Power distribution network fault location method and device based on BAS-PSO algorithm, equipment and medium
By combining edge computing, deep learning, and the BAS-PSO algorithm, the fault location method solves the problem of inaccurate fault location under complex and extreme conditions of traditional methods, and realizes rapid, accurate location and efficient recovery of distribution network faults.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- STATE GRID BEIJING ELECTRIC POWER CO
- Filing Date
- 2025-06-18
- Publication Date
- 2026-06-16
AI Technical Summary
Traditional fault location methods are difficult to achieve accurate and rapid fault location under complex and extreme weather conditions, especially in new energy power plants where fault location is inaccurate and response is slow.
A fault location method based on the BAS-PSO algorithm is adopted, which combines edge computing and deep learning. It uses a deep neural network model for fault diagnosis and the BAS-PSO algorithm for accurate location. This includes real-time data acquisition of edge nodes, feature extraction of a lightweight deep learning model, and local optimization search of the BAS-PSO algorithm.
It enables rapid and accurate fault location in the distribution network within milliseconds, improving the accuracy and response speed of fault location and adapting to the stability and reliability of complex network environments.
Smart Images
Figure CN120610108B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of power system fault diagnosis technology, specifically relating to a method, device, equipment and medium for fault location in distribution networks based on the BAS-PSO algorithm. Background Technology
[0002] The reliability and stability of the distribution network are crucial for the normal operation of the power system. However, with the continuous expansion of the power grid and the increase in the integration of new energy sources, the structure of the distribution network is becoming increasingly complex, which gradually increases the difficulty of fault location.
[0003] Traditional fault location methods typically rely on human experience or simple algorithms, which have significant limitations when dealing with complex environments and extreme weather conditions, easily leading to inaccurate fault location and slow response times. Especially in the application scenarios of renewable energy power plants, extreme weather conditions (such as high temperatures, heavy rain, and typhoons) have a significant impact on the stability of the distribution network, increasing the probability of system failures. Existing fault location technologies are insufficient to effectively address these challenges, necessitating a more advanced fault location method to improve the system's fault response capabilities and recovery efficiency. Summary of the Invention
[0004] The purpose of this invention is to provide a method, device, equipment, and medium for fault location in power distribution networks based on the BAS-PSO algorithm (BeetleAntennae Search Particle Swarm Optimization), so as to achieve rapid and accurate location of faults in power distribution networks and improve the intelligence level of fault location.
[0005] To achieve the above objectives, the present invention employs the following technical solution:
[0006] According to one aspect of the present invention, a method for fault location in a distribution network based on the BAS-PSO algorithm is provided, comprising the following steps:
[0007] Obtain the time-series characteristic sequence of the edge node of the distribution network, wherein the time-series characteristic sequence includes voltage, current and zero-sequence component;
[0008] The time-series feature sequence is input into a pre-trained deep neural network model embedded in the edge node, and the fault probability distribution result is output; candidate fault segments are selected based on the fault probability distribution result.
[0009] The candidate faulty section is uploaded to the central node, and the central node obtains the temporal feature sequence of the candidate faulty section and activates the BAS-PSO algorithm.
[0010] Using candidate fault sections as the initial search area, the BAS-PSO algorithm is used to obtain the predicted fault location and corresponding fault probability based on the temporal feature sequence of the candidate fault sections.
[0011] By employing the above technical solution, edge intelligent computing nodes are deployed at distribution network feeder terminals, switching stations, and distributed energy access points to collect waveform data such as voltage, current, and zero-sequence components in real time. A deep neural network model is then used for fault diagnosis of the distribution network, resulting in a more accurate and realistic model. A multi-layered fault response architecture combining edge computing, deep learning, and the BAS-PSO algorithm is constructed, enabling rapid local inference and determining the existence and possible location of faults within milliseconds. The deep neural network model for fault diagnosis is more accurate and closely reflects engineering realities. Furthermore, the use of statistical methods based on fault characteristics ensures accurate extraction of fault signals, improving the accuracy of fault location.
[0012] According to one embodiment of the present invention, the pre-trained deep neural network model is a CNN (Convolutional Neural Network) model or an LSTM (Long Short-Term Memory) neural network model.
[0013] By utilizing lightweight deep learning models embedded in edge nodes to rapidly extract features and identify anomalies in the obtained time-series feature sequences, the presence of power disturbances or potential fault signals can be determined, thereby improving response speed and location accuracy. Before using CNN or LSTM neural network models, historical time-series feature sequences of the distribution network are used, combined with historical fault locations of the distribution network, to train the deep neural network model.
[0014] According to one embodiment of the present invention, the BAS-PSO algorithm initializes the search region with candidate fault segments and obtains the predicted fault location and corresponding fault probability based on the temporal feature sequence of the candidate fault segments, including the following steps:
[0015] In the initial stage of fault location, the PSO algorithm is used for global search to obtain the fault range, and then the BAS algorithm is used for local optimization to obtain the fault point from the fault range.
