CN-122001882-A - Cloud edge cooperation-based industrial equipment computing task dynamic allocation method

CN122001882ACN 122001882 ACN122001882 ACN 122001882ACN-122001882-A

Abstract

The invention relates to the technical field of cloud edge coordination, in particular to a cloud edge coordination-based industrial equipment computing task dynamic allocation method, which comprises the following steps of constructing a computing task global allocation model comprising a decentralization privacy protection mechanism, decomposing the computing task global allocation model into a plurality of equipment terminal optimization problems and a cloud global aggregation problem, realizing full-network computing power resource collaborative optimization on the premise of not collecting bottom data by a mechanism of iteratively interacting and updating global coordination variables and local intermediate variables, dynamically adjusting allocation strategies by using convergence residual errors as feedback indexes, ensuring that a final generating task allocation scheme is converged to a global optimal solution mathematically, reducing network bandwidth congestion pressure caused by real-time uploading of mass equipment state data, avoiding task processing delay caused by overload of a single-point data center, and improving response speed of an industrial Internet system when coping with large-scale concurrent requests.

Inventors

LIU SHIYU
DING YI
LIU JUN
YANG LU
JIANG HONGZHENG
GUAN JINGYU

Assignees

无锡学院

Dates

Publication Date: 20260508
Application Date: 20260211

Claims (10)

1. The cloud edge cooperation-based industrial equipment computing task dynamic allocation method is characterized by comprising the following steps of: Constructing a computing task global distribution model containing a decentralization privacy protection mechanism, and decomposing the computing task global distribution model into a plurality of equipment terminal optimization problems and a cloud global aggregation problem; Initializing a current global coordination variable; Obtaining private state data of each industrial device, and solving each device terminal optimization problem in parallel based on the current global coordination variable and the private state data to obtain a local intermediate variable corresponding to each industrial device; The local intermediate variable is sent to a cloud server, the cloud server utilizes the local intermediate variable to solve the cloud global aggregation problem, an updated global coordination variable is obtained, and the updated global coordination variable is broadcasted to each industrial device; And calculating a convergence residual of the updated global coordination variable, and if the convergence residual does not meet a preset convergence condition, taking the updated global coordination variable as the current global coordination variable, and returning to execute the step of parallelly solving the equipment terminal optimization problem based on the current global coordination variable and the private state data until the convergence residual meets the preset convergence condition to obtain a target task allocation scheme.
2. The cloud-edge collaboration-based dynamic allocation method of computing tasks for industrial equipment of claim 1, wherein the step of constructing a global allocation model of computing tasks including a decentralised privacy protection mechanism comprises: Constructing a deterministic network operation model, and generating time delay constraint conditions aiming at all industrial equipment by using the deterministic network operation model; And introducing the time delay constraint condition into a preset global objective function, decomposing the global objective function introduced with the time delay constraint condition by using an alternate direction multiplier method, and establishing the plurality of equipment terminal optimization problems and the cloud global aggregation problem.
3. The cloud-edge collaboration-based industrial equipment computing task dynamic allocation method according to claim 2, wherein the step of generating the time delay constraint condition for each industrial equipment by using the deterministic network evolution model comprises the following steps: modeling the computing task arrival flow of each industrial device as an arrival curve, and modeling the processing capacity of the industrial device for the computing task arrival flow as a service curve; calculating the maximum horizontal deviation between the arrival curve and the service curve, and determining the maximum horizontal deviation as a maximum time delay boundary; And establishing the time delay constraint condition according to the maximum time delay boundary.
4. The cloud edge collaboration-based dynamic allocation method for industrial equipment computing tasks according to claim 3, wherein the step of computing the maximum horizontal deviation between the arrival curve and the service curve is performed according to the following maximum time delay computation formula: Wherein Representing the maximum delay boundary in question, Representing an up-bound operation of the device, Representing the operation of the down-bound, A time of day variable is indicated and, A variable representing the time-offset is indicated, The arrival curve is represented by a graph of the arrival curve, Is shown at the moment The service profile at that time.
5. The cloud-edge collaboration-based industrial equipment computing task dynamic allocation method of claim 3, wherein the step of generating latency constraints for the industrial equipment using the deterministic network evolution model further comprises: Acquiring the task hard real-time deadlines of the industrial equipment; Based on the task hard real-time deadline and the arrival curve, reversely pushing by utilizing a reverse constraint meeting strategy to obtain a minimum service curve; And screening a candidate industrial equipment set meeting the time delay constraint condition according to the minimum service curve, and taking the candidate industrial equipment set as a solving domain of the calculation task global allocation model.
6. The cloud-edge collaboration-based dynamic allocation method of computing tasks for industrial equipment according to claim 1, wherein the step of solving each of the equipment terminal optimization problems in parallel based on the current global coordination variable and the private-state data comprises: Receiving the current global coordination variable broadcasted by the cloud server; Reading the private-state data, wherein the private-state data comprises a local task amount parameter, a local computing power parameter and a local privacy constraint parameter; constructing a local Lagrange function by using the private-state data, and carrying out minimization solution on the local Lagrange function on the premise of fixing the current global coordination variable to obtain an optimal unloading decision; generating the local intermediate variable comprising a lagrangian multiplier vector based on the optimal offload decision, the local intermediate variable not comprising the private-state data.
7. The cloud-edge collaboration-based industrial equipment computing task dynamic allocation method according to claim 1, wherein the step of solving the cloud global aggregation problem at the cloud server by using the local intermediate variables comprises: collecting all the local intermediate variables uploaded by the industrial equipment; Performing aggregation calculation on the local intermediate variables to obtain an aggregation result vector; And iteratively updating the current global coordination variable based on the aggregation result vector to generate the updated global coordination variable.
8. The cloud computing edge collaboration-based industrial equipment computing task dynamic allocation method of claim 1, further comprising: After the convergence residual meets the preset convergence condition, analyzing the target task allocation scheme to obtain a target unloading strategy of each industrial device; And dispatching the calculation tasks to be processed of the industrial equipment to the corresponding industrial equipment or the cloud server according to the target unloading strategy.
9. The cloud edge collaboration-based industrial equipment calculation task dynamic allocation method according to claim 2, wherein the global objective function is a weighted linear combination of minimizing the sum of energy consumption and processing time delay of all industrial equipment, the weight coefficient of energy consumption optimization in the weighted linear combination is 0.6, the weight coefficient of time delay performance is 0.4, a dual variable and a punishment parameter rho are introduced into the alternating direction multiplier method, and the initial value of the punishment parameter rho is 1.5 and is dynamically adjusted according to iteration times.
10. The cloud edge collaboration-based industrial equipment computing task dynamic allocation method according to claim 1, wherein the convergence residual comprises an original residual and a dual residual, and the original residual passes through a formula Calculating, wherein the dual residual error is calculated by a formula Calculating, wherein x k is a device local decision vector of the kth round of iteration, z k is a global coordination variable of the kth round of iteration, ρ is a penalty parameter, and the preset convergence condition is that the original residual is smaller than And the dual residual is less than 。

