Automatic non-inductive transmission control system for image data of transformer substation

Abstract

The invention relates to the field of automatic control, in particular to a transformer substation image data automatic non-inductive transmission control system which comprises an image acquisition module, an intelligent video analysis module, an automatic transmission control module, a safety and privacy protection module and a non-inductive interaction design module. According to the scheme, the image information is deeply analyzed by adopting the multi-scale Flow model, the requirements of image anomaly detection and pixel-level anomaly positioning are distinguished, a multi-scale aggregation strategy is flexibly applied, and the anomaly recognition precision and efficiency are improved; an optimized self-adaptive data packet scheduling method is adopted to ensure continuity and stability of video and image data in a transmission process, and an improved intelligent optimizer algorithm is used to optimize parameter vectors of the optimized self-adaptive data packet scheduling method, so that the parameter convergence speed is accelerated, the reliability of data transmission is enhanced, and the transmission efficiency is improved. And smooth and unobstructed user experience is brought, and existing network resources are utilized to the maximum extent.

Description

Technical Field

[0001] The present invention relates to the field of automatic control, and in particular to an automatic and sensorless transmission control system for substation image data. Background Art

[0002] In recent years, with the deepening reform of power system management and the development of automation technology, many substations have been unmanned, which has promoted the image monitoring system as an important supplement to unmanned substation monitoring. The amount of image data generated by substation monitoring is huge, which makes it more difficult to quickly and accurately identify anomalies from massive images, which may delay the system response time and thus weaken the real-time monitoring efficiency; especially in remote and harsh areas, the instability of network connection will further interfere with the continuous transmission of data, which constitutes another obstacle; in addition, the transmission of high-definition images greatly occupies bandwidth, and the limited network resources may become a bottleneck restricting the overall performance of the system. Summary of the invention

[0003] In view of the above situation, the present invention provides an automatic and senseless transmission control system for substation image data. The huge amount of image data generated by substation monitoring leads to the problem that it is more difficult to quickly and accurately identify anomalies from massive images. This solution adopts a multi-scale Flow model to deeply analyze image information. By distinguishing the needs of image anomaly detection and pixel-level anomaly positioning, the multi-scale aggregation strategy is flexibly used to not only achieve comprehensive image anomaly scanning, but also accurately locate abnormal pixels, thereby improving the accuracy and efficiency of anomaly recognition. For remote and harsh areas, the instability of network connection will further interfere with the problem of continuous data transmission and the transmission of high-definition images greatly occupies bandwidth, but network resources are limited. This solution adopts an optimized adaptive data packet scheduling method to ensure the continuity and stability of video and image data during transmission, and uses an improved intelligent optimizer algorithm to optimize the parameter vector of the optimized adaptive data packet scheduling method, thereby accelerating the parameter convergence speed, enhancing the reliability of data transmission, and bringing a smooth and unimpeded user experience, thereby maximizing the use of existing network resources.

[0004] The present invention provides a substation image data automatic and senseless transmission control system, which includes an image acquisition module, an intelligent video analysis module, an automatic transmission control module, a security and privacy protection module, and a senseless interaction design module;

[0005] The image acquisition module deploys high-definition cameras in key areas of the substation to capture high-definition video images around the clock and from multiple angles, ensuring clear images under all lighting conditions;

[0006] The intelligent video analysis module is responsible for analyzing the collected high-definition video images using a multi-scale Flow model, identifying normal data and abnormal data in the high-definition video images, obtaining abnormal data information and triggering abnormal alarms;

[0007] The automated transmission control module transmits the high-definition video images as video data and uses an optimized adaptive data packet scheduling method to automatically adjust transmission parameters according to network conditions;

[0008] The security and privacy protection module has built-in strict security measures. At the same time, for high-definition video images involving personal privacy, anonymization or masking of specific areas is performed;

[0009] The seamless interaction design module provides a visual interface for automatic scheduling and execution of preset rules, reducing the need for manual operations, minimizing the impact of the entire monitoring process on daily operation and maintenance work, and achieving a truly seamless experience.

[0010] Furthermore, the high-definition cameras in the image acquisition module include fixed cameras, panoramic cameras, and thermal imaging cameras, and drones are regularly used to conduct inspections within the substation to collect images in places where ordinary equipment is difficult to deploy. At the same time, infrared sensors and light intensity sensors are used to assist in image acquisition to optimize the environmental conditions for image acquisition;

[0011] Furthermore, the network devices required for the automatic transmission control module to be used during transmission include switches and routers for building a stable local area network or wide area network connection; the wireless communication module includes Wi-Fi and 4G / 5G modules for realizing remote transmission; and the software-defined network controller is used for intelligently managing network resources to ensure the security and reliability of data transmission;

[0012] Furthermore, the visualization interface includes a system status overview module, a real-time monitoring view module, a historical data playback module, an event management module, and a rule configuration module. The system status overview module displays the overall operating status of the system, including whether data is being transmitted and whether there is a fault alarm;

[0013] The real-time monitoring view module provides real-time video stream display of the camera, so that users can instantly view the on-site situation of the substation, support simultaneous display of multiple videos, and monitor key locations;

[0014] The historical data playback module allows users to retrieve and play video records within a certain period of time in the past, and provides a search function that can quickly locate the required data segment based on date and time;

[0015] The event management and recording module displays abnormal events and alarm triggering events occurring in the substation, and supports classification and screening of events;

[0016] The rule configuration module provides an interface for users to define automatic scheduling and preset rules, and provides a test mode for verifying the validity of the rules.

[0017] Furthermore, the intelligent video analysis module is responsible for analyzing the collected high-definition video images using a multi-scale Flow model, which specifically includes the following steps:

[0018] Step S1: primary feature extraction, using the pre-trained ViT model as a feature extractor, forward propagating the high-definition video image to extract feature maps at different levels to obtain primary feature maps;

[0019] Step S2: multi-scale feature extraction, using an average pooling layer to downsample the primary feature map to multiple scales to obtain a multi-scale feature map;

[0020] Step S3: Multi-scale Flow model, using asymmetric parallel flow to enhance the feature expression of each scale, and using fusion flow to fuse multi-scale information to build a multi-scale Flow model;

[0021] Step S4: Learning and optimization, setting the minimization of negative log-likelihood as the learning goal, operating the Jacobian determinants of parallel flows and fusion flows, and optimizing the multi-scale Flow model to distinguish between normal data and abnormal data distribution;

[0022] Step S5: Anomaly detection and scoring, using pixel-level anomaly localization and image-level anomaly detection methods to calculate anomaly scores;

[0023] Furthermore, in step S3, the multi-scale Flow model specifically includes the following steps:

[0024] Step S31: parallel flow design, designing a parallel flow with an asymmetric structure, constructing a parallel flow including L flow blocks, each flow block is constructed with an asymmetric architecture;

[0025] Step S32: In each parallel stream, a 3*3 convolutional ST network is used to automatically capture spatial context information, and position encoding is performed on each feature map in the multi-scale feature map to convert it into a multi-scale potential feature map;

[0026] Step S33: multi-scale information fusion, using the fusion flow to integrate the multi-scale potential feature maps, the fusion flow uses average pooling to adjust all input potential feature maps to the minimum size and then splices them, and uses two convolutional layers with similar structures to perform feature fusion to generate a fused feature map;

[0027] Step S34: feature segmentation, segmenting the fused feature map and re-adjusting it into a multi-scale feature map to obtain a segmented feature map;

[0028] Step S35: scale and shift weight learning, using group convolution to learn the scale weight and shift weight of each scale, adjust the features of the segmentation feature map, group convolution divides the input channels into N groups, and performs convolution operations on each group independently, reducing the amount of calculation and the number of parameters;

[0029] Furthermore, in step S5, anomaly detection and scoring specifically include the following steps:

[0030] Step S51: Log-likelihood estimation calculation, using log-likelihood to calculate the Jacobian determinant of any pixel on the segmentation feature map to obtain a log-likelihood map. The formula used is as follows: ;

[0031] In the formula, , and is a natural number, Represents the first On the feature map of scales, The pixel value at the position is the transformed potential feature; Representing potential features The Euclidean square of represents the latent feature space, Represents the latent feature space middle The log-likelihood estimate of ;

[0032] Step S52: log-likelihood map processing, using bilinear interpolation to enlarge the log-likelihood maps of all scales to the original image size to obtain a rescaled log-likelihood map;

[0033] Step S53: Pixel-level anomaly location, converting the rescaled log-likelihood map into a probability map, and calculating the pixel-level anomaly map through additive aggregation. The formula used is as follows: ; ;

[0034] In the formula, Indicates The probability map of the scale, is the additive aggregation probability map of pixel-level anomalies, Indicates to the maximum value, is a pixel-level anomaly map;

[0035] Step S54: Image-level anomaly detection, the rescaled log-likelihood maps are first added and then converted into probability maps, and the global anomaly map is calculated. The formula used is as follows: ; ;

[0036] In the formula, Indicates that all Multiply the values ​​from 1 to 3, is the multiplicative aggregation probability graph, is the global anomaly map at the image level;

[0037] Step S55: Calculate the anomaly score using the following formula: ;

[0038] In the formula, is the set parameter, Indicates from Select the largest K values ​​from Represents the final anomaly score obtained.

[0039] The automatic transmission control module uses an optimized adaptive data packet scheduling method to automatically adjust transmission parameters according to network conditions. The optimized adaptive data packet scheduling method specifically includes the following steps:

[0040] Step M1: Network monitoring, continuously monitoring the round-trip time, congestion window, number of bytes in transit, and packet loss rate of each path during data transmission and updating them in real time;

[0041] Step M2: video analysis, extracting the video characteristics of the video data during data transmission, calculating the real-time bit rate, and setting the frame priority;

[0042] Step M3: define the context and action space, construct the state vector, and define the action set. The state vector includes network status information and video characteristics. The action set includes four actions: default scheduling, copying frames, discarding frames, and packet loss compensation.

[0043] Step M4: Initialization: Initialize a linear model parameter vector for each action and set the initial confidence boundary value;

[0044] Step M5: Predict the action value. For each action, use the parameter vector corresponding to the current state and the action, use linear regression to predict the expected reward, get the predicted result, get the action corresponding to the expected reward, calculate and add the confidence boundary;

[0045] Step M6: Select the best action. According to the last prediction result, select the action with the highest expected reward as the output of this time.

[0046] Step M7: Execute scheduling and calculate actual rewards, schedule the data packet according to the action with the highest expected reward, and calculate the actual reward value of this decision;

[0047] Step M8: Update optimization, record the actual reward and the state vector when executing the scheduling, and use the improved intelligent optimizer algorithm to optimize and update the parameter vector of the corresponding action based on the actual reward and state vector obtained, and adjust the confidence boundary;

[0048] Furthermore, in step M8, the parameter vector of the corresponding action is optimized and updated using the improved intelligent optimizer algorithm, and the confidence boundary is adjusted, which specifically includes the following steps:

[0049] Step M81: Initialization phase, setting the upper and lower bounds of the search space, the total number of iterations, and the balance factor, and randomly generating initial candidate solutions and the positions of candidate solutions;

[0050] Step M82: Evaluate the initial candidate solution, calculate the fitness value of the initial candidate solution, set the initial candidate solution as the global optimal solution, and calculate and record the fitness value of the global optimal solution;

[0051] Step M83: Exploration phase, during the exploration phase Iterations, position update, where Less than the total number of iterations;

[0052] Furthermore, in step M83, the exploration phase specifically includes the following steps:

[0053] Step M831: Position update: Use the update rule to update the position of the candidate solution. The formula used is as follows: ; ;

[0054] In the formula, represents a natural number, represents the global optimal solution, Indicates The location of the candidate solution, is a random number in the range [0,1], represents the weight, is the balance factor, is the current iteration number, is the total number of iterations, is the natural exponential function, Indicates the number of iterations The weight factor when ;

[0055] Step M832: Check the boundaries. If the updated position exceeds the upper and lower bounds of the search space, set the updated position to the upper and lower bounds of the search space. Otherwise, keep it unchanged.

[0056] Step M833: Update the global best solution, calculate the fitness value of the updated candidate solution, if f(x) is better than f(gb), update the fitness value of the global best solution and the global best solution, otherwise, keep it unchanged, where f(x) is the fitness value of the updated candidate solution, and f(gb) is the fitness value of the global best solution;

[0057] Step M834: The iteration ends, and steps M831 to M833 are repeated. After iterations, the iteration ends;

[0058] Step M84: Development and dynamic adjustment, and the remaining Iterations are performed to dynamically adjust and update the position. add Equal to the total number of iterations;

[0059] Furthermore, in step M84, development and dynamic adjustment specifically include the following steps:

[0060] Step M841: Update the position of the candidate solution. The formula used is as follows: ;

[0061] In the formula, is a random number in the range [0,1], and are the upper and lower bounds of the search space respectively;

[0062] Step M842: Check fitness stagnation, and if the fitness value does not improve in m consecutive iterations, perform dynamic adjustment, where m is less than the total number of iterations;

[0063] Step M843: Dynamically adjust the position of the candidate solution. The formula used is as follows: ;

[0064] In the formula, is a random number in the range [0,1];

[0065] Step M844: Check the boundary and update it to confirm that the position of the candidate solution is still within the search space after the update, calculate the fitness value of the updated candidate solution, and update the fitness value of the global best solution and the global best solution;

[0066] Step M845: iterative optimization, repeating steps M841 to M844, and ending the iteration when the total number of iterations is reached;

[0067] Step M85: Output the final global optimal solution and the fitness value of the optimal solution.

[0068] The beneficial results achieved by the present invention using the above scheme are as follows:

[0069] (1) The huge amount of image data generated by substation monitoring has led to the problem that it is more difficult to quickly and accurately identify anomalies from massive images. This solution uses a multi-scale Flow model to deeply analyze image information, cleverly distinguishing and meeting the dual requirements of image anomaly detection and pixel-level precise positioning. By adopting a dynamic multi-scale aggregation strategy, not only can a wide range of image anomaly screening be achieved, but also a single abnormal pixel can be accurately located, greatly enhancing the accuracy and response speed of anomaly identification. In addition, this solution incorporates a pre-trained ViT model as a feature extractor, which effectively shortens the feature learning stage, ensures that feature extraction is both efficient and accurate, and further optimizes the overall performance.

[0070] (2) For remote and harsh areas, the instability of network connections will further interfere with the continuous transmission of data and the transmission of high-definition images will greatly occupy bandwidth, but network resources are limited. This solution adopts an optimized adaptive packet scheduling method to ensure the continuity and stability of video and image data during transmission. This method has carefully designed four sets of mechanisms, each of which is targeted to solve different network problems, thereby ensuring the continuity and reliability of video and image data in complex transmission environments. Through continuous self-optimization to adapt to the dynamic changes of the network environment, the reliability of data transmission is effectively enhanced, while achieving the optimization of network resource utilization efficiency, and bringing a smooth and unimpeded user experience, maximizing the use of existing network resources.

[0071] (3) In the optimized adaptive packet scheduling method, an improved intelligent optimizer algorithm is used to adjust the parameter vector of the action, improve network performance and data transmission efficiency, and adjust the confidence boundary. The SCO algorithm continuously adjusts the parameter vector through an iterative optimization process until it converges to the global optimal solution, thereby dynamically adjusting the confidence boundary to ensure that in an uncertain network environment, the scheduling strategy can maintain robustness and respond quickly to real-time changes, ultimately achieving more efficient and stable data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic diagram of an automatic and sensorless transmission control system for substation image data provided by the present invention;

[0073] Figure 2 It is a flow chart of the multi-scale Flow model;

[0074] Figure 3 is a flow chart of step 83;

[0075] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0076] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0077] Example 1, see Figure 1 , the present invention provides a substation image data automatic and senseless transmission control system, including an image acquisition module, an intelligent video analysis module, an automatic transmission control module, a security and privacy protection module and a senseless interaction design module;

[0078] The image acquisition module deploys high-definition cameras in key areas of the substation to capture high-definition video images around the clock and from multiple angles, ensuring clear images under all lighting conditions;

[0079] The intelligent video analysis module is responsible for analyzing the collected high-definition video images using a multi-scale Flow model, identifying normal data and abnormal data in the high-definition video images, obtaining abnormal data information and triggering abnormal alarms;

[0080] The automated transmission control module transmits the high-definition video images as video data and uses an optimized adaptive data packet scheduling method to automatically adjust transmission parameters according to network conditions;

[0081] The security and privacy protection module has built-in strict security measures. At the same time, for high-definition video images involving personal privacy, anonymization or masking of specific areas is performed;

[0082] The seamless interaction design module provides a visual interface for automatic scheduling and execution of preset rules, reducing the need for manual operations, minimizing the impact of the entire monitoring process on daily operation and maintenance work, and achieving a truly seamless experience.

[0083] Embodiment 2: This embodiment is based on the above embodiment. The high-definition camera in the image acquisition module includes a fixed camera, a panoramic camera, and a thermal imaging camera. A drone is regularly used to conduct inspections within the substation, and image acquisition is performed in places where ordinary equipment is difficult to deploy. At the same time, infrared sensors and light intensity sensors are used to assist in image acquisition to optimize the environmental conditions for image acquisition.

[0084] The network equipment required by the automatic transmission control module for transmission includes switches and routers for building a stable local area network or wide area network connection; the wireless communication module includes Wi-Fi and 4G / 5G modules for realizing remote transmission; the software-defined network controller is used to intelligently manage network resources to ensure the security and reliability of data transmission;

[0085] The visualization interface includes a system status overview module, a real-time monitoring view module, a historical data playback module, an event management module, and a rule configuration module. The system status overview module displays the overall operating status of the system, including whether data is being transmitted and whether there is a fault alarm;

[0086] The real-time monitoring view module provides real-time video stream display of the camera, so that users can instantly view the on-site situation of the substation, support simultaneous display of multiple videos, and monitor key locations;

[0087] The historical data playback module allows users to retrieve and play video records within a certain period of time in the past, and provides a search function that can quickly locate the required data segment based on date and time;

[0088] The event management and recording module displays abnormal events and alarm triggering events occurring in the substation, and supports classification and screening of events;

[0089] The rule configuration module provides an interface for users to define automatic scheduling and preset rules, and provides a test mode for verifying the validity of the rules.

[0090] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. The intelligent video analysis module is responsible for analyzing the collected high-definition video images using a multi-scale Flow model, which specifically includes the following steps:

[0091] Step S1: primary feature extraction, using the pre-trained ViT model as a feature extractor, forward propagating the high-definition video image to extract feature maps at different levels to obtain primary feature maps;

[0092] Step S2: multi-scale feature extraction, using an average pooling layer to downsample the primary feature map to multiple scales to obtain a multi-scale feature map;

[0093] Step S3: Multi-scale Flow model, using asymmetric parallel flow to enhance the feature expression of each scale, and using fusion flow to fuse multi-scale information to build a multi-scale Flow model;

[0094] Step S4: Learning and optimization, setting the minimization of negative log-likelihood as the learning goal, operating the Jacobian determinants of parallel flows and fusion flows, and optimizing the multi-scale Flow model to distinguish between normal data and abnormal data distribution;

[0095] Step S5: Anomaly detection and scoring, using pixel-level anomaly localization and image-level anomaly detection methods to calculate anomaly scores.

[0096] Embodiment 4: This embodiment is based on the above embodiment. In step S3, the multi-scale Flow model specifically includes the following steps:

[0097] Step S31: parallel flow design, designing a parallel flow with an asymmetric structure, constructing a parallel flow including L flow blocks, each flow block is constructed with an asymmetric architecture;

[0098] Step S32: In each parallel stream, a 3*3 convolutional ST network is used to automatically capture spatial context information, and position encoding is performed on each feature map in the multi-scale feature map to convert it into a multi-scale potential feature map;

[0099] Step S33: multi-scale information fusion, using the fusion flow to integrate the multi-scale potential feature maps, the fusion flow uses average pooling to adjust all input potential feature maps to the minimum size and then splices them, and uses two convolutional layers with similar structures to perform feature fusion to generate a fused feature map;

[0100] Step S34: feature segmentation, segmenting the fused feature map and re-adjusting it into a multi-scale feature map to obtain a segmented feature map;

[0101] Step S35: scale and shift weight learning, use group convolution to learn the scale weight and shift weight of each scale, adjust the features of the segmentation feature map, group convolution divides the input channels into N groups, and performs convolution operations on each group independently, reducing the amount of calculation and the number of parameters.

[0102] Embodiment 5, based on the above embodiment, in step S5, abnormality detection and scoring specifically include the following steps:

[0103] Step S51: Log-likelihood estimation calculation, using log-likelihood to calculate the Jacobian determinant of any pixel on the segmentation feature map to obtain a log-likelihood map. The formula used is as follows: ;

[0104] In the formula, , and is a natural number, Represents the first On the feature map of scales, The pixel value at the position is the transformed potential feature; Representing potential features The Euclidean square of represents the latent feature space, Represents the latent feature space middle The log-likelihood estimate of ;

[0105] Step S52: log-likelihood map processing, using bilinear interpolation to enlarge the log-likelihood maps of all scales to the original image size to obtain a rescaled log-likelihood map;

[0106] Step S53: Pixel-level anomaly location, converting the rescaled log-likelihood map into a probability map, and calculating the pixel-level anomaly map through additive aggregation. The formula used is as follows: ;

[0107] ;

[0108] In the formula, Indicates The probability map of the scale, is the additive aggregation probability map of pixel-level anomalies, Indicates to the maximum value, is a pixel-level anomaly map;

[0109] Step S54: Image-level anomaly detection, the rescaled log-likelihood maps are first added and then converted into probability maps, and the global anomaly map is calculated. The formula used is as follows: ; ;

[0110] In the formula, Indicates that all Multiply the values ​​from 1 to 3, is the multiplicative aggregation probability graph, is the global anomaly map at the image level;

[0111] Step S55: Calculate the anomaly score using the following formula: ;

[0112] In the formula, is the set parameter, Indicates from Select the largest K values ​​from Represents the final anomaly score obtained.

[0113] Embodiment 6: This embodiment is based on the above embodiment. In a specific embodiment, the parallel stream cascades more stream blocks for high-level feature maps with more channels. In the process of feature extraction, 2, 5 and 8 full convolution stream blocks are stacked on the primary feature map as asymmetric parallel streams. Each full convolution block includes two 3*3 convolutions, a RELU activation function and a layer normalization layer, and a 2D position encoding with 64 channels is inserted to capture absolute position information.

[0114] For information exchange, a fusion stream is connected to fuse the outputs of the three parallel streams. The stream model is trained from scratch on a 2080Ti GPU with a batch size of 16. In terms of optimization, we use the Adam optimizer with an initial learning rate of 1e-4 and reduce the learning rate by a factor of 3 at 70% and 90% of the training cycle. Different training rounds are set according to the size of the data set. In this embodiment, the training round is 100 rounds.

[0115] During the scale and shift weight learning process, group convolution divides the input channels into 10 groups. The number of channels contained in each group depends on the total number of original input channels. For example, if the original input has 100 channels, then each group will have 10 channels. Each group performs convolution operations independently, thereby performing fast convolution operations on multi-scale feature maps.

[0116] The huge amount of image data generated by substation monitoring has led to the problem of increasing difficulty in quickly and accurately identifying anomalies from massive images. This solution uses a multi-scale Flow model to deeply analyze image information, cleverly distinguishing and meeting the dual requirements of image anomaly detection and pixel-level precise positioning. By adopting a dynamic multi-scale aggregation strategy, it not only achieves extensive anomaly screening of images, but also accurately locates a single abnormal pixel, greatly enhancing the accuracy and response speed of anomaly identification. In addition, the solution incorporates a pre-trained ViT model as a feature extractor, which effectively shortens the feature learning stage, ensures that feature extraction is both efficient and accurate, and further optimizes the overall performance.

[0117] Embodiment 7, based on the above embodiment, the automatic transmission control module uses an optimized adaptive data packet scheduling method to automatically adjust transmission parameters according to network conditions. The optimized adaptive data packet scheduling method specifically includes the following steps:

[0118] Step M1: Network monitoring, continuously monitoring the round-trip time, congestion window, number of bytes in transit, and packet loss rate of each path during data transmission and updating them in real time;

[0119] Step M2: video analysis, extracting the video characteristics of the video data during data transmission, calculating the real-time bit rate, and setting the frame priority;

[0120] Step M3: define the context and action space, construct the state vector, and define the action set. The state vector includes network status information and video characteristics. The action set includes four actions: default scheduling, copying frames, discarding frames, and packet loss compensation.

[0121] Step M4: Initialization: Initialize a linear model parameter vector for each action and set the initial confidence boundary value;

[0122] Step M5: Predict the action value. For each action, use the parameter vector corresponding to the current state and the action, use linear regression to predict the expected reward, get the predicted result, get the action corresponding to the expected reward, calculate and add the confidence boundary;

[0123] Step M6: Select the best action. According to the last prediction result, select the action with the highest expected reward as the output of this time.

[0124] Step M7: Execute scheduling and calculate actual rewards, schedule the data packet according to the action with the highest expected reward, and calculate the actual reward value of this decision;

[0125] Step M8: Update optimization, record the actual reward and the state vector when executing the scheduling, and use the improved intelligent optimizer algorithm to optimize and update the parameter vector of the corresponding action based on the actual reward and state vector obtained, and adjust the confidence boundary.

[0126] Embodiment 8, this embodiment is based on the above embodiment. In step M3, the action set includes four actions of default scheduling, copying frames, discarding frames and packet loss compensation, corresponding to four mechanisms, including scheduling mechanism, repetition mechanism, discarding mechanism and compensation mechanism. In step M7, data packets are scheduled according to the mechanisms corresponding to different actions. The four mechanisms include the following contents:

[0127] Scheduling mechanism: First, collect the round-trip time of all established transmission paths, and sort these paths from small to large according to the round-trip time of the transmission paths. For each pair of adjacent paths after sorting, calculate the time interval ratio, and calculate the amount of data that needs to be pre-allocated for the path based on the time interval ratio and the bytes in transit and congestion window of the path to ensure that the data can arrive in order; according to the calculated pre-allocated data amount, set the pre-allocated data size for each path, while considering the current amount of data to be sent S, and update S to reflect the allocated data amount. At the same time, create a buffer for each path to store the pre-allocated data, ensuring that the data can be continuously taken out of the buffer and sent, optimizing the path utilization efficiency;

[0128] Replication mechanism: Analyze the packet loss rate of each path, identify the path with poor network conditions, determine the path with the lowest packet loss rate as the main path, and use it to transmit the original data packet. Other paths with available congestion windows are set as backup paths to send redundant copies of data. Only high-priority data packets are replicated and distributed to backup paths to increase the probability of data packets reaching the receiving end. Redundant data packets are marked as semi-reliable to avoid triggering unnecessary retransmissions when they are lost, thereby saving bandwidth resources.

[0129] Abandonment mechanism: Determine which data is critical and which is non-critical based on the priority of the video frame. When bandwidth is limited, the abandonment mechanism is enabled, and only the header information of the M frame is transmitted, and the rest is discarded. The header is retained to ensure that the application layer can process the data correctly. When the bandwidth is insufficient to support all data transmission, the complete transmission of high-priority data is prioritized, and non-critical data is discarded;

[0130] Compensation mechanism: When the packet loss rate of a path is extremely high and the bandwidth is insufficient, consider taking compensation measures. When packaging video data, mark the reliability attribute of the QUIC packet according to the data priority. High-priority data and control information are marked as reliable, and the rest are unreliable. For packets marked as unreliable, retransmission is not triggered when they are lost, thereby saving bandwidth resources. Set a deadline based on the longest round-trip time of all paths. When the time waiting for data to fill the holes exceeds this deadline, fill these holes with zeros to reduce the impact on users.

[0131] Embodiment 9: This embodiment is based on the above embodiment. In step M8, an improved intelligent optimizer algorithm is used to optimize and update the parameter vector of the corresponding action and adjust the confidence boundary. Specifically, the following steps are included:

[0132] Step M81: Initialization phase, setting the upper and lower bounds of the search space, the total number of iterations, and the balance factor, and randomly generating initial candidate solutions and the positions of candidate solutions;

[0133] Step M82: Evaluate the initial candidate solution, calculate the fitness value of the initial candidate solution, set the initial candidate solution as the global optimal solution, and calculate and record the fitness value of the global optimal solution;

[0134] Step M83: Exploration phase, during the exploration phase Iterations, position update, where Less than the total number of iterations;

[0135] Step M84: Development and dynamic adjustment, and the remaining Iterations are performed to dynamically adjust and update the position. add Equal to the total number of iterations;

[0136] Step M85: Output the final global optimal solution and the fitness value of the optimal solution.

[0137] Embodiment 10, see Figure 3 This embodiment is based on the above embodiment. In step M83, the exploration phase specifically includes the following steps:

[0138] Step M831: Position update: Use the update rule to update the position of the candidate solution. The formula used is as follows: ; ;

[0139] In the formula, represents a natural number, represents the global optimal solution, Indicates The location of the candidate solution, is a random number in the range [0,1], represents the weight, is the balance factor, is the current iteration number, is the total number of iterations, is the natural exponential function, Indicates the number of iterations The weight factor when ;

[0140] Step M832: Check the boundaries. If the updated position exceeds the upper and lower bounds of the search space, set the updated position to the upper and lower bounds of the search space. Otherwise, keep it unchanged.

[0141] Step M833: Update the global best solution, calculate the fitness value of the updated candidate solution, if f(x) is better than f(gb), update the fitness value of the global best solution and the global best solution, otherwise, keep it unchanged, where f(x) is the fitness value of the updated candidate solution, and f(gb) is the fitness value of the global best solution;

[0142] Step M834: The iteration ends, and steps M831 to M833 are repeated. After iterations, the iteration ends;

[0143] In step M84, development and dynamic adjustment specifically include the following steps:

[0144] Step M841: Update the position of the candidate solution. The formula used is as follows: ;

[0145] In the formula, is a random number in the range [0,1], and are the upper and lower bounds of the search space respectively;

[0146] Step M842: Check fitness stagnation, and if the fitness value does not improve in m consecutive iterations, perform dynamic adjustment, where m is less than the total number of iterations;

[0147] Step M843: Dynamically adjust the position of the candidate solution. The formula used is as follows: ;

[0148] In the formula, is a random number in the range [0,1];

[0149] Step M844: Check the boundary and update it to confirm that the position of the candidate solution is still within the search space after the update, calculate the fitness value of the updated candidate solution, and update the fitness value of the global best solution and the global best solution;

[0150] Step M845: Iterative optimization, repeat steps M841 to M844, and the iteration ends when the total number of iterations is reached.

[0151] In view of the problem that the instability of network connection will further interfere with the continuous transmission of data in remote areas and areas with harsh environments, and the transmission of high-definition images will greatly occupy bandwidth, but network resources are limited, this solution adopts an optimized adaptive data packet scheduling method to ensure the continuity and stability of video and image data during transmission. This method has carefully designed four sets of mechanisms, each of which is targeted to solve different network problems, thereby ensuring the continuity and reliability of video and image data in complex transmission environments. Through continuous self-optimization to adapt to the dynamic changes of the network environment, the reliability of data transmission is effectively enhanced, while achieving the optimization of network resource utilization efficiency, and bringing a smooth and unimpeded user experience, maximizing the use of existing network resources.

[0152] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A substation image data automatic and non-sensing transmission control system, characterized by: It includes image acquisition module, intelligent video analysis module, automatic transmission control module, security and privacy protection module and non-sensing interaction design module; The image acquisition module deploys high-definition cameras in key areas of the substation to collect high-definition video images around the clock and from multiple angles; The intelligent video analysis module is responsible for analyzing the collected high-definition video images using a multi-scale Flow model, identifying normal data and abnormal data in the high-definition video images, obtaining abnormal data information and triggering abnormal alarms; The automated transmission control module transmits the high-definition video images as video data and uses an optimized adaptive data packet scheduling method to automatically adjust transmission parameters according to network conditions; The security and privacy protection module has built-in strict security measures. At the same time, for high-definition video images involving personal privacy, anonymization or masking of specific areas is performed; The seamless interaction design module provides a visual interface for automatic scheduling and execution of preset rules, reducing the need for manual operation.

2. According to claim 1, a substation image data automatic non-sensing transmission control system is characterized by: The high-definition cameras in the image acquisition module include fixed cameras, panoramic cameras, and thermal imaging cameras. UAVs are regularly used to conduct inspections within the substation and to collect images in places where ordinary equipment is difficult to deploy. At the same time, infrared sensors and light intensity sensors are used to assist in image acquisition and optimize the environmental conditions for image acquisition. The network equipment required by the automatic transmission control module for transmission includes switches and routers for building a stable local area network or wide area network connection; the wireless communication module includes Wi-Fi and 4G / 5G modules for realizing remote transmission; the software-defined network controller is used to intelligently manage network resources to ensure the security and reliability of data transmission; The visualization interface includes a system status overview module, a real-time monitoring view module, a historical data playback module, an event management module, and a rule configuration module. The system status overview module displays the overall operating status of the system, including whether data is being transmitted and whether there is a fault alarm; The real-time monitoring view module provides real-time video stream display of the camera, so that users can instantly view the on-site situation of the substation, support simultaneous display of multiple videos, and monitor key locations; The historical data playback module allows users to retrieve and play video records within a certain period of time in the past, and provides a search function that can quickly locate the required data segment based on date and time; The event management and recording module displays abnormal events and alarm triggering events occurring in the substation, and supports classification and screening of events; The rule configuration module provides an interface for users to define automatic scheduling and preset rules, and provides a test mode for verifying the validity of the rules.

3. According to claim 1, the automatic and non-conductive transmission control system for substation image data is characterized in that: The intelligent video analysis module is responsible for analyzing the collected high-definition video images using a multi-scale Flow model, which specifically includes the following steps: Step S1: primary feature extraction, using the pre-trained ViT model as a feature extractor, forward propagating the high-definition video image to extract feature maps at different levels to obtain primary feature maps; Step S2: multi-scale feature extraction, using an average pooling layer to downsample the primary feature map to multiple scales to obtain a multi-scale feature map; Step S3: Multi-scale Flow model, using asymmetric parallel flow to enhance the feature expression of each scale, and using fusion flow to fuse multi-scale information to build a multi-scale Flow model; Step S4: Learning and optimization, setting the minimization of negative log-likelihood as the learning goal, operating the Jacobian determinants of parallel flows and fusion flows, and optimizing the multi-scale Flow model to distinguish between normal data and abnormal data distribution; Step S5: Anomaly detection and scoring, using pixel-level anomaly localization and image-level anomaly detection methods to calculate anomaly scores.

4. According to claim 3, a substation image data automatic non-sensing transmission control system is characterized by: In step S3, the multi-scale Flow model specifically includes the following steps: Step S31: parallel flow design, designing a parallel flow with an asymmetric structure, constructing a parallel flow including L flow blocks, each flow block is constructed with an asymmetric architecture; Step S32: In each parallel stream, a 3*3 convolutional ST network is used to automatically capture spatial context information, and position encoding is performed on each feature map in the multi-scale feature map to convert it into a multi-scale potential feature map; Step S33: multi-scale information fusion, using the fusion flow to integrate the multi-scale potential feature maps, the fusion flow uses average pooling to adjust all input potential feature maps to the minimum size and then splices them, and uses two convolutional layers with similar structures to perform feature fusion to generate a fused feature map; Step S34: feature segmentation, segmenting the fused feature map and re-adjusting it into a multi-scale feature map to obtain a segmented feature map; Step S35: scale and shift weight learning, using group convolution to learn the scale weight and shift weight of each scale, and adjust the features of the segmentation feature map.

5. According to claim 3, the automatic and non-conductive transmission control system for substation image data is characterized in that: In step S5, anomaly detection and scoring specifically includes the following steps: Step S51: Log-likelihood estimation calculation, using the log-likelihood to calculate the Jacobian determinant of any pixel on the segmentation feature map to obtain a log-likelihood map; Step S52: log-likelihood map processing, using bilinear interpolation to enlarge the log-likelihood maps of all scales to the original image size to obtain a rescaled log-likelihood map; Step S53: pixel-level anomaly location, converting the rescaled log-likelihood map into a probability map, and calculating a pixel-level anomaly map through additive aggregation; Step S54: image-level anomaly detection, adding the rescaled log-likelihood maps and then converting them into probability maps, and calculating a global anomaly map; Step S55: Calculate the anomaly score.

6. The automatic and non-conductive transmission control system for substation image data according to claim 1 is characterized by: The automatic transmission control module uses an optimized adaptive data packet scheduling method to automatically adjust transmission parameters according to network conditions. The optimized adaptive data packet scheduling method specifically includes the following steps: Step M1: Network monitoring, continuously monitoring the round-trip time, congestion window, number of bytes in transit, and packet loss rate of each path during data transmission and updating them in real time; Step M2: video analysis, extracting the video characteristics of the video data during data transmission, calculating the real-time bit rate, and setting the frame priority; Step M3: define the context and action space, construct the state vector, and define the action set. The state vector includes network state information and video characteristics. Step M4: Initialization: Initialize a linear model parameter vector for each action and set the initial confidence boundary value; Step M5: Predict the action value. For each action, use the parameter vector corresponding to the current state and the action, use linear regression to predict the expected reward, get the prediction result, get the action corresponding to the expected reward, calculate and add the confidence boundary; Step M6: Select the best action. According to the last prediction result, select the action with the highest expected reward as the output of this time. Step M7: Execute scheduling and calculate actual rewards, schedule the data packet according to the action with the highest expected reward, and calculate the actual reward value of this decision; Step M8: Update optimization, record the actual reward and the state vector when executing the scheduling, and use the improved intelligent optimizer algorithm to optimize and update the parameter vector of the corresponding action based on the actual reward and state vector obtained, and adjust the confidence boundary.

7. The automatic and non-conductive transmission control system for substation image data according to claim 6 is characterized by: In step M8, the parameter vector of the corresponding action is optimized and updated using the improved intelligent optimizer algorithm, and the confidence boundary is adjusted, which specifically includes the following steps: Step M81: Initialization phase, setting the upper and lower bounds of the search space, the total number of iterations, and the balance factor, and randomly generating initial candidate solutions and the positions of candidate solutions; Step M82: Evaluate the initial candidate solution, calculate the fitness value of the initial candidate solution, set the initial candidate solution as the global optimal solution, and calculate and record the fitness value of the global optimal solution; Step M83: Exploration phase, during the exploration phase Iterations, position update, where Less than the total number of iterations; Step M84: Development and dynamic adjustment, and the remaining Iterations are performed to dynamically adjust and update the position. add Equal to the total number of iterations; Step M85: Output the final global optimal solution and the fitness value of the optimal solution.

8. The automatic and non-conductive transmission control system for substation image data according to claim 7 is characterized by: In step M83, the exploration phase specifically includes the following steps: Step M831: position update, using the update rule to update the position of the candidate solution; Step M832: Check the boundaries. If the updated position exceeds the upper and lower bounds of the search space, set the updated position to the upper and lower bounds of the search space. Otherwise, keep it unchanged. Step M833: Update the global best solution, calculate the fitness value of the updated candidate solution, if f(x) is better than f(gb), update the fitness value of the global best solution and the global best solution, otherwise, keep it unchanged, where f(x) is the fitness value of the updated candidate solution, and f(gb) is the fitness value of the global best solution; Step M834: The iteration ends, and steps M831 to M833 are repeated. After iterations, the iteration ends.

9. The automatic and non-conductive transmission control system for substation image data according to claim 7 is characterized by: In step M84, development and dynamic adjustment specifically include the following steps: Step M841: Update the position of the candidate solution; Step M842: Check fitness stagnation, and if the fitness value does not improve in m consecutive iterations, perform dynamic adjustment, where m is less than the total number of iterations; Step M843: dynamically adjust the position of the candidate solution; Step M844: Check the boundary and update it to confirm that the position of the candidate solution is still within the search space after the update, calculate the fitness value of the updated candidate solution, and update the fitness value of the global best solution and the global best solution; Step M845: iterative optimization, repeating steps M841 to M844, and ending the iteration when the total number of iterations is reached; Step M85: Output the final global optimal solution and the fitness value of the optimal solution.

Images