Multi-equipment linkage inspection method and system based on electrical equipment

Abstract

The invention discloses a multi-equipment linkage inspection method and system based on electrical equipment, and relates to the field of inspection of the electrical equipment, and the method comprises the following steps: S1, preparing inspection equipment; s2, constructing a dynamic Mesh network; s3, collecting and processing data; s4, dynamic task arrangement; s5, task decomposition; and S6, multi-modal data fusion is carried out. According to the multi-equipment linkage inspection method and system based on the electrical equipment, the problem that a single inspection terminal cannot cover a complex scene is effectively solved through collaborative operation of the multi-modal equipment, the inspection comprehensiveness is improved, the space-time database and the data fusion technology are utilized, and the multi-equipment linkage inspection method and system based on the electrical equipment are obtained. According to the method, accurate alignment and comprehensive analysis of multi-modal data such as vision, infrared and voiceprint are realized, the accuracy of power equipment state evaluation is improved, the inspection response speed is greatly improved based on dynamic Mesh networking and 5G edge calculation, the emergency defect processing delay is greatly shortened, and the timeliness requirement of power equipment operation and maintenance is met.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electrical equipment inspection, in particular to a multi-device linkage inspection method and system based on electrical equipment. BACKGROUND

[0002] Power equipment is a power production and consumption system composed of power generation, power transmission, power transformation, power distribution and power consumption. It converts primary energy in nature into electric power through power generation power plants, and then supplies electric power to users through power transmission, power transformation and power distribution. Mainly including two major categories of power generation equipment and power supply equipment, power generation equipment mainly is power station boiler, steam turbine, gas turbine, water turbine, generator, notching machine, transformer, etc., power supply equipment mainly is various voltage level transmission line, mutual inductor, contactor, etc.

[0003] In the field of operation and maintenance of power equipment, the traditional inspection method faces many challenges, and the existing technology has the problem of insufficient device cooperation. Single inspection terminal is difficult to cover complex scenes such as high-altitude GIS equipment and underground cable trench, and its data processing is fragmented, and visual, infrared, voiceprint and other data lack space-time alignment, such as isolated analysis of temperature and vibration data, which cannot fully reflect the equipment state; and the response timeliness is poor, the traditional work order system relies on manual intervention, and the emergency defect processing is delayed for a long time.

[0004] Therefore, it is necessary to propose a multi-device linkage inspection method and system based on electrical equipment to solve the above problems. SUMMARY

[0005] The main purpose of the present application is to provide a multi-device linkage inspection method and system based on electrical equipment, which can effectively solve the problems in the background art.

[0006] To achieve the above purpose, the technical scheme adopted by the present application is: A multi-device linkage inspection method based on electrical equipment, comprising the following operation steps: S1: preparing inspection equipment, including four-legged robot, unmanned aerial vehicle, wheeled robot; S2: constructing dynamic Mesh networking, adjusting the communication frequency band in real time according to the environmental electromagnetic interference, preferentially selecting the frequency band with small interference for data transmission to ensure the stability of communication; S3: data collection and processing, collecting the data collected by the inspection equipment in the inspection process, and performing timestamp alignment processing, and then storing it for subsequent retrieval and use; S4: dynamic task arrangement, constructing a three-dimensional matrix of device state, and introducing device operating condition parameters to dynamically correct the risk model; S5: task decomposition, discretize the space of power equipment into 200-200 millimeter grid units, and generate an optimal path using an improved ant colony algorithm; S6: multi-modal data fusion, fuse the data collected by the inspection equipment about temperature and voiceprint, and judge and alarm the state of the current power equipment according to the fused data.

[0007] Preferably, the quadruped robot is equipped with a 6-DOF mechanical arm, a 3D laser radar, a force sensor installed at the end of the mechanical arm, a temperature and humidity sensor, and a current transformer; the unmanned aerial vehicle is equipped with a double-light gimbal and a high-resolution panoramic camera; the wheeled robot integrates an ultrasonic array, a microphone matrix, a gas sensor, and a vibration sensor.

