Medium-voltage power distribution network fault intelligent first-aid repair command system and method

Abstract

The invention provides a medium-voltage power distribution network fault intelligent first-aid repair command system and method. The system comprises a data acquisition module, a data analysis module, a decision scheduling module and a field execution module. The data acquisition module collects operation and topological data in real time; the analysis module locates a fault point through a fault probability prediction model and a Dijkstra algorithm; the decision scheduling module constructs a resource portrait library and formulates a scheduling scheme by using a multi-objective optimization algorithm; the field execution module carries out first-aid repair according to instructions and feeds back progress. According to the system and the method, faults can be accurately positioned, resource optimization scheduling is realized, and the first-aid repair efficiency is improved. Tests show that the fault positioning accuracy is high, the response time, the path cost and the resource waste rate are reduced, the average first-aid repair time is shortened, the user satisfaction degree is improved, and an efficient and intelligent solution is provided for medium-voltage power distribution network fault first-aid repair.

Description

Technical Field

[0001] This invention relates to the field of power system fault repair technology, and more specifically, to an intelligent fault repair command system and method for medium-voltage distribution networks. Background Technology

[0002] As a crucial link connecting high-voltage transmission networks and low-voltage users in the power system, the stable operation of medium-voltage distribution networks is essential for ensuring normal electricity supply for social production and residential life. With rapid economic development and continuously increasing electricity demand, the scale of medium-voltage distribution networks is expanding, and their structure is becoming more complex. Simultaneously, users are placing increasingly higher demands on power supply reliability and power quality. However, medium-voltage distribution networks are inevitably affected by various factors during operation, such as natural disasters, equipment aging, and external damage, leading to faults. Once a fault occurs, quickly and accurately locating the fault point, rationally allocating repair resources, and efficiently completing repair tasks become significant challenges for the power sector.

[0003] Current fault location methods mainly rely on traditional fault indicators and manual inspection. While fault indicators can roughly indicate the area where a fault occurs, their accuracy is poor for complex fault situations, such as intermittent or multi-point faults. Manual inspection requires a significant amount of time and manpower, and in some geographically complex or concealed areas, it is difficult to locate faults in a timely manner. In medium-voltage distribution networks in mountainous or forested areas, the complex terrain and dense vegetation make manual inspection difficult, potentially leading to faults remaining unaddressed for extended periods and affecting power supply reliability.

[0004] Traditional fault location methods often require manual, segment-by-segment line inspection, a tedious and inefficient process. When a fault occurs on a long line, the inspection area is wide, requiring a significant amount of time to locate the fault point. Furthermore, due to the lack of real-time monitoring data and effective analysis methods, fault location response times are long, failing to meet the demand for rapid power restoration. In medium-voltage distribution networks in bustling urban areas, a fault can cause prolonged power outages that severely impact commercial activities and residents' lives, but existing fault location technologies struggle to pinpoint the fault location quickly.

[0005] Currently, some power departments suffer from incomplete and untimely information management in emergency repair resources. The skill levels of repair teams, the status of material inventory, and the real-time location of vehicles are not accurately and promptly reflected in the dispatch system. This prevents optimal decision-making based on the actual situation when dispatching resources. Furthermore, inaccurate inventory information may lead to insufficient quantities or incorrect types of materials being allocated, impacting the progress of repairs.

[0006] Existing emergency repair resource scheduling schemes are mostly based on experience and simple rules, lacking scientific optimization algorithms. During the scheduling process, various factors are often not comprehensively considered, such as the urgency of the fault, the skill matching of the repair teams, and the route costs of material transportation. This makes the scheduling scheme potentially suboptimal, leading to long response times and wasted resources. Furthermore, when assigning repair teams, the distance between the teams and the fault location, as well as their skill levels, may not be taken into account, resulting in excessively long arrival times for repair personnel or an inability to effectively handle the fault.

[0007] During on-site emergency repairs, communication between repair personnel and the dispatch center was often delayed and untimely. Repair personnel could not provide real-time updates on the fault conditions and repair progress to the dispatch center, which in turn could not promptly understand the actual needs on-site and make corresponding adjustments. This could lead to untimely resource allocation and decision-making errors, impacting repair efficiency. When new fault points were discovered on-site or necessary materials were insufficient, the delayed communication prevented the dispatch center from allocating additional resources in a timely manner, prolonging the repair time.

[0008] There is a lack of effective supervision and evaluation mechanisms for on-site emergency repair work. It is impossible to comprehensively and accurately assess the work quality, efficiency, and resource utilization of repair personnel. This makes it difficult to summarize lessons learned, improve work methods, and enhance repair capabilities in subsequent repair work. In cases where repairs take too long or resources are wasted significantly, the lack of an evaluation mechanism prevents timely identification of problems and implementation of corrective measures.

[0009] Existing systems have limited capacity to process operational data from medium-voltage distribution networks, failing to fully uncover the potential information hidden within the data. A large amount of operational data is simply stored without effective analysis or utilization. Historical equipment fault data lacks in-depth statistical analysis, making it impossible to identify patterns and trends in fault occurrence and hindering the implementation of preventative measures.

