Distributed edge computing system and method based on power smart station scenario

Abstract

This solution belongs to the field of information technology of distribution network, and specifically relates to a distributed edge computing system and method based on the scenario of smart power station. The distributed edge computing method based on the scenario of smart power station includes the following steps: S10: deploying edge computing nodes in each station, collecting sensor monitoring data of each power equipment as first data, collecting each monitoring image data as second data, and sending the first data and the second data to the edge computing node; S20: the edge computing node identifies the preset marked area in the second data, extracts the pixel color information in the marked area as the first information, and compares the first information extracted from the current marked area with the first information extracted from other marked areas. This solution solves the problems of inaccurate leakage monitoring data and low upload speed of the station under the condition of limited operation and maintenance and inspection costs, and at the same time, optimizes the operation and maintenance efficiency.

Description

Technical Field

[0001] This solution belongs to the field of distribution network information management technology, and specifically involves a distributed edge computing system and method based on a smart power station scenario. Background Art

[0002] With the acceleration of urbanization and the continuous expansion of urban construction, the number of power stations under the jurisdiction of power supply companies is also growing rapidly. These stations are numerous, widely distributed, operate in complex environments, and present high safety risks. Some stations have aging equipment and harsh surrounding environments, posing safety hazards such as electric shock to inspectors while performing their inspections. This poses numerous challenges to traditional manual operations and inspections, including high labor costs, difficulty in inspections, long inspection cycles, and low efficiency, making it difficult to accurately and timely monitor the operating conditions of the stations.

[0003] To address the above issues, existing technologies install sensors in power stations to collect monitoring data from various power equipment and upload this data to the cloud. The cloud performs statistics and calculations on the monitoring data. When the monitoring data shows an anomaly, an abnormality signal is generated and sent to the operation and maintenance terminal. This method reduces the cost of manual inspections and improves the efficiency of operation and maintenance inspections. However, leakage from power equipment and electromagnetic signals generated during operation not only interfere with the sensor's normal detection of current (affecting the accuracy of the detection data) but also affect the speed at which the sensor uploads monitoring data (reducing the upload speed of monitoring data). If each power station is equipped with a device specifically designed to counteract electromagnetic interference, the economic cost of operation and maintenance inspections will increase significantly. Summary of the Invention

[0004] The purpose of this solution is to provide a distributed edge computing system and method based on the power smart station scenario to solve the problems of inaccurate leakage monitoring data and low upload speed in the station under limited operation and maintenance and inspection costs.

[0005] To achieve the above objectives, this solution provides a distributed edge computing method based on the power smart station scenario, including the following steps:

[0006] S10: Deploy an edge computing node in each station building, collect sensor monitoring data of each power device as first data, collect each monitoring image data as second data, and send the first data and the second data to the edge computing node;

[0007] S20: The edge computing node identifies a preset annotated area in the second data, extracts pixel color information within the annotated area as first information, compares the first information extracted from the current annotated area with the first information extracted from other annotated areas, and calculates pixel color similarity values ​​between the current annotated area and the other annotated areas;

[0008] S30: The edge computing node performs statistics on the first data and generates an abnormality signal based on the statistical results. After generating the abnormality signal, the edge computing node obtains several annotated areas corresponding to the first data, obtains pixel color similarity values ​​corresponding to each annotated area, generates a maintenance instruction based on the pixel color similarity values ​​and the positions of the annotated areas, and uploads the maintenance instruction to the cloud.

[0009] S40: The edge computing node obtains data of other edge computing nodes with the smallest communication delay as an assisting node, splits the first data and the second data corresponding to the abnormal signal into multiple data packets according to the number of assisting nodes, and sends the data packets to the assisting nodes; after receiving the data packets, the assisting node uploads the data packets to the cloud.

[0010] And, a distributed edge computing system based on the electric power smart station scenario that uses a distributed edge computing method based on the electric power smart station scenario.

[0011] The principles and technical benefits of this solution are as follows: First, it deploys edge computing nodes within smart power stations to collect power equipment sensor data (first data) and monitoring image data (second data) locally, and completes local data processing. This mechanism effectively avoids the latency associated with long-distance data transmission, significantly improving system response speed and ensuring timely data processing.

