A data acquisition and cooperative processing method for low-voltage power distribution information

Abstract

The application discloses a kind of low voltage distribution power utilization information data acquisition collaborative processing methods, method includes constructing cloud edge computing architecture, in each time slot of data processing total optimization cycle under the cloud edge computing architecture, by each high-frequency acquisition terminal acquisition low voltage distribution power utilization data, to each high-frequency acquisition terminal is carried out edge data unloading optimization processing, obtains each selected edge server and selected transmission power, unload low voltage distribution power utilization data to each selected edge server, obtain unloading data;Through each selected edge server receives unloading data, to each selected edge server is carried out edge-cloud data shunting optimization processing, obtains selected data shunting ratio, shunts unloading data to cloud server;Through cloud server, shunted data is carried out cloud-edge collaborative data processing.This embodiment realizes the joint optimization of high-frequency acquisition terminal transmission power selection, edge server selection and data shunting three, improves data processing performance.

Description

Technical Field

[0001] This invention relates to the field of low-voltage power distribution information data processing, and in particular to a collaborative data acquisition and processing method for low-voltage power distribution information. Background Technology

[0002] With a high proportion of distributed renewable energy, energy storage, and controllable loads being connected to low-voltage distribution substations, a large number of high-frequency data acquisition terminals are deployed in these areas to collect multi-dimensional operational data such as voltage and current, supporting continuous monitoring, unmanned control, and fault detection, in order to achieve transparent monitoring and stable operation of distribution terminals. Compared with traditional acquisition terminals, the data acquisition range and frequency have increased dramatically. However, due to limited computing and energy resources on the terminal side, it is difficult to meet the requirements of stringent and differentiated data processing for power services.

[0003] Cloud-edge collaboration can effectively improve data processing performance, allowing high-frequency acquisition terminals to offload collected data to edge servers or cloud servers for remote processing. However, traditional cloud-edge collaboration involves joint optimization of acquisition terminal transmission power selection, edge server selection, and data offloading. Solving this joint optimization is difficult, lacks consideration of the data offloading ratio to edge servers, and employs traditional binary full offloading strategies, failing to fully utilize the computing resources of both cloud and edge servers. Furthermore, traditional cloud-edge collaboration lacks consideration of the impact of data acquisition frequency on data processing performance, failing to adjust data processing strategies based on the acquisition terminal's acquisition frequency, resulting in poor data processing performance and difficulty in meeting the differentiated data processing needs of high-frequency acquisition terminals. Summary of the Invention

[0004] This invention provides a collaborative data acquisition and processing method for low-voltage power distribution information, which achieves joint optimization of high-frequency acquisition terminal transmission power selection, edge server selection, and data diversion, thereby improving data processing performance.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a data acquisition and collaborative processing method for low-voltage power distribution information, comprising:

[0006] Based on the hierarchical devices of the low-voltage power distribution and consumption information high-frequency acquisition system, a cloud-edge computing architecture is constructed; wherein, the hierarchical devices of the high-frequency acquisition system include one or more high-frequency acquisition terminals in the terminal layer, one or more edge servers in the edge layer, and one cloud server in the cloud layer.

[0007] In each time slot of the overall data processing optimization cycle under the cloud-edge computing architecture, low-voltage power distribution data is collected by each high-frequency acquisition terminal. Based on the low-voltage power distribution data and the end-edge data offloading model, end-edge data offloading optimization processing is performed on each high-frequency acquisition terminal to obtain a first selected configuration. Then, according to the first selected configuration, the low-voltage power distribution data is offloaded to each selected edge server by each high-frequency acquisition terminal to obtain offloaded data. The first selected configuration includes the selected edge servers and the selected transmission power.

[0008] The system receives unloaded data from each selected edge server. Based on the cloud-edge data splitting model and the unloaded data, it performs edge-cloud data splitting optimization on each selected edge server to obtain a second selected configuration. Then, based on the second selected configuration, the unloaded data is split to the cloud server through each selected edge server to obtain split data. The second selected configuration includes the selected data splitting ratio.

[0009] The cloud server performs cloud-edge collaborative data processing based on the cloud-edge collaborative data processing model and the second selection configuration.

[0010] Implementing this invention, the cloud-edge computing architecture and collaborative mechanism of the low-voltage power distribution information high-frequency acquisition system decomposes joint optimization into a transmission power selection and edge server selection stage (first selection configuration) and a data offloading stage (second selection configuration). In the first stage, the transmission power of the high-frequency acquisition terminal and the edge server are selected through an end-to-edge data offloading model and end-to-edge data offloading optimization. In the second stage, the data offloading ratio of the edge server is selected through a cloud-to-edge data offloading model and edge-to-cloud data offloading optimization, achieving joint optimization of the high-frequency acquisition terminal transmission power selection, edge server selection, and data offloading, and enabling collaborative data processing in the cloud server. During the joint optimization process, the acquisition frequency and transmission power of the high-frequency acquisition terminal are fully considered. Through continuous learning and updating, more reasonable server and transmission power selection decisions are made to meet the differentiated data processing needs of the high-frequency acquisition terminal and improve data processing performance.

[0011] As a preferred option, based on low-voltage power distribution data and the end-to-edge data offloading model, the end-to-edge data offloading optimization process is performed on each high-frequency acquisition terminal to obtain the first selected configuration, specifically:

[0012] In the end-edge data offloading model, each high-frequency acquisition terminal is modeled as a decision-maker based on UCB frequency sensing end-edge data offloading optimization processing. The transmission power level of each edge server and each high-frequency acquisition terminal during data offloading is modeled as a rocker arm. The selection problem of edge server and transmission power when each high-frequency acquisition terminal offloads low-voltage power distribution data is solved to obtain the first preferred configuration.

[0013] As a preferred solution, the end-edge data offloading model is as follows:

[0014] Within the current time slot, based on the amount of data collected at the current high-frequency acquisition terminal, the data stored at the current high-frequency acquisition terminal is modeled as the data backlog queue for the current time slot. The data backlog queue for the next time slot is then dynamically updated, specifically as follows:

[0015]

[0016] Among them, Q n (t+1) represents the data backlog queue for the (t+1)th time slot, Q n (t) represents the data backlog queue of the t-th time slot, where t is the current t-th time slot, t+1 is the (t+1)-th time slot, and U n (t) represents the current high-frequency acquisition terminal d in the t-th time slot. n The amount of data currently collected, x n,m (t) is the variable for selecting the edge server. The current unloaded data volume; where the current unloaded data volume is the current high-frequency acquisition terminal d in the t-th time slot. n Unload to the current edge server m The amount of data;

[0017] The current unloaded data volume is obtained based on the data backlog queue of the current time slot, the maximum duration of the end-to-edge data unloading phase, and the data transmission rate, specifically:

[0018]

[0019] Where τ1 is the maximum time length of the end-to-edge data unloading phase, R n,m (t) represents the current high-frequency acquisition terminal d. n and current edge servers m The current data transfer rate between them;

[0020] The current data transmission rate is obtained based on the current data transmission bandwidth, the current data transmission channel gain, the current Gaussian white noise, the current electromagnetic interference power, and the current transmission power, specifically as follows:

[0021]

[0022] Among them, B n,m (t) represents the current high-frequency acquisition terminal d. n and current edge servers m The bandwidth of current data transmission between them, g n,m (t) represents the current high-frequency acquisition terminal d. n and current edge serversm The channel gain for current data transmission between them, where δ0 is the current high-frequency acquisition terminal d. n and current edge servers m The current Gaussian white noise between, I n,m (t) represents the current high-frequency acquisition terminal d. n and current edge servers m The current electromagnetic interference power between them, P n (t)∈P n For the current high-frequency acquisition terminal d n The current transmission power; where P n The set of transmission power is as follows:

[0023]

[0024] Where L is the transmit power level number, P n,min and P n,max These represent the current high-frequency acquisition terminal d. n The minimum and maximum transmission power, where l is the power level;

[0025] The terminal transmission delay is calculated based on the current data transmission rate, the maximum duration of the end-to-side data offloading phase, the data backlog queue of the current time slot, and the amount of data currently collected.

[0026]

[0027] in, Let be the terminal transmission delay, representing the time interval from the current high-frequency acquisition terminal d in the t-th time slot. n To the current edge server m Data offloading transmission delay; R n,m (t) represents the current data transmission rate, τ1 represents the maximum time length of the end-to-side data offloading phase, and U n (t) represents the current high-frequency acquisition terminal d in the t-th time slot. n The amount of data currently collected, Q n (t) represents the data backlog queue for the t-th time slot;

[0028] Based on the current transmission power and terminal transmission delay, the data transmission energy consumption during data offloading is calculated as follows:

[0029]

[0030] Among them, E n,m (t) represents the data transmission energy consumption during data offloading, indicating the energy consumed in the t-th time slot to transfer data from the current high-frequency acquisition terminal d. n To the current edge server mData transmission energy consumption during data offloading.