[0010] Existing systems rely heavily on human experience and simple rules when making fault repair decisions, lacking intelligent decision support systems. They cannot base decisions on real-time operational data and fault conditions. Furthermore, in complex fault situations, human decision-making may be influenced by subjective factors, leading to inaccurate and untimely decisions. Therefore, this paper proposes an intelligent fault repair command system and method for medium-voltage distribution networks. Summary of the Invention

[0011] The purpose of this invention is to address the problems raised in the existing background technology. To achieve the above-mentioned objective, this invention provides the following technical solution: an intelligent emergency repair command system for medium-voltage distribution network faults, comprising a data acquisition module, a data analysis module, a decision-making and scheduling module, and a field execution module; The data acquisition module is used to collect real-time power distribution network operation data, including three-phase voltage. Three-phase current Active power and reactive power The sampling frequency is once per minute, i.e., the sampling period. ; The data analysis module receives data from the data acquisition module and outputs a device failure probability distribution map through a failure probability prediction model. ,in This represents the i-th power distribution equipment node. Indicates at time Lower device The probability value of a failure occurring; The failure probability prediction model uses the equipment's historical failure dataset. and the current running parameter vector , As input, a nonlinear mapping function is obtained by training using a machine learning algorithm: in For trained machine learning models (such as random forests, or Neural networks These are model parameters; The decision-making and scheduling module formulates a repair resource scheduling plan based on the fault probability distribution map output by the data analysis module and the confirmed fault events. The on-site execution module receives the dispatch instructions issued by the decision-making and dispatching module and executes the emergency repair task.

[0012] As a preferred embodiment of the present invention, the data acquisition module further includes a device spatial topology data acquisition unit based on a Geographic Information System (GIS), used to establish a device spatial topology database G=(V,E,W), where V is a set of device nodes. For the set of connecting edges, The edge weight represents the line length or impedance. pass The shortest path algorithm locates critical fault points on the fault propagation path. Its search range is limited to a radius centered on the alarm source. Within the circular area; The algorithm iteration process is as follows: in From the starting node to the node The shortest distance, for The set of adjacent nodes, For the edge The weights; The final three-dimensional geographic coordinate labels of the fault points And it is visualized through a GIS platform.

[0013] As a preferred technical solution of the present invention, the decision scheduling module constructs a multi-dimensional emergency repair resource profile database. It is used to record the following information: Repair team skill level These correspond to beginner (1), intermediate (2), and advanced (3), respectively. Material inventory status This represents the available quantity of resource type m at time t, updated hourly, satisfying the following: in To replenish stock, Consumption amount; Real-time vehicle location information It is refreshed every 30 seconds by the BeiDou system; All resource information is presented in vector form. This is used for subsequent scheduling and matching.

[0014] As a preferred embodiment of the present invention, the decision scheduling module establishes a dynamic priority evaluation model to optimize the priority sequence of emergency repair tasks. Sort in descending order; The overall priority score for each task T_k is calculated as a weighted sum: in: The severity level of the fault: If the power outage time And it affects the number of users. (Level 1 Emergency); like and ,but (Level 2 Emergency); like and ,but (Level 3 Emergency) To influence user level weights: Residents = 1, Commercial = 2, Industrial = 3, Important Users (Hospitals, Government, etc.) = 4; Key performance indicators for equipment are defined as follows: in To bear the load of the equipment, Its betweenness centrality in the network topology reflects its importance in the network; The coefficients satisfy the normalization condition: .

[0015] As a preferred embodiment of the present invention, the decision scheduling module employs a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. The objective function is defined as minimizing a weighted combination of the following three objectives: in: The first item is a penalty for failing to meet the response time target. Let k be the actual arrival time of the k-th task. The time of failure is required. Indicator function Determine if a timeout has occurred; The second item is the path cost. Let v be the total distance traveled by vehicle v. This represents the total travel time. This is the unit cost coefficient; Third item Resource waste rate is defined as: That is, the proportion of unused resources to the total allocated resources shall not exceed 10%; The weighting coefficients satisfy ,and ; The multi-objective optimization problem is solved by using NSGA-II, MOEA / D or other evolutionary algorithms to obtain the Pareto optimal solution set, and the scheduling scheme with the highest satisfaction is selected and issued.

[0016] A method for intelligent emergency repair command of medium-voltage distribution network faults includes the following steps: Step 1: Collect power distribution network operation data, including voltage. Current Active power reactive power The data is collected once per minute; the data is then input into the trained fault probability prediction model. Output device failure probability distribution diagram; Step 2: Based on Geographic Information System Establish a device space topology database ,use The algorithm in the search radius Internal location fault point and mark its three-dimensional geographic coordinates. ; Step 3: Construct a multi-dimensional emergency repair resource profile database Record the skill levels of each emergency repair team Inventory status of materials Updated hourly, and vehicle location in real time. ; Step 4: Establish a dynamic priority evaluation model and calculate a comprehensive priority score for each fault task T_k: Optimize task priority sequence ; Step 5: Use a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. The objective function is: Constraints include: response time resource waste rate Finally, the optimal scheduling instruction is optimized and sent to the field execution module.

[0017] As a preferred technical solution of the present invention, when collecting power distribution network operation data, the data is also preprocessed, including data cleaning and data normalization operations. Data cleaning removes outliers and missing values, and data normalization scales the data to the range of 0 to 1.

