Full space-time patrol method based on multistage cooperation and intelligent decision

Abstract

The invention relates to a full space-time patrol method based on multistage cooperation and intelligent decision making. The method comprises the following steps: collecting multi-source data; constructing a time sequence prediction model, predicting, generating and automatically issuing a substation patrol task, and updating a patrol result after the patrol process is finished; converting patrol data in the patrol result into a multi-modal data vector through an intelligent decision support method, and performing key feature extraction and fusion of the multi-modal data to generate an intelligent decision result; and displaying the intelligent decision result in real time. According to the invention, through multi-stage cooperative processing and intelligent decision making, the fault discovery and response time is greatly shortened, and the omission ratio and the false alarm rate are reduced; intelligent scheduling inspection and task maintenance are carried out, resource waste is effectively avoided, and the operation and maintenance cost benefits are improved; through the multi-dimensional inspection terminal, all-weather and all-dimensional monitoring is achieved, potential safety hazards are found and processed in time, stable operation of a power grid is guaranteed, and a set of complete intelligent solution is provided for operation and maintenance management of the transformer substation.

Description

Technical Field

[0001] The present invention relates to the field of substation operation and maintenance management, and in particular to a full-time and space inspection method based on multi-level collaboration and intelligent decision-making. Background Art

[0002] Traditional substation inspections mainly rely on regular manual inspections. Manual inspections are limited by the experience and skill level of personnel, and it is difficult to ensure the accuracy and consistency of inspection results. Different inspectors may have different judgments on the operating status of the same equipment, resulting in misleading or missing information. With the continuous expansion of the scale of the power grid and the increasing complexity of the grid structure, the number and distribution of substations are also showing a rapid growth trend, which has led to a sharp increase in the workload of manual inspections. The inspection cycle has to be extended, further reducing the timeliness and effectiveness of the inspection. In addition, manual inspections are difficult to achieve comprehensive monitoring of the operating status of substations. Due to the limited number and energy of inspectors, they can often only inspect key equipment and key areas, while ignoring some potential safety hazards, which may lead to failures not being discovered and handled in time in the early stages, leading to greater safety accidents.

[0003] In addition, the existing intelligent patrol technology lacks the ability to make intelligent decisions based on comprehensive data fusion. Due to the lack of effective algorithm and model support, the system can only provide simple data display and alarm functions, but cannot conduct in-depth analysis and prediction of the equipment operation status. This makes it difficult for operation and maintenance personnel to extract valuable information from massive data and make timely and accurate decisions. Summary of the invention

[0004] In order to solve the problems of delayed fault detection response, waste of operation and maintenance resources, difficulty in timely detection and handling of safety hazards, and lack of complete intelligent inspection solutions in the existing substation equipment operation and maintenance technology, the purpose of the present invention is to provide a full-time and space inspection method based on multi-level collaboration and intelligent decision-making, which can achieve optimal allocation and efficient utilization of operation and maintenance resources through intelligent scheduling and inspection strategies, reduce operation and maintenance costs, improve operation and maintenance efficiency, and enhance the operating flexibility and response speed of the power system.

[0005] To achieve the above object, the present invention adopts the following technical solution: a full-time and space patrol method based on multi-level collaboration and intelligent decision-making, the method comprising the following steps in order:

[0006] (1) Collect multi-source data: Use a variety of monitoring equipment to collect data on the substation's internal and external environment and equipment operating status around the clock and in multiple dimensions to obtain multi-source data;

[0007] (2) The collected multi-source data are sorted by timestamp and stored in the substation-level data center for data preprocessing, and a time series prediction model is constructed to predict and generate substation inspection tasks and automatically issue them. After the inspection process is completed, the inspection results are updated;

[0008] (3) Through intelligent decision support methods, the inspection data in the inspection results are converted into multimodal data vectors, and the key features of the multimodal data are extracted and fused to generate intelligent decision results;

[0009] (4) The results of intelligent decision-making are displayed in real time so that operation and maintenance personnel can quickly grasp the overall operating status of the substation, respond to potential risks in a timely manner, and make operation and maintenance decisions.

