A real-time monitoring method and system for power engineering

Abstract

The application provides a real-time monitoring method and system for electric power engineering, and relates to the technical field of data processing, which constructs the relative correlation between the equipment and the construction task and the twin identification model based on the equipment data and the construction task data of the electric power engineering, performs image collection based on the continuous monitoring track node, generates an image set; input the image set into the twin identification model to obtain a feature set; perform feature set similarity evaluation through a discrimination subnetwork to obtain initial warning information; generate a track correlation factor based on the track image set; generate real-time warning information based on the track correlation factor and the initial warning information. The technical problem that the existing technology relies on the manual experience of construction personnel for electric power engineering construction risk identification and avoidance, resulting in insufficient timeliness and effectiveness of electric power construction risk warning and risk elimination, is solved. The technical effects of improving the accuracy and immediacy of electric power engineering risk identification and improving the safety of electric power engineering construction are achieved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and specifically to a real-time monitoring method and system for power engineering. Background Technology

[0002] Because current hazard identification and avoidance in power engineering construction relies on the personal experience of construction workers, this reliance on human experience is easily affected by the subjective abilities and cognitive limitations of construction workers. The quality and completeness of individual experience also vary, leading to inconsistencies in safety risk assessment. This may result in the neglect or misjudgment of potential hazards, causing significant deficiencies in the timeliness of hazard warnings and the effectiveness of hazard elimination in power construction, thereby increasing the risks in power engineering construction.

[0003] In summary, the current reliance on the manual experience of construction workers for hazard identification and avoidance in power engineering construction leads to technical problems such as insufficient timeliness of hazard warnings and inadequate effectiveness of hazard elimination. Summary of the Invention

[0004] This application provides a real-time monitoring method and system for power engineering, which addresses the technical problem that existing technologies rely on the manual experience of construction personnel for identifying and avoiding construction hazards, resulting in insufficient timeliness of hazard warnings and effectiveness of hazard elimination in power construction.

[0005] In view of the above problems, this application provides a real-time monitoring method and system for power engineering.

[0006] The first aspect of this application provides a real-time monitoring method for power engineering projects. The method includes: collecting construction task data of the power engineering project and calling equipment data corresponding to the construction task data, wherein the equipment data includes equipment model, equipment location, and equipment application; associating and mapping the equipment data with the construction task data to construct a dynamic-static relative association between the equipment and the construction task; performing big data control matching based on the equipment data to construct a control image set, wherein each image in the control image set has an attention distribution identifier; building a twin recognition model based on the control image set, wherein the twin recognition model includes a first recognition subnetwork and a second recognition subnetwork; reading real-time construction tasks and identifying the dynamic-static status of equipment based on the real-time construction tasks and the dynamic-static relative association, and classifying the data accordingly. Continuous monitoring trajectory nodes are deployed; image acquisition devices are configured through these continuous monitoring trajectory nodes to acquire device control images and device trajectory images, generating a real-time control image set and a trajectory image set; the real-time control image set and the control image set are respectively input into the first recognition subnetwork and the second recognition subnetwork, outputting a first feature set and a second feature set; the discriminative subnetwork of the twin recognition model performs similarity evaluation on the first feature set and the second feature set based on the attention distribution identifier, outputting a similarity evaluation image, and generating initial warning information based on the similarity evaluation image; the trajectory image set is used to perform device trajectory warning recognition, generating trajectory correlation factors; the trajectory correlation factors and the initial warning information are used to perform superimposed evaluation of warnings, generating real-time warning information.

[0007] A second aspect of this application provides a real-time monitoring system for power engineering projects. The system includes: a data acquisition and retrieval module for acquiring construction task data of the power engineering project and retrieving equipment data corresponding to the construction task data, wherein the equipment data includes equipment model, equipment location, and equipment application; a relative association construction module for mapping the equipment data to the construction task data to construct a dynamic-static relative association between the equipment and the construction task; an image set construction module for performing large-scale control matching based on the equipment data to construct a control image set, wherein each image in the control image set has an attention distribution identifier; a twin model construction module for building a twin recognition model based on the control image set, wherein the twin recognition model includes a first recognition subnetwork and a second recognition subnetwork; and a monitoring node distribution module for reading real-time construction tasks and identifying the dynamic-static status of equipment based on the real-time construction tasks and the dynamic-static relative association, and then distributing the data accordingly. The system comprises: a continuous monitoring trajectory node; an image acquisition execution module, configured with an image acquisition device through the continuous monitoring trajectory node, and executing the acquisition of device control images and device trajectory images to generate a real-time control image set and a trajectory image set; a feature recognition output module, inputting the real-time control image set and the control image set into the first recognition sub-network and the second recognition sub-network respectively, and outputting a first feature set and a second feature set; a similarity evaluation execution module, using the discriminant sub-network of the twin recognition model to perform similarity evaluation on the first feature set and the second feature set based on the attention distribution identifier, outputting a similarity evaluation image, and generating initial warning information based on the similarity evaluation image; a trajectory warning recognition module, used to perform trajectory warning recognition on the trajectory image set and generate trajectory correlation factors; and a warning overlay evaluation module, used to perform overlay evaluation of warnings using the trajectory correlation factors and the initial warning information to generate real-time warning information.