[0012] Secondly, when key equipment in the station building generates an induced potential, the electromagnetic signal interference from this potential is less pronounced in surveillance image acquisition than with traditional sensors, resulting in higher data accuracy. Although the station building's outer shell is equipped with a grounding wire to prevent safety hazards caused by leakage, this wire is easily disconnected due to environmental influences. While it may appear to be connected to the outer shell, it is actually disconnected, making grounding wire issues easily missed. This solution accurately identifies pre-defined annotated areas in surveillance images (such as key areas of the station building's outer shell), extracts pixel color information, and calculates pixel color similarity values ​​between these annotated areas to quantify image differences. This is because when a leakage occurs in the station building, the outer shells of some equipment carry an electric charge, forming various forms of capacitors, which attract dust from the environment. By quantitatively analyzing image differences at different locations on the equipment shells, this solution can pinpoint the annotated areas with the highest dust absorption. Furthermore, multi-area comparative analysis effectively eliminates interference from environmental factors such as lighting, accurately identifying areas of abnormally charged electricity in the station building. This technology not only significantly reduces the interference of electromagnetic signals on data collection, but also significantly improves the accuracy of leakage monitoring through image recognition technology, greatly shortens the time for manual investigation of leakage locations, and significantly optimizes operation and maintenance efficiency.

[0013] Furthermore, this solution leverages the redundant communication capabilities of other edge nodes (assisting nodes) to split data packets and upload them in parallel. This approach effectively avoids the high latency associated with excessive data volumes on a single node. The collaborative efforts of assisting nodes significantly increase data upload speeds and reduce transmission loss rates, effectively ensuring reliable data transmission. Furthermore, this collaborative transmission model reduces the hardware performance requirements of individual nodes, effectively saving hardware procurement and maintenance costs.

[0014] In summary, this solution solves the problems of inaccurate leakage monitoring data and low upload speed in station buildings while limiting operation and maintenance and inspection costs. At the same time, it also optimizes operation and maintenance efficiency.

[0015] Furthermore, when the edge computing node extracts the first information, it performs color space conversion on the pixels in the marked area, converts the RGB value into HSV space, takes the maximum value of the three components of red, green, and blue as the brightness value, calculates the saturation by the difference between the brightness and the minimum component and the ratio of the brightness, and determines the hue according to the component corresponding to the maximum value of the three RGB components, as shown in the following formula (1):

[0016] (1).

[0017] By converting the RGB color space to the HSV color space, not only can the color information in the image be processed and analyzed more intuitively and effectively, but by calculating the maximum value of the red, green, and blue components as the brightness value, the brightness information of the image can be more accurately represented, which is of great significance for subsequent image analysis and processing. In addition, using the difference between the brightness and the minimum component and the ratio of the brightness to calculate the saturation can more accurately reflect the purity or intensity of the color, and determining the hue according to the component corresponding to the maximum value of the three RGB components can more accurately represent the type of color in the image. The precise calculation of these color features not only optimizes the color space conversion and color feature extraction process of the edge computing node, but also reduces the amount of data transmission and reduces the dependence on cloud computing resources, thereby improving the efficiency and real-time performance of data processing.

[0018] Furthermore, when calculating the pixel color similarity between the current marked area and other marked areas, for the current marked area , and each other marked area Compare one by one and calculate the color feature distance between the two areas , The calculation formula is as follows:

[0019] (3),

[0020] in, , , and , , Respectively for regions and The mean of hue, saturation, and value in HSV color space; , , is the weight coefficient of each color component, and satisfies ; According to the region With the current area Calculate spatial distance from pixel coordinates ,according to Setting weights , The calculation formula is shown in the following formula (4):

[0021] (4),

[0022] in, is the distance attenuation coefficient; the current area is calculated by combining the similarity and weight of all other annotated areas with the current area The overall similarity value of , The calculation formula is as follows:

[0023] (5).