[0010] The step (1) specifically refers to: the multiple monitoring devices include visible light cameras, infrared cameras, voiceprint collection devices, inspection drones, inspection robots and online monitoring devices, the effective pixels of the visible light camera are greater than 10 million, and the zoom factor is greater than or equal to 2 times; the temperature measurement error of the infrared camera is ±2%; the voiceprint collection device has real-time time-frequency domain map upload; the inspection drone simultaneously provides dual-view fusion and machine nest sensor data upload, which is uploaded to the substation-level data center.

[0011] The step (2) specifically comprises the following steps in order:

[0012] (2a) Various monitoring devices transmit the collected multi-source data to the substation-level data center in real time through the Internet of Things, and store the collected multi-source data in the order of timestamps as historical data. In the substation-level data center, the multi-source data is first pre-processed by the edge computing node, and the pre-processing includes data cleaning, anomaly detection and data fusion to obtain pre-processed data;

[0013] (2b) A time series prediction model is constructed. Historical data and pre-processed multi-source data collected at the current moment are input into the encoder in the time series prediction model. The final state of the substation-level data center is used to initialize the encoder in the time series prediction model. The decoder in the time series prediction model outputs the predicted substation patrol task based on the current real-time observation results on site. The time series construction process is as follows:

[0014]

[0015] Among them, S t , X t They represent the predicted output and actual input of the patrol points at time t respectively; α is the smoothing coefficient, 0≤α≤1; T represents the set of all predicted patrol points, and the set of all predicted patrol points forms the substation patrol task;

[0016] (2c) Perform multi-level collaborative data processing: After the inspection task is generated in step (2b), the inspection task data is saved in the State Grid substation intelligent inspection system, and the State Grid substation remote intelligent inspection system sends the inspection task data to the regional remote intelligent inspection system, the centralized control level remote intelligent inspection centralized monitoring system, and the provincial substation panoramic smart management and control platform. The regional remote intelligent inspection system reviews, executes, and issues the inspection tasks; the data processed by the regional remote intelligent inspection system is uploaded to the centralized control level remote intelligent inspection centralized monitoring system, which identifies potential risks and operation and maintenance requirements and automatically generates inspection results; the inspection results generated by the centralized control level remote intelligent inspection centralized monitoring system are further uploaded to the provincial substation panoramic smart management and control platform, and the provincial substation panoramic smart management and control platform reviews, confirms, and updates the inspection results.

[0017] The step (3) specifically comprises the following steps in order:

[0018] (3a) Multimodal classification of inspection results: Obtain the inspection results generated in step (2), classify the inspection results according to the type of monitoring equipment, and obtain three types of multimodal data: image type, audio type, and text type;

[0019] (3b) Perform key feature extraction to extract the key features of multimodal data and uniformly convert them into numerical feature vectors:

[0020] For text types, they are displayed in text serialization form. The representation of sequence t is:

[0021] t=c1,c2,…,cj,…,cm

[0022] Among them, cj is the jth word in the text sequence. The word2vec function is used to model the sequence t and obtain the vector A of the sequence t:

[0023] a=word2vec(t)

[0024] For the image type, use image2vec to vectorize and get the vector representation P of image p:

[0025] P = image2vec(p)

[0026] For the audio type, use sound2vec to vectorize it and get the vector representation V of the voiceprint v:

[0027] V = sound2vec(v)

[0028] (3c) Perform key feature fusion, input vector A, vector P and vector V into the encoder, obtain single-modal vector representations αt, αp and αv, and calculate the multi-modal joint representation α:

[0029]

[0030] (3d) Calculate the empirical state evaluation health level Y based on the multimodal joint representation α:

[0031] Y=F(α)

[0032] The empirical status evaluation health level Y includes level I, level II, level III, level IV, and level V. The health statuses corresponding to level I, level II, level III, level IV, and level V are: excellent, good, average, poor, and very poor, respectively.

[0033] The step (4) specifically refers to: constructing a visualization display layer to provide an intuitive user interface, wherein the visualization display layer includes a three-dimensional simulation model, dynamic reports and report management; wherein the three-dimensional simulation model is constructed based on the actual layout and equipment information of the substation, and can display equipment status and patrol path information in real time; the dynamic reports include equipment status statistical reports, fault analysis reports, and maintenance record reports; the report management includes the generation and export of intelligent decision-making result reports, the automatic generation and export of patrol results, and the generation and export of multi-source data acquisition reports.