[0008] Preferably, in S2, the steps specifically include: S201: node discovery, in the power equipment inspection scene, the quadruped robot, the unmanned aerial vehicle, the wheeled robot, and the sensors and working equipment equipped thereon are all nodes of dynamic mesh networking, after each node is started, it actively sends a probe signal containing its own device identifier, supported communication frequency band, and transmission power, after the surrounding nodes receive the probe signal, they record the information of the sending node in the neighbor node table, and complete the discovery process of the neighbor nodes; S202: based on the discovered neighbor node information, each node starts to establish a communication link, and calculates the communication time slot of each node according to a TDMA time slot allocation algorithm, when there are nodes in the area of the power equipment, and it is known that the maximum distance , bandwidth , and protection interval of signal reliable transmission in the dynamic mesh networking under the current environment, then the time slot of each node is calculated by the formula: ; wherein is the time slot of each node, and the node sends data in the time slot allocated to itself, so as to avoid communication conflict between nodes, thereby establishing a stable communication link; S203: link maintenance and optimization, during the operation of the dynamic mesh networking, the nodes continuously monitor the link quality, when a link is caused to be degraded due to factors such as device movement and electromagnetic interference, the nodes at both ends of the link broadcast link abnormal information to other nodes, after the other nodes in the dynamic mesh networking receive the information, they try to find an alternative path according to their own neighbor node information and network topology, recalculate the time slot allocation, and adjust the network for self-healing, so as to maintain the communication stability of the entire dynamic mesh networking; S204: Data transmission, after establishing a stable link and completing time slot allocation, each node then transmits data to the target node in its own time slot; S205: 5G edge gateway, deploy LSTM traffic prediction model, dynamically adjust QoS priority, at the same time, integrate blockchain technology, encrypt and store important inspection data, each data block contains timestamp, device identification, data content and hash value of previous data block, form a chain structure that cannot be tampered with, used to ensure the security and traceability of data.

[0009] Preferably, in the S4, the following steps are specifically included: S401: Construction of three-dimensional matrix, based on real-time state parameters and historical failure weight of power equipment, construct risk assessment matrix, formula is: ; Wherein is the risk value, is the temperature and humidity coefficient of the power equipment, is the vibration value of the power equipment, is the voiceprint energy of the power equipment, F is the historical failure weight of the power equipment, , , , temperature, vibration, voiceprint and historical weight respectively; S402: Adjustment of dynamic weight, introduce equipment operating condition parameters to correct the weight, the wheeled robot monitors the load rate of the power equipment through the current transformer , the temperature and humidity data of the power equipment are collected in real time through the temperature and humidity sensor, when >80%, the temperature weight adjustment is started; when the humidity >70%RH, humidity compensation is started, wherein the adjustment formula of temperature weight is: ; Wherein is the adjusted temperature weight, is the correction linear unit; The compensation formula of humidity is: ; When the humidity increases from 60% to 80%, the vibration weight increases by 1%; S403: Recalculate the adjusted , every five minutes, adjust the adjusted , The risk assessment immediately impacts the current inspection task, and the results of each adjustment are recorded. When a power equipment actually fails, the effectiveness of the reverse adjustment results is verified.

[0010] Preferably, step S5 specifically includes the following steps: S501: Calculated based on S401 Label the corresponding grid cells; S502: Path planning for risk inspection based on ant colony algorithm. First, parameters are initialized: initial value of pheromone. Heuristic factors Risk preference coefficient Then, the probability of path selection is calculated, that is, the path chosen by a single ant. The probability is given by the formula: ; in For path pheromone concentration, As a heuristic factor, The risk value of the target grid. The pheromone importance coefficient. For heuristic factor coefficients, This is the risk preference coefficient. For the set of networks that are allowed to be accessed, Representing a path Historical appeal Representing a path Even the attraction, This indicates the driving effect of the target grid's risk level on path selection. This represents the sum of the attractiveness of all allowed paths, ensuring probability normalization. The probability of choosing a path for the ant; After the ant selects a path based on probability, it updates its pheromones using the following formula: ; in For path The amount of pheromone added in this iteration , This refers to the number of inspection devices that participate in path planning simultaneously. Where is the pheromone intensity constant. For the first The total length of the path traveled by the ant. For pheromone evaporation rate; S503: Task Priority Matrix, which categorizes risk levels into high, medium, and low. When the risk level is high, the task type is determined as urgent maintenance, the execution device is selected as a drone and a quadruped robot, and the time constraint is <5 minutes; when the risk level is medium, the task type is determined as key inspection, the execution device is selected as a wheeled robot, and the time constraint is <30 minutes; and when the risk level is low, the task type is determined as routine inspection, the execution device is selected as a drone, and the time constraint is a two-hour period. S504: Conflict resolution mechanism, when multiple devices compete for the same task, a priority formula is used for arbitration, and the formula is: The task is selected for the device with a high priority value. S505: Dynamic task scheduling, when an urgent maintenance task type is encountered, the time slot allocation of the dynamic Mesh network is adjusted, 50% bandwidth is reserved for the urgent maintenance task type data, and the current routine inspection task is interrupted. After the completion of the urgent maintenance, the interrupted task is re-added to the queue, and the risk level is increased by 1 level.