[0016] Obtain the electrical parameters near the fault location and establish an electrical equivalent model based on the electrical parameters;
[0017] The error of the electrical equivalent model is minimized, and the predicted fault location and corresponding fault probability are obtained based on the result of the error minimization.
[0018] Furthermore, the electrical parameters near the fault location include the resistance and inductance values between the fault location and the converter stations on both sides; the electrical parameters near the fault location also include the impedance values between the fault location and the converter stations on both sides.
[0019] Let the resistance between the fault point and the left converter station be R1, and the inductance be L1; and the resistance between the fault point and the right converter station be R2, and the inductance be L2. The impedance between the fault point and both the left and right converter stations is... , In the formula, The angular frequency between the fault point and the left-side converter station. This is the angular frequency between the fault point and the right-side converter station. Due to the short circuit in the line... and Since they are difficult to measure, the four parameters R1, R2, L1, and L2 during a short circuit can be identified and measured.
[0020] According to one embodiment of the present invention, in the initial stage of fault location, the step of obtaining the fault range using PSO global search and then obtaining the fault point from the fault range using BAS local optimization includes the following steps:
[0021] In each iteration, the inertia weight and learning factor of PSO, as well as the step size and perturbation direction of BAS, are dynamically adjusted based on the current search results of PSO and BAS.
[0022] According to one embodiment of the present invention, in each iteration, the particle velocity of the PSO algorithm in the next iteration is updated according to the perturbation direction in the current iteration number of the BAS algorithm; the step size of the BAS algorithm in the next iteration is updated according to the current global optimal position obtained by the PSO algorithm and the current global optimal position obtained by the BAS algorithm.
[0023] According to one embodiment of the present invention, the real-time time-series feature sequence of the candidate fault segment is input into the BAS-PSO algorithm to obtain the predicted fault location and the corresponding fault probability step.
[0024] Distributed energy resources and backup power sources in the distribution network are dispatched based on the predicted fault location and corresponding fault probability.
[0025] According to one embodiment of the present invention, the step of scheduling distributed energy resources and backup power sources in a distribution network based on predicted fault locations and corresponding fault probabilities includes:
[0026] Using the actual power supply of distributed energy, the power demand of the distribution network load, and the transmission loss of the power supply path as decision variables, a power deviation loss objective function is constructed. With the goal of minimizing the sum of power deviation and transmission loss, a scheduling scheme for distributed energy and backup power is obtained.
[0027] By adopting the above technical solution and combining the BAS algorithm with the PSO algorithm, the convergence speed can be improved, effectively solving the problem of low search efficiency in the fault location process of distribution network. The location method comprehensively considers fault characteristics and network topology, ensuring stability and reliability in complex network environments.
[0028] According to one aspect of the present invention, a distribution network fault location device based on the BAS-PSO algorithm is provided, comprising:
[0029] The data acquisition module is used to obtain the time-series characteristic sequence of the power distribution network;
[0030] The neural network analysis module is used to input the time-series feature sequence into a pre-trained deep neural network model and output the fault probability distribution result; and to filter candidate fault sections based on the fault probability distribution result.
[0031] The fault location module is used to initialize the search area with candidate fault sections and use the BAS-PSO algorithm to obtain the predicted fault location and corresponding fault probability based on the temporal feature sequence of the candidate fault sections.
[0032] According to one aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the power distribution network fault location method based on the BAS-PSO algorithm of any of the above embodiments.
[0033] According to one aspect of the present invention, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the distribution network fault location method based on the BAS-PSO algorithm of any of the above embodiments.
[0034] Compared with the prior art, the present invention has at least the following beneficial effects:
[0035] 1. This invention utilizes a deep neural network model for fault diagnosis in distribution networks, constructing a multi-layered fault response architecture of "edge computing + deep learning + BAS-PSO algorithm." This architecture enables rapid local inference, determining the existence of faults and their possible location within milliseconds. The deep neural network model for distribution network fault diagnosis is more accurate and closely reflects engineering realities. Furthermore, the use of statistical methods based on fault characteristics ensures accurate extraction of fault signals, improving the accuracy of fault location.
[0036] 2. When performing fault location, the fault area of the distribution network is divided by combining the BAS algorithm and the PSO algorithm, and the fault location in each area is optimized and evaluated multiple times to achieve accurate fault location. This method can make full use of the global search capability of the PSO algorithm and combine the local optimization characteristics of the BAS algorithm, thereby improving the real-time performance and accuracy of fault response and location within milliseconds.
[0037] 3. The method of combining the BAS algorithm and the PSO algorithm can improve the convergence speed and effectively solve the problem of low search efficiency in the fault location process of distribution network. The location method takes into account the fault characteristics and network topology, ensuring stability and reliability in complex network environments. Attached Figure Description
[0038] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:
[0039] Figure 1 This is a flowchart of a power distribution network fault location method based on the BAS-PSO algorithm according to the present invention. Detailed Implementation
[0040] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.
[0041] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.
