Patrol auxiliary decision-making method based on multi-scene perception

Abstract

The invention relates to a patrol auxiliary decision-making method based on multi-scene perception, and the method comprises the steps: constructing a standard point location label library, mapping the standard point location label library into a patrol point location label library with scene label information, and making a task generation rule under each scene; processing and analyzing the collected data through a deep learning model and scene label information, and identifying a scene where the current transformer substation is located; and generating a patrol task conforming to the scene, automatically executing the patrol task and generating a patrol report, continuously collecting the patrol report, and optimizing the deep learning model through feedback loop. According to the method, the working efficiency of operation and maintenance personnel is improved, the task is automatically created according to the current substation scene, the current substation scene and the real-time state of the patrol host can be quickly analyzed, the optimal patrol plan is generated, and the timeliness and continuity of the patrol task are ensured; according to the real-time environment change, the inspection key point can be rapidly adjusted, the safety of key nodes and lines is preferentially guaranteed, and the continuity and reliability of power supply are ensured.

Description

Technical Field

[0001] The present invention relates to the field of intelligent patrol technology, and in particular to a patrol auxiliary decision-making method based on multi-scenario perception. Background Art

[0002] Smart substation inspection refers to the use of modern information technology and automation to conduct automated and intelligent inspection and maintenance of substations. This inspection method significantly improves the reliability and safety of the power system, while also reducing labor costs and inspection time.

[0003] At present, the formulation of inspection plans often relies on manual experience judgment and routine inspections at fixed periods, resulting in many substations relying solely on repeated and frequent inspections by intelligent inspection equipment in actual applications, causing intelligent inspection equipment such as robots, cameras, and drones to malfunction or wear out. When a substation encounters severe weather or emergencies, the existing inspection plan may no longer meet the inspection needs, and manual updates of the inspection plan are not flexible enough. Especially in large or complex substations, configuring inspection points and inspection tasks is very time-consuming and difficult to achieve rapid response, resulting in potential problems not being handled in a timely manner. Summary of the invention

[0004] In order to solve the problem that the creation of patrol tasks in the prior art requires manual participation, the process is cumbersome and prone to errors, the purpose of the present invention is to provide a patrol decision-making auxiliary method based on multi-scenario perception that can automatically formulate the optimal patrol plan in combination with the current substation scenario, reduce the loss of patrol equipment due to repeated and frequent patrols, and quickly generate or adjust patrol tasks when facing severe weather or emergencies, thereby achieving rapid response.

[0005] To achieve the above object, the present invention adopts the following technical solution: a patrol auxiliary decision method based on multi-scene perception, the method comprising the following steps in order:

[0006] (1) Build a standard point label library, map the standard point label library into a patrol point label library with scene label information, and formulate task generation rules for each scene;

[0007] (2) Collect substation environment information and intelligent patrol system status information in real time, process and analyze the collected data through deep learning models and scene label information, and identify the current scene of the substation;

[0008] (3) Based on the identified current substation scenario and the task generation rules under each scenario, generate patrol tasks that meet the scenario, automatically execute patrol tasks and generate patrol reports, continuously collect patrol reports, and optimize the deep learning model through feedback loops.

[0009] The step (1) specifically includes the following steps in order:

[0010] (1a) Based on the characteristics of patrol points and the correlation between scenes, a standard point label library is constructed. The standard point label library contains patrol point ID, name, coordinates and scene label information;

[0011] (1b) Constructing a patrol point label library based on the punctuation point label library: Mapping the punctuation point label library to a patrol point label library with scene label information through the mapping relationship between standard points and real substations. The mapping relationship includes patrol content, data collection requirements and standard thresholds. The patrol content refers to the equipment status, environmental conditions, fire protection facilities and electrical connections that need to be checked, observed and recorded during the substation patrol. The data collection requirements refer to the specifications and standards followed when collecting the data to be recorded during the patrol process, including accuracy, completeness, timeliness and standardization, to ensure that the collected data is accurate. The standard threshold refers to the critical value used to judge whether the equipment status or data is normal during patrol and data collection.

[0012] (1c) Formulate task generation rules for each scenario: The task generation rules include task name, execution time, patrol point, priority, task type, frequency, patrol terminal, patrol path and patrol content.

[0013] The step (2) specifically refers to: using a temperature sensor, a humidity sensor, a partial discharge monitoring system, and a SF 6 The gas online monitoring device and infrared thermometer monitor the appearance, operating status and changes in the surrounding environment of the equipment in real time, and collect environmental information in the substation and status information of the intelligent patrol system; the deep learning model adopts the ResNeXt model.

[0014] The step (3) specifically includes the following steps in order:

[0015] (3a) According to the identified current substation scene and patrol point tag library, the patrol point information is extracted from the patrol point tag library; according to the matched task generation rules, the patrol tasks that meet the current scene are automatically generated; the patrol tasks include equipment inspection, environmental monitoring and abnormality troubleshooting;

[0016] (3b) assigning the generated inspection tasks to corresponding inspection entities, including inspection robots, drones, and manual inspection teams; during the inspection process, continuously collecting data, including images, videos, temperature, and humidity;

[0017] (3c) Analyze the collected inspection data to obtain analysis results, identify anomalies and potential problems, including equipment failure warnings, environmental change analysis, and equipment operation status assessment; automatically generate an inspection report based on the analysis results, and the inspection report includes an overview of the inspection task, a summary of the inspection data, an analysis of abnormal problems, and recommended measures;

[0018] (3d) Continuously collect patrol reports and feed back task execution results to the system for optimizing the deep learning model. The optimized deep learning model includes updating the scene recognition algorithm, adjusting the task generation rules, and improving the patrol path.

[0019] It can be seen from the above technical scheme that the beneficial effects of the present invention are: first, the present invention improves the work efficiency of operation and maintenance personnel. Automatic creation of tasks according to the current substation scene can quickly analyze the real-time status of the current substation scene and the patrol host, generate the optimal patrol plan, avoid the time consumption of manual planning, and improve efficiency. In addition, the automation system can run uninterruptedly 24 hours a day, ensuring the timeliness and continuity of patrol tasks; second, it improves emergency response capabilities. In the face of natural disasters, extreme weather and other sudden scenarios, it can quickly adjust the patrol focus according to real-time environmental changes, give priority to the safety of key nodes and lines, and ensure the continuity and reliability of power supply. This rapid The response mechanism is crucial to maintaining the stable operation of the power grid in emergency situations; third, it promotes intelligent upgrades. The automation system can not only perform established patrol tasks, but also continuously learn from historical data and self-optimize deep learning models; as time goes by, the system will become more and more "smart", able to more accurately predict equipment status and plan patrol plans more finely, ultimately achieving highly intelligent and autonomous operation and maintenance work, and promoting the power industry to develop in a smarter and greener direction; fourth, the substation automatically generates patrol task strategies based on the current scenario, which not only greatly improves the efficiency and safety of operation and maintenance, but also strengthens the stability and emergency response capabilities of the power grid, laying a solid foundation for the sustainable development of the power industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0021] like Figure 1 As shown, a patrol auxiliary decision method based on multi-scene perception includes the following steps in sequence:

[0022] (1) Build a standard point label library, map the standard point label library into a patrol point label library with scene label information, and formulate task generation rules for each scene;

[0023] (2) Collect substation environment information and intelligent patrol system status information in real time, process and analyze the collected data through deep learning models and scene label information, and identify the current scene of the substation;

[0024] (3) Based on the identified current substation scenario and the task generation rules under each scenario, generate patrol tasks that meet the scenario, automatically execute patrol tasks and generate patrol reports, continuously collect patrol reports, and optimize the deep learning model through feedback loops.

[0025] The step (1) specifically includes the following steps in order:

[0026] (1a) Based on the characteristics of patrol points and the correlation between scenarios, a standard point label library is constructed. The standard point label library contains patrol point ID, name, coordinates and scenario label information; standard point configuration specifications are formulated. The standard point configuration specifications include point naming rules, description information, and location information; based on the general management regulations and rules for substation operation and maintenance, common scenarios are defined, such as peak load, power supply guarantee, weather, and fault tripping.

[0027] (1b) Constructing a patrol point label library based on the punctuation point label library: Mapping the punctuation point label library to a patrol point label library with scene label information through the mapping relationship between the standard points and the real substation. The mapping relationship includes patrol content, data collection requirements and standard thresholds. The patrol content refers to the equipment status, environmental conditions, fire protection facilities and electrical connections that need to be checked, observed and recorded during the substation patrol. The data collection requirements refer to the specifications and standards followed when collecting the data to be recorded during the patrol process, including accuracy, completeness, timeliness and standardization, to ensure that the collected data is accurate and avoid errors and misleading. According to the requirements of the patrol content, the relevant data is fully collected without missing any important information. The data collection is completed within the prescribed patrol cycle to ensure the timeliness of the data. The data is recorded in a unified format and standard to facilitate subsequent analysis and processing. The standard threshold refers to the critical value used to judge whether the equipment status or data is normal during patrol and data collection. These thresholds are determined according to the equipment's operating specifications, safety standards and historical data.

[0028] Ensure that the inspection task can be accurately carried out at each specific inspection point. This step establishes the corresponding relationship between the standard points and the actual inspection points of the substation.

[0029] (1c) Formulate task generation rules for each scenario: The task generation rules include task name, execution time, patrol point, priority, task type, frequency, patrol terminal, patrol path and patrol content. Customizing task generation rules according to different scenarios is to meet the diversity and flexibility of substation patrol. The system can customize the corresponding patrol task generation rules according to different scenarios and needs to ensure the pertinence and effectiveness of patrol tasks.

[0030] The step (2) specifically refers to: using a temperature sensor, a humidity sensor, a partial discharge monitoring system, and a SF 6 The gas online monitoring device and infrared thermometer monitor the appearance, operation status and changes in the surrounding environment of the equipment in real time, and collect environmental information in the substation and the status information of the intelligent patrol system; the deep learning model adopts the ResNeXt model. This model is based on the use of multiple parallel paths (or groups) in the same layer, enabling the network to learn a wider and more diverse set of features. It combines the two basic methods of increasing the depth of the network and increasing the width of the network by increasing the number of groups in each layer, and can use more parameters to achieve better performance.

[0031] The step (3) specifically includes the following steps in order:

[0032] (3a) According to the identified current substation scene and patrol point tag library, the patrol point information is extracted from the patrol point tag library; according to the matched task generation rules, the patrol tasks that meet the current scene are automatically generated; the patrol tasks include equipment inspection, environmental monitoring and abnormality troubleshooting;

[0033] (3b) assigning the generated inspection tasks to corresponding inspection entities, including inspection robots, drones, and manual inspection teams; during the inspection process, continuously collecting data, including images, videos, temperature, and humidity;

[0034] (3c) Analyze the collected inspection data to obtain analysis results, identify anomalies and potential problems, including equipment failure warnings, environmental change analysis, and equipment operation status assessment; automatically generate an inspection report based on the analysis results, and the inspection report includes an overview of the inspection task, a summary of the inspection data, an analysis of abnormal problems, and recommended measures;

[0035] (3d) Continuously collect patrol reports and feed back task execution results to the system for optimizing the deep learning model. The optimized deep learning model includes updating the scene recognition algorithm, adjusting the task generation rules, and improving the patrol path.

