CN110677300B - Electric power safety management video intelligent shunting device and method based on mobile edge calculation - Google Patents
Electric power safety management video intelligent shunting device and method based on mobile edge calculation Download PDFInfo
- Publication number
- CN110677300B CN110677300B CN201910953248.2A CN201910953248A CN110677300B CN 110677300 B CN110677300 B CN 110677300B CN 201910953248 A CN201910953248 A CN 201910953248A CN 110677300 B CN110677300 B CN 110677300B
- Authority
- CN
- China
- Prior art keywords
- module
- shunting
- value
- evaluation module
- calculation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000004364 calculation method Methods 0.000 title claims abstract description 63
- 238000000034 method Methods 0.000 title claims abstract description 31
- 238000011156 evaluation Methods 0.000 claims abstract description 85
- 238000004458 analytical method Methods 0.000 claims abstract description 81
- 238000005457 optimization Methods 0.000 claims abstract description 55
- 238000005265 energy consumption Methods 0.000 claims abstract description 41
- 230000011218 segmentation Effects 0.000 claims abstract description 39
- 238000012821 model calculation Methods 0.000 claims abstract description 30
- 238000012544 monitoring process Methods 0.000 claims abstract description 30
- 230000008859 change Effects 0.000 claims abstract description 29
- 238000009826 distribution Methods 0.000 claims abstract description 28
- 238000001514 detection method Methods 0.000 claims abstract description 27
- 238000007906 compression Methods 0.000 claims abstract description 19
- 230000006835 compression Effects 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 15
- 230000005540 biological transmission Effects 0.000 claims description 73
- 238000004422 calculation algorithm Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 13
- 238000003860 storage Methods 0.000 claims description 13
- 238000005562 fading Methods 0.000 claims description 10
- 239000013307 optical fiber Substances 0.000 claims description 9
- 238000013135 deep learning Methods 0.000 claims description 7
- 238000013210 evaluation model Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 2
- 238000012549 training Methods 0.000 description 18
- 238000012360 testing method Methods 0.000 description 12
- 230000000694 effects Effects 0.000 description 8
- 238000010521 absorption reaction Methods 0.000 description 4
- 238000013136 deep learning model Methods 0.000 description 4
- 238000007689 inspection Methods 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 238000004519 manufacturing process Methods 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000009430 construction management Methods 0.000 description 2
- 230000001934 delay Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010248 power generation Methods 0.000 description 1
- 230000003449 preventive effect Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 231100000331 toxic Toxicity 0.000 description 1
- 230000002588 toxic effect Effects 0.000 description 1
- 238000009423 ventilation Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 238000002759 z-score normalization Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/10—Architectures or entities
- H04L65/1013—Network architectures, gateways, control or user entities
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/08—Configuration management of networks or network elements
- H04L41/0803—Configuration setting
- H04L41/0823—Configuration setting characterised by the purposes of a change of settings, e.g. optimising configuration for enhancing reliability
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/12—Discovery or management of network topologies
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/60—Network streaming of media packets
- H04L65/75—Media network packet handling
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/02—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP]
- H04L67/025—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP] for remote control or remote monitoring of applications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1001—Protocols in which an application is distributed across nodes in the network for accessing one among a plurality of replicated servers
- H04L67/1004—Server selection for load balancing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Multimedia (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Remote Monitoring And Control Of Power-Distribution Networks (AREA)
- Image Analysis (AREA)
Abstract
The invention provides an electric power safety management video intelligent shunting device and method based on mobile edge calculation, wherein the device comprises a shunting strategy optimization evaluation module, a scene detection module, a dynamic change region segmentation module, a model calculation overhead evaluation module, a model energy consumption overhead evaluation module, an equipment state monitoring unit, a local analysis module and a compression shunting module, wherein the scene detection module, the dynamic change region segmentation module, the model calculation overhead evaluation module, the model energy consumption overhead evaluation module, the equipment state monitoring unit, the local analysis module and the compression shunting module are connected with the shunting strategy optimization evaluation module; the shunting strategy optimization evaluation module is used for storing the scene detection and region segmentation pictures, determining and optimizing a shunting strategy, and shunting the shunting strategy to the local analysis module, the adjacent edge analysis unit or the video processing cloud analysis unit for intelligent identification. The method can adaptively adjust the distribution strategy according to the running state of the edge equipment, realize the convergence of multi-channel videos, compress distribution contents, optimize the distribution strategy and schedule distribution tasks by moving the edge equipment, solve the video time delay, balance the energy consumption of the terminal, reduce the flow overhead and improve the management efficiency.
Description
Technical Field
The invention relates to the field of power systems and video safety management, in particular to a power safety management video intelligent shunting device and method based on mobile edge calculation.
Background
The safe and stable operation of the power has a particularly important meaning, and relates to each link of power generation, power transmission, power transformation, power distribution, power utilization and the like, which is determined by the characteristics of power production and the serious consequences of power grid accidents, the power grid accidents cause the power failure of users, and the economic and political influences are very great. With the development of society, the demand for electric energy is increasing day by day, and various devices of electric power are promoted to grow continuously, so that the monitoring of electric power is strengthened, and the method is a main preventive measure for ensuring the safe and stable operation of electric power.
Due to the characteristics of high risk, safety risk and the like of an electric power operation place, the intelligent inspection system of the inspection robot and the transmission line unmanned helicopter and the access of mobile video equipment such as vehicles and individual soldiers can replace manual work to be on duty, and the security and protection requirements of monitoring and inspecting the inside and the outside of the transformer substation anytime and anywhere are met. Especially in the electric power tunnel with water accumulation, dark light, insufficient ventilation and even toxic and harmful gas, the inspection robot can play an important application value, can replace personnel to carry out conventional inspection, ensures the normal work of equipment in the electric power tunnel, and prevents illegal behaviors such as stealing, private wire pulling and the like. The national power grid introduces a unified video monitoring platform of a power system to realize comprehensive video access of a transformer substation, a power transmission line, an information communication machine room, a business place and the like, so that manual duty is replaced. In consideration of the problems of limited coverage, insufficient bandwidth, poor flexibility and the like of the existing video monitoring system, the electric power safety construction management gradually adopts a more flexible portable field operation monitoring terminal as supplement, and video return is realized through a 3G/4G wireless network. In addition, the rear-end coding processing equipment of the electric power safety construction management system must integrate and uniformly manage all subsystem information such as video data, environmental data, power data, alarm data and the like, linkage alarm among data is realized, an omnibearing three-dimensional safety control system for realizing equipment safety, personnel safety and environmental safety of a transformer substation is established, and the safe, stable and reliable operation of the whole intelligent power grid system can be finally guaranteed.
