A power distribution network monitoring method and system under robot inspection

Abstract

The application provides a power distribution network monitoring method and system under robot inspection, and relates to the technical field of power systems.The method comprises the following steps: configuring a sensor group;uploading a network node data set to a central cloud platform;after calling the network structure of the power distribution network on the central cloud platform, performing global anomaly identification according to the uploaded data;performing anomaly clustering according to the global anomaly nodes with the communication range of the robot as a constraint;after controlling the robot to move to a relay center point, building a local communication network, reading the window collection data of the network nodes;using the robot to perform power distribution network data collection;performing anomaly verification and reporting monitoring anomalies.The application solves the technical problem that the anomaly identification credibility is insufficient due to uneven distribution of anomalies in large-scale power distribution networks and high false alarm rates in the prior art.The introduction of robot inspection and cloud-based global analysis improves the credibility of anomaly identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power systems, and particularly relates to a power distribution network monitoring method and system under robot inspection. BACKGROUND

[0002] In a large-scale power distribution network, due to the complex network structure, a large number of nodes and a wide distribution range, combined with multiple factors such as dynamic load changes and environmental disturbances, the abnormal state has the characteristics of uneven distribution, high randomness, and often being limited to a specific local area or a short time window. The traditional power distribution network monitoring method mainly relies on manual periodic inspection or remote collection of fixed-point sensors. When facing a large amount of distributed data, there are problems of insufficient sensing coverage and long response period, and it is difficult to distinguish between occasional disturbances and actual faults, resulting in low reliability of abnormal identification. Especially in the large-scale power distribution network scenario, the abnormal state often presents a fragmented distribution, and it is difficult to cover all potential abnormal areas in time by relying on a single-point trigger verification strategy, resulting in problems of insufficient positioning accuracy and low response efficiency.

[0003] In summary, in the prior art, there is a technical problem that the abnormal state is unevenly distributed in a large-scale power distribution network, the false positive rate is high, the reliability of abnormal identification is insufficient, and the reliability of the power distribution network is further affected. SUMMARY

[0004] The purpose of the present application is to provide a power distribution network monitoring method and system under robot inspection, to solve the technical problem in the prior art that the abnormal state is unevenly distributed in a large-scale power distribution network, the false positive rate is high, the reliability of abnormal identification is insufficient, and the reliability of the power distribution network is further affected.

[0005] In view of the above problems, the present application provides a power distribution network monitoring method and system under robot inspection.

[0006] In a first aspect, the application provides a power distribution network monitoring method under robot patrol, which is implemented by a power distribution network monitoring system under robot patrol. The power distribution network monitoring method under robot patrol comprises the following steps: configuring a sensor group at a network node of a power distribution network, performing data collection based on the sensor group at the network node to establish a network node dataset; uploading the network node dataset to a central cloud platform after preprocessing by the network node; identifying global abnormal nodes according to the uploaded data after calling the network structure of the power distribution network at the central cloud platform; performing abnormal clustering according to the global abnormal nodes to configure relay center points after reading the communication range of the robot and taking the communication range as a constraint; performing local area communication network construction after controlling the robot to move to the relay center points, reading window collection data of network nodes in the local area communication network to establish a first verification dataset after successful construction; performing power distribution network data collection in the local area communication network by the robot to establish a second verification dataset; and reporting monitoring abnormalities after performing global abnormal node abnormal verification based on the first verification dataset and the second verification dataset in the local area communication network.

[0007] Optionally, a graph model is established by using the structural information of the power distribution network, the graph nodes of the graph model represent switch stations, transformers and junction points, and the edges of the graph model represent cable lines; a communicable perception domain of each global abnormal node is established by taking the communication range as a constraint condition, the communicable perception domain is mapped into the graph model to identify a set of coverage points centered on each global abnormal node; inter-node correlation analysis is performed on the global abnormal nodes to establish a node abnormal correlation value; after synchronizing the node abnormal correlation value into the graph model, joint abnormal clustering is performed according to the node abnormal correlation value and the set of coverage points, and a relay center point is configured according to the joint abnormal clustering result.

[0008] Optionally, the identification of coverable relay center candidate points is performed by taking the communication range as a constraint condition in the joint abnormal clustering result, and a set of coverable relay center candidate points is configured; the set of coverable relay center candidate points is adaptively analyzed based on communication cost and abnormal radiation, and the relay center point is selected according to the adaptive analysis result.

[0009] Optionally, the robot initiates a local area broadcast packet when the position information of the robot satisfies the preset area positioning; an initial local area network topology is established after reading the connection response of the global abnormal nodes in the joint abnormal clustering result; and the local area communication network is constructed by performing topology adjustment on the local area network topology under response quality optimization.

[0010] Optionally, a communication delay index between the robot and network nodes within the local area network topology is configured, and the communication delay index is used as a first evaluation index; a signal strength index of the network nodes within the local area network topology is configured, and the signal strength index is used as a second evaluation index; a node load index of the network nodes within the local area network topology is configured, and the node load index is used as a third evaluation index; a response quality function is constructed based on the first evaluation index, the second evaluation index, and the third evaluation index to perform topology optimization adjustment.

[0011] Optionally, after unifying the time nodes of the local communication network, a window zero point is configured; after configuring the acquisition window of the network node based on the window zero point and the preset window length, the network node in the local communication network is controlled to perform data acquisition based on the acquisition window to establish window acquisition data; the window acquisition data is sent back to the robot and saved as the first verification dataset.

[0012] Optionally, obtain the historical data stability score of the network node, establish a first window enhancement factor based on the historical data stability score; perform data stability verification of the collected data at the verification node within the collection window, and establish a second window enhancement factor; after enhancing the collection window according to the first window enhancement factor and the second window enhancement factor, construct the window collection data.

[0013] Optionally, the robot's sensing unit is activated, an inspection task is generated based on the distribution of network nodes within the local communication network, path planning is performed according to task priority, and data collection by the sensing unit is performed based on the path planning results to establish a sensing dataset; the sensing dataset is then saved as a second verification dataset.

[0014] Optionally, after unifying the features of the first verification dataset and the second verification dataset, a multi-dimensional verification feature space is constructed; cross-validation of global abnormal nodes is performed in the multi-dimensional verification feature space, and abnormality verification is completed based on the cross-validation results, and monitoring abnormalities are reported.