[0042] Example 1
[0043] A method for fault location in distribution networks based on the BAS-PSO algorithm is provided, such as... Figure 1As shown, it includes the following steps:
[0044] Obtain the time-series characteristic sequence of the edge node of the distribution network, which includes voltage, current and zero-sequence components;
[0045] The temporal feature sequence is input into a pre-trained deep neural network model embedded in the edge node, and the fault probability distribution result is output; candidate fault segments are selected based on the fault probability distribution result.
[0046] The candidate faulty section is uploaded to the central node, and the central node obtains the temporal feature sequence of the candidate faulty section and activates the BAS-PSO algorithm.
[0047] Using candidate fault sections as the initial search area, the BAS-PSO algorithm is used to obtain the predicted fault location and corresponding fault probability based on the temporal feature sequence of the candidate fault sections.
[0048] The pre-trained deep neural network model is either a CNN model or an LSTM neural network model.
[0049] The above method utilizes a deep neural network model for fault diagnosis in the distribution network. This model is more accurate and closely reflects engineering realities. Furthermore, statistical methods based on fault characteristics ensure accurate extraction of fault signals, improving the accuracy of fault location. A multi-layered fault response architecture combining edge computing, deep learning, and the BAS-PSO algorithm is constructed, enabling rapid local inference and determining the existence and possible location of faults within milliseconds. During fault location, the distribution network fault areas are divided using a combination of the BAS (Beetle Beard Algorithm) and PSO (Particle Swarm Optimization) algorithms. Multiple optimization evaluations are performed on the fault locations within each area, ultimately achieving precise fault location. This method fully leverages the global search capability of the PSO algorithm while combining the local optimization characteristics of the BAS algorithm, thereby improving the real-time performance and accuracy of fault response and location within milliseconds.
[0050] Specifically, taking a six-terminal distribution network as an example, the distribution network fault location method based on the BAS-PSO algorithm in this embodiment is as follows:
[0051] S1. Obtain the time-series characteristic sequence of the edge nodes of the distribution network.
[0052] Edge intelligent terminals are deployed at edge nodes such as distribution network feeder terminals, switching stations, and distributed energy access points. These terminals, combined with high-frequency sampling devices, collect real-time operating data such as local voltage, current, and zero-sequence components. The obtained real-time operating data is then normalized to obtain a time-series feature sequence.
[0053] S2. Use a deep neural network model to obtain and screen candidate fault sections.
[0054] By using a lightweight deep neural network model (CNN model or LSTM neural network model) embedded in the edge nodes and pre-trained, rapid feature extraction and anomaly identification are performed on time-series feature sequences to determine whether there are power disturbances or potential fault signals.
[0055] During the training phase, the lightweight deep neural network model (CNN model or LSTM neural network model) uses the historical time-series feature sequences of edge nodes (including voltage, current, and zero-sequence components) as the training set. The input of the lightweight deep neural network model contains the historical time-series feature sequences of multi-channel voltage, current, and zero-sequence components within a short time window. The output result is the judgment result of whether there is power disturbance or fault symptoms. The cross-entropy loss function is used during the training process.
[0056] The trained lightweight deep neural network model, when used, takes the time-series feature sequence of the distribution network edge node as input and outputs the fault probability distribution result, including the area where a fault may occur and the probability of a fault occurring in that area, etc., and filters out high-risk areas and marks candidate fault sections; generally, areas with a fault probability higher than 60% are considered high-risk areas and marked as candidate fault sections.
[0057] S3. The BAS-PSO algorithm is used to obtain the predicted fault location and the corresponding fault probability.
[0058] The candidate fault region is uploaded to the central node, which then activates the BAS-PSO algorithm. The central node's SCADA (Supervisory Control and Data Acquisition) system acquires the temporal feature sequence of the candidate fault region. The temporal feature sequence of the candidate fault region is then input into the BAS-PSO algorithm, which uses the candidate fault region as the initial search area to perform a search and obtain the predicted fault location and the corresponding fault probability.
[0059] In the BAS-PSO algorithm, during the initial stage of fault location, the PSO algorithm is used to perform a global search on the initial search area to obtain the fault range. The PSO algorithm has a strong global search capability in the early stage and can cover a wider solution space.
[0060] Specifically, the particle positions in the PSO algorithm are initialized to correspond to the candidate fault segments. The PSO algorithm iterates according to the following formula:
[0061] ;
[0062] ;
[0063] = + ( );
[0064] = + ( );
[0065] = + ( );
[0066] In the formula: T This represents the total number of iterations for the PSO algorithm. t This represents the current iteration number of the PSO algorithm. This represents the proportion of the current iteration number to the total number of iterations. As an intermediate variable; These are control parameters used to control the convergence rate. >1, In this embodiment, the following settings are provided =10;
[0067] Let be the inertia weight in the t-th iteration. , These are the upper and lower bounds of the inertia weight, typically set to 1.2 and 0.9 respectively;
[0068] Let be the cognitive coefficient in the t-th iteration. and These are the upper and lower bounds of the cognitive coefficient, typically set to 2.5 and 0.1, respectively.
