Electric power operation scene management and control method and system based on artificial intelligence algorithm

Abstract

The invention relates to the technical field of intelligent power grids, in particular to an electric power operation scene management and control method and system based on an artificial intelligence algorithm, and aims to solve the problem of how to timely process wrong operation behaviors occurring in a construction process in an electric power operation process so as to ensure standard safety of electric power operation. In order to solve the problem, the invention provides a management and control method which comprises the following steps: deploying sensor equipment according to the implementation content of the electric power operation; collecting working data to obtain a control data set; labeling the management and control data set to obtain an initial management and control model; calculating the detection accuracy; when the detection accuracy is less than the lowest accuracy, optimizing the initial management and control model; when an illegal behavior exists, an early warning signal is sent out; and when the final management and control model cannot identify the working behavior of the electric power operation, marking the working behavior as a newly added behavior, and processing the newly added behavior.

Description

Technical Field

[0001] The technical field of the present invention relates to the field of smart grid technology, and specifically, to a method and system for controlling power operation scenarios based on artificial intelligence algorithms. Background Art

[0002] With the continuous development of society, cables play a vital role in modern society and are widely used in many fields such as electricity, communications and construction. However, there are many safety hazards in the laying, construction and use of cables. These safety hazards not only affect the progress of cable construction, but also easily lead to personal injury or safety accidents. Therefore, the construction safety of power operation scenes is very important. For example, when working in a humid environment, insulation protection measures such as laying insulating pads must be taken to ensure construction safety. However, the traditional method of illegal control of power operations mainly relies on manual supervision and regular inspections, which has many limitations; manual supervision is not only inefficient, but also difficult to achieve real-time and comprehensive monitoring of the operation site, and it is impossible to detect and correct violations in time. Regular inspections have time intervals and cannot stop violations at the moment they occur. It may cause safety hazards to accumulate during this period and increase the risk of accidents. Therefore, how to strengthen the construction safety control of the operation site to improve the efficiency and safety of power operations is one of the problems that technicians in this field urgently need to solve. Summary of the invention

[0003] The problem solved by the present invention is: during the electric power operation, how to promptly handle the erroneous operation behaviors occurring during the construction process to ensure the safety of the electric power operation.

[0004] To solve the above problems, an embodiment of the present invention provides a method for controlling power operation scenes based on an artificial intelligence algorithm, and the method comprises: deploying corresponding sensor equipment for the places, equipment and personnel of the power operation according to the specific implementation content of the power operation; obtaining all construction specifications that need to be complied with by the power operation, conducting tests, collecting the working data of each sensor equipment, and obtaining a control data set; marking the control data set, and training the artificial intelligence model to obtain an initial control model; detecting the corresponding recognition of the behaviors that violate various construction specifications by the initial control model, and calculating the detection accuracy; when the detection accuracy is less than the minimum accuracy, optimizing the initial control model to obtain the final control model; when there is a violation of the rules in the power operation, the sensor sends a warning signal according to the violation; when the final control model cannot identify the working behavior of the power operation, the working behavior is recorded as a new behavior, and the safety officer processes the new behavior according to the stored data of the sensor equipment.

[0005] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By configuring sensor devices for the sites, equipment, and personnel of power operations, the construction process can be controlled in real time. The acquisition of the control data set can clarify the objects and behaviors that need to be controlled during the construction process. By annotating the control data set, the understanding and recognition ability of the artificial intelligence model for data can be improved, and the training efficiency and training effect of the model can be enhanced. By calculating and comparing the detection accuracies of various items, the recognition effects of each construction specification of the initial control model can be intuitively reflected, providing a screening criterion for model optimization and improving the training efficiency of the model. The setting of the audible and visual alarm device can improve the response efficiency when violations occur. When it is impossible to determine whether there are violations, it is handed over to the safety officer for handling, avoiding construction hazards caused by fuzzy recognition and improving the overall construction efficiency and construction safety of power operations.

[0006] In an embodiment of the present invention, according to the specific implementation content of power operations, corresponding sensor devices are deployed for the sites, equipment, and personnel of power operations, specifically including: According to the construction plan of power operations, information such as the construction area, construction equipment, and construction personnel is obtained to obtain operation information; According to the operation information and construction specifications, corresponding standards are set for all control objects to obtain control standards; According to the operation information and control standards, the types and quantities of the sensor devices required for each control object are calculated to obtain a deployment plan; According to the deployment plan, corresponding sensor devices are deployed for all control objects of power operations.

[0007] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By obtaining the work information, the information for subsequent formulation of control standards and determination of deployment plans can be screened out, improving the deployment efficiency. By formulating control standards, the content and specific requirements that need to be controlled can be clarified, providing a standard for subsequent control. By calculating the types and quantities of sensor devices, the operation cost of the control system is reduced. The determination of the deployment plan can help the staff quickly carry out the deployment work and improve the deployment efficiency of the sensor devices.