[0031] As a preferred solution, each high-frequency acquisition terminal is modeled as a decision-maker based on UCB frequency sensing end-edge data offloading optimization processing. The transmission power levels of each edge server during data offloading from each high-frequency acquisition terminal are modeled as rocker arms. The selection problem of edge servers and transmission power when each high-frequency acquisition terminal offloads low-voltage power distribution data is solved to obtain the first preferred configuration, specifically:

[0032] Based on the first optimization variable, each high-frequency acquisition terminal optimizes the end-to-edge data offloading process by minimizing the weighted sum of the total data processing latency and the data transmission energy consumption during data offloading; wherein, the first optimization variable is the current transmission power and the edge server selection variable; the optimization of the end-to-edge data offloading process is as follows:

[0033]

[0034]

[0035]

[0036]

[0037]

[0038] Among them, SP1 optimizes the end-to-edge data offloading process, x n,m (t) is the variable for selecting the edge server. E represents the total latency of data processing. n,m (t) represents the energy consumption for data transmission during data unloading, V E (t) is the weight of energy consumption, d n For the current high-frequency acquisition terminal d n s m For the current edge servers s m D is the set of N high-frequency acquisition terminals, S is the set of M edge servers, T is the set of time slots for the total data processing optimization cycle, and P n (t) represents the current high-frequency acquisition terminal d. n The current transmission power, where P is the set of transmission powers. For the current edge servers s m The maximum number of terminals that can be processed; C1, C2, and C3 are constraints for selecting edge servers; C4 is a constraint for selecting transmit power.

[0039] The optimization of the edge-to-edge data offloading process is transformed into solving the MAB data offloading problem based on UCB to obtain the first preferred configuration. The MAB data offloading problem is the selection of edge servers and transmission power when each high-frequency acquisition terminal offloads low-voltage power distribution data, specifically including:

[0040] Each high-frequency acquisition terminal is taken as the first decision-maker, and decisions are made on edge server selection and power selection for data offloading.

[0041] The first jib arm is formed by combining the power level with each edge server, specifically:

[0042]

[0043] Among them, A n For the set of first rocker arms; The first rocker arm represents the power level l and the current edge server s. m The combination;

[0044] The first rocker arm selects the action indicator variable as the current unloading action; where... For the current uninstallation action, This indicates that the current high-frequency acquisition terminal selects the current transmission power to transmit data to the current edge server, where the current transmission power is calculated using the formula:

[0045]

[0046] Among them, P n (t) represents the current high-frequency acquisition terminal d. n The current transmission power, L is the transmit power level, P n,min and P n,max These represent the current high-frequency acquisition terminal d. n The minimum and maximum transmission power, where l is the power level;

[0047] Based on the first rocker arm selected by the current high-frequency acquisition terminal within the current time slot as the first reward, specifically:

[0048]

[0049] Where t is the current t-th time slot, As the first reward, V E (t) represents the weight of energy consumption. E represents the total latency of data processing. n,m (t) represents the energy consumption for data transmission during data unloading.

[0050] As a preferred solution, the optimization of the end-to-edge data offloading process is transformed into solving the MAB data offloading problem based on UCB to obtain the first-choice configuration, specifically:

[0051] The current unloading action, edge server selection variable, and first reward are initialized. When the first preset condition is met, the current high-frequency acquisition terminal sequentially selects the set of first rocker arms to obtain the initial reward, which is then used as the first reward for the first solution.

[0052] Within the current time slot, the current high-frequency acquisition terminal calculates the upper bound of the confidence level based on the current number of selected first rocker arms, specifically as follows:

[0053]

[0054] in, This is the upper bound of the confidence level. β represents the average reward over the first t-1 time slots. n Indicates the current high-frequency acquisition terminal d n The sampling frequency weight, Indicates the current first rocker arm The confidence interval, The current number of selections for the first rocker arm, t represents the current t-th time slot, and t-1 represents the previous t-1-th time slot;

[0055] Based on the upper bound of confidence, the current high-frequency acquisition terminal selects the first rocker arm with the highest confidence upper bound to perform the action, selects the current transmission power to transmit the data to the current edge server, and obtains the first selected configuration.

[0056] Based on the total latency of data processing, the energy consumption of data transmission during data unloading, the weight of energy consumption, and the first selection configuration, the current first reward is obtained. Based on the current first reward, the average reward of the previous time slot and the current selection number of the first rocker arm are updated until the calculation of all time slots in the time slot set of the total data processing optimization cycle is completed.

[0057] As a preferred option, based on the cloud-edge data offloading model and offloaded data, edge-cloud data offloading optimization is performed on each selected edge server to obtain the second preferred configuration, specifically:

[0058] In the cloud-edge data offloading model, each selected edge server is modeled as a decision-maker for edge-cloud data offloading optimization based on EXP3, and the data offloading ratio of each selected edge server is modeled as a rocker arm. The problem of choosing the offloading ratio when each selected edge server will unload data offloading is solved to obtain the second choice configuration.

[0059] As the preferred solution, the cloud-edge data offloading model is as follows:

[0060] Within the current time slot, based on the current maintenance data backlog queue where the edge server is unloading data from the current high-frequency acquisition terminal, the maintenance data backlog queue for the next time slot is dynamically updated, specifically as follows:

[0061]

[0062] in, For the maintenance data backlog queue of the (t+1)th time slot, For the current edge servers s m From the current high-frequency acquisition terminal d n The current maintenance data backlog queue for unloading data in the middle. This represents the current amount of data being unloaded. For the current edge servers s m The amount of data processed, y n,m (t) represents the current data split ratio, y n,m (t)∈Y, where Y is the set of data split ratios, specifically:

[0063]

[0064] Among them, y min (t) and y max (t) represents the minimum and maximum values ​​of the data split ratio, respectively; H represents the number of levels included in the data split ratio set; and h represents the data split ratio selected by the current edge server as the h-th level.

[0065] The current data splitting ratio is calculated based on the selected data splitting ratio level of the current edge server, the minimum value of the data splitting ratio, and the maximum value of the data splitting ratio. The formula is as follows:

[0066]

[0067] Among them, y n,m (t) represents the current data split ratio;

[0068] Based on the current data splitting ratio, server transmission rate, maximum duration of cloud-edge data transmission phase, and current offloaded data volume, calculate the edge offloaded data volume and edge server transmission latency, specifically:

[0069]

[0070]

[0071] in, The amount of data offloaded at the edge represents the amount of data transferred from the current edge server s in the t-th time slot. mThe current high-frequency acquisition terminal d is unloaded to cloud server s0. n The amount of data, τ2 represents the current unloaded data volume, τ2 represents the maximum time duration of the cloud-edge data transmission phase, and y represents the data volume currently unloaded. n,m (t) is the current data split ratio. For the current edge servers s m The server transfer rate between the cloud server s0 and the server s0 For edge server transmission latency, it represents the latency from the current edge server s within the t-th time slot. m The transmission latency of data transmitted to cloud server s0.

[0072] As a preferred approach, each selected edge server is modeled as a decision-maker in the edge-cloud data offloading optimization process based on EXP3, and the data offloading ratio of each selected edge server is modeled as a rocker arm. The problem of choosing the offloading ratio when each selected edge server unloads data offloading is solved to obtain the second alternative configuration, specifically:

[0073] Based on the second optimization variable, each selected edge server optimizes the edge-cloud data offloading process by minimizing the latency required for the server to process all data. The second optimization variable is the current data offloading ratio, and the latency required for the server to process all data includes the edge server transmission latency, the cloud server processing latency, and the edge server processing latency. The optimization of the edge-cloud data offloading process specifically involves:

[0074]

[0075]

[0076] Among them, SP2 is the edge-cloud data offloading process, C5 is the data offloading ratio selection constraint, and y n,m (t) represents the current data split ratio. For edge server transmission latency, To handle latency for cloud servers, For edge server latency processing, Y represents the set of data splitting ratios, and d n For the current high-frequency acquisition terminal d n s m For the current edge servers s m D is the set of N high-frequency acquisition terminals, and S is the set of M edge servers;

[0077] The edge-cloud data offloading process is transformed into solving the MAB data offloading problem based on EXP3 to obtain the second alternative configuration. The MAB data offloading problem is specifically designed to determine the offloading ratio when offloading data from edge servers, and includes:

[0078] Each selected edge server is designated as a secondary decision-maker to determine the distribution ratio when offloading data from the edge to the cloud.

[0079] Use the total segmentation ratio of the data splitting ratio set as the second rocker arm;

[0080] The second rocker arm selector action indicator variable is used as the current diversion action; where... Current traffic diversion action, The current data split ratio represents the current edge server s. m In the t-th time slot, the high-frequency acquisition terminal d will be used. n The data splitting rate setting is set by the following formula:

[0081]

[0082] Among them, y n,m (t) represents the current data split ratio, y min and y max These represent the minimum and maximum values ​​of the data splitting ratio, respectively. H represents the number of levels included in the data splitting ratio set, and h represents the data splitting ratio selected by the current edge server as the h-th level.

[0083] The edge server of the second crane, selected based on the latency required for the server to process all data, is given a second bonus, specifically:

[0084]

[0085] in, As the second reward, For edge server transmission latency, To handle latency for cloud servers, To handle latency for edge servers.

[0086] As a preferred solution, the edge-cloud data splitting process is transformed into solving the MAB data splitting problem based on EXP3, resulting in the second alternative configuration, specifically:

[0087] Initialize uniform distribution parameters and preset performance-related empirical distribution parameters;

[0088] Within the current time slot, calculate the probability of selecting the current second rocker arm, specifically as follows:

[0089]

[0090] in, Let ξ∈(0,1] be the probability of selecting the current second rocker arm, and let ξ∈(0,1] be a uniform distribution parameter. Here, H represents the empirical distribution parameters related to the performance of the current t-th time slot, and H is the number of levels included in the data split ratio set.

