Transformer electrified intelligent maintenance device and efficient cleaning method

Abstract

The invention discloses a transformer electrified intelligent maintenance device and an efficient cleaning method, and the method comprises the steps: S1, multi-dimensional state perception and data fusion, S2, intelligent diagnosis and risk assessment, S3, adaptive maintenance strategy generation, S4, resource dynamic scheduling and path planning, S5, cooperative control and precise execution, and S6, operation closed-loop feedback and model optimization. And S7, a man-machine interaction and decision supervision step. According to the invention, a scheme integrating real-time monitoring, intelligent diagnosis, dynamic decision making and accurate execution is constructed, so that the improvement of live maintenance of the transformer is realized; microscopic state changes of equipment and the environment are sensed in real time by using a multi-source sensor network, and early-stage accurate early warning of defects is realized based on edge calculation and cloud platform collaborative analysis; by introducing a self-adaptive optimization algorithm, the system can dynamically generate and optimize a cleaning maintenance strategy and a resource allocation scheme, and it is ensured that operation is executed at the best opportunity with the highest efficiency, the lowest cost and the lowest risk.

Description

Technical Field

[0001] This invention relates to the field of power equipment operation and maintenance technology, specifically to a live-line intelligent maintenance device for transformers and an efficient cleaning method. Background Technology

[0002] Currently, the traditional methods for live-line maintenance and cleaning of large power equipment such as transformers mainly rely on regular manual inspections and pre-established fixed procedures. These methods have significant shortcomings: First, they lack real-time, multi-dimensional perception of equipment operating status, making it impossible to accurately identify and locate potential defects such as insulator contamination and overheating at connection points. Second, maintenance decisions depend on experience; the timing, frequency, and resource allocation of cleaning operations lack scientific optimization, easily leading to insufficient maintenance causing faults or over-maintenance resulting in resource waste. Third, the automation level of the operation process is low, personnel safety risks are high, and it is difficult to adapt to complex and changing environments and equipment operating conditions. With the development of smart grid construction, there is an urgent need for an operation and maintenance management method that can achieve full-process intelligentization, adaptive optimization, and efficient collaboration. Summary of the Invention

[0003] To address this, the present invention provides a live-line intelligent maintenance device and an efficient cleaning method for transformers, thereby solving the problem in the prior art of lacking real-time, multi-dimensional sensing capabilities of equipment operating status and being unable to accurately identify and locate potential defects such as insulator contamination and connection point overheating.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] The intelligent live-line maintenance device and efficient cleaning method for transformers include the following steps:

[0006] S1: Multi-dimensional state perception and data fusion step: Through a multi-source sensor network deployed on the transformer equipment body and its surrounding environment, multi-dimensional operating status parameters and working condition data, including leakage current, infrared thermal imaging temperature, partial discharge signal, ambient temperature and humidity and high-definition pollution images, are collected in real time, and the data is transmitted to the edge computing unit and the central cloud platform for synchronization and fusion processing.

[0007] S2: Intelligent diagnosis and risk assessment step. Based on the multi-dimensional data fused in step S1, it calls the equipment health assessment model and defect feature library in the central cloud platform. Through comparative analysis and pattern recognition, it diagnoses the pollution level of transformer insulators, overheated parts of connection points and potential discharge defects in real time, and quantitatively assesses their current risk level and evolution trend.

[0008] S3: Adaptive maintenance strategy generation step. Based on the risk assessment results obtained in step S2, combined with preset maintenance cost constraints, safety procedures and historical operation efficiency data, the optimal live-line maintenance strategy is dynamically generated using an adaptive optimization algorithm. The optimal live-line maintenance strategy includes at least the priority of maintenance operations, target area, recommended cleaning method and expected time window.

[0009] S4: Resource dynamic scheduling and path planning step. Based on the maintenance strategy generated in step S3, available resources are scheduled, including autonomous mobile cleaning robots, fixed cleaning devices, working periods and energy quotas; based on the three-dimensional structural model of the transformer equipment and real-time operating conditions, motion paths and work sequences are planned for the mobile execution units.

[0010] S5: Collaborative control and precise execution steps, which send the detailed operation instructions and path information determined in step S4 to the corresponding field operation units and cleaning execution mechanisms; under the collaborative control of the edge computing unit, drive the cleaning execution mechanism to perform precise cleaning of the target area according to the predetermined plan, and at the same time monitor the changes in equipment status and cleaning effect in real time through the sensor network.