[0008] In an embodiment of the present invention, all construction specifications that need to be complied with in power operations are obtained, tested, and the working data of each sensor device is collected to obtain a control data set, specifically including: Testing each control object and collecting the data of the sensor device when the control object is operating normally according to the construction specifications to obtain safety data; Obtaining the data of the sensor device when each control object has a violation to obtain violation data; According to the safety data and violation data, they are packaged to obtain a control data set, which is sent to the artificial intelligence model and stored in the database.

[0009] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By obtaining security data and violation data, the training of the artificial intelligence model is more targeted and more applicable to the target power operation scenario, improving the working effect in the actual power operation scenario. By packaging the control data set, the artificial intelligence model can be trained more efficiently, accelerating the training efficiency of the model.

[0010] In an embodiment of the present invention, the control data set is labeled and handed over to the artificial intelligence model for training to obtain an initial control model, which specifically includes: arranging technical personnel to label the valid content in the control data set to obtain the processed data, denoted as the control training set; setting the highest confidence level and the lowest confidence level according to the control standards of each control object to obtain the confidence level standard; training the artificial intelligence model according to the control training set and the confidence level to obtain the initial control model.

[0011] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By labeling the valid content of the control data set, it can help the model better understand the image content, clarify the control object, improve the training efficiency and accuracy of the model. By setting the highest confidence level and the lowest confidence level, the trained artificial intelligence model can be adapted to the data of different types of sensor devices, enhancing the recognition ability of the model and enabling effective recognition of various control objects, strengthening the control effect.

[0012] In an embodiment of the present invention, the corresponding recognition situation of the initial control model for behaviors violating various construction specifications is detected, and the detection accuracy is calculated, which specifically includes: using the test data that has not been recognized to detect the recognition effect of each violation behavior to obtain the detection result; handing over the detection result to the technical personnel for processing to judge whether the detection result is correct to obtain the recognition result; calculating the detection accuracy of the initial control model for each violation behavior according to the recognition result combined with the construction scenario parameters.

[0013] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By calculating the detection accuracy, it can quickly understand the recognition effect of the initial control model for each control object, provide data support for the subsequent optimization of the initial control model, quickly locate the part that needs to be optimized, improve the training efficiency of the model, accelerate the subsequent development of the control work, and further shorten the overall deployment time of the control system. The setting of the construction scenario parameters can also reasonably convert the accuracy of the violation result according to different construction situations.

[0014] In an embodiment of the present invention, if the accuracy of each detection is greater than or equal to the corresponding minimum accuracy, the initial control model is not optimized, and the final control model is directly obtained. If there is a detection accuracy less than the corresponding minimum accuracy, the initial control model is optimized to obtain the final control model, which specifically includes: setting the corresponding minimum accuracy for each control object according to the operation information and control standards; comparing each detection accuracy with the corresponding minimum accuracies to obtain a screening result; when there is a corresponding minimum accuracy greater than the detection accuracy in the screening result, the control object is recorded as an object to be optimized; reprocessing the misidentified data according to the identification result and the object to be optimized to obtain an optimized data set; correcting the identification mode in the initial control model for identifying the object to be optimized according to the optimized data set until the detection accuracy is greater than or equal to the corresponding minimum accuracy; combining the corrected identification mode with the initial control model to obtain the final control model.

[0015] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By comparing each detection accuracy with the minimum accuracy, the control objects whose identification effects need to be optimized can be quickly located, accelerating the training speed of the model. By optimizing the data set and optimizing the initial control model, the overall identification effect of the model can be improved, ensuring that each control object can be effectively controlled during the actual power operation process, improving the control effect of the control system, and further improving the operation efficiency and operation safety of the power operation.

[0016] In an embodiment of the present invention, the final control model is deployed to all sensor devices and connected to the server; when the power operation starts, the data of the sensor devices are identified by the final control model to obtain a credibility; judging according to the credibility and the credibility standard, when the credibility is less than or equal to the minimum credibility, continue to identify and control the power operation; when the credibility is greater than the maximum credibility, obtain the data of the sensor device that identifies the violation behavior, judge the type of the violation behavior, emit a sound and light signal through the sound and light alarm device, and process it according to the construction specification; when the credibility is greater than or equal to the minimum credibility and less than the maximum credibility, record the continuous duration of the credibility within the credibility interval to obtain a fuzzy time; if the fuzzy time is less than the safety time, continue the control, and if the fuzzy time is greater than or equal to the safety time, obtain the recent data of the sensor device and identify and control the power operation according to the recent data.