[0091] Based on the probability of selecting the current second rocker arm, calculate the current cumulative distribution function, as follows:

[0092]

[0093] Among them, F n,m (h) is the current cumulative distribution function;

[0094] Based on the current cumulative distribution function, the second alternative configuration is obtained as follows:

[0095]

[0096] in, This represents the current traffic splitting action, indicating the second-choice configuration. hour,

[0097] Each selected edge server executes the second selection configuration. Based on the latency required for the server to process all data, the selected edge server of the second jib arm with the data split ratio obtains the current second reward. Based on the current second reward, the performance-related empirical distribution parameters of the next time slot are updated until the calculation of all time slots in the time slot set of the total data processing optimization cycle is completed.

[0098] Specifically, based on the current second reward, the performance-related empirical distribution parameters for the next time slot are updated as follows:

[0099]

[0100]

[0101] in, The empirical distribution parameters related to the performance of the (t+1)th time slot are... To determine the probability of selecting the current second rocker arm, The empirical distribution parameters related to the performance of the current time slot t are... To estimate the reward, η is an adjustment factor. For the current second reward, H represents the number of levels included in the data splitting ratio set.

[0102] As a preferred solution, the cloud-edge collaborative data processing model is as follows:

[0103] Based on the current data backlog queue, the current data splitting ratio, the current amount of data unloaded, the edge server's computing power, the terminal data processing complexity, and the maximum duration of the cloud-edge collaborative data processing phase, the data processing volume and latency of the edge server are calculated using the following formulas:

[0104]

[0105]

[0106] in, The amount of data processed by the edge server, representing the current edge server s. m The amount of data processed For the current edge servers s m From the current high-frequency acquisition terminal d n The current maintenance data backlog queue for unloading data in the middle, y n,m (t) represents the current data split ratio. This represents the current amount of data being unloaded. The current edge server s m The computing power in the t-th time slot, χ n For the current high-frequency acquisition terminal d n The terminal data processing complexity is τ3, where τ3 is the maximum time length of the cloud-edge collaborative data processing stage. For edge server latency handling, it represents the current edge server s m Processing delay in the t-th time slot;

[0107] Based on the current data backlog queue, the data volume processed by edge servers, and the data volume processed by cloud servers, the data backlog queue for the next time slot is dynamically updated, specifically as follows:

[0108]

[0109] in, This is the backlog queue for the offloaded data in the (t+1)th time slot. For cloud server s0, from the current edge server s m The current high-frequency acquisition terminal d is being diverted. n The current backlog of diverted data, To handle the amount of data for edge servers, The amount of data processed by the cloud server; specifically, the amount of data processed by the cloud server is as follows:

[0110]

[0111] in, For edge server transmission latency, To offload data volume at the edge, χ is the computing power of cloud server s0 in the t-th time slot. n For the current high-frequency acquisition terminal d n The terminal data processing complexity, τ3 is the maximum time length of the cloud-edge collaborative data processing stage;

[0112] Based on the current backlog of diverted data, edge server transmission latency, edge offloaded data volume, cloud server computing power, terminal data processing complexity, and the maximum duration of the cloud-edge collaborative data processing phase, the cloud server processing latency is calculated as follows:

[0113]

[0114] in, To handle latency for cloud servers. Attached Figure Description

[0115] Figure 1 This is a flowchart illustrating an embodiment of a collaborative data acquisition and processing method for low-voltage power distribution information provided by the present invention.

[0116] Figure 2 This is a cloud-edge computing architecture diagram of a high-frequency acquisition system for low-voltage power distribution information, which is an embodiment of a collaborative data acquisition method for low-voltage power distribution information provided by the present invention.

[0117] Figure 3 This is a comparison chart of different algorithms for the weighted sum of data processing latency and terminal energy consumption in one embodiment of a collaborative data acquisition method for low-voltage power distribution information provided by the present invention.

[0118] Figure 4 This is a comparison chart of different algorithms for cumulative energy consumption in one embodiment of a collaborative data acquisition and processing method for low-voltage power distribution information provided by the present invention. Detailed Implementation

[0119] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0120] Example 1

[0121] Please refer to Figure 1This is a flowchart illustrating a collaborative data acquisition method for low-voltage power distribution information provided in an embodiment of the present invention. The collaborative data acquisition method of this embodiment is applicable to a high-frequency data acquisition system for low-voltage power distribution information. This embodiment achieves joint optimization of high-frequency acquisition terminal transmission power selection, edge server selection, and data offloading through end-to-edge data offloading optimization and edge-to-cloud data diversion optimization, thereby improving data processing performance. The collaborative data acquisition method includes steps 101 to 104, each step as follows:

[0122] Step 101: Construct a cloud-edge computing architecture based on the hierarchical devices of the low-voltage power distribution information high-frequency acquisition system; wherein, the hierarchical devices of the high-frequency acquisition system include one or more high-frequency acquisition terminals in the terminal layer, one or more edge servers in the edge layer, and one cloud server in the cloud layer.

[0123] In this embodiment, the cloud-edge computing architecture of the low-voltage power distribution information high-frequency acquisition system is as follows: Figure 2 As shown, at the terminal layer, high-frequency acquisition terminals are deployed on electrical equipment such as photovoltaic and distributed energy storage systems. They collect data at high frequencies to support different services. Assuming there are N high-frequency acquisition terminals, their set is represented as D = {d1,...,d...}. n ,...,d N The edge layer and the cloud layer each consist of M edge servers and one cloud server. The set of edge servers is represented as S = {s1,...,s...}. m ,...,s M The cloud server is represented as s0. The high-frequency data acquisition terminal offloads data to the edge server to reduce data processing latency. Simultaneously, the edge server diverts data offloaded from the terminal to the cloud server to alleviate the data processing pressure caused by the surge in terminal-acquired data. Through cloud-edge collaborative data processing, the processing requirements for high-frequency power distribution information acquisition can be met.

[0124] Step 102: In each time slot of the total data processing optimization cycle under the cloud-edge computing architecture, low-voltage power distribution data is collected through each high-frequency acquisition terminal. Based on the low-voltage power distribution data and the end-edge data offloading model, end-edge data offloading optimization processing is performed on each high-frequency acquisition terminal to obtain a first selected configuration. Then, according to the first selected configuration, the low-voltage power distribution data is offloaded to each selected edge server through each high-frequency acquisition terminal to obtain offloaded data. The first selected configuration includes each selected edge server and the selected transmission power.

[0125] In this embodiment, in the cloud-edge computing architecture and collaborative mechanism of the low-voltage power distribution information high-frequency acquisition system, the total data processing optimization cycle is divided into T time slots, the set of which is T={1,...,t,...,T}. Each time slot includes three stages: end-edge data offloading, edge-cloud data transmission, and cloud-edge collaborative data processing, corresponding to the end-edge data offloading model, the edge-cloud data diversion model, and the cloud-edge collaborative data processing model, respectively. Simultaneously, the maximum time lengths of the three stages are set as the maximum time length τ1 for the end-edge data offloading stage, the maximum time length τ2 for the cloud-edge data transmission stage, and the maximum time length τ3 for the cloud-edge collaborative data processing stage, respectively. x n,m (t) represents the edge server selection variable, where x n,m (t) = 1 indicates that terminal d n Select edge servers m Used for data offloading in the t-th time slot; otherwise, x n,m (t) = 0. It should be noted that, in this embodiment, the terminal refers to the high-frequency acquisition terminal.

[0126] Optionally, based on low-voltage power distribution data and the end-to-edge data offloading model, end-to-edge data offloading optimization processing is performed on each high-frequency acquisition terminal to obtain the first selected configuration, specifically:

[0127] In the end-edge data offloading model, each high-frequency acquisition terminal is modeled as a decision-maker based on UCB frequency sensing end-edge data offloading optimization processing. The transmission power level of each edge server and each high-frequency acquisition terminal during data offloading is modeled as a rocker arm. The selection problem of edge server and transmission power when each high-frequency acquisition terminal offloads low-voltage power distribution data is solved to obtain the first preferred configuration.

[0128] In this embodiment, the selection of high-frequency acquisition terminal transmission power and edge server, i.e., the first selection configuration, is achieved through a frequency-sensing end-to-edge data offloading optimization algorithm based on UCB. During the end-to-edge data offloading stage, i.e., within the end-to-edge data offloading model, the high-frequency acquisition terminal is modeled as the decision-maker in the UCB-based frequency-sensing end-to-edge data offloading optimization algorithm, and the transmission power levels of the edge server during terminal data offloading are modeled as rocker arms to solve the problem of edge server and transmission power selection during terminal data offloading. The high-frequency acquisition terminal has different acquisition frequencies, and the terminal acquires different amounts of data in each time slot, offloading the data to the selected edge server for data processing.