[0015] In a second aspect, the application further provides a power distribution network monitoring system under robot inspection, configured to perform the power distribution network monitoring method under robot inspection as described in the first aspect. The power distribution network monitoring system under robot inspection comprises: a data collection module, configured to configure a sensor group at a network node of a power distribution network, perform sensor group-based data collection at the network node, and establish a network node dataset; a data processing module, configured to upload the network node dataset to a central cloud platform after node preprocessing; a global anomaly identification module, configured to identify global anomalies according to uploaded data after the central cloud platform calls a network structure of the power distribution network, and configure global anomaly nodes; an anomaly clustering module, configured to perform anomaly clustering according to the global anomaly nodes with the communication range of a robot as a constraint, and configure relay center points; a first verification module, configured to perform local communication network construction after the robot moves to the relay center points, read window collection data of network nodes in the local communication network after successful construction, and establish a first verification dataset; a second verification module, configured to perform power distribution network data collection in the local communication network by using the robot, and establish a second verification dataset; and an anomaly verification module, configured to perform global anomaly node anomaly verification based on the first verification dataset and the second verification dataset in the local communication network, and report monitoring anomalies.

[0016] The one or more technical solutions provided in the application have at least the following beneficial effects:

[0017] The sensor group is configured at the network node of the power distribution network, the sensor group-based data collection is performed at the network node, and the network node dataset is established. The network node dataset is uploaded to the central cloud platform after node preprocessing. Global anomalies are identified according to uploaded data after the central cloud platform calls the network structure of the power distribution network, and global anomaly nodes are configured. The communication range of the robot is read, the communication range is used as a constraint, anomaly clustering is performed according to the global anomaly nodes, relay center points are configured, the robot moves to the relay center points, local communication network construction is performed, window collection data of network nodes in the local communication network is read after successful construction, a first verification dataset is established, power distribution network data collection in the local communication network is performed by using the robot, a second verification dataset is established, and global anomaly node anomaly verification based on the first verification dataset and the second verification dataset is performed in the local communication network, and monitoring anomalies are reported. That is, by combining global data analysis of the central cloud platform and local fine detection of the mobile robot, the first verification dataset and the second verification dataset are constructed for double anomaly verification, the accuracy of anomaly identification is improved, and the safety and reliability of the power distribution network are improved.

[0018] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating a method for monitoring power distribution networks using robotic inspection, as described in this application.

[0021] Figure 2 This is a schematic diagram of the structure of a power distribution network monitoring system under robot inspection according to this application.

[0022] Figure labeling: Data acquisition module 11, data processing module 12, global anomaly identification module 13, anomaly clustering module 14, first verification module 15, second verification module 16, anomaly verification module 17. Detailed Implementation

[0023] This application provides a method and system for monitoring power distribution networks using robotic inspection. It addresses the technical problem in existing technologies where uneven anomaly distribution in large-scale power distribution networks leads to high false alarm rates, resulting in insufficient reliability of anomaly identification and further impacting the reliability of the power distribution network. By combining global data analysis from a central cloud platform with fine-grained local detection by a mobile robot, a first and second verification dataset are constructed for dual anomaly verification, improving the accuracy of anomaly identification and thus enhancing the security and reliability of the power distribution network.

[0024] Below, the technical solutions in the present application will be described clearly and completely with reference to the drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the example embodiments described herein. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application. In addition, it should be noted that, for the convenience of description, only the parts related to the present application are shown in the drawings, not all.

[0025] Embodiment one, please refer to the attached Figure 1 The present application provides a power distribution network monitoring method under robot inspection, wherein the power distribution network monitoring method under robot inspection is executed by a power distribution network monitoring system under robot inspection, and the power distribution network monitoring method under robot inspection specifically includes the following steps:

[0026] A sensor group is configured at a network node of the power distribution network, data acquisition based on the sensor group is performed at the network node, and a network node data set is established.

[0027] Specifically, a sensor group is installed at a key position (i.e. a network node) of the power distribution network, and the monitored physical quantity is periodically or real-time collected. In the power distribution network, the network node is a physical device point with topological connection relationship, such as a switch station, a transformer, a terminal box, a cable junction point, etc., or a specific position on a line. The sensor group refers to a plurality of types of sensors installed at the network node of the power distribution network, such as temperature sensors, current / voltage sensors, partial discharge sensors, humidity sensors, vibration sensors, infrared temperature measurement modules, etc.

[0028] Key nodes are selected for deployment, such as selecting a ring network cabinet node every 500 meters, or distributing each transformer station area, to complete the sensor group distribution of each network node. For example, voltage / current sensors are used to monitor the load running state, infrared or thermocouple temperature sensors are used to judge the heating abnormality, humidity sensors are used to identify condensation or water intrusion, and partial discharge sensors are used to identify equipment insulation deterioration. The sensor group collects data at the network node according to the set sampling period, such as sampling once every 15 seconds, and forming a group of data upload packets every 5 minutes to obtain a network node data set. The network node data set includes all the data collected by the sensor group at each network node within a period of time.

[0029] Exemplarily, a certain 110kV substation is provided with a power distribution network covering about 120 switching stations and distribution box nodes, and a sensor group is deployed at each node as follows: the temperature sensor has an accuracy of ±0.1℃, a sampling period of 15 seconds, and a single-point data volume of 10B; the voltage sensor has an accuracy of ±0.5%, a sampling period of 1 second, and a single-point data volume of 20B; the current sensor has an accuracy of ±0.5%, a sampling period of 1 second, and a single-point data volume of 20B; the humidity sensor has an accuracy of 1%RH, a sampling period of 30 seconds, and a single-point data volume of 10B; the partial discharge ultrasonic sensor has a sensitivity of 65dB, a sampling period of 10ms, and a single-point data volume of 512B; and the vibration sensor has an accuracy of ±0.1g, a sampling period of 100ms, and a single-point data volume of 15B. The partial discharge sensor sampling period is explained as follows: the aggregated characteristic value (such as peak value and frequency) is uploaded once every 60 seconds, and the original waveform is sampled once every 10ms locally. The sensor group deployed at each node generates about 71MB of original data per day.

[0030] By configuring a sensor group at a key node and performing data collection, various parameters such as temperature, voltage, current, humidity, partial discharge, and vibration are monitored at the same node, the running state of the equipment is comprehensively grasped, and real-time, multi-point, and multi-parameter sensing of the network running state is achieved. Not only does it contain rich raw running information, but it also reflects the trend of the node state over time.

[0031] The network node data set is uploaded to the central cloud platform after being preprocessed by the node.

[0032] Specifically, the network node data set is uploaded to the central cloud platform data preprocessing after being preprocessed by the node. Data cleaning (removing obvious errors or invalid values), data filtering (smoothing noise), data compression (reducing data volume), feature extraction (extracting more representative parameters from raw data), etc. improve data quality, reduce redundancy, reduce transmission burden, and may perform preliminary anomaly filtering.