[0008] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0009] The method provided in this application embodiment acquires construction task data of a power engineering project and calls the corresponding equipment data, wherein the equipment data includes equipment model, equipment location, and equipment application; it associates and maps the equipment data with the construction task data to construct a dynamic-static relative association between equipment and construction tasks; it performs big data control matching based on the equipment data to construct a control image set, wherein each image in the control image set has an attention distribution identifier. By setting the attention distribution identifier for each feature region of the control image, a reference is provided for the computing power distribution of subsequent similarity comparison of various feature regions of the same control image; a twin recognition model is built based on the control image set, wherein the twin recognition model includes a first recognition sub-network and a second recognition sub-network; it reads real-time construction tasks and, based on the real-time construction tasks and... The dynamic and static correlation is used to identify the dynamic and static status of equipment and distribute continuous monitoring trajectory nodes. Image acquisition devices are configured through these continuous monitoring trajectory nodes to acquire equipment control images and equipment trajectory images, generating a real-time control image set and a trajectory image set. The real-time control image set and the control image set are respectively input into the first and second recognition sub-networks, outputting a first feature set and a second feature set. The discriminative sub-network of the twin recognition model performs similarity evaluation on the first and second feature sets based on the attention distribution identifier, outputting a similarity evaluation image and generating initial warning information based on the similarity evaluation image. The trajectory image set is used for equipment trajectory warning identification, generating trajectory correlation factors. The trajectory correlation factors and the initial warning information are used for superimposed evaluation of the warnings to generate real-time warning information. This achieves the technical effect of improving the accuracy and timeliness of hazard identification in power engineering and enhancing the safety of power engineering construction. Attached Figure Description

[0010] Figure 1 This application provides a schematic diagram of a real-time monitoring method for power engineering.

[0011] Figure 2 This is a schematic diagram of the process for generating image attention distribution markers in a real-time monitoring method for power engineering provided in this application.

[0012] Figure 3 This is a flowchart illustrating the process of generating trajectory correlation factors in a real-time monitoring method for power engineering provided in this application.

[0013] Figure 4 This application provides a schematic diagram of the structure of a real-time monitoring system for power engineering.

[0014] Figure labeling: 1. Data acquisition and calling module; 2. Relative association construction module; 3. Image set construction module; 4. Twin model construction module; 5. Monitoring node distribution module; 6. Image acquisition execution module; 7. Feature recognition output module; 8. Similarity evaluation execution module; 9. Trajectory early warning recognition module; 10. Early warning overlay evaluation module. Detailed Implementation

[0015] This application provides a real-time monitoring method and system for power engineering projects, addressing the technical problem that existing technologies rely on the manual experience of construction workers for hazard identification and avoidance in power engineering construction, resulting in insufficient timeliness of hazard warnings and effectiveness of hazard elimination. It achieves the technical effect of improving the accuracy and timeliness of hazard identification in power engineering projects, thereby enhancing the safety of power engineering construction.

[0016] The acquisition, storage, use, and processing of data in the technical solution of this invention all comply with relevant regulations.

[0017] The technical solutions of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. It should be understood that the present invention is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, not all of them.

[0018] Example 1, as Figure 1 As shown, this application provides a real-time monitoring method for power engineering, the method comprising:

[0019] S100: Collect and obtain construction task data of power engineering, and call up the equipment data corresponding to the construction task data, wherein the equipment data includes equipment model, equipment location and equipment application.

[0020] S200: Associate and map the equipment data with the construction task data to establish a dynamic-static relative association between the equipment and the construction task.

[0021] Specifically, in this embodiment, the power engineering project includes, but is not limited to, the installation and connection of power equipment such as generator sets, substations, transformers, switchgear, and cable lines, as well as power construction activities such as the excavation of underground cable trenches, cable laying, and installation of connection boxes. The construction task data divides the overall power engineering project into multiple days of construction tasks. By correctly executing several construction tasks in the construction task data on time and in sequence, the power engineering project can be completed within the construction time limit.

[0022] The construction task data includes several construction tasks corresponding to several construction dates. Based on the different construction tasks, the equipment data for performing the construction tasks will vary. For example, the equipment model, quantity of the same model, and equipment layout on the construction site required for the underground cable trench excavation construction task will be different from the equipment model, quantity, and equipment layout on the construction site required for the cable laying construction task.

[0023] Based on this, this embodiment calls the construction task data to obtain equipment data for several construction tasks corresponding to several construction dates. The equipment data includes equipment model, equipment location, and equipment application. The equipment data in a construction task includes multiple sets of equipment application-equipment model-equipment location. The equipment model represents the specific purpose name and model parameters of the equipment. The equipment location is the location information of the power construction equipment at the construction site. The equipment application is used to distinguish multiple power engineering equipment of the same equipment model in a construction task.

[0024] It should be understood that in a construction task, there are equipment such as excavators and material transport vehicles that move around the construction site, as well as equipment such as cement mixing plants that are fixed in a location on the construction site. Therefore, when the equipment is mobile and can move around on the construction site, the equipment location is the area where the equipment moves, and when the equipment is a facility that is fixed in a location on the construction site, the equipment location is the stationary location.

[0025] Based on whether the equipment location in the equipment data is a stationary position or a moving area, the equipment in several construction tasks in the construction task data is associated and mapped to construct a dynamic-static relative association between the equipment and the construction task, thereby obtaining the dynamic-static status of the equipment in each of the several construction tasks in the construction task data. It should be understood that the dynamic-static association relationship between the same equipment and the construction task may differ in different construction tasks.

[0026] S300: Perform big data control matching based on the device data to construct a control image set, wherein each image in the control image set has an attention distribution identifier.