[0011] Preferably, in the S6, the following steps are specifically included: S601: Aligning the data collected by different devices at different times in space and time to form a time-aligned data set; S602: Mapping the temperature distribution of the power equipment to the digital twin, using the U-Net++ network to segment the overheating area, combining the three-dimensional structure model of the equipment to spatially locate and calculate the volume of the overheating area, and evaluating the severity of the thermal defect; S603: Setting the trigger conditions for three levels of alarms, including level one: voiceprint energy > 60 dB + temperature mutation > 5 ℃ / s; level two: voiceprint energy > 50 dB + temperature mutation > 3 ℃ / s; and level three: voiceprint energy > 45 dB + temperature mutation > 1 ℃ / s; S604: Dynamic weight adjustment based on load rate to calculate the current alarm level, and the formula is: .

[0012] A multi-device joint inspection system based on electrical equipment includes a hardware layer, a network layer, a data layer, a dynamic task arrangement module, and a multi-modal data fusion module. The hardware layer includes a quadruped robot, a drone, and a wheeled robot for collecting state data of power equipment. The network layer includes a dynamic Mesh networking module and a 5G edge gateway module. The data layer is a multi-modal spatio-temporal database for storing the collected data after time stamp alignment processing for subsequent retrieval and use. ​​​​The dynamic task arrangement module comprises a risk modeling submodule and a task decomposition and scheduling submodule; The multi-modal data fusion module comprises a data preprocessing submodule and a data analysis and decision submodule.

[0013] Preferably, the dynamic Mesh networking module is used to take each inspection device and sensor as a node, initiatively sends a detection signal containing device identification, communication frequency band and transmission power after starting, the surrounding nodes receive and record information to complete neighbor node discovery, calculates a communication time slot according to a TDMA time slot allocation algorithm, establishes a stable communication link based on the time slot, in the running process, the node monitors the link quality, when the link has a problem, other nodes find an alternative path and recalculate the time slot allocation, for realizing network self-healing, when data transmission, each node transmits data to the target node in the allocated time slot. The 5G edge gateway module is used for deploying an LSTM traffic prediction model, dynamically adjusting QoS priority, integrating blockchain technology, and encrypting and sharing the inspection data.

[0014] Preferably, the risk modeling submodule is used to construct a risk assessment matrix based on real-time state parameters and historical fault weights of electrical equipment, and introduce working condition parameters of the electrical equipment operation to dynamically correct the weights, and recalculate the adjusted weights every five minutes. The task decomposition and scheduling submodule is used to discretize the electrical equipment space into 200*200mm grid units according to the risk assessment results, label the risk values, adopt an improved ant colony algorithm for path planning, and establish a task priority matrix to determine the task type, execution device and time constraint according to the risk level.

[0015] Preferably, the data preprocessing submodule is used to perform space-time alignment on different time data collected by different electrical equipment, and form a time-aligned data set. The data analysis and decision submodule is used to map the temperature distribution of the electrical equipment to a digital twin, adopt a U-Net++ network to segment the overheating area, combine a three-dimensional structure model for spatial positioning and volume calculation, evaluate the severity of the thermal defect, and based on the set three-level alarm triggering conditions, timely issue an alarm and perform corresponding processing.

[0016] Compared with the prior art, the application provides a multi-device linkage inspection method and system based on electrical equipment, which has the following beneficial effects: This multi-device linkage inspection method and system based on electrical equipment effectively solves the problem that a single inspection terminal cannot cover complex scenarios through multi-modal device collaborative operation, improving the comprehensiveness of inspection. By utilizing spatiotemporal databases and data fusion technology, it achieves accurate alignment and comprehensive analysis of multi-modal data such as vision, infrared, and soundprint, improving the accuracy of power equipment status assessment. Based on dynamic mesh networking and 5G edge computing, it greatly improves the inspection response speed and significantly shortens the delay in emergency defect handling, meeting the timeliness requirements of power equipment operation and maintenance. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the steps of the present invention. Detailed Implementation

