Method, system and equipment for generating fault first-aid repair scheme of low-voltage power distribution network and medium

Abstract

The invention discloses a low-voltage power distribution network fault first-aid repair scheme generation method, system, equipment and medium, and the method comprises the steps: collecting the current, voltage and power data of a power grid in real time, carrying out the anomaly detection through employing an isolated forest algorithm, and generating a high-precision initial fault signal; performing intelligent classification on fault types and candidate fault points in combination with a convolutional neural network, and constructing a power distribution network model based on historical topological data; dynamically analyzing a fault propagation path by using a minimum spanning tree algorithm, and accurately extracting a fault influence range and key nodes; a first-aid repair resource allocation scheme is optimized through a genetic algorithm, and time and cost are balanced; an A * algorithm and real-time GIS data are fused to plan an optimal first-aid repair path, and the path efficiency is improved; and finally, constructing a multi-objective optimization model, planning the first-aid repair time, the resource cost and the power failure loss as a whole, and outputting a global optimal scheme. The whole process is intelligent, and the fault positioning precision, the resource allocation rationality and the path planning efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power systems and smart grids, and in particular to a method, system, equipment and medium for generating a fault repair plan for a low-voltage distribution network. Background Art

[0002] As low-voltage distribution networks expand and become increasingly complex, traditional fault repair methods rely primarily on human experience and manual troubleshooting, resulting in low efficiency and a response speed that is insufficient to meet the demands of modern power systems. Existing technologies, some of which employ rule-based expert systems or simple data analysis (such as threshold alarms) for fault detection, present the following problems: Inaccurate fault location: Traditional methods struggle to handle complex network topologies and real-time multi-source data (current, voltage, power), leading to a high rate of misjudgment of fault points and waste of repair resources; Irrational resource allocation: Existing technologies often employ static scheduling strategies that fail to consider the dynamic relationship between the fault impact range and key nodes, easily leading to uneven or redundant repair resource allocation; Inefficient path planning: Path planning based on fixed road networks fails to integrate real-time traffic conditions and grid topology constraints, resulting in actual repair times far exceeding expectations.

[0003] To alleviate the above problems, existing technologies attempt to introduce a single algorithm (such as Dijkstra algorithm or simple genetic algorithm) to optimize some links, but lack systematic integration and do not solve the global optimization problem of multi-objective conflicts (time, cost, resource utilization). Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method and system for generating emergency repair plans for low-voltage distribution network faults, which solve the technical problems of inaccurate fault positioning, static redundancy in resource configuration, and low efficiency in path planning caused by traditional reliance on manual experience. Through multi-algorithm collaborative optimization, the entire process is intelligentized, significantly improving emergency repair efficiency and resource rationality.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for generating a low-voltage distribution network fault repair plan, comprising:

[0008] Collect grid operation data in real time and input it into the algorithm for anomaly detection, generating a first-level fault signal;

[0009] Based on the primary fault signal and the collected historical data, secondary fault information is obtained through classification and modeling.

[0010] The algorithm analyzes and processes the secondary fault information to obtain the tertiary fault information, and the fault repair resource allocation plan is obtained through algorithm processing;

[0011] The fault repair resource allocation plan and the collected GIS data are processed through algorithms to obtain the repair path;

[0012] Based on the repair path and fault impact range, the target repair plan is obtained through model optimization.

[0013] As a preferred embodiment of the method for generating a low-voltage distribution network fault repair plan according to the present invention, the method includes: collecting grid operation data in real time and inputting it into an algorithm for abnormality detection to generate a first-level fault signal, including:

[0014] The grid operation data is collected in real time through sensors and processed into a standardized data set;

[0015] Based on the standardized data set, an isolation tree is constructed and the anomaly score of the data point is calculated after training;

[0016] Threshold determination is performed on the anomaly score to obtain an anomaly point set, and a preliminary fault time window is obtained through cluster analysis;

[0017] According to the preliminary fault time window and the line segment identification corresponding to the abnormal point, a first-level fault signal is obtained through correlation analysis, including the abnormal time point and abnormal line segment.

[0018] As a preferred embodiment of the method for generating a low-voltage distribution network fault repair plan according to the present invention, the second-level fault information is obtained through classification and modeling based on the first-level fault signal and the collected historical data, including:

[0019] The first-level fault signal and the collected historical data are converted and features are extracted to obtain a classification model;

[0020] After fault classification and topology modeling, secondary fault information is obtained, including fault type, candidate fault set and distribution network topology map.

[0021] As a preferred embodiment of the method for generating a low-voltage distribution network fault repair plan according to the present invention, the algorithm analyzes and processes the secondary fault information to obtain the tertiary fault information, including:

[0022] Among them, the third-level fault information includes the fault impact range and key nodes;

[0023] According to the secondary fault information, the minimum spanning tree is obtained after algorithm processing;

[0024] Based on the minimum spanning tree and candidate fault points, the fault impact range is obtained through analysis and processing;

[0025] According to the fault impact range and the distribution network topology, key nodes are obtained through extraction and processing.

[0026] As a preferred solution of the method for generating a low-voltage distribution network fault repair plan according to the present invention, the fault repair resource configuration plan obtained by algorithm processing includes:

[0027] The emergency repair resource allocation problem is modeled as a multi-objective optimization problem, where the objectives include minimum emergency repair time and resource cost;

[0028] Through algorithms, emergency repair time and resource costs are optimized simultaneously, resource allocation plans are found, and emergency repair resource allocation plans are dynamically adjusted.