[0011] S6: Operation closed-loop feedback and model optimization steps. After the maintenance operation is completed, collect and analyze the data of the entire operation, including actual energy consumption, time consumption, comparison of state parameters before and after cleaning, and target achievement. Feed the analysis results back to the central cloud platform to update the equipment historical status records and iteratively optimize the parameters of the equipment health assessment model and adaptive optimization algorithm.

[0012] Preferably, in step S1, the deployment of the multi-source sensor network adopts the principle of redundancy configuration, and at least two types of sensors based on different principles are deployed at the monitoring points for cross-validation; the data fusion processing adopts timestamp alignment and spatial coordinate matching technology, and uses the Kalman filter algorithm to reduce noise and estimate dynamic parameters.

[0013] Preferably, the equipment health assessment model in step S2 is a classification and regression hybrid model built on a deep convolutional neural network, and the equipment health assessment model is updated by combining offline training and online incremental learning; the defect feature library is constructed from historical defect case data, simulation data and expert experience rules.

[0014] Preferably, the path planning in step S4 specifically adopts a hybrid algorithm that combines the A search algorithm and the artificial potential field method; wherein, the A search algorithm is used to plan the global initial path in the grid map of the device's three-dimensional model, and the artificial potential field method is used to dynamically avoid unforeseen minor obstacles or risk areas caused by electromagnetic field fluctuations based on sensor feedback during real-time execution.

[0015] Preferably, the adaptive optimization algorithm in step S3 is a multi-objective optimization algorithm. The adaptive optimization algorithm takes risk reduction and profit maximization as its primary objective, and minimizes the weighted sum of operation cost and time cost as its secondary objective. It sets safety constraints, obtains the Pareto optimal solution set by solving the multi-objective optimization problem, and then selects the final execution strategy from the solution set based on real-time resource availability. The decision variables of the algorithm include the start time of the cleaning operation, the selected combination of execution units, the operation intensity of each execution unit, and the amount of cleaning agent used. The safety constraints include that the temperature rise of the critical connection points of the equipment must not exceed the threshold during the operation, the minimum safe distance must always be maintained between the operation unit and the live parts, and the operation must not cause the equipment protection system to malfunction.

[0016] Preferably, the adaptive optimization algorithm incorporates current power grid load forecast information and environmental weather forecast data as boundary conditions at each decision, and dynamically adjusts the parameter weights in the optimization model, so that the maintenance strategy has both long-term equipment health maintenance benefits and short-term power grid operation stability requirements.

[0017] Preferably, the collaborative control in step S5 adopts a master-slave control architecture and an event-driven mechanism; the central cloud platform acts as the master controller to issue macro commands, and the edge computing unit acts as the slave controller to be responsible for micro motion control and emergency obstacle avoidance; when the sensor detects a sudden change in the equipment state or an abnormality in the working unit in real time, the event-driven mechanism is triggered, the edge computing unit starts the locally preset safety response program, and at the same time, it alerts the central cloud platform.

[0018] Preferably, the safety response procedure includes at least: immediately suspending the current operation, controlling the operation unit to move to a preset safe standby point, and activating backup monitoring sensors for verification.

[0019] Preferably, it also includes S7: human-computer interaction and decision supervision step, which displays the process data, diagnostic results and generated strategy suggestions of steps S1 to S3 to the operation and maintenance personnel in real time through a visual interface, and provides interactive interfaces for manual confirmation, parameter fine-tuning or strategy rejection.

[0020] Preferably, the central cloud platform adopts a microservice architecture, decoupling and deploying data access services, model analysis services, optimization computing services, and instruction issuance services.

[0021] The present invention has the following advantages: The intelligent warehouse management and resource optimization method provided by the present invention improves the live-line maintenance of transformers by constructing a closed-loop system integrating real-time monitoring, intelligent diagnosis, dynamic decision-making and precise execution; by utilizing a multi-source sensor network to perceive the micro-state changes of equipment and environment in real time, and by conducting collaborative analysis based on edge computing and cloud platforms, it achieves early and accurate warning of defects; by introducing an adaptive optimization algorithm, the system can dynamically generate and optimize cleaning and maintenance strategies and resource allocation plans, ensuring that operations are performed at the best time, with the highest efficiency, lowest cost and lowest risk. This not only significantly improves the pertinence and effectiveness of maintenance operations, significantly reduces equipment failure rates and extends their service life, but also minimizes the intensity of manual labor and safety risks through intelligent scheduling and remote collaborative control, achieving efficient and lean utilization of operation and maintenance resources. Attached Figure Description

[0022] To more intuitively illustrate the prior art and this application, exemplary drawings are provided below. It should be understood that the specific shapes and structures shown in the drawings should not generally be regarded as limiting conditions for implementing this application; for example, based on the technical concept disclosed in this application and the exemplary drawings, those skilled in the art are able to easily make conventional adjustments or further optimizations to the addition / reduction / classification, specific shapes, positional relationships, connection methods, size ratios, etc. of certain units (components).

[0023] Figure 1 A flowchart of the intelligent live-line maintenance device and efficient cleaning method for transformers provided in the embodiments of this application. Detailed Implementation

[0024] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. It should be understood that these embodiments are merely for further explanation of the present invention and should not be construed as limiting the scope of protection of the present invention. Technical engineers in the field can make some non-essential improvements and adjustments to the present invention based on the above-described content. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1 The intelligent live-line maintenance device and efficient cleaning method for transformers include the following steps:

[0026] S1: Multi-dimensional state perception and data fusion step: Through a multi-source sensor network deployed on the transformer equipment body and its surrounding environment, multi-dimensional operating status parameters and working condition data, including leakage current, infrared thermal imaging temperature, partial discharge signal, ambient temperature and humidity and high-definition pollution images, are collected in real time, and the data is transmitted to the edge computing unit and the central cloud platform for synchronization and fusion processing.

[0027] S2: Intelligent diagnosis and risk assessment step. Based on the multi-dimensional data fused in step S1, it calls the equipment health assessment model and defect feature library in the central cloud platform. Through comparative analysis and pattern recognition, it diagnoses the pollution level of transformer insulators, overheated parts of connection points and potential discharge defects in real time, and quantitatively assesses their current risk level and evolution trend.

[0028] S3: Adaptive maintenance strategy generation step. Based on the risk assessment results obtained in step S2, combined with the preset maintenance cost constraints, safety procedures and historical operation efficiency data, the optimal live-line maintenance strategy is dynamically generated using an adaptive optimization algorithm. The strategy at least specifies the priority of the maintenance operation, the target area, the recommended cleaning method and the expected time window.

[0029] S4: Resource dynamic scheduling and path planning step. Based on the maintenance strategy generated in step S3, available resources are scheduled, including autonomous mobile cleaning robots, fixed cleaning devices, work periods and energy quotas; based on the three-dimensional structural model and real-time operating conditions of the transformer equipment, a safe and efficient spatial movement path and work sequence are planned for the mobile execution unit to avoid high electromagnetic risk areas and equipment protrusions.

[0030] S5: Collaborative control and precise execution steps, which send the detailed operation instructions and path information determined in step S4 to the corresponding field operation units and cleaning execution mechanisms; under the collaborative control of the edge computing unit, drive the cleaning execution mechanism to perform precise cleaning of the target area according to the predetermined plan, and at the same time monitor the changes in equipment status and cleaning effect in real time through the sensor network.

[0031] S6: Operation closed-loop feedback and model optimization steps. After the maintenance operation is completed, collect and analyze the data of the entire process of this operation, including actual energy consumption, time consumption, comparison of state parameters before and after cleaning, and target achievement. Feed the analysis results back to the central cloud platform to update the equipment historical status records and iteratively optimize the parameters of the equipment health assessment model and adaptive optimization algorithm to continuously improve the accuracy of subsequent maintenance decisions.

[0032] In implementation, step S1 involves comprehensively collecting equipment and environmental data through a multi-source sensor network and fusing it to provide a unified and reliable data foundation for subsequent decision-making. Step S2 utilizes advanced models and knowledge bases to deeply mine the fused data, transforming raw data into risk knowledge for precise problem identification. Step S3 uses optimization algorithms to dynamically generate optimal strategies based on risk assessment results, changing the rigid model of traditional fixed-cycle maintenance and achieving precise and personalized maintenance. Step S4 transforms abstract strategies into executable tasks and paths, ensuring the efficient and safe use of limited resources (such as robots and time windows). Step S5, through cloud and edge collaboration, accurately translates the plan into physical actions to complete the cleaning and maintenance operation. Step S6 continuously optimizes the model and algorithm using execution feedback data, enabling the system to self-improve. This method constructs a highly automated, intelligent, and self-learning complete operation and maintenance ecosystem. Through the iterative cycle of six steps from S1 to S6, it achieves continuous and lean management of transformer equipment status, achieving optimal equipment health with minimal resource consumption and operational risks, and improving the predictability, accuracy, and economy of operation and maintenance.

[0033] In step S1, the deployment of the multi-source sensor network adopts the principle of redundancy configuration, and at least two types of sensors with different principles are deployed at key monitoring points for cross-validation. The data fusion processing adopts timestamp alignment and spatial coordinate matching technology, and uses the Kalman filter algorithm to reduce noise and estimate dynamic parameters in order to improve the real-time performance and reliability of the data.

[0034] By employing redundant configurations of sensors based on different principles at key monitoring points, misjudgments caused by potential failures or errors of single-point sensors can be effectively avoided, thus significantly improving data reliability. Advanced data processing techniques such as timestamp alignment, spatial matching, and Kalman filtering can effectively eliminate noise interference and accurately fuse data from different spatiotemporal dimensions to obtain smoother and more accurate estimates of equipment status. This greatly reduces the risk of misdiagnosis, missed diagnosis, or invalid operations due to inaccurate data, ensuring the reliability and stability of the entire system's decision-making.

[0035] The equipment health assessment model in step S2 is a hybrid classification and regression model built on a deep convolutional neural network, which is updated through a combination of offline training and online incremental learning; the defect feature library is constructed from historical defect case data, simulation data, and expert experience rules.

[0036] The equipment health assessment model employs a hybrid model based on deep convolutional neural networks (CNNs). These models excel at automatically extracting deep features from images (such as images of dirt and thermal images) and sequential data, thereby achieving more accurate identification and classification of complex defect patterns. The model combines offline training with online incremental learning, enabling it to gain strong generalization capabilities based on historical big data while continuously adapting to slow changes in equipment characteristics or new defects based on newly generated data. The defect feature library integrates case studies, simulations, and expert rules, forming a multi-dimensional knowledge system, which further enhances the accuracy and adaptability of intelligent diagnosis.

[0037] The path planning in step S4 specifically employs a hybrid algorithm combining the A search algorithm (providing global navigation for the work unit) and the artificial potential field method (guiding the work unit to flexibly avoid unforeseen minor obstacles or electromagnetic interference that may occur during execution). The A search algorithm is used to plan the initial global path in the grid map of the equipment's 3D model, while the artificial potential field method is used to dynamically avoid risk areas caused by unforeseen minor obstacles or electromagnetic field fluctuations based on sensor feedback during real-time execution. This balances the global optimality of path planning with the flexibility of local real-time obstacle avoidance, ensuring the overall efficiency of the work path, avoiding blind searching, and empowering the execution unit to cope with complex dynamic environments on-site, significantly improving the safety and success rate of mobile cleaning operations.

[0038] The adaptive optimization algorithm in step S3 is a multi-objective optimization algorithm that simultaneously considers the timeliness, economy, and safety objectives of maintenance operations. The algorithm prioritizes risk reduction and maximizing benefits (ensuring the fundamental value of maintenance operations), with the secondary objective of minimizing the weighted sum of operation costs and time costs (balancing the economy and efficiency of operation and maintenance). It sets strict safety constraints, obtains the Pareto optimal solution set by solving this multi-objective optimization problem, and then selects the final execution strategy from the solution set based on real-time resource availability. The decision variables of the algorithm include the start time of the cleaning operation, the selected combination of execution units, the operation intensity of each unit, and the amount of cleaning agent used. The safety constraints include, but are not limited to: the temperature rise of critical connection points of the equipment during operation must not exceed the threshold, the minimum safe distance must always be maintained between the operation unit and the live parts, and the operation must not cause the equipment protection system to malfunction.

[0039] When implemented, this scheme transforms the abstract traditional optimization concept into a rigorous mathematical model that can be computed and executed. Through multi-objective trade-offs, the strategy generated by the algorithm is no longer an extreme solution of a single index, but a balanced solution with the best overall benefits, thus achieving safe, economical and efficient collaborative optimization.

[0040] The adaptive optimization algorithm incorporates current power grid load forecast information and environmental weather forecast data as boundary conditions at each decision, and dynamically adjusts the parameter weights in the optimization model so that the maintenance strategy can meet both the long-term equipment health maintenance benefits and the short-term power grid operation stability requirements.

[0041] Adaptive optimization algorithms incorporate external dynamic information such as grid load forecasting and weather forecasting as boundary conditions when making decisions. For example, when heavy loads or thunderstorms are predicted in the future, the algorithm can dynamically adjust weights, tending to advance or strengthen maintenance to improve the equipment's ability to withstand risks; conversely, when the grid load is light and the weather is good, it may focus more on economic optimization, thereby improving the overall resilience and operating efficiency of the power system.

[0042] The collaborative control in step S5 adopts a master-slave control architecture and an event-driven mechanism. The central cloud platform acts as the master controller, issuing macro-level instructions (responsible for macro-level strategies and resource scheduling), while the edge computing unit acts as the slave controller, responsible for micro-level motion control and emergency obstacle avoidance (responsible for real-time, low-latency micro-level motion control and emergency response). When the sensor detects a sudden change in the device state or an abnormality encountered by the operating unit in real time, the event-driven mechanism is triggered, and the edge computing unit starts its locally preset safety response program within milliseconds, while simultaneously alerting the central cloud platform.

[0043] The collaborative control in step S5 balances the needs of centralized intelligence and distributed agility. It not only ensures the advantages of global optimization and unified management, but also effectively addresses the security issues that may be caused by network latency by giving the edge side sufficient autonomous response capabilities, thus greatly improving the reliability and security of the entire system in the face of emergencies.

[0044] The safety response procedure includes at least: immediately suspending the current operation, controlling the operation unit to move to the preset safe standby point, and activating the backup monitoring sensor for verification. This forms a standardized emergency operation process, which greatly standardizes and strengthens the system's safety handling capabilities, enabling a rapid, orderly, and effective response in emergency situations.

[0045] It also includes S7: Human-Computer Interaction and Decision Supervision Steps. Through a visual interface, the process data, diagnostic results, and generated strategy recommendations from steps S1 to S3 are displayed to operations and maintenance personnel in real time. Interactive interfaces for manual confirmation, parameter fine-tuning, or strategy rejection are provided to ensure the intelligent decision-making process remains under control. In other words, at key decision-making stages (from data display to strategy generation), a transparent visual interface and manual intervention interface are provided to operations and maintenance personnel, making all intelligent analysis processes and decision recommendations visible, understandable, and superviseable to users. Operations and maintenance personnel can confirm, fine-tune, or reject system recommendations based on their own experience.

[0046] The central cloud platform adopts a microservice architecture, decoupling data access services, model analysis services, optimization calculation services, and command issuance services to support high-concurrency data processing and independent upgrades and expansions of algorithm modules. In this architecture, different functions such as data access, model analysis, optimization calculation, and command issuance are decoupled into independent, loosely coupled service modules, resulting in extremely high flexibility, maintainability, and scalability for the entire system.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A live-line intelligent maintenance device and efficient cleaning method for transformers, characterized in that, Includes the following steps: S1: Multi-dimensional state perception and data fusion step: Through a multi-source sensor network deployed on the transformer equipment body and its surrounding environment, multi-dimensional operating status parameters and working condition data, including leakage current, infrared thermal imaging temperature, partial discharge signal, ambient temperature and humidity and high-definition pollution images, are collected in real time, and the data is transmitted to the edge computing unit and the central cloud platform for synchronization and fusion processing. S2: Intelligent diagnosis and risk assessment step. Based on the multi-dimensional data fused in step S1, it calls the equipment health assessment model and defect feature library in the central cloud platform. Through comparative analysis and pattern recognition, it diagnoses the pollution level of transformer insulators, overheated parts of connection points and potential discharge defects in real time, and quantitatively assesses their current risk level and evolution trend. S3: Adaptive maintenance strategy generation step. Based on the risk assessment results obtained in step S2, combined with preset maintenance cost constraints, safety procedures and historical operation efficiency data, the optimal live-line maintenance strategy is dynamically generated using an adaptive optimization algorithm. The optimal live-line maintenance strategy includes at least the priority of maintenance operations, target area, recommended cleaning method and expected time window. S4: Resource dynamic scheduling and path planning step. Based on the maintenance strategy generated in step S3, available resources are scheduled, including autonomous mobile cleaning robots, fixed cleaning devices, working periods and energy quotas; based on the three-dimensional structural model of the transformer equipment and real-time operating conditions, motion paths and work sequences are planned for the mobile execution units. S5: Collaborative control and precise execution steps, which send the detailed operation instructions and path information determined in step S4 to the corresponding field operation units and cleaning execution mechanisms; under the collaborative control of the edge computing unit, drive the cleaning execution mechanism to perform precise cleaning of the target area according to the predetermined plan, and at the same time monitor the changes in equipment status and cleaning effect in real time through the sensor network. S6: Operation closed-loop feedback and model optimization steps. After the maintenance operation is completed, collect and analyze the data of the entire operation, including actual energy consumption, time consumption, comparison of state parameters before and after cleaning, and target achievement. Feed the analysis results back to the central cloud platform to update the equipment historical status records and iteratively optimize the parameters of the equipment health assessment model and adaptive optimization algorithm.

2. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 1, characterized in that, In step S1, the deployment of the multi-source sensor network adopts the principle of redundancy configuration, and at least two types of sensors based on different principles are deployed at the monitoring points for cross-validation; the data fusion processing adopts timestamp alignment and spatial coordinate matching technology, and uses the Kalman filter algorithm to reduce noise and estimate dynamic parameters.

3. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 1, characterized in that, The equipment health assessment model in step S2 is a classification and regression hybrid model built on a deep convolutional neural network. The equipment health assessment model is updated by combining offline training and online incremental learning. The defect feature library is constructed from historical defect case data, simulation data, and expert experience rules.

4. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 1, characterized in that, The path planning in step S4 specifically adopts a hybrid algorithm that combines the A search algorithm and the artificial potential field method. The A search algorithm is used to plan the global initial path in the grid map of the device's 3D model, while the artificial potential field method is used to dynamically avoid unforeseen minor obstacles or risk areas caused by electromagnetic field fluctuations based on sensor feedback during real-time execution.

5. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 1, characterized in that, The adaptive optimization algorithm in step S3 is a multi-objective optimization algorithm. The primary objective of the adaptive optimization algorithm is to reduce risk and maximize benefits, while the secondary objective is to minimize the weighted sum of operation cost and time cost. Safety constraints are set, and Pareto optimal solution set is obtained by solving the multi-objective optimization problem. Then, the final execution strategy is selected from the solution set based on real-time resource availability. The decision variables of the algorithm include the start time of the cleaning operation, the selected combination of execution units, the operation intensity of each execution unit and the amount of cleaning agent used. The safety constraints include that the temperature rise of the critical connection points of the equipment must not exceed the threshold during the operation, the minimum safe distance must always be maintained between the operation unit and the live parts, and the operation must not cause the equipment protection system to malfunction.

6. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 5, characterized in that, The adaptive optimization algorithm incorporates current power grid load forecast information and environmental weather forecast data as boundary conditions at each decision, and dynamically adjusts the parameter weights in the optimization model, so that the maintenance strategy can meet both the long-term equipment health maintenance benefits and the short-term power grid operation stability requirements.

7. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 1, characterized in that, The collaborative control in step S5 adopts a master-slave control architecture and an event-driven mechanism. The central cloud platform acts as the master controller and issues macro-level instructions, while the edge computing unit acts as the slave controller and is responsible for micro-motion control and emergency obstacle avoidance. When the sensor detects a sudden change in the equipment status or an abnormality in the working unit in real time, the event-driven mechanism is triggered, and the edge computing unit starts the locally preset safety response program and sends an alarm to the central cloud platform.

8. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 7, characterized in that, The safety response procedure includes at least: immediately suspending the current operation, controlling the operation unit to move to the preset safe standby point, and activating the backup monitoring sensor for verification.

9. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 1, characterized in that, It also includes S7: Human-computer interaction and decision supervision steps, which uses a visual interface to display the process data, diagnostic results and generated strategy suggestions of steps S1 to S3 to the operation and maintenance personnel in real time, and provides interactive interfaces for manual confirmation, parameter fine-tuning or strategy rejection.

10. The intelligent live-line maintenance device and efficient cleaning method for transformers according to claim 1, characterized in that, The central cloud platform adopts a microservice architecture, decoupling and deploying data access services, model analysis services, optimization computing services, and instruction issuance services.

Images