Distributed power equipment collaborative maintenance path planning method and system

Abstract

The application provides a distributed power equipment cooperative maintenance path planning method and system, relates to the technical field of power equipment maintenance, constructs a power fault node perception network and performs maintenance guidance feature analysis, constructs a multi-dimensional maintenance guidance vector to guide a candidate maintenance robot network to perform cooperative maintenance path planning, obtains a first space of maintenance path planning, performs loss prediction optimization according to the first space, obtains a second space of maintenance path planning, performs maintenance path optimization maximum optimization on the second space, outputs a maintenance path optimization result, and controls the candidate maintenance robot network to perform maintenance. The application solves the technical problems that the prior art relies on static path planning or manual scheduling, lacks a cooperative scheduling mechanism based on global optimization, and causes maintenance path planning response lag and unreasonable resource allocation, and achieves the technical effects of improving power equipment maintenance efficiency and resource utilization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power equipment maintenance, in particular to a distributed power equipment collaborative maintenance path planning method and system. BACKGROUND

[0002] With the continuous development of new power systems, emerging technologies such as distributed power generation and smart grids are widely used, and the structure of the power grid tends to be complex, with equipment becoming increasingly widespread and dispersed. Under this background, intelligent maintenance of distributed power equipment has become a key link to ensure the safe and stable operation of the power grid. Existing distributed power equipment maintenance path planning methods mostly use static path planning strategies or rely on manual scheduling rules for task allocation and path planning. Such methods can still be applied in small-scale, relatively stable equipment networks, but when faced with large-scale distributed power equipment sets and sudden, multi-point fault situations, the overall task distribution, robot capability matching, equipment priority, and other multi-factor coupling relationships are ignored, resulting in suboptimal path planning results at the global level, which can cause path failure or task delays, significantly reducing the response speed of maintenance, and easily causing problems such as repeated work, path conflicts, and insufficient work coverage for multiple robots, which affects the rationality of resource allocation and is not conducive to reducing system energy consumption and work time costs. SUMMARY

[0003] The present application provides a distributed power equipment collaborative maintenance path planning method and system, which solves the technical problems of existing technologies that rely on static path planning or manual scheduling, lack of collaborative scheduling mechanisms based on global optimization, and result in delayed maintenance path planning response and unreasonable resource allocation, achieving the technical effect of improving power equipment maintenance efficiency and resource utilization.

[0004] In view of the above problems, on the one hand, the present application provides a distributed power equipment collaborative maintenance path planning method, which comprises: constructing a distributed power equipment network according to a set of distributed power equipment of a power grid; performing fault perception according to the distributed power equipment network, establishing a power fault node perception network, and performing maintenance guidance feature analysis on the power fault node perception network to construct a multi-dimensional maintenance guidance vector; performing multi-dimensional perception analysis on a maintenance robot library to establish a candidate maintenance robot network; guiding the candidate maintenance robot network to perform collaborative maintenance path planning on the power fault node perception network according to the multi-dimensional maintenance guidance vector to obtain a first space of maintenance path planning; performing loss prediction optimization according to the first space of maintenance path planning to obtain a second space of maintenance path planning; performing maintenance path optimization maximization optimization on the second space of maintenance path planning according to a maintenance path optimization fitness analysis model to output a maintenance path optimization result; and controlling the candidate maintenance robot network to perform maintenance on the power fault node perception network according to the maintenance path optimization result.

[0005] In another aspect, the application also provides a distributed power equipment cooperative maintenance path planning system, which comprises: an equipment network construction module, configured to construct a distributed power equipment network according to a set of distributed power equipment of a power grid; a fault awareness module, configured to perform fault awareness according to the distributed power equipment network, establish a power fault node awareness network, and analyze maintenance guidance features of the power fault node awareness network to construct a multi-dimensional maintenance guidance vector; a maintenance network construction module, configured to perform multi-dimensional awareness analysis on a maintenance robot library to establish a candidate maintenance robot network; a maintenance path planning module, configured to guide the candidate maintenance robot network to perform cooperative maintenance path planning on the power fault node awareness network according to the multi-dimensional maintenance guidance vector to obtain a maintenance path planning first space; a first optimization module, configured to perform loss prediction optimization according to the maintenance path planning first space to obtain a maintenance path planning second space; a second optimization module, configured to perform maintenance path optimization maximum optimization on the maintenance path planning second space according to a maintenance path optimization fitness analysis model to output a maintenance path optimization result; and an execution module, configured to control the candidate maintenance robot network to perform maintenance on the power fault node awareness network according to the maintenance path optimization result.

[0006] One or more technical solutions provided in the application have at least the following beneficial effects:

[0007] By constructing a distributed power equipment network according to a distributed power equipment set of a power grid, the connection relationship and topological structure between each distributed power equipment in the power grid can be clearly sorted out, providing a comprehensive equipment relationship framework for subsequent fault perception, maintenance guidance and robot path planning operations, so that the subsequent operations can be carried out in an orderly manner within this framework. According to the distributed power equipment network, a power fault node perception network is established to accurately locate the fault node. A multi-dimensional maintenance guidance vector is constructed by analyzing the characteristics of the maintenance guidance, which can describe the characteristics and maintenance requirements of the fault node from multiple dimensions, providing detailed guidance information for the path planning of the maintenance robot, so that the maintenance robot can plan the path more targetedly to the fault node for maintenance. A candidate maintenance robot network is established by multi-dimensional perception analysis of the maintenance robot library, which can comprehensively consider various factors such as the characteristics, functions and states of the robots, to provide a suitable candidate set of robot resources for subsequent collaborative maintenance path planning, and to ensure that the most suitable robot is selected to participate in the maintenance task. The candidate maintenance robot network performs collaborative maintenance path planning on the power fault node perception network according to the multi-dimensional maintenance guidance vector, realizes the initial collaborative matching from the robot capability to the task target, and forms a coarse-grained path solution, providing a basis for subsequent optimization. By considering various losses that may occur during the maintenance process, the first space of the maintenance path planning is optimized to obtain the second space of the maintenance path planning, which improves the path quality on the basis of ensuring the feasibility of execution. According to the maintenance path optimization and fitness analysis model, the maintenance path optimization second space is optimized to maximize the fitness of the maintenance path, which evaluates the fitness of the maintenance path from a more comprehensive perspective, maximizes the optimization of the maintenance path planning second space, and selects the final path that best suits the current scenario to ensure that the various requirements of the maintenance are met to the greatest extent. According to the maintenance path optimization result, the candidate maintenance robot network is controlled to maintain the power fault node perception network according to the optimal path, which converts the planning and optimization achievements into actual maintenance actions, realizes the closed-loop feedback from path planning to actual operation, and improves the response speed and execution accuracy of maintenance.

