithm based on stochastic gradient descent, and adjust the model parameters through the backpropagation algorithm so that the value of the loss function of the target fault location model on the training set gradually decreases.

[0144] The fault location device for the injection type fault of the distribution network overhead line provided by the embodiment of the present application can execute the technical solutions shown in the embodiment of the above-mentioned fault location method for the injection type fault of the distribution network overhead line. The implementation principle and beneficial effects are similar, and will not be elaborated here.

[0145] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0146] Furthermore, it should be noted that although the steps in the flowchart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0147] It should be noted that the above device embodiments are merely illustrative, and the devices of the present application can also be implemented in other ways; moreover, it should be understood that the division of each module of the above device is only a division of logical functions, and in actual implementation, it can be fully or partially integrated into a physical entity, or physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also all be implemented in the form of hardware; they can also be partially implemented in the form of software called by processing elements and partially implemented in the form of hardware. For example, the determination module can be a separately established processing element, or can be integrated in a certain chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and the function of the above determination module can be called and executed by a certain processing element of the above device. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together or can be independently implemented. Here, the processing element can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by the integrated logic circuit in the processor element or the instruction in the form of software.

[0148] For example, the above modules can be one or more integrated circuits configured to implement the above method, such as: one or more application specific integrated circuits (ASICs), or, one or more microprocessors (DSPs), or, one or more field programmable gate arrays (FPGAs), etc. Again, when a certain module above is implemented in the form of a processing element scheduling program code, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processors that can call program code. Again, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0149] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from a website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, Digital Subscriber Line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a Digital Video Disc (DVD)), or a semiconductor medium (such as a solid state disk (SSD)).

[0150] Figure 6 The structural schematic diagram of the electronic device provided by the exemplary embodiment of the present application. As Figure 6 shown, the electronic device 60 in this embodiment includes:

[0151] At least one processor 61; and a memory 62 communicatively connected to the at least one processor;

[0152] Wherein, the memory 62 stores instructions executable by the at least one processor 61, and the instructions are executed by the at least one processor 61 to cause the electronic device to execute the method described in any of the above embodiments.

[0153] Optionally, the memory 62 can be either independent or integrated with the processor 61.

[0154] The memory 62 may include a high-speed random access memory (Random Access Memory, RAM) and may also include non-volatile memory, such as at least one disk memory.

[0155] The processor 61 may be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. Specifically, when implementing the positioning method for the injection-type fault of the distribution overhead line described in the foregoing method embodiments, the electronic device may be an electronic device with processing functions such as a server.

[0156] Optionally, the electronic device may further include a communication interface 63. In a specific implementation, if the communication interface 63, the memory 62, and the processor 61 are implemented independently, the communication interface 63, the memory 62, and the processor 61 may be interconnected through a bus and communicate with each other. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc., but it does not mean that there is only one bus or one type of bus.

[0157] Optionally, in a specific implementation, if the communication interface 63, the memory 62, and the processor 61 are integrated on a chip for implementation, the communication interface 63, the memory 62, and the processor 61 may complete communication through an internal interface.

[0158] For the implementation principle and technical effects of the electronic device provided in this embodiment, reference may be made to the foregoing embodiments, and details are not described herein again.

[0159] The embodiments of the present application further provide a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed, they are used to implement the method steps in the method embodiments as described above. The specific implementation manners and technical effects are similar and will not be described herein again.

[0160] The above computer-readable storage medium can be implemented by any type of volatile or non-volatile storage 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 memory, flash memory, a magnetic disk, or an optical disk. The readable storage medium can be any available medium accessible by a general-purpose or special-purpose computer.

[0161] An exemplary readable storage medium is coupled to the processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application-specific integrated circuit. Of course, the processor and the readable storage medium can also exist as discrete components in the positioning device for injection-type faults in distribution network overhead lines.

[0162] The embodiments of the present application also provide a computer program product, including a computer program, which implements the method steps in the above method embodiments when the computer program is executed. The specific implementation manners and technical effects are similar and will not be elaborated here.

[0163] In the above embodiments, the descriptions of the respective embodiments have their own focuses. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as falling within the scope described in this specification.

[0164] Those skilled in the art will readily conceive of other implementations of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any variations, uses, or adaptations of the present application, which follow the general principles of the present application and include known common general knowledge or conventional technical means in the technical field not disclosed in the present application. The specification and the embodiments are only regarded as exemplary, and the true scope and spirit of the present application are pointed out by the following claims.

[0165] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present application is only limited by the appended claims.