[0034] It can be seen from the above technical scheme that the beneficial effects of the present invention are as follows: First, the present invention realizes real-time monitoring and accurate analysis of the operation status of the power grid through multi-level collaborative processing and intelligent decision-making. When an abnormal situation occurs in the power grid, the present invention can quickly identify and trigger the corresponding response mechanism, significantly shortening the fault discovery and response time. This efficient fault handling mode not only greatly reduces the missed detection rate and false alarm rate, effectively avoids larger-scale safety accidents caused by the spread of faults, but also fundamentally guarantees the safety and stability of the power grid operation; Second, the present invention is aimed at intelligent scheduling inspection and task maintenance, and through intelligent scheduling and inspection strategies, it realizes the optimal configuration and efficient utilization of operation and maintenance resources, and can automatically adjust the inspection plan and task allocation according to actual needs and power grid status, avoiding excessive consumption and waste of resources. This intelligent operation and maintenance mode not only reduces the operation and maintenance cost and improves the operation and maintenance efficiency, but also improves the operation flexibility and response speed of the power system to a certain extent; Third, the present invention timely discovers and handles potential safety hazards through all-weather and all-round monitoring. At the same time, the application of multi-dimensional inspection terminals also greatly reduces the workload of operation and maintenance personnel and improves work efficiency and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0036] like Figure 1As shown, a full-time and space inspection method based on multi-level collaboration and intelligent decision-making includes the following steps in sequence:

[0037] (1) Collect multi-source data: Use a variety of monitoring equipment to collect data on the substation's internal and external environment and equipment operating status around the clock and in multiple dimensions to obtain multi-source data;

[0038] (2) The collected multi-source data are sorted by timestamp and stored in the substation-level data center for data preprocessing, and a time series prediction model is constructed to predict and generate substation inspection tasks and automatically issue them. After the inspection process is completed, the inspection results are updated;

[0039] (3) Through intelligent decision support methods, the inspection data in the inspection results are converted into multimodal data vectors, and the key features of the multimodal data are extracted and fused to generate intelligent decision results;

[0040] (4) The results of intelligent decision-making are displayed in real time so that operation and maintenance personnel can quickly grasp the overall operating status of the substation, respond to potential risks in a timely manner, and make operation and maintenance decisions.

[0041] The step (1) specifically refers to: the multiple monitoring devices include visible light cameras, infrared cameras, voiceprint collection devices, inspection drones, inspection robots and online monitoring devices, the effective pixels of the visible light camera are greater than 10 million, and the zoom factor is greater than or equal to 2 times; the temperature measurement error of the infrared camera is ±2%; the voiceprint collection device has real-time time-frequency domain map upload; the inspection drone simultaneously provides dual-view fusion and machine nest sensor data upload, which is uploaded to the substation-level data center.

[0042] The step (2) specifically comprises the following steps in order:

[0043] (2a) Various monitoring devices transmit the collected multi-source data to the substation-level data center in real time through the Internet of Things, and store the collected multi-source data in the order of timestamps as historical data. In the substation-level data center, the multi-source data is first pre-processed by the edge computing node. The pre-processing includes data cleaning, anomaly detection and data fusion to reduce the amount of data transmission and improve processing efficiency, thereby obtaining pre-processed data;

[0044] (2b) A time series prediction model is constructed. Historical data and pre-processed multi-source data collected at the current moment are input into the encoder in the time series prediction model. The final state of the substation-level data center is used to initialize the encoder in the time series prediction model. The decoder in the time series prediction model outputs the predicted substation patrol task based on the current real-time observation results on site. The time series construction process is as follows:

[0045]

[0046] Among them, S t , X t They represent the predicted output and actual input of the patrol points at time t respectively; α is the smoothing coefficient, 0≤α≤1; T represents the set of all predicted patrol points, and the set of all predicted patrol points forms the substation patrol task;

[0047] (2c) Perform multi-level collaborative data processing: After the inspection task is generated in step (2b), the inspection task data is saved in the State Grid substation intelligent inspection system, and the State Grid substation remote intelligent inspection system sends the inspection task data to the regional remote intelligent inspection system, the centralized control level remote intelligent inspection centralized monitoring system, and the provincial substation panoramic smart management and control platform. The regional remote intelligent inspection system reviews, executes, and issues the inspection tasks; the data processed by the regional remote intelligent inspection system is uploaded to the centralized control level remote intelligent inspection centralized monitoring system, which identifies potential risks and operation and maintenance requirements and automatically generates inspection results; the inspection results generated by the centralized control level remote intelligent inspection centralized monitoring system are further uploaded to the provincial substation panoramic smart management and control platform, and the provincial substation panoramic smart management and control platform reviews, confirms, and updates the inspection results.

[0048] The step (3) specifically comprises the following steps in order:

[0049] (3a) Multimodal classification of inspection results: Obtain the inspection results generated in step (2), classify the inspection results according to the type of monitoring equipment, and obtain three types of multimodal data: image type, audio type, and text type;

[0050] (3b) Perform key feature extraction to extract the key features of multimodal data and uniformly convert them into numerical feature vectors:

[0051] For text types, they are displayed in text serialization form. The representation of sequence t is:

[0052] t=c1,c2,…,cj,…,cm

[0053] Among them, cj is the jth word in the text sequence. The word2vec function is used to model the sequence t and obtain the vector A of the sequence t:

[0054] A = word2vec(t)

[0055] For the image type, use image2vec to vectorize and get the vector representation P of image p:

[0056] P = image2vec(p)

[0057] For the audio type, use sound2vec to vectorize it and get the vector representation V of the voiceprint v:

[0058] V = sound2vec(v)

[0059] (3c) Perform key feature fusion, input vector A, vector P and vector V into the encoder, obtain single-modal vector representations αt, αp and αv, and calculate the multi-modal joint representation α:

[0060]

[0061] (3d) Calculate the empirical state evaluation health level Y based on the multimodal joint representation α:

[0062] Y=F(α)

[0063] The experience status evaluation health level Y includes level I, level II, level III, level IV, and level V. The health status corresponding to level I, level II, level III, level IV, and level V are: excellent, good, average, poor, and very poor, respectively.

[0064] The step (4) specifically refers to: constructing a visualization display layer to provide an intuitive user interface, wherein the visualization display layer includes a three-dimensional simulation model, dynamic reports and report management; wherein the three-dimensional simulation model is constructed based on the actual layout and equipment information of the substation, and can display equipment status and patrol path information in real time; the dynamic reports include equipment status statistical reports, fault analysis reports, and maintenance record reports; the report management includes the generation and export of intelligent decision-making result reports, the automatic generation and export of patrol results, and the generation and export of multi-source data acquisition reports.

[0065] In summary, the present invention greatly shortens the fault detection and response time and reduces the missed detection rate and false alarm rate through multi-level collaborative processing and intelligent decision-making; performs intelligent scheduling inspections and task maintenance to effectively avoid waste of resources and improve the cost-effectiveness of operation and maintenance; uses multi-dimensional inspection terminals to monitor around the clock and in all directions, timely discovers and handles potential safety hazards, and ensures stable operation of the power grid; and provides a complete set of intelligent solutions for substation operation and maintenance management through multi-level collaborative and multi-system linkage.

Claims

1. A full-time and space inspection method based on multi-level collaboration and intelligent decision-making, characterized by: The method comprises the following steps in order: (1) Collect multi-source data: Use a variety of monitoring equipment to collect data on the substation's internal and external environment and equipment operating status around the clock and in multiple dimensions to obtain multi-source data; (2) The collected multi-source data are sorted by timestamp and stored in the substation-level data center for data preprocessing, and a time series prediction model is constructed to predict and generate substation inspection tasks and automatically issue them. After the inspection process is completed, the inspection results are updated; (3) Through intelligent decision support methods, the inspection data in the inspection results are converted into multimodal data vectors, and the key features of the multimodal data are extracted and fused to generate intelligent decision results; (4) The results of intelligent decision-making are displayed in real time so that operation and maintenance personnel can quickly grasp the overall operating status of the substation, respond to potential risks in a timely manner, and make operation and maintenance decisions.

2. The full-time and space inspection method based on multi-level collaboration and intelligent decision-making according to claim 1 is characterized by: The step (1) specifically refers to: the multiple monitoring devices include visible light cameras, infrared cameras, voiceprint collection devices, inspection drones, inspection robots and online monitoring devices, the effective pixels of the visible light camera are greater than 10 million, and the zoom factor is greater than or equal to 2 times; the temperature measurement error of the infrared camera is ±2%; the voiceprint collection device has real-time time-frequency domain map upload; the inspection drone simultaneously provides dual-view fusion and machine nest sensor data upload, which is uploaded to the substation-level data center.

3. The full-time and space inspection method based on multi-level collaboration and intelligent decision-making according to claim 1 is characterized by: The step (2) specifically comprises the following steps in order: (2a) Various monitoring devices transmit the collected multi-source data to the substation-level data center in real time through the Internet of Things, and store the collected multi-source data in the order of timestamps as historical data. In the substation-level data center, the multi-source data is first pre-processed by the edge computing node, and the pre-processing includes data cleaning, anomaly detection and data fusion to obtain pre-processed data; (2b) A time series prediction model is constructed. Historical data and pre-processed multi-source data collected at the current moment are input into the encoder in the time series prediction model. The final state of the substation-level data center is used to initialize the encoder in the time series prediction model. The decoder in the time series prediction model outputs the predicted substation patrol task based on the current real-time observation results on site. The time series construction process is as follows: Among them, S t , X t They represent the predicted output and actual input of the patrol points at time t respectively; α is the smoothing coefficient, 0≤α≤1; T represents the set of all predicted patrol points, and the set of all predicted patrol points forms the substation patrol task; (2c) Perform multi-level collaborative data processing: After the inspection task is generated in step (2b), the inspection task data is saved in the State Grid substation intelligent inspection system, and the State Grid substation remote intelligent inspection system sends the inspection task data to the regional remote intelligent inspection system, the centralized control level remote intelligent inspection centralized monitoring system, and the provincial substation panoramic smart management and control platform. The regional remote intelligent inspection system reviews, executes, and issues the inspection tasks; the data processed by the regional remote intelligent inspection system is uploaded to the centralized control level remote intelligent inspection centralized monitoring system, which identifies potential risks and operation and maintenance requirements and automatically generates inspection results; the inspection results generated by the centralized control level remote intelligent inspection centralized monitoring system are further uploaded to the provincial substation panoramic smart management and control platform, and the provincial substation panoramic smart management and control platform reviews, confirms, and updates the inspection results.

4. The full-time and space inspection method based on multi-level collaboration and intelligent decision-making according to claim 1 is characterized by: The step (3) specifically comprises the following steps in order: (3a) Multimodal classification of inspection results: Obtain the inspection results generated in step (2), classify the inspection results according to the type of monitoring equipment, and obtain three types of multimodal data: image type, audio type, and text type; (3b) Perform key feature extraction to extract the key features of multimodal data and uniformly convert them into numerical feature vectors: For text types, they are displayed in text serialization form. The representation of sequence t is: t=c1,c2,…,cj,…,cm Among them, cj is the jth word in the text sequence. The word2vec function is used to model the sequence t and obtain the vector A of the sequence t: A=word2vec(t) For the image type, use image2vec to vectorize and get the vector representation P of image p: P = image2vec(p) For the audio type, use sound2vec to vectorize it and get the vector representation V of the voiceprint v: V = sound2vec(v) (3c) Perform key feature fusion, input vector A, vector P and vector V into the encoder, obtain single-modal vector representations αt, αp and αv, and calculate the multi-modal joint representation α: (3d) Calculate the empirical state evaluation health level Y based on the multimodal joint representation α: Y=F(α) The experience status evaluation health level Y includes level I, level II, level III, level IV, and level V. The health statuses corresponding to level I, level II, level III, level IV, and level V are: excellent, good, average, poor, and very poor.

5. The full-time and space inspection method based on multi-level collaboration and intelligent decision-making according to claim 1 is characterized by: The step (4) specifically refers to: constructing a visualization display layer to provide an intuitive user interface, wherein the visualization display layer includes a three-dimensional simulation model, dynamic reports and report management; wherein the three-dimensional simulation model is constructed based on the actual layout and equipment information of the substation, and can display equipment status and patrol path information in real time; the dynamic reports include equipment status statistical reports, fault analysis reports, and maintenance record reports; the report management includes the generation and export of intelligent decision-making result reports, the automatic generation and export of patrol results, and the generation and export of multi-source data acquisition reports.

Images