[0027] In one embodiment, such as Figure 2 As shown, the method steps provided in this application further include:

[0028] S310: Perform image classification of the control image set according to the control pattern, and generate an initial classification result with control pattern identifier.

[0029] S320: Perform weight distribution evaluation of in-image features under the initial classification result, and generate weight distribution evaluation results.

[0030] S330: Perform weighted similarity aggregation based on the weighted distribution evaluation results, and generate attention distribution labels for the image based on the similarity aggregation results.

[0031] Specifically, in this embodiment, several sets of equipment model information are extracted from several equipment data of several construction tasks in the construction task data, and multiple sets of equipment use control images corresponding to multiple equipment models are obtained by traversing big data matching based on several sets of equipment model information.

[0032] Based on multiple sets of equipment usage control images corresponding to various equipment models, and combined with several sets of construction tasks from the construction task data, the control image set is constructed. The control image set includes several subsets of control images corresponding to several equipment data for several construction tasks, and several equipment usage control images in each control image subset have equipment model identifiers.

[0033] By extracting control images from the control image set based on the equipment model identifier, a set of equipment control images corresponding to several equipment data for several construction tasks in the construction task data can be obtained.

[0034] In this embodiment, each device obtains several control images based on big data matching. These control images include both images with incorrect control actions and images with correct control actions. Therefore, in this embodiment, the device categorizes the control image set into image classes corresponding to the control modes performed, generating initial classification results with control mode identifiers. These control mode identifiers include two types: correct control behavior and incorrect control behavior. In this embodiment, image classification for control modes is preferably performed manually to reduce the economic cost of image classification for control modes. Alternatively, it can be based on building an image recognition model to improve recognition efficiency.

[0035] After classifying images based on control modes, this embodiment obtains the initial classification result. The initial classification result is a subset of control images corresponding to a subset of equipment data for a subset of construction tasks in the construction task data. Each control image in the subset of control images of each equipment carries a control mode identifier that indicates whether the control is correct or not.

[0036] It should be understood that, in this embodiment, the control image can be a driver's cab image taken from the driver's perspective, and the image composition features include the steering wheel, driver's arm, driver's fingers, dashboard, windshield, rearview mirror, and sun visor. Therefore, each control image contains background area features that lack control recognition capabilities, such as the windshield, rearview mirror, and sun visor; low-reference area features with weaker control recognition effectiveness, such as the driver's arm and steering wheel; and high-reference area features with higher control recognition effectiveness, such as the driver's fingers and dashboard.

[0037] Therefore, in this embodiment, the effectiveness of each region feature of the control image under the initial classification result in control recognition is evaluated by weight distribution, and a weight distribution evaluation result is generated. The weight distribution evaluation is preferably based on the expert evaluation method. After setting the control recognition effectiveness level for each region feature of the control image, the weight evaluation of the control recognition effectiveness of each region feature is performed by multiple experts in the field, thereby obtaining several groups of region weight distribution evaluation results for the same region feature division given by multiple experts. Further, the weight similarity aggregation is performed based on the weight distribution value of each region feature to obtain the weight distribution value with the highest frequency among multiple weight distribution values ​​of each region feature, which is used as the weight distribution value of that region, thereby generating the similarity aggregation result. The similarity aggregation result includes weight distribution data of several regions. Based on the similarity aggregation result, the attention distribution of each feature region is performed, thereby generating the attention distribution label of the image. The attention distribution label is used to divide the image according to feature regions and to mark the attention of feature regions. For example, if the weight distribution value of a certain region is 30%, then the attention label of that region is 30%.

[0038] This embodiment achieves the technical effect of providing a reference for the computing power distribution of similarity comparison of various feature regions of the same control image by setting the attention distribution identifier for each feature region of the control image.

[0039] S400: A twin recognition model is built based on the control image set, wherein the twin recognition model includes a first recognition subnetwork and a second recognition subnetwork.

[0040] Specifically, it should be understood that the twin recognition model is used to identify the similarity between two images. In this embodiment, the twin recognition model is used to identify the similarity between the real-time control image of the equipment being controlled by unspecified personnel at the power engineering construction site and the control image, thereby identifying and determining whether there are any errors in the equipment control behavior.

[0041] The twin recognition model consists of an input layer, a recognition layer, and an output layer. The recognition layer includes a first recognition subnetwork, a second recognition subnetwork, and a discriminant subnetwork. The image recognition object of the first recognition subnetwork is the real-time control image generated during the power engineering construction process. The image recognition object of the second recognition subnetwork is the control image. The output of the first recognition subnetwork is a first image feature set, and the output of the second recognition subnetwork is a second image feature set. The first image feature set and the second image feature set are used to calculate feature similarity based on the discriminant subnetwork, thereby outputting the similarity between the real-time control image and the control image based on the output layer.

[0042] Furthermore, based on whether the control mode of the current input control image is correct or incorrect, and the similarity between the real-time control image and the control image, it is determined whether the current real-time operation is correct or not.

[0043] In this embodiment, the recognition sub-network uses existing image feature extraction methods to extract image features from the real-time control image.

[0044] S500: Reads real-time construction tasks, identifies equipment movement and statics based on the real-time construction tasks and the relative correlation between movement and statics, and distributes continuous monitoring trajectory nodes.

[0045] In this embodiment, the current time node is obtained, and the construction task data is traversed based on the current time node to obtain the real-time construction task corresponding to the current time node. It should be understood that the real-time construction task is a non-specific construction task in the construction task data.