[0018] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0019] Example 1: like Figure 1 As shown, a multi-device coordinated inspection method based on electrical equipment includes the following steps: S1: Prepare inspection equipment, including a quadruped robot, a drone, and a wheeled robot. The quadruped robot is equipped with a 6-DOF robotic arm, a 3D LiDAR, a force sensor, a temperature and humidity sensor, and a current transformer installed at the end of the robotic arm. When performing touch inspections of equipment in confined spaces, the force sensor can detect the contact force in real time to avoid damage to the equipment due to improper force. At the same time, it can accurately collect the physical parameters of the equipment surface to provide richer data for equipment condition assessment. The drone is equipped with a dual-light gimbal and a high-resolution panoramic camera. The panoramic camera is used to acquire overall environmental information of the equipment during hovering inspections. Combined with the fine detection of the dual-light gimbal, it can achieve all-round observation of the equipment. The wheeled robot integrates an ultrasonic array, a microphone matrix, a gas sensor, and a vibration sensor. The gas sensor is used to detect the specific gas composition and concentration around the equipment. Combined with the partial discharge acoustic signature characteristics detected by the ultrasonic waves and microphones, it can more comprehensively judge the insulation status of the equipment.

[0020] S2: Construct a dynamic mesh network, adjust the communication frequency band in real time according to the environmental electromagnetic interference, and prioritize the frequency band with less interference for data transmission to ensure stable communication. This includes the following steps: S201: Node discovery, in the power equipment inspection scene, the quadruped robot, unmanned aerial vehicle, wheeled robot and its equipped sensors and working equipment are all nodes of dynamic mesh networking, after each node starts, it actively sends a probe signal containing its own device identifier, supported communication frequency band and transmission power, after the surrounding nodes receive the probe signal, the information of the sending node is recorded in the neighbor node table, and the neighbor node discovery process is completed; S202: Based on the discovered neighbor node information, each node starts to establish a communication link, and calculates the communication time slot of each node according to the TDMA time slot allocation algorithm, when there are nodes in the area of the power equipment, it is known that the maximum distance , bandwidth and guard interval of the signal can be reliably transmitted in the current environment, then the formula is: ; Wherein is the time slot of each node, the node sends data in the time slot allocated to itself, which is used to avoid communication conflict between nodes, so as to establish a stable communication link; S203: Link maintenance and optimization, during the operation of dynamic mesh networking, the node continuously monitors the link quality, when a link is caused by factors such as device movement and electromagnetic interference, the nodes at both ends of the link will broadcast link abnormal information to other nodes, after receiving the information, other nodes in the dynamic mesh network try to find alternative paths, recalculate the time slot allocation, and adjust the network to maintain the communication stability of the entire dynamic mesh network; S204: Data transmission, after establishing a stable link and completing time slot allocation, each node then transmits data to the target node in its own time slot, such as the quadruped robot collecting device detection data, sending the data to the 5G edge gateway and other nodes responsible for data transfer in the allocated time slot, and finally transmitting the data to the data processing center for analysis; S205: 5G edge gateway, deploy LSTM traffic prediction model, dynamically adjust QoS priority, at the same time, integrate blockchain technology, encrypt important inspection data for storage and sharing, each data block contains timestamp, device identifier, data content and hash value of the previous data block, forming a chain structure that cannot be tampered with, used to ensure the security and traceability of the data.

[0021] S3: Data collection and processing, based on the data collected by the inspection equipment during the inspection process, the data is collected and time stamped, and then stored for subsequent retrieval and use.

[0022] S4: Dynamic task arrangement, construct a three-dimensional matrix of device state, introduce device operating condition parameters, dynamically correct risk model, including the following steps: S401: Construction of three-dimensional matrix, based on real-time state parameters and historical failure weight of power equipment, construct risk assessment matrix, formula is: ; Wherein is the risk value, is the temperature and humidity coefficient of power equipment, is the vibration value of power equipment, is the voiceprint energy of power equipment, F is the historical failure weight of power equipment, , , , respectively for temperature, vibration, voiceprint and historical weight; S402: Dynamic weight adjustment, introduce device operating condition parameters to correct the weight, the wheeled robot monitors the load rate of power equipment through the current transformer , through the temperature and humidity sensor to collect the temperature and humidity data of power equipment, when >80%, start temperature weight adjustment; when the humidity >70%RH, start humidity compensation, wherein the adjustment formula of temperature weight is: ; Wherein is the adjusted temperature weight, is the correction linear unit; The compensation formula of humidity is: ; When the humidity rises from 60% to 80%, the vibration weight increases by 1%; S403: Recalculate the adjusted , every five minutes, the adjusted , immediately affect the risk assessment of the current inspection task, and record the result after each adjustment, when the power equipment actually fails, the effectiveness of the reverse serious adjustment result.