[0018] As a preferred embodiment of the present invention, the fault probability prediction model is trained using a neural network algorithm, specifically a feedforward neural network. Convolutional Neural Networks or Long Short-Term Memory Network ; Let the input feature vector be The real label is (Whether a fault has occurred), the model output is the predicted probability. ,in Activation function; The training process uses the binary cross-entropy loss function as the target: The model parameters θ are updated using gradient descent, with a learning rate set to θ = 0.01, and a total of T = 1000 training iterations. The AUC value on the validation set after each iteration is used to determine early stopping.

[0019] As a preferred technical solution of the present invention, after the optimal scheduling instruction is issued, the field execution module provides real-time feedback on the repair progress information via a mobile terminal device, with a feedback cycle of once every 15 minutes. ; Each feedback includes the following state vector: in: The vehicle's current location; Indicates the task completion status (0 = not started, 1 = completed); The estimated remaining repair time; The system dynamically adjusts subsequent scheduling strategies based on feedback, constructing a closed-loop control mechanism: in For the amount of scheduling instruction updates, This is an adaptive rescheduling function.

[0020] As a preferred technical solution of the present invention, after the emergency repair task is completed, a comprehensive evaluation of the emergency repair task is carried out, and the evaluation indicators include emergency repair time, resource utilization rate and user satisfaction. The definitions of each indicator are as follows: Actual repair time : in For power restoration time, The time of the fault occurrence; resource utilization rate : in The actual quantity of materials used. Allocate quantities for scheduling; User satisfaction The survey was conducted via telephone follow-up, with a follow-up sample rate of no less than 30% of the affected users. Satisfaction rating was conducted using a five-point Likert scale (1-5 points), and the final score was the average. All evaluation results are stored in the historical database. This is used for subsequent model optimization and performance analysis.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: The data acquisition module of this invention collects voltage, current, and power operating parameters at a frequency of once per minute. This high-frequency data acquisition can capture subtle changes in the operating status of the distribution network. At the moment a fault occurs, the system can promptly obtain the parameter fluctuations before and after the fault, providing an accurate data foundation for subsequent fault analysis. Compared with traditional systems with lower acquisition frequencies, it can detect fault signs earlier, significantly shortening the fault detection time.

[0022] Fault location is achieved using a Geographic Information System (GIS) spatial topology data acquisition unit combined with Dijkstra's algorithm. This algorithm, with a search radius of 5 kilometers, can quickly and accurately determine the three-dimensional geographic coordinates of fault points in complex power distribution network topologies. Compared to traditional fault diagnosis methods, it eliminates the need for manual segment-by-segment line inspection, saving significant time and labor costs.

[0023] This invention's decision-making and scheduling module constructs a multi-dimensional emergency repair resource profile database, which records in detail the skill level of emergency repair teams, material inventory status, and real-time vehicle location information. When formulating a scheduling plan, the system can accurately match the most suitable emergency repair team based on the type and complexity of the fault. For complex electrical equipment faults, the system will prioritize scheduling highly skilled repair teams to ensure that the fault can be handled professionally and efficiently. Simultaneously, based on the material inventory status and the location of the fault, the system selects the nearest warehouse with sufficient inventory to allocate materials, reducing material transportation time and costs.

[0024] This invention employs a multi-objective optimization algorithm to formulate a resource scheduling plan for emergency repairs, aiming to minimize response time, path cost, and resource waste rate. This algorithm comprehensively considers various factors; when scheduling the routes for repair vehicles, it avoids traffic congestion and selects the shortest and most convenient routes, enabling repair personnel to reach the fault site as quickly as possible. Simultaneously, it rationally allocates resources, avoiding excessive stockpiling and waste, and improving resource utilization efficiency.

[0025] In this invention, on-site repair personnel use mobile terminals to report repair progress to the dispatch center every 15 minutes. The dispatch center can then monitor and dynamically adjust the entire repair process based on this real-time feedback. If unforeseen circumstances arise during repairs, such as the discovery of new fault points or a shortage of necessary materials, the dispatch center can promptly allocate additional resources to the site to ensure the smooth progress of the repair work. This real-time communication and dynamic adjustment mechanism significantly improves repair efficiency and reduces repair time.

[0026] After a repair task is completed, the system summarizes and evaluates the entire repair process, calculating repair time, resource utilization, and user satisfaction indicators. By analyzing these indicators, lessons learned are summarized, and corresponding improvement measures are formulated. If frequent failures are found in a certain area, the system can analyze the causes and take preventative measures in advance, such as strengthening equipment maintenance and upgrading equipment. Simultaneously, these experiences and improvement measures are incorporated into the system's knowledge base, providing a reference for future fault repairs and continuously improving the system's performance and fault handling capabilities.

[0027] This invention significantly shortens power outage time following medium-voltage distribution network faults by enabling rapid fault location, optimized emergency repair resource scheduling, and improved repair efficiency. For users, reduced outage time means minimal impact on production and daily life. In commercial areas, power outages can lead to business losses, but this system can quickly restore power, minimizing losses and increasing user satisfaction. During repairs, the system communicates with the dispatch center in real-time via mobile terminal devices, providing timely feedback to users on fault handling progress and estimated power restoration time. Users can access this information via mobile app or SMS, reducing anxiety and unease. Furthermore, after fault handling is completed, user feedback and suggestions are collected through telephone follow-ups to further improve service quality and enhance user experience and satisfaction.