In view of the problems of mobile network coverage, network transmission delay, network stability, network congestion degree and the like, the adoption of mobile video equipment for monitoring inevitably causes transmission delay, transmission interruption, transmission energy consumption and flow cost, and affects video quality, safety management efficiency, operation cost and video analysis performance.
Disclosure of Invention
Aiming at the defects of video distribution through mobile edge equipment and the requirements of power safety production management monitoring at present, the invention provides an intelligent video distribution device and method for power safety management based on mobile edge computing.
In order to realize the purpose, the invention adopts the following technical scheme:
a power safety management video intelligent shunting device based on mobile edge calculation comprises a shunting strategy optimization evaluation module, a scene detection module, a dynamic change region segmentation module, a model calculation overhead evaluation module, a model energy consumption overhead evaluation module, an equipment state monitoring unit, a local analysis module and a compression shunting module, wherein the scene detection module, the dynamic change region segmentation module, the model calculation overhead evaluation module, the model energy consumption overhead evaluation module, the device state monitoring unit and the compression shunting module are connected with the shunting strategy optimization evaluation module; the shunting strategy optimization evaluation module is used for storing the scene detection and region segmentation pictures and calculating the cost S by using a target identification model1Energy consumption overhead, available computing resource value T of edge computing device1Storing resource value T2Power supply state value T3And a network state value T4Determining and optimizing a shunting strategy according to the information, shunting the shunting strategy to a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit for intelligent identification, wherein the energy consumption overhead comprises transmission loss of a wired local area network
And fading Lbf of electromagnetic waves of the wireless wide area network in a transmission path.
Further, the optimization objective of the shunting strategy is as follows:
wherein the value of theta is 1 or 0, indicating selection or non-selection of the local analysis module, Si1Is the target recognition model calculation cost of the ith device, S1The calculation cost of the target identification model is the device cost grade value and the available calculation resource value T transmitted by the model calculation cost evaluation module1Storing resource value T2Power supply state value T3Network state value T4Is obtained by a device state monitoring unit; t isi1Is the value of the available computing resource, T, of the ith devicei2Is the storage resource value, T, of the ith devicei3Is the power supply state value, T, of the ith devicei4Is the network status value of the ith device; the Loss function describes the transmission Loss of the network, and utilizes the calculation formula of the transmission Loss of the electromagnetic wave in the wired local area network and the fading formula of the electromagnetic wave in the free space, wherein
Attenuation of electromagnetic waves transmitted along optical fibers, L, is described for transmission loss in wired local area networksjDevice-to-device distances are described, corresponding to edges in the network topology graph; and selecting K devices in the topological graph until the three target functions reach the optimal values, namely completing the task of making a shunting strategy, wherein the K devices are randomly selected and comprise a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit.
Further, the scene detection module is configured to extract key features from a video frame background according to the power safety management characteristics and differences between the management and control target objects and management and control tasks in different scenes, determine a current operation scene, detect a target category in the current scene, and feed a target category detection result back to the shunting policy optimization evaluation module to reduce the number of target identification classifications.
Further, the dynamic change region segmentation module is configured to implement dynamic change identification and region segmentation through background and foreground differences, forward a plurality of segmented change region pictures to the distribution policy optimization evaluation module, and the distribution policy optimization module selects K analysis units according to an optimization target and distributes the region segmentation pictures to each unit for analysis.
Further, the model calculation overhead evaluation module is used for measuring and constructing a model calculation overhead evaluation model according to historical classification performance on the equipment, establishing a calculation overhead performance evaluation table, describing corresponding calculation time delays under the conditions of different identification algorithms, instruction sets, to-be-identified category sizes, types and quantities and electric quantity grades, and taking the time delay corresponding to the performance index with the highest correlation coefficient absolute value in the table as the estimated calculation overhead S1And the value is transmitted to a shunting strategy optimization evaluation module.
Further, the model energy consumption overhead evaluation module is used for calculating local area network transmission energy consumption and wide area wireless network transmission energy consumption.
Further, the device state monitoring unit is configured to monitor the remaining power, the available computational resources, the transmission rate, and the analysis delay on the edge device with a fixed sampling period, and specifically, sample the device with a specified period T to obtain the remaining power Q (unit: degree), the remaining memory SP (unit: MB), the available computational resource U (unit: MiB) of the CPU/GPU, the device transmission rate P (unit: Mbps), and the analysis delay Danl(units: sheets/second), available computing resources T1I.e. U, storage resource T2SP, power supply state T3Represented by a residual capacity Q, network state T4Expressed by the transmission rate P.
Furthermore, the local analysis module is used for receiving scene pictures, target classification and dynamic segmentation results, and identifying and classifying the safety management events by adopting a deep learning algorithm; and the compression and distribution module is used for transferring the scene picture, the target classification and the dynamic segmentation result set to adjacent edge equipment or a cloud server for event identification and classification according to a distribution strategy.
A power safety management video intelligent distribution method based on mobile edge calculation is characterized by comprising the following steps: the method is carried out by adopting the device and comprises the following steps:
step 1: the scene detection module extracts key features from the background of the video picture, determines the current operation scene, and feeds the result back to the shunting strategy optimization evaluation module to reduce the number of target identification classifications and reduce the amount of analysis tasks of edge equipment;
step 2: the dynamic change region segmentation module analyzes various safety events and hidden dangers from the dynamic change of a video picture, realizes dynamic change identification and region segmentation through background and foreground difference on the basis of a reference scene background, and transfers a plurality of segmented change regions to a shunting strategy unit for evaluation and analysis;
and step 3: obtaining calculation cost through a model calculation cost evaluation module, measuring and constructing a model calculation cost evaluation model according to historical classification performance on the equipment by the module, establishing a calculation cost performance evaluation table, describing corresponding calculation time delay under the conditions of different identification algorithms, instruction sets, category sizes, types and quantities to be identified and electric quantity grades, obtaining the calculation time delay of the current state of the equipment according to the table, and transmitting the calculation time delay to a shunting strategy optimization evaluation module;
and 4, step 4: obtaining model energy consumption overhead through a model energy consumption overhead evaluation unit module, and transmitting the energy consumption overhead to a shunting strategy optimization evaluation module;
and 5: through equipment state monitoring unit, establish equipment state grade table, acquire equipment state through looking up the table: available computing resources T1I.e. U, storage resource T2SP, power supply state T3Represented by a residual capacity Q, network state T4Expressed by a transmission rate P and transmitted to a shunting strategy optimization evaluation module;
step 6: processing data transmitted by other modules through a shunting strategy optimization evaluation module so as to decide a shunting strategy, and shunting a picture analysis task to a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit for intelligent identification;
and 7: and the local analysis module analyzes the transmitted data.