[0069] Let be the social coefficient in the t-th iteration. and These are the upper and lower bounds of the social coefficient, usually set to 3.2 and 0.8 respectively.
[0070] and This is the learning factor of the PSO algorithm.
[0071] When the preset number of iterations is reached, the PSO algorithm terminates the iteration. The PSO algorithm dynamically adjusts the inertia weight. and learning factors and Perform a global search to quickly locate the approximate location of the fault.
[0072] After using the PSO algorithm to find the approximate location of the fault, the BAS algorithm is used to further refine the local area obtained by the PSO algorithm search. This further refinement search obtains the fault point from the fault area, improving the accuracy of fault location.
[0073] Specifically, the position of the longhorn beetle in the BAS algorithm is initialized to correspond to the approximate fault location obtained by the PSO algorithm. The BAS algorithm simulates the beetle's left and right whisker sensing behavior, calculates the evaluation function values of the left and right whisker positions, determines the direction with smaller error through the sign function, and performs a local fine search by combining the decay step size. The BAS algorithm iterates according to the following formula:
[0074] + +(1 ;
[0075] + ( ( );
[0076] sign ;
[0077] =eta ;
[0078] Longhorn beetle left and right search behavior:
[0079] ;
[0080] ;
[0081] In the formula: i Representing the first in the longhorn beetle herd i Only a longhorn beetle; s The dimension of each longhorn beetle is set to 1 in this embodiment; This represents the current iteration number of the BAS algorithm; Representative at the k In the nth iteration, the 1st i Only longhorn beetles in the dimension s The lower position; Representative refers to the first k The position of the beetle's left whisker in the next iteration Representative refers to the first k The position of the beetle's right whisker in the next iteration. and This is used to simulate how longhorn beetles adjust their position by sensing information from their left and right whiskers in order to search for the optimal solution.
[0082] Indicates the first kIn the next iteration, the position of the longhorn beetle's left whisker was targeted. The calculated objective function value;
[0083] Indicates the first k In the next iteration, the position of the longhorn beetle's left whisker was targeted. The calculated objective function value;
[0084] In the first k In the nth iteration i The speed of a longhorn beetle; Representative at the k In the nth iteration i The extreme value of only longhorn beetles; Representative at the k The extreme value of the longhorn beetle population in the next iteration; In the first k The position increment factor of the longhorn beetle in the next iteration;
[0085] is a random number in the range [0,1]; eta is the step size decay factor, which is 0.95 in this embodiment; Step size; The search distance is set to 0.5 in this embodiment; Inertial weight; For the first k Cognitive coefficient in the next iteration For the first k Social coefficient in the next iteration; and These are random numbers in the range [0, 1].
[0086] When the preset number of iterations is reached, the BAS algorithm terminates the iteration, obtaining a more precise fault point within the approximate range of the fault.
[0087] Furthermore, during the optimization process using the BAS-PSO algorithm, a joint fitness function can be used to coordinate the optimization objectives of BAS and PSO. The specific form of the joint fitness function is as follows:
[0088] = ( x )+ ( x );
[0089] in: x Indicates the location of the fault;
[0090] This is the evaluation function of PSO for the global solution;
[0091] This is the evaluation function of BAS for local optimal solutions;
[0092] For dynamic fusion coefficients, ∈ The contribution in the current iteration is updated dynamically.
[0093] ( x );
[0094] in:
[0095] Multiple energy sources in location Top load Actual power supply;
[0096] :load Power requirements;
[0097] Electrical energy travels from the source through the path Losses transmitted to the load;
[0098] n : Indicates the number of loads involved in the power deviation calculation;
[0099] M : Indicates the number of power supply paths participating in the scheduling, that is, the number of paths that generate transmission losses and are included in the optimization model;
[0100] : A loss weighting coefficient set manually.
[0101] ;
[0102] That is, the current fault location The left and right step lengths are The objective function values are calculated for each small range, and then the absolute difference is taken as the local evaluation index for the current point. The smaller the value, the closer the current position is to a local minimum, indicating a stable point with good convergence; conversely, a larger value indicates a more stable point. The larger the value, the more drastic the local fluctuations, and the further away it may be from the optimal position.
[0103] Furthermore, the PSO and BAS algorithms are optimized using a two-way information sharing mechanism. Specifically, after each iteration, the PSO and BAS algorithms share the following information: current optimal position, fitness value, historical trajectory, search density, etc. This information is passed through an intermediate buffer structure to adjust each other's initialization or parameter settings, thereby avoiding invalid searches and overlapping solution spaces.
[0104] Repeat the above process until the termination accuracy or maximum algebra is met, to obtain the predicted fault location and the corresponding fault probability.