[0029] As a preferred embodiment of the method for generating a low-voltage distribution network fault repair plan according to the present invention, the fault repair resource configuration plan and the collected GIS data are processed by an algorithm to obtain a repair path, including:

[0030] The starting and ending points in the fault repair resource allocation plan, as well as the road network and edge weights in the GIS data, are input into the A* algorithm to calculate the total cost and select the minimum node for expansion until the end point is reached.

[0031] Combine the road network and real-time traffic conditions in GIS data to generate the optimal repair path and output the specific route.

[0032] As a preferred embodiment of the method for generating a low-voltage distribution network fault repair plan according to the present invention, a target repair plan is obtained through model optimization processing based on the repair path and the fault impact range, including:

[0033] By building a multi-objective optimization model and considering the objective function, the final emergency repair plan is generated;

[0034] Among them, the objective function includes emergency repair time, resource cost and power outage loss.

[0035] In a second aspect, the present invention provides a system for generating a low-voltage distribution network fault repair plan, comprising:

[0036] An acquisition module is used to collect real-time grid operation data and input it into the isolation forest algorithm for anomaly detection to generate a preliminary fault signal, wherein the grid operation data includes current, voltage and power, and the preliminary fault signal includes the abnormal time point and abnormal line section;

[0037] The first processing module is used to obtain the fault type, candidate fault points and distribution network topology map based on the preliminary fault signal and the collected historical topology database through convolutional neural network fault classification and topology modeling;

[0038] The second processing module is used to analyze and process the fault type, candidate fault points and distribution network topology using the minimum spanning tree algorithm to obtain the fault impact range and key nodes, and then obtain the low-voltage distribution network fault repair resource allocation plan through genetic algorithm processing;

[0039] The third processing module is used to process the emergency repair resource allocation plan and the collected GIS data through the A* algorithm to obtain the optimal emergency repair path;

[0040] The generation module is used to obtain the final low-voltage distribution network fault repair plan based on the optimal repair path and fault impact range through multi-objective optimization model processing, where the objective function is to minimize the repair time, resource cost and power outage loss.

[0041] In a third aspect, the present invention provides an electronic device, comprising:

[0042] memory and processor;

[0043] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method for generating a low-voltage distribution network fault emergency repair plan are implemented.

[0044] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for generating a low-voltage distribution network fault repair plan.

[0045] Compared with the existing technology, the beneficial effects of the present invention are as follows: the present invention realizes accurate detection of current, voltage and power anomalies by real-time collection of power grid operation data and adopts the isolation forest algorithm to generate preliminary fault signals; combines convolutional neural networks to intelligently classify fault types and candidate fault points, and dynamically analyzes fault propagation paths based on the minimum spanning tree algorithm to generate fault impact ranges and key nodes; optimizes emergency repair resource allocation plans through genetic algorithms, integrates the A* algorithm with real-time GIS data to plan the optimal emergency repair path, and constructs a multi-objective optimization model to balance emergency repair time, resource cost and power outage losses; finally realizes the intelligentization of the entire process of low-voltage distribution network fault repair, significantly improves fault location accuracy, resource allocation rationality and path planning efficiency, is suitable for complex scenarios, and ensures reliable operation of the power grid. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 The figure is a schematic diagram of the overall process of a method for generating a low-voltage distribution network fault repair plan according to an embodiment of the present invention.

[0048] Figure 2 The present invention is a schematic structural diagram of a system for generating a low-voltage distribution network fault repair plan according to an embodiment of the present invention.

[0049] Figure 3 The present invention is a schematic diagram of the device structure of a method for generating a low-voltage distribution network fault emergency repair plan according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0051] Example 1, with reference to Figure 1 , as one embodiment of the present invention, provides a method for generating a low-voltage distribution network fault repair plan, comprising:

[0052] S100: Collects grid operation data in real time and inputs it into the algorithm for anomaly detection, generating a first-level fault signal;

[0053] S200: Based on the primary fault signal and the collected historical data, secondary fault information is obtained through classification and modeling.

[0054] S300: Analyze and process the second-level fault information through an algorithm to obtain third-level fault information, and obtain a fault repair resource allocation plan through algorithm processing;

[0055] S400: Processing the fault repair resource configuration plan and the collected GIS data through an algorithm to obtain a repair path;

[0056] S500: Based on the emergency repair path and the fault impact range, a target emergency repair plan is obtained through model optimization processing.

[0057] It should be noted that traditional low-voltage distribution network fault repair methods rely on manual judgment and experience, resulting in inaccurate fault location, static redundant resource allocation, and inefficient path planning. This leads to slow repair response, wasted resources, and difficulty in effectively controlling power outage losses. Existing technologies, such as rule-based expert systems or single algorithms, struggle to handle complex network topologies and multi-source data, and are unable to dynamically analyze the scope of fault impact and implement multi-objective collaborative optimization, severely impacting repair efficiency and grid reliability.