[0024] By calculating color similarity per region and using a spatial distance weighting mechanism, the accuracy and anti-interference capabilities of power equipment anomaly detection are significantly improved. Furthermore, by calculating HSV hue distance and assigning weights to adjacent regions, the solution can effectively distinguish between true equipment failures (such as color changes caused by leakage) and environmental interference (such as light fluctuations), significantly reducing false alarm rates. Furthermore, because the weight coefficients and distance attenuation coefficients can be adjusted for different station building scenarios, this solution is dynamically adaptable and compatible with diverse equipment layouts and monitoring conditions. Furthermore, by performing pixel-by-pixel analysis only on candidate anomaly regions, computational redundancy in global image processing is avoided, making it particularly suitable for resource-constrained edge nodes. Furthermore, this solution can establish a baseline for equipment color changes by recording regional similarity values ​​over a long period of time, thereby assisting in predicting potential failures. Furthermore, by comparing similarity values ​​across stations, it can ultimately identify systemic failures (such as the collective degradation of a batch of equipment).

[0025] Furthermore, in step S30, the edge computing node also extracts the motion trajectory of the target device in the second data. After the edge computing node generates an abnormal signal, it combines the running trajectory with the pixel color similarity value to determine whether the marked area is abnormally charged, and then generates a maintenance instruction based on the judgment result and the position of the marked area; the target device includes a hanging rod and a plurality of feathers, the hanging rod is fixedly connected to the station building, and multiple feathers are hung on the hanging rod one by one; when the edge computing node identifies the motion trajectory of the target device, it first calculates the swing arc of the feather according to the relative position between the feather and the target area in the second data, and then calculates the swing trajectory of the feather according to the swing arc of the feather in combination with the acquisition time of the second data, and uses the swing trajectory of the feather as the motion trajectory of the target device.

[0026] Furthermore, when the edge computing node identifies the motion trajectory of the target device, the following steps are included:

[0027] S31: For The collaborative detection model is established by the following formula (6):

[0028] (6),

[0029] in, Indicates the A feather in The swing angle at the moment, is the allowed angle deviation threshold;

[0030] S32: The three-dimensional motion trajectory of the feather is reconstructed based on the three-dimensional coordinates of the binocular vision. The reconstruction formula is shown in the following formula (7):

[0031] (7),

[0032] in is the baseline distance, is the focal length, for parallax;

[0033] S33: Calculate the dynamic characteristic parameters of the swing trajectory. The calculation formula of the characteristic parameters is shown in the following formula (8):

[0034] (8);

[0035] S34: Use support vector machine classifier to perform abnormality judgment. The judgment formula is shown in the following formula (9):

[0036] (9),

[0037] Among them, the eigenvector ;

[0038] S35: Establish the spatiotemporal correlation matrix between feathers. The matrix is ​​shown in the following formula (10):

[0039] (10),

[0040] When the matrix Rank When the regional abnormality alarm is triggered.

[0041] This solution significantly improves the accuracy and reliability of live anomaly detection through the multimodal fusion of mechanical and visual signals. Furthermore, due to the introduction of high-precision three-dimensional motion modeling and dynamic feature analysis, it can achieve submillimeter-level micro-motion detection and millisecond-level transient response. Furthermore, through multi-feather collaborative detection and probabilistic fusion mechanisms, it effectively suppresses environmental interference and single-point false triggering. Furthermore, by combining frequency-domain energy entropy analysis and spatiotemporal correlation matrices, it further enhances its ability to resist power frequency interference and fault location accuracy. Furthermore, because this solution supports long-term trajectory feature recording and self-learning optimization, it not only meets real-time monitoring needs but also provides data support for predictive maintenance. Furthermore, thanks to standardized parameter design and a flexible expansion architecture, this solution can adapt to diverse power equipment monitoring scenarios while reducing operation and maintenance costs.

[0042] Furthermore, when a regional anomaly alarm is triggered, the edge computing node combines the feather swing trajectory with the pixel color similarity value to establish a joint decision function. The joint decision function is shown in the following formula (11):

[0043] (11),

[0044] in, is the Sigmoid activation function; is the normalized swing amplitude characteristic; is the pixel color similarity value; is the estimated electric field strength vector, and is the trainable weight;

[0045] The priority weight is calculated according to the position of the marked area. The calculation formula is shown in the following formula (12):

[0046] (12),

[0047] in, For the The center coordinates of the labeled area, The center of the feather swing;

[0048] The abnormality level function is defined according to the joint decision function, and the definition formula is shown in the following formula (13):

[0049] L= (13),

[0050] The threshold Obtained through training with historical data;

[0051] Finally, the maintenance instruction is generated according to the maintenance instruction parameters. The calculation formula of the maintenance instruction parameters is shown in the following formula (14):

[0052] (14),

[0053] in is the required response time threshold.