Description

Cloud edge cooperation-based industrial equipment computing task dynamic allocation method Technical Field The invention relates to the technical field of cloud edge coordination, in particular to a cloud edge coordination-based industrial equipment computing task dynamic allocation method. Background The cloud edge cooperative technology belongs to the cross front category of distributed computing and industrial Internet, and aims to construct a computing framework of cloud global management and control and edge agile response by integrating strong storage and mass data processing capability of a central cloud and low-time-delay and high-real-time computing capability of an edge side (namely a gateway, a base station or a local server close to a data source). In the prior art, a centralized management and control architecture is generally adopted, and industrial equipment which is required to be distributed at different physical positions must upload the total amount of original information such as self calculation force state, task queue length, energy consumption data, production parameters and the like to a central cloud or an edge management node in real time for unified processing. The data aggregation mode causes overlong and wide exposure of a core production data transmission link, and is easy to intercept or tamper maliciously in the transmission process, so that enterprise business confidentiality is leaked. As the number of access devices grows exponentially, massive original data concurrent uploading can instantaneously occupy limited backhaul network bandwidth, so that data queuing congestion and transmission delay are caused, and a cloud decision center is difficult to acquire the latest field state in time. Therefore, improvements are needed. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a Yun Bian-collaboration-based dynamic allocation method for computing tasks of industrial equipment. In order to achieve the above purpose, the invention adopts the following technical scheme that the cloud-edge collaboration-based industrial equipment computing task dynamic allocation method comprises the following steps: Constructing a computing task global distribution model containing a decentralization privacy protection mechanism, and decomposing the computing task global distribution model into a plurality of equipment terminal optimization problems and a cloud global aggregation problem; Initializing a current global coordination variable; Obtaining private state data of each industrial device, and solving each device terminal optimization problem in parallel based on the current global coordination variable and the private state data to obtain a local intermediate variable corresponding to each industrial device; The local intermediate variable is sent to a cloud server, the cloud server utilizes the local intermediate variable to solve the cloud global aggregation problem, an updated global coordination variable is obtained, and the updated global coordination variable is broadcasted to each industrial device; And calculating a convergence residual of the updated global coordination variable, and if the convergence residual does not meet a preset convergence condition, taking the updated global coordination variable as the current global coordination variable, and returning to execute the step of parallelly solving the equipment terminal optimization problem based on the current global coordination variable and the private state data until the convergence residual meets the preset convergence condition to obtain a target task allocation scheme. Preferably, the step of constructing a global distribution model of computing tasks including a decentralised privacy protection mechanism comprises: Constructing a deterministic network operation model, and generating time delay constraint conditions aiming at all industrial equipment by using the deterministic network operation model; And introducing the time delay constraint condition into a preset global objective function, decomposing the global objective function introduced with the time delay constraint condition by using an alternate direction multiplier method, and establishing the plurality of equipment terminal optimization problems and the cloud global aggregation problem. Preferably, the step of generating a time delay constraint condition for each industrial device by using the deterministic network evolution model includes: modeling the computing task arrival flow of each industrial device as an arrival curve, and modeling the processing capacity of the industrial device for the computing task arrival flow as a service curve; calculating the maximum horizontal deviation between the arrival curve and the service curve, and determining the maximum horizontal deviation as a maximum time delay boundary; And establishing the time delay constraint condition according to the maximum time delay boundar