[0033] In each network node (such as a distribution box or ring main unit), the data collected by the local processing device is preliminarily processed to ensure the quality and effectiveness of the data. For example, for the collected analog data such as current and voltage, digital filtering is performed to remove transient spikes caused by electromagnetic interference. Assuming that the current reading at a certain time in the original data is Ia=120.5A, but after filtering it is stabilized at Ia=120.2A. For temperature and humidity data, simple linear calibration is performed, and differential encoding is used for compression, only transmitting the difference between the current value and the previous time value, rather than the entire absolute value. If the original data set has 720 data points, after compression only 200 difference points need to be transmitted.

[0034] The preprocessed data is sent to the central cloud platform via a communication network (such as Wi-Fi, LoRa, etc.). The data is formatted, for example, into JSON format, with each data packet including node ID, timestamp, sensor type, and measurement value. Data from different sampling periods are aligned (e.g., voltage 1 second / sample, temperature 15 seconds / sample). Data is encrypted during transmission to ensure security and tamper-proofing. Node preprocessing significantly improves data quality and reduces interference introduced by transmission errors or noise in the raw data.

[0035] After the central cloud platform calls the network structure of the power distribution network, it performs global anomaly identification based on the uploaded data and configures global anomaly nodes.

[0036] Specifically, the central cloud platform is a cloud computing platform responsible for receiving, storing, and processing sensor data from various nodes. It provides powerful computing capabilities and supports functions such as data analysis, intelligent identification, and alarms. The central cloud platform invokes the network structure of the power distribution network, which refers to the node topology of the power distribution network, including various equipment such as substations, distribution boxes, and switching stations, as well as the electrical connections between these devices. This is typically represented as a graph model, where nodes are vertices and lines are edges. The network structure of the power distribution network includes the type and location of nodes, their connection relationships, and the impedance parameters of the lines.

[0037] The topology information of the power distribution network is obtained through the central cloud platform, including the state of each network node (such as substations, switches, distribution boxes, etc.) and the connection relationship between these nodes. The topology of the power distribution network is represented as a graph model, with nodes representing nodes and edges representing power lines. Through uploading data combined with network structure, the central cloud platform analyzes the running state of all nodes in real time. Based on historical data and model thresholds, the current node state is compared with the standard value during normal operation (such as temperature not exceeding 85°C, current not exceeding rated value, etc.). For example, if the current of a certain node suddenly exceeds the set threshold, it is marked as an abnormal node. Global anomaly recognition not only determines whether a single node is abnormal based on its data, but also considers adjacent nodes, line load, historical data, and other information to determine whether a node or a section of line is abnormal. Global anomaly recognition emphasizes analyzing the entire network, rather than focusing on a single node. For example, a transformer's current temperature is 84°C, although it does not exceed the set threshold, but the central cloud platform identifies that the temperature has risen from 72°C to 84°C in the past three minutes, that is, the temperature rising rate (slope) is 4°C / min, which is higher than the set warning slope threshold of 2°C / min. At the same time, the transformer also produces partial discharge pulses, which increase from 2 times per minute to 12 times. Accordingly, although the temperature does not exceed the standard, the equipment is deteriorating rapidly and needs to be marked as a global abnormal node. In addition, from the overall perspective, if the current of a node increases from 29A to 30.5A, an increase of about 5%, the value does not appear to be abnormal, but if the current of the adjacent 3 nodes also increases by 5% to 7% in a short period of time, it means that load migration has occurred - perhaps a nearby line fault, the load is redistributed to these nodes, so it is also identified as an abnormal node.

[0038] When the anomaly identification is completed, the central cloud platform will generate a global abnormal node list, which contains all nodes that are determined to have potential faults or abnormal operation, which may cause the device to run overload due to faults, aging, external factors, etc. For each abnormal node, record the specific abnormal type, time of occurrence, impact range and other related information. For example, assume that a 110 kV substation includes 120 nodes in the distribution network under its jurisdiction, and after analyzing the data uploaded within 10 minutes, the following situations are identified: the temperature of node A7 is 72℃, 76℃, 84℃ (ΔT=12℃, Δt=3min, slope=4℃ / min), the local discharge pulse frequency is from 2 / min to 12 / min, the current remains within the rated value, and it is identified as an overheating insulation deterioration anomaly with a confidence of 0.94, and the impact range includes 3 downstream distribution boxes. Nodes B12, B13, B14, the current is synchronized from 28A to 30A, 29A to 31A, 30A to 32A, and the voltage remains stable, which is identified as a group load migration anomaly, with a confidence of B12=0.83, B13=0.80, B14=0.80, and the three nodes are located on the same feeder, with overlapping impact ranges (such as affecting the same bus). The output global abnormal node configuration is as follows: the abnormal type of A07 is temperature rise plus local discharge anomaly, the abnormal confidence is 0.94, the impact range is 3 downstream nodes, and the inspection priority is 10 (emergency); the abnormal type of B12-B14 is group load migration, the abnormal confidence is 0.83 / 0.80 / 0.80, the impact range is 6 nodes of feeder segment equipment, and the inspection priority is 8 (high).

[0039] By combining network structure and multi-node data, global anomaly identification can discover anomalies that are difficult to detect with single-node data and do not conform to the overall operation mode of the network, improving the accuracy and coverage of anomaly detection.

[0040] After reading the communication range of the robot, the relay center point is configured according to the global abnormal node and the communication range as a constraint.

[0041] Further, the application further includes the following steps: establishing a graph model using the structural information of the power distribution network, the graph nodes of the graph model representing switch stations, transformers, and junction points, and the edges of the graph model representing cable lines; establishing a communicable perception domain for each global abnormal node with the communication range as a constraint, mapping the communicable perception domain to the graph model, and identifying a set of coverage points centered on each global abnormal node; performing inter-node correlation analysis on the global abnormal nodes to establish node abnormal correlation values; after synchronizing the node abnormal correlation values to the graph model, performing joint anomaly clustering according to the node abnormal correlation values and the set of coverage points, and configuring a relay center point according to the joint anomaly clustering result.

[0042] Further, the application further comprises the following steps: performing the coverable relay center candidate point identification and configuring the coverable relay center candidate point set within the joint anomaly clustering result with the communication range as the constraint condition; performing the adaptive analysis of the coverable relay center candidate point set based on the communication cost and the abnormal radiation, and selecting the relay center point according to the adaptive analysis result.

[0043] Specifically, the structure information of the power distribution network is used to establish a graph model, i.e., the structure information describing the physical connection mode of the power distribution network, including the positions, types of all devices (such as switch stations, transformers, connection points) and the connection relationship between them through cable lines, is converted into a graph model. Each node in the graph represents a physical location, such as a switch station, a transformer, a connection point, etc., and each edge of the graph model represents the connection relationship between the nodes, mainly cable lines, usually with weights (such as cable length, load capacity).