[0036] In summary, the present invention improves the work efficiency of operation and maintenance personnel. Automatically creating tasks according to the current substation scene can quickly analyze the current substation scene and the real-time status of the patrol host, generate the optimal patrol plan, avoid the time consumption of manual planning, and improve efficiency. In addition, the automation system can run 24 hours a day without interruption, ensuring the timeliness and continuity of patrol tasks; improving emergency response capabilities, in the face of natural disasters, extreme weather and other sudden scenarios, can quickly adjust the patrol focus according to real-time environmental changes, give priority to the safety of key nodes and lines, and ensure the continuity and reliability of power supply. This rapid response mechanism is very important for maintenance The stable operation of the power grid in emergency situations is crucial; it has promoted intelligent upgrades, and the automation system can not only perform established patrol tasks, but also continuously learn from historical data and self-optimize deep learning models; as time goes by, the system will become more and more "smart", able to more accurately predict equipment status, and plan patrol plans more finely, ultimately achieving highly intelligent and autonomous operation and maintenance work, and promoting the power industry to develop in a smarter and greener direction; the substation automatically generates patrol task strategies based on the current scenario, which not only greatly improves the operation and maintenance efficiency and safety, but also strengthens the stability and emergency response capabilities of the power grid, laying a solid foundation for the sustainable development of the power industry.

Claims

1. A patrol auxiliary decision-making method based on multi-scene perception, characterized by: The method comprises the following steps in order: (1) Build a standard point label library, map the standard point label library into a patrol point label library with scene label information, and formulate task generation rules for each scene; (2) Collect substation environment information and intelligent patrol system status information in real time, process and analyze the collected data through deep learning models and scene label information, and identify the current scene of the substation; (3) Based on the identified current substation scenario and the task generation rules under each scenario, generate patrol tasks that meet the scenario, automatically execute patrol tasks and generate patrol reports, continuously collect patrol reports, and optimize the deep learning model through feedback loops.

2. The remote upgrade method of the intelligent patrol system based on the cascade architecture according to claim 1 is characterized in that: The step (1) specifically includes the following steps in order: (1a) Based on the characteristics of patrol points and the correlation between scenes, a standard point label library is constructed. The standard point label library contains patrol point ID, name, coordinates and scene label information; (1b) Constructing a patrol point label library based on the punctuation point label library: Mapping the punctuation point label library to a patrol point label library with scene label information through the mapping relationship between standard points and real substations. The mapping relationship includes patrol content, data collection requirements and standard thresholds. The patrol content refers to the equipment status, environmental conditions, fire protection facilities and electrical connections that need to be checked, observed and recorded during the substation patrol. The data collection requirements refer to the specifications and standards followed when collecting the data to be recorded during the patrol process, including accuracy, completeness, timeliness and standardization, to ensure that the collected data is accurate. The standard threshold refers to the critical value used to judge whether the equipment status or data is normal during patrol and data collection. (1c) Formulate task generation rules for each scenario: The task generation rules include task name, execution time, patrol point, priority, task type, frequency, patrol terminal, patrol path and patrol content.

3. The remote upgrade method of the intelligent patrol system based on the cascade architecture according to claim 1 is characterized in that: The step (2) specifically refers to: using temperature sensors, humidity sensors, partial discharge monitoring systems, SF6 gas online monitoring devices, and infrared thermometers to monitor the appearance, operating status, and changes in the surrounding environment of the equipment in real time, and to collect environmental information in the substation and status information of the intelligent patrol system; the deep learning model adopts the ResNeXt model.

4. The remote upgrade method of the intelligent patrol system based on the cascade architecture according to claim 1 is characterized in that: The step (3) specifically includes the following steps in order: (3a) According to the identified current substation scene and patrol point tag library, the patrol point information is extracted from the patrol point tag library; according to the matched task generation rules, the patrol tasks that meet the current scene are automatically generated; the patrol tasks include equipment inspection, environmental monitoring and abnormality troubleshooting; (3b) assigning the generated inspection tasks to corresponding inspection entities, including inspection robots, drones, and manual inspection teams; during the inspection process, continuously collecting data, including images, videos, temperature, and humidity; (3c) Analyze the collected patrol data to obtain analysis results and identify anomalies and potential problems, including equipment failure warning, environmental change analysis, and equipment operation status assessment; According to the analysis results, an inspection report is automatically generated, which includes an overview of the inspection tasks, a summary of the inspection data, an analysis of abnormal problems and suggested measures; (3d) Continuously collect patrol reports and feed back task execution results to the system for optimizing the deep learning model. The optimized deep learning model includes updating the scene recognition algorithm, adjusting the task generation rules, and improving the patrol path.

Images