Further, the optimization objective of the shunting strategy is as follows:
wherein the value of theta is 1 or 0, indicating selection or non-selection of the local analysis module, Si1Is the target recognition model calculation cost of the ith device, S1The calculation cost of the target identification model is the device cost grade value and the available calculation resource value T transmitted by the model calculation cost evaluation module1Storage resource value T2Power supply state value T3Network state value T4Is obtained by a device state monitoring unit; t isi1Is the value of the available computing resource, T, of the ith devicei2Is the storage resource value, T, of the ith devicei3Is the power supply state value, T, of the ith devicei4Is the network status value of the ith device; the Loss function describes the transmission Loss of the network, and utilizes the calculation formula of the transmission Loss of the electromagnetic wave in the wired local area network and the fading formula of the electromagnetic wave in the free space, wherein
Electromagnetism is described for transmission loss in wired local area networksAttenuation of wave transmission along optical fiber, LjDevice-to-device distances are described, corresponding to edges in the network topology graph; and selecting K devices in the topological graph until the three target functions reach the optimal values, namely completing the task of making a shunting strategy, wherein the K devices are randomly selected and comprise a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit.
In general, compared with the prior art, the above technical solutions conceived by the present invention can achieve the following beneficial effects:
1. the convergence of multi-channel videos, the compression of shunting content, the optimization of shunting strategies and the scheduling of shunting tasks are realized through the mobile edge equipment, the video time delay is solved, the energy consumption of a terminal is balanced, the flow overhead is reduced, and the management efficiency is improved;
2. the invention provides a comprehensive monitoring system combined with portable mobile video equipment for solving the comprehensive monitoring of national power grid transformer substations, power transmission lines, information communication machine rooms, business places and the like, and the system is simple and complete and can meet the supervision tasks under multiple scenes;
3. the innovation point of the intelligent shunting strategy is that the intelligent shunting strategy ensures the stability, real-time performance and accuracy of the system, reduces network transmission delay, transmission interruption and transmission energy consumption, can distribute tasks to other edge devices and cloud ends, and is high in speed, high in stability and low in power consumption;
4. the invention also has the innovation points that the deep learning is applied to the power grid scene, and the high detection accuracy of the deep learning algorithm is utilized to improve the detection accuracy of the power grid safety event;
5. in order to reduce the task of model analysis, the invention provides a region segmentation module, and an interested security event and hidden danger picture are dynamically segmented from a video picture, so that the shunting burden is reduced;
6. in order to reduce the target identification and classification tasks, the invention provides a scene detection module for detecting the scene of the image from the dynamic image in advance, thereby improving the accuracy of the detection task and reducing the burden of the detection task.
Drawings
FIG. 1 is a schematic structural diagram of an electric power safety management video intelligent shunting device based on mobile edge calculation according to the present invention;
FIG. 2 is a schematic flow chart of the intelligent video distribution method for power security management based on mobile edge calculation according to the present invention;
FIG. 3 is a network topology diagram of a video analytics device constructed in accordance with the present invention;
FIG. 4 is a device status level table constructed in accordance with the device historical status in accordance with the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The embodiment of the invention provides an intelligent shunting device for electric power safety management video based on mobile edge calculation, which adaptively adjusts shunting strategies according to the running states of edge equipment, and comprises a shunting strategy optimization evaluation module, a scene detection module, a dynamic change region segmentation module, a model calculation overhead evaluation module, a model energy consumption overhead evaluation module, an equipment state monitoring unit, a local analysis module and a compression shunting module, wherein the scene detection module, the dynamic change region segmentation module, the model calculation overhead evaluation module, the model energy consumption overhead evaluation module, the equipment state monitoring unit, the local analysis module and the compression shunting module are connected with the shunting strategy optimization evaluation module, as shown in figure 1. The function and function of each module will be described in detail below.
The shunting strategy optimization evaluation module:
the module is a core module of the branching frame and is used for storing pictures of scene detection and region segmentation and calculating the cost S by using a target identification model1Power consumption overhead (including transmission losses in wired local area networks)
And fading Lbf of electromagnetic waves of the wireless wide area network in the transmission path), edge computing devices (in particular with LCS edge)Computing device, Link Edge computing device, etc.) of the available computing resource value T1Storing resource value T2Power supply state value T3And a network state value T4For the basis, a shunting strategy is determined and optimized, and shunted to a Local _ Block (corresponding to a Local edge computing device) and a neighboring edge analysis unit Neighbor _ Block ═ Block { (Block)1、Block2、...、BlockMPerforming intelligent identification on (corresponding to a mobile edge gateway in fig. 1) or a video processing Cloud analysis unit Cloud _ Block (corresponding to a video analysis Cloud server in fig. 1) to reduce network transmission overhead and traffic to the maximum extent, and balancing load and energy consumption of edge equipment.
The shunting strategy is as follows:
1. according to the distribution condition of the equipment, constructing a network topological graph as shown in FIG. 3, wherein the weight of the edge is described by distance;
2. the optimization target of the shunting strategy is as follows:
wherein the value of theta is 1 or 0, indicating selection or non-selection of the local analysis module, Si1Is the target recognition model calculation cost of the ith device, S1The calculation cost of the local target recognition model is a device cost grade value, T, transmitted by the model calculation cost evaluation modulei1Is the value of the available computing resource, T, of the ith devicei2Is the storage resource value, T, of the ith devicei3Is the power supply state value, T, of the ith devicei4Is the network status value of the ith device; locally available computing resource value T1Local storage resource value T2Local power supply state value T3Local network state value T4These values are obtained by the device status monitoring unit. The Loss function describes the transmission Loss of the network, and utilizes the calculation formula of the transmission Loss of the electromagnetic wave in the wired local area network and the fading formula of the electromagnetic wave in the free space, wherein
Attenuation of electromagnetic waves transmitted along optical fibers, L, is described for transmission loss in wired local area networksjDevice-to-device distances are depicted, corresponding to the edges in FIG. 3; and selecting K devices in the topological graph (randomly selecting K devices including local analysis devices and cloud analysis devices) until the three target functions reach the optimal values, and finishing the task of making the shunting strategy.