[0046] Based on the obtained real-time construction task, the dynamic and static identification of several devices in the real-time construction task is further performed according to the real-time construction task and the dynamic-static relative association. For example, the real-time construction task includes K devices. Among the K devices, T devices are statically associated with the real-time construction task, and M devices are dynamically associated with the real-time construction task, where T+M=K.

[0047] A first continuous monitoring trajectory node is distributed among the T devices, and a second continuous trajectory monitoring node is distributed among the M devices. The first continuous trajectory monitoring node is a periodic monitoring node that uses a single image acquisition device in the control room to acquire control images, for example, every 15 seconds. The second continuous trajectory monitoring node is a periodic monitoring node that uses both the image acquisition device in the control room and the image acquisition device of the construction site drone to acquire images of the movement trajectory of the equipment in the control room and at the construction site, for example, every 15 seconds.

[0048] This embodiment sets different image acquisition and monitoring trajectory nodes for the equipment based on the relative correlation between the equipment's static and dynamic states, thereby providing a technical reference for the effective acquisition of control images of various types of equipment at the construction site.

[0049] S600: Configure the image acquisition device through the continuous monitoring trajectory node, and perform equipment control image and equipment trajectory image acquisition to generate a real-time control image set and a trajectory image set.

[0050] In one embodiment, the method steps provided in this application further include:

[0051] S610: Historical abnormal data of human-computer interaction, and generate initial human-computer interaction data based on the historical abnormal data.

[0052] S620: Build a device operation profile for the user, and generate a stable correlation coefficient for the user based on the device operation profile.

[0053] S630: Generate collection constraints using the initial number of human-organizational connections and the stable correlation coefficient of users.

[0054] S640: Control the image acquisition of the image acquisition device according to the acquisition constraints.

[0055] Specifically, in this embodiment, the T devices are distributed with first continuous monitoring trajectory nodes, and the M devices are distributed with second continuous trajectory monitoring nodes. Image acquisition devices are configured to acquire device control images and device trajectory images based on the configured image acquisition devices, generating T sets of real-time control images for the T devices, and generating M sets of real-time control images and M sets of trajectory images for the M devices. The M sets of trajectory images represent the driving trajectories of the M devices at the construction site.

[0056] To improve the scientific validity of setting the interval period for the first and second continuous trajectory monitoring nodes, in this embodiment, after setting the interval period for the first and second continuous trajectory monitoring nodes, the interval period is further adjusted adaptively according to the frequency of safety accidents occurring for the users controlling the equipment. This allows for the configuration of higher frequency connection trajectory monitoring nodes for equipment of users with high frequency of safety accidents, and higher frequency connection trajectory monitoring nodes for equipment of users with low frequency of safety accidents.

[0057] The specific method for optimizing the interval period of the connection trajectory monitoring nodes in the equipment configuration is as follows:

[0058] The historical anomaly data refers to records of control incidents that occur during the operation of unspecified devices by unspecified device control users. The historical anomaly data for the human-machine interface includes multiple sets of historical device anomalies and historical operating users. Based on the historical operating users and device models, the historical anomaly data is summarized and integrated to generate an initial human-machine interface contact number, which consists of multiple sets of operating users, device models, and incident frequencies.

[0059] Based on the types of devices a user can operate, a device operation profile is constructed for each user. This profile consists of multiple user-device model pairs. A stable correlation coefficient is generated based on the device models a user can operate. For example, if the device operation profile indicates that a user can operate three different device models, then that user's stable correlation coefficient is 3.

[0060] K users who operate and control K devices in the real-time construction task are obtained. Further, the initial human-machine interface contact number is traversed according to the username to obtain the K failure frequencies of the K devices controlled by the K users in history, which are used as the K initial human-machine interface contact numbers. The user stable correlation coefficients are traversed according to the username to obtain the K stable correlation coefficients of the K users.

[0061] Based on the initial number of human-organizational connections and the stable correlation coefficient, a normalized product is calculated to obtain K collection constraints.

[0062] A preset acquisition constraint adjustment threshold is set. The acquisition constraint adjustment threshold is used to determine whether an adaptive adjustment of the interval period is needed. If the actual calculated acquisition constraint is higher than the acquisition constraint adjustment threshold, the multiple by which the acquisition constraint is higher than the acquisition constraint adjustment threshold is calculated. Based on the multiple multiplied by the original interval period, the adaptive adjustment of the interval period is completed.

[0063] Conversely, if the actual calculated acquisition constraint is lower than the acquisition constraint adjustment threshold, the original interval period is used. In this embodiment, the acquisition constraint adjustment threshold is not numerically limited and can be manually set according to the nature of the construction.

[0064] The continuous monitoring trajectory nodes of the image acquisition device are optimized based on the acquisition constraint adjustment threshold and the acquisition constraint control.

[0065] This embodiment adapts and optimizes the continuous monitoring trajectory nodes based on the diversity of operators' equipment operation skills and the frequency of failures of unspecified equipment in the past. This enables frequent image acquisition and analysis for users with low operation skills, ensuring the effectiveness of construction process monitoring and the timeliness of equipment operation and construction detection.

[0066] S700: Input the real-time control image set and the control image set into the first recognition subnetwork and the second recognition subnetwork respectively, and output the first feature set and the second feature set.

[0067] In one embodiment, before inputting the real-time control image set and the control image set into the first recognition subnetwork and the second recognition subnetwork respectively, the method step S700 provided in this application further includes:

[0068] S711: Determine the subject authentication region of each image in the control image set based on the attention distribution identifier.

[0069] S712: Input the subject authentication region corresponding to the control image set and the real-time control image set into the fuzzy matching network, wherein the fuzzy matching network is a sub-network of the twin recognition model.

[0070] S713: Output fuzzy matching results, and execute the input constraints of the real-time control image set and the control image set based on the fuzzy matching results.

[0071] Specifically, in this embodiment, based on the equipment data of the real-time construction task, a set of K control images for K devices is obtained, and further, the attention distribution identifier of the K device control images is obtained. As can be seen from the foregoing, the attention distribution identifier is an identifier for the attention allocation value of each feature region of a control image.

[0072] Based on the correspondence between the attention distribution identifier and the feature region, the subject authentication region of each image in the control image set is determined. The subject authentication region is the region with the highest attention allocation value in the attention distribution identifier.

[0073] The fuzzy matching network is a sub-network of the twin recognition model. The function of the fuzzy matching network is to divide and delineate the image region corresponding to the subject authentication region in the real-time control image with reference to the subject authentication region, so as to provide an indication of the image extraction focus for subsequent image feature extraction based on the first recognition sub-network and the second recognition sub-network.

[0074] Therefore, in this embodiment, the fuzzy matching network is set between the input layer and the recognition layer of the twin recognition model. The main authentication region corresponding to the control image set and the real-time control image set are input into the fuzzy matching network, and the fuzzy matching result is output. The fuzzy matching result is multiple sets of control images and real-time images, and the real-time images are marked with the main image authentication region division mark.

[0075] The real-time control image set and the input constraints of the control image set in the first recognition subnetwork and the second recognition subnetwork are executed based on the fuzzy matching result.

[0076] This embodiment constructs a fuzzy matching network, which enables the division and identification of key regions for image feature extraction before image feature extraction based on the recognition subnetwork. This indirectly improves the effectiveness of image feature extraction and the accuracy of image similarity comparison results based on the image feature extraction results.

[0077] S800: The discriminant subnetwork of the twin recognition model performs similarity evaluation on the first feature set and the second feature set based on the attention distribution identifier, outputs a similarity evaluation image, and generates initial warning information based on the similarity evaluation image.

[0078] Specifically, based on step S300, the discriminant subnetwork is used to calculate the feature similarity between the first image feature set and the second image feature set, thereby outputting the similarity between the real-time control image and the control image based on the output layer.

[0079] The first feature set consists of image features of the main authentication region and image features of the non-main authentication region, and the second feature set consists of image features of the main authentication region and image features of the non-main authentication region, wherein the image features of all authentication regions have attention distribution values ​​for the corresponding regions.

[0080] The discriminant subnetwork of the twin recognition model performs similarity evaluation on the first feature set and the second feature set based on the attention distribution identifier (attention distribution values ​​of multiple regions), and outputs a similarity evaluation image. The similarity evaluation image is obtained by assigning the image similarity identifier to the corresponding real-time control image.

[0081] It should be understood that this embodiment performs image similarity evaluation based on existing technology and image features. It only multiplies the obtained regional image similarity evaluation result by the corresponding attention distribution value to obtain the similarity result between the real-time control image and the control image with the attention distribution identifier as a weighted reference.

[0082] In this embodiment, the similarity evaluation process of the twin recognition model is as follows:

[0083] Randomly obtain the first real-time control image set of the first device among the K devices and the corresponding first control image set, wherein the first control image set includes multiple control images with control mode identifiers.

[0084] The first real-time control image from the first real-time control image set is input into the twin recognition model for multiple rounds of similarity evaluation with several control images in the first control image set, thereby obtaining several similar evaluation images. These similar evaluation images are then sorted, and the control image in the first control image set corresponding to the highest similarity value is obtained. The control mode identifier of this control image is used as the first initial warning result of the first real-time control image. The same method is used to obtain the initial warning results of K devices, constituting the initial warning information.

[0085] This embodiment achieves the technical effect of quickly verifying and determining the initial early warning information of the correct or incorrect control of all equipment in the equipment data of real-time construction tasks based on image-based similarity comparison calculation between real-time acquired images and control images.

[0086] S900: Perform device trajectory warning recognition on the trajectory image set and generate trajectory association factors.

[0087] In one embodiment, such as Figure 3 As shown, the method steps provided in this application further include:

[0088] S910: Perform initial image acquisition for the real-time construction task before construction to construct the initial construction scene.

[0089] S920: Perform data update of the initial construction scenario based on the device's dynamic and static identifiers, and generate trajectory constraints based on the update results.

[0090] S930: Based on the trajectory constraints, perform trajectory warning recognition on the trajectory image set and generate the trajectory association factor.

[0091] Specifically, before executing the real-time construction task, the K devices in the real-time construction task have been deployed at the construction site. In this embodiment, the initial image of the construction site from a top-down angle is acquired before executing the real-time construction task to complete the construction of the initial construction scene.

[0092] Based on the static-relative association of T devices with the real-time construction task and the dynamic-relative association of M devices with the real-time construction task, the data of K devices in the initial construction scenario is updated using the static-dynamic relative association, thereby completing the identification of the static-dynamic relative association of K devices in the initial construction scenario.

[0093] Furthermore, based on the travel trajectories of the M devices at the construction site and the radius range of possible travel deviations of the M devices on their travel trajectories, trajectory constraints for the M devices are generated.

[0094] Based on the set of trajectory images of M devices, M driving trajectories of the M devices at the construction site are obtained.

[0095] Calculate the M trajectory constraint areas of the M devices, and perform trajectory overlap on the M driving trajectories based on the corresponding comparison of the trajectory constraints of the M devices to obtain the trajectory deviation areas of the M devices. Then, calculate the area ratio of the M trajectory deviation areas to the M trajectory constraint areas to generate the trajectory association factor. The trajectory association factor includes M trajectory association factors. The more severely the driving trajectory exceeds the trajectory constraints, the larger the value of the corresponding trajectory association factor.

[0096] This embodiment achieves the technical effect of obtaining a trajectory correlation factor that characterizes the degree of deviation of the driving trajectory of the dynamic relative associated equipment from the original trajectory in real-time construction tasks by performing trajectory deviation verification, thereby indirectly improving the technical effect of improving the accuracy of motion trajectory deviation analysis of dynamic equipment.

[0097] S1000: Real-time warning information is generated by superimposing and evaluating the warning based on the trajectory correlation factor and the initial warning information.

[0098] In one embodiment, the method steps provided in this application further include:

[0099] S1010: Determine the execution status of the device based on the real-time early warning information and generate a task execution delay result.

[0100] S1020: Configure the task tolerance period of the real-time construction task according to the construction task data.

[0101] S1030: Based on the task tolerance period and the result of the task execution delay, replan the task execution.

[0102] S1040: Implement construction control for power engineering projects based on the results of the replanning.

[0103] In one embodiment, the method steps provided in this application further include:

[0104] S1110: Construct an early warning response instruction library.

[0105] S1120: Match the real-time early warning information to the early warning response instruction library.

[0106] S1130: Perform early warning processing based on the response matching result, and record the feedback identifier of the real-time early warning information.

[0107] S1110: Control feedback for real-time monitoring based on the feedback identifier.

[0108] Specifically, in this embodiment, the superposition evaluation involves superimposing the equipment operation analysis results with the equipment travel analysis results to obtain real-time early warning information that reflects whether several pieces of equipment have operational defects in the real-time construction task of power engineering from two perspectives: whether the equipment operation is correct and whether the equipment travel is normal.

[0109] Specifically, the real-time early warning information includes T initial early warning results from T devices and M sets of initial early warning results – trajectory correlation factors – from M devices. Based on this real-time early warning information, it is possible to intuitively determine whether the operation of all equipment in the real-time construction task is correct and whether the travel trajectory deviates, thereby determining whether all equipment poses a danger in power engineering construction.

[0110] This embodiment achieves the technical effect of improving the accuracy and timeliness of hazard identification in power engineering and enhancing the safety of power engineering construction.

[0111] Furthermore, in this embodiment, the method for determining the execution status of the device and generating the task execution delay result based on the real-time early warning information is as follows: a first device is randomly extracted from the K devices. If the first device is statically associated with the real-time construction task, the initial early warning result of the first device is obtained. If the control mode identifier of the initial early warning result is incorrect, the execution status determination result of the first device is suspended execution.

[0112] Conversely, if the first device is dynamically associated with the real-time construction task, the initial warning result and trajectory association factor of the first device are obtained, and a preset trajectory association threshold is set. When the trajectory association factor exceeds the preset trajectory association threshold, it is considered that the driving direction control system of the first device is faulty. When the initial warning result and trajectory association factor show an error in the control mode or the trajectory association factor exceeds the preset trajectory association threshold, the execution status of the first device is considered to be suspended.

[0113] Using the same method as obtaining the execution status determination result of the first device, the execution status determination results of K devices are obtained, among which W devices are in a state of pending maintenance. Then, the maintenance time of W devices is obtained as W, and the longest maintenance time is obtained by serializing the W maintenance times, which is used as the result of the task execution delay.

[0114] The allowable grace period for the overall power engineering task is determined based on the construction task data, and this allowable grace period is used as the task tolerance period for the real-time construction task.

[0115] Determine whether the task tolerance period is greater than the task execution delay result. If the task tolerance period is greater than the task execution delay result, then replan the task execution. The replanning is to carry out emergency repairs on W devices.

[0116] If the task tolerance period is less than the task execution delay result, then the task execution is replanned. This replanning involves replacing the healthy devices of W equipment. Based on the replanning result, construction control of the real-time construction task in the power engineering is executed.

[0117] This embodiment achieves the technical effect of generating a replanning result that ensures the construction task has a low degree of impact on the overall power engineering construction by analyzing equipment maintenance time and the actual given task extension grace period.

[0118] Furthermore, this embodiment constructs an early warning response instruction library, which includes the equipment maintenance priorities for K equipment failures in the real-time construction task. The equipment maintenance priorities represent the importance of the equipment in the real-time construction task, and the equipment maintenance priorities can be set based on the work experience of power engineering construction personnel.

[0119] The real-time early warning information is used to extract W devices with faults that need to be repaired. Based on the W devices, the early warning response instruction library is matched to obtain the W fault repair priorities of the W devices, which constitute the response matching result.

[0120] Based on the W fault repair priorities in the response matching results, fault warnings are processed for W devices one by one, and the repair results of each of the W devices are recorded as feedback identifiers. Control feedback for real-time monitoring of the corresponding devices is then performed based on these feedback identifiers. This real-time monitoring occurs when all equipment faults are eliminated during the real-time construction task, and the control feedback verifies whether the faulty equipment has been successfully repaired.

[0121] This embodiment achieves orderly operation and maintenance of faulty equipment by prioritizing fault elimination in real-time construction tasks, thereby improving the effectiveness and orderliness of operation and maintenance.

[0122] Example 2, based on the same inventive concept as the real-time monitoring method for power engineering in the foregoing examples, such as... Figure 4 As shown, this application provides a real-time monitoring system for power engineering, wherein the system includes:

[0123] The data acquisition and retrieval module 1 is used to acquire construction task data of the power project and retrieve the equipment data corresponding to the construction task data. The equipment data includes equipment model, equipment location and equipment application.

[0124] The relative association construction module 2 is used to associate and map the equipment data with the construction task data to construct a dynamic and static relative association between the equipment and the construction task.

[0125] Image set construction module 3 is used to perform big data control matching based on the device data to construct a control image set, wherein each image in the control image set has an attention distribution identifier.

[0126] The twin model construction module 4 is used to build a twin recognition model based on the control image set, wherein the twin recognition model includes a first recognition sub-network and a second recognition sub-network.

[0127] The monitoring node distribution module 5 is used to read real-time construction tasks, identify equipment movement and statics based on the real-time construction tasks and the relative correlation between movement and statics, and distribute continuous monitoring trajectory nodes.

[0128] The image acquisition execution module 6 is used to configure the image acquisition device through the continuous monitoring trajectory node, and to perform the acquisition of device control images and device trajectory images, generating a set of real-time control images and a set of trajectory images.

[0129] The feature recognition output module 7 is used to input the real-time control image set and the control image set into the first recognition sub-network and the second recognition sub-network respectively, and output the first feature set and the second feature set.

[0130] The similarity evaluation execution module 8 is used to perform similarity evaluation of the first feature set and the second feature set based on the attention distribution identifier through the discriminant sub-network of the twin recognition model, output a similarity evaluation image, and generate initial warning information based on the similarity evaluation image.

[0131] The trajectory warning and recognition module 9 is used to perform trajectory warning and recognition of the device on the trajectory image set and generate trajectory correlation factors.

[0132] The early warning overlay evaluation module 10 is used to perform overlay evaluation of early warnings using the trajectory correlation factor and the initial early warning information, and generate real-time early warning information.

[0133] In one embodiment, the system further includes:

[0134] The image classification execution unit is used to perform image classification of the control image set according to the control pattern and generate an initial classification result with the control pattern identifier.

[0135] The weight distribution evaluation unit is used to perform weight distribution evaluation of the in-image features under the initial classification result and generate weight distribution evaluation results.

[0136] The weighted similarity aggregation unit is used to perform weighted similarity aggregation based on the weight distribution evaluation results, and generate attention distribution labels for the image based on the similarity aggregation results.

[0137] In one embodiment, the system further includes:

[0138] The authentication region identification unit is used to determine the main authentication region of each image in the control image set based on the attention distribution identifier.

[0139] An information input execution unit is used to input the subject authentication region corresponding to the control image set and the real-time control image set into a fuzzy matching network, wherein the fuzzy matching network is a sub-network of the twin recognition model.

[0140] The matching result output unit is used to output fuzzy matching results and execute the input constraints of the real-time control image set and the control image set based on the fuzzy matching results.

[0141] In one embodiment, the system further includes:

[0142] The initial image acquisition unit is used to acquire initial images of the real-time construction task before construction and to construct the initial construction scene.

[0143] The data update execution unit is used to perform data updates for the initial construction scenario based on the device's dynamic and static identifiers, and to generate trajectory constraints based on the update results.

[0144] The trajectory warning and recognition unit is used to perform trajectory warning and recognition of the device on the trajectory image set based on the trajectory constraints, and generate the trajectory association factor.

[0145] In one embodiment, the system further includes:

[0146] The execution status determination unit is used to determine the execution status of the device based on the real-time early warning information and generate a task execution delay result.

[0147] The tolerance period configuration unit is used to configure the task tolerance period of the real-time construction task according to the construction task data.

[0148] The task execution planning unit is used to replan the task execution based on the task tolerance period and the result of the task execution delay.

[0149] The construction control unit is used to execute construction control of power engineering projects based on the results of the replanning.

[0150] In one embodiment, the system further includes:

[0151] An abnormal data interaction unit is used to interact with historical abnormal data of the human-machine interface and generate an initial number of human-machine connections based on the historical abnormal data.

[0152] The correlation coefficient generation unit is used to construct a device operation profile for the user and generate a stable correlation coefficient for the user based on the device operation profile.

[0153] The data collection constraint generation unit is used to generate data collection constraints using the initial number of human-organizational connections and the stable correlation coefficient of users.

[0154] An image acquisition execution unit is used to control the image acquisition of the image acquisition device according to the acquisition constraints.

[0155] In one embodiment, the system further includes:

[0156] The early warning instruction library building unit is used to build an early warning response instruction library.

[0157] The response matching execution unit is used to perform response matching on the warning response instruction library based on the real-time warning information.

[0158] The early warning processing execution unit is used to perform early warning processing based on the response matching result and record the feedback identifier of the real-time early warning information.

[0159] The monitoring and control feedback unit is used for real-time monitoring and control feedback based on the feedback identifier.

[0160] In summary, any of the methods or steps described above can be stored as computer instructions or programs in various types of computer memory, and the computer instructions or programs can be recognized by various types of computer processors to implement any of the above methods or steps.

[0161] Based on the above specific embodiments of the present invention, any improvements and modifications made to the present invention by those skilled in the art without departing from the principle of the present invention shall fall within the patent protection scope of the present invention.

Claims

1. A real-time monitoring method for power engineering, characterized in that, The method includes: The system collects and obtains construction task data for power engineering projects, and retrieves the corresponding equipment data, which includes equipment model, equipment location, and equipment application. The equipment data is associated and mapped with the construction task data to establish a dynamic and static relative association between the equipment and the construction task; Based on the device data, big data control matching is performed to construct a control image set, wherein each image in the control image set has an attention distribution identifier; A twin recognition model is built based on the control image set, wherein the twin recognition model includes a first recognition subnetwork and a second recognition subnetwork; Read real-time construction tasks, identify equipment movement and statics based on the real-time construction tasks and the relative correlation between movement and statics, and distribute continuous monitoring trajectory nodes; The image acquisition device is configured through the continuously monitored trajectory nodes, and the acquisition of equipment control images and equipment trajectory images is performed to generate a set of real-time control images and a set of trajectory images. The real-time control image set and the control image set are respectively input into the first recognition subnetwork and the second recognition subnetwork, and the first feature set and the second feature set are output. The discriminant subnetwork of the twin recognition model performs similarity evaluation on the first feature set and the second feature set based on the attention distribution identifier, outputs a similarity evaluation image, and generates initial warning information based on the similarity evaluation image; The device performs trajectory warning recognition on the trajectory image set and generates trajectory correlation factors; Real-time warning information is generated by superimposing and evaluating the warnings using the trajectory correlation factors and the initial warning information.

2. The method as described in claim 1, characterized in that, The method further includes: The control image set is classified according to control patterns to generate an initial classification result with control pattern identifiers; Perform a weight distribution evaluation of the image features under the initial classification result, and generate a weight distribution evaluation result; Weighted similarity aggregation is performed based on the weighted distribution evaluation results, and attention distribution labels for the image are generated based on the similarity aggregation results.

3. The method as described in claim 2, characterized in that, Before inputting the real-time control image set and the control image set into the first recognition subnetwork and the second recognition subnetwork respectively, the method further includes: The subject authentication region of each image in the control image set is determined based on the attention distribution identifier; The subject authentication region corresponding to the control image set and the real-time control image set are input into the fuzzy matching network, wherein the fuzzy matching network is a sub-network of the twin recognition model; Output fuzzy matching results, and execute the input constraints of the real-time control image set and the control image set based on the fuzzy matching results.

4. The method as described in claim 1, characterized in that, The method further includes: The real-time construction task is performed by acquiring initial images before construction to construct an initial construction scene; The initial construction scenario data is updated based on the device's static / dynamic indicators, and trajectory constraints are generated based on the update results. Based on the trajectory constraints, the device performs trajectory warning recognition on the trajectory image set and generates the trajectory association factor.

5. The method as described in claim 1, characterized in that, The method further includes: Based on the real-time early warning information, the execution status of the device is determined, and a task execution delay result is generated; Configure the task tolerance period for the real-time construction task based on the construction task data; Based on the task tolerance period and the result of the task execution delay, the task execution is replanned; Construction control of the power project will be implemented based on the results of the replanning.

6. The method as described in claim 1, characterized in that, The method further includes: Historical abnormal data of human-computer interaction, and generate an initial number of human-computer interaction connections based on the historical abnormal data; Create device operation profiles for users, and generate stable correlation coefficients for users based on the device operation profiles; Data collection constraints are generated using the initial number of human-organizational contacts and the stable user correlation coefficient. The image acquisition device is controlled to acquire images based on the acquisition constraints.

7. The method as described in claim 1, characterized in that, The method further includes: Build an early warning and response instruction library; The real-time early warning information is used to match the early warning response command library; Based on the response matching results, an early warning is issued, and the feedback identifier of the real-time early warning information is recorded. Control feedback is monitored in real time based on the feedback identifier.

8. A real-time monitoring system for power engineering, characterized in that, The system includes: The data acquisition and retrieval module is used to acquire construction task data of power engineering and retrieve the equipment data corresponding to the construction task data, wherein the equipment data includes equipment model, equipment location and equipment application; The relative association construction module is used to associate and map the equipment data with the construction task data to construct a dynamic and static relative association between the equipment and the construction task; An image set construction module is used to perform big data control matching based on the device data and construct a control image set, wherein each image in the control image set has an attention distribution identifier; A twin model construction module is used to build a twin recognition model based on the control image set, wherein the twin recognition model includes a first recognition subnetwork and a second recognition subnetwork; The monitoring node distribution module is used to read real-time construction tasks, identify equipment movement and statics based on the real-time construction tasks and the relative correlation between movement and statics, and distribute continuous monitoring trajectory nodes. The image acquisition execution module is used to configure the image acquisition device through the continuous monitoring trajectory nodes, and to perform the acquisition of device control images and device trajectory images, generating a real-time control image set and a trajectory image set; The feature recognition output module is used to input the real-time control image set and the control image set into the first recognition subnetwork and the second recognition subnetwork respectively, and output the first feature set and the second feature set; The similarity evaluation execution module is used to evaluate the similarity between the first feature set and the second feature set based on the attention distribution identifier through the discriminant sub-network of the twin recognition model, output a similarity evaluation image, and generate initial warning information based on the similarity evaluation image; The trajectory warning and recognition module is used to perform trajectory warning and recognition on the trajectory image set and generate trajectory correlation factors; The early warning overlay evaluation module is used to perform overlay evaluation of early warnings using the trajectory correlation factors and the initial early warning information, and generate real-time early warning information.

Images