[0054] This solution not only significantly improves the accuracy and reliability of live anomaly detection through multimodal feature fusion and spatial position weighting mechanisms, but also enables accurate fault classification and priority determination due to the introduction of anomaly level classification and dynamic weight adjustment. At the same time, by combining feather swing features with visual similarity analysis, it not only effectively enhances the anti-interference ability in complex environments, but also significantly improves the operation and maintenance response efficiency due to the adoption of intelligent maintenance instruction generation strategies. In addition, because this solution supports online parameter optimization and real-time response, it can not only adapt to changing on-site conditions, but also provide decision support for resource scheduling. Furthermore, thanks to the modular instruction architecture and scalable fusion algorithm, this solution can be flexibly applied to intelligent operation and maintenance scenarios of various power equipment while ensuring detection accuracy.

[0055] Furthermore, in step S40, when the edge computing node splits the abnormal data into multiple data packets according to the number of assisting nodes, the following steps are included:

[0056] S41: Establish the node selection scoring function shown in the following formula (15):

[0057] (15),

[0058] in, To the node The round trip delay, is the available bandwidth, is the CPU load rate, is a configurable weight;

[0059] S42: Calculate the optimal number of shards using the following formula (16):

[0060] (16),

[0061] in, is the total amount of abnormal data, is the maximum transmission unit of the network, is the baseline bandwidth requirement;

[0062] S43: Allocate data packet size based on node score. The allocation formula is shown in the following formula (17):

[0063] (17);

[0064] S44: Fountain code is used for encoding. The encoding formula is shown in the following formula (18):

[0065] (18),

[0066] in, Shard the original data. To randomly generate matrix elements, .

[0067] This solution significantly improves the efficiency and reliability of abnormal data transmission through a dynamic scoring model and adaptive sharding algorithm. Furthermore, due to its real-time load sensing mechanism, it can intelligently optimize network resource allocation. Furthermore, through fountain code encoding and dynamic sharding strategies, it not only ensures transmission success rates in high-concurrency environments, but also enhances the scalability of this solution by supporting elastic node scaling. Furthermore, the introduction of periodic load balancing adjustments not only adapts to dynamic changes in network status but also effectively reduces task processing latency. Furthermore, thanks to the flexible configuration of the scoring model and the adaptive nature of the sharding algorithm, this solution can be widely applied to the collaborative processing needs of various edge computing scenarios while ensuring transmission quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 This is a flow chart of a distributed edge computing method based on a smart power station scenario in an embodiment of the present invention. DETAILED DESCRIPTION

[0069] The following will clearly and completely describe the concept and technical effects of the present invention in conjunction with the embodiments to fully understand the purpose, features and effects of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative work are all within the scope of protection of the present invention:

[0070] like Figure 1 As shown, a distributed edge computing method based on a power smart station scenario includes the following steps:

[0071] S10: Deploy an edge computing node in each station building, collect sensor monitoring data of each power device as first data, collect each monitoring image data as second data, and send the first data and the second data to the edge computing node;

[0072] S20: The edge computing node identifies a preset annotated area in the second data, extracts pixel color information within the annotated area as first information, compares the first information extracted from the current annotated area with the first information extracted from other annotated areas, and calculates pixel color similarity values ​​between the current annotated area and the other annotated areas;

[0073] S30: The edge computing node performs statistics on the first data and generates an abnormality signal based on the statistical results. After generating the abnormality signal, the edge computing node obtains several annotated areas corresponding to the first data, obtains pixel color similarity values ​​corresponding to each annotated area, generates a maintenance instruction based on the pixel color similarity values ​​and the positions of the annotated areas, and uploads the maintenance instruction to the cloud.

[0074] S40: The edge computing node obtains data of other edge computing nodes with the smallest communication delay as an assisting node, splits the first data and the second data corresponding to the abnormal signal into multiple data packets according to the number of assisting nodes, and sends the data packets to the assisting nodes; after receiving the data packets, the assisting node uploads the data packets to the cloud.