[0058] Therefore, in order to address the above-mentioned problems of inaccurate fault location, unreasonable resource allocation and unexpected repair time, steps S100-S500 are used to collect power grid operation data in real time and generate a first-level fault signal based on the isolation forest algorithm; the fault type and candidate points are dynamically determined by combining convolutional neural network classification modeling and topological analysis; the minimum spanning tree algorithm is used to analyze the fault propagation path and extract key nodes, and the resource allocation plan is optimized through the genetic algorithm; the A* algorithm is integrated with real-time GIS data to generate the optimal repair path; finally, a multi-objective optimization model is used to balance repair time, resource cost and power outage losses, realizing the intelligentization of the entire process of low-voltage distribution network fault repair, significantly improving positioning accuracy, rationality of resource allocation and path planning efficiency, and ensuring the reliable operation of the power grid.

[0059] Example 2, reference Figure 1 , which is an embodiment of the present invention, provides a method for generating a low-voltage distribution network fault repair plan based on the above embodiment.

[0060] In an embodiment of the present application, in step S100, grid operation data is collected in real time and input into an algorithm for anomaly detection. Real-time current, voltage, and power are collected by sensors, and after data cleaning and normalization, a standardized real-time electrical measurement data set is obtained. Based on the standardized real-time electrical measurement data, an isolation tree is constructed, which includes selecting ψ samples as training data for each iTree, recursively splitting the data until all samples are isolated or the maximum depth of the tree is reached, and calculating the path length in each iTree to obtain an anomaly score for each data point. The anomaly score is subjected to a threshold judgment to obtain an anomaly point set. Based on the anomaly point set, a preliminary fault time window is obtained through time clustering analysis. Based on the preliminary fault time window and the line segment identifier corresponding to the anomaly point, a preliminary fault signal is obtained through space-time correlation analysis.

[0061] In an optional embodiment, the real-time collection of grid operation data in step S100 and the input into the algorithm for anomaly detection can also be carried out through a static rule detection method of traditional threshold alarm, that is, a fixed threshold range of current, voltage and power is preset. When the real-time data exceeds the threshold, an anomaly mark is triggered, and the anomaly frequency is counted through a time window sliding statistics, and potential fault signals are screened in combination with the line topology relationship.

[0062] In another alternative embodiment, the real-time acquisition of power grid operation data in step S100 and inputting it into the algorithm for anomaly detection can also be achieved through the dynamic threshold detection method of statistical process control (SPC). The mean and standard deviation of each electrical quantity are calculated using historical data to construct dynamic control limits (such as ±3σ), and it is monitored in real time whether the data exceeds the control limits and anomalies are marked. At the same time, the anomaly trend is analyzed in combination with a sliding window, and the threshold offset is corrected through linear regression.

[0063] In the embodiment of the present application, in step S100, the real-time acquisition of power grid operation data and inputting it into the algorithm for anomaly detection to generate a primary fault signal further includes the following steps:

[0064] Collect real-time current, voltage, and power through sensors, and after data cleaning and normalization processing, obtain a standardized real-time electrical quantity measurement data set;

[0065] Based on the standardized real-time electrical quantity measurement data, construct isolation trees, where for each iTree, ψ samples are selected as training data, and the data is recursively divided until all samples are isolated or the maximum depth of the tree is reached, and the path length in each iTree is calculated to obtain the anomaly score for each data point;

[0066] After threshold determination of the anomaly scores, obtain a set of anomaly points; based on the set of anomaly points, through time clustering analysis, obtain a preliminary fault time window;

[0067] Based on the preliminary fault time window and the line segment identifier corresponding to the anomaly point, through space-time correlation analysis, obtain a preliminary fault signal.

[0068] Specifically, in step S100, according to the original current I(t), voltage V(t), and power P(t) collected by the smart meter and sensors, obtain the standardized real-time electrical quantity measurement data set X = {x i , i , i , i , i , i , , i ,

[0069] , ∣x i =(I i ,V i ,P i )}. For each iTree, randomly select ψ samples (usually ψ = 256) as training data; recursively divide the data until all samples are isolated or the maximum depth of the tree is reached Then randomly select a feature F ∈ {I, V, P}; randomly select a split value p between the minimum and maximum values of the feature F; divide the data into left and right subtrees: the left subtree contains samples with F < p, and the right subtree contains samples with F ≥ p;

[0069] Then calculate the path length h(x i in each iTree for each data point x i), the path length is the number of edges from the root node to the leaf node, if x i Isolated in iTree, the path length is the depth of the current tree;

[0070] Calculate the anomaly score, that is, take the average of the path lengths of all iTrees E(h(x i )), the normalization factor c(N) is calculated as:

[0071]

[0072] Where H(k) is the harmonic number, H(k) = ln(k) + 0.5772 (Euler constant), N is the number of samples in the dataset, and H(N-1) is the N-1th harmonic number.

[0073] Calculate the anomaly score, which is:

[0074]

[0075] When s(x i ) is close to 1, indicating that x i It is very likely to be an outlier; when s(x i ) is close to 0, indicating that x i Most likely normal;

[0076] According to the anomaly score s(x i ), after threshold determination (set s th =0.7), we get the abnormal point set A = {x i ∣s(x i )>s th}, that is, if s(x i )>s th , then mark x i For abnormal points.

[0077] Cluster the continuous abnormal points to form the fault time period T f =[t start ,t end ], t start ,t end Represent the start time point and the end time point respectively, according to the line segment L where the abnormal point is located j , generating a preliminary fault signal:

[0078] Initial fault signal = {(t start ,t end )}

[0079] Among them, t i is the abnormal time point, L j It is an abnormal line segment.