[0129] Optional, edge-to-edge data offloading model, specifically:

[0130] Within the current time slot, based on the amount of data collected at the current high-frequency acquisition terminal, the data stored at the current high-frequency acquisition terminal is modeled as the data backlog queue for the current time slot. The data backlog queue for the next time slot is then dynamically updated, specifically as follows:

[0131]

[0132] Among them, Q n (t+1) represents the data backlog queue for the (t+1)th time slot, Q n (t) represents the data backlog queue for the t-th time slot, i.e., the high-frequency acquisition terminal d for the current time slot. n The stored data backlog queue, where t is the current t-th time slot, t+1 is the t+1-th time slot, and t+1 is the next time slot after t, U n (t) represents the current high-frequency acquisition terminal d in the t-th time slot. n The amount of data currently collected, x n,m (t) is the variable for selecting the edge server. The current unloaded data volume; where the current unloaded data volume is the current high-frequency acquisition terminal d in the t-th time slot. n Unload to the current edge server m The amount of data;

[0133] The current unloaded data volume is obtained based on the data backlog queue of the current time slot, the maximum duration of the end-to-edge data unloading phase, and the data transmission rate, specifically:

[0134]

[0135] Where τ1 is the maximum time length of the end-to-edge data unloading phase, R n,m (t) represents the current high-frequency acquisition terminal d. n and current edge servers m The current data transfer rate between them, i.e., d n and s m Data transfer rate between them;

[0136] The current data transmission rate is obtained based on the current data transmission bandwidth, the current data transmission channel gain, the current Gaussian white noise, the current electromagnetic interference power, and the current transmission power, specifically as follows:

[0137]

[0138] Among them, B n,m (t) represents the current high-frequency acquisition terminal d. n and current edge servers m The bandwidth of current data transmission between them, g n,m(t) represents the current high-frequency acquisition terminal d. n and current edge servers m The channel gain for current data transmission between; δ0 is the current high-frequency acquisition terminal d. n and current edge servers m The current Gaussian white noise between, I n,m (t) represents the current high-frequency acquisition terminal d. n and current edge servers m The current electromagnetic interference power between them, P n (t)∈P n For the current high-frequency acquisition terminal d n The current transmission power, i.e., d n The transmission power; where P n The set of transmission power is as follows:

[0139]

[0140] Where L is the transmit power level number, P n It contains L levels; P n,min and P n,max These represent the current high-frequency acquisition terminal d. n The minimum and maximum transmission power, where l is the power level;

[0141] The terminal transmission delay is calculated based on the current data transmission rate, the maximum duration of the end-to-side data offloading phase, the data backlog queue of the current time slot, and the amount of data currently collected.

[0142]

[0143] in, Let be the terminal transmission delay, representing the time interval from the current high-frequency acquisition terminal d in the t-th time slot. n To the current edge server m Data offloading transmission delay; R n,m (t) represents the current data transmission rate, τ1 represents the maximum time length of the end-to-side data offloading phase, and U n (t) represents the current high-frequency acquisition terminal d in the t-th time slot. n The amount of data currently collected, Q n (t) represents the data backlog queue for the t-th time slot;

[0144] Based on the current transmission power and terminal transmission delay, the data transmission energy consumption during data offloading is calculated as follows:

[0145]

[0146] Among them, En,m (t) represents the data transmission energy consumption during data offloading, indicating the energy consumed in the t-th time slot to transfer data from the current high-frequency acquisition terminal d. n To the current edge server m Data transmission energy consumption during data offloading.

[0147] It should be noted that the optimization of data processing needs to address the issues of low latency and low energy consumption in cloud-edge collaborative computing data processing. The total data processing latency includes the latency of edge data offloading, and the maximum value among the edge server data processing latency, cloud-edge data offloading latency, and the sum of cloud server data processing latency. The formula is:

[0148]

[0149] in, This represents the total latency of data processing. This refers to the terminal transmission delay (the delay of data offloading from the end to the edge). To address edge server latency (edge ​​server data processing latency), For edge server transmission latency (cloud-edge data offloading latency), This refers to the latency of cloud server processing (cloud server data processing latency).

[0150] It should be noted that this addresses the problem of low-latency, low-energy cloud-edge collaborative computing data processing. The goal is to minimize the weighted sum of data processing latency and terminal energy consumption by jointly optimizing the selection of transmission power, edge server, and cloud-edge data offloading ratio. The joint optimization problem is expressed by the following formula:

[0151]

[0152]

[0153]

[0154]

[0155]

[0156]

[0157] Among them, V E These are the weights for energy consumption; C1, C2, and C3 are edge server selection constraints, meaning each terminal can only select one edge server for data offloading, and s m The maximum number of terminals that can be processed is C4 is the transmit power selection constraint; C5 is the data split ratio selection constraint.

[0158] In this embodiment, a cloud-edge collaborative data processing optimization algorithm based on machine learning is used to solve the joint optimization problem. To decouple the optimization variables, P1 is decomposed into two sub-problems, SP1 and SP2, where SP1 involves the current transmission power P. n (t) and edge server selection variable x n,m (t) is the end-to-edge data offloading subproblem; SP2 is related to the current data splitting ratio y. n,m The cloud-edge data splitting problem (t)

[0159] Optionally, each high-frequency acquisition terminal is modeled as a decision-maker based on UCB frequency sensing end-edge data offloading optimization processing, and the transmission power level of each edge server during data offloading from each high-frequency acquisition terminal is modeled as a rocker arm. The selection problem of edge server and transmission power when each high-frequency acquisition terminal offloads low-voltage power distribution data is solved to obtain the first preferred configuration, specifically:

[0160] Based on the first optimization variable, each high-frequency acquisition terminal optimizes the end-to-edge data offloading process by minimizing the weighted sum of the total data processing latency and the data transmission energy consumption during data offloading; wherein, the first optimization variable is the current transmission power and the edge server selection variable; the optimization of the end-to-edge data offloading process is as follows:

[0161]

[0162]

[0163]

[0164]

[0165]

[0166] Among them, SP1 optimizes the end-to-edge data offloading process, x n,m (t) is the variable for selecting the edge server. E represents the total latency of data processing. n,m (t) represents the energy consumption for data transmission during data unloading, V E (t) is the weight of energy consumption, d n For the current high-frequency acquisition terminal d n s m For the current edge servers s m D is the set of N high-frequency acquisition terminals, S is the set of M edge servers, T is the set of time slots for the total data processing optimization cycle, and P n (t) represents the current high-frequency acquisition terminal d. n The current transmission power, where P is the set of transmission powers. For the current edge servers s m The maximum number of terminals that can be processed; C1, C2, and C3 are constraints for selecting edge servers; C4 is a constraint for selecting transmit power.

[0167] In this embodiment, SP1 optimizes the end-to-side data offloading process, and the optimization variables include the current transmission power P. n (t) and edge server selection variable x n,m (t), the terminal minimizes the total latency of data processing. Data transfer energy consumption E during data offloading n,m The optimal offloading decision is obtained by weighting the sum of (t). However, global state information such as channel quality and edge server computing resources is unknown, making it difficult for the terminal to make the optimal offloading decision. The terminal should optimize edge server selection and power selection based on local state information. The Multi-armed Bandit (MAB) problem is an effective method for solving incomplete information combination optimization problems. In each time slot, the decision-maker selects an arm, and the selected arm generates a reward. The decision-maker's goal is to maximize the cumulative reward. SP1 is transformed into the MAB problem (MAB data offloading problem), in which the decision-maker, arm, action, and reward are determined.

[0168] Optionally, the optimized end-to-edge data offloading process can be transformed into solving the MAB data offloading problem based on UCB to obtain the first selected configuration; wherein, the MAB data offloading problem is the selection problem of edge servers and transmission power when each high-frequency acquisition terminal offloads low-voltage power distribution data, specifically including:

[0169] Optionally, each high-frequency acquisition terminal can be designated as the primary decision-maker to make decisions regarding edge server selection and power allocation for data offloading.

[0170] In this embodiment, the decision-maker for the MAB data offloading problem is defined as the acquisition terminal, which makes decisions on edge server selection and power control for data offloading.

[0171] Optionally, the power level and the combination of each edge server can be used as the first jib arm, specifically:

[0172]

[0173] Among them, A n For the set of first rocker arms; The first rocker arm represents the power level l and the current edge server s. m The combination;

[0174] In this embodiment, the rocker arm for the MAB data offloading problem: [This refers to the power P...] n Combined with edge servers (S) to reduce the spatial complexity of actions. Definition As a set of rocker arms, it satisfies the condition |A n |=L×M. Rocker arm Represents power level l and edge server s m The combination of. For selection The number of times.

[0175] Optionally, the first rocker arm selection action indicator variable can be used as the current unloading action; where, For the current uninstallation action, This indicates that the current high-frequency acquisition terminal selects the current transmission power to transmit data to the current edge server, where the current transmission power is calculated using the formula:

[0176]

[0177] Among them, P n (t) represents the current high-frequency acquisition terminal d. n The current transmission power, L is the transmit power level, P n,min and P n,max These represent the current high-frequency acquisition terminal d. n The minimum and maximum transmission power, where l is the power level;

[0178] In this embodiment, the action for the MAB data unloading problem is: the rocker arm selection action indicator variable is... in, Terminal d m Select transmission power P n (t) Transmit data to the edge server s m P n (t) is represented as:

[0179]

[0180] Based on the first rocker arm selected by the current high-frequency acquisition terminal within the current time slot as the first reward, specifically:

[0181]

[0182] Where t is the current t-th time slot, As the first reward, V E (t) represents the weight of energy consumption. E represents the total latency of data processing. n,m (t) represents the energy consumption for data transmission during data unloading.