[0075] Among them, in step S20, when the edge computing node extracts the first information, it performs color space conversion on the pixels in the marked area, converts the RGB value into HSV space, takes the maximum value of the three components of red, green, and blue as the brightness value, and uses the difference between the brightness and the minimum component and the ratio of the brightness to calculate the saturation. The calculation formula of saturation is shown in the following formula (2); the hue is determined according to the component corresponding to the maximum value of the three RGB components, as shown in the following formula (1):

[0076] (1),

[0077] (2).

[0078] Where R, G, and B are the red, green, and blue components of the pixel (range 0-255). V is the max(R,G,B), which represents the brightness. When V = R (red is dominant), use (GB) to calculate the hue; when V = G (green is dominant), use (BR) and add 120° (2 corresponds to 120° in the formula); when V = B (blue is dominant), use (RG) and add 240° (4 corresponds to 240° in the formula). V - min(R,G,B) indicates color purity and avoids division by zero (saturation is 0 when V = 0).

[0079] Among them, when calculating the pixel color similarity value of the current marked area and other marked areas, for the current marked area , and each other marked area Compare one by one and calculate the color feature distance between the two areas , The calculation formula is as follows:

[0080] (3),

[0081] in, , , and , , Respectively for regions and The mean of hue, saturation, and value in HSV color space; , , is the weight coefficient of each color component, and satisfies ;

[0082] According to the region With the current area Calculate spatial distance from pixel coordinates ,according to Setting weights , The calculation formula is shown in the following formula (4):

[0083] (4),

[0084] in, is the distance attenuation coefficient, which is used to adjust the influence of spatial distance on weight;

[0085] Combine the similarity and weight of all other annotated areas with the current area to calculate the current area The overall similarity value of , The calculation formula is as follows:

[0086] (5),

[0087] in, Map distance to similarity (range [0,1]), weight Used to emphasize the influence of neighboring areas.

[0088] Among them, in step S30, the edge computing node also extracts the motion trajectory of the target device in the second data. After the edge computing node generates an abnormal signal, it combines the running trajectory with the pixel color similarity value to determine whether the marked area is abnormally charged, and then generates a maintenance instruction based on the judgment result and the position of the marked area; the target device includes a hanging rod and a plurality of feathers, and the hanging rod is fixedly connected to the station building, and multiple feathers are hung on the hanging rod one by one; when the edge computing node identifies the motion trajectory of the target device, it first calculates the swing arc of the feather according to the relative position between the feather and the target area in the second data, and then calculates the swing trajectory of the feather according to the swing arc of the feather in combination with the acquisition time of the second data, and uses the swing trajectory of the feather as the motion trajectory of the target device.

[0089] Specifically, when the edge computing node identifies the motion trajectory of the target device, the following steps are included:

[0090] S31: For The collaborative detection model is established by the following formula (6):

[0091] (6),

[0092] in, Indicates the A feather in The swing angle at the moment, is the allowed angle deviation threshold;

[0093] S32: The three-dimensional motion trajectory of the feather is reconstructed based on the three-dimensional coordinates of the binocular vision. The reconstruction formula is shown in the following formula (7):

[0094] (7),

[0095] in is the baseline distance, is the focal length, for parallax; and is the pixel coordinate of the feather in the left camera image (unit: pixel); ,and The principal point coordinates of the left camera (image center, unit: pixel), used to correct optical center offset.

[0096] S33: Calculate the dynamic characteristic parameters of the swing trajectory. The calculation formula of the characteristic parameters is shown in the following formula (8):

[0097] (8);

[0098] in, Swing angle signal The Fourier transform of is the amplitude spectrum; is the frequency corresponding to the maximum value of the amplitude spectrum (reflecting the main frequency of the swing); Frequency The energy proportion of ); is the energy entropy, which quantifies the uncertainty of the motion (the higher the entropy, the more disordered the oscillations); A small angular perturbation at the initial moment; for The disturbance amplitude at the moment; is the Lyapunov index, if >0 indicates chaotic motion (abnormal oscillation).