[0028] This invention's precise fault location and efficient resource scheduling avoid unnecessary investment of manpower and resources. Traditional fault repair methods may require a large number of personnel for blind inspections and material stockpiling, while this system can accurately allocate resources according to actual needs, reducing waste of manpower and materials. In terms of material allocation, the system can allocate resources according to the fault type and the exact quantity of materials required, avoiding excessive stockpiling and idle materials, and reducing inventory costs.

[0029] This invention reduces the frequency of equipment damage and failures by monitoring the operating status of the power distribution network in real time and handling faults promptly. Equipment operating in good condition can have its service life extended, reducing replacement and maintenance costs. For critical equipment like transformers, timely detection and handling of potential faults can prevent damage, extend their service life, and thus lower maintenance costs. Attached image description: Figure 1 The system logic block diagram provided for this invention; Figure 2 A schematic diagram of the data parameters provided for this invention; Figure 3 This is a schematic diagram of the method flow provided by the present invention; Figure 4This is a schematic diagram of the method data parameters provided by the present invention; Figure 5 A schematic diagram of the grade data parameters provided by this invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are specific implementations of the present invention and are not limited to all embodiments.

[0031] Therefore, the following detailed description of embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely illustrates some embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0032] It should be noted that, in the absence of conflict, the embodiments and features and technical solutions in the embodiments of the present invention can be combined with each other. It should be noted that similar reference numerals and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0033] Example 1: An intelligent emergency repair command system for medium-voltage distribution network faults, comprising a data acquisition module, a data analysis module, a decision-making and scheduling module, and a field execution module; The data acquisition module is used to collect real-time operating data of the power distribution network, including three-phase voltage. Three-phase current Active power and reactive power The sampling frequency is once per minute, i.e., the sampling period. ; The data analysis module receives data from the data acquisition module and outputs a fault probability distribution map of the equipment through a fault probability prediction model. ,in This represents the i-th power distribution equipment node. Indicates at time Lower device The probability value of a failure occurring; Failure probability prediction models use historical failure datasets of the equipment. and the current running parameter vector , As input, a nonlinear mapping function is obtained by training using a machine learning algorithm: in For trained machine learning models (such as random forests, or Neural networks These are model parameters; The decision-making and scheduling module formulates a repair resource scheduling plan based on the fault probability distribution map output by the data analysis module and the confirmed fault events. The on-site execution module receives the dispatch instructions issued by the decision-making and dispatching module and executes the emergency repair task.

[0034] The data acquisition module also includes a device spatial topology data acquisition unit based on a Geographic Information System (GIS), used to establish a device spatial topology database G=(V,E,W), where V is the set of device nodes. For the set of connecting edges, The edge weight represents the line length or impedance. pass The shortest path algorithm locates critical fault points on the fault propagation path. Its search range is limited to a radius centered on the alarm source. Within the circular area; The algorithm iteration process is as follows: in From the starting node to the node The shortest distance, for The set of adjacent nodes, For the edge The weights; The final three-dimensional geographic coordinate labels of the fault points And it is visualized through a GIS platform.

[0035] The decision-making and scheduling module constructs a multi-dimensional emergency repair resource profile database. It is used to record the following information: Repair team skill level These correspond to beginner (1), intermediate (2), and advanced (3), respectively. Material inventory status This represents the available quantity of resource type m at time t, updated hourly, satisfying the following: in To replenish stock, Consumption amount; Real-time vehicle location information It is refreshed every 30 seconds by the BeiDou system; All resource information is presented in vector form. This is used for subsequent scheduling and matching.

[0036] The decision-making and scheduling module establishes a dynamic priority evaluation model to optimize the priority sequence of emergency repair tasks. Sort in descending order; The overall priority score for each task T_k is calculated as a weighted sum: in: The severity level of the fault: If the power outage time And it affects the number of users. (Level 1 Emergency); like and ,but (Level 2 Emergency); like and ,but (Level 3 Emergency) To influence user level weights: Residents = 1, Commercial = 2, Industrial = 3, Important Users (Hospitals, Government, etc.) = 4; Key performance indicators for equipment are defined as follows: in To bear the load of the equipment, Its betweenness centrality in the network topology reflects its importance in the network; The coefficients satisfy the normalization condition: .

[0037] The decision-making and scheduling module employs a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. The objective function is defined as minimizing the weighted combination of the following three objectives: in: The first item is a penalty for failing to meet the response time target. Let k be the actual arrival time of the k-th task. The time of failure is required. Indicator function Determine if a timeout has occurred; The second item is the path cost. Let v be the total distance traveled by vehicle v. This represents the total travel time. This is the unit cost coefficient; Third item Resource waste rate is defined as: That is, the proportion of unused resources to the total allocated resources shall not exceed 10%; The weighting coefficients satisfy ,and ; The multi-objective optimization problem is solved by using NSGA-II, MOEA / D or other evolutionary algorithms to obtain the Pareto optimal solution set, and the scheduling scheme with the highest satisfaction is selected and issued.