[0183] In this embodiment, the reward for the MAB data offloading problem is: in the t-th time slot, d n choose To get rewards Represented as:

[0184]

[0185] It should be noted that the frequency-aware end-side data offloading optimization processing algorithm based on UCB incorporates the sampling frequency weight into the confidence upper bound calculation formula to achieve frequency awareness and solves the MAB problem of end-side data offloading. UCB is a low-complexity learning-based algorithm used to balance development and exploration. The proposed algorithm enables the acquisition terminal to take actions based on local state information such as latency, and combines optimization variables... and E n,m (t) , obtains the reward and updated state information obtained by the terminal perception, for the next selection. The implementation process of the frequency perception end-side data offloading optimization processing algorithm based on UCB is to transform the optimization end-side data offloading process into solving the MAB data offloading problem based on UCB to obtain the first selection configuration.

[0186] Optionally, the optimized end-to-edge data offloading process is transformed into solving the MAB data offloading problem based on UCB to obtain the first selected configuration, including steps S11-S14, specifically:

[0187] Step S11: Initialize the current unloading action, edge server selection variable, and first reward. When the first preset condition is met, the current high-frequency acquisition terminal sequentially selects the set of first rocker arms to obtain the initial reward, and uses the initial reward as the current first reward for the first solution.

[0188] In this embodiment, initialization x n,m (t) = 0, When t≤|A n Under the first preset condition, the terminal Select the rocker arm in turn and receive the initial reward.

[0189] Step S12: Within the current time slot, the current high-frequency acquisition terminal calculates the upper bound of the confidence level based on the current number of first rocker arms selected, specifically as follows:

[0190]

[0191] in, This is the upper bound of the confidence level. β represents the average reward over the first t-1 time slots. n Indicates the current high-frequency acquisition terminal dn The sampling frequency weight, Indicates the current first rocker arm The confidence interval, The current number of selections for the first rocker arm, t represents the current t-th time slot, and t-1 represents the previous t-1-th time slot;

[0192] In this embodiment, in the t-th time slot, d n according to Number of choices The upper bound of the confidence level is calculated as follows:

[0193]

[0194] in, β represents the average reward over the first t-1 time slots; n d n The sampling frequency weight; express The confidence interval.

[0195] It should be noted that the high-frequency data acquisition terminal d n The higher the sampling frequency weight, the more data can be collected in each time slot. To ensure transmission latency and energy efficiency, optimal decision-making is essential. If β... n If β is large and the reward value of the selected rocker arm is high, then the confidence interval is small, and the terminal tends to utilize the currently selected rocker arm. If β n If the value is smaller, the selected rocker arm will have a lower reward value, resulting in a larger confidence interval, and the terminal will tend to explore other rocker arms.

[0196] Step S13: Based on the upper bound of the confidence level, the current high-frequency acquisition terminal selects the first rocker arm with the highest confidence level to perform the action, selects the current transmission power to transmit the data to the current edge server, and obtains the first selection configuration;

[0197] In this embodiment, after obtaining After that, d n Choose the rocker arm with the highest confidence upper bound. To perform an action, represented as:

[0198]

[0199] Step S14: Based on the total latency of data processing, the energy consumption of data transmission during data unloading, the weight of energy consumption, and the first selection configuration, obtain the current first reward, and based on the current first reward, update the average reward of the previous time slot and the current selection number of the first rocker arm, until the calculation of all time slots in the time slot set of the total data processing optimization cycle is completed.

[0200] In this embodiment, the terminal comprehensively observes latency and energy efficiency performance, including total data processing latency, data transmission energy consumption during data offloading, energy consumption weights, and the first selected configuration, and based on... Receive rewards therefore, and Updated to:

[0201]

[0202] Step 103: Receive unloaded data through the selected edge servers, perform edge-cloud data splitting optimization processing on the selected edge servers according to the cloud-edge data splitting model and the unloaded data, obtain the second selection configuration, and split the unloaded data to the cloud server through the selected edge servers according to the second selection configuration to obtain the split data; wherein, the second selection configuration includes the selected data splitting ratio.

[0203] Optionally, based on the cloud-edge data offloading model and offloaded data, edge-cloud data offloading optimization is performed on each selected edge server to obtain a second alternative configuration, specifically:

[0204] In the cloud-edge data offloading model, each selected edge server is modeled as a decision-maker for edge-cloud data offloading optimization based on EXP3, and the data offloading ratio of each selected edge server is modeled as a rocker arm. The problem of choosing the offloading ratio when each selected edge server will unload data offloading is solved to obtain the second choice configuration.

[0205] In this embodiment, a traffic splitting ratio is used in the cloud-edge data transmission and cloud-edge collaborative data processing stages. For the selection of the splitting ratio, the edge server is modeled as the decision-maker in the EXP3-based edge-cloud data splitting optimization algorithm, and the data splitting ratio of the edge server is modeled as a rocker arm to solve the problem of selecting the splitting ratio when splitting data from the edge server. The selection of the edge server data splitting ratio, i.e., the second selection configuration, is achieved through the EXP3-based edge-cloud data splitting optimization algorithm.

[0206] Optional cloud-edge data offloading model, specifically:

[0207] Within the current time slot, based on the current maintenance data backlog queue of the edge server unloading data from the current high-frequency acquisition terminal, the maintenance data backlog queue for the next time slot is dynamically updated, i.e., the edge server s m To from terminal d n The unloaded data maintains the data backlog queue (currently maintaining the data backlog queue). Dynamic queue updates are performed as follows:

[0208]

[0209] in, For the maintenance data backlog queue of the (t+1)th time slot, For the current edge servers s m From the current high-frequency acquisition terminal d n The current maintenance data backlog queue for unloading data in the middle. This represents the current amount of data being unloaded. For the current edge servers s m The amount of data processed, y n，m (t) represents the current data split ratio, indicating that d n From s m Data splitting ratio to s0; y n,m (t)∈Y, where Y is the set of data split ratios, containing H levels, specifically:

[0210]

[0211] Among them, y min (t) and y max (t) represents the minimum and maximum values ​​of the data split ratio, respectively; H represents the number of levels included in the data split ratio set; and h represents the data split ratio selected by the current edge server as the h-th level.

[0212] The current data splitting ratio is calculated based on the selected data splitting ratio level of the current edge server, the minimum value of the data splitting ratio, and the maximum value of the data splitting ratio. The formula is as follows:

[0213]

[0214] Among them, y n,m (t) represents the current data split ratio, i.e., if s m If the data splitting ratio is selected as the h-th level, then

[0215]

[0216] Based on the current data splitting ratio, server transmission rate, maximum duration of cloud-edge data transmission phase, and current offloaded data volume, calculate the edge offloaded data volume and edge server transmission latency, specifically:

[0217]

[0218]

[0219] in, The amount of data offloaded at the edge represents the amount of data transferred from the current edge server s in the t-th time slot. mThe current high-frequency acquisition terminal d is unloaded to cloud server s0. n The amount of data, τ2 represents the current unloaded data volume, τ2 represents the maximum time duration of the cloud-edge data transmission phase, and y represents the data volume currently unloaded. n,m (t) is the current data split ratio. For the current edge servers s m The transmission rate between the cloud server s0 and the server (server transmission rate) For edge server transmission latency, it represents the latency from the current edge server s within the t-th time slot. m The transmission latency of data sent to cloud server s0. In this embodiment, the server transmission rate... The calculation method and the current data transmission rate R n,m The calculation method for (t) is the same.

[0220] Optionally, each selected edge server is modeled as a decision-maker in the edge-cloud data offloading optimization process based on EXP3, and the data offloading ratio of each selected edge server is modeled as a rocker arm. The problem of choosing the offloading ratio when each selected edge server unloads data offloading is solved to obtain the second alternative configuration, specifically:

[0221] Based on the second optimization variable, each selected edge server optimizes the edge-cloud data offloading process by minimizing the latency required for the server to process all data. The second optimization variable is the current data offloading ratio, and the latency required for the server to process all data includes the edge server transmission latency, the cloud server processing latency, and the edge server processing latency. The optimization of the edge-cloud data offloading process specifically involves:

[0222]

[0223]

[0224] Among them, SP2 is the edge-cloud data offloading process, C5 is the data offloading ratio selection constraint, and y n,m (t) represents the current data split ratio. For edge server transmission latency, To handle latency for cloud servers, For edge server latency processing, Y represents the set of data splitting ratios, and d n For the current high-frequency acquisition terminal d n s m For the current edge servers s m D is the set of N high-frequency acquisition terminals, and S is the set of M edge servers;

[0225] In this embodiment, based on the offloading decision obtained by optimizing SP1, SP2 optimizes the edge-cloud data offloading process, and its optimization variables include y. n,m (t). Edge servers achieve optimal offloading decisions by minimizing the latency required to process all data. Similarly, SP2 is transformed into the MAB problem (MAB Data Offloading Problem), which involves the selection of decision-makers, cranes, actions, and rewards.

[0226] The edge-cloud data offloading process is transformed into solving the MAB data offloading problem based on EXP3 to obtain the second alternative configuration. The MAB data offloading problem is specifically designed to determine the offloading ratio when offloading data from edge servers, and includes:

[0227] Optionally, each selected edge server can be used as a second decision-maker to make decisions on the offloading ratio when offloading data from the edge to the cloud.

[0228] In this embodiment, the decision-maker for the MAB data offloading problem is defined as the edge server, which makes the offloading ratio decision when performing edge-cloud data offloading.

[0229] Optionally, the total split ratio of the data split ratio set can be used as the second rocker arm;

[0230] In this embodiment, the rocker arm for the MAB data splitting problem is defined as the total split ratio based on Y. Represents edge server s m The specified d n The data routing level is h.