Claims

1. A method for locating an injection-type fault in a distribution network overhead line, characterized in that: include: When a target signal is injected into an overhead line of a distribution network, a reflected signal of the target signal is obtained; The reflected signal is input into a pre-trained target fault location model to obtain the location information of the fault point of the distribution network overhead line, wherein the target fault location model is constructed after optimizing the hyperparameters of a target deep learning model based on the lotus effect algorithm LEA, and the target deep learning model includes a long short-term memory network model, a gated recurrent unit or a bidirectional long short-term memory network model.

2. The method for locating injection-type faults of distribution network overhead lines according to claim 1, characterized in that: The target signal includes at least one of a high-frequency pulse signal, a sinusoidal wave signal of a specific frequency, a multi-frequency modulation signal and a chirp signal.

3. The method for locating an injection-type fault of a distribution network overhead line according to claim 1 or 2, characterized in that: The hyperparameters include the number of hidden layers, the number of neurons and the learning rate, and the hyperparameters are optimized in the following way: Randomly generate an initial population of the LEA based on a search range corresponding to the hyperparameter, wherein each individual in the initial population represents a set of hyperparameter combinations; Based on the fault location mean square error of the target deep learning model on the training set, construct a fitness function, wherein the fitness function is used to evaluate the performance of each set of hyperparameter combinations; For each individual in the initial population, based on the fitness function and according to the exploration phase principle of the LEA, the hyperparameter combination of the individual is updated according to the target factor; According to the development phase principle of the LEA, a local search is performed with the best solution obtained in the exploration phase as the core to update the individual hyperparameter combination; According to the development phase strengthening principle of the LEA, taking the best solution obtained in the development phase as the core, the movement of water droplets in the lotus leaf pits is simulated to perform local search and update the individual hyperparameter combination; The current number of iterations is determined. If the current number of iterations reaches the target number of iterations, the hyperparameters corresponding to the best solution obtained by strengthening the development phase are determined as the optimized hyperparameters.

4. The method for locating injection-type faults of distribution network overhead lines according to claim 3, characterized in that: Also includes: If the current number of iterations does not reach the target number of iterations, the step of returning to execute for each individual in the initial population, based on the fitness function, according to the exploration phase principle of the LEA, and updating the hyperparameter combination of the individual according to the target factors, until the target number of iterations is reached.

5. The method for locating injection-type faults of distribution network overhead lines according to claim 3, characterized in that: The target factors include separation behavior, alignment behavior, cohesion behavior, food attraction and enemy distraction. According to the exploration phase principle of the LEA, the hyperparameter combination of the individual is updated according to the target factors, including: For each factor in the target factors, determine the product of the factor value corresponding to the factor and the weight coefficient corresponding to the factor to obtain the weighted value corresponding to the factor; Determining a velocity vector of the individual based on the sum of weighted values ​​corresponding to each of the factors; According to the velocity vector and a current hyperparameter combination corresponding to the individual, the position of the individual is adjusted to obtain an updated hyperparameter combination.

6. The method for locating an injection-type fault of a distribution network overhead line according to claim 1 or 2, characterized in that: The step of inputting the reflected signal into a pre-trained target fault location model comprises: Inputting the reflected signal into a target filter for filtering to obtain filtered data, wherein the target filter includes a Butterworth filter or a Chebyshev filter; Normalizing the filtered data to obtain normalized data; Formatting the normalized data according to the input format requirements of the target fault location model to obtain formatted data; The formatted data is input into the target fault location model.

7. The method for locating an injection-type fault of a distribution network overhead line according to claim 1 or 2, characterized in that: The target fault location model is trained in the following way: Acquire a training set, wherein the training set includes reflection signal samples and actual fault locations corresponding to the reflection signal samples; Inputting the reflected signal sample into a target fault location model to obtain a fault prediction position corresponding to the reflected signal sample; Based on the predicted fault location and the actual fault location, calculating a mean square error as a loss function; An optimization algorithm based on stochastic gradient descent is adopted, and model parameters are adjusted through a back-propagation algorithm so that the loss function value of the target fault location model on the training set is gradually reduced.

8. A device for locating injection-type faults in distribution network overhead lines, characterized in that: include: An acquisition module, used for acquiring a reflected signal of the target signal when the target signal is injected into the overhead line of the distribution network; A fault location module is used to input the reflected signal into a pre-trained target fault location model to obtain the location information of the fault point of the distribution network overhead line, wherein the target fault location model is constructed after optimizing the hyperparameters of the target deep learning model based on the lotus effect algorithm LEA, and the target deep learning model includes a long short-term memory network model, a gated recurrent unit or a bidirectional long short-term memory network model.

9. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory is used to store computer-executable instructions; The processor is configured to execute the computer-executable instructions to implement the method according to any one of claims 1 to 7.

10. 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 7 when executed.

Images