[0023] S5: Task decomposition, discretize the space of power equipment into 200-200 millimeter grid units, and generate the optimal path by using improved ant colony algorithm, including the following steps: S501: based on the calculation of S401 , the corresponding grid unit is labeled; S502: path planning for risk inspection based on ant colony algorithm, first, initialize parameters: initial value of pheromone ; heuristic factor ; risk preference coefficient ; then calculate the probability of path selection, i.e. the probability of a single ant selecting a path , the formula is: ; Where is the pheromone concentration of the path , the heuristic factor , the risk value of the target grid , the importance coefficient of pheromone , the heuristic factor coefficient , the risk preference coefficient , the set of allowed networks , the historical attractiveness of path , the immediate attractiveness of path , the driving effect of the risk level of the target grid on path selection , the sum of the attractiveness of all allowed paths to ensure probability normalization , the probability of an ant selecting a path ; After the ant selects a path according to the probability, update the pheromone, the formula is: ; ; Where is the amount of pheromone newly added to path in this round of iteration , is the number of inspection devices participating in path planning at the same time is the pheromone intensity constant is the total length of the path of the th ant is the pheromone evaporation rate S503: task priority matrix, divide the risk level into high, medium and low, when , judge as high, task type is defined as urgent maintenance, execution equipment selects unmanned aerial vehicle and quadruped robot, time constraint is <5 minutes; when , judge as medium, task type is defined as key inspection, execution equipment selects wheeled robot, time constraint is <30 minutes; when ​When the temperature is low, the task type is determined to be regular inspection, and the execution device is selected as a drone, and the time constraint is two hours; S504: Conflict resolution mechanism, when multiple devices compete for the same task, priority formula is used for arbitration, formula is: ; The task is selected for a device with a high priority value; S505: Dynamic task scheduling, when an emergency maintenance task type is encountered, the time slot allocation of the dynamic Mesh network is adjusted, 50% bandwidth is reserved for the emergency maintenance task type data, and the current regular inspection task is interrupted, after the emergency maintenance is completed, the interrupted task is re-added to the queue, and the risk level is increased by 1 level.

[0024] S6: Multimodal data fusion, the data collected by the inspection device about temperature and voiceprint are fused, and the state of the current power equipment is judged and alarm processing is performed according to the fused data, which specifically includes the following steps: S601: The data collected by different devices at different times are spatio-temporally aligned to form a time-aligned data set, wherein the visual adopts a cropped YOLOv7-tiny model, realizes meter identification on Jetson AGXXavier, introduces a DeepSORT algorithm, continuously tracks the identified meter, and automatically records the change process when the meter value changes, which is used to provide data support for power equipment operation trend analysis; the infrared eliminates thermal drift through a non-uniformity correction algorithm to generate an isotherm map, on the basis of which an infrared image enhancement algorithm based on deep learning is used to highlight the details of the thermal anomaly area of the power equipment, which is used to improve the accuracy of thermal defect detection; the voiceprint uses Mel cepstrum coefficient and GRU neural network to identify partial discharge type, introduces transfer learning technology, uses existing large amount of electrical equipment voiceprint data to train a general model, and then fine-tunes the model for specific substation equipment to improve the adaptability of the model in different scenarios; S602: The temperature distribution of the power equipment is mapped to the digital twin, the overheat area is segmented by using a U-Net++ network, and the overheat area is spatially located and volumetrically calculated in combination with the three-dimensional structure model of the equipment, which is used to evaluate the severity of the thermal defect; S603: The trigger conditions of three-level alarms are set, which are divided into first level: voiceprint energy > 60 dB + temperature mutation > 5 ℃ / s; second level: voiceprint energy > 50 dB + temperature mutation > 3 ℃ / s; third level: voiceprint energy > 45 dB + temperature mutation > 1 ℃ / s; S604: Dynamic weight adjustment, based on load rate The current alarm level is calculated, formula is: .

[0026] Embodiment two: A multi-device linkage inspection system based on electrical equipment includes a hardware layer, a network layer, a data layer, a dynamic task arrangement module, a multi-modal data fusion module, and the hardware layer.