[0231] Optionally, the second rocker arm selection action indicator variable can be used as the current diversion action; where, Current traffic diversion action, The current data split ratio represents the current edge server s. m In the t-th time slot, the high-frequency acquisition terminal d will be used. n The data splitting rate setting is set by the following formula:

[0232]

[0233] Among them, y n,m (t) represents the current data split ratio, y min and y max These represent the minimum and maximum values ​​of the data splitting ratio, respectively. H represents the number of levels included in the data splitting ratio set, and h represents the data splitting ratio selected by the current edge server as the h-th level.

[0234] In this embodiment, the action for the MAB data offloading problem is: the action indicator variable is... in Represents edge server s m The terminal d will be in the t-th time slot n Data splitting rate set to

[0235]

[0236] Optionally, based on the latency required for the server to process all data, the edge server of the second crane, selected with the highest data splitting ratio, is given as a second bonus, specifically:

[0237]

[0238] in, As the second reward, For edge server transmission latency, To handle latency for cloud servers, To handle latency for edge servers.

[0239] In this embodiment, the reward for the MAB data offloading problem is: Indicates the choice y n,m (t) the rocker arm's s m The reward obtained is represented as:

[0240]

[0241] Optionally, the edge-cloud data splitting process can be transformed into solving the MAB data splitting problem based on EXP3 to obtain the second alternative configuration, specifically including steps S21-S23, as follows:

[0242] Step S21: Initialize the uniform distribution parameters and preset the performance-related empirical distribution parameters;

[0243] In this embodiment, the uniform distribution parameter ξ∈(0,1) is initialized. Empirical distribution parameters related to performance are then set.

[0244]

[0245] Step S22: Within the current time slot, calculate the probability of selecting the current second rocker arm, specifically as follows:

[0246]

[0247] in, Let ξ∈(0,1] be the probability of selecting the current second rocker arm, and let ξ∈(0,1] be a uniform distribution parameter. Here, H represents the empirical distribution parameters related to the performance of the current t-th time slot, and H is the number of levels included in the data split ratio set.

[0248] Based on the probability of selecting the current second rocker arm, calculate the current cumulative distribution function, as follows:

[0249]

[0250] Among them, F n,m (h) is the current cumulative distribution function;

[0251] Based on the current cumulative distribution function, the second alternative configuration is obtained as follows:

[0252]

[0253] in, This represents the current traffic splitting action, indicating the second-choice configuration. hour,

[0254] In this embodiment, firstly, the selection is calculated in the t-th time slot. The probability of:

[0255]

[0256] Next, calculate the cumulative distribution function:

[0257]

[0258] Finally, a random value is generated, and the optimal split ratio decision, the second choice configuration, is obtained, represented as:

[0259]

[0260] In particular, when hour,

[0261] Step S23: Each selected edge server performs the second selection configuration. Based on the latency required for the server to process all data, the selected edge server of the second rocker arm with the data split ratio obtains the current second reward. Based on the current second reward, the performance-related empirical distribution parameters of the next time slot are updated until the calculation of all time slots in the time slot set of the total data processing optimization cycle is completed.

[0262] Specifically, based on the current second reward, the performance-related empirical distribution parameters for the next time slot are updated as follows:

[0263]

[0264]

[0265] in, The empirical distribution parameters related to the performance of the (t+1)th time slot are... To determine the probability of selecting the current second rocker arm, The empirical distribution parameters related to the performance of the current time slot t are... To estimate the reward, η is an adjustment factor. For the current second reward, H represents the number of levels included in the data splitting ratio set.

[0266] In this embodiment, the edge server performs the traffic splitting ratio decision. And according to the formula Receive rewards Update the performance-related empirical distribution parameters using the following formula:

[0267]

[0268] Where η>0 is an adjustment factor for the empirical performance-related distribution parameter. This is the estimated reward, expressed as:

[0269]

[0270] Finally, the algorithm terminates when t > T.

[0271] Step 104: The offloaded data is processed using the cloud server according to the cloud-edge collaborative data processing model and the second selection configuration.

[0272] In this embodiment, the edge-cloud data offloading optimization algorithm based on EXP3 is used to select the data offloading ratio for the edge server. The edge server then offloads data to the cloud server, where the offloaded data undergoes cloud-edge collaborative data processing. The cloud-edge computing architecture and collaborative mechanism of the low-voltage power distribution information high-frequency acquisition system fully considers the acquisition frequency of the high-frequency acquisition terminal during the optimization process. Through continuous learning and updating, it makes more reasonable server and transmission power selection decisions to meet the differentiated data processing needs of the high-frequency acquisition terminal.

[0273] Optional, cloud-edge collaborative data processing model, specifically:

[0274] Based on the current data backlog queue, the current data splitting ratio, the current amount of data unloaded, the edge server's computing power, the terminal data processing complexity, and the maximum duration of the cloud-edge collaborative data processing phase, the data processing volume and latency of the edge server are calculated using the following formulas:

[0275]

[0276]

[0277] in, The amount of data processed by the edge server, representing the current edge server s. m The amount of data processed (edge ​​servers) m Terminal d processed in the t-th time slot n (data volume) For the current edge servers s m From the current high-frequency acquisition terminal d n The current maintenance data backlog queue for unloading data in the middle, y n,m (t) represents the current data split ratio. This represents the current amount of data being unloaded. The current edge server s m The computing power in the t-th time slot, χ n For the current high-frequency acquisition terminal d n The terminal data processing complexity is τ3, where τ3 is the maximum time length of the cloud-edge collaborative data processing stage. For edge server latency handling, it represents the current edge server s m Processing delay in the t-th time slot;

[0278] Based on the current data backlog queue, the data volume processed by edge servers, and the data volume processed by cloud servers, the data backlog queue for the next time slot is dynamically updated, specifically as follows:

[0279]

[0280] in, This is the backlog queue for the offloaded data in the (t+1)th time slot. For cloud server s0, from the current edge server s m The current high-frequency acquisition terminal d is being diverted. n The current backlog of diverted data (cloud server s0 is from edge server s) m The terminal d of the traffic split n Data maintenance data backlog queue), To handle the amount of data for edge servers, The amount of data processed by the cloud server; specifically, the amount of data processed by the cloud server is as follows:

[0281]

[0282] in, For edge server transmission latency, To offload data volume at the edge, χ is the computing power of cloud server s0 in the t-th time slot. n For the current high-frequency acquisition terminal d nThe terminal data processing complexity, τ3 is the maximum time length of the cloud-edge collaborative data processing stage;

[0283] Based on the current backlog of diverted data, edge server transmission latency, edge offloaded data volume, cloud server computing power, terminal data processing complexity, and the maximum duration of the cloud-edge collaborative data processing phase, the cloud server processing latency is calculated as follows:

[0284]

[0285] in, For cloud server latency processing, it represents d in s0. n Data processing latency.

[0286] In this embodiment, to further illustrate the technical effect of data acquisition collaborative processing under the cloud-edge computing architecture and collaborative mechanism of the low-voltage power distribution information high-frequency acquisition system of the present invention, simulation latency is performed to demonstrate the technical effect of the present invention through simulation results. A cloud-edge collaborative data processing scenario is considered for a low-voltage power distribution substation area, including 10 power distribution acquisition devices, 3 edge servers, and one cloud server. The number of data acquisitions per device per time slot is distributed within [1.2-1.8] Mbits. Transmission power and data splitting ratio are divided into 5 and 6 levels, respectively. Simultaneously, two innovative algorithms are compared with the algorithm of the present invention, as follows:

[0287] (1) Multi-index evaluation learning-based computation offloading algorithm (MINCO) takes minimizing the average total data processing delay as the optimization objective, but lacks energy consumption control of the equipment.

[0288] (2) UCB-Advantage actor-critic-based dataoffloading algorithm (UCB-A3C), which takes into account energy consumption management and transmission delay optimization.

[0289] Meanwhile, both comparison algorithms employ the traditional binary complete offloading strategy for edge server data offloading, that is, for and The data splitting ratios all satisfy y n,m (t)∈{0,1}.

[0290] The data was simulated using three different algorithms. A comparison of the weighted sum of data processing latency and terminal power consumption for each algorithm was presented, such as... Figure 3 As shown, the simulation results illustrate the variation of the weighted sum of data processing latency and terminal power consumption over time slots. Compared to MINCO and UCB-A3C, the proposed algorithm (the algorithm of this invention) has the lowest weighted sum, with weighted sum performance reduced by 19.69% and 16.05%, respectively. This is because the proposed algorithm can coordinate the balance between total latency and power consumption by adjusting the device's transmission power and the data offloading ratio of the edge server, thereby reducing power consumption while ensuring low latency. However, MINCO only focuses on reducing latency and ignores power consumption balance; while UCB-A3C considers power management, insufficient utilization of cloud-edge computing resources leads to poor weighted sum performance.

[0291] Different algorithms are used to compare cumulative energy consumption charts, such as Figure 4 As shown, the relationship between cumulative energy consumption and time slots is illustrated. Compared to MINCO and UCB-A3C, the proposed algorithm (the algorithm of this invention) reduces cumulative energy consumption by 15.04% and 9.52%, respectively. This is because the proposed algorithm can coordinate the balance between total latency and energy consumption through joint optimization of task offloading and computational resource allocation. MINCO only considers the optimization of data offloading latency, ignoring the coupling relationship between data transmission power and data offloading latency, which leads to the highest energy consumption. UCB-A3C lacks optimization for the edge cloud data splitting process, failing to fully utilize the computational resources of cloud servers and edge servers, resulting in severe data backlog queues and increased energy consumption for device data offloading.