[0017] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By connecting to the server, centralized management and control of all access sensor devices can be realized, the data transmission efficiency can be improved, which helps the final control model to identify data faster, thus enhancing the recognition speed and control efficiency. Through the sound and light alarm device for alarming, it can handle violations in a timely manner when they occur, improving the response efficiency, reducing the risk of subsequent operation progress being affected due to operation errors or safety accidents caused by untimely warnings, and further improving the operation efficiency and operation safety of power operations. By setting fuzzy time and safety time, it can avoid incorrect responses of the control system caused by short-term recognition anomalies, reducing the workload of safety officers, improving the stability and processing efficiency of the control system, and strengthening the overall control strength of power operation control.

[0018] In an embodiment of the present invention, when a new behavior appears, the real-time image of the power operation and the real-time data of the sensor device are sent to the safety officer, and the safety officer conducts risk control based on the real-time image and real-time data; after the new behavior ends, the safety officer analyzes the new behavior according to the construction specifications and calculates the risk probability corresponding to each new operation of the new behavior; corresponding warning conditions are set for each of the new operations according to the risk probability, and the warning conditions are added to the final control model.

[0019] Compared with the prior art, the technical effects achieved by adopting this technical solution are as follows: By obtaining the new behavior, it can avoid the situation where the final control model fails to effectively supervise when a new power operation behavior appears, improving the safety of power operations. By obtaining the risk probability and warning conditions, the new operations in power operations can be analyzed, and corresponding construction standards can be formulated according to the actual risks, enhancing the rationality and effectiveness of the final control model during control.

[0020] In an embodiment of the present invention, the present invention also provides a power operation scenario control system based on an artificial intelligence algorithm. The control system includes: a storage module, in which a control data set is stored; an acquisition module, which is used to acquire various data obtained by the sensor device; an intelligent module, which is used to train the artificial intelligence model and identify violation behaviors; an execution module, which is used to control the sound and light alarm device and send the recent data of the sensor device to the safety officer. The above-described control method is applied to the control system, and this control system has all the technical features of the above control method, which will not be elaborated here one by one. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is one of the flow diagrams of the power operation scenario control method of the present invention; Figure 2This is the second flowchart of the power operation scenario control method of the present invention; Figure 3 This is the third flowchart of the power operation scenario control method of the present invention; Figure 4 This is the fourth flowchart of the power operation scenario control method of the present invention; Figure 5 This is the control system schematic diagram of the power operation scenario control system of the present invention; Explanation of reference numerals: 100 - Control system; 110 - Storage module; 120 - Acquisition module; 130 - Intelligent module; 140 - Execution module. Detailed implementation manners

[0022] In order to make the above - mentioned objects, features and advantages of the present invention more obvious and understandable, the following detailed description of the specific embodiments of the present invention will be given with reference to the accompanying drawings.

[0023]

First Embodiment

[0024] In step S100, the sensor devices include, but are not limited to, vision sensors, temperature sensors, and displacement sensors, depending on the specific supervision requirements of the power operation. For example, under normal circumstances, vision sensors are deployed at the sites of power operations, temperature sensors and acceleration sensors are deployed for equipment, and displacement sensors are deployed for personnel, with the aim of obtaining various types of required data in real time through various types of sensor devices to achieve real-time control of the construction process of the power operation.

[0025] In step S200, under normal circumstances, the construction specifications include two aspects: operation progress and operation safety. For example, if two construction teams install a high-speed cable simultaneously from both ends, on the one hand, it is necessary to consider whether the daily construction progress of the two construction teams meets the standards and whether the installation route of the cable conforms to the power operation plan. On the other hand, it is necessary to consider whether the construction personnel wear relevant safety equipment and whether they construct in accordance with safety specifications during work.

[0026] It should be noted that in actual power operations, there are usually more aspects of construction specifications and they are not fixed. There are often corresponding special construction specification requirements in special operation environments. For example, when installing cables in a humid environment, construction personnel should be equipped with insulating pads, and when repairing high-voltage wires, workers should wear metal protective clothing, etc., which will not be elaborated here.

[0027] In step S300, since there is a large amount of irrelevant data in the process of sensor devices obtaining working data, which will affect the training of the artificial intelligence model. Therefore, in order to extract the data in the control dataset that can reflect whether the construction process conforms to the construction specifications, the control dataset is labeled. For example, if an image sensor is used to obtain images of the construction site to determine whether construction personnel are operating construction equipment normally, it is necessary to arrange technicians to use means such as the labelme image annotation tool to label the feature points of the images of normal operation of the construction equipment, such as both hands must be placed in fixed positions and the equipment indicator light is red, etc.