[0044] The communication range of the robot, i.e., the maximum distance at which the robot can effectively send and receive data through its wireless communication module, determines which network nodes or devices the robot can directly communicate with and interact with data. The communication range of the robot is used as a constraint condition to determine the communicable perception domain of each global anomaly node, i.e., considering the communication range limitation of the robot, a set of all nodes (switch stations, transformers, connection points) that the robot can reach and directly communicate with is established around each global anomaly node, representing the geographical or topological area where the robot can perform local inspection and data collection around the anomaly node. In the graph model, the communication reachable area of each anomaly node is circled, i.e., the communicable perception domain, and the adjacent points that each anomaly node can directly perceive are marked in the graph structure, thereby forming a coverage structure in space. The coverage point set refers to the set of all adjacent communicable nodes circled in the graph model with each global anomaly node as the center and the communication range of the robot as the constraint.

[0045] The inter-node correlation analysis of the global abnormal nodes is performed to explore whether there is some kind of connection or common mode between different abnormal nodes, including physical proximity (such as adjacent nodes on the same line), common power supply, similar load characteristics, or the result of abnormal propagation on the network. The goal of correlation analysis is to find that the abnormality does not occur in isolation, but there may be a causal, propagating or symbiotic relationship. By considering the abnormal type similarity, time proximity (whether two abnormalities occur in a short time), graph topology distance (whether two nodes are physically adjacent or the number of interval nodes is less than or equal to 2), etc., the abnormal correlation between nodes is calculated to obtain the node abnormal correlation value. The node abnormal correlation value refers to the quantitative value of whether there is a possible fault correlation between two abnormal nodes, which is usually calculated based on abnormal type similarity, time proximity, topology proximity, etc. The node abnormal correlation value is usually between 0 and 1 (or other specific range), and the larger the value, the stronger the correlation. For example, nodes P1 and P2 are both temperature rise + partial discharge abnormalities, the time difference is only 2 minutes, and the topology distance is 1 (only one intermediate node in the graph), and their correlation value is set to 0.92 (close to 1, indicating strong correlation).

[0046] Synchronizing the calculated node abnormal correlation value to the graph model means that the node abnormal correlation value is used as the edge weight of the graph model, which means that not only the physical connection information of each node (especially those marked as global abnormal nodes) in the graph model is stored, but also the correlation strength data with other abnormal nodes is attached. According to the node abnormal correlation value and the coverage point set, joint abnormal clustering is performed to group the global abnormal nodes and form multiple clusters to obtain the joint abnormal clustering result. All abnormal nodes are divided into several groups, and the similarity between nodes in each group is high (behavior or coverage point), and the similarity between groups is low. The joint abnormal clustering result, each group is called an abnormal cluster, representing a local fault area. Combined with the coverage point set, joint abnormal clustering is performed to identify node groups that can be physically cooperated and have close abnormal correlation, avoiding the error of grouping distant and low correlation abnormal nodes. In short, according to the global abnormal node set, the abnormal correlation value matrix between each other, and the coverage point set of each node, clustering is performed to divide the global abnormal nodes into multiple abnormal clusters. Joint abnormal clustering divides those nodes with high abnormal correlation and physically cooperative coverage by a group of robots into the same cluster (or cluster).

[0047] Within the joint abnormal clustering results, identify the coverable relay center candidate points as the constraint condition. That is, in each abnormal cluster, select a batch of candidate points that can be directly reached by robots and can cover most of the abnormal nodes in communication. Extract the subgraph corresponding to each abnormal cluster in the graph model, traverse each node in the subgraph, and calculate the number of abnormal nodes that can be covered within the communication radius to obtain the coverable relay center candidate points. After completing the joint abnormal clustering, each cluster represents a group of related abnormal nodes that can be physically cooperatively covered. Within each cluster, one or more positions need to be identified as candidate positions for the center point of the robot communication relay or cooperative work. This center point itself does not necessarily have to be an abnormal node, but needs to be located within the communication coverage range of the nodes in the cluster so that the robot can communicate with it or communicate with other robots through it.

[0048] Calculate the communication cost and abnormal radiation for each candidate point, the communication cost represents the resources consumed for communication between robots or between robots and the center platform, mainly including time cost (communication delay) and energy cost (communication power consumption), and the abnormal radiation can be understood as the influence range or severity of abnormal nodes. An abnormal node may not only affect itself, but also affect its adjacent nodes. When selecting a relay center point, the overall influence range of the abnormal in the cluster served by the center point needs to be considered to ensure coverage of critical areas.

[0049] The candidate points are scored and ranked by considering multiple dimensions (such as communication cost and abnormal radiation intensity), so as to select the final optimal relay center point. The communication cost is modeled as a weighted function, for example, communication cost = a * (moving distance / maximum moving distance) + b * (network delay / maximum delay time) + c * (hop count / maximum hop count), where a, b, and c are weights of moving distance, network delay, and hop count, respectively, and a + b + c = 1. The abnormal radiation is calculated by the sum of the number of abnormal nodes in the communication range of the point * node weight. The adaptation score = a * (1 - communication cost) + β * (abnormal radiation / maximum abnormal radiation value) is obtained, where a and β are configurable weights, such as a = 0.4 and β = 0.6. For example, two abnormal clusters are obtained after clustering: the abnormal nodes of cluster 1 are A1, A2, and A3, and the abnormal nodes of cluster 2 are B1 and B2. The candidate point set (screening result) in cluster 1 is as follows: P1 can cover abnormal nodes A1 and A2, the moving distance is 60 meters, the network hop count is 2, the communication cost (normalized) is 0.5, and the abnormal radiation is 6; P2 can cover abnormal nodes A1, A2, and A3, the moving distance is 80 meters, the network hop count is 1, the communication cost (normalized) is 0.6, and the abnormal radiation is 9; P3 can cover abnormal node A2, the moving distance is 30 meters, the network hop count is 1, the communication cost (normalized) is 0.3, and the abnormal radiation is 3. After adaptation analysis, the adaptation score of P1 is 0.6, the adaptation score of P2 is 0.76, and the adaptation score of P3 is 0.48. Finally, P2 is selected as the relay center point of cluster 1. The calculation of cluster 2 is similar.

[0050] The originally dispersed abnormal nodes that may be related to each other are organized by combining abnormal pattern analysis and physical network layout, which not only makes subsequent robot inspection task planning more accurate and efficient, avoids blind and dispersed inspection, but also helps to understand the root cause and propagation path of the abnormality by identifying related abnormalities.