[0080] In the embodiment of the present application, in step S200, according to the primary fault signal and the collected historical data, secondary fault information is obtained through classification and modeling processing. The preliminary fault signal and the historical topology database are converted into a two-dimensional feature map, and multiple convolution kernels are used to extract features from the two-dimensional feature map to obtain a classification model, wherein the historical topology database includes the ring network cabinet, the branch box location and the line connection relationship. After convolutional neural network fault classification and topological structure modeling, the fault type, the candidate fault set and the distribution network topology map are obtained; wherein the convolutional neural network fault classification is a set dimension, and the spatiotemporal-electrical features are extracted through the convolution kernel K, and the fault type probability distribution is output;

[0081] Construct a distribution network topology map based on the device coordinates and connection relationships in the historical topology database.

[0082] In an optional embodiment, in step S200, the secondary fault information is obtained through classification and modeling based on the primary fault signal and the collected historical data. Traditional feature engineering machine learning methods can also be used, that is, statistical features such as line impedance mean, current fluctuation variance, and voltage deviation trend are manually extracted from historical topology data, and a feature vector is constructed in combination with the spatiotemporal distribution of the fault signal, and a support vector machine (SVM) or random forest (RF) classification model is used to distinguish the fault type.

[0083] In another optional embodiment, in step S200, the secondary fault information is obtained through classification and modeling based on the primary fault signal and the collected historical data. A rule-matching expert system can also be used, that is, a rule base of current, voltage and power change thresholds under predefined typical fault scenarios. The fault type and candidate areas are determined by matching the real-time fault signal with the rule base entries, and the fault impact range is manually marked based on the topological connection relationship.

[0084] In the embodiment of the present application, step S200 obtains the second-level fault information through classification and modeling based on the first-level fault signal and the collected historical data, and further includes the following steps:

[0085] The preliminary fault signal and historical topology database are converted into a two-dimensional feature map. Multiple convolution kernels are used to extract features from the two-dimensional feature map to obtain a classification model. The historical topology database includes the location of the ring main unit and branch box and the line connection relationship. After convolutional neural network fault classification and topology structure modeling, the fault type, candidate fault set and distribution network topology map are obtained. The convolutional neural network fault classification is used as the setting dimension, and the convolution kernel K is used to extract spatiotemporal-electrical features and output the probability distribution of the fault type.

[0086] According to the equipment coordinates and connection relationships in the historical topology database, the distribution network topology graph G = (V, E, W) is constructed, where the edge weight is the line impedance Zline (e), which is calculated as follows:

[0087]

[0088] Where R is resistance, X is reactance, and E represents the set of all edges in the topology graph;

[0089] It should be noted that the fault types are:

[0090] y k ∈{short circuit, ground, overload}

[0091] The candidate fault point set is expressed as:

[0092] Z={z m |z m =(v m ,l m )}

[0093] Among them, v m For device nodes, l m is the line segment, z m is the mth candidate fault point;

[0094] The distribution network topology is G = (V, E, W), where the node V is the ring main unit / branch box, the edge E is the line, and the weight W is the line impedance Z. line (e);

[0095] In this step, the CNN classification model is generated by concatenating the fault signal and topology data as the input feature map X, with a dimension of H*W*3 (time*space*electrical quantity)

[0096] The feature extraction of convolution kernel K is expressed as:

[0097]

[0098] Among them, X' h,w,c is the output feature map, σ is the activation function, Perform weighted summation on the k×k local area covered by the convolution kernel, K i,j,c K is the weight parameter of the convolution kernel at position (i, j) and channel c, h+i-1,w+j-1,c is the local data aligned with the convolution kernel in the input feature map, b c is the bias term.

[0099] Among them, the output fault type probability P(y k ) is expressed as:

[0100] P(y k )=softmax(x′ fc )

[0101] Among them, x' fc is the original output vector of the fully connected layer, and softmax is the probability distribution function.

[0102] In the embodiment of the present application, in step S300, the secondary fault information is analyzed and processed by the algorithm according to the fault type y k , candidate fault points and topology graph G = (V, E, W), processed by the minimum spanning tree algorithm, the minimum spanning tree is obtained

[0103] Based on the minimum spanning tree and candidate fault points, the fault impact range is obtained through fault propagation analysis and processing.

[0104] According to the fault impact range and topology diagram, the key nodes are extracted and processed to obtain the key nodes.

[0105] In an optional embodiment, the analysis and processing of the secondary fault information by the algorithm in step S300 can also be carried out through a breadth-first search (BFS) path traversal method, that is, based on the candidate fault points and the topology diagram, the adjacent lines are traversed layer by layer with the substation as the root node, and the set of lines affected by the fault current is counted as the fault impact range, and the key nodes are screened by the number of node connections.

[0106] In another optional embodiment, the analysis and processing of the secondary fault information by the algorithm in step S300 can also be carried out through a fixed weight path selection rule, that is, the physical length or rated capacity of the predefined line is used as the basis for path selection, and the shortest physical distance or highest capacity line is preferentially selected through a greedy algorithm to construct the fault propagation path, and the key nodes are extracted based on the node degree threshold.