[0027] The hardware layer includes a quadruped robot, a drone, and a wheeled robot, which are used to collect state data of power equipment.

[0028] The network layer includes a dynamic mesh networking module and a 5G edge gateway module. The dynamic mesh networking module uses each inspection device and sensor as a node, actively sends a detection signal containing device identification, communication frequency band, and transmission power after being started, and the surrounding nodes receive and record the information to complete neighbor node discovery, calculate the communication time slot based on the TDMA time slot allocation algorithm, and establish a stable communication link based on this. During operation, the node monitors the link quality, and when the link has a problem, it broadcasts abnormal information, and other nodes find an alternative path and recalculate the time slot allocation to realize network self-healing. During data transmission, each node transmits data to the target node within the allocated time slot. The 5G edge gateway module is used to deploy an LSTM traffic prediction model, dynamically adjust the QoS priority, integrate blockchain technology, and encrypt and store and share the inspection data.

[0029] The data layer is a multi-modal spatio-temporal database that is used to store the collected data after time stamp alignment processing for subsequent retrieval and use.

[0030] The dynamic task arrangement module includes a risk modeling submodule and a task decomposition and scheduling submodule. The risk modeling submodule constructs a risk assessment matrix based on real-time state parameters and historical fault weights of electrical equipment, and dynamically corrects the weights by introducing operating condition parameters of power equipment. The adjusted weights are recalculated every five minutes. The task decomposition and scheduling submodule is used to discretize the power equipment space into 200x200mm grid units according to the risk assessment results, label the risk values, use an improved ant colony algorithm for path planning, and establish a task priority matrix to determine the task type, execution device, and time constraints according to the risk level.

[0031] The multi-modal data fusion module includes a data preprocessing submodule and a data analysis and decision-making submodule. The data preprocessing submodule is used to align the time and space of different time data collected from different power equipment to form a time-aligned data set. The data analysis and decision sub-module is used for mapping the temperature distribution of the power equipment to the digital twin, using the U-Net++ network to segment the overheating area, combining the three-dimensional structure model for spatial positioning and volume calculation, evaluating the severity of the thermal defects, and based on the set three-level alarm trigger conditions, timely issuing an alarm and corresponding processing.

[0032] The inspection system is deployed in the area where the power equipment is located, after starting, the hardware device starts collecting data, the data is transmitted through dynamic Mesh networking, the dynamic task scheduling module plans the inspection task and schedules the device to execute according to the state and risk evaluation result of the power equipment, the multi-modal data fusion module processes and analyzes the collected data, judges the state of the power equipment, if there is an abnormality, an alarm is issued and the corresponding processing flow is started, in the inspection process, the device state and network condition are continuously monitored, the task and network configuration are dynamically adjusted according to the actual situation.

[0033] The above shows and describes the basic principles and main features of the present application and the advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present application, without departing from the spirit and scope of the present application, the present application can have various changes and improvements, these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A multi-device coordinated inspection method based on electrical equipment, characterized in that: The following steps are included: S1: Prepare inspection equipment, including quadruped robots, drones, and wheeled robots; S2: Construct a dynamic mesh network, adjust the communication frequency band in real time according to the environmental electromagnetic interference, and prioritize the frequency band with less interference for data transmission to ensure stable communication; S3: Data collection and processing. Data is collected based on the inspection equipment during the inspection process, and then timestamped and aligned before being stored for subsequent retrieval and use. S4: Dynamic task orchestration, constructing a three-dimensional matrix of equipment status, and introducing equipment operating condition parameters to dynamically correct the risk model; S5: Task decomposition, the space of the power equipment is discretized into 200-200 mm grid cells, and the optimal path is generated by an improved ant colony algorithm; S6: Multimodal data fusion, which integrates temperature and acoustic data collected by inspection equipment, and judges and alarms based on the fused data.

2. The method for multi-device coordinated inspection based on electrical equipment according to claim 1, characterized in that: The quadruped robot is equipped with a 6-DOF robotic arm, a 3D LiDAR, a force sensor, a temperature and humidity sensor, and a current transformer installed at the end of the robotic arm; the drone is equipped with a dual-light gimbal and a high-resolution panoramic camera; the wheeled robot integrates an ultrasonic array, a microphone matrix, a gas sensor, and a vibration sensor.