3. And compressing the data by using a compression unit, and then transmitting the data to the K devices for analysis according to a shunting strategy.
A scene detection module: according to the characteristics of electric power safety management and the differences of the management and control target objects and management and control tasks in different scenes, key features are extracted from a video picture background, the current operation scene is determined, the target category in the current scene is detected, and the target category detection result is fed back to a shunting strategy optimization evaluation module to reduce the number of target identification classifications and reduce the amount of analysis tasks of edge equipment.
The method for determining the operation scene comprises the following steps:
1. the power grid operation scene comprises: the device comprises a transformer, a high-voltage switch, a lightning arrester, a disconnecting link, a pole tower, a power transmission line, a transformer substation, a ring main unit, a branch box and a cable well ditch;
2. making a data set under each scene, and identifying scene markers;
3. making a deep learning model, and training the deep learning model by using scene data;
4. and identifying scenes in the current video stream by using the trained deep learning model, classifying video pictures belonging to the same scene into the same data, and transmitting the data to the shunting strategy optimization evaluation module.
A dynamic change region segmentation module: in the process of monitoring the electric power safety management, various safety events and hidden dangers only need to be analyzed from dynamic changes in a video picture. Based on a reference scene background, dynamic change identification and region segmentation are realized through background and foreground difference, a plurality of segmented change region pictures are transferred to a distribution strategy optimization evaluation module, the distribution strategy optimization module selects K analysis units according to an optimization target, and the region segmentation pictures are distributed to each unit for analysis.
The region segmentation method using region growing is as follows:
1. selecting n pixel points in the image as Seed points Seed ═ Sd1、Sd2、...、Sdn}；
2. At the seed point SdiSelecting 8 nearest pixel points as points to be expanded, calculating the similarity between the points and field pixel points, and if the similarity is smaller than a threshold value Thrd, determining that the pixels belong to the same region of the seed points, and forming n regions through the step;
3. and expanding the n regions, gradually expanding the points outside the regions into the regions according to the expanding mode of the step 2, and stopping growing the regions until the similarity value between the pixel points outside the regions and the pixel points in the regions is no longer smaller than a threshold value Thrd.
Specifically, the similarity determination criterion in step 2 is as follows:
1. subjecting RGB space to nonlinear transformation to obtain YCbCrThe color space is specifically as follows:
2. calculating similarity by using color characteristic values of pixel points, taking a certain pixel point A outside the area as an example, taking a pixel area formed by m points nearest to the point A, and calculating a local average value of the area Y
CbLocal partMean value of
CrLocal mean of
Calculating the color similarity between the pixel point A and the pixel point B in the region by using the following formula;
by using a region growing algorithm and pixel color characteristics, the picture can be divided into a plurality of different regions, and the divided regions are transmitted to a shunting strategy optimization evaluation module.
A model calculation overhead evaluation module: the module measures and constructs a model calculation overhead evaluation model according to historical classification performance on the equipment, and establishes a calculation overhead performance evaluation table which describes corresponding calculation time delays under the conditions of different identification algorithms, instruction sets, to-be-identified category sizes, types, quantities and electric quantity grades.
The specific calculation method is as follows:
1. calculating expense is estimated by combining the performance during historical classification, and the performance indexes comprise: recognition algorithm MT { MT1、MT2、...MTTThe method comprises the steps of (1) calculating time delay D, the number Num of instruction sets, the Size of classes to be identified, the number Class of classes to be identified, the number All of targets to be identified and the residual electric quantity Q of equipmentcal；
2. And constructing a calculation overhead performance evaluation table according to the indexes, wherein data in the table is used for calculating the similarity, so normalization is needed to form dimensionless data, and the normalization is specifically carried out by adopting a z-score method:
wherein x represents each performance index in the table,
denotes the index mean, σ is the index variance, and x' is the index value after normalization.
3. Obtaining each performance index of the current equipment to obtain a performance index vector u ═ MTiNum ', Size ', Class ', All ', Q ' are normalized according to the formula in step 2, then according to the following formula, the similarity S (u, v) between the vector and each row in the table is calculated by using the Pearson correlation coefficient, and the time delay corresponding to the performance index with the highest absolute value of the correlation coefficient in the table is taken as the estimated calculation cost S1And the value is transmitted to a shunting strategy optimization evaluation module.
Wherein r isu，iAn i-th index value representing u,
means, r, representing the u index vectorv，iThe value of the i-th index representing a certain row v in the table,
represents the mean of the v-index vector. s (u, v) has a value range of [ -1, 1 [ ]]The larger the absolute value, the higher the correlation of u with v.
A model energy consumption overhead evaluation module: under the framework, the energy consumption overhead specifically comprises the calculation of local area network transmission energy consumption, wide area wireless network transmission energy consumption and the like. Different shunting strategies cause large energy consumption difference, but considering that signals are influenced by various factors in the transmission process, the shunting strategy needs to balance several kinds of energy consumption in the process of making the shunting strategy so as to reduce the loss in the data transmission process as much as possible.
The specific energy consumption calculation method comprises the following steps:
1. in a Local Area Network (LAN), there are many factors that affect the energy consumption of data transmission, such as: the optical fiber absorption loss, intrinsic absorption loss, scattering loss and the like all cause signals to generate different signal strengths when reaching different receiving ends, so that small transmission loss needs to be selected to ensure that the signals are not distorted, and the calculation formula of the transmission loss of the wired local area network is as follows:
2. There are many ways to consume transmission energy in a Wireless Wide Area Network (WWAN), for example: shadow effect, multipath effect, near-far effect, Doppler effect and the like, and the free space loss is used for describing the fading of the electromagnetic wave of the wireless wide area network in a transmission path, and the specific calculation formula is as follows:
Lbf＝32.5+20lgF+20lgD
where Lbf describes the loss of electromagnetic waves, in dB; f is the electromagnetic wave transmission frequency, in MHZ; d is the electromagnetic wave transmission distance, in km.
An equipment state monitoring unit: the monitoring unit monitors the residual electric quantity, available computing resources, transmission rate and analysis time delay on the edge equipment in a fixed sampling period, and the shunting unit adaptively adjusts a scheduling strategy according to the equipment state.