[0028] In step S400, in order to ensure that the model can effectively identify behaviors that violate construction specifications, after obtaining the initial control model, it is necessary to test the initial control model, and decide whether to adjust the final control model subsequently according to the detection accuracy of the initial control model for various violation behaviors.

[0029] In step S500, the minimum accuracy refers to the minimum recognition accuracy required to ensure that the model can correctly identify violations in order to meet the control standards. For example, when using image sensors to control the laying of cables, since the image is continuous and the position of the laid cables usually does not change, it is only necessary to set a minimum accuracy of 50% to ensure that the cable can be correctly identified when there is an error in the laying position.

[0030] Furthermore, if it is necessary to identify whether construction workers are wearing aerial work ropes correctly when working at heights, due to the safety risks involved and the posture and position of construction workers constantly changing during the construction process, in order to minimize the risk of safety accidents, a minimum accuracy of 92% needs to be set to ensure that the behavior of incorrectly wearing aerial work ropes can be identified in time.

[0031] In step S600, the sound and light alarm device is deployed in the construction site to ensure that all construction personnel within the construction scope can receive the alarm issued by the sound and light alarm device. When a violation occurs, the alarm is sounded by flashing the alarm light and announcing the specific content of the violation.

[0032] In step S700, due to emergencies and special circumstances in the actual construction process, the data obtained by the sensor equipment may be abnormal or it may be impossible to determine whether there is any violation through the final management and control model. Therefore, it is necessary to arrange a safety officer to manually judge the data that cannot be judged to ensure the safety of the construction.

[0033] For example, when dust appears during the cable laying process, the image acquired by the image sensor device becomes blurred, resulting in the final control model being unable to correctly identify the feature points in the image. It is necessary to obtain nearly ten minutes of video images acquired by the image sensor and send them to the safety officer, who will then go to the construction site for inspection and processing.

[0034] By configuring sensor devices for the sites, equipment and personnel of power operations, the construction process can be controlled in real time. The acquisition of control data sets can clarify the objects and behaviors that need to be controlled during the construction process. Labeling the control data sets can improve the artificial intelligence model's ability to understand and recognize data, improve the model's training efficiency and training effect. By calculating and comparing the accuracy of various tests, the recognition effect of each construction specification of the initial control model can be intuitively reflected, providing screening criteria for model optimization and improving the model's training efficiency. The setting of sound and light alarm devices can improve the response efficiency when violations occur. When it is impossible to determine whether there is a violation, it will be handed over to the safety officer for handling, avoiding construction hazards caused by fuzzy identification and improving the overall construction efficiency and safety of power operations.

[0035]

Second Embodiment

[0036] In step S110, the construction plan refers to the specific implementation plan formulated according to the power operation project, including relevant information such as the personnel composition plan, technical plan, and safety plan. The operation information refers to the relevant information extracted from the construction plan for subsequent formulation of control standards and deployment plans, such as the location and size of the construction area, the quantity and type of construction equipment, and the quantity and identity information of construction personnel.

[0037] In step S120, the controlled object refers to the items that need to be controlled during the construction process of the power operation. Since the content of the construction specifications for different power operations is usually different, it is necessary to clarify all the controlled objects of the power operation according to the specific construction specifications of the power operation and set corresponding control standards.

[0038] It should be noted that the controlled object does not refer to a single operation. A single operation may have multiple controlled objects, and there may also be different standards for the same controlled object of different power operations. For example, a power operation of laying cables requires operating a cable excavator. This operation includes multiple controlled objects such as the operator wearing a safety helmet, the cable excavator traveling in a specified area, and no construction personnel in the blind area of the cable excavator's vision.

[0039] Furthermore, for the controlled object that there should be no construction personnel in the blind area (expandable point) of the cable excavator, since different power operations may use cable excavator equipment of different specifications, it is necessary to determine the corresponding control standards according to the specific situation. For example, for cable excavator equipment of model A, it is necessary to ensure that there are no construction personnel within a radius of 10m in the blind area, and for cable excavator equipment of model B, it is necessary to ensure that there are no construction personnel within 15m in the blind area.

[0040] In step S130, different sensor devices are selected according to different control objects, the number of required sensor devices is determined according to the control requirements, and a plan for deploying each sensor device is determined to obtain a deployment plan. For example, positioning devices are worn by construction workers and cameras are installed in the construction area, etc.

[0041] In step S140, after obtaining the deployment plan, in order to avoid affecting the normal construction of power operations, usually, according to the time required for deployment, the sensor devices of the control objects need to be deployed in advance.

[0042] By obtaining work information, information for subsequent formulation of control standards and determination of deployment plans can be screened out, improving the deployment efficiency. By formulating control standards, the content and specific requirements that need to be controlled can be clarified, providing a standard for subsequent control. By calculating the types and quantities of sensor devices, the operating cost of the control system is reduced. The determination of the deployment plan can help the staff quickly carry out the deployment work and improve the deployment efficiency of the sensor devices.