[0099] S34: Use support vector machine classifier to perform abnormality judgment. The judgment formula is shown in the following formula (9):

[0100] ) (9),

[0101] Among them, the eigenvector (The four parameters are the swing amplitude, energy entropy, main frequency component, and Lyapunov exponent extracted from formula (8); is the Lagrange multiplier (obtained through training); is the sample label (+1 normal, -1 abnormal); is a kernel function (such as Gaussian kernel) to calculate the similarity between samples; is the bias term (obtained through training); It is a sign function, and the output +1 or -1 indicates the classification result.

[0102] S35: Establish the spatiotemporal correlation matrix between feathers. The matrix is ​​shown in the following formula (10):

[0103] (10),

[0104] For the A feather and The motion correlation coefficient of each feather (such as the mutual correlation coefficient), i and j are both positive integers; When the matrix Rank When the regional abnormality alarm is triggered.

[0105] More specifically, when a regional anomaly alarm is triggered, the edge computing node combines the feather swing trajectory with the pixel color similarity value to establish a joint decision function. The joint decision function is shown in the following formula (11):

[0106] (11),

[0107] in, is the Sigmoid activation function; is the normalized swing amplitude characteristic; is the pixel color similarity value; is the estimated electric field strength vector, and is the trainable weight;

[0108] The priority weight is calculated according to the position of the marked area. The calculation formula is shown in the following formula (12):

[0109] (12),

[0110] in, For the The center coordinates of the labeled area, The center of the feather swing; is the standard deviation of the Gaussian function (controls the weight decay rate), The larger it is, the flatter the weight distribution is (the farther away the area, the higher the weight); The smaller it is, the more weight is concentrated in the center of the feather. nearby;

[0111] The abnormality level function is defined according to the joint decision function, and the definition formula is shown in the following formula (13):

[0112] L= (13),

[0113] The threshold Obtained through training with historical data;

[0114] Finally, the maintenance instruction is generated according to the maintenance instruction parameters. The calculation formula of the maintenance instruction parameters is shown in the following formula (14):

[0115] (14),

[0116] in is the required response time threshold, The index of the marked area (such as number 1 to m); Indicates selection · The area with the largest value As the first inspection point, For the region The joint decision value of (output of formula (11)), For the region The priority weight of (output of formula (12)).

[0117] In step S40, when the edge computing node splits the abnormal data into multiple data packets according to the number of assisting nodes, the following steps are included:

[0118] S41: Establish the node selection scoring function shown in the following formula (15):

[0119] (15),

[0120] in, To the node The round trip delay, is the available bandwidth, is the CPU load rate, is a configurable weight;

[0121] S42: Calculate the optimal number of shards using the following formula (16):

[0122] (16),

[0123] in, is the total amount of abnormal data, is the maximum transmission unit of the network, is the baseline bandwidth requirement;

[0124] S43: Allocate data packet size based on node score. The allocation formula is shown in the following formula (17):

[0125] (17);

[0126] in, To assist in node indexing; For nodes The score of (calculated by formula (15)); The sum of the scores of all assisting nodes; is the total amount of abnormal data to be transmitted; The sum of the CPU load rates of all nodes; For nodes CPU load rate. The higher the value (better performance), the more data is allocated; the load The lower it is, the more data is allocated;

[0127] S44: Fountain code is used for encoding. The encoding formula is shown in the following formula (18):

[0128] (18),

[0129] in, The original data shard index (1≤ ≤k, where k is the total number of shards); is the code packet index (jth code packet); For the Original data shards, is a randomly generated 0 / 1 matrix element ( Package corresponds to coefficient of the shard), .

[0130] This embodiment also includes a distributed edge computing system based on a smart power station scenario that uses a distributed edge computing method based on a smart power station scenario.

[0131] In specific implementation, multiple smart power substations are widely distributed across a city's power supply network. According to the plan, edge computing nodes were deployed in each substation. These nodes act as "smart butlers," continuously collecting sensor monitoring data from power equipment, such as device temperature, current and voltage values, as well as monitoring image data (secondary data), and rapidly processing this information.