Specifically, sampling is carried out on the equipment in a specified period T, and the residual electric quantity Q (unit: degree), the residual memory SP (unit: MB), the available computing resource U (unit: MiB) of the CPU/GPU, the transmission rate P (unit: Mbps) of the equipment and the analysis time delay D are obtainedanl(unit: sheet/sec). Available computing resources T1I.e. U, storage resource T2SP, power supply state T3Represented by a residual capacity Q, network state T4Expressed by the transmission rate P, for convenience T1、T2、T3、T4And (4) participating in calculation, namely, grading the state of the equipment and participating in calculation by using the grade value.
Specifically, the value ranges of U, SP, Q and P are obtained according to the historical state of the equipment, and are respectively used as [ Umin and Umax]、[SPmin，SPmax]、[Qmin，Qmax]、[Pmin，Pmax]The expression is that the interval is divided into N parts, and the corresponding grades are [1, 2., N]And making a grade table, wherein the specific table is shown in FIG. 4, and after the current state of the equipment is obtained, the table is looked up to obtain T1、T2、T3、T4And transmitting the grade value to a shunting strategy optimization evaluation module.
A local analysis module: and local analysis receives scene pictures, target classification and dynamic segmentation results, and the safety management events are identified and classified by adopting a deep learning algorithm.
Specifically, the local identification process is as follows:
1. for different scenes, determining the target of the safety management event to be detected as { Obj1、Obj2、......、ObjmAccording to the standard of the VOC data set, a training set Train (the number of pictures is set to be L1), a verification set Val (the number of pictures is set to be L2), and a Test set Test (the number of pictures is set to be L3) are made, so that L1: L2: L3 is approximately equal to 6: 2;
2. selecting a proper target detection algorithm, setting training parameters, inputting a training set and a verification set into a deep learning network for training, obtaining a training model through N training rounds, using a test set test model, and finishing model training when the average accuracy of the test, Arc ═ Thrd; when the test average accuracy Arc < Thrd, the training parameters are adjusted and retrained until when the test average accuracy Arc > -Thrd.
3. Configuring the running environment of the local equipment, configuring the model to the local equipment, and identifying and classifying the scene picture and the dynamic segmentation picture.
A compression shunting module: and compressing and shunting, namely transferring the scene picture, the target classification and the dynamic segmentation result set to adjacent edge equipment or a cloud server for event identification and classification according to a shunting strategy.
Specifically, as shown in fig. 3, the compression and splitting method is as follows:
1. compressing the scene picture, the target classification and the dynamic segmentation result set according to scenes to obtain compressed data G { G1, G2,. and Gn }, wherein n is the number of scenes, and Gi is the compression result of the scene picture, the target classification and the dynamic segmentation result set under the scene i;
2. the specific compression method can adopt a Run-Length Encoding method, pictures contain the same color in many places, so that many same data exist, and the adoption of the Run-Length Encoding method can reduce the data transmission amount.
As shown in fig. 2, an embodiment of the present invention further provides an electric power safety management video intelligent shunting method based on mobile edge calculation, which can be performed by using the foregoing apparatus, and the method includes the following steps:
step 1: the scene detection module extracts key features from the background of the video picture, determines the current operation scene, and feeds the result back to the shunting strategy optimization evaluation module to reduce the number of target identification classifications and reduce the amount of analysis tasks of edge equipment;
specifically, the specific steps of step 1 are as follows:
(1-1) the power grid operation scene comprises the following steps: the device comprises a transformer, a high-voltage switch, a lightning arrester, a disconnecting link, a pole tower, a power transmission line, a transformer substation, a ring main unit, a branch box and a cable well ditch;
(1-2) making a data set under each scene, and identifying scene markers;
(1-3) manufacturing a deep learning model, and training the model by using scene data;
and (1-4) identifying scenes in the current video stream by using the trained model, classifying video pictures belonging to the same scene into the same data, and transmitting the data to the shunting strategy module.
And 2, step: the dynamic change region segmentation module analyzes various safety events and hidden dangers from the dynamic change of the video picture. Based on the background of a reference scene, dynamic change identification and region segmentation are realized through the difference between the background and the foreground, and a plurality of segmented change regions are transferred to a shunting strategy unit for evaluation and analysis;
specifically, the specific steps of step 2 are as follows:
(2-1) selecting n pixel points in the image as Seed points Seed ═ Sd1、Sd2、...、Sdn}；
(2-2) at the seed Point SdiSelecting 8 nearest pixel points as points to be expanded, calculating the similarity between the points and field pixel points, and if the similarity is smaller than a threshold value Thrd, determining that the pixels belong to the same region of the seed points, and forming n regions through the step;
(2-3) expanding the n regions, gradually expanding the points outside the regions into the regions according to the expanding mode of the step 2, and stopping growing the regions until the similarity value between the pixel points outside the regions and the pixel points in the regions is no longer smaller than a threshold value Thrd.
Specifically, the similarity determination criterion in step (2-2) is as follows:
(2-2-1) subjecting the RGB space to nonlinear transformation to obtain YCbCrThe color space is specifically as follows:
(2-2-2) calculating similarity by using color characteristic values of pixel points, taking a certain pixel point A outside the area as an example, taking a pixel area formed by m points nearest to A, and calculating a local average value of the area Y
CbLocal mean value
CrLocal mean of
Calculating the color similarity between the pixel point A and the pixel point B in the region by using the following formula;
by using a region growing algorithm and pixel color characteristics, the picture can be divided into a plurality of different regions, and the divided regions are transmitted to a shunting strategy optimization evaluation module.
And step 3: obtaining calculation cost through a model calculation cost evaluation module, measuring and constructing a model calculation cost evaluation model according to historical classification performance on the equipment by the module, establishing a calculation cost performance evaluation table, describing corresponding calculation time delay under the conditions of different identification algorithms, instruction sets, category sizes, types and quantities to be identified and electric quantity grades, obtaining the calculation time delay of the current state of the equipment according to the table, and transmitting the calculation time delay to a shunting strategy optimization evaluation module;
specifically, the specific steps of step 3 are as follows:
(3-1) estimating the calculation cost by combining the performance in the historical classification, wherein the performance index comprises: recognition algorithm MT { MT1、MT2、...MTTThe method comprises the steps of (1) calculating time delay D, the number Num of instruction sets, the Size of classes to be identified, the number Class of classes to be identified, the number All of targets to be identified and the residual electric quantity Q of equipmentcal；
(3-2) constructing a calculation overhead performance evaluation table according to the indexes, wherein since the data in the table is used for calculating the similarity, normalization is needed to form dimensionless data, and specifically, z-score normalization is adopted:
(3-3) obtaining each performance index of the current equipment to obtain a performance index vector u ═ { MTiNum ', Size ', Class ', All ', Q ' are normalized first, then according to the following formula, the similarity between the vector and each line in the table is calculated by using the Pearson correlation coefficient, and the time delay corresponding to the performance index with the highest absolute value of the correlation coefficient in the table is taken as the time delay corresponding to the performance indexPre-estimated computational cost S1And the value is transmitted to a shunting strategy optimization evaluation module.