3. The method for multi-device coordinated inspection based on electrical equipment according to claim 1, characterized in that: S2 specifically includes the following steps: S201: Node discovery. In the power equipment inspection scenario, quadruped robots, drones, wheeled robots, and their equipped sensors and working equipment all serve as nodes in a dynamic mesh network. After each node starts up, it actively sends a detection signal containing its own device identifier, supported communication frequency bands, and transmission power. After receiving the detection signal, the surrounding nodes record the information of the sending node in the neighbor node table, thus completing the process of discovering neighbor nodes. S202: Based on the discovered neighbor node information, each node begins to establish a communication link. The communication time slot for each node is calculated according to the TDMA time slot allocation algorithm. When there are power equipment within the area... Given a set of nodes, the longest distance over which signals can be reliably transmitted in a dynamic mesh network under the current environment is known. ,bandwidth and protection interval Then, through the formula: ； in Each node has a time slot, and each node sends data within its assigned time slot to avoid communication conflicts between nodes, thereby establishing a stable communication link. S203: Link maintenance and optimization. During the operation of dynamic mesh networking, nodes continuously monitor link quality. When the signal quality of a link deteriorates due to factors such as device movement or electromagnetic interference, the nodes at both ends of the link will broadcast link abnormality information to other nodes. After receiving the information, other nodes in the dynamic mesh network will try to find alternative paths, recalculate time slot allocation, and perform self-healing adjustments to the network based on their neighbor node information and network topology to maintain the communication stability of the entire dynamic mesh network. S204: Data transmission. After a stable link is established and time slots are allocated, each node transmits data to the target node within its own time slot. S205: A 5G edge gateway that deploys an LSTM traffic prediction model to dynamically adjust QoS priorities. It also integrates blockchain technology to encrypt and share important inspection data. Each data block contains a timestamp, device identifier, data content, and the hash value of the previous data block, forming an immutable chain structure to ensure data security and traceability.

4. The method for multi-device coordinated inspection based on electrical equipment according to claim 1, characterized in that: S4 specifically includes the following steps: S401: Construction of a three-dimensional matrix: Based on the real-time status parameters of power equipment and historical fault weights, a risk assessment matrix is ​​constructed, with the following formula: ； in This is the risk value. The temperature and humidity coefficient of electrical equipment. The vibration value of the power equipment. For the acoustic energy of electrical equipment, F represents the historical fault weight of the power equipment. , , , These are temperature, vibration, voiceprint, and historical weights, respectively. S402: Dynamic weight adjustment, introducing equipment operating condition parameters to correct the weights, the wheeled robot monitors the load rate of power equipment through a current transformer. Temperature and humidity data of power equipment are collected in real time through temperature and humidity sensors. When >80%, temperature weighting is activated. Adjustment; when humidity When RH > 70%, humidity compensation is activated, with temperature weighting. The adjustment formula is: ； in For the adjusted temperature weights, To correct the linear unit; humidity The compensation formula is: ； When the humidity increases from 60% to 80%, the vibration weight Increase by 1%; S403: Recalculate and adjust every five minutes. , Adjusted , The risk assessment immediately impacts the current inspection task, and the results of each adjustment are recorded. When a power equipment failure actually occurs, the effectiveness of the reverse adjustment results is verified.

5. The method for multi-device coordinated inspection based on electrical equipment according to claim 1, characterized in that: S5 specifically includes the following steps: S501: Calculated based on S401 Label the corresponding grid cells; S502: Path planning for risk inspection based on ant colony algorithm. First, parameters are initialized: initial value of pheromone. Heuristic factors Risk preference coefficient Then, the probability of path selection is calculated, i.e., the path chosen by a single ant. The probability is given by the formula: ； in For path pheromone concentration, As a heuristic factor, The risk value of the target grid. The importance coefficient of pheromones. For heuristic factor coefficients, This is the risk preference coefficient. For the set of networks that are allowed to be accessed, Representing a path Historical appeal Representing a path Even the attraction, This indicates the driving effect of the target grid's risk level on path selection. This represents the sum of the attractiveness of all allowed paths, ensuring probability normalization. The probability of choosing a path for the ant; After the ant selects a path based on probability, it updates its pheromones using the following formula: ； in For path The amount of pheromone added in this iteration , This refers to the number of inspection devices that participate in path planning simultaneously. Where is the pheromone intensity constant. For the first The total length of the path traveled by the ant. For pheromone evaporation rate; S503: Task Priority Matrix, which categorizes risk levels into high, medium, and low. When the time limit is reached, the task type is designated as emergency repair, and the execution equipment is selected as drones and quadruped robots, with a time constraint of <5 minutes; When the time limit is reached, the task type is determined to be a key inspection, the execution equipment is a wheeled robot, and the time constraint is <30 minutes; When the time is low, the task type is set as routine inspection, the execution equipment is a drone, and the time constraint is a two-hour cycle. S504: Conflict resolution mechanism. When multiple devices compete for the same task, a priority formula is used for arbitration. The formula is as follows: ； The task should use equipment with a high priority value; S505: Dynamic task scheduling. When encountering an emergency repair task, the time slot allocation of the dynamic Mesh network is adjusted to reserve 50% bandwidth for the emergency repair task data. At the same time, the current regular inspection task is interrupted. After the emergency repair is completed, the interrupted task is added back to the queue, and the risk level is increased by 1 level.