[0132] When the station building is operational, the edge computing nodes quickly go to work. They accurately identify pre-annotated areas in the surveillance image, such as key equipment parts, the station building's insulated handles, and the station building's outer shell, and carefully extract the pixel color information from these areas. By comparing the color information of different annotated areas, they can keenly detect subtle changes.

[0133] During data processing, edge computing nodes also handle the statistical analysis of sensor monitoring data. If an anomaly is detected, such as a sudden, significant fluctuation in current, it immediately generates an anomaly signal. Next, it integrates pixel color similarity and the location of the annotated area to quickly generate maintenance instructions and upload them to the cloud. This efficient and accurate process significantly shortens the time from problem discovery to scheduled maintenance.

[0134] It's worth noting that the solution utilizes a target device consisting of a hanging rod and a feather to aid in detection. Under normal circumstances, the feather will naturally sway in the air. However, if an abnormally charged area appears within the station, the electric field changes, causing the feather's swaying trajectory to change. Edge computing nodes utilize binocular vision technology to precisely capture the feather's swaying motion. Then, through a series of intelligent algorithm analysis, they can more accurately identify the abnormally charged area, making the detection results more reliable.

[0135] When a station detects abnormal data that needs to be uploaded, the edge computing node quickly searches for other edge computing nodes with the lowest communication latency to serve as assisting nodes. It then cleverly splits the abnormal data into multiple packets based on the number of assisting nodes and sends them to the assisting nodes. These assisting nodes work together to quickly upload the packets to the cloud, ensuring timely and accurate data transmission and avoiding information loss due to excessive data volume or transmission failures.

[0136] The above is only an embodiment of the present invention, and the common knowledge such as the specific structure and characteristics of the scheme is not described in detail here. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A distributed edge computing method based on the electric power intelligent station scenario is characterized by: The following steps are involved: S10: Deploy an edge computing node in each station building, collect sensor monitoring data of each power device as first data, collect each monitoring image data as second data, and send the first data and the second data to the edge computing node; S20: The edge computing node identifies a preset annotated area in the second data, extracts pixel color information within the annotated area as first information, compares the first information extracted from the current annotated area with the first information extracted from other annotated areas, and calculates pixel color similarity values ​​between the current annotated area and the other annotated areas; S30: The edge computing node performs statistics on the first data and generates an abnormality signal based on the statistical results. After the edge computing node generates the abnormality signal, the edge computing node also extracts the motion trajectory of the target device in the second data. After the edge computing node generates the abnormality signal, the motion trajectory is combined with the pixel color similarity value to determine whether the marked area is abnormally charged, and then a maintenance instruction is generated based on the judgment result and the position of the marked area; the target device includes a hanging rod and a plurality of feathers, the hanging rod is fixedly connected to the station building, and the plurality of feathers are hung one by one on the hanging rod; When the edge computing node identifies the motion trajectory of the target device, it first calculates the swing arc of the feather based on the relative position between the feather and the target area in the second data. Then, it calculates the swing trajectory of the feather based on the swing arc of the feather in combination with the collection time of the second data. The swing trajectory of the feather is used as the motion trajectory of the target device, and the maintenance instruction is uploaded to the cloud. S40: The edge computing node obtains data of other edge computing nodes with the smallest communication delay as an assisting node, splits the first data and the second data corresponding to the abnormal signal into multiple data packets according to the number of assisting nodes, and sends the data packets to the assisting nodes; after receiving the data packets, the assisting node uploads the data packets to the cloud.