[0051] After the robot is controlled to move to the relay center point, a local communication network is built, and after the local communication network is successfully built, the window collection data of the network nodes in the local communication network are read to establish a first verification data set.

[0052] Further, the application further includes the following steps: when the position information of the robot satisfies the preset area positioning, the robot is controlled to initiate a local broadcast packet; after reading the connection response of the global abnormal nodes in the joint abnormal clustering result, an initial local network topology is established; and the local network topology is adjusted under response quality optimization to complete the local communication network building.

[0053] Further, the application further comprises the following steps: configuring a communication delay index of the robot and the network nodes in the local network topology, taking the communication delay index as a first evaluation index; configuring a signal strength index of the network nodes in the local network topology, taking the signal strength index as a second evaluation index; configuring a node load index of the network nodes in the local network topology, taking the node load index as a third evaluation index; constructing a response quality function according to the first evaluation index, the second evaluation index, and the third evaluation index, to perform topology optimization adjustment.

[0054] Specifically, the robot is controlled to move to a relay center point, and the robot moves accurately by inertial navigation (IMU), visual SLAM or GPS (if outdoors), and confirms that the position of the robot is within a preset deviation (such as ±1 meter), that is, the robot is positioned in a preset area. The preset area positioning refers to that after the robot reaches the vicinity of the relay center point, the robot positions itself in a smaller, pre-set geographical area centered on the relay center point. The robot autonomously navigates to the selected relay center point according to the previously planned path. For example, the robot may need to start from the last task point, walk about 150 meters along the preset cable path, and finally position itself within a range of 10 meters around the relay center point (such as a connection point numbered #012), which is achieved by scanning through the integrated GPS and laser radar. After reaching the predetermined area, the robot confirms the accurate position, and then controls the wireless communication module to initiate a local broadcast packet to the area within a radius of 50 meters (its communication range), which contains the ID of the robot and the information of requesting to establish communication. The local broadcast packet refers to a special data packet sent by the robot, which includes the robot number and identity, the current coordinate and task number, the signaling of requesting to establish local communication, the currently supported communication protocol parameters, and the broadcast of this data packet through the communication module around the robot, the purpose of which is to announce its existence and try to establish contact with other devices within its communication range.

[0055] After the surrounding nodes (including abnormal nodes or relay auxiliary nodes) receive the broadcast, they respond and return basic information (including RSSI, delay, and state), and establish a graph structure between all responding nodes and the robot, forming an initial local communication topology. The initial local network topology is a preliminary network connection relationship graph established between the robot and the devices within its communication range according to the received connection response information. The robot usually serves as the center node, and the surrounding responding devices serve as the connected nodes.

[0056] For each link of the robot-node in the initial topology, the communication delay index of the robot and the network node in the local network topology is configured, and the communication delay index is obtained by sending test data packets and recording the round trip time. The communication delay index is the time required for data to be sent from the robot to a node in the local network and returned to the robot from the node, reflecting the real-time performance of network communication. The signal strength index is the strength of the wireless signal received between the robot and each node in the local network, and the signal strength directly affects the reliability of communication. Low strength leads to data loss or errors. The node load index is the amount of data currently being processed or forwarded by each node in the local network, i.e. the amount of data forwarding or task processing currently carried by the node, which can be calculated by buffer length, CPU occupancy, etc. High-load nodes may respond slowly or fail, affecting network performance. For example, the communication delay of node 1 is 38 ms, the signal strength is -62 dBm, the node load is CPU usage 65%, and the buffer occupancy is 70%; the communication delay of node 2 is 45 ms, the signal strength is -75 dBm, and the node load is relatively light.

[0057] The communication delay index is used as the first evaluation index, the signal strength index is used as the second evaluation index, and the node load index is used as the third evaluation index, which are normalized to eliminate dimensional effects. According to the first evaluation index, the second evaluation index, and the third evaluation index, a response quality function is constructed, i.e. quality score = x*(1-first evaluation index / maximum evaluation index) + y*[(second evaluation index-minimum second evaluation index) / (maximum second evaluation index-minimum second evaluation index)] + z*(1-third evaluation index / 100), where x, y, and z are weight coefficients, such as 0.3, 0.4, and 0.3. The higher the score, the better. According to the score, high-score node links are retained, unstable nodes are removed, and relay chains between multiple nodes are optimized for connection (such as using hop scheduling and forwarding compression), and finally a stable, low-delay, and anti-interference communication network topology structure is formed. The robot serves as a temporary relay to complete the construction of the local communication network. For example, the delay of node 1 is 35 ms, the signal strength is -62 dBm, and the load is 30%, so the quality score is 0.81; the delay of node 2 is 48 ms, the signal strength is -68 dBm, and the load is 75%, so the quality score is 0.59; the delay of node 3 is 70 ms, the signal strength is -75 dBm, and the load is 50%, so the quality score is 0.41, and so on.

[0058] The topology adjustment under the response quality optimization of the local area network topology, that is, according to the delay, signal quality, load and other indicators of each communication link of the local area network topology, scoring analysis and structure adjustment are performed, and selectively, links are added or deleted, connection order is reset or relay path is configured, so as to improve the overall communication performance. The connection node is selected through multi-index optimization, the node with large delay, weak signal or overload is avoided to access, the communication stability and timeliness are improved, and before the robot performs a specific inspection task, a local communication network with excellent performance and adaptability to the on-site environment can be quickly and automatically established. Through accurate positioning, active broadcast connection establishment and quantitative evaluation and optimization adjustment based on multi-dimensional indicators (delay, signal strength and load), task failure or low efficiency caused by unstable communication or unreasonable network structure is effectively avoided.

[0059] Further, the application further includes the following steps: after the time node of the local communication network is unified, a window zero point is configured; after the acquisition window of the network node is configured based on the window zero point and the preset window length, the network node in the local communication network performs data acquisition based on the acquisition window to establish window acquisition data; the window acquisition data is returned to the robot and saved as a first verification data set.

[0060] Further, the application further includes the following steps: obtaining a historical acquisition data stability score of the network node, establishing a first window enhancement factor based on the historical acquisition data stability score; performing data stability verification of the acquisition data by the verification node in the acquisition window to establish a second window enhancement factor; constructing window acquisition data according to the first window enhancement factor and the second window enhancement factor after the acquisition window is enhanced.