[0105] Furthermore, in each iteration, a dynamic parameter coordination mechanism is used to optimize the PSO and BAS algorithms. Based on the current search results of the BAS and PSO algorithms, the inertia weight of the PSO algorithm is dynamically adjusted. Learning factors and and the step size of the BAS algorithm And the direction of perturbation. If the BAS algorithm obtains a high-quality local optimum in the current region, the inertia weight of the PSO algorithm is increased to expand the search range; if the PSO algorithm gets stuck in a local optimum, the BAS algorithm adjusts the step size to improve its ability to escape and prevent it from falling into a convergence trap.
[0106] Step size of the BAS algorithm The algorithm dynamically updates the results based on the global search information from the PSO algorithm, as shown in the following formula:
[0107] ;
[0108] in:
[0109] This is the step size decay coefficient;
[0110] For PSO global feedback coefficients;
[0111] This is the current global optimal solution position of PSO (i.e., the global optimal position in the particle swarm).
[0112] This represents the current local optimum location of BAS.
[0113] To further refine the identification of faulty equipment or faulty nodes, after determining the fault point using the BAS algorithm, electrical parameters near the fault location, such as voltage, current, resistance, and inductance, are obtained from sensors or databases in the distribution network. Based on these electrical parameters, mathematical models (such as resistance-inductance models and total impedance models) are used to model the distribution network and obtain an electrical equivalent model. Then, an error minimization algorithm (such as least squares method) is used to optimize the electrical equivalent model to reduce the error between predicted and actual measured values. The optimized electrical equivalent model is then used to further determine the fault location and fault probability.
[0114] The resistor-inductor model is: Z = R + jL;
[0115] In the formula, R is resistance, L is inductance, and j is the imaginary unit, representing the phase of the inductance;
[0116] Total impedance model: ;
[0117] In the formula, It is the voltage at the fault point. It is the current at the fault point. and These represent the impedance values from the fault point to the two converter stations, respectively.
[0118] Specifically, let the resistance between the fault point and the left converter station be R1, and the inductance be L1; and the resistance between the fault point and the right converter station be R2, and the inductance be L2. The impedance between the fault point and both the left and right converter stations is... , ,in, The angular frequency between the fault point and the left-side converter station. This is the angular frequency between the fault point and the right-side converter station. Due to the short circuit in the line... and Since they are not easy to measure, the four parameters R1, R2, L1, and L2 during a short circuit can be identified and measured.
[0119] Preferably, the following formula is discretized:
[0120] ;
[0121] get:
[0122] ;
[0123] ;
[0124] In the formula, P represents the parameter when the system fails. , The sampling constant;
[0125] , These represent the currents in different branches of the circuit, and are physical quantities used in circuit analysis to describe the directional movement of charges.
[0126] These represent the inductor elements on the left and right sides of the fault point, respectively. They are used to store magnetic field energy, and their characteristics affect the circuit behavior through their inductance. , They represent inductor components. , The voltage across the terminals reflects the voltage drop or rise caused by the inductor's opposition to changes in current.
[0127] These are the identifiers for two DC voltage sources, used to distinguish between different DC voltage sources, which helps to clarify the working path and relationship of each power source in circuit analysis and equation establishment; , These represent the voltage values of two DC voltage sources, providing a stable DC voltage input to the circuit.
[0128] and Let A be the system matrix for unipolar short-circuit fault and bipolar fault, respectively, where:
[0129] ;
[0130] ;
[0131] And from:
[0132] ;
[0133] ;
[0134] ;
[0135] We can obtain:
[0136] ;
[0137] ;
[0138] in: , r’ and l’ These are the resistance and inductance values per unit length of the line, respectively. ，D This refers to the total length of the lines in each section of the six-terminal distribution network. The distance from the fault location to the left converter station. This represents the distance from the location of the fault to the right-side converter station.
[0139] Since the above method has only one unknown. Therefore, solving for the above parameter only requires the voltage and current information at a certain moment after the short circuit. However, since the voltage and current information may have errors, the impedance value may have a large error. Therefore, it is necessary to combine multiple points. Based on the established electrical equivalent model, with the fault location as the only variable, the algorithm optimizes the cumulative error of multiple sampling points to achieve high-precision positioning.
[0140] Then, with the goal of minimizing the error, the electrical equivalent model is optimized. The error minimization model is as follows:
[0141] ;
[0142] In the formula, and These are actual measured voltage and current data. and The voltage and current are predicted based on the electrical equivalent model.
[0143] With the goal of minimizing error, the electrical model is optimized according to the above formula to obtain the specific location of the fault. Based on the time-series feature sequence and combined with the electrical parameter model, the probability of the fault occurrence can be calculated.
[0144] To minimize voltage and current information errors at individual moments, an evaluation function is established, and the location and probability are obtained based on the electrical equivalent model after error minimization. The BAS-PSO algorithm is then used for optimization, integrating information from multiple sampling points to minimize the error and achieve precise positioning.
[0145] The evaluation function is:
[0146] ;
[0147] ;
[0148] In the formula,
[0149] : Yes f from arrive A comprehensive error measure that sums and takes the absolute value; it assists in evaluating the overall error in the BAS-PSO algorithm optimization by aggregating multiple... This information reduces the impact of errors at individual sampling points on fault location.
[0150] The upper limit of the summation operation, indicating the number of values involved. Calculated The quantity, i.e., the number of sampling points or the number of calculations, reflects the scope of the comprehensive information. The larger the value, the more data it can use to reduce error interference.
[0151] The index variable for the summation operation, with a value range of 1 to 1. Used to iterate through each Accumulation is the construction The identifier that integrates information from a single sampling point.
[0152] : Indicates the location of the fault to be optimized;
[0153] The function value of a single sampling point or calculation unit reflects the fault location. Error-related information at a certain sampling point constitutes The basic elements, multiple pass The accumulation of To assist in algorithm optimization.
[0154] : Evaluation function, for From 1 to Calculate the measured voltage value at each sampling point. Compared with model calculated values The square of the error, plus the measured current value Compared with model calculated values The squared errors are then summed. The error is minimized using the BAS-PSO algorithm. By integrating information from multiple sampling points, errors in voltage and current information at a single moment are eliminated, and fault location is determined. Precise positioning.
[0155] , : respectively represent the first Voltage and current measured at a specific moment or sampling point;
[0156] , Based on the constructed electrical equivalent model, assuming the fault occurs at location... The voltage and current are calculated below.
[0157] The aforementioned distribution network fault location method based on the BAS-PSO algorithm enables the prediction of fault locations and probabilities in the distribution network. Subsequently, distributed energy resources and backup power supplies in the distribution network can be dispatched based on the predicted fault locations and corresponding probabilities. Generally, the BAS algorithm is used to prioritize the identification of critical equipment and load recovery sequences, while the PSO algorithm is combined to dynamically optimize the joint dispatch scheme of distributed energy resources (wind power, photovoltaic, energy storage) and backup power supplies, quickly restoring the system's power supply capacity and reducing the impact of power shortages on critical loads.
[0158] Specifically, the method for dispatching distributed energy resources and backup power sources in a distribution network includes the following steps:
[0159] (1) Use the BAS algorithm to identify key equipment and key loads first, and determine the priority order of power restoration.
[0160] (2) Combine the PSO algorithm to dynamically optimize the joint scheduling of distributed energy (wind power, photovoltaic, energy storage) and backup power to ensure rapid power restoration and reduce the power gap of critical loads.
[0161] The particle velocity update formula in the PSO algorithm is:
[0162] = + ( )+ ( ) ;
[0163] = + ;
[0164] In the formula:
[0165] :particle i exist t +1 generation speed; :particle i exist t The speed of generation;
[0166] Inertia weight;
[0167] Individual learning factor, representing the degree to which a particle is attracted by its own historical best position, i.e., cognitive coefficient;
[0168] : Group learning factor, representing the degree to which a particle is attracted to the global optimal position, i.e., the social coefficient;
[0169] The range of values is Random numbers are used to introduce randomness;
[0170] :particle i The historical best position;
[0171] The current global optimal position among all particles;
[0172] :particle i exist t The current position of the generation.
[0173] (3) Monitor power demand and energy status in real time and adjust recovery plans dynamically.
[0174] In the process of real-time monitoring of power demand and energy status, and dynamic adjustment of recovery plans, to optimize power allocation, a power deviation loss objective function is constructed based on the monitored power demand and energy status. The objective is to minimize the sum of power deviation and transmission loss. Solving the objective function yields the scheduling scheme for distributed energy resources and backup power sources in the distribution network. The objective function is as follows:
[0175] = ;
[0176] This function is equivalent to the PSO evaluation function for the global solution defined earlier. ,Right now: In the formula:
[0177] : Power deviation loss objective function;
[0178] Multiple energy sources i Actual power supply;
[0179] :load i Power requirements;
[0180] Power supply path j Transmission loss;
[0181] n : Indicates the number of loads involved in the power deviation calculation;
[0182] M : Indicates the number of power supply paths participating in the scheduling, that is, the number of paths that generate transmission losses and are included in the optimization model;
[0183] : Represents the weighting coefficient, The value can be determined empirically; for example, in cases where supply and demand balance is prioritized, a smaller value can be set. Value, such as ∈[0.01,0.1]; When prioritizing reducing transmission loss, a larger value can be set. Value, such as ∈[1,10].
[0184] To improve the speed and reliability of fault location in distribution networks, this embodiment proposes a fault location method based on BAS-PSO. The BAS and PSO algorithms are fundamental to fault location methods. By combining global search capabilities and local optimization characteristics, this method adapts to the high-dimensional search requirements of multi-objective optimization and complex power grid environments, enabling efficient fault location and recovery under extreme weather conditions. This method employs a phased closed-loop optimization process, with the algorithm running iteratively in the following stages:
[0185] Phase 1: The PSO algorithm dominates the global search to identify suspected fault areas;
[0186] The second stage: The BAS algorithm performs a fine search in the target region to determine the local optimum;
[0187] Phase 3: The PSO algorithm and BAS algorithm exchange results and reinitialize parameters, then provide feedback for optimization.
[0188] Repeat the above process until the termination precision or maximum algebra is satisfied.
[0189] First, the distribution network is divided into fault zones, and a network topology analysis system is used to accurately assess fault nodes in each zone, ensuring the accuracy and reliability of fault location. Then, using fault signal strength, equipment damage severity, and system recovery time as optimization objective functions, the BAS-PSO algorithm is applied to obtain a comprehensive fault location strategy for the distribution network, improving the fault recovery capability and reliability of the distribution network. Furthermore, the BAS algorithm prioritizes the identification of critical equipment and load recovery sequences, and the PSO algorithm dynamically optimizes the joint scheduling of wind power, photovoltaic power, energy storage, and backup power. For critical load demands during system recovery, priority is given to ensuring the power supply reliability of critical loads, significantly improving system recovery efficiency.
[0190] Example 2
[0191] A method for fault location in distribution networks based on the BAS-PSO algorithm is provided, which differs from Example 1 in that:
[0192] In the step of obtaining the predicted fault location and corresponding fault probability using the BAS-PSO algorithm, in order to improve the adaptive capability and global optimization efficiency of the BAS-PSO algorithm in complex search spaces, this embodiment proposes a fusion strategy of "dynamic parameter coordination mechanism" and "hybrid search incremental mechanism".
[0193] First, the dynamic parameter coordination mechanism can dynamically adjust the parameters of PSO and BAS in each iteration based on their current search states, including the inertia weight of PSO. Learning factor , and the step size of BAS directional disturbance All of them can adaptively adjust according to changes in fitness.
[0194] Inertia weight updates follow a linearly decreasing function:
[0195] ;
[0196] in This represents the maximum number of iterations for the PSO algorithm. This represents the current generation of the PSO algorithm. Secondly, the hybrid search incremental mechanism introduces a local perturbation vector provided by BAS into the PSO velocity update formula, improving the particle's fine-tuning ability in high-density regions. The corrected particle velocity update formula is:
[0197] ;
[0198] in:
[0199] :particle The speed in generation t+1;
[0200] : The particle's current position;
[0201] : Best historical position : Global optimal position;
[0202] : Local optimal direction vector of BAS;
[0203] Directional gain factor, which adjusts the influence of BAS information on velocity updates;
[0204] : Random number.
[0205] By introducing feedback from the BAS local perturbation direction into the velocity update of PSO particles, the particles will combine global and local information during the update, making the search more efficient.
[0206] BAS step size The update is performed dynamically based on the global search information of PSO, as shown in the following formula:
[0207] ;
[0208] in:
[0209] This is the step size decay coefficient;
[0210] The global feedback coefficient for PSO is used to adjust the intensity of the feedback influence of the optimal PSO solution on the local jump step size of BAS.
[0211] This is the current global optimal solution position of PSO (i.e., the global optimal position in the particle swarm).
[0212] This represents the current local optimum location of BAS.
[0213] The collaborative mechanism driver function is designed as follows:
[0214] ;
[0215] in:
[0216] : Coordination mechanism drives function values;
[0217] PSO optimization objective function;
[0218] BAS optimizes the objective function;
[0219] ;
[0220] ;
[0221] , : No. Voltage and current measured at a specific moment or sampling point;
[0222] , Based on the constructed electrical equivalent model, assuming the fault occurs at location... The voltage and current obtained from the calculation;
[0223] Dynamic fusion coefficients control the weight of the search strategy;
[0224] Fusion coefficient The dynamic adjustment formula is:
[0225] ;
[0226] in:
[0227] : Rate factor that controls the switching of the search phase;
[0228] : The current iteration number of the PSO algorithm;
[0229] : Maximum number of iterations for the PSO algorithm.
[0230] Example 3
[0231] A distribution network fault location device based on the BAS-PSO algorithm is provided, comprising:
[0232] The data acquisition module is used to obtain the time-series characteristic sequence of the distribution network, which includes voltage, current and zero-sequence components;
[0233] The neural network analysis module is used to input the time-series feature sequence into a pre-trained deep neural network model and output the fault probability distribution result; and to filter candidate fault sections based on the fault probability distribution result.
[0234] The fault location module is used to initialize the search area with candidate fault sections and use the BAS-PSO algorithm to obtain the predicted fault location and corresponding fault probability based on the temporal feature sequence of the candidate fault sections.
[0235] Example 4
[0236] A computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the power distribution network fault location method based on the BAS-PSO algorithm of Embodiment 1 or Embodiment 2.
[0237] Example 5
[0238] A computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the power distribution network fault location method based on the BAS-PSO algorithm of Embodiment 1 or Embodiment 2.