[0292] Through simulation comparison of the three algorithms, the advantages of this invention are as follows: (1) The joint optimization of terminal transmission power selection, edge server selection and data diversion is achieved by using the frequency sensing end-edge data offloading optimization algorithm based on UCB and the edge-cloud data diversion optimization algorithm based on EXP3. A cloud-edge computing architecture and collaborative mechanism for a low-voltage power distribution information high-frequency acquisition system is proposed, and the joint optimization is decomposed into the transmission power selection and edge server selection stages and the data diversion stage. In the first stage, the transmission power of the high-frequency acquisition terminal and the edge server are selected by using the frequency sensing end-edge data offloading optimization algorithm based on UCB; in the second stage, the edge-cloud data diversion optimization algorithm based on EXP3 is used to select the data diversion ratio of the edge server. (2) Differentiated offloading decisions are made for terminals with different acquisition frequencies to meet the processing needs of different acquisition terminals. When processing the acquisition data in the cloud-edge computing architecture and collaborative mechanism of the low-voltage power distribution information high-frequency acquisition system, the acquisition frequency of the high-frequency acquisition terminal is fully considered in the optimization process. Through continuous learning and updating, more reasonable server and transmission power selection decisions are made to meet the differentiated data processing needs of the high-frequency acquisition terminal.

[0293] Implementing this invention, the cloud-edge computing architecture and collaborative mechanism of the low-voltage power distribution information high-frequency acquisition system decomposes joint optimization into a transmission power selection and edge server selection stage (first selection configuration) and a data offloading stage (second selection configuration). In the first stage, the transmission power of the high-frequency acquisition terminal and the edge server are selected through an end-to-edge data offloading model and end-to-edge data offloading optimization. In the second stage, the data offloading ratio of the edge server is selected through a cloud-to-edge data offloading model and edge-to-cloud data offloading optimization, achieving joint optimization of the high-frequency acquisition terminal transmission power selection, edge server selection, and data offloading, and enabling collaborative data processing in the cloud server. During the joint optimization process, the acquisition frequency and transmission power of the high-frequency acquisition terminal are fully considered. Through continuous learning and updating, more reasonable server and transmission power selection decisions are made to meet the differentiated data processing needs of the high-frequency acquisition terminal and improve data processing performance.

[0294] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention in detail. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for collaborative data acquisition and processing of low-voltage power distribution information, characterized in that, include: Based on the hierarchical devices of the low-voltage power distribution information high-frequency acquisition system, a cloud-edge computing architecture is constructed; wherein, the hierarchical devices of the high-frequency acquisition system include one or more high-frequency acquisition terminals in the terminal layer, one or more edge servers in the edge layer, and one cloud server in the cloud layer. In each time slot of the overall data processing optimization cycle under the cloud-edge computing architecture, low-voltage power distribution data is collected by each of the high-frequency acquisition terminals. Based on the low-voltage power distribution data and the end-edge data offloading model, end-edge data offloading optimization processing is performed on each of the high-frequency acquisition terminals to obtain a first selected configuration. The low-voltage power distribution data is then offloaded to selected edge servers by each of the high-frequency acquisition terminals according to the first selected configuration to obtain offloaded data. The first selected configuration includes the selected edge servers and the selected transmission power. The process of obtaining the first selected configuration is as follows: In the end-edge data offloading model, each high-frequency acquisition terminal is modeled as a decision-maker based on UCB frequency sensing end-edge data offloading optimization processing, and the transmission power level of each edge server and each high-frequency acquisition terminal during data offloading is modeled as a rocker arm. The selection problem of edge server and transmission power when each high-frequency acquisition terminal unloads the low-voltage power distribution data is solved to obtain the first selected configuration. The selected edge servers receive the offloaded data, and based on the cloud-edge data splitting model and the offloaded data, the selected edge servers undergo edge-cloud data splitting optimization processing to obtain a second selection configuration. The selected edge servers then split the offloaded data to the cloud server according to the second selection configuration to obtain split data. The second selection configuration includes the selected data splitting ratio. The process of obtaining the second selection configuration is as follows: In the cloud-edge data diversion model, each selected edge server is modeled as a decision-maker for edge-cloud data diversion optimization based on EXP3, and the data diversion ratio of each selected edge server is modeled as a rocker arm. The problem of selecting the diversion ratio when each selected edge server diverts the unloaded data is solved to obtain the second selection configuration. The cloud server performs cloud-edge collaborative data processing on the offloaded data according to the cloud-edge collaborative data processing model and the second selected configuration.

2. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 1, characterized in that, The end-to-edge data offloading model is specifically as follows: Within the current time slot, based on the amount of data collected at the current high-frequency acquisition terminal, the data stored at the current high-frequency acquisition terminal is modeled as the data backlog queue for the current time slot, and the data backlog queue for the next time slot is dynamically updated, specifically as follows: in, For the first Data backlog queue for each time slot For the first Data backlog queue for each time slot For the current number Each time slot For the first Each time slot For the first The current high-frequency acquisition terminal within each time slot The amount of data currently collected. Choose variables for the edge server. The current amount of data uninstalled; wherein, the current amount of data uninstalled is the amount of data uninstalled at the time. The current high-frequency acquisition terminal within each time slot Unload to the current edge server The amount of data; The current unloaded data volume is obtained based on the data backlog queue of the current time slot, the maximum time length of the end-to-side data unloading phase, and the data transmission rate, specifically: in, This refers to the maximum time length of the end-to-edge data unloading phase. The current high-frequency acquisition terminal and the current edge server The current data transfer rate between them; The current data transmission rate is obtained based on the current data transmission bandwidth, the current data transmission channel gain, the current Gaussian white noise, the current electromagnetic interference power, and the current transmission power, specifically: in, The current high-frequency acquisition terminal and the current edge server The bandwidth of current data transmission between them. The current high-frequency acquisition terminal and the current edge server The channel gain of the current data transmission between them. The current high-frequency acquisition terminal and the current edge server The current Gaussian white noise between, The current high-frequency acquisition terminal and the current edge server The current electromagnetic interference power between them The current high-frequency acquisition terminal The current transmission power; where, The set of transmission power is as follows: in, It is the transmit power level. and These respectively represent the current high-frequency acquisition terminals. Minimum and maximum transmission power, Power rating; The terminal transmission delay is calculated based on the current data transmission rate, the maximum duration of the end-to-side data offloading phase, the data backlog queue of the current time slot, and the amount of data currently collected. Specifically: in, The terminal transmission delay is represented by the delay at the 1st minute. Within a time slot, from the current high-frequency acquisition terminal To the current edge server The transmission delay of data offloading; The current data transmission rate, This refers to the maximum time length of the end-to-edge data unloading phase. For the first The current high-frequency acquisition terminal within each time slot The amount of data currently collected. For the first Data backlog queue for each time slot; Based on the current transmission power and the terminal transmission delay, the data transmission energy consumption during data offloading is calculated as follows: in, The data transmission energy consumption during data offloading is represented in the first... Within a time slot, data is collected from the current high-frequency acquisition terminal. To the current edge server Data transmission energy consumption during data offloading.

3. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 2, characterized in that, The process involves modeling each high-frequency acquisition terminal as a decision-maker based on UCB frequency sensing end-edge data offloading optimization processing, and modeling the transmission power levels of each edge server and each high-frequency acquisition terminal during data offloading as a rocker arm. The process then solves the problem of selecting the edge server and transmission power when each high-frequency acquisition terminal offloads the low-voltage power distribution data, obtaining the first selected configuration, specifically: Based on the first optimization variable, each high-frequency acquisition terminal optimizes the end-to-side data offloading process by minimizing the weighted sum of the total data processing latency and the data transmission energy consumption during data offloading; wherein, the first optimization variable is the current transmission power and the edge server selection variable; the optimized end-to-side data offloading process specifically includes: in, For the optimized end-to-side data offloading process, Select variables for the edge server. The total latency of the data processing is [missing information]. The data transmission energy consumption during data unloading. It is the weight of energy consumption. The current high-frequency acquisition terminal , For the current edge server , for A collection of high-frequency acquisition terminals. for A collection of edge servers The set of time slots for the total optimization cycle of data processing. The current high-frequency acquisition terminal The current transmission power, For the collection of transmission power, For the current edge server Maximum number of terminals that can be processed , and Choose constraints for edge servers. It is a transmit power selection constraint; The optimized edge-to-edge data offloading process is transformed into solving the MAB data offloading problem based on UCB to obtain the first selected configuration; wherein, the MAB data offloading problem is the selection problem of edge servers and transmission power when each of the high-frequency acquisition terminals offloads the low-voltage power distribution data, specifically including: Each of the aforementioned high-frequency acquisition terminals is taken as the first decision-maker, making decisions on edge server selection and power selection for data offloading; The combination of the power level and each of the edge servers is used as the first rocker arm, specifically: in, This refers to the set of the first rocker arms; The first rocker arm represents the power level. and the current edge server The combination; The first rocker arm selection action indicator variable is used as the current unloading action; where... For the current uninstallation action, This indicates that the current high-frequency acquisition terminal selects the current transmission power to transmit data to the current edge server, wherein the current transmission power is defined by the formula: in, The current high-frequency acquisition terminal The current transmission power, It is the transmit power level. and These respectively represent the current high-frequency acquisition terminals. Minimum and maximum transmission power, Power rating; Based on the first rocker arm selected by the current high-frequency acquisition terminal within the current time slot as the first reward, specifically: in, For the current first Each time slot For the first reward, The weight of the energy consumption. The total latency of the data processing is [missing information]. The energy consumption for data transmission during data unloading.

4. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 3, characterized in that, The process of optimizing the edge-to-edge data offloading process is transformed into solving the MAB data offloading problem based on UCB to obtain the first selected configuration, specifically as follows: The current unloading action, the edge server selection variable, and the first reward are initialized. When the first preset condition is met, the current high-frequency acquisition terminal sequentially selects the set of the first rocker arms to obtain the initial reward, and the initial reward is used as the current first reward for the first solution. Within the current time slot, the current high-frequency acquisition terminal calculates the upper bound of the confidence level based on the current number of selected first rocker arms, specifically as follows: in, This is the upper bound of the confidence level. Indicates the preceding Average reward per time slot This indicates the current high-frequency acquisition terminal. The sampling frequency weight, Indicates the current first rocker arm The confidence interval, The number of selected first rocker arms is currently... Indicates the current first Each time slot Indicates the previous one One time slot; Based on the upper bound of the confidence level, the current high-frequency acquisition terminal selects the first rocker arm with the highest upper bound of the confidence level to perform the action, selects the current transmission power to transmit the data to the current edge server, and obtains the first selection configuration. Based on the total latency of the data processing, the data transmission energy consumption during data unloading, the weight of the energy consumption, and the first selection configuration, the current first reward is obtained, and based on the current first reward, the average reward of the previous time slot and the current number of selections of the first rocker arm are updated until the calculation of all time slots in the time slot set of the total data processing optimization cycle is completed.

5. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 1, characterized in that, The second selected configuration is obtained by performing edge-cloud data splitting optimization processing on each selected edge server based on the cloud-edge data splitting model and the offloaded data, specifically as follows: In the cloud-edge data offloading model, each selected edge server is modeled as a decision-maker for edge-cloud data offloading optimization processing based on EXP3, and the data offloading ratio of each selected edge server is modeled as a rocker arm. The problem of choosing the offloading ratio when each selected edge server offloads the unloaded data is solved to obtain the second selection configuration.

6. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 5, characterized in that, The cloud-edge data offloading model is specifically as follows: Within the current time slot, based on the current maintenance data backlog queue where the edge server is unloading data from the current high-frequency acquisition terminal, the maintenance data backlog queue for the next time slot is dynamically updated, specifically as follows: in, For the first Maintenance data backlog queue for each time slot, For the current edge server From the current high-frequency acquisition terminals The current maintenance data backlog queue for unloading data in the middle. This represents the current amount of data being unloaded. For the current edge server The amount of data processed Given the current data split ratio, , The data splitting ratio set is specifically defined as follows: in, and These represent the minimum and maximum values ​​of the data splitting ratio, respectively. This indicates the number of levels included in the data splitting ratio set. This indicates that the data splitting ratio selected by the current edge server is the [number]th [percentage]. Each level; The current data splitting ratio is calculated based on the data splitting ratio level selected by the current edge server, the minimum value of the data splitting ratio, and the maximum value of the data splitting ratio, using the following formula: in, The current data split ratio; Based on the current data splitting ratio, server transmission rate, maximum cloud-edge data transmission phase duration, and current offloaded data volume, calculate the edge offloaded data volume and edge server transmission latency, specifically as follows: in, The amount of data unloaded at the edge represents the amount of data unloaded at the edge. Within a time slot, from the current edge server Unload to the cloud server The current high-frequency acquisition terminal The amount of data, The amount of data currently being unloaded. This refers to the maximum time length of the cloud-edge data transmission phase. This is the current data splitting ratio. For the current edge server and the cloud server The server transfer rate between them The transmission latency of the edge server represents the first... Within a time slot, from the current edge server Transmitted to the cloud server The data transmission latency.

7. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 6, characterized in that, The selected edge servers are modeled as decision-makers for edge-cloud data offloading optimization based on EXP3, and the data offloading ratio of the selected edge servers is modeled as a rocker arm. The problem of choosing the offloading ratio when the selected edge servers offload the offloaded data is solved to obtain the second selection configuration, specifically: Based on the second optimization variable, the selected edge servers optimize the edge-cloud data offloading process by minimizing the latency required for the servers to process all data; wherein, the second optimization variable is the current data offloading ratio, and the latency required for the servers to process all data includes the edge server transmission latency, the cloud server processing latency, and the edge server processing latency; the optimization of the edge-cloud data offloading process specifically involves: in, For the edge-cloud data offloading process, Choose constraints for the data split ratio. The current data split ratio, For the transmission latency of the edge server, To handle latency for cloud servers, To handle latency for the edge server, For data split ratio set, The current high-frequency acquisition terminal , For the current edge server , for A collection of high-frequency acquisition terminals. for A collection of edge servers; The edge-cloud data offloading process is transformed into solving the MAB data offloading problem based on EXP3 to obtain the second selected configuration; wherein, the MAB data offloading problem is the problem of selecting the offloading ratio when offloading data from the edge server, specifically including: Each of the selected edge servers is used as a second decision-maker to make a decision on the diversion ratio when diverting data between the edge and cloud. The total segmentation ratio of the data splitting ratio set is used as the second rocker arm; The second rocker arm selection action indicator variable is used as the current diversion action; wherein... The current traffic splitting action, The current data split ratio represents the current edge server ratio. In the Each time slot will come from the current high-frequency acquisition terminal. The data splitting rate setting is set by the following formula: in, This indicates the current data splitting ratio. and These represent the minimum and maximum values ​​of the data splitting ratio, respectively. This indicates the number of levels included in the data splitting ratio set. This indicates that the data splitting ratio selected by the current edge server is the [number]th [percentage]. Each level; Based on the latency required for the server to process all data, the edge server of the second crane, selected based on the data splitting ratio, is given as the second reward, specifically: in, For the second reward, For the transmission latency of the edge server, To handle latency for cloud servers, To handle latency for the edge server.

8. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 7, characterized in that, The process of transforming the edge-cloud data splitting process into solving the MAB data splitting problem based on EXP3 to obtain the second selection configuration is as follows: Initialize uniform distribution parameters and preset performance-related empirical distribution parameters; Within the current time slot, the probability of selecting the current second rocker arm is calculated as follows: in, The probability of selecting the current second rocker arm. The uniform distribution parameter is... For the current number Empirical distribution parameters related to the performance of each time slot The number of levels included in the data splitting ratio set; Based on the probability of selecting the current second rocker arm, calculate the current cumulative distribution function, as shown in the formula: in, This is the current cumulative distribution function; Based on the current cumulative distribution function, the second selection configuration is obtained as follows: in, The current traffic splitting action represents the second selected configuration, when hour, ; Each of the selected edge servers executes the second selection configuration. Based on the latency required for the server to process all data, the edge server of the second jib arm with the selected data split ratio obtains the current second reward. Based on the current second reward, the performance-related empirical distribution parameters of the next time slot are updated until the calculation of all time slots in the time slot set of the total data processing optimization cycle is completed. Specifically, based on the current second reward, the performance-related empirical distribution parameters for the next time slot are updated as follows: in, For the first Empirical distribution parameters related to the performance of each time slot The probability of selecting the current second rocker arm. For the current number Empirical distribution parameters related to the performance of each time slot To estimate the reward, To adjust the factor, For the current second reward, This indicates the number of levels included in the data splitting ratio set.

9. The data acquisition and collaborative processing method for low-voltage power distribution information as described in claim 6, characterized in that, The cloud-edge collaborative data processing model is specifically as follows: Based on the current maintenance data backlog queue, the current data splitting ratio, the current unloaded data volume, the edge server computing power, the terminal data processing complexity, and the maximum time length of the cloud-edge collaborative data processing phase, the edge server data processing volume and edge server processing latency are calculated using the following formula: in, The amount of data processed by the edge server represents the current edge server. The amount of data processed For the current edge server From the current high-frequency acquisition terminals The current maintenance data backlog queue for unloading data in the middle. The current data split ratio, The amount of data currently being unloaded. It is the current edge server In the Computing capacity within a time slot The current high-frequency acquisition terminal The complexity of terminal data processing This refers to the maximum time length of the cloud-edge collaborative data processing phase. The processing latency for the edge server indicates the current edge server. In the Processing delay within each time slot; Based on the current data backlog queue, the amount of data processed by the edge server, and the amount of data processed by the cloud server, the data backlog queue for the next time slot is dynamically updated, specifically as follows: in, For the first The offloaded data backlog queue for each time slot For the cloud server From the current edge server The current high-frequency acquisition terminal that is diverting the flow The current backlog of diverted data. The amount of data processed by the edge server The amount of data processed by the cloud server; specifically, the amount of data processed by the cloud server is: in, For the transmission latency of the edge server, For the amount of data unloaded at the edge, The cloud server mentioned In the Computing capacity within a time slot The current high-frequency acquisition terminal The complexity of terminal data processing This refers to the maximum time length of the cloud-edge collaborative data processing phase. The cloud server processing latency is calculated based on the current backlog of diverted data, the edge server transmission latency, the amount of data offloaded at the edge, the cloud server computing power, the terminal data processing complexity, and the maximum time length of the cloud-edge collaborative data processing phase. Specifically: in, To handle latency for the cloud server.

Images