[0038] Example 2: An intelligent emergency repair command system for medium-voltage distribution network faults includes a data acquisition module, a data analysis module, a decision-making and scheduling module, and a field execution module. The data acquisition module is used to collect distribution network operation data in real time, including voltage, current, and power parameters, with a collection frequency of once per minute. The data analysis module receives the data from the data acquisition module and outputs a fault probability distribution map of the equipment through a fault probability prediction model. This model is trained using historical fault data and real-time operating parameters of the equipment as input and employs a machine learning algorithm. The decision-making and scheduling module formulates an emergency repair resource scheduling plan based on the results of the data analysis module. The field execution module receives instructions from the decision-making and scheduling module and executes the emergency repair tasks.

[0039] The data acquisition module also includes a device spatial topology data acquisition unit based on Geographic Information System (GIS), which is used to establish a device spatial topology database, locate fault points and mark three-dimensional geographic coordinate labels using the Dijkstra algorithm, where the search radius of the Dijkstra algorithm is set to 5 kilometers.

[0040] The decision-making and scheduling module constructs a multi-dimensional emergency repair resource profile database, recording the skill level of the emergency repair team, the status of material inventory, and the real-time location information of vehicles. The skill level of the emergency repair team is divided into three levels: primary, intermediate, and advanced. The status of material inventory is updated in real time, with an update cycle of once per hour.

[0041] The decision-making and scheduling module establishes a dynamic priority evaluation model, which prioritizes emergency repair tasks based on the urgency of the fault, the number of affected users, and the criticality of the equipment. The urgency of the fault is divided into three levels according to the power outage time and the scope of the power outage: Level 1 is a power outage with a power outage time of more than 2 hours and a power outage scope of more than 10 users; Level 2 is a power outage with a power outage time of 1 to 2 hours and a power outage scope of 5 to 10 users; and Level 3 is a power outage with a power outage time of less than 1 hour and a power outage scope of less than 5 users.

[0042] The decision-making and scheduling module uses a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. The objective function is set to minimize response time, path cost and resource waste rate. The response time objective is set to arrive at the site within 1 hour after the fault occurs, the path cost is calculated based on vehicle mileage and time, and the resource waste rate objective is to control it within 10%.

[0043] A method for intelligent emergency repair command of medium-voltage distribution network faults includes the following steps: Step 1: Collect power distribution network operation data, including voltage, current, and power parameters, at a frequency of once per minute; output the equipment fault probability distribution map through the fault probability prediction model. This model is trained using machine learning algorithms with historical fault data and real-time operating parameters of the equipment as input. Step 2: Establish a spatial topology database for the equipment based on a Geographic Information System (GIS), locate the fault point using the Dijkstra algorithm and mark it with three-dimensional geographic coordinate labels, with the search radius set to 5 kilometers; Step 3: Construct a multi-dimensional emergency repair resource profile database, recording the skill level of the emergency repair team, the status of material inventory, and the real-time location information of vehicles. The skill level of the emergency repair team is divided into three levels: primary, intermediate, and advanced. The status of material inventory is updated in real time, with an update cycle of once per hour. Step 4: Establish a dynamic priority assessment model. Based on the urgency of the fault, the number of users affected, and the priority sequence of critical equipment repair tasks, the urgency of the fault is divided into three levels according to the power outage time and the scope of the power outage. Level 1 is a power outage with a power outage time of more than 2 hours and a power outage scope of more than 10 users. Level 2 is a power outage with a power outage time of 1 to 2 hours and a power outage scope of 5 to 10 users. Level 3 is a power outage with a power outage time of less than 1 hour and a power outage scope of less than 5 users. Step 5: Use a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. Set the objective function in the direction of minimizing response time, path cost and resource waste rate. The response time objective is to arrive at the site within 1 hour after the fault occurs. The path cost is calculated based on the vehicle's mileage and time. The resource waste rate objective is to control it within 10%. Issue the optimal scheduling instruction and continuously monitor the task execution status.

[0044] When collecting power distribution network operation data, the data is also preprocessed, including data cleaning and data normalization. Data cleaning removes outliers and missing values, and data normalization scales the data to the range of 0 to 1.

[0045] The failure probability prediction model was trained using a neural network algorithm with 1000 iterations and a learning rate of 0.01.

[0046] After the optimal scheduling command is issued, the on-site execution module provides real-time feedback on the repair progress via mobile terminal devices, with a feedback cycle of once every 15 minutes.

[0047] After the emergency repair task is completed, the task will be evaluated. The evaluation indicators include repair time, resource utilization rate, and user satisfaction. User satisfaction will be surveyed by telephone follow-up, with a follow-up rate of no less than 30%.

[0048] A method for intelligent emergency repair command of medium-voltage distribution network faults includes the following steps: Step 1: Collect power distribution network operation data, including voltage. Current Active power reactive power The data is collected once per minute; the data is then input into the trained fault probability prediction model. Output device failure probability distribution diagram; Step 2: Based on Geographic Information System Establish a device space topology database ,use The algorithm in the search radius Internal location fault point and mark its three-dimensional geographic coordinates. ; Step 3: Construct a multi-dimensional emergency repair resource profile database Record the skill levels of each emergency repair team Inventory status of materials Updated hourly, and vehicle location in real time. ; Step 4: Establish a dynamic priority evaluation model and calculate a comprehensive priority score for each fault task T_k: Optimize task priority sequence ; Step 5: Use a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. The objective function is: Constraints include: response time resource waste rate Finally, the optimal scheduling instruction is optimized and sent to the field execution module.