Wherein r isu，iAn i-th index value representing u,
means, r, representing the u index vectorv，iThe value of the i-th index representing a certain row v in the table,
represents the mean of the v-index vector. The value range of the correlation coefficient is [ -1, 1]. The larger the absolute value of the correlation coefficient, the higher the correlation of u with v.
And 4, step 4: obtaining model energy consumption overhead through a model energy consumption overhead evaluation unit module, and transmitting the energy consumption overhead to a shunting strategy optimization evaluation module;
specifically, the specific steps of step 4 are as follows:
(4-1) in a Local Area Network (LAN), there are many factors that affect the energy consumption of data transmission, for example: the optical fiber absorption loss, intrinsic absorption loss, scattering loss and the like all cause signals to generate different signal strengths when reaching different receiving ends, so that small transmission loss needs to be selected to ensure that the signals are not distorted, and the calculation formula of the transmission loss of the wired local area network is as follows:
(4-2) transmission energy consumption in a Wireless Wide Area Network (WWAN) has many methods, for example: shadow effect, multipath effect, near-far effect, Doppler effect and the like, and the free space loss is used for describing the fading of the electromagnetic wave of the wireless wide area network in a transmission path, and the specific calculation formula is as follows:
Lbf＝32.5+20lgF+20lgD
where Lbf describes the loss of electromagnetic waves in dB; f is the electromagnetic wave transmission frequency, in MHZ; d is the electromagnetic wave transmission distance, in km.
And 5: an equipment state grade table is constructed through an equipment state monitoring unit, the equipment state is obtained through table lookup and is transmitted to a shunting strategy optimization evaluation module;
specifically, the specific steps of step 5 are as follows:
(5-1) sampling the equipment in a specified period T to obtain the residual electric quantity Q (unit: degree), the residual memory SP (unit: MB), the available computing resource U (unit: MiB) of the CPU/GPU, the transmission rate P (unit: Mbps) of the equipment and the analysis delay Danl(unit: sheet/second).
(5-2) available computing resources T1I.e. U, storage resource T2SP, power supply state T3Network state T represented by remaining capacity Q4Expressed by the transmission rate P, for convenience T1、T2、T3、T4And (4) participating in calculation, namely, grading the state of the equipment and participating in calculation by using the grade value.
(5-3) specifically, acquiring value ranges of U, SP, Q and P according to the historical state of the equipment, wherein the value ranges are respectively used as [ Umin and Umax]、[SPmin，SPmax]、[Qmin，Qmax]、[Pmin，Pmax]The expression is that the interval is divided into N parts, and the corresponding grades are [1, 2., N]And making a grade table, wherein the specific table is shown in FIG. 4, and after the current state of the equipment is obtained, the table is looked up to obtain T1、T2、T3、T4The grade value of the flow distribution strategy is transmitted to a flow distribution strategy optimization evaluation module.
And 6: processing data transmitted by other modules through a shunting strategy optimization evaluation module so as to decide a shunting strategy, and shunting a picture analysis task to a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit for intelligent identification;
specifically, the specific steps of step 6 are as follows:
(6-1) constructing a network topology graph as shown in FIG. 3 according to the distribution condition of the devices, wherein the weights of the edges are described by distances;
(6-2) the optimization target of the shunting strategy is as follows:
and selecting K devices in the topological graph until the three target functions reach the optimal values, namely finishing the strategy formulation task.
And (6-3) calling a compression and distribution unit to compress the data, dividing the data into K parts, and transmitting the K parts to each analysis unit for analysis.
Specifically, the compression and diversion unit in the step (6-3) includes the following steps:
(6-3-1) compressing the scene picture, the target classification and the dynamic segmentation result set according to the scene to obtain compressed data G { G1, G2,. and Gn }, wherein n is the number of scenes, and Gi is the compression result of the scene picture, the target classification and the dynamic segmentation result set under the scene i;
(6-3-2) the specific compression method may adopt a Run-Length Encoding method, pictures will contain the same color in many places, and thus there will be many same data, and the use of the Run-Length Encoding method can reduce the amount of transmission data.
And 7: and the local analysis module analyzes the transmitted data.
Specifically, the specific steps of step 7 are as follows:
(7-1) determining that the target of the to-be-detected safety management event is { Obj1、Obj2、......、ObjmAccording to the standard of the VOC data set, a training set Train (the number of pictures is set to be L1), a verification set Val (the number of pictures is set to be L2), and a Test set Test (the number of pictures is set to be L3) are made, so that L1: L2: L3 is approximately equal to 6: 2;
(7-2) selecting a proper target detection algorithm, setting training parameters, inputting a training set and a verification set into a deep learning network for training, obtaining a training model through N training rounds, using a test set test model, and finishing model training when testing average accuracy Arc > -Thrd; when the average accuracy, Arc < Thrd, is tested, the training parameters are adjusted and retrained until when the average accuracy, Arc > is tested, Thrd.
And (7-3) configuring the running environment of the equipment, configuring the model to the equipment, and identifying and classifying the scene picture and the dynamic segmentation picture.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. The utility model provides an electric power ann manages video intelligence diverging device based on remove edge calculation which characterized in that: the dynamic change area division module is connected with the shunting strategy optimization evaluation module, and comprises a shunting strategy optimization evaluation module, a scene detection module, a dynamic change area division module, a model calculation overhead evaluation module, a model energy consumption overhead evaluation module, an equipment state monitoring unit, a local analysis module and a compression shunting module; the shunting strategy optimization evaluation module is used for storing pictures of scene detection and region segmentation and calculating the cost S by using a target recognition model1Energy consumption overhead, available computing resource value T of edge computing device1Storing resource value T2Power supply state value T3And network statusValue T4Determining and optimizing a distribution strategy according to the data, distributing the distribution strategy to a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit for intelligent identification, wherein the energy consumption overhead comprises transmission loss of a wired local area network
And fading Lbf of electromagnetic waves of the wireless wide area network in a transmission path.