[0043]

Third Embodiment

[0044] In steps S210 to S230, since training an artificial intelligence model requires a large amount of data, and the increase in the amount of data can improve the accuracy and generalization ability of the model, and since there are differences in the sensor data of different power operation scenarios, in order to ensure that the model adapts to different power operation scenarios, data when operating normally according to the construction specifications and data when performing violation behaviors need to be obtained according to the actual power operation scenario, and at the same time, they are handed over to the artificial intelligence model for training. The amount of data in the control data set depends on specific requirements. For control objects with higher recognition accuracy requirements, usually more relevant safety data and violation data should be collected.

[0045] It should be noted that if the control object cannot be obtained through the actual power operation scenario, similar data is obtained through other conventional means for training. For example, if the control object is whether the cable laying position conforms to the construction plan, since the cable has not been laid and the actual cable laying image in this power operation cannot be obtained, the cable laying images of other similar power operations are used as training data.

[0046] By obtaining safety data and violation data, the training of the artificial intelligence model is more targeted, more applicable to the target power operation scenario, and the working effect in the actual power operation scenario is improved. By packaging the control data set, the artificial intelligence model can be trained more efficiently, and the training efficiency of the model is accelerated.

[0047]

Fourth Embodiment

[0048] In step S310, labeling the valid content means cleaning and processing the data in the control data set, and screening out the data that can reflect the true state of the control object. For example, for the image data of whether the cable is installed at the planned position, technicians need to label the actual position and the planned position of the cable in all images to obtain the control training set.

[0049] In steps S320 to S330, the confidence level reflects the degree of certainty of the model's recognition result after processing the data. The highest confidence level refers to the confidence level requirement for determining that there is a violation behavior of the control object, and the lowest confidence level refers to the confidence level requirement for determining that there is no violation behavior of the control object. Due to the differences in control objects and the data types in the control training set, correspondingly, the recognition difficulties of different control objects also vary. Therefore, it is necessary to set appropriate confidence level standards according to the control standards of the control objects to ensure that the recognition results of different control objects can be correctly obtained subsequently.

[0050] For example, if the control object is whether the working hours of construction workers today exceed 8 hours, it can be directly judged through the timer data, with a relatively low recognition difficulty and no fuzzy results. Therefore, both the highest confidence level and the lowest confidence level are set to 1.0, and there are only two results: recognized as exceeding 8 hours and not exceeding 8 hours. Further, if the control object is whether construction workers are holding construction equipment correctly, usually images are used for recognition, with a relatively large recognition difficulty and fuzzy results. Therefore, the highest confidence level is set to 0.58 and the lowest confidence level is set to 0.05.

[0051] By annotating the valid content of the control data set, it can help the model better understand the image content, clarify the control object, improve the training efficiency and accuracy of the model. By setting the highest confidence level and the lowest confidence level, the trained artificial intelligence model can be adapted to the data of different types of sensor devices, enhancing the recognition ability of the model, enabling effective recognition of various control objects, and strengthening the control effect.

[0052]

Fifth Embodiment

[0053] In step S410, the test data that has not been recognized refers to the data that has not been used for the training of the artificial intelligence model. To ensure that the detection result can reflect the true recognition effect of the initial control model, the test data of each control object needs to be collected separately for detection. At the same time, to ensure the training efficiency of the model and reduce the workload of technical personnel, an appropriate amount of data needs to be selected according to the type of test data. For example, if detecting the recognition effect of the control object in the image, usually, two hundred to three hundred image recognition tests are carried out.

[0054] In steps S420 to S430, according to the credibility criteria, the detection results are divided into three types. When the credibility is less than or equal to the lowest credibility, the detection result is non-violation. When the credibility is greater than the highest credibility, the detection result is violation. When the credibility is greater than or equal to the lowest credibility and less than the highest credibility, the detection result is ambiguous. The construction scenario parameters are comprehensively judged based on the construction weather, the construction experience of the construction personnel, and the construction environment. The maximum value of the construction scenario parameters is 1.05, and the minimum value is 1.

[0055] The correctness of the detection results is judged by technicians to obtain the recognition results. It should be noted that when judging the data with ambiguous detection results, the accuracy is the proportion of violation results in the ambiguous results. For example, if the detection results of 10 test data are ambiguous, and after being judged by technicians, 3 of the test data are found to have violations, then the detection accuracy of the ambiguous result part is 30%.

[0056] It should be noted that since the proportion of violations in the ambiguous results is relatively low, in order to avoid affecting the calculation of the detection accuracy and causing the calculation results to not reflect the true recognition ability of the model, when calculating the detection accuracy, the ambiguous result part is adjusted according to the weight and type of the ambiguous results.