[0107] In the embodiment of the present application, in step S300, the second-level fault information is analyzed and processed by an algorithm to obtain the third-level fault information, and the fault repair resource configuration plan is obtained by the algorithm processing, and the following steps are also included:

[0108] According to the fault type k , candidate fault points and topology graph G = (V, E, W), processed by the minimum spanning tree algorithm, the minimum spanning tree is obtained The calculation formula that T satisfies is expressed as:

[0109] Connected and loop-free

[0110] Among them, T' is the candidate edge set;

[0111] Based on the minimum spanning tree and candidate fault points, the fault impact range is obtained through fault propagation analysis and processing. Where R is the number of nodes from candidate fault point z m The path from edge e to the root node is calculated as follows:

[0112]

[0113] Wherein, the range represents the set of lines affected by the fault current, and e' represents the path from edge e' to the root node of the minimum spanning tree (usually a substation);

[0114] According to the fault impact range and topology diagram, the key nodes are extracted and processed to obtain the key nodes. Among them, V key is a node with a degree greater than or equal to 2 in R, V represents the vertex set in the distribution network topology graph, and its calculation formula is expressed as:

[0115] V key ={v∈V|deg(v)≥2∧V∈R}

[0116] Where v is a vertex in the distribution network topology graph, and deg(v) is the degree of node v.

[0117] It should be noted that the minimum spanning tree algorithm is a greedy algorithm that sorts edges in ascending order and gradually selects edges to construct a minimum spanning tree (MST) to ensure that there is no loop and the total impedance is minimized. Specifically, all edges E of the topology graph G are sorted according to the impedance Z. line Sort in ascending order and initialize an empty set T to store the edges of the minimum spanning tree. Traverse the sorted edges and select e in turn to add to T to ensure that no loops are formed. This process continues until T contains |V|-1 edges, ultimately resulting in a minimum spanning tree. By constructing a minimum spanning tree, the impedance of the fault propagation path is minimized and the complex distribution network is simplified into a tree structure, facilitating subsequent fault propagation analysis and key node extraction.

[0118] In low-voltage distribution networks, fault currents propagate along the path of least impedance, with the impact range depending on the fault type and network topology. Minimum spanning tree and fault propagation analysis can accurately identify affected lines and equipment, avoiding blind repairs. This reduces troubleshooting time and improves repair efficiency.

[0119] When extracting key nodes, we prioritize equipment nodes (such as ring main units and branch boxes) that connect multiple lines in the distribution network. Failures there could lead to wider power outages. Nodes with a degree ≥ 2 are considered key nodes. Key nodes are ring main units or branch boxes on the critical power supply path and require priority repair. By extracting key nodes, we clarify repair priorities and ensure that critical equipment and lines are restored first, thus avoiding wasting resources on non-critical nodes and improving repair efficiency.

[0120] Therefore, fault propagation models can further incorporate dynamic load data and real-time weather conditions (such as temperature and humidity) to optimize the accuracy of fault impact prediction. For example, high temperatures may cause line impedance changes, necessitating dynamic adjustment of the minimum spanning tree weights. Furthermore, in actual distribution networks, multiple concurrent faults may exist. The minimum spanning tree algorithm can be extended to support propagation analysis across multiple fault points. For example, a multi-source shortest path algorithm (such as a multi-source version of the Dijkstra algorithm) can be used to account for the impact of multiple fault points.

[0121] It is understandable that the emergency repair resource allocation problem is modeled as a multi-objective optimization problem, with the objectives of minimizing the emergency repair time t(S) and resource cost c(S). The emergency repair resource allocation plan is encoded as a chromosome, with each gene representing a resource allocation decision. For example, chromosome C can be expressed as:

[0122] C={(p i ,q j ,r k )→v m |p i ∈P,q j ∈Q,r k ∈R,v m ∈V}

[0123] Among them, p i For personnel, q j For equipment, r k For the vehicle, v m is the key point, P is the personnel set;

[0124] The fitness function f(S) is defined to evaluate the quality of resource allocation schemes as follows:

[0125]

[0126] Where, t(v m ) is the emergency repair task v m The completion time includes the driving time and the repair time; c(q j ,r k ) is the device q j and vehicle r k α and β are weight coefficients used to balance the priorities of time and cost.

[0127] Initialize the population and, based on the fitness function f(S), select the best individuals to advance to the next generation. Crossover operations are performed on these selected individuals to generate new offspring, and the offspring chromosomes are randomly mutated to increase population diversity. For example, the resource allocation decision for a gene is randomly changed. The selection, crossover, and mutation operations are repeated until the maximum number of iterations is reached or the fitness function converges. The optimal emergency repair resource allocation plan Sresource is then output: it represents the optimal allocation plan for personnel, equipment, and vehicles. Through a genetic algorithm, both emergency repair time and resource costs are optimized to ensure the efficiency and cost-effectiveness of the resource allocation plan, avoid falling into local optimal solutions, find a more optimal resource allocation plan, and dynamically adjust the resource allocation plan to adapt to complex and changing emergency repair scenarios.

[0128] In the embodiment of the present application, in step S400, the fault repair resource configuration plan and the collected GIS data are processed by an algorithm to obtain a repair path, including:

[0129] The starting point and end point in the emergency repair resource allocation plan, as well as the road network and edge weights in the GIS data, are input into the A* algorithm to calculate the total cost. The smallest node is selected for expansion until the end point is reached. The formula for calculating the total cost f(v) is expressed as:

[0130] f(v)=g(v)+h(v)

[0131] Where g(v) is the actual cost from the starting point to v, and h(v) is the Euclidean distance from node v to the end point g in the heuristic function;

[0132] Combine the road network and real-time traffic conditions in GIS data to generate the optimal repair path and output the specific route.