[0008] In summary, the present application constructs a comprehensive distributed power equipment maintenance path planning system by constructing a distributed power equipment network and integrating power fault perception, multi-dimensional vector guidance, robot resource analysis, path multi-stage optimization and fitness driven control, so that the maintenance robot can quickly and accurately reach the fault node for maintenance, greatly improving the response efficiency of the maintenance task, the rationality of path execution and the comprehensive utilization rate of robot resources, reducing resource waste and time loss in the maintenance process, and realizing efficient, intelligent and collaborative maintenance of distributed power equipment in complex environments.

[0009] The above description is only a summary of the technical solutions of the present application. In order to enable the technical means of the present application to be more clearly understood, and to be implemented according to the content of the description, and in order to enable the above and other purposes, characteristics and advantages of the present application to be more apparent and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 A flowchart of a distributed power equipment cooperative maintenance path planning method provided by an embodiment of the present application is shown.

[0011] Figure 2 A flowchart of obtaining a first space of a maintenance path planning in a distributed power equipment cooperative maintenance path planning method provided by an embodiment of the present application is shown.

[0012] Figure 3 A structure diagram of a distributed power equipment cooperative maintenance path planning system provided by an embodiment of the present application is shown.

[0013] Marked with a figure: equipment network construction module 10, fault sensing module 20, maintenance network construction module 30, maintenance path planning module 40, first optimization module 50, second optimization module 60, execution module 70. DETAILED DESCRIPTION

[0014] The embodiments of the present application provide a distributed power equipment cooperative maintenance path planning method and system, solve the technical problems that the prior art relies on static path planning or manual scheduling, lacks a cooperative scheduling mechanism based on global optimization, causes the maintenance path planning to respond with lag, and the resource configuration is unreasonable, and achieve the technical effects of improving the power equipment maintenance efficiency and resource utilization rate.

[0015] Embodiment one, as shown in the figure, the embodiments of the present application provide a distributed power equipment cooperative maintenance path planning method, the method comprises: Figure 1

[0016] Step S1: constructing a distributed power equipment network according to a distributed power equipment set of a power grid.

[0017] ​Specifically, the distributed power equipment set refers to a set of power infrastructure equipment such as distributed substations, wind turbines, solar photovoltaic arrays, and distribution switches distributed in various regions of the power system. The static information (device type, geographic coordinates, connection relationship) and dynamic information (operation state, load capacity, communication parameters, etc.) of all distributed equipment in the power grid are uniformly modeled, and a distributed power equipment network is constructed using a graph structure (Graph). For example, each device is abstracted as a node in the network, and the connection relationship between devices is abstracted as an edge, and a distributed power equipment network is constructed.

[0018] The construction of the distributed power equipment network provides a basic framework for subsequent fault awareness, maintenance planning, and other operations, making the structure of the entire power system clear and facilitating subsequent operations on specific equipment.

[0019] Step S2: Fault awareness is performed according to the distributed power equipment network, a power fault node awareness network is established, and maintenance guidance feature analysis is performed on the power fault node awareness network to construct a multi-dimensional maintenance guidance vector.

[0020] Specifically, based on the constructed distributed power equipment network, the operation state of the equipment is monitored in real time using fault detection sensors such as current sensors and voltage sensors. Sensor data and historical operation data are used in combination with deep learning models such as gated recurrent unit (GRU) for fault prediction to identify current fault or high-risk equipment, which are determined as fault nodes to establish a power fault node awareness network. The power fault node awareness network identifies all fault nodes and includes related connection device information of the fault nodes. Then, each fault node is analyzed in detail, and feature engineering methods are used to analyze maintenance guidance features based on historical maintenance records and current fault alarm information of the fault equipment. Four-dimensional feature vectors of maintenance urgency (such as device heat, risk level), topological dependence (whether the device is a hub node), maintenance value (degree of influence on overall system power supply), and location association (such as device geographic clustering) are extracted, and the above vectors are spliced or integrated to form a final multi-dimensional maintenance guidance vector. The multi-dimensional maintenance guidance vector is an important input data for subsequent path planning, guiding the path planning of the maintenance robot.

[0021] Through multi-dimensional feature extraction of fault nodes, the target guidance of path planning and the pertinence of task allocation are effectively improved, and the understanding and response capability to complex situations of the power grid are enhanced.

[0022] Step S3: Multi-dimensional perception analysis is performed on the maintenance robot library to establish a candidate maintenance robot network.

[0023] Specifically, the maintenance robot library is a collection of all schedulable robot-related data in the power equipment maintenance process. The data of each robot in the maintenance robot library is analyzed, and the current state information of the robot is extracted, such as the executable time period, geographic location, and energy consumption state, to obtain multi-dimensional robot data. The multi-dimensional robot data is filtered using a rule engine or a fuzzy logic decision model, and robots that match the current maintenance task and scene in terms of time margin (such as whether they can complete the task within the time period), robot state (such as health status and whether the maintenance capability is intact), and feature adaptation (such as whether the tools carried are suitable for the fault type) are selected. A graph structure is used to construct a candidate maintenance robot network based on the robots and possible collaboration relationships between the robots.

[0024] By performing multi-dimensional perception analysis on the maintenance robot library, the resource situation of the robots is comprehensively understood, and a candidate maintenance robot network is established, which realizes efficient and refined robot resource screening, avoids blind scheduling and resource waste, and ensures that the scheduling scheme is feasible and adaptive.

[0025] Step S4: According to the multi-dimensional maintenance guidance vector, guide the candidate maintenance robot network to perform collaborative maintenance path planning on the power fault node perception network to obtain a first space of maintenance path planning.

[0026] Specifically, the candidate maintenance robot network contains information about robots suitable for participating in maintenance tasks and their collaboration relationships. Based on the multi-dimensional maintenance guidance vector obtained in step S2 and the candidate maintenance robot network constructed in step S3, a path planning algorithm (such as the A-star algorithm or the Dijkstra algorithm) is used to plan a path for each robot from its current location to the fault node while considering the collaboration relationships between multiple robots (such as avoiding collisions between robots and reasonably allocating tasks to improve efficiency). These path solutions collectively constitute a first space of maintenance path planning, providing a rich set of candidate path combinations for subsequent optimization and improving the diversity and global optimization potential of the scheduling results.