[0057] The calculation formula for the detection accuracy is: A c =(A h +S m *A f +A u )÷3×M. A c is the detection accuracy, A h is the accuracy of the violation results, S m is the adjustment coefficient, A f is the accuracy of the ambiguous results, A u is the accuracy of the non-violation results, and M is the construction scenario parameter.

[0058] For example, in the case of good construction environment and construction weather, the construction scenario parameter corresponding to a regular construction worker is 1. At this time, in the recognition result of the violation behavior that there are other construction workers in the blind area of the construction worker's vision, the accuracy of the violation results is 90%, the accuracy of the ambiguous results is 35%, the determined adjustment coefficient is 2, and the accuracy of the non-violation results is 95%. Then calculate the detection accuracy A c =(90% + 2 * 35% + 95%)÷3×1 = 85%.

[0059] By calculating the detection accuracy, it is possible to quickly understand the recognition effect of the initial control model for each control object, provide data support for the subsequent optimization of the initial control model, quickly locate the parts that need to be optimized, improve the training efficiency of the model, accelerate the subsequent control work, and further shorten the overall deployment time of the control system. The setting of construction scenario parameters can also reasonably convert the accuracy of violation results according to different construction situations.

[0060]

Sixth Embodiment

[0061] In step S510, the minimum accuracy refers to the minimum detection accuracy requirement that the model needs to achieve for each control object in order to meet the progress requirements and safety control requirements of power operations. Usually, based on the principle that the control intensity reaches or exceeds the manual control intensity, it is determined in combination with the actual situation. For example, in a certain power operation, the minimum accuracy for whether the equipment is operating normally is set to 85%, and the minimum accuracy for whether the construction personnel are working in the designated area is set to 93%.

[0062] In steps S520 to S540, compare the detection accuracy and the minimum accuracy for screening. If the detection accuracy of the control object is greater than or equal to the corresponding minimum accuracy, no optimization is required. If the detection accuracy of the control object is less than the corresponding minimum accuracy, optimization is required. The optimization object refers to the control object that needs to retrain the model to improve the detection accuracy.

[0063] In steps S550 to S560, the data with recognition errors of the object to be optimized is sent to relevant technical personnel to analyze the reasons for the recognition errors, and the recognition logic of the model is optimized. The data is re-labeled and processed to obtain an optimized control training set, that is, an optimized data set, which is sent to the initial control model for retraining to meet the minimum accuracy requirement, and a final control model is obtained.

[0064] It should be noted that if the detection accuracy still cannot meet the minimum accuracy requirement after retraining, the above steps need to be repeated to continuously optimize the initial control model until the minimum accuracy requirement is met.

[0065] By comparing the detection accuracies of each item with the minimum accuracy, the control objects whose recognition effects need to be optimized can be quickly located, the training speed of the model can be accelerated, and the initial control model can be optimized through the optimized data set, which can improve the overall recognition effect of the model, ensure that each control object can be effectively controlled during the actual power operation process, improve the control effect of the control system, and further improve the operation efficiency and operation safety of the power operation.

[0066]

Seventh Embodiment

[0067] In steps S610 to S620, the server needs to meet the operation requirements of the final control model and all sensor devices to ensure that the overall control system can be controlled and supervised through the server, and the credibility of each controlled object obtained in real time is transmitted in real time through the server.

[0068] In steps S630 to S640, when the credibility of the controlled object is less than or equal to the minimum credibility, it indicates that the controlled object has committed no violation, and no processing is required. Continue to control the controlled object. When the credibility of the controlled object is greater than the maximum credibility, it indicates that a violation of the controlled object has been identified. Obtain the specific content of the violation, send an alarm through the nearest audible and visual alarm device, and announce the violation behavior by voice until it is re-identified that there is no violation, stop the alarm, and record all information about this violation.

[0069] For example, when construction worker A uses a ground drill to dig a cable trench, it is necessary to ensure that there are no other construction workers within a blind area radius of 1m behind construction worker A. The maximum credibility is 0.6. The image sensor identifies that there is construction worker B within a blind area radius of 1m behind construction worker A, and the credibility is 0.7. Since the credibility is greater than the minimum credibility, an alarm is sent through the nearest audible and visual alarm device, the lights flash, and a voice announcement is made to notify construction worker B to quickly move away from construction worker A. After construction worker B is in a safe position, stop the alarm and record the image data of this alarm, as well as the time, personnel, location, and type of violation of the violation behavior.

[0070] In step S650, the fuzzy time refers to the time length during which the credibility of the controlled object is continuously greater than or equal to the minimum credibility and less than the maximum credibility.