6. The method for multi-device coordinated inspection based on electrical equipment according to claim 1, characterized in that: S6 specifically includes the following steps: S601: Spatiotemporally align data collected from different devices at different times to form a time-aligned dataset; S602: Maps the temperature distribution of power equipment to a digital twin, uses U-Net++ network to segment overheated areas, and combines the three-dimensional structural model of the equipment to spatially locate and calculate the volume of overheated areas to assess the severity of thermal defects. S603: Set the trigger conditions for three levels of alarms, which are: Level 1: Voiceprint energy > 60dB + temperature change > 5℃ / s; Level 2: Voiceprint energy > 50dB + temperature change > 3℃ / s; Level 3: Voiceprint energy > 45dB + temperature change > 1℃ / s. S604: Dynamic weight adjustment based on load rate The current alarm level is calculated using the following formula: 。 7. A multi-device linkage inspection system based on electrical equipment, employing a multi-device linkage inspection method based on electrical equipment as described in any one of claims 1-6, comprising a hardware layer, a network layer, a data layer, a dynamic task orchestration module, and a multi-modal data fusion module, characterized in that: The hardware layer; The hardware layer includes quadruped robots, drones, and wheeled robots, used to collect status data of power equipment; The network layer includes a dynamic Mesh networking module and a 5G edge gateway module; The data layer is a multimodal spatiotemporal database used to store the collected data after timestamp alignment for subsequent retrieval and use. The dynamic task orchestration module includes a risk modeling submodule and a task decomposition and scheduling submodule; The multimodal data fusion module includes a data preprocessing submodule and a data analysis and decision-making submodule.

8. A multi-device coordinated inspection method based on electrical equipment according to claim 7, characterized in that: The dynamic mesh networking module is used to treat each inspection device and sensor as a node. After startup, it actively sends a detection signal containing the device identifier, communication frequency band, and transmission power. Surrounding nodes receive and record the information to complete the discovery of neighbor nodes. The communication time slot is calculated according to the TDMA time slot allocation algorithm, and a stable communication link is established based on this. During operation, the nodes monitor the link quality. When a link problem occurs, the nodes broadcast the abnormal information, and other nodes find alternative paths and recalculate the time slot allocation to achieve network self-healing. During data transmission, each node transmits data to the target node within the allocated time slot. The 5G edge gateway module is used to deploy an LSTM traffic prediction model, dynamically adjust QoS priorities, integrate blockchain technology, and encrypt and share inspection data.

9. A multi-device coordinated inspection method based on electrical equipment according to claim 7, characterized in that: The risk modeling submodule constructs a risk assessment matrix based on real-time status parameters of electrical equipment and historical fault weights. It also introduces operating condition parameters of power equipment to dynamically correct the weights, recalculating the adjusted weights every five minutes. The task decomposition and scheduling submodule is used to discretize the power equipment space into 200×200mm grid cells based on the risk assessment results, label the risk values, use an improved ant colony algorithm for path planning, and establish a task priority matrix to determine the task type, execution equipment, and time constraints according to the risk level.

10. A multi-device linkage inspection method and system based on electrical equipment according to claim 7, characterized in that: The data preprocessing submodule is used to perform spatiotemporal alignment on different time data collected from different power devices to form a time-aligned dataset. The data analysis and decision-making submodule is used to map the temperature distribution of power equipment to a digital twin, use U-Net++ network to segment overheated areas, combine three-dimensional structural models for spatial positioning and volume calculation, assess the severity of thermal defects, and issue alarms and take corresponding actions in a timely manner based on the set three-level alarm trigger conditions.

Images