[0027] Step S5: Loss prediction optimization is performed based on the first space of maintenance path planning to obtain a second space of maintenance path planning.

[0028] Specifically, each path in the first space of the maintenance path planning is executed in a simulation environment, simulation maintenance data corresponding to each path is obtained, the simulation maintenance data is input into a pre-constructed maintenance loss evaluation model, and risks or system burdens such as energy consumption, time consumption, task interruption rate, etc. that may be caused by the execution of each path are evaluated. According to the path evaluation results, paths that meet the expected loss requirements are screened out, and these paths collectively constitute the second space of the maintenance path planning. Through loss prediction optimization, high-risk, unreliable or high-execution-cost schemes are eliminated, various losses in the maintenance process are reduced, and the maintenance path planning is more economical and efficient.

[0029] Step S6: performing maintenance path optimization degree maximization optimization on the second space of the maintenance path planning according to the maintenance path optimization degree analysis model, and outputting a maintenance path optimization result.

[0030] Specifically, the maintenance path optimization degree analysis model is a function model for comprehensively evaluating maintenance efficiency and path loss, and is used to measure the degree of optimization of the maintenance path. All path schemes in the second space of the maintenance path planning are input into the maintenance path optimization degree analysis model one by one for optimization degree evaluation. According to the evaluation result of each path, an optimization algorithm (such as a linear programming algorithm) is used to maximize the optimization of the path schemes in the second space of the maintenance path planning, and the optimal (i.e. maximum optimization degree) maintenance path scheme under comprehensive evaluation is found, and the output is the maintenance path optimization result.

[0031] The maintenance path optimization degree analysis model comprehensively evaluates the pros and cons of the maintenance path, and the optimal maintenance path scheme is obtained by combining maximization optimization, which realizes the dynamic balance between efficiency and stability, and the finally output path scheme has higher cost performance and task execution controllability.

[0032] Step S7: controlling the candidate maintenance robot network to perform maintenance on the power fault node perception network according to the maintenance path optimization result.

[0033] Specifically, according to the maintenance path optimization result obtained in step S6, control instructions are sent to each robot in the candidate maintenance robot network through a scheduling system (such as a ROS robot operating system or a SCADA interface). These control instructions contain information such as the specific action path of the robot, the target device, and the time window, so that the robot can perform maintenance on the fault nodes in the power fault node perception network according to the optimal path. After receiving the instructions, the robot follows the specified path to the fault node and performs maintenance operations, realizing an automatic execution mechanism from planning to control, improving the automation level of the maintenance process, and reducing manual intervention. At the same time, the local task scheduling module of the robot is started, and the execution state is fed back to the main control system in real time during the maintenance process, forming a closed-loop control.

[0034] Further, the embodiment of the present application analyzes the maintenance guidance features of the power failure node perception network in step S2, constructs a multi-dimensional maintenance guidance vector, including:

[0035] Step S24: The maintenance urgency of the power failure node perception network is evaluated, and a maintenance urgency guidance vector is established.

[0036] Step S25: The topology dependence analysis of the power failure node perception network is performed, and a topology dependence maintenance guidance vector is established.

[0037] Step S26: The maintenance value of the power failure node perception network is evaluated, and a maintenance value guidance vector is established.

[0038] Step S27: According to the power failure node perception network, a position correlation analysis is performed, and a position correlation maintenance guidance vector is established.

[0039] Step S28: The maintenance urgency guidance vector, the topology dependence maintenance guidance vector, the maintenance value guidance vector and the position correlation maintenance guidance vector are output as the multi-dimensional maintenance guidance vector.

[0040] Specifically, the node data of each node in the power failure node perception network is collected, including load level, influence radius, historical failure frequency, etc. Using a failure influence analysis model, the influence range and cascading reaction possibility of the failure are evaluated according to the topological structure of the power grid, power load distribution and other factors, the urgency score of each failure node is calculated, then the standardization method (such as Z-score or Min-Max normalization) is used to standardize the urgency score, and the urgency score is arranged in order according to the failure node, forming a maintenance urgency guidance vector, which is used to guide the priority of the maintenance work and guide the robot to respond to high urgency nodes first.

[0041] The topological structure of the power failure node perception network is analyzed in detail using a topological analysis tool (such as a topological analysis software based on graph theory), the degree centrality and betweenness centrality indexes of each failure node are calculated, the two indexes are weighted according to the self-defined weight, the topological dependence analysis result is obtained, and all the topological dependence analysis results of the failure nodes are combined into a topological dependence guidance vector. The topological dependence guidance vector reflects the influence degree of the failure node on the network structure stability.

[0042] The overhaul value refers to the benefit brought by overhauling a node, including economy and system recovery contribution. The overhaul value of each fault node is evaluated by using a cost-benefit analysis model, comprehensively considering factors such as the cost of equipment, the recovery power supply capacity, and the cost of avoiding accident cascades, and the evaluation result is converted into a numerical value or a numerical range. The overhaul values of all fault nodes are combined to form an overhaul value guide vector. The overhaul value guide vector contains the overhaul value related information of each fault node, which is used to reasonably allocate the overhaul resources to the fault nodes with higher value in the case of limited resources.

[0043] The position correlation refers to the proximity of different fault nodes in space. According to the geographic coordinates or relative topological coordinates of the equipment corresponding to each fault node in the power grid, the node distance between each node is determined by using the Euclidean distance or map distance calculation method. According to these node distances, the K-means clustering or DBSCAN clustering is used for clustering analysis of the positions of each fault node, and the equipment with similar positions are clustered into the same overhaul task group, which is convenient for unified scheduling and path merging. According to the clustering result, a position label is assigned to each node, and the average distance of each node to other nodes in the task group to which it belongs is calculated as the position correlation degree of the node. After normalization processing, the position correlation degrees of all nodes are arranged into a vector as a position correlation overhaul guide vector, which is used to guide the robot to perform nearby combined overhaul.

[0044] The foregoing constructed overhaul urgency guide vector, topological dependence overhaul guide vector, overhaul value guide vector, and position correlation overhaul guide vector are combined in order to form a multi-dimensional overhaul guide vector, which realizes comprehensive quantitative evaluation of the fault nodes in multiple dimensions, provides a clear guidance basis for robot collaborative path planning, and makes the overhaul work more scientific, efficient, and reasonable.