2. The distributed edge computing method based on the electric power intelligent station scenario according to claim 1 is characterized in that: When the edge computing node extracts the first information, it performs color space conversion on the pixels in the marked area, converts the RGB value into HSV space, takes the maximum value of the three components of red, green, and blue as the brightness value, and uses the difference between the brightness and the minimum component and the ratio of the brightness to calculate the saturation. The hue is determined according to the component corresponding to the maximum value of the three RGB components, as shown in the following formula (1): （1）。 3. The distributed edge computing method based on the electric power intelligent station scenario according to claim 2 is characterized in that: When calculating the pixel color similarity value of the current annotation area and other annotation areas, for the current annotation area , and each other marked area Compare one by one and calculate the color feature distance between the two areas , The calculation formula is as follows: （3）， in, , , and , , Respectively for regions and The mean of hue, saturation, and value in HSV color space; , , is the weight coefficient of each color component, and satisfies ; According to the region With the current area Calculate spatial distance from pixel coordinates ,according to Setting weights , The calculation formula is shown in the following formula (4): （4）， in, is the distance attenuation coefficient; Combine the similarity and weight of all other annotated areas with the current area to calculate the current area The overall similarity value of , The calculation formula is as follows: （5）。 4. The distributed edge computing method based on the electric power intelligent station scenario according to claim 3 is characterized by: In step S30, the edge computing node further extracts the motion trajectory of the target device in the second data. After generating an abnormal signal, the edge computing node combines the motion trajectory with the pixel color similarity value to determine whether the marked area is abnormally charged, and then generates a maintenance instruction based on the judgment result and the location of the marked area. The target device includes a hanging rod and a plurality of feathers, the hanging rod is fixedly connected to the station building, and the plurality of feathers are hung one by one on the hanging rod; When the edge computing node identifies the motion trajectory of the target device, it first calculates the swing arc of the feather based on the relative position between the feather and the target area in the second data, and then calculates the swing trajectory of the feather based on the swing arc of the feather in combination with the collection time of the second data, and uses the swing trajectory of the feather as the motion trajectory of the target device.

5. The distributed edge computing method based on the electric power intelligent station scenario according to claim 4 is characterized in that: When the edge computing node identifies the motion trajectory of the target device, the following steps are included: S31: For The collaborative detection model is established by the following formula (6): （6）， in, Indicates the A feather in The swing angle at the moment, is the allowed angle deviation threshold; S32: The three-dimensional motion trajectory of the feather is reconstructed based on the three-dimensional coordinates of the binocular vision. The reconstruction formula is shown in the following formula (7): （7）， in is the baseline distance, is the focal length, for parallax; S33: Calculate the dynamic characteristic parameters of the swing trajectory. The calculation formula of the characteristic parameters is shown in the following formula (8): （8）； S34: Use support vector machine classifier to perform abnormality judgment. The judgment formula is shown in the following formula (9): （9）， Among them, the eigenvector ; S35: Establish the spatiotemporal correlation matrix between feathers. The matrix is ​​shown in the following formula (10): （10）， When the matrix Rank When the regional abnormality alarm is triggered.

6. The distributed edge computing method based on the electric power intelligent station scenario according to claim 5 is characterized by: When a regional anomaly alarm is triggered, the edge computing node combines the feather swing trajectory with the pixel color similarity value to establish a joint decision function. The joint decision function is shown in the following formula (11): （11）， in, is the Sigmoid activation function; is the normalized swing amplitude characteristic; is the pixel color similarity value; is the estimated electric field strength vector, and is the trainable weight; The priority weight is calculated according to the position of the marked area. The calculation formula is shown in the following formula (12): （12）， in, For the The center coordinates of the labeled area, The center of the feather swing; The abnormality level function is defined according to the joint decision function, and the definition formula is shown in the following formula (13): L= （13）， The threshold Obtained through training with historical data; Finally, the maintenance instruction is generated according to the maintenance instruction parameters. The calculation formula of the maintenance instruction parameters is shown in the following formula (14): （14）， in is the required response time threshold.

7. The distributed edge computing method based on the electric power intelligent station scenario according to claim 6 is characterized in that: In step S40, the edge computing node splits the abnormal data into multiple data packets according to the number of assisting nodes, including the following steps: S41: Establish the node selection scoring function shown in the following formula (15): （15）， in, To the node The round trip delay, is the available bandwidth, is the CPU load rate, is a configurable weight; S42: Calculate the optimal number of shards using the following formula (16): （16）， in, is the total amount of abnormal data, is the maximum transmission unit of the network, is the baseline bandwidth requirement; S43: Allocate data packet size based on node score. The allocation formula is shown in the following formula (17): （17）； S44: Fountain code is used for encoding. The encoding formula is shown in the following formula (18): （18）， in, Shard the original data. To randomly generate matrix elements, .

8. The distributed edge computing system based on the power intelligent station scene is characterized by: The distributed edge computing method based on the electric power smart station scenario described in any one of claims 1-7 is used.

Images