[0071] In step S660, due to the fact that in the actual control operation of the final control model, during the identification process of some sensor devices such as image sensors, it is impossible to ensure continuous and correct identification of the controlled object. The situation where the controlled object cannot be identified within a short period of time does not affect the overall control. Further, in order to avoid an excessive workload for the safety officer and ensure timely handling, a safety time is set. When the fuzzy time is greater than or equal to the safety time, it is determined that recent sensor data, that is, data to be processed, needs to be obtained and handed over to the administrator for judgment and processing.

[0072] By connecting to the server, centralized management and control of all connected sensor devices can be achieved, improving the data transmission efficiency, helping the final control model to identify data faster, enhancing the recognition speed and control efficiency. By using the acoustic and optical alarm device for alarming, timely handling can be carried out when violations occur, improving the response efficiency, reducing the risk of subsequent operation progress being affected due to operation errors or safety accidents caused by untimely warnings, and further improving the operation efficiency and operation safety of power operations. By setting fuzzy time and safety time, false responses of the control system caused by short-term recognition anomalies can be avoided, reducing the workload of safety officers, improving the stability and processing efficiency of the control system, and strengthening the overall control strength of power operation control.

[0073]

Eighth Embodiment

[0074] In steps S710 to S720, in actual power operations, there may be a situation where there is no corresponding control dataset for a new operation behavior, and it has not been trained by the artificial intelligence model, so the final control model cannot control this behavior, that is, a new behavior. Therefore, the safety officer needs to perform manual control through the real-time images of power operations and the real-time data of sensor devices until the new behavior ends. Further, the safety officer analyzes the new behavior, decomposes the new behavior into multiple new actions according to the construction environment and construction standards, and calculates the risk probability according to the construction standards corresponding to the new actions. For example, in a power operation, a new behavior appears, and it is required that the staff operate equipment A within a specified range. Then, this new behavior can be decomposed into multiple new actions such as the usage standards of equipment A and the working position. For the new action of the working position, when the staff is within a radius of 0 to 5 m from the specified location, the risk probability is 0%. When the staff is within a radius of 5 to 6 m from the specified location, the risk probability is 30%. When the staff is within a radius of 6 to 7 m from the specified location, the risk probability is 70%. When the staff is outside a radius of 7 m from the specified location, the risk probability is 100%.

[0075] In step S730, the warning condition refers to the risk probability required for a new behavior to trigger an alarm. After obtaining the warning condition, the warning condition and the new behavior are used as new control objects for training and added to the final control model. For example, for a certain new behavior, it is necessary to ensure that there are no staff within 5 m around equipment B. When there are staff within 5 to 6 m around, the risk probability is 70%. When there are staff within 6 to 7 m around, the risk probability is 40%. When there are staff within 7 to 8 m around, the risk probability is 10%. When there are no staff within a radius of 8 m, the risk probability is 0%. Then, the warning condition is set as the risk probability being greater than 70%, that is, the area within a radius of 6 m around equipment B during operation is prohibited from having staff as the control object for training and added to the final control model.

[0076] By obtaining the new behavior, it is possible to avoid the situation where the final control model cannot effectively supervise when a new power operation behavior appears, improving the safety of power operations. By obtaining the warning condition of the risk probability, it is possible to analyze the new operations in power operations and formulate corresponding construction standards according to the actual risks, enhancing the rationality and effectiveness of the final control model during control.

[0077]

Ninth Embodiment

[0078] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention should be subject to the scope defined by the claims.

Claims

1. A method for controlling power operation scenarios based on artificial intelligence algorithms, characterized in that, The control method includes: According to the specific implementation content of the power operation, deploy corresponding sensor equipment for the location, equipment and personnel of the power operation; Obtain all construction specifications that must be followed for the power operation, collect working data of each of the sensor devices through testing, and obtain a control data set; The control data set is labeled and trained by an artificial intelligence model to obtain an initial control model; Detecting the identification of violations of the construction specifications by the initial control model and calculating the detection accuracy; When the detection accuracy is less than the minimum accuracy, optimizing the initial control model to obtain a final control model; When there is a violation of the electrical operation, the sensor sends out a warning signal according to the violation; When the final control model cannot identify the working behavior of the power operation, the working behavior is recorded as a new behavior, and the safety officer processes the new behavior according to the storage data of the sensor device.

2. The power operation scenario control method based on artificial intelligence algorithm according to claim 1, wherein According to the specific implementation content of the power operation, corresponding sensor equipment is deployed for the place, equipment and personnel of the power operation, specifically including: According to the construction plan of the power operation, information such as the construction area, construction equipment and construction personnel is obtained to obtain operation information; According to the operation information and the construction specifications, control standards are set for all control objects; Determine the type and quantity of the sensor devices required for each of the controlled objects according to the operation information and the control standard, and obtain a deployment plan; According to the deployment plan, corresponding sensor devices are deployed for all the controlled objects of the power operation.