[0045] Further, as shown in Figure 2 , the step S4 includes:

[0046] Step S41: According to each guide vector in the multi-dimensional overhaul guide vector, the candidate overhaul robot network is respectively guided to perform collaborative overhaul path planning on the power fault node perception network, and a collaborative overhaul path planning first domain is obtained.

[0047] Step S42: The multi-dimensional overhaul guide vector is randomly combined and fused to obtain a plurality of fusion overhaul guide vectors.

[0048] Step S43: According to the plurality of fusion overhaul guide vectors, the candidate overhaul robot network is respectively guided to perform collaborative overhaul path planning on the power fault node perception network, and a collaborative overhaul path planning second domain is obtained.

[0049] Step S44: merging the collaborative maintenance path planning first domain and the collaborative maintenance path planning second domain to generate the maintenance path planning first space.

[0050] Specifically, for each guidance vector (maintenance urgency guidance vector, topology-dependent maintenance guidance vector, maintenance value guidance vector, and location-dependent maintenance guidance vector) in the multi-dimensional maintenance guidance vector, the dimension is used as a priority indicator for path guidance, and the candidate maintenance robot network is guided to perform collaborative maintenance path planning on the power fault node perception network, obtaining a set of path sets generated under the guidance of each guidance vector, which together constitute the collaborative maintenance path planning first domain. Specifically, for the maintenance urgency guidance vector, nodes with high processing urgency are preferentially scheduled. For the topology-dependent maintenance guidance vector, the path is planned according to the topology-dependent relationship of the fault nodes, ensuring that the associated nodes are considered when repairing the fault nodes. For the maintenance value guidance vector, the fault nodes with high maintenance value are visited first. For the location-dependent maintenance guidance vector, the path is planned according to the location relationship of the fault nodes, such as preferentially selecting fault nodes that are close or in the same area for maintenance. Through the guidance based on each guidance vector, the path planning algorithm (such as A-star algorithm, Dijkstra algorithm, etc.) is used to plan the path for the robot, obtaining the collaborative maintenance path planning first domain.

[0051] The multiple vectors are randomly combined in different weight modes such as balanced fusion or biased fusion to form multiple fusion maintenance guidance vectors. The combination scheme can be a random combination of two vectors, three vectors, or four vectors. Among them, balanced fusion means equal-weight combination of each dimension guidance vector, and biased fusion means setting a specific weight for each dimension guidance vector.

[0052] For each fusion guidance vector, it is used as a task priority guide to perform collaborative maintenance path planning on the power fault node perception network using a path planning algorithm (such as A-star algorithm, Dijkstra algorithm, etc.), obtaining a set of paths corresponding to each fusion guidance vector, which together constitute the collaborative maintenance path planning second domain.

[0053] All path schemes in the collaborative maintenance path planning first domain and all path schemes in the collaborative maintenance path planning second domain are combined together to form a complete maintenance path planning first space, obtaining a more comprehensive maintenance path preliminary planning result, which provides a richer basic path scheme for subsequent loss prediction optimization. This process can simply combine the path schemes in the two sets into a new set.

[0054] Further, the step S5 of the embodiment of the present application comprises:

[0055] Step S51: Plan a first space according to the maintenance path, and extract a first maintenance path.

[0056] Step S52: Loss prediction is performed on the first maintenance path to determine a first path loss prediction coefficient.

[0057] Step S53: Determine whether the first path loss prediction coefficient is less than a path loss threshold.

[0058] Step S54: If the first path loss prediction coefficient is less than the path loss threshold, add the first maintenance path to a second space of the maintenance path planning.

[0059] Step S55: If the first path loss prediction coefficient is greater than or equal to the path loss threshold, eliminate the first maintenance path.

[0060] Step S56: Continue loss prediction optimization on the first space of the maintenance path planning according to the path loss threshold to generate the second space of the maintenance path planning.

[0061] Specifically, a planning path is extracted from the first space of the maintenance path planning as the first maintenance path. The extraction process can be performed in the order of the storage of the paths or randomly.

[0062] A maintenance loss evaluation model is constructed in advance. The first maintenance path is simulated for multiple rounds to collect path simulation data such as energy consumption, delay, and task failure rate. The path simulation data are input into the maintenance loss evaluation model, and the model outputs a first path loss prediction coefficient corresponding to the first maintenance path. The loss prediction coefficient is used to measure the comprehensive situation of various losses that may be generated by the path in the maintenance process. The larger the value, the greater the loss of the corresponding path.

[0063] The path loss threshold is a pre-set value used as a standard for determining whether the maintenance path is acceptable. The first path loss prediction coefficient calculated is compared with the pre-set path loss threshold to determine whether the first path loss prediction coefficient is less than the path loss threshold.

[0064] When the first path loss prediction coefficient is less than the path loss threshold, it means that the first maintenance path is acceptable in terms of loss, and the first maintenance path is directly added to the second space of the maintenance path planning.

[0065] When the first path loss prediction coefficient is greater than or equal to the path loss threshold, it means that the loss of the first maintenance path is large and does not meet the requirements. At this time, the first maintenance path is directly eliminated from the consideration range, such as being marked as an eliminated state in the first space of the maintenance path planning or being directly deleted from the data structure of the first space.

[0066] According to the foregoing path loss prediction and threshold screening method, the other maintenance paths in the maintenance path planning first space are looped, the next maintenance path (such as the second maintenance path, the third maintenance path, etc.) is continuously extracted, the path loss prediction coefficient thereof is calculated, and the path loss threshold is compared. According to the comparison result, it is decided whether to add to the maintenance path planning second space or to eliminate. Through processing all the paths in the maintenance path planning first space, the maintenance path planning second space is finally generated.

[0067] Through the above steps, the paths with high loss are removed from the maintenance path planning first space, the actual scheduling is avoided to have too high energy consumption or too high failure rate, and a higher quality candidate path set is provided for the next stage of optimal degree optimization.

[0068] Further, step S52 includes:

[0069] Step S521: According to the first maintenance path, multiple simulation maintenance is performed to obtain multiple simulation maintenance data sets.

[0070] Step S522: The multiple simulation maintenance data sets are input into a maintenance loss evaluation model to obtain multiple maintenance loss evaluation coefficients.

[0071] Step S523: The multiple maintenance loss evaluation coefficients are mean calculated to generate the first path loss prediction coefficient.

[0072] Specifically, using MATLAB Simulink, AnyLogic, and other simulation simulation software, according to the robot moving route, operation content, and other information specified by the first maintenance path, multiple simulation maintenance is performed, and the maintenance process data of each time is recorded, such as the time of the robot from the starting point to each fault node, the energy consumption at each operation link, the wear value of the key components of the robot, etc., to obtain the simulation maintenance data set.

[0073] The maintenance loss evaluation model is a pre-established model, which can be constructed based on random forest, neural network, or weighted scoring function, etc. It can quantitatively evaluate various losses in the maintenance process according to the input simulation maintenance data set, and output a comprehensive maintenance loss evaluation coefficient. For example, a weighted scoring function is designed, such as Loss=ω1*T+ω2*E+ω3*F, wherein Loss is the maintenance loss, T is the time consumption, E is the energy consumption, F is the number of intermediate faults, ω1, ω2, ω3 are the weights of the time consumption, energy consumption, and intermediate fault number, respectively, reflecting the influence degree of the time consumption, energy consumption, and intermediate fault number on the maintenance loss, and are determined according to the expert experience.

[0074] The multiple sets of simulated maintenance data corresponding to the first maintenance path are input into the maintenance loss evaluation model one by one, and the model outputs a maintenance loss evaluation coefficient corresponding to each set of simulated data according to the internal operation rules. The average value of the multiple maintenance loss evaluation coefficients corresponding to the first maintenance path is calculated as the first path loss prediction coefficient.

[0075] Through the above steps, the risk prediction and quantification of the path execution result are effectively realized, providing data support for subsequent path screening. Multiple simulations reduce the risk of misjudgment caused by occasional abnormalities, ensuring that the paths input into the next stage have high stability and execution reliability.

[0076] Further, the maintenance path optimization degree analysis model comprises a maintenance path optimization degree analysis function, and the maintenance path optimization degree analysis function is: ; wherein MPO represents the maintenance path optimization degree, exp represents the exponential function with the natural constant e as the base, PEW represents the maintenance efficiency predetermined weight, g(MPE) represents the normalized predicted path maintenance efficiency, PLW represents the path loss predetermined weight, and g(MPL) represents the normalized path loss prediction coefficient.

[0077] Specifically, the maintenance path optimization degree analysis function is constructed based on an exponential weighting function. The model comprehensively considers two key dimensions of path efficiency and path loss, sets efficiency weight (PEW) and loss weight (PLW) to balance the influence of the two on the optimization degree. The model first normalizes the predicted path maintenance efficiency (MPE) and the path loss prediction coefficient (MPL) to ensure comparability of the two types of data. Then, the maintenance path optimization degree (MPO) is calculated by an exponential function. The predicted path maintenance efficiency (MPE) is a value that measures the efficiency of a robot executing a path in completing a maintenance task. The higher the value, the more reasonable the robot distribution and scheduling under the path, the faster the task completion, and the more optimized the resource use. It can be calculated based on the aforementioned simulated maintenance data set. Exemplarily, , wherein T avg represents the average completion time in multiple rounds of simulated maintenance, T max represents the maximum acceptable completion time in historical maintenance tasks, which serves as a normalization reference, R s represents the success rate in simulated maintenance (i.e., the number of task completions divided by the number of simulations), T wait represents the average cooperative waiting time, i.e., the idle waiting time in multi-robot interactive scheduling, and ω1, ω2, and ω3 are weight coefficients that can be set according to actual scheduling goals.

[0078] The maintenance path optimization degree analysis function has the ability of nonlinear sensitivity regulation and control, and can significantly distinguish the advantages and disadvantages of different maintenance paths while maintaining the calculation efficiency. Based on the calculated maintenance path optimization degree, the path with the highest optimization degree is selected as the final maintenance path scheme, realizing the precise decision of path optimization.

[0079] Further, the step S3 of the embodiment of the application comprises:

[0080] Step S31: Time margin awareness optimization is performed according to the maintenance robot library to obtain a first candidate space of maintenance robots meeting the predetermined time margin.

[0081] Step S32: State awareness optimization is performed according to the first candidate space of maintenance robots to obtain a second candidate space of maintenance robots.

[0082] Step S33: Basic feature awareness is performed according to the second candidate space of maintenance robots to build a candidate maintenance robot network.

[0083] Specifically, the time margin refers to the time available for a new task after a maintenance robot completes a current task in the maintenance task arrangement. The current task and the estimated completion time of the current task of each robot in the maintenance robot library are collected, and a path planning algorithm is used to calculate the shortest time path of each robot to the target task point (i.e. the fault node). Let the required completion time of the task at the target task point be T deadline , the time required for the robot to reach the maintenance point from the current position be T travel , and the time required to complete the current maintenance task be T maint , then the total time T total =T travel +T maint , whether the total time T total is less than or equal to T deadline is determined, and robots meeting T total ≤T deadline are added to the first candidate space of maintenance robots, realizing the pre-selection of robots under the task time constraint. The first candidate space of maintenance robots contains all maintenance robots in the maintenance robot library that meet the predetermined time margin requirement.

[0084] For each maintenance robot in the first candidate space of the maintenance robot, state detection is performed, sensor state information and operation logs of each robot are read, and a state monitoring model or a preset state judgment rule is used to identify whether the robot has operation abnormalities such as communication interruption, insufficient power, mechanical failure, motion deviation, etc. For example, if the preset state judgment rule is used, state abnormalities such as power less than 30%, temperature anomaly, and intermittent communication frequency can be set. If the state monitoring model is used, a supervised learning model can be trained based on historical operation data to distinguish between "normal / abnormal" labels. If an abnormal condition is detected in a robot, it is excluded from the first candidate space of the maintenance robot, and the remaining robots form a second candidate space of the maintenance robot, avoiding the arrangement of unstable state robots to critical maintenance tasks and preventing the failure of maintenance tasks.

[0085] For each maintenance robot in the second candidate space of the maintenance robot, the basic feature information of the robot is extracted by querying the equipment information database of the robot, including model, specification, real-time position, software / hardware compatibility, etc. Then, based on these basic feature information, the robots are connected and organized to build a candidate maintenance robot network. For example, the robots can be classified according to the model, and robots of the same model have similar maintenance capabilities. In combination with the real-time position, the roles and relationships of different models of robots in different positions in the network are determined, thereby constructing a complete candidate maintenance robot network.

[0086] The above steps use a three-step screening mechanism of time margin perception, state anomaly detection, and basic capability matching to accurately construct a candidate maintenance robot network suitable for the current fault maintenance task from the robot library, effectively ensuring path availability and maintenance execution success rate, and achieving optimal utilization of scheduling resources and overall improvement of maintenance efficiency.

[0087] Further, the embodiment of the present application performs fault perception according to the distributed power equipment network in step S2 to establish a power fault node perception network, including:

[0088] Step S21: Perform fault prediction according to the distributed power equipment network to determine the fault prediction coefficient of each device.

[0089] Step S22: Determine whether the fault prediction coefficient of each device is greater than or equal to the fault threshold of each device to obtain the fault judgment result of each device.

[0090] Step S23: According to the fault judgment result of each device, the fault node of the distributed power equipment network is sorted to generate the power fault node perception network.

[0091] Specifically, operation data of each device in the distributed power equipment network is collected, including historical operation data and real-time operation data of the device. Supervised learning is performed on the device operation data to train a fault prediction model. The fault prediction model can predict a fault prediction coefficient of the power equipment according to the device operation data, the fault prediction coefficient being a quantitative representation of the possibility of each device in the distributed power equipment network failing, and the greater the fault prediction coefficient value, the greater the possibility of the corresponding power equipment failing.

[0092] A device fault threshold is set for each power equipment in advance to determine whether the device is in a state of high risk of failure. The device fault prediction coefficient of each device is compared with the corresponding device fault threshold to determine whether the device fault prediction coefficient is greater than or equal to the device fault threshold, and a device fault determination result is obtained. If the device fault prediction coefficient is greater than or equal to the corresponding device fault threshold, the device fault determination result is "device failure"; if the device fault prediction coefficient is less than the corresponding device fault threshold, the device fault determination result is "device normal".

[0093] According to the obtained device fault determination results, in the distributed power equipment network, devices with a device fault determination result of "device failure" (i.e., a device fault prediction coefficient greater than or equal to a device fault threshold) are screened out and marked as power failure nodes. These power failure nodes are combined according to the device connection relationship to construct a power failure node perception network.

[0094] The above steps establish a power failure node perception network through fault prediction, threshold determination, and fault node sorting, which can monitor and identify fault devices in the network in real time, provide a clear target and basis for subsequent maintenance path planning, and significantly improve the efficiency and pertinence of maintenance work.

[0095] Further, step S21 includes:

[0096] Step S211: Real-time monitoring of the distributed power equipment network is performed to obtain each power equipment monitoring data.

[0097] Step S212: Supervised learning is performed on the gate recurrent unit according to the corresponding power equipment monitoring records and power equipment fault detection records of the distributed power equipment network to generate each power equipment fault prediction model.

[0098] Step S213: The each power equipment monitoring data is input into the each power equipment fault prediction model to output the each device fault prediction coefficient.

[0099] Specifically, a sensor detection network for various types of power equipment (such as distributed transformers, circuit breakers, busbars, cable nodes, etc.) is established to collect real-time operation data of each power equipment, including but not limited to current, voltage, temperature, operating load, communication signal quality, packet loss rate, etc., and obtain each power equipment monitoring data.

[0100] From the operation records and maintenance records of each power equipment, the monitoring records and fault detection records of each power equipment are extracted, including the monitoring data of each power equipment in the past period of time, whether each power equipment has failed in the past, and detailed information such as fault type and fault occurrence time. The power equipment monitoring records and the power equipment fault detection records are sorted and preprocessed to make them suitable as input data for the gated recurrent unit. For example, the time series data in the monitoring records is divided according to a certain time interval, and the fault detection records are converted into corresponding fault labels (such as 0 for no fault and 1 for fault). Then, using the preprocessed monitoring records as input and the fault detection records as output labels, a gated recurrent unit (GRU) neural network is used to supervise the learning of each power equipment monitoring record and each power equipment fault detection record. During the learning process, the parameters (such as weights and biases) of the gated recurrent unit are adjusted to make the predicted output of the model as close as possible to the actual fault detection record. After multiple iterations of training, when the accuracy of the model reaches the expected standard, the power equipment fault prediction model is output.

[0101] The obtained monitoring data of each power equipment is input into the corresponding power equipment fault prediction model, and the neurons inside the model will calculate according to the pre-learned weights and biases to output a value representing the possibility of equipment failure, i.e. the equipment failure prediction coefficient.

[0102] The above steps can accurately capture the feature patterns of equipment failure through real-time monitoring, GRU model training and prediction, and realize the fault prediction of distributed power equipment, providing a scientific basis for subsequent fault handling.

[0103] In summary, the distributed power equipment cooperative maintenance path planning method provided by the embodiments of the present application has the following beneficial effects:

[0104] By constructing a distributed power equipment network, a system modeling and structured representation of distributed equipment in the power grid is achieved, providing a networked framework for subsequent fault identification and path planning. This serves as the starting point and data foundation for the entire process. Fault perception is performed based on the distributed power equipment network, establishing a power fault node perception network. Maintenance guidance feature analysis is then performed on this network to construct a multi-dimensional maintenance guidance vector. This enables intelligent identification and multi-dimensional feature extraction of power equipment faults, transforming fault nodes into vectorized expressions that can be used to guide path planning, providing guidance for subsequent decision-making. Multi-dimensional perception analysis is performed on the maintenance robot library to establish a schedulable robot resource candidate set, ensuring the accuracy of task allocation and path matching. Based on the multi-dimensional maintenance guidance vector, the candidate maintenance robot network is guided to perform collaborative maintenance path planning with the power fault node perception network, obtaining a first space for maintenance path planning. This achieves initial collaborative matching from robot capabilities to task objectives, forming a coarse-grained path solution. Loss prediction and optimization are then performed based on the first space for maintenance path planning, further optimizing the initial path to obtain a second space for maintenance path planning. This improves path quality while ensuring feasibility. A optimum-appropriate evaluation model is introduced to further filter the second space of maintenance path planning. Based on overall efficiency, the most suitable final path for the current scenario is selected to ensure optimal execution results. The optimization results are then used for robot control execution, achieving closed-loop feedback from path planning to actual operation, thereby improving maintenance response speed and execution accuracy.

[0105] Overall, the embodiments of this application significantly improve the response efficiency of maintenance tasks, the rationality of path execution, and the comprehensive utilization rate of robot resources through multi-dimensional information fusion and adaptive path optimization, realizing efficient, intelligent, and collaborative maintenance of distributed power equipment in complex environments.

[0106] Example 2, as Figure 3 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides a distributed power equipment collaborative maintenance path planning system, the system comprising:

[0107] The device network construction module 10 is used to construct a distributed power device network based on the distributed power device set of the power grid.

[0108] The fault perception module 20 is used to perceive faults based on the distributed power equipment network, establish a power fault node perception network, and perform maintenance guidance feature analysis on the power fault node perception network to construct a multi-dimensional maintenance guidance vector.

[0109] The maintenance network construction module 30 is used to perform multi-dimensional perception analysis on the maintenance robot library and establish a candidate maintenance robot network.

[0110] The maintenance path planning module 40 is configured to guide the candidate maintenance robot network to cooperatively plan a maintenance path for the power fault node perception network according to the multi-dimensional maintenance guidance vector, and obtain a first space of maintenance path planning.

[0111] The first optimization module 50 is configured to perform loss prediction optimization according to the first space of maintenance path planning, and obtain a second space of maintenance path planning.

[0112] The second optimization module 60 is configured to perform maintenance path optimization according to the second space of maintenance path planning, and output a maintenance path optimization result.

[0113] The execution module 70 is configured to control the candidate maintenance robot network to perform maintenance on the power fault node perception network according to the maintenance path optimization result.

[0114] Further, the fault perception module 20 is further configured to perform the following steps:

[0115] The maintenance urgency of the power fault node perception network is evaluated, and a maintenance urgency guidance vector is established. The topological dependency of the power fault node perception network is analyzed, and a topological dependency maintenance guidance vector is established. The maintenance value of the power fault node perception network is evaluated, and a maintenance value guidance vector is established. The position correlation of the power fault node perception network is analyzed, and a position correlation maintenance guidance vector is established. The maintenance urgency guidance vector, the topological dependency maintenance guidance vector, the maintenance value guidance vector, and the position correlation maintenance guidance vector are output as the multi-dimensional maintenance guidance vector.

[0116] Further, the maintenance path planning module 40 is further configured to perform the following steps:

[0117] According to each guidance vector in the multi-dimensional maintenance guidance vector, the candidate maintenance robot network is guided to cooperatively plan a maintenance path for the power fault node perception network, and a first domain of cooperative maintenance path planning is obtained. The multi-dimensional maintenance guidance vector is randomly combined and fused to obtain a plurality of fused maintenance guidance vectors. According to the plurality of fused maintenance guidance vectors, the candidate maintenance robot network is guided to cooperatively plan a maintenance path for the power fault node perception network, and a second domain of cooperative maintenance path planning is obtained. The first domain of cooperative maintenance path planning and the second domain of cooperative maintenance path planning are merged to generate the first space of maintenance path planning.

[0118] Further, the first optimization module 50 is further configured to perform the following steps:

[0119] According to the first space of the maintenance path planning, a first maintenance path is extracted; loss prediction is performed on the first maintenance path to determine a first path loss prediction coefficient; it is judged whether the first path loss prediction coefficient is less than a path loss threshold; if the first path loss prediction coefficient is less than the path loss threshold, the first maintenance path is added to the second space of the maintenance path planning; if the first path loss prediction coefficient is greater than or equal to the path loss threshold, the first maintenance path is eliminated; according to the path loss threshold, loss prediction optimization is continuously performed on the first space of the maintenance path planning to generate the second space of the maintenance path planning.

[0120] Further, the first optimization module 50 is further used to perform the following steps:

[0121] According to the first maintenance path, multiple simulation maintenances are performed to obtain multiple simulation maintenance data sets; the multiple simulation maintenance data sets are input into a maintenance loss evaluation model to obtain multiple maintenance loss evaluation coefficients; mean calculation is performed on the multiple maintenance loss evaluation coefficients to generate the first path loss prediction coefficient.

[0122] Further, the maintenance path optimization fitness analysis model comprises a maintenance path optimization fitness analysis function, and the maintenance path optimization fitness analysis function is: ; wherein, MPO represents maintenance path optimization fitness, exp represents an exponential function with a natural constant e as a base, PEW represents a maintenance efficiency predetermined weight, g(MPE) represents a normalized predicted path maintenance efficiency, PLW represents a path loss predetermined weight, and g(MPL) represents a normalized path loss prediction coefficient.

[0123] Further, the maintenance network construction module 30 is further used to perform the following steps:

[0124] According to the maintenance robot library, time margin awareness optimization is performed to obtain a first candidate space of maintenance robots satisfying a predetermined time margin; according to the first candidate space of maintenance robots, state awareness optimization is performed to obtain a second candidate space of maintenance robots; according to the second candidate space of maintenance robots, basic feature awareness is performed to build the candidate maintenance robot network.

[0125] Further, the fault awareness module 20 is further used to perform the following steps:

[0126] According to the distributed power equipment network, fault prediction is performed to determine device fault prediction coefficients; it is judged whether the device fault prediction coefficients are greater than or equal to device fault thresholds to obtain device fault judgment results; according to the device fault judgment results, fault nodes of the distributed power equipment network are combed to generate the power fault node awareness network.

[0127] Further, the fault perception module 20 is further used to execute the following steps:

[0128] Real-time monitoring is performed on the distributed power equipment network to obtain power equipment monitoring data; a gated recurrent unit is supervised and learned according to each power equipment monitoring record and each power equipment fault detection record corresponding to the distributed power equipment network, to generate a power equipment fault prediction model; and the power equipment monitoring data is input into the power equipment fault prediction model to output a device fault prediction coefficient.

[0129] Through the foregoing detailed description of the distributed power equipment cooperative maintenance path planning method, those skilled in the art can clearly know the distributed power equipment cooperative maintenance path planning system in the embodiment. For the system disclosed in Embodiment 2, since it corresponds to the method disclosed in Embodiment 1, it has corresponding functional modules and beneficial effects, and the related parts can be referred to the method part.

[0130] The above description of the disclosed embodiments enables a person skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for distributed power equipment collaborative maintenance path planning, characterized in that, The method comprises the following steps: According to the distributed power equipment set of the power grid, a distributed power equipment network is constructed; According to the distributed power equipment network, a power fault node perception network is established, and a multi-dimensional maintenance guidance vector is constructed by analyzing the maintenance guidance characteristics of the power fault node perception network; A candidate maintenance robot network is established by performing multi-dimensional perception analysis on a maintenance robot library; According to the multi-dimensional maintenance guidance vector, the candidate maintenance robot network is guided to perform collaborative maintenance path planning on the power fault node perception network, and a first space of maintenance path planning is obtained; Loss prediction optimization is performed according to the first space of maintenance path planning, and a second space of maintenance path planning is obtained; The maintenance path optimization result is output by performing maintenance path optimization maximization on the second space of maintenance path planning according to a maintenance path optimization fitness analysis model; The candidate maintenance robot network is controlled to perform maintenance on the power fault node perception network according to the maintenance path optimization result; According to the multi-dimensional maintenance guidance vector, the candidate maintenance robot network is guided to perform collaborative maintenance path planning on the power fault node perception network, and a first space of maintenance path planning is obtained, which comprises the following steps: According to each guidance vector in the multi-dimensional maintenance guidance vector, the candidate maintenance robot network is guided to perform collaborative maintenance path planning on the power fault node perception network, and a first domain of collaborative maintenance path planning is obtained; A plurality of fusion maintenance guidance vectors are obtained by randomly combining and fusing the multi-dimensional maintenance guidance vectors; According to the plurality of fusion maintenance guidance vectors, the candidate maintenance robot network is guided to perform collaborative maintenance path planning on the power fault node perception network, and a second domain of collaborative maintenance path planning is obtained; The first domain of collaborative maintenance path planning and the second domain of collaborative maintenance path planning are merged to generate the first space of maintenance path planning.

2. The distributed power equipment collaborative maintenance path planning method of claim 1, wherein, The power fault node perception network is analyzed to construct a multi-dimensional maintenance guidance vector, which comprises the following steps: A maintenance urgency guidance vector is established by performing maintenance urgency evaluation on the power fault node perception network; A topological dependence maintenance guidance vector is established by performing topological dependence analysis on the power fault node perception network; A maintenance value guidance vector is established by performing maintenance value evaluation on the power fault node perception network; A location correlation maintenance guidance vector is established by performing location correlation analysis on the power fault node perception network; The maintenance urgency guidance vector, the topological dependence maintenance guidance vector, the maintenance value guidance vector, and the location correlation maintenance guidance vector are output as the multi-dimensional maintenance guidance vector. 3.The distributed power equipment collaborative maintenance path planning method of claim 1, wherein, Loss prediction optimization is performed according to the first space of maintenance path planning to obtain a second space of maintenance path planning, which comprises the following steps: A first maintenance path is extracted according to the first space of maintenance path planning; A first path loss prediction coefficient is determined by performing loss prediction on the first maintenance path; It is judged whether the first path loss prediction coefficient is less than a path loss threshold value; if the first path loss prediction coefficient is less than the path loss threshold, adding the first maintenance path to the maintenance path planning second space; if the first path loss prediction coefficient is greater than or equal to the path loss threshold, eliminating the first maintenance path; continue loss prediction optimization on the maintenance path planning first space according to the path loss threshold, to generate the maintenance path planning second space.

4. The distributed power equipment collaborative maintenance path planning method of claim 3, wherein, loss prediction on the first maintenance path to determine a first path loss prediction coefficient, including: multiple times of simulated maintenance according to the first maintenance path to obtain multiple simulated maintenance data sets; inputting the multiple simulated maintenance data sets into a maintenance loss evaluation model to obtain multiple maintenance loss evaluation coefficients; mean calculation on the multiple maintenance loss evaluation coefficients to generate the first path loss prediction coefficient.

5. The distributed power equipment collaborative maintenance path planning method of claim 1, wherein, The maintenance path optimization fitness analysis model includes a maintenance path optimization fitness analysis function, and the maintenance path optimization fitness analysis function is: ； wherein, MPO represents maintenance path optimization fitness, exp represents an exponential function with natural constant e as the base, PEW represents maintenance efficiency predetermined weight, g(MPE) represents normalized predicted path maintenance efficiency, PLW represents path loss predetermined weight, and g(MPL) represents normalized path loss prediction coefficient.

6. The distributed power equipment collaborative maintenance path planning method of claim 1, wherein, multi-dimensional perception analysis on a maintenance robot library to establish a candidate maintenance robot network, including: time margin perception optimization according to the maintenance robot library to obtain a first candidate space of maintenance robots satisfying a predetermined time margin; state perception optimization according to the first candidate space of maintenance robots to obtain a second candidate space of maintenance robots; basic feature perception according to the second candidate space of maintenance robots to build the candidate maintenance robot network.

7. The distributed power equipment collaborative maintenance path planning method of claim 1, wherein, fault perception according to the distributed power equipment network to establish a power fault node perception network, including: fault prediction according to the distributed power equipment network to determine each equipment fault prediction coefficient; determination of whether each equipment fault prediction coefficient is greater than or equal to each equipment fault threshold to obtain each equipment fault determination result; fault node analysis of the distributed power equipment network according to each equipment fault determination result to generate the power fault node perception network.

8. The distributed power equipment collaborative maintenance path planning method of claim 7, wherein, fault prediction according to the distributed power equipment network to determine each equipment fault prediction coefficient, including: real-time monitoring of the distributed power equipment network to obtain each power equipment monitoring data; supervised learning of a gating cycle unit according to each power equipment monitoring record and each power equipment fault detection record of the distributed power equipment network to generate each power equipment fault prediction model; inputting each power equipment monitoring data into each power equipment fault prediction model to output each equipment fault prediction coefficient.

9. A distributed power equipment collaborative maintenance path planning system, characterized in that, The system is used to execute the distributed power equipment collaborative maintenance path planning method of any one of claims 1-8, including: a device network construction module, configured to construct a distributed power equipment network according to a distributed power equipment set of a power grid; A fault perception module is configured to perform fault perception based on the distributed power equipment network, establish a power fault node perception network, perform maintenance guidance feature analysis on the power fault node perception network, and construct a multi-dimensional maintenance guidance vector; A maintenance network construction module is configured to perform multi-dimensional perception analysis on a maintenance robot library, and establish a candidate maintenance robot network; A maintenance path planning module is configured to guide the candidate maintenance robot network to perform collaborative maintenance path planning on the power fault node perception network based on the multi-dimensional maintenance guidance vector, and obtain a maintenance path planning first space; A first optimization module is configured to perform loss prediction optimization based on the maintenance path planning first space, and obtain a maintenance path planning second space; A second optimization module is configured to perform maintenance path optimization and fitness maximization optimization on the maintenance path planning second space based on a maintenance path optimization and fitness analysis model, and output a maintenance path optimization result; An execution module is configured to control the candidate maintenance robot network to perform maintenance on the power fault node perception network based on the maintenance path optimization result.

Images