2. The intelligent shunting device for power safety management video based on mobile edge computing as claimed in claim 1, wherein: the optimization target of the shunting strategy is as follows:
wherein the value of theta is 1 or 0, indicating selection or non-selection of the local analysis module, Si1Is the target recognition model calculation cost of the ith device, S1The calculation cost of the target identification model is the device cost grade value and the available calculation resource value T transmitted by the model calculation cost evaluation module1Storing resource value T2Power supply state value T3Network state value T4Is obtained by a device state monitoring unit; t is a unit ofi1Is the value of the available computing resource, T, of the ith devicei2Is the storage resource value, T, of the ith devicei3Is the power supply status value, T, of the ith devicei4Is the network status value of the ith device; the Loss function describes the transmission Loss of the network, and utilizes the transmission Loss of electromagnetic waves in the wired local area networkA formula of power consumption calculation and a formula of electromagnetic wave fading in free space, wherein
Attenuation of electromagnetic waves transmitted along optical fibers, L, is described for transmission loss in wired local area networksjDevice-to-device distances are described, corresponding to edges in the network topology graph; and selecting K devices in the topological graph until the three target functions reach the optimal values, namely completing the task of making a shunting strategy, wherein the K devices are randomly selected and comprise a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit.
3. The intelligent shunting device for power safety management video based on mobile edge computing as claimed in claim 1, wherein: the scene detection module is used for extracting key features from a video picture background according to the characteristics of electric power safety management and the differences of a control target object and a control task in different scenes, determining a current operation scene, detecting a target category in the current scene, and feeding a target category detection result back to the shunting strategy optimization evaluation module to reduce the number of target identification classifications.
4. The intelligent shunting device for power safety management video based on mobile edge computing as claimed in claim 1, wherein: the dynamic change region segmentation module is used for realizing dynamic change identification and region segmentation through background and foreground difference, transferring a plurality of segmented change region pictures to the distribution strategy optimization evaluation module, and the distribution strategy optimization module selects K analysis units according to an optimization target and distributes the region segmentation pictures to each unit for analysis.
5. The intelligent shunting device for power safety management video based on mobile edge computing as claimed in claim 1, wherein: the model calculation expense evaluation module is used for measuring and constructing a model calculation expense evaluation model according to historical classification performance on the equipment and establishing a calculation expense performance evaluation table which describes different identificationsCalculating time delay corresponding to the algorithm, the instruction set, the size, the type and the number of the to-be-identified category and the electric quantity grade, and taking the time delay corresponding to the performance index with the highest absolute value of the correlation coefficient in the table as the estimated calculation expense S of the table1And the value is transmitted to a shunting strategy optimization evaluation module.
6. The intelligent shunting device for power safety management video based on mobile edge computing as claimed in claim 1, wherein: and the model energy consumption overhead evaluation module is used for calculating the local area network transmission energy consumption and the wide area wireless network transmission energy consumption.
7. The intelligent shunting device for power safety management video based on mobile edge computing as claimed in claim 1, wherein: the device state monitoring unit is used for monitoring the residual electric quantity, the available computing resources, the transmission rate and the analysis time delay on the edge device in a fixed sampling period, and specifically, the device is sampled in a specified period T to obtain the residual electric quantity Q (unit: degree), the residual memory SP (unit: MB), the CPU/GPU available computing resources U (unit: MiB), the device transmission rate P (unit: Mbps) and the analysis time delay Danl(units: sheets/second), available computing resources T1I.e. U, storage resource T2SP, power supply state T3Represented by a residual capacity Q, network state T4Expressed by the transmission rate P.
8. The intelligent power safety management video shunting device based on mobile edge computing as claimed in claim 1, characterized in that: the local analysis module is used for receiving scene pictures, target classification and dynamic segmentation results, and identifying and classifying the safety management events by adopting a deep learning algorithm; and the compression and distribution module is used for transferring the scene picture, the target classification and the dynamic segmentation result set to adjacent edge equipment or a cloud server for event identification and classification according to a distribution strategy.
9. A power safety management video intelligent distribution method based on mobile edge calculation is characterized by comprising the following steps: performed with the apparatus of any one of claims 1-8, the method comprising the steps of:
step 1: the scene detection module extracts key features from the background of the video picture, determines the current operation scene, and feeds the result back to the shunting strategy optimization evaluation module to reduce the number of target identification classifications and reduce the amount of analysis tasks of edge equipment;
step 2: the dynamic change region segmentation module analyzes various safety events and hidden dangers from the dynamic change of a video picture, realizes dynamic change identification and region segmentation through background and foreground difference on the basis of a reference scene background, and transfers a plurality of segmented change regions to a shunting strategy unit for evaluation and analysis;
and step 3: obtaining calculation cost through a model calculation cost evaluation module, measuring and constructing a model calculation cost evaluation model according to historical classification performance on the equipment by the model calculation cost evaluation module, establishing a calculation cost performance evaluation table, describing corresponding calculation time delay under the conditions of different identification algorithms, instruction sets, category sizes, types and quantities to be identified and electric quantity grades, obtaining the calculation time delay of the current state of the equipment according to the table, and transmitting the calculation time delay to a shunting strategy optimization evaluation module;
and 4, step 4: obtaining model energy consumption overhead through a model energy consumption overhead evaluation module, and transmitting the energy consumption overhead to a shunting strategy optimization evaluation module;
and 5: an equipment state grade table is constructed through an equipment state monitoring unit, and the equipment state is obtained through table look-up: available computing resources T1I.e. U, storage resource T2SP, power supply state T3Represented by a residual capacity Q, network state T4Expressed by a transmission rate P and transmitted to a shunting strategy optimization evaluation module;
step 6: processing data transmitted by other modules through a shunting strategy optimization evaluation module so as to decide a shunting strategy, and shunting a picture analysis task to a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit for intelligent identification;
and 7: and the local analysis module analyzes the transmitted data.