[0061] Specifically, after the local communication network is built, in order to ensure that the internal clocks of all nodes (including robots, sensor nodes, etc.) participating in communication in the local communication network are synchronized to a common time reference, time node unification is performed, all node acquisition time references in the local area network are unified through a time synchronization protocol, and data timing consistency is ensured. The window zero point is configured to set an explicit starting time point for the upcoming data acquisition period. The preset window length is a time period length that the data acquisition will last, such as 10 minutes, 30 minutes or 1 hour, etc. The specific length depends on the type of abnormality to be monitored and the change rate. According to the window zero point and the preset window length, the acquisition window of the network node is configured, that is, starting from the window zero point, the time of the preset window length is continued. In the acquisition window, the node needs to perform a data acquisition task.

[0062] The historical data acquisition stability score of the network node is obtained, that is, the stability evaluation value of the data collected by the network node in the past period of time, which measures whether the data fluctuates frequently, such as standard deviation or volatility. The past data of each node is called to calculate the fluctuation of the key parameters (such as current, voltage, temperature), and the first window enhancement factor is obtained by calculating the standard deviation or coefficient of variation. The first window enhancement factor is a coefficient adjusted according to the historical stability score. If the historical score is high (the node has performed stably in the past), this factor may encourage the node to continue collecting data at a higher frequency or quality in the current window; if the historical score is low, it may trigger additional calibration or reduce the collection frequency of certain non-critical indicators to improve the quality of the current window data. For example, the voltage standard deviation of node A is 0.2V, and the current coefficient of variation is 3%. Normalize the stability score to get the first enhancement factor 1.05.

[0063] At the same time, the designated verification node in the local communication network will check the real-time quality of the data in real time during the collection, and generate the second window enhancement factor. The second window enhancement factor is a coefficient adjusted according to the data stability verification results of the verification node in the current collection window. If the verification finds that the data is unstable (such as abnormal fluctuation, data loss), the second window enhancement factor will prompt to take more forceful measures (such as increasing verification, requiring retransmission, adjusting collection parameters) to improve data quality. For example, node C has a sudden 5% current increase in the collection window, so the second window enhancement factor is set to 1.15 to further enhance its data reliability.

[0064] According to the first window enhancement factor and the second window enhancement factor, the collection window enhancement is carried out, and the strategy of data collection in the current collection window is dynamically adjusted, such as increasing the collection frequency of key data, starting redundant collection, marking or discarding unstable data, or triggering the self-calibration program of the node, etc. After the enhancement based on the historical and real-time feedback, the data collected at the end of the window forms the window collection data. After each node completes the collection, the data in the collection window is packaged and returned to the robot through the local communication network. The robot as the controller stores it in the central cloud platform as the first verification data set. By introducing time synchronization, dynamic enhancement mechanism based on history and real-time feedback, the quality and reliability of the data collected in a specific time window are significantly improved, the possibility of false positives is reduced, and the credibility of anomaly identification is improved.

[0065] The robot is used to perform power distribution network data collection in the local communication network to establish a second verification data set.

[0066] Further, the application further comprises the following steps: activating the sensing unit of the robot, generating an inspection task according to the network node distribution in the local area communication network, performing path planning according to the task priority, performing sensing unit data collection according to the path planning result, and establishing a sensing data set; saving the sensing data set as a second verification data set.

[0067] Specifically, the sensing unit of the robot is activated, including various environmental / electrical sensing devices integrated on the robot, such as infrared thermal imager, ultrasonic partial discharge probe, optical camera, laser ranging, vibration sensor, etc., so that it enters the working state and prepares to collect environmental data. According to the network node distribution in the local area communication network, an inspection task is generated, that is, according to the abnormal node distribution information of the local area communication network, the abnormal intensity is scored and the priority is allocated, the inspection task is determined, including the specified node, the inspection order, the sensing content, etc.

[0068] According to the task priority, the target point distribution, combined with the geographic map and obstacle information, path planning is performed to determine the optimal motion path. An optimal or suboptimal movement route is designed for the robot to start from the current position, pass through all nodes or regions that need to be inspected, and finally return (or reach the next target point). Optimal usually refers to the shortest time, the lowest energy consumption or the smallest risk. Specifically, the information of the current node to be inspected is obtained, including node ID, position coordinates, abnormal type and severity, and the task priority is determined. Using a heuristic graph search algorithm, path planning is performed according to the task priority, the node with high priority is visited first, when the priority is the same, the nearest distance is preferred, the remaining power and action range of the robot are considered, the terrain restrictions or inaccessible areas are adapted, and the path planning result is obtained. According to the task order and geographic location, a travel path is calculated for the robot, taking into account the shortest distance, minimum energy consumption or minimum time consumption.

[0069] The robot advances according to the path planning result, collects data through the sensing unit, records all the data collected by the sensors (such as infrared images, connector images, environmental sound waveforms, etc.) during the entire movement and inspection process, and obtains a sensing data set. The sensing data set refers to the original data set collected after the robot completes the inspection task, including images, waveforms, temperature curves, signal strengths, etc. After preprocessing the sensing data set, it is saved as a second verification data set for cross comparison with the first verification data set (such as temperature difference, frequency offset, etc.), to determine whether the anomaly is real, persistent, false or exaggerated. By combining network topology, abnormal information and priority for task planning and path planning, the robot completes key inspections in the optimal order, improves efficiency, reduces response time, and improves the accuracy and credibility of anomaly identification.

[0070] Performing global abnormal node anomaly verification based on the first verification data set and the second verification data set in the local communication network, and reporting monitoring anomalies.

[0071] Further, the application further comprises the following steps: after aligning the features of the first verification data set and the second verification data set, constructing a multi-dimensional verification feature space; performing cross verification of the global abnormal nodes in the multi-dimensional verification feature space, and completing anomaly verification according to the cross verification result to report monitoring anomalies.

[0072] Specifically, the first verification data set (sensor historical enhancement data) and the second verification data set (robot field perception data) are aligned in features, and the corresponding monitoring attributes (such as temperature, current, vibration) in the two data sets are unified in data structure, so that they can be compared and analyzed. After aligning the features, a multi-dimensional verification feature space is constructed, the dimensions of which are determined by the key features extracted from the two data sets. Each abnormal node or related node can be represented as a point in this space, and the coordinates of the point are determined by the feature values.

[0073] In the multi-dimensional verification feature space, cross verification of the global abnormal nodes is performed to compare whether the feature vectors of the global abnormal nodes in the two data sets are consistent or have expected correlations. For example, if the first data set shows that the sensors around node Z detect an abnormally high temperature, and the second data set (robot infrared thermal imager) also captures that the surface of node Z indeed has an overheating point (such as 85℃) far beyond the normal range, and the vibration sensor also detects abnormal vibration, then the information of the two data sets is highly consistent. Cross verification uses the constructed multi-dimensional verification feature space to compare and comprehensively analyze the information from the first verification data set (sensor data) and the second verification data set (robot perception data).

[0074] According to the cross verification result, it is finally determined whether a node is indeed abnormal. For those nodes confirmed to be abnormal, a monitoring anomaly is reported, and an alarm information containing node ID, abnormal type, severity, evidence summary, etc. is generated to notify the operation and maintenance personnel. By fusing multi-source heterogeneous data, the accuracy and reliability of anomaly judgment are greatly improved, the limitations of a single data source (such as sensor false positives, robot visual blind spots, etc.) are effectively overcome, the false positive rate and the false negative rate are significantly reduced, and the final reported monitoring anomaly has high credibility.

[0075] In summary, the power distribution network monitoring method provided by the application has the following beneficial effects:

[0076] The network node data set is uploaded to a central cloud platform after node preprocessing; global abnormal nodes are configured after global abnormality identification according to uploaded data after the network structure of the power distribution network is called on the central cloud platform; after the communication range of the robot is read, abnormal clustering is performed according to the global abnormal nodes, and a relay center point is configured, taking the communication range as a constraint; after the robot is controlled to move to the relay center point, a local area communication network is built, and window collection data of network nodes in the local area communication network is read to establish a first verification data set after successful building; the robot is used to perform power distribution network data collection in the local area communication network to establish a second verification data set; global abnormal node abnormal verification is performed in the local area communication network based on the first verification data set and the second verification data set, and monitoring abnormalities are reported. That is, by combining global data analysis of the central cloud platform and local fine detection of the mobile robot, the first verification data set and the second verification data set are constructed for double abnormal verification, the accuracy of abnormal identification is improved, and the safety and reliability of the power distribution network are improved.

[0077] Embodiment two, based on the same inventive concept as the power distribution network monitoring method under robot patrol in the foregoing embodiment one, the present application also provides a power distribution network monitoring system under robot patrol, please refer to the attached Figure 2 , the power distribution network monitoring system under robot patrol comprises:

[0078] The data collection module 11 is configured to configure a sensor group at a network node of a power distribution network, perform sensor group-based data collection at the network node, and establish a network node data set; the data processing module 12 is configured to upload the network node data set to a central cloud platform after node preprocessing; the global abnormality identification module 13 is configured to perform global abnormality identification according to uploaded data after the network structure of the power distribution network is called on the central cloud platform; the abnormal clustering module 14 is configured to read the communication range of the robot, take the communication range as a constraint, perform abnormal clustering according to the global abnormal nodes, and configure a relay center point; the first verification module 15 is configured to control the robot to move to the relay center point, perform local area communication network building, and read window collection data of network nodes in the local area communication network to establish a first verification data set after successful building; the second verification module 16 is configured to use the robot to perform power distribution network data collection in the local area communication network to establish a second verification data set; and the abnormal verification module 17 is configured to perform global abnormal node abnormal verification in the local area communication network based on the first verification data set and the second verification data set, and report monitoring abnormalities.

[0079] Further, the abnormal clustering module 14 in the power distribution network monitoring system under the robot inspection is further configured to: establish a graph model by using the structure information of the power distribution network, wherein the graph nodes of the graph model represent switch stations, transformers, and connection points, and the edges of the graph model represent cable lines; take the communication range as a constraint condition to establish a communicable sensing domain of each global abnormal node, map the communicable sensing domain to the graph model, and identify a set of coverage points centered on each global abnormal node; perform inter-node correlation analysis on the global abnormal nodes to establish a node abnormal correlation value; after synchronizing the node abnormal correlation value to the graph model, perform joint abnormal clustering according to the node abnormal correlation value and the set of coverage points, and configure a relay center point according to the joint abnormal clustering result.

[0080] Further, the abnormal clustering module 14 in the power distribution network monitoring system under the robot inspection is further configured to: perform coverable relay center candidate point identification in the joint abnormal clustering result by taking the communication range as a constraint condition, and configure a set of coverable relay center candidate points; perform adaptive analysis on the set of coverable relay center candidate points based on communication cost and abnormal radiation, and select a relay center point according to the adaptive analysis result.

[0081] Further, the first verification module 15 in the power distribution network monitoring system under the robot inspection is further configured to: control the robot to initiate a local area broadcast packet when the position information of the robot satisfies a preset area positioning; read the connection responses of the global abnormal nodes in the joint abnormal clustering result to establish an initial local area network topology; and perform topology adjustment on the local area network topology under response quality optimization to complete the construction of a local area communication network.

[0082] Further, the first verification module 15 in the power distribution network monitoring system under the robot inspection is further configured to: configure a communication delay index of the robot and the network nodes in the local area network topology as a first evaluation index; configure a signal strength index of the network nodes in the local area network topology as a second evaluation index; configure a node load index of the network nodes in the local area network topology as a third evaluation index; and construct a response quality function according to the first evaluation index, the second evaluation index, and the third evaluation index to perform topology optimization adjustment.

[0083] Further, the first verification module 15 in the power distribution network monitoring system under the robot inspection is further configured to: after the time node of the local communication network is unified, configure a window zero point; after the window zero point and the preset window length are used to configure the collection window of the network node, the network node in the local communication network is controlled to perform data collection based on the collection window, and the window collection data is established; the window collection data is returned to the robot and saved as the first verification data set.

[0084] Further, the first verification module 15 in the power distribution network monitoring system under the robot inspection is further configured to: obtain the historical collection data stability score of the network node, establish a first window enhancement factor based on the historical collection data stability score; the data stability verification of the collection data of the verification node in the collection window is performed, and a second window enhancement factor is established; after the first window enhancement factor and the second window enhancement factor are used for collection window enhancement, the window collection data is constructed.

[0085] Further, the second verification module 16 in the power distribution network monitoring system under the robot inspection is further configured to: activate the perception unit of the robot, generate an inspection task according to the network node distribution in the local communication network, perform path planning according to the task priority, perform data collection of the perception unit according to the path planning result, and establish a perception data set; the perception data set is saved as the second verification data set.

[0086] Further, the abnormal verification module 17 in the power distribution network monitoring system under the robot inspection is further configured to: after the first verification data set and the second verification data set are uniformly aligned in features, a multi-dimensional verification feature space is constructed; cross verification of the global abnormal node is performed in the multi-dimensional verification feature space, the abnormal verification is completed according to the cross verification result, and the monitoring abnormality is reported.

[0087] The various embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The foregoing Figure 1 The power distribution network monitoring method and specific examples in embodiment one are also applicable to the power distribution network monitoring system under the robot inspection in the present embodiment. Through the foregoing detailed description of the power distribution network monitoring method under the robot inspection, those skilled in the art can clearly know the power distribution network monitoring system under the robot inspection in the present embodiment. Therefore, in order to make the specification brief, it will not be described in detail here.

[0088] The above description of disclosed embodiments enables one of ordinary skill in the art to make or use the application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0089] Obviously, many modifications and changes can be made to the application without departing from the spirit and scope of the application. It is understood that the application is not to be limited to the particular embodiments disclosed, but it is intended to cover all modifications which are within the scope of the application as defined by the language of the claims.

Claims

1. A method for monitoring power distribution networks under robot inspection, characterized in that, The application relates to a method for monitoring an electric power distribution network, comprising the following steps: a sensor group is configured at a network node of the electric power distribution network, sensor group-based data collection is performed at the network node, a network node dataset is established; after the network node dataset is preprocessed at the network node, the network node dataset is uploaded to a central cloud platform; after the network structure of the electric power distribution network is called at the central cloud platform, global anomaly identification is performed according to the uploaded data, and a global anomaly node is configured; after the communication range of a robot is read, the communication range is taken as a constraint, anomaly clustering is performed according to the global anomaly node, and a relay center point is configured; after the robot is controlled to move to the relay center point, a local area communication network is built, window collection data of network nodes in the local area communication network is read after the local area communication network is successfully built, and a first verification dataset is established; the step of reading the window collection data of the network nodes in the local area communication network after the local area communication network is successfully built and establishing the first verification dataset comprises the following steps: after time nodes of the local area communication network are unified, a window zero point is configured; after a network node is configured with a window zero point and a preset window length based on the window zero point, the network node in the local area communication network is controlled to perform data collection based on the collection window, and window collection data is established; the window collection data is returned to the robot and saved as the first verification dataset; the step of controlling the network node in the local area communication network to perform data collection based on the collection window comprises the following steps: a first window enhancement factor is established based on a historical collection data stability score of the network node; data stability verification of collected data of a verification node in the collection window is performed, and a second window enhancement factor is established; after the collection window is enhanced according to the first window enhancement factor and the second window enhancement factor, the window collection data is constructed; the robot is used to perform electric power distribution network data collection in the local area communication network, and a second verification dataset is established; global anomaly node anomaly verification based on the first verification dataset and the second verification dataset is performed in the local area communication network, and monitoring anomalies are reported.

2. The method for monitoring power distribution network under robot patrol of claim 1, wherein, the step of taking the communication range as a constraint and performing anomaly clustering according to the global anomaly node to configure a relay center point comprises the following steps: a graph model is established by using structure information of the electric power distribution network, graph nodes of the graph model represent switch stations, transformers and wiring points, and edges of the graph model represent cable lines; a communicable perception domain of each global anomaly node is established by taking the communication range as a constraint condition, the communicable perception domain is mapped into the graph model, and a set of covering points centered on each global anomaly node is identified; inter-node correlation analysis of the global anomaly nodes is performed, and a node anomaly correlation value is established; after the node anomaly correlation value is synchronized into the graph model, joint anomaly clustering is performed according to the node anomaly correlation value and the set of covering points, and a relay center point is configured according to a joint anomaly clustering result.

3. The method of claim 2, wherein the method further comprises: the step of configuring the relay center point according to the joint anomaly clustering result comprises the following steps: in the joint anomaly clustering result, communicable relay center candidate point identification is performed by taking the communication range as a constraint condition, and a set of communicable relay center candidate points is configured. Adaptively analyze the coverable relay center candidate point set based on communication cost and abnormal radiation, and select a relay center point according to the adaptively analyzed result.

4. The method of claim 1, wherein the method further comprises: After the control robot moves to the relay center point, a local area communication network is built, including: When the position information of the robot meets the preset area positioning, the robot is controlled to initiate a local area broadcast packet; After reading the connection response of the global abnormal node in the joint abnormal clustering result, an initial local area network topology is established; The local area network topology is adjusted under the response quality optimization to complete the local area communication network building.

5. The method of claim 4, wherein the method further comprises: The local area network topology is adjusted under the response quality optimization to complete the local area communication network building, including: The communication delay index of the robot and the network node in the local area network topology is configured, and the communication delay index is used as a first evaluation index; The signal strength index of the network node in the local area network topology is configured, and the signal strength index is used as a second evaluation index; The node load index of the network node in the local area network topology is configured, and the node load index is used as a third evaluation index; The response quality function is constructed according to the first evaluation index, the second evaluation index and the third evaluation index to perform topology optimization adjustment.

6. The method of claim 1, wherein the method further comprises: The robot is used to perform power distribution network data collection in the local area communication network, and a second verification data set is established, including: The perception unit of the robot is activated, the network node distribution in the local area communication network is generated to generate an inspection task, the path planning is performed according to the task priority, the perception unit data collection is performed according to the path planning result, and a perception data set is established; The perception data set is saved as the second verification data set.

7. The method of claim 1, wherein the method further comprises: identifying a location of the fault in the power distribution network; and providing a notification of the location of the fault to a user of the mobile device. 7 The global abnormal node anomaly verification based on the first verification data set and the second verification data set is performed in the local area communication network, and a monitoring anomaly is reported, including: After the first verification data set and the second verification data set are aligned in the unified feature, a multi-dimensional verification feature space is constructed; Cross-checking of the global abnormal node is performed in the multi-dimensional verification feature space, and the anomaly verification is completed according to the cross-checking result, and a monitoring anomaly is reported.

8. A power distribution network monitoring system under robotic patrol, characterized in that, The steps of the power distribution network monitoring method under the robot inspection according to any one of claims 1 to 7 are implemented, and the power distribution network monitoring system under the robot inspection includes: A data collection module is configured to configure a sensor group at a network node of a power distribution network, perform data collection based on the sensor group at the network node, and establish a network node data set; A data processing module is configured to upload the network node data set to a central cloud platform after node preprocessing; A global anomaly identification module is configured to call the network structure of the power distribution network on the central cloud platform, perform global anomaly identification according to the uploaded data, and configure a global abnormal node; An abnormal clustering module is configured to read the communication range of the robot, and perform abnormal clustering according to the global abnormal node under the constraint of the communication range, and configure a relay center point; The first verification module is configured to control the robot to move to the relay center point, perform local communication network construction, read window collection data of network nodes in the local communication network after the construction is successful, and establish a first verification data set; The second verification module is configured to use the robot to perform power distribution network data collection in the local communication network, and establish a second verification data set; The anomaly verification module is configured to perform global anomaly node anomaly verification based on the first verification data set and the second verification data set in the local communication network, and report monitoring anomalies.

Images