[0133] It is understandable that the channels for collecting GIS data include: the power company's internal geographic information system (GIS) database: storing the location of distribution network equipment, line topology and related road information; open road network data shared by municipal traffic management departments: including road grades, real-time traffic status and closed road section information; and third-party map service platforms (such as Open StreetMap): providing high-precision road network topology and dynamic path weights (such as travel time).

[0134] In the embodiment of the present application, in step S500, based on the emergency repair path and the fault impact range, a target emergency repair plan is obtained through model optimization processing, including:

[0135] By constructing a multi-objective optimization model, the final emergency repair plan is generated by comprehensively considering emergency repair time, resource cost and power outage losses, achieving comprehensive optimization of emergency repair efficiency, economy and user satisfaction, and solving the local optimal problem caused by single-objective optimization in traditional methods.

[0136] In summary, the present invention significantly solves the core problems of large fault location deviation, static redundancy of resource configuration and low path planning efficiency in traditional methods through the integration of full-process intelligent technology, effectively improves the emergency response speed, rationality of resource utilization and reliability of power grid operation, and is suitable for complex scenarios such as old communities and high-load areas, providing efficient technical guarantee for the safe and stable operation of low-voltage distribution networks.

[0137] Example 3. The above is a schematic scheme of a method for generating a low-voltage distribution network fault repair plan. It should be noted that the technical scheme of the system for generating a low-voltage distribution network fault repair plan is the same as the technical scheme of the method for generating a low-voltage distribution network fault repair plan. For details not described in detail in the technical scheme of the system for generating a low-voltage distribution network fault repair plan in this embodiment, please refer to the description of the technical scheme of the method for generating a low-voltage distribution network fault repair plan.

[0138] Reference Figure 2 This embodiment also provides a low-voltage distribution network fault repair plan generation system, including:

[0139] Acquisition module 701: used to collect real-time grid operation data and input it into the isolation forest algorithm for anomaly detection to generate preliminary fault signals, wherein the grid operation data includes current, voltage and power, and the preliminary fault signals include the abnormal time point and abnormal line section;

[0140] Specifically, the acquisition module 701 further includes:

[0141] The first processing unit is used to collect real-time current, voltage and power through sensors, and obtain a standardized real-time electrical measurement data set through data cleaning and normalization;

[0142] The first construction unit is used to construct an isolation tree based on the standardized real-time electrical measurement data. This includes selecting ψ samples as training data for each iTree, recursively splitting the data until all samples are isolated or the maximum depth of the tree is reached, and calculating the path length in each iTree to obtain the anomaly score of each data point.

[0143] The first analysis unit is used to determine the anomaly score by threshold value to obtain an anomaly point set; based on the anomaly point set, a time cluster analysis is performed to obtain a preliminary fault time window;

[0144] The second analysis unit is used to obtain a preliminary fault signal through space-time correlation analysis based on the preliminary fault time window and the line segment identifier corresponding to the abnormal point.

[0145] The first processing module 702 is used to obtain the fault type, candidate fault points and distribution network topology map based on the preliminary fault signal and the collected historical topology database through convolutional neural network fault classification processing and topology modeling processing;

[0146] Specifically, the first processing module 702 further includes:

[0147] Model building unit: used to convert preliminary fault signals and the historical topology database into a two-dimensional feature map, and use multiple convolution kernels to extract features from the two-dimensional feature map to obtain a classification model. The historical topology database includes the locations of ring main units and branch boxes and line connections. After convolutional neural network fault classification and topology structure modeling, the fault type, candidate fault set, and distribution network topology map are obtained. The convolutional neural network fault classification is used as a set dimension, and the convolution kernel is used to extract spatiotemporal-electrical features to output the probability distribution of the fault type.

[0148] The second construction unit is used to construct a distribution network topology map according to the device coordinates and connection relationships in the historical topology database.

[0149] The second processing module 703 is used to analyze and process the fault type, candidate fault points and distribution network topology using a minimum spanning tree algorithm to obtain the fault impact range and key nodes, and then obtain a low-voltage distribution network fault repair resource allocation plan using a genetic algorithm.

[0150] Specifically, the second processing module 703 further includes:

[0151] The second processing unit is used to obtain a minimum spanning tree based on the fault type, candidate fault points and topology diagram through a minimum spanning tree algorithm;

[0152] The third processing unit is used to obtain the fault impact range through fault propagation analysis and processing based on the minimum spanning tree and candidate fault points;

[0153] The fourth processing unit is used to obtain key nodes through key node extraction processing according to the fault impact range and the topology map.

[0154] The third processing module 704 is used to process the emergency repair resource allocation plan and the collected GIS data using the A* algorithm to obtain the optimal emergency repair path;

[0155] Specifically, the third processing module 704 further includes:

[0156] Calculation unit: used to input the starting point and end point in the emergency repair resource allocation plan and the road network and edge weights in the GIS data into the A* algorithm, calculate the total cost, and select the smallest node to expand until the end point is reached;

[0157] Generation unit: used to combine the road network in GIS data and real-time traffic conditions to generate the optimal repair path and output the specific route.

[0158] Generation module 705: is used to obtain the final low-voltage distribution network fault repair plan based on the optimal repair path and fault impact range through multi-objective optimization model processing, where the objective function is to minimize the repair time, resource cost and power outage loss.

[0159] Reference Figure 3 This embodiment also provides an electronic device suitable for generating a low-voltage distribution network fault repair plan, comprising: a low-voltage distribution network fault repair plan intelligent generation device 800, a processor 801, and a memory 802; the low-voltage distribution network fault repair plan intelligent generation device 800 also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805; the processor 801 is used to control the overall operation of the low-voltage distribution network fault repair plan intelligent generation device 800 to complete all or part of the steps in the above-mentioned low-voltage distribution network fault repair plan intelligent generation method. The memory 802 is used to store various types of data to support the operation of the low-voltage distribution network fault repair plan intelligent generation device 800; the multimedia component 803 may include a screen and an audio component; the I / O interface 804 provides an interface between the processor 801 and other interface modules; and the communication component 805 is used to conduct wired or wireless communication between the low-voltage distribution network fault repair plan intelligent generation device 800 and other devices, thereby implementing the low-voltage distribution network fault repair plan generation method proposed in the above-mentioned embodiment.

[0160] This embodiment further provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for generating a low-voltage distribution network fault emergency repair plan proposed in the above embodiment is implemented.

[0161] The storage medium proposed in this embodiment and the method for generating a low-voltage distribution network fault repair plan proposed in the above embodiment belong to the same inventive concept. For technical details not described in detail in this embodiment, please refer to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0162] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general hardware, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0163] Example 4 is an embodiment of the present invention, which provides a method for generating a low-voltage distribution network fault repair plan. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0164] Verify the performance of the intelligent generation method for low-voltage distribution network fault repair plan proposed in this application in the following aspects:

[0165] Anomaly detection accuracy: The ability of the isolation forest algorithm to detect anomalies in current, voltage, and power.

[0166] Fault classification accuracy: The classification accuracy of the CNN model for short circuit, grounding, and overload faults.

[0167] Resource allocation optimization: Cost comparison between the resource allocation scheme generated by genetic algorithm and traditional methods.

[0168] Path planning efficiency: Comparison of path length and time between the A* algorithm and the traditional Dijkstra algorithm.

[0169] Overall solution effectiveness: Optimize the entire process from fault occurrence to power restoration.

[0170] The experimental steps are as follows:

[0171] Step 1: Data preparation and simulation environment construction

[0172] Dataset:

[0173] Grid topology data: simulates an old community distribution network, including 10 ring main units, 15 branch boxes, and 50 user access points.

[0174] Electrical measurement data: 24-hour current, voltage, and power data were generated using Open DSS with a sampling frequency of 1 Hz. The following faults were injected:

[0175] Short circuit fault (20 times), ground fault (15 times), overload fault (25 times).

[0176] GIS road network data: Generate old community road networks based on Open Street Map, including road grades, real-time traffic status and closed road section information.

[0177] Resource Information:

[0178] Personnel (3 emergency repair teams), equipment (5 categories such as insulation testers, cable locators, etc.), vehicles (4 categories such as engineering vehicles, live working vehicles, etc.).

[0179] Step 2: Model training and parameter setting

[0180] Isolation forest model: number of trees ψ = 100, maximum depth log2(N), anomaly score threshold s th =0.7.

[0181] CNN model: Input: 20×20 spatiotemporal feature map of 3 channels (current, voltage, power).

[0182] Architecture: 2 convolutional layers (3×3 kernel) + 2 fully connected layers, ReLU activation function, Softmax output layer. Training: 80% data for training, 20% for validation, 50 iterations, learning rate 0.001.

[0183] Genetic algorithm: population size 100, crossover rate 0.8, mutation rate 0.05, and iteration 200 times.

[0184] A* algorithm: The heuristic function is Euclidean distance, and the real-time traffic weight α∈[1.0,2.0].

[0185] Step 3: Experimental run

[0186] Scenario 1: Single-point short circuit fault (Ring Main Unit No. 5).

[0187] Scenario 2: Multiple ground faults (branch boxes No. 3 and No. 7).

[0188] Scenario 3: Overload failure (area with dense user access points).

[0189] Step 4: Record performance indicators

[0190] Anomaly detection: precision, recall, F1 score;

[0191] Fault classification: confusion matrix, overall accuracy;

[0192] Resource allocation: cost (10,000 yuan), time (hours);

[0193] Path planning: path length (km), planning time (seconds).

[0194] The experimental results are shown in Tables 1 to 4:

[0195] Table 1 Anomaly detection performance comparison

[0196] method Accuracy Recall F1 score isolated forest 98.2％ 96.5％ 97.3％ Threshold method 85.4％ 72.1％ 78.1％

[0197] Table 2 Fault classification accuracy

[0198] Fault type Accuracy Short Circuit 95.6％ Grounding 92.3％ Overload 89.7％ overall 92.5％

[0199] Table 3 Comparison of resource allocation schemes

[0200] method Average cost (10,000 yuan) Average time (hours) Genetic Algorithm 3.2 1.5 Random assignment 5.1 2.8

[0201] Table 4 Comparison of path planning efficiency

[0202] method Average path length (km) Planning time (seconds) A* algorithm 4.7 0.8 Dijkstra 5.2 2.3

[0203] In summary, the simulation conclusions through experiments are as follows:

[0204] 1. Anomaly Detection: The Isolation Forest algorithm performs well in complex data from old neighborhoods, with an F1 score of 97.3%.