[0049] When collecting power distribution network operation data, the data is also preprocessed, including data cleaning and data normalization. Data cleaning removes outliers and missing values, and data normalization scales the data to the range of 0 to 1.

[0050] The failure probability prediction model is trained using a neural network algorithm, specifically a feedforward neural network. Convolutional Neural Networks or Long Short-Term Memory Network ; Let the input feature vector be The real label is (Whether a fault has occurred), the model output is the predicted probability. ,in Activation function; The training process uses the binary cross-entropy loss function as the target: The model parameters θ are updated using gradient descent, with a learning rate set to θ = 0.01, and a total of T = 1000 training iterations. The AUC value on the validation set after each iteration is used to determine early stopping.

[0051] After issuing the optimal scheduling command, the field execution module provides real-time feedback on the repair progress via mobile terminal devices, with a feedback cycle of once every 15 minutes. ; Each feedback includes the following state vector: in: The vehicle's current location; Indicates the task completion status (0 = not started, 1 = completed); The estimated remaining repair time; The system dynamically adjusts subsequent scheduling strategies based on feedback, constructing a closed-loop control mechanism: in For the amount of scheduling instruction updates, This is an adaptive rescheduling function.

[0052] After the emergency repair task is completed, a comprehensive evaluation will be conducted, and the evaluation indicators will include repair time, resource utilization rate and user satisfaction. The definitions of each indicator are as follows: Actual repair time : in For power restoration time, The time of the fault occurrence; resource utilization rate : in The actual quantity of materials used. Allocate quantities for scheduling; User satisfaction The survey was conducted via telephone follow-up, with a follow-up sample rate of no less than 30% of the affected users. Satisfaction rating was conducted using a five-point Likert scale (1-5 points), and the final score was the average. All evaluation results are stored in the historical database. This is used for subsequent model optimization and performance analysis.

[0053] This intelligent emergency repair command system for medium-voltage distribution networks mainly consists of a data acquisition module, a data analysis module, a decision-making and scheduling module, and a field execution module. The data acquisition module is responsible for collecting real-time operational data of the distribution network, including key parameters such as voltage, current, and power, with a collection frequency set to once per minute. This module achieves data acquisition by installing numerous sensors at various key nodes in the medium-voltage distribution network. Voltage and current sensors are installed at the outgoing lines of substations to monitor the voltage and current values ​​of the lines in real time; power sensors are installed at important load nodes to obtain the power consumption of those nodes. These sensors convert the collected analog signals into digital signals and transmit them to the data acquisition center via wired or wireless communication networks.

[0054] The equipment spatial topology data acquisition unit, based on a Geographic Information System (GIS), is a crucial component of the data acquisition module. It establishes an equipment spatial topology database by collecting and processing information on the geographical locations and connections of various devices in the power distribution network. Specifically, it uses high-precision GPS positioning technology to assign precise geographical coordinates to each device. Simultaneously, by analyzing the electrical connections between devices, it constructs the topology of the power distribution network. When locating fault points, it employs Dijkstra's algorithm, using the fault location as the center and setting a search radius of 5 kilometers, to quickly and accurately determine the three-dimensional geographical coordinates of the faulty equipment and label it accordingly, providing precise location information for subsequent repair work.

[0055] After receiving data from the data acquisition module, the data analysis module first analyzes the failure probability of the equipment using a failure probability prediction model. This model takes historical failure data and real-time operating parameters of the equipment as input and is trained using a machine learning algorithm. Using a neural network algorithm, the failure type, failure time, and equipment operating status information from the historical failure data are used as training samples to train the network. The training iterations are 1000 times, and the learning rate is set to 0.01. By continuously adjusting the network weights and biases, the model can accurately predict the failure probability of the equipment. Finally, the model outputs a failure probability distribution map, intuitively showing the likelihood of failure for each piece of equipment in the distribution network.

[0056] Based on a Geographic Information System (GIS) spatial topology database of equipment, the data analysis module can further pinpoint the location of faults. Utilizing Dijkstra's algorithm, combined with the spatial topology relationships of the equipment and real-time operational data, it rapidly searches for faulty equipment within a 5-kilometer search radius. The algorithm determines the location of the fault by calculating the shortest path from the fault location to each piece of equipment and labels it with three-dimensional geographic coordinates, providing accurate location information for subsequent emergency repair scheduling.

[0057] The decision-making and scheduling module employs a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. This algorithm aims to minimize response time, path cost, and resource waste rate, comprehensively considering factors such as the skill level of the repair teams, material inventory status, real-time vehicle location, and the location of the fault point. When assigning repair teams, it prioritizes teams closest to the fault point with matching skill levels; when allocating materials, it selects the nearest warehouse with sufficient inventory based on the fault type and the required type and quantity of materials. The equipment spatial topology data acquisition unit based on a Geographic Information System (GIS) then comes into play. It utilizes high-precision GPS positioning technology to assign accurate geographical coordinates to every piece of equipment in the distribution network. For utility poles and transformers, it accurately records their latitude and longitude information.

[0058] Simultaneously, this unit conducts in-depth analysis of the electrical connections between devices to construct the power distribution network topology. During this process, it meticulously records the connection methods and sequences between devices, forming a complete spatial topology database. When the system needs to locate a fault point, it activates the Dijkstra algorithm. Using the fault location as the center and setting a search radius of 5 kilometers, the algorithm rapidly searches within this range, calculating the shortest paths from the fault location to each device, determining the three-dimensional geographic coordinates of the fault point, and labeling it accordingly.

[0059] After receiving data from the acquisition module, the data acquisition center first preprocesses the data. This step includes data cleaning and data normalization. During data cleaning, the system automatically identifies and removes outliers and missing values. If a voltage value collected by a sensor significantly exceeds the normal range, the system will identify it as an outlier and remove it; if a data point is missing, the system will perform reasonable interpolation based on the preceding and following data to fill in the gaps.

[0060] Data normalization scales all data to the range of 0-1. This eliminates dimensional differences between different data points, making subsequent analysis more accurate and effective.

[0061] The preprocessed data is then input into the fault probability prediction model. This model uses historical fault data and real-time operating parameters of the equipment as input and is trained using a machine learning algorithm; here, a neural network algorithm is used as an example. During training, fault type, fault time, and equipment operating status information from the historical fault data are used as training samples to repeatedly train the network. The number of iterations is set to 1000, and the learning rate is set to 0.01.

[0062] As training progresses, the model continuously adjusts its weights and biases, gradually learning the intrinsic relationship between equipment failures and operating parameters. Ultimately, the model outputs a probability distribution map of equipment failures, visually demonstrating the likelihood of failure for each piece of equipment in the power distribution network.

[0063] Based on a Geographic Information System (GIS) spatial topology database of equipment, the system further refines the location of the fault. When a fault is detected, the system immediately activates Dijkstra's algorithm. This algorithm rapidly searches for the faulty equipment within a 5-kilometer search radius using information from the equipment spatial topology database. By calculating the shortest path from the fault location to each piece of equipment, the algorithm accurately determines the fault's location and labels it with three-dimensional geographic coordinates. This label is then sent to the decision-making and scheduling module, providing accurate location information for subsequent emergency repair scheduling.

[0064] The decision-making and dispatching module first constructs a multi-dimensional emergency repair resource profile database. This database records the skill level of the emergency repair team, the status of material inventory, and the real-time location information of vehicles. For emergency repair teams, their skill levels are divided into three levels: basic, intermediate, and advanced, based on the professional skills and work experience of the repair personnel. The status of material inventory is updated in real time through sensors installed in the material warehouse and the inventory management system, with an update cycle of once per hour. The system monitors the quantity, type, and storage location of materials in real time to ensure that materials can be quickly allocated when needed. The real-time location of vehicles is tracked through a GPS positioning system, and dispatchers can view the vehicle's dynamics in the system at any time.

[0065] The urgency of the fault is categorized into three levels based on the outage time and affected area: Level 1 is an emergency if the outage lasts more than 2 hours and affects more than 10 users; Level 2 is an emergency if the outage lasts 1-2 hours and affects 5-10 users; and Level 3 is an emergency if the outage lasts less than 1 hour and affects fewer than 5 users. The model also considers the criticality of the faulty equipment; for example, if the faulty equipment is a critical transformer in the substation, its priority will be relatively high. Through a comprehensive evaluation of these factors, the model can quickly and accurately determine the priority of each repair task.

[0066] The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described in the present invention. Although the present invention has been described in detail with reference to the above embodiments, the present invention is not limited to the specific embodiments described above. Therefore, any modifications or substitutions to the present invention, and all technical solutions and improvements that do not depart from the spirit and scope of the invention, are covered within the scope of the claims of the present invention.

Claims

1. A smart emergency repair command system for medium-voltage distribution network faults, characterized in that, It includes a data acquisition module, a data analysis module, a decision-making and scheduling module, and an on-site execution module; The data acquisition module is used to collect real-time power distribution network operation data, including three-phase voltage. Three-phase current Active power and reactive power The sampling frequency is once per minute, i.e., the sampling period. ; The data analysis module receives data from the data acquisition module and outputs a device failure probability distribution map through a failure probability prediction model. ,in This represents the i-th power distribution equipment node. Indicates at time Lower device The probability value of a failure occurring; The failure probability prediction model uses the equipment's historical failure dataset. and the current running parameter vector , As input, a nonlinear mapping function is obtained by training using a machine learning algorithm: in For trained machine learning models (such as random forests, or Neural networks These are model parameters; The decision-making and scheduling module formulates a repair resource scheduling plan based on the fault probability distribution map output by the data analysis module and the confirmed fault events. The on-site execution module receives the dispatch instructions issued by the decision-making and dispatching module and executes the emergency repair task.

2. The intelligent emergency repair command system for medium-voltage distribution network faults according to claim 1, characterized in that, The data acquisition module also includes a device spatial topology data acquisition unit based on a Geographic Information System (GIS), used to establish a device spatial topology database G=(V,E,W), where V is the set of device nodes. For the set of connecting edges, The edge weight represents the line length or impedance. pass The shortest path algorithm locates critical fault points on the fault propagation path. Its search range is limited to a radius centered on the alarm source. Within the circular area; The algorithm iteration process is as follows: in From the starting node to the node The shortest distance, for The set of adjacent nodes, For the edge The weights; The final three-dimensional geographic coordinate labels of the fault points And it is visualized through a GIS platform.

3. The intelligent emergency repair command system for medium-voltage distribution network faults according to claim 1, characterized in that, The decision-making and scheduling module constructs a multi-dimensional emergency repair resource profile database. It is used to record the following information: Repair team skill level These correspond to beginner (1), intermediate (2), and advanced (3), respectively. Material inventory status This represents the available quantity of resource type m at time t, updated hourly, satisfying the following: in To replenish stock, Consumption amount; Real-time vehicle location information It is refreshed every 30 seconds by the BeiDou system; All resource information is presented in vector form. This is used for subsequent scheduling and matching.

4. The intelligent emergency repair command system for medium-voltage distribution network faults according to claim 1, characterized in that, The decision-making and scheduling module establishes a dynamic priority evaluation model to optimize the priority sequence of emergency repair tasks. Sort in descending order; The overall priority score for each task T_k is calculated as a weighted sum: in: The severity level of the fault: If the power outage time And it affects the number of users. (Level 1 Emergency); like and ,but (Level 2 Emergency); like and ,but (Level 3 Emergency) To influence user level weights: Residents = 1, Commercial = 2, Industrial = 3, Important Users (Hospitals, Government, etc.) = 4; Key performance indicators for equipment are defined as follows: in To bear the load of the equipment, Its betweenness centrality in the network topology reflects its importance in the network; The coefficients satisfy the normalization condition: .

5. The intelligent emergency repair command system for medium-voltage distribution network faults according to claim 1, characterized in that, The decision-making and scheduling module employs a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. The objective function is defined as minimizing a weighted combination of the following three objectives: in: The first item is a penalty for failing to meet the response time target. Let k be the actual arrival time of the k-th task. The time of failure is required. Indicator function Determine if a timeout has occurred; The second item is the path cost. Let v be the total distance traveled by vehicle v. This represents the total travel time. This is the unit cost coefficient; Third item Resource waste rate is defined as: That is, the proportion of unused resources to the total allocated resources shall not exceed 10%; The weighting coefficients satisfy ,and ; The multi-objective optimization problem is solved by using NSGA-II, MOEA / D or other evolutionary algorithms to obtain the Pareto optimal solution set, and the scheduling scheme with the highest satisfaction is selected and issued.

6. A method for intelligent emergency repair command of medium-voltage distribution network faults, characterized in that, Includes the following steps: Step 1: Collect power distribution network operation data, including voltage. Current Active power reactive power The data is collected once per minute; the data is then input into the trained fault probability prediction model. Output device failure probability distribution diagram; Step 2: Based on Geographic Information System Establish a device space topology database ,use The algorithm in the search radius Internal location fault point and mark its three-dimensional geographic coordinates. ; Step 3: Construct a multi-dimensional emergency repair resource profile database Record the skill levels of each emergency repair team Inventory status of materials Updated hourly, and vehicle location in real time. ; Step 4: Establish a dynamic priority evaluation model and calculate a comprehensive priority score for each fault task T_k: Optimize task priority sequence ; Step 5: Use a multi-objective optimization algorithm to solve the emergency repair resource scheduling scheme. The objective function is: Constraints include: response time resource waste rate Finally, the optimal scheduling instruction is optimized and sent to the field execution module.

7. The intelligent emergency repair command method for medium-voltage distribution network faults according to claim 6, characterized in that, When collecting power distribution network operation data, the data is also preprocessed, including data cleaning and data normalization. Data cleaning removes outliers and missing values, and data normalization scales the data to the range of 0 to 1.

8. The intelligent emergency repair command method for medium-voltage distribution network faults according to claim 6, characterized in that, The fault probability prediction model is trained using a neural network algorithm, specifically a feedforward neural network. Convolutional Neural Networks or Long Short-Term Memory Network ; Let the input feature vector be The real label is (Whether a fault has occurred), the model output is the predicted probability. ,in Activation function; The training process uses the binary cross-entropy loss function as the target: The model parameters θ are updated using gradient descent, with a learning rate set to θ = 0.01, and a total of T = 1000 training iterations. The AUC value on the validation set after each iteration is used to determine early stopping.

9. The intelligent emergency repair command method for medium-voltage distribution network faults according to claim 6, characterized in that, After issuing the optimal scheduling command, the field execution module provides real-time feedback on the repair progress via mobile terminal devices, with a feedback cycle of once every 15 minutes. ; Each feedback includes the following state vector: in: The vehicle's current location; Indicates the task completion status (0 = not started, 1 = completed); The estimated remaining repair time; The system dynamically adjusts subsequent scheduling strategies based on feedback, constructing a closed-loop control mechanism: in For the amount of scheduling instruction updates, This is an adaptive rescheduling function.

10. The intelligent emergency repair command method for medium-voltage distribution network faults according to claim 6, characterized in that, After the emergency repair task is completed, a comprehensive evaluation will be conducted, and the evaluation indicators will include repair time, resource utilization rate and user satisfaction. The definitions of each indicator are as follows: Actual repair time : in For power restoration time, The time of the fault occurrence; resource utilization rate : in The actual quantity of materials used. Allocate quantities for scheduling; User satisfaction The survey was conducted via telephone follow-up, with a follow-up sample rate of no less than 30% of the affected users. Satisfaction rating was conducted using a five-point Likert scale (1-5 points), and the final score was the average. All evaluation results are stored in the historical database. This is used for subsequent model optimization and performance analysis.

Images