[0239] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A method for fault location in distribution networks based on the BAS-PSO algorithm, characterized in that, Includes the following steps: Obtain the time-series characteristic sequence of the edge node of the distribution network; the time-series characteristic sequence includes voltage, current and zero-sequence component; The time-series feature sequence is input into a pre-trained deep neural network model embedded in the edge node, and the fault probability distribution result is output; candidate fault segments are selected based on the fault probability distribution result. Using candidate fault segments as the initial search area, in the initial stage of fault localization, the PSO algorithm is used for global search to obtain the fault range, and then the BAS algorithm is used for local optimization to obtain the fault point from the fault range. In each iteration, based on the current search results of PSO and BAS, the inertia weight and learning factor of PSO, as well as the step size and perturbation direction of BAS, are dynamically adjusted; the step size of the BAS algorithm... The algorithm dynamically updates the results based on the global search information from the PSO algorithm, as shown in the following formula: ;in: This is the step size decay coefficient; For PSO global feedback coefficients; This represents the current global optimal solution position for PSO. The current local optimum location of BAS is determined; the electrical parameters between the fault point and the converter stations on both sides are obtained, and an electrical equivalent model is established based on the electrical parameters; the error of the electrical equivalent model is minimized, and the predicted fault location and corresponding fault probability are obtained based on the result of the error minimization; in the process of optimization using the BAS-PSO algorithm, a joint fitness function is used to coordinate the optimization objectives of BAS and PSO. The specific form of the joint fitness function is as follows: = ( x )+ ( x );in: x Indicates the location of the fault; This is the evaluation function of PSO for the global solution; This is the evaluation function of BAS for local optimal solutions; For dynamic fusion coefficients, ∈ It is dynamically updated based on the contributions of BAS and PSO in the current iteration.
2. The distribution network fault location method based on the BAS-PSO algorithm according to claim 1, characterized in that, The pre-trained deep neural network model is either a CNN neural network model or an LSTM neural network model.
3. The distribution network fault location method based on the BAS-PSO algorithm according to claim 1, characterized in that, In each iteration, the particle velocity of the PSO algorithm in the next iteration is updated according to the perturbation direction in the current iteration number of the BAS algorithm; the step size of the BAS algorithm in the next iteration is updated according to the current global optimal position obtained by the PSO algorithm and the current global optimal position obtained by the BAS algorithm.
4. The distribution network fault location method based on the BAS-PSO algorithm according to claim 1, characterized in that, The BAS-PSO algorithm initializes the search area with candidate fault sections and obtains the predicted fault location and corresponding fault probability based on the temporal feature sequence of the candidate fault sections. Distributed energy resources and backup power sources in the distribution network are dispatched based on the predicted fault location and corresponding fault probability.
5. The distribution network fault location method based on the BAS-PSO algorithm according to claim 4, characterized in that, The step of scheduling distributed energy resources and backup power sources in the distribution network based on the predicted fault location and corresponding fault probability includes: Using the actual power supply of distributed energy, the power demand of the distribution network load, and the transmission loss of the power supply path as decision variables, a power deviation loss objective function is constructed. With the goal of minimizing the sum of power deviation and transmission loss, a scheduling scheme for distributed energy and backup power is obtained.
6. A distribution network fault location device based on the BAS-PSO algorithm, characterized in that, include: The data acquisition module is used to obtain the time-series characteristic sequences of the edge nodes of the distribution network; The time-series characteristic sequence includes voltage, current, and zero-sequence components; The neural network analysis module is used to input the time-series feature sequence into a pre-trained deep neural network model embedded in the edge node, output the fault probability distribution result, and filter candidate fault segments based on the fault probability distribution result; The fault location module uses candidate fault segments as the initial search area. In the initial stage of fault location, it uses the PSO algorithm for global search to obtain the fault range, and then uses the BAS algorithm for local optimization to obtain the fault point from the fault range. In each iteration, based on the current search results of PSO and BAS, it dynamically adjusts the inertia weight and learning factor of PSO, as well as the step size and perturbation direction of BAS; the step size of the BAS algorithm... The algorithm dynamically updates the results based on the global search information from the PSO algorithm, as shown in the following formula: ;in: This is the step size decay coefficient; For PSO global feedback coefficients; This represents the current global optimal solution position for PSO. The current local optimum location of BAS is determined; the electrical parameters between the fault point and the converter stations on both sides are obtained, and an electrical equivalent model is established based on the electrical parameters; the error of the electrical equivalent model is minimized, and the predicted fault location and corresponding fault probability are obtained based on the result of the error minimization; in the process of optimization using the BAS-PSO algorithm, a joint fitness function is used to coordinate the optimization objectives of BAS and PSO. The specific form of the joint fitness function is as follows: = ( x )+ ( x );in: x Indicates the location of the fault; This is the evaluation function of PSO for the global solution; This is the evaluation function of BAS for local optimal solutions; For dynamic fusion coefficients, ∈ It is dynamically updated based on the contributions of BAS and PSO in the current iteration.
7. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the distribution network fault location method based on the BAS-PSO algorithm as described in any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the power distribution network fault location method based on the BAS-PSO algorithm as described in any one of claims 1 to 5.