[0205] 2. Fault classification: The CNN model has the highest accuracy in classifying short-circuit faults (95.6%), while overload faults are slightly lower (89.7%).

[0206] 3. Resource allocation: Genetic algorithms reduce repair costs by 37% and time by 46%.

[0207] 4. Path planning: The A* algorithm reduces the path length by 9.6% compared to Dijkstra and the planning time is 65% faster.

[0208] 5. Overall efficiency: The entire process time from fault occurrence to power restoration is shortened from 4.2 hours with traditional methods to 2.1 hours.

[0209] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for generating a low-voltage distribution network fault repair plan, characterized in that: include: Collect grid operation data in real time and input it into the algorithm for anomaly detection, generating a first-level fault signal; Based on the primary fault signal and the collected historical data, secondary fault information is obtained through classification and modeling. The algorithm analyzes and processes the secondary fault information to obtain the tertiary fault information, and the fault repair resource allocation plan is obtained through algorithm processing; The fault repair resource allocation plan and the collected GIS data are processed through algorithms to obtain the repair path; Based on the repair path and fault impact range, the target repair plan is obtained through model optimization.

2. The method for generating a low-voltage distribution network fault repair plan according to claim 1, wherein: Collect grid operation data in real time and input it into the algorithm for anomaly detection, generating a first-level fault signal, including: The grid operation data is collected in real time through sensors and processed into a standardized data set; Based on the standardized data set, an isolation tree is constructed and the anomaly score of the data point is calculated after training; Threshold determination is performed on the anomaly score to obtain an anomaly point set, and a preliminary fault time window is obtained through cluster analysis; According to the preliminary fault time window and the line segment identification corresponding to the abnormal point, a first-level fault signal is obtained through correlation analysis, including the abnormal time point and abnormal line segment.

3. The method for generating a low-voltage distribution network fault repair plan according to claim 2, wherein: Based on the primary fault signal and the collected historical data, secondary fault information is obtained through classification and modeling, including: The first-level fault signal and the collected historical data are converted and features are extracted to obtain a classification model; After fault classification and topology modeling, secondary fault information is obtained, including fault type, candidate fault set and distribution network topology map.

4. The method for generating a low-voltage distribution network fault repair plan according to claim 3, wherein: The algorithm analyzes and processes the secondary fault information to obtain the tertiary fault information. include: Among them, the third-level fault information includes the fault impact range and key nodes; According to the secondary fault information, the minimum spanning tree is obtained after algorithm processing; Based on the minimum spanning tree and candidate fault points, the fault impact range is obtained through analysis and processing; According to the fault impact range and the distribution network topology, key nodes are obtained through extraction and processing.

5. The method for generating a low-voltage distribution network fault repair plan according to claim 4, wherein: The fault repair resource configuration plan obtained through algorithm processing includes: The emergency repair resource allocation problem is modeled as a multi-objective optimization problem, where the objectives include minimum emergency repair time and resource cost; Through algorithms, emergency repair time and resource costs are optimized simultaneously, resource allocation plans are found, and emergency repair resource allocation plans are dynamically adjusted.

6. The method for generating a low-voltage distribution network fault repair plan according to claim 5, wherein: The algorithm processes the fault repair resource allocation plan and the collected GIS data to obtain the repair path, including: The starting and ending points in the fault repair resource allocation plan, as well as the road network and edge weights in the GIS data, are input into the A* algorithm to calculate the total cost and select the minimum node for expansion until the end point is reached. Combine the road network and real-time traffic conditions in GIS data to generate the optimal repair path and output the specific route.

7. The method for generating a low-voltage distribution network fault repair plan according to claim 6, wherein: Based on the repair path and fault impact range, the target repair plan is obtained through model optimization, including: By building a multi-objective optimization model and considering the objective function, the final emergency repair plan is generated; Among them, the objective function includes emergency repair time, resource cost and power outage loss.

8. A system for generating a low-voltage distribution network fault repair plan, applying the method according to any one of claims 1 to 7, characterized in that: include: An acquisition module is used to collect real-time grid operation data and input it into the isolation forest algorithm for anomaly detection to generate a preliminary fault signal, wherein the grid operation data includes current, voltage and power, and the preliminary fault signal includes the abnormal time point and abnormal line section; The first processing module is used to obtain the fault type, candidate fault points and distribution network topology map based on the preliminary fault signal and the collected historical topology database through convolutional neural network fault classification and topology modeling; The second processing module is used to analyze and process the fault type, candidate fault points and distribution network topology using the minimum spanning tree algorithm to obtain the fault impact range and key nodes, and then obtain the low-voltage distribution network fault repair resource allocation plan through genetic algorithm processing; The third processing module is used to process the emergency repair resource allocation plan and the collected GIS data through the A* algorithm to obtain the optimal emergency repair path; The generation module is used to obtain the final low-voltage distribution network fault repair plan based on the optimal repair path and fault impact range through multi-objective optimization model processing, where the objective function is to minimize the repair time, resource cost and power outage loss.

9. An electronic device comprising: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method for generating a low-voltage distribution network fault repair plan as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the method for generating a low-voltage distribution network fault repair plan according to any one of claims 1 to 7.

Images