3. The power operation scenario control method based on the artificial intelligence algorithm according to claim 2, wherein The acquisition of all construction specifications that the power operation needs to comply with, and the collection of working data of each of the sensor devices through testing, to obtain a control data set specifically includes: Testing each of the controlled objects, collecting data information of the sensor equipment when the controlled objects meet the construction specifications, and obtaining safety data; Acquire the data of the sensor device when each of the controlled objects has a violation, and obtain violation data; The control data set is packaged according to the security data and the violation data, and the control data set is sent to the artificial intelligence model and stored in a database.

4. The power operation scenario control method based on artificial intelligence algorithm according to claim 3, characterized in that, The labeling of the control data set and training of the artificial intelligence model to obtain the initial control model specifically includes: Arrange technical personnel to mark the valid content in the control data set to obtain processed data, which is recorded as the control training set; According to the control standards of each of the control objects, a maximum credibility and a minimum credibility are set to obtain a credibility standard; The artificial intelligence model is trained according to the control training set and the credibility to obtain the initial control model.

5. The method for controlling power operation scenarios based on artificial intelligence algorithms according to claim 4, characterized in that, The detecting of the identification of the violation of each of the construction specifications by the initial control model and the calculation of the detection accuracy specifically include: Using the unidentified test data, the identification effect of each of the illegal behaviors is tested to obtain the test results; The detection results are handed over to technicians for processing to determine whether the detection results are correct, and recognition results are obtained; Based on the recognition results and combined with construction scenario parameters, calculate the detection accuracy of the initial control model for each of the violation behaviors.

6. The method for controlling a power operation scenario based on an artificial intelligence algorithm according to claim 5, characterized in that, When there is a situation where the detection accuracy is less than the minimum accuracy, optimize the initial control model to obtain the final control model, which specifically includes: According to the operation information and the control standards, set the corresponding minimum accuracy for each of the control objects; Screen the detection accuracies with the corresponding minimum accuracies to obtain a screening result; When there is a situation in the screening result where the corresponding minimum accuracy is greater than the detection accuracy, mark the control object as an object to be optimized; According to the recognition results and the objects to be optimized, reprocess the misrecognized data to obtain an optimized data set; According to the optimized data set, correct the recognition pattern in the initial control model for identifying the objects to be optimized until the detection accuracy is greater than or equal to the corresponding minimum accuracy; Combine the corrected recognition pattern with the initial control model to obtain the final control model.

7. The power operation scenario control method based on the artificial intelligence algorithm according to claim 6, wherein When there is a violation behavior in the power operation, the sensor issues a warning signal according to the violation behavior, which specifically includes: Deploy the final control model to all the sensor devices and connect it to the database; When the power operation starts, identify the data of the sensor devices through the final control model to obtain a credibility; Make a judgment according to the credibility and the credibility standard. When the credibility is less than or equal to the minimum credibility, continue to identify and control the power operation; When the credibility is greater than the maximum credibility, obtain the data of the sensor device that identified the violation behavior, judge the type of the violation behavior, issue a sound and light signal through the sound and light alarm device, and handle it according to the construction specifications; When the credibility is greater than or equal to the minimum credibility and less than the maximum credibility, record the continuous duration of the credibility within the credibility interval to obtain a fuzzy time; If the fuzzy time is less than the safety time, continue the control. If the fuzzy time is greater than or equal to the safety time, obtain the recent data of the sensor device and perform identification and control on the power operation according to the recent data.

8. The method for controlling a power operation scenario based on an artificial intelligence algorithm according to claim 7, wherein, When the final control model cannot identify the working behavior of the power operation, mark the working behavior as a new behavior, and the safety officer processes the new behavior according to the stored data of the sensor device, which specifically includes: When the new behavior appears, send the real-time image of the power operation and the real-time data of the sensor device to the safety officer, and the safety officer performs risk control according to the real-time image and the real-time data; After the new behavior ends, the safety officer analyzes the new behavior according to the construction specifications and calculates the risk probability corresponding to each new operation of the new behavior; Set corresponding warning conditions for each of the new operations according to the risk probability, and add the warning conditions to the final control model.

9. A power operation scenario control system based on artificial intelligence algorithms, characterized in that, The control method according to any one of claims 1 to 8 is applied to the control system, and the control system includes: A storage module, where the control data set is stored in the storage module; An acquisition module, which is used to acquire various data acquired by the sensor device; An intelligent module, which is used to train the artificial intelligence model and identify the violation behaviors; An execution module, which is used to process the new behaviors according to the stored data of the sensor device.

Images