10. The intelligent video distribution method for power safety management based on mobile edge computing as claimed in claim 9, wherein: the optimization target of the shunting strategy is as follows:
wherein the value of theta is 1 or 0, indicating selection or non-selection of the local analysis module, Si1Is the target recognition model calculation cost of the ith device, S1The calculation cost of the target identification model is the device cost grade value and the available calculation resource value T transmitted by the model calculation cost evaluation module1Storing resource value T2Power supply state value T3Network state value T4Is obtained by a device state monitoring unit; t isi1Is the value of the available computing resource, T, of the ith devicei2Is the storage resource value, T, of the ith devicei3Is the power supply state value, T, of the ith devicei4Is the network status value of the ith device; the Loss function describes the transmission Loss of the network, and utilizes the calculation formula of the transmission Loss of the electromagnetic wave in the wired local area network and the fading formula of the electromagnetic wave in the free space, wherein
Attenuation of electromagnetic waves transmitted along optical fibers, L, is described for transmission loss in wired local area networksjDescribes device to deviceThe distance between, corresponding to an edge in the network topology map; and selecting K devices in the topological graph until the three target functions reach the optimal values, namely completing the task of making a shunting strategy, wherein the K devices are randomly selected and comprise a local analysis module, an adjacent edge analysis unit or a video processing cloud analysis unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910953248.2A CN110677300B (en) | 2019-10-09 | 2019-10-09 | Electric power safety management video intelligent shunting device and method based on mobile edge calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910953248.2A CN110677300B (en) | 2019-10-09 | 2019-10-09 | Electric power safety management video intelligent shunting device and method based on mobile edge calculation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110677300A CN110677300A (en) | 2020-01-10 |
| CN110677300B true CN110677300B (en) | 2022-05-31 |
Family
ID=69081089
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910953248.2A Active CN110677300B (en) | 2019-10-09 | 2019-10-09 | Electric power safety management video intelligent shunting device and method based on mobile edge calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110677300B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111740904B (en) * | 2020-06-27 | 2021-06-04 | 常州天正信息科技有限公司 | Intelligent industrial gateway communication method and system |
| CN112134361A (en) * | 2020-09-29 | 2020-12-25 | 南京邦州电力自动化有限公司 | Intelligent monitoring system for power distribution network |
| CN112486667B (en) * | 2020-11-03 | 2022-03-18 | 深圳市中博科创信息技术有限公司 | Method and device for accurately processing data based on edge calculation |
| CN113191654A (en) * | 2021-05-12 | 2021-07-30 | 余绍祥 | Distributed resource scheduling system and method based on multi-objective optimization |
| CN113691893B (en) * | 2021-08-09 | 2022-09-06 | 武汉理工大学 | Remote intelligent monitoring platform and method for energy consumption of press machine |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108809695A (en) * | 2018-04-28 | 2018-11-13 | 国网浙江省电力有限公司电力科学研究院 | A kind of distribution uplink unloading strategy towards mobile edge calculations |
| CN110262337A (en) * | 2019-06-20 | 2019-09-20 | 中国南方电网有限责任公司 | Electric Intelligent Monitoring System and method |
| CN110311376A (en) * | 2019-07-31 | 2019-10-08 | 三峡大学 | A kind of Electrical Power System Dynamic security evaluation collective model and space-time method for visualizing |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050108414A1 (en) * | 2003-11-14 | 2005-05-19 | Taylor Thomas M. | System and method for transmitting data in computer systems using virtual streaming |
| US8126685B2 (en) * | 2006-04-12 | 2012-02-28 | Edsa Micro Corporation | Automatic real-time optimization and intelligent control of electrical power distribution and transmission systems |
-
2019
- 2019-10-09 CN CN201910953248.2A patent/CN110677300B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108809695A (en) * | 2018-04-28 | 2018-11-13 | 国网浙江省电力有限公司电力科学研究院 | A kind of distribution uplink unloading strategy towards mobile edge calculations |
| CN110262337A (en) * | 2019-06-20 | 2019-09-20 | 中国南方电网有限责任公司 | Electric Intelligent Monitoring System and method |
| CN110311376A (en) * | 2019-07-31 | 2019-10-08 | 三峡大学 | A kind of Electrical Power System Dynamic security evaluation collective model and space-time method for visualizing |
Non-Patent Citations (2)
| Title |
|---|
| 基于边缘计算的电力末端融合系统的优化;谢玮，夏水斌，何行等;《电测与仪表》;20181212;全文 * |
| 电网灾害监测与精准预报系统的设计与实现;李源林，林忠，尹沁;《电脑知识与技术》;20190805;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110677300A (en) | 2020-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110677300B (en) | Electric power safety management video intelligent shunting device and method based on mobile edge calculation | |
| CN113543068B (en) | Forest area unmanned aerial vehicle network deployment method and system based on hierarchical clustering | |
| CN113408087B (en) | Substation inspection method based on cloud side system and video intelligent analysis | |
| CN110012508A (en) | A kind of resource allocation methods and system of the edge calculations towards super-intensive network | |
| CN113422803B (en) | Seamless migration method for intelligent unmanned aerial vehicle inspection task based on end edge cloud cooperation | |
| CN114980169A (en) | Unmanned aerial vehicle auxiliary ground communication method based on combined optimization of track and phase | |
| CN212785420U (en) | Unmanned aerial vehicle system of patrolling and examining based on cloud limit is in coordination with 5G network | |
| CN116664359A (en) | Intelligent fire early warning decision system and method based on multi-sensor fusion | |
| CN115209379A (en) | Power grid cloud edge cooperative inspection system and method based on 5G intelligent networked unmanned aerial vehicle | |
| CN114494916A (en) | Black-neck crane monitoring and tracking method based on YOLO and DeepsORT | |
| CN116827389B (en) | Medium-voltage carrier communication system based on edge calculation depth fusion | |
| Al-Quraan et al. | A hybrid data manipulation approach for energy and latency-efficient vision-aided UDNs | |
| CN113242066A (en) | Multi-cell large-scale MIMO communication intelligent power distribution method | |
| CN106657889A (en) | Intelligent dynamic environment and video monitoring cloud service system based on Internet of things | |
| Sun et al. | Semantic-driven computation offloading and resource allocation for uav-assisted monitoring system in vehicular networks | |
| CN110072197A (en) | A kind of preferably intelligent switching and selecting method of emergency communication transmission channel | |
| Fu et al. | An edge computing framework for digital grid | |
| CN113452961A (en) | Water surface monitoring alarm system, method and medium based on edge calculation | |
| CN103198325A (en) | Cyanophyta algal bloom monitoring system | |
| Weifu et al. | Research on the application of target detection based on deep learning technology in power grid operation inspection | |
| Shen et al. | Recognition and Optimization of Atmospheric Duct in TD-LTE System Based on Convolutional Neural Network | |
| CN117279019B (en) | Method and device for predicting spectrum efficiency, electronic equipment and storage medium | |
| CN114531193B (en) | Bridge state monitoring method based on unmanned aerial vehicle cellular topology networking and mobile edge calculation | |
| CN109450674B (en) | Detector DIS system man-machine interaction method based on multicast optimization | |
| Wang et al. | Mean-field power allocation for UDN |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |