Edge cloud resource allocation method and system

Abstract

This invention discloses an edge-cloud resource allocation method and system, applied in the field of edge-cloud resource allocation technology. The method includes: acquiring real-time operating data of insulators; determining each electric field analysis task and its priority based on anomaly analysis results of the real-time operating data; determining the edge-cloud resource requirements of each electric field analysis task based on the simulation processing results of each task; determining the resource scarcity of the electric field analysis system; determining an initial edge-cloud resource allocation strategy based on the edge-cloud resource requirements and task priorities corresponding to each electric field analysis task; optimizing the initial edge-cloud resource allocation strategy using a multi-objective optimization algorithm to obtain a target edge-cloud allocation strategy; and allocating corresponding edge-side resources and cloud-side resources to each electric field analysis task based on the target edge-cloud allocation strategy. The edge-cloud resource allocation method and system provided by this invention effectively improve the utilization efficiency of edge-cloud resources.

Description

Technical Field

[0001] This invention relates to the field of edge-cloud resource allocation technology, and in particular to an edge-cloud resource allocation method and system. Background Technology

[0002] Electric field analysis of insulators is a crucial step in ensuring the safe and stable operation of high-voltage power transmission and transformation equipment, as the electric field distribution directly reflects the insulation performance status. An edge-cloud collaborative insulator electric field analysis system significantly improves the real-time performance, accuracy, and robustness of insulation status perception by rapidly performing data preprocessing and initial anomaly detection at the edge, and then executing high-fidelity numerical calculations and multi-source fusion diagnostics in the cloud.

[0003] Existing real-time insulator electric field analysis technology processes the insulator electric field analysis task in the cloud. For example, the cloud needs to receive all the raw electric field data and 3D modeling information. This method not only leads to cloud overload, but also wastes the edge computing potential in the distributed architecture. As a result, the system as a whole cannot achieve efficient coordination of computing, storage and communication resources, ultimately leading to a significantly low overall utilization efficiency of edge and cloud resources. Summary of the Invention

[0004] This invention provides an edge-cloud resource allocation method and system to solve the technical problem of low overall utilization efficiency of existing edge-cloud resource allocation methods, thereby improving the utilization efficiency of edge-cloud resources.

[0005] To address the aforementioned technical problems, this invention provides an edge-cloud resource allocation method and system, the method comprising: Obtain the real-time operating data of the insulator; Based on the anomaly analysis results of the real-time operating data, each electric field analysis task and the task priority of each electric field analysis task are determined. Based on the simulation results of each electric field analysis task, the edge-cloud resource requirements of each electric field analysis task are determined. Determine the resource scarcity level of the electric field analysis system; Based on the edge cloud resource requirements and task priority corresponding to each electric field analysis task, an initial edge cloud resource allocation strategy is determined. Using the resource scarcity level as a constraint, the initial edge cloud resource allocation strategy is optimized using a multi-objective optimization algorithm to obtain the edge cloud objective allocation strategy; Based on the edge-cloud target allocation strategy, corresponding edge-side resources and cloud-side resources are allocated to each of the electric field analysis tasks.

[0006] Preferably, determining each electric field analysis task and its priority based on the anomaly analysis results of the real-time operating data includes: Based on the anomaly analysis results, each electric field analysis task is determined; Each of the electric field analysis tasks was analyzed to determine the influencing factors; A task priority quantification model is constructed based on each of the electric field analysis tasks and the corresponding influencing factors. The task priority of each electric field analysis task is obtained according to the task priority quantification model.

[0007] Preferably, determining the edge-cloud resource requirements for each electric field analysis task based on the simulation results of each electric field analysis task includes: Cluster analysis is performed on the simulation results to obtain the resource requirement feature vector for each electric field analysis task; By using the edge-cloud collaborative resource mapping model, the resource demand feature vector is subjected to heterogeneous resource adaptation processing to obtain the edge-cloud resource requirements of each of the electric field analysis tasks.

[0008] Preferably, determining the resource scarcity level of the electric field analysis system includes: Acquire the current edge-cloud resource usage status data of the electric field analysis system; Based on the edge cloud resource usage status data, a resource occupancy vector is obtained; The resource occupancy vector is subjected to a stress mapping process to obtain the resource stress level.

[0009] Preferably, determining the initial edge-cloud resource allocation strategy based on the edge-cloud resource requirements and task priorities corresponding to each of the electric field analysis tasks includes: Based on the edge cloud resource requirements and the task priorities, a resource request list is obtained; Construct a resource state space based on the task priorities: A delay calculation model is established based on the resource request list and resource state space to obtain the task delay of each electric field analysis task. Construct a multi-objective priority reward function based on the task delay; The resource request list is matched item by item based on the multi-objective priority reward function to obtain the initial edge cloud resource allocation strategy.

[0010] Another aspect of the present invention provides an edge-cloud resource allocation system, comprising: The acquisition module is used to acquire the real-time operating data of the insulator; The determination module is used to determine each electric field analysis task and the task priority of each electric field analysis task based on the anomaly analysis results of the real-time running data. The requirement module is used to determine the edge-cloud resource requirements of each electric field analysis task based on the simulation processing results of each electric field analysis task. The resource stress level module is used to determine the resource stress level of the electric field analysis system; The initial module is used to determine the initial edge cloud resource allocation strategy based on the edge cloud resource requirements and task priority corresponding to each electric field analysis task. The target module is used to optimize the initial edge cloud resource allocation strategy using a multi-objective optimization algorithm, with the resource scarcity level as a constraint, to obtain the edge cloud target allocation strategy. The allocation module is used to allocate corresponding edge-side resources and cloud-side resources to each of the electric field analysis tasks based on the edge-cloud target allocation strategy.

[0011] Preferably, the determining module includes: An anomaly analysis unit is used to determine each electric field analysis task based on the anomaly analysis results; The influence factor unit is used to analyze each of the electric field analysis tasks and determine the influence factors; The priority quantization model unit is used to construct a task priority quantization model based on each of the electric field analysis tasks and the corresponding influencing factors. The priority unit is used to obtain the task priority of each electric field analysis task according to the task priority quantization model.

[0012] Preferably, the requirement module includes: Clustering unit, used to perform cluster analysis on the simulation processing results to obtain the resource requirement feature vector of each electric field analysis task; The resource requirement unit is used to perform heterogeneous resource adaptation processing on the resource requirement feature vector using the edge-cloud collaborative resource mapping model to obtain the edge-cloud resource requirements of each of the electric field analysis tasks.

[0013] Preferably, the tension level module includes: The status data unit is used to acquire the current edge cloud resource usage status data of the electric field analysis system. The occupancy vector unit is used to obtain a resource occupancy vector based on the edge cloud resource usage status data; The mapping unit is used to perform stress mapping processing on the resource occupancy vector to obtain the resource stress level.

[0014] Preferably, the initial module includes: The resource request unit is used to obtain a resource request list based on the edge cloud resource requirements and the task priority; State space unit, used to construct resource state space according to the task priority: The task delay unit is used to establish a delay calculation model based on the resource request list and the resource state space to obtain the task delay of each of the electric field analysis tasks. The reward function unit is used to construct a multi-objective priority reward function based on the task delay. The allocation strategy unit is used to perform item-by-item resource matching on the resource request list based on the multi-objective priority reward function to obtain the initial edge cloud resource allocation strategy.

[0015] Compared with the prior art, the beneficial effects of the present invention are at least one of the following: This invention, by acquiring the real-time operational data of the insulator, firstly avoids uploading all raw electric field data and 3D modeling information to the cloud indiscriminately. Furthermore, based on the anomaly analysis results of the real-time operational data, it determines each electric field analysis task and its priority, thus triggering subsequent high-overhead simulation tasks only for insulators exhibiting anomalies or high risks, reducing unnecessary cloud computing load at the source. On this basis, based on the simulation processing results of each electric field analysis task, its edge-cloud resource requirements are determined, and an initial allocation strategy is constructed in conjunction with the current resource scarcity of the system, enabling lightweight tasks to be identified and directed to edge execution. Subsequently, using resource scarcity as a constraint, a multi-objective optimization algorithm is used to optimize the initial strategy, resulting in an edge-cloud target allocation strategy. This ensures that, under resource-constrained conditions, priority resource supply is guaranteed for high-priority tasks, and tasks are rationally distributed to the edge or cloud. Finally, based on this target allocation strategy, corresponding edge-side and cloud-side resources are allocated to each task, achieving on-demand, energy-efficient, and optimal task deployment. This invention effectively solves the problems of cloud overload and edge computing power idleness caused by full deployment and cloud-only operation in the prior art, and truly activates the distributed computing potential of the edge-cloud collaborative architecture. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the edge-cloud resource allocation method in one embodiment of the present invention; Figure 2 This is a schematic diagram of the edge-cloud resource allocation system in one embodiment of the present invention; Figure label: The modules are: 11. Acquisition module; 12. Determination module; 13. Requirement module; 14. Tension level module; 15. Initial module; 16. Target module; and 17. Allocation module. Detailed Implementation

[0017] 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. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0018] In the description of this invention, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0019] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0020] In the description of this invention, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0021] Electric field analysis of insulators is crucial for assessing their insulation performance and ensuring the safe operation of high-voltage power transmission and transformation equipment. Current technologies typically upload all raw data and 3D models to the cloud for centralized processing, leading to cloud computing overload and idle computing power at edge nodes, resulting in low efficiency of edge-cloud resource collaboration.

[0022] One embodiment of the present invention provides a method for edge-cloud resource allocation. For details, please refer to [link to relevant documentation]. Figure 1 , Figure 1 The diagram shown is a flowchart illustrating an edge-cloud resource allocation method according to one embodiment of the present invention, including: S1. Obtain real-time operating data of the insulator; S2. Based on the anomaly analysis results of real-time running data, determine each electric field analysis task and the task priority of each electric field analysis task. S3. Based on the simulation results of each electric field analysis task, determine the edge cloud resource requirements of each electric field analysis task. S4. Determine the resource constraints of the electric field analysis system; S5. Based on the edge cloud resource requirements and task priority corresponding to each electric field analysis task, determine the initial edge cloud resource allocation strategy. S6. Using the resource scarcity level as a constraint, optimize the initial edge cloud resource allocation strategy using a multi-objective optimization algorithm to obtain the edge cloud objective allocation strategy. S7. Based on the edge-cloud target allocation strategy, allocate corresponding edge-side resources and cloud-side resources to each electric field analysis task respectively.

[0023] This invention relates to an electric field analysis system for insulators. This system is a dedicated intelligent system for power system equipment condition monitoring and analysis, consisting of a hardware layer and a software layer. The hardware layer includes various sensing and monitoring devices deployed at the insulator site, edge computing nodes, cloud servers, and communication transmission equipment. The software layer encompasses a data acquisition module, a data preprocessing module, an anomaly analysis module, an electric field simulation analysis module, a resource scheduling module, and a result display module. The system can realize the full-process processing of insulator-related data. It collects insulator operating data in real time from the site, performs data anomaly analysis after transmission and preprocessing, and then initiates corresponding electric field analysis tasks based on the analysis results. Simultaneously, it intelligently schedules and allocates computing, storage, and network resources on the edge and cloud sides. Through professional electric field simulation algorithms, it completes core analyses such as insulator electric field distribution and electric field intensity changes. Finally, it visualizes the analysis results and insulator condition assessment information. The core service of this system is the condition monitoring of insulators in power transmission lines. It can accurately determine changes in the insulation performance of insulators through electric field analysis, promptly detect insulation degradation and potential faults, and provide scientific and technical support for the operation and maintenance of insulators.

[0024] First, real-time operational data of the insulators must be acquired. Acquiring this data is the fundamental prerequisite for the entire insulator electric field analysis system. Its core is to collect operational status parameters of the insulators in actual working scenarios through various monitoring devices, providing reliable data support for subsequent electric field analysis tasks. Insulators are insulating components in power systems used to fix conductors and isolate live conductors from grounded conductors. Common types include porcelain insulators and composite insulators. Real-time operational data refers to various status indicators of the insulator at the current operating moment, mainly including leakage current, surface temperature, partial discharge, ambient temperature and humidity, and electric field strength. Specific implementation methods need to be combined with the on-site deployment of the power system, employing a fusion of multiple monitoring technologies. Firstly, various sensors are deployed at the insulator installation location. Leakage current can be collected using Rogowski coil sensors and leakage current monitoring modules; surface temperature can be captured in real time using infrared thermometers; and partial discharge can be measured using ultra-high frequency (UHF) sensors. The system uses a combination of frequency (UHF) and ultrasonic sensors for detection. Ambient temperature and humidity are collected by temperature and humidity sensors, while electric field strength is obtained by electric field sensors. These sensors transmit the collected analog signals to an edge computing gateway. The edge computing gateway filters, amplifies, and performs A / D conversion on the analog signals, converting them into digital signals. These digital signals are then transmitted to the data acquisition module of the electric field analysis system via industrial Ethernet and 5G communication technology. The data acquisition module performs noise reduction and synchronous calibration on the digital signals to ensure the accuracy and timeliness of the data. Simultaneously, the system periodically verifies the integrity of the collected data, supplementing and correcting missing or abnormal data in real time. This ensures the continuous and stable acquisition of real-time insulator operating data. The advantage of this approach is that it allows for timely monitoring of the insulator's actual operating status, providing data support for subsequent anomaly analysis and electric field analysis tasks, ensuring the accuracy and reliability of the electric field analysis results, and ultimately guaranteeing the safe operation of the insulator.

[0025] Preferably, based on the anomaly analysis results of real-time operational data, each electric field analysis task and its priority are determined. Based on the anomaly analysis results, each electric field analysis task is identified; each task is analyzed to determine influencing factors; a task priority quantification model is constructed based on each task and its corresponding influencing factors; and the task priority of each electric field analysis task is obtained according to the task priority quantification model.

[0026] Based on the anomaly analysis results of the insulator's real-time operating data, the various electric field analysis tasks and their priorities are determined. The core is to first identify the key links in the insulator's electric field that may have anomalies through anomaly analysis, then clarify the targeted analysis direction and classify the tasks according to their importance, ensuring efficient resource allocation and accurate analysis. Anomaly analysis results refer to the judgment of abnormal states obtained after processing the insulator's real-time operating data, including specific situations such as excessive leakage current, abnormal surface temperature, and abnormal partial discharge. Electric field analysis tasks refer to the electric field-related simulation calculations and analyses carried out on the abnormal parts of the insulator, specifically including simulation of the electric field distribution in the abnormal area, calculation of the peak electric field intensity, and analysis of the degree of electric field distortion. Task priority refers to the execution order level divided according to the urgency and importance of the electric field analysis tasks. Influence factors refer to the key indicators that determine the priority of tasks, mainly including the severity of the anomaly, the timeliness requirements for task processing, and the computational complexity of the task.

[0027] First, based on the anomaly analysis results, various electric field analysis tasks are determined. For different anomaly types, corresponding analysis directions are clarified. If the leakage current exceeds the standard, a simulation task of the electric field distribution in the corresponding leakage current region is determined. If the surface temperature is abnormal, a task of calculating the electric field intensity in the temperature anomaly region is determined. If partial discharge is abnormal, a task of analyzing the electric field distortion at the partial discharge point is determined. Next, each electric field analysis task is comprehensively analyzed. The influencing factors corresponding to each task are extracted and quantified. The severity of the anomaly is divided into three levels—slight, moderate, and severe—based on the magnitude of exceeding the standard, and assigned values ​​of 1, 2, or 3 respectively. The timeliness requirements of task processing are divided into urgent, general, and non-urgent levels, and assigned values ​​of 1, 2, or 3 respectively. The computational complexity of the task is divided into high, medium, and low levels based on the required computational resources and time, and assigned values ​​of 1, 2, or 3 respectively. Finally, based on each electric field analysis task and the corresponding quantified influencing factors, a task priority quantification is constructed. The model is constructed using a weighted summation method. The model calculation formula is: Task Priority Score = Anomaly Severity Score × 0.5 + Task Processing Timeliness Requirement Score × 0.3 + Task Computational Complexity Score × 0.2. Finally, the quantitative scores of each influencing factor for each electric field analysis task are substituted into the model to calculate the priority score for each task. The tasks are ranked from highest to lowest score to determine the task priority of each electric field analysis task, with the highest score being the highest priority, and so on. The advantage of this approach is that it ensures that critical and urgent electric field analysis tasks are executed first, improving the efficiency and relevance of electric field analysis, and guaranteeing the timeliness and accuracy of insulator condition assessment. The task priority quantification model refers to a mathematical model that calculates task priority scores to determine the task execution order by quantifying influencing factors and assigning corresponding weights. The weighted summation method is a commonly used quantification method that calculates a comprehensive score by assigning different weights to multiple quantitative indicators.

[0028] Furthermore, based on the simulation results of each electric field analysis task, the edge-cloud resource requirements for each task are determined. Cluster analysis is performed on the simulation results to obtain the resource requirement feature vector for each electric field analysis task. Using an edge-cloud collaborative resource mapping model, heterogeneous resource adaptation processing is applied to the resource requirement feature vector to obtain the edge-cloud resource requirements for each electric field analysis task.

[0029] Based on the simulation results of each electric field analysis task, the edge-cloud resource requirements of each task are determined. The core is to clarify the specific usage and specifications of edge and cloud resources for each task by analyzing the simulation results, so as to provide a precise basis for subsequent resource allocation. The simulation results refer to various output data obtained after the electric field analysis task is simulated, including electric field distribution simulation data, electric field intensity calculation results, and simulation error parameters. The edge-cloud resource requirements refer to the specific quantity and specifications of various edge and cloud resources required to complete the electric field analysis task, mainly including computing resources, storage resources, and communication resources. Cluster analysis is a data analysis method that classifies similar data. The resource requirement feature vector is a multi-dimensional vector formed by quantifying various resource requirements of the electric field analysis task. The edge-cloud collaborative resource mapping model is a mathematical model used to match the resource requirement characteristics of the task with heterogeneous edge and cloud resources.

[0030] First, cluster analysis is performed on the simulation results of each electric field analysis task using the K-means clustering algorithm. Core indicators related to resource requirements are extracted from the simulation results, including simulation computation time, data processing volume, data transmission rate, and storage usage. These indicators are standardized to eliminate the influence of units. Then, a reasonable number of clusters is determined, and through iterative calculation, the simulation results of different electric field analysis tasks are divided into different categories based on the similarity of resource requirements. The common characteristics of resource requirements for each category are then extracted and quantified into a resource requirement feature vector for each electric field analysis task. Each component in the vector corresponds to a specific quantified value of a resource requirement indicator. Next, a heterogeneous resource adaptation process is performed on the resource requirement feature vector using an edge-cloud collaborative resource mapping model. This model pre-inputs the performance parameters of edge-side and cloud-side resources. This includes the CPU computing power, memory capacity, and communication bandwidth of edge nodes, as well as the CPU computing power, memory capacity, storage capacity, and communication bandwidth of cloud servers. The model's built-in adaptation algorithm matches the resource requirement feature vector with the edge-cloud resource performance parameters, calculating the edge-side resource usage and cloud-side resource usage corresponding to each resource requirement. This ultimately yields the edge-cloud resource requirements for each electric field analysis task. The advantage of this approach is that it accurately matches the resource requirements of each task, avoiding resource waste and shortages, and improving the utilization efficiency of edge-cloud resources. K-means clustering is an unsupervised clustering algorithm based on distance metrics. It achieves accurate data classification by continuously iterating and updating cluster centers. Standardization is a preprocessing method that converts indicators of different dimensions into the same dimension for easier subsequent calculations.

[0031] Furthermore, the resource stress level of the electric field analysis system is determined. Current edge-cloud resource usage status data of the electric field analysis system is acquired; based on the edge-cloud resource usage status data, a resource occupancy vector is obtained; and the resource occupancy vector is processed by stress mapping to obtain the resource stress level.

[0032] Determining the resource scarcity level of the electric field analysis system is a crucial constraint for subsequent optimization of edge-cloud resource allocation strategies. The core is to quantitatively assess the degree of resource supply and demand matching by collecting and analyzing the current usage of edge-cloud resources in the system. Resource scarcity level refers to the degree of imbalance between the current availability of edge and cloud resources in the electric field analysis system and the total resource requirements of various electric field analysis tasks. Its levels can be divided into relaxed, moderately strained, and extremely strained. Edge-cloud resource usage status data refers to various data that reflect the current operating status of edge and cloud resources in real time. These mainly include the CPU utilization rate, memory utilization rate, storage utilization rate, and communication bandwidth utilization rate of edge nodes and cloud servers. The resource occupancy vector is a multi-dimensional vector formed by quantifying various edge-cloud resource usage status data, used to centrally represent the current resource occupancy status of the system.

[0033] First, the current edge-cloud resource usage status data of the electric field analysis system is acquired. The system's built-in resource monitoring module collects various resource usage data from edge nodes and cloud servers in real time. CPU utilization is collected using a process scheduling monitoring tool, memory utilization is read from the memory management module, storage utilization is obtained from a storage monitoring tool, and communication bandwidth utilization is collected using a network traffic monitoring tool. The collection frequency is set to once per second to ensure data real-time performance and continuity. After collection, various data types are denoised and standardized to eliminate the impact of data anomalies and dimensional differences. Then, based on the processed edge-cloud resource usage status data, a resource occupancy vector is obtained. Each component of the vector corresponds to a quantified value of resource usage status. For example, the quantified value of CPU utilization ranges from 0 to 1, where 0 represents completely idle and 1 represents fully occupied. The quantified values ​​of CPU utilization, memory utilization, storage utilization, and communication bandwidth utilization are then combined as vector components to form the resource occupancy vector. The resource occupancy vector is then mapped to a stress level using a fuzzy comprehensive evaluation method. This involves pre-establishing a mapping relationship between the resource occupancy vector components and resource stress levels, determining the weight of each component, and calculating a comprehensive quantitative value of resource stress through fuzzy transformation. This comprehensive quantitative value is then mapped to the corresponding resource stress level based on a preset threshold range, ultimately yielding the current resource stress level of the electric field analysis system. This approach allows for precise understanding of the system's resource operation status, providing reasonable constraints for subsequent resource allocation optimization and preventing resource imbalances that could lead to task execution delays or resource waste. The fuzzy comprehensive evaluation method is a comprehensive evaluation method based on fuzzy mathematics, effectively addressing the fuzziness issues in resource stress assessment. Noise denoising refers to removing outliers and interference data from the collected data to ensure data accuracy. Standardization converts resource usage data of different dimensions into the same dimension, facilitating subsequent vector construction and mapping calculations.

[0034] Next, based on the edge-cloud resource requirements and task priorities corresponding to each electric field analysis task, an initial edge-cloud resource allocation strategy is determined. A resource request list is obtained based on the edge-cloud resource requirements and task priorities; a resource state space is constructed according to the task priorities; a latency calculation model is established based on the resource request list and resource state space to obtain the task latency of each electric field analysis task; a multi-objective priority reward function is constructed based on the task latency; and the resource request list is matched item by item using the multi-objective priority reward function to obtain the initial edge-cloud resource allocation strategy.

[0035] The initial edge-cloud resource allocation strategy is determined based on the edge-cloud resource requirements and task priorities corresponding to each electric field analysis task. The core is to combine the resource usage requirements and execution priorities of each task to initially formulate an allocation scheme for edge and cloud resources, providing a basic framework for subsequent optimization. The initial edge-cloud resource allocation strategy refers to the allocation scheme for edge and cloud resources based solely on task resource requirements and priorities, without considering system resource constraints. It clarifies the edge resource usage, cloud resource usage, and resource allocation sequence for each task. The resource request list is a unified list of edge-cloud resource requirements and task priorities for each electric field analysis task. The formatted list includes task identifiers, edge cloud resource requirements details, and priority levels. The resource state space is a spatial model built based on task priorities to describe the possibilities of system resource allocation, covering the resource allocation range and constraints corresponding to tasks with different priorities. The delay calculation model is a mathematical model used to calculate the execution delay of each electric field analysis task under different resource allocation scenarios. Task delay refers to the total time taken for an electric field analysis task from receiving resource allocation to completing simulation processing. The multi-objective priority reward function is an evaluation function built by comprehensively considering task delay and task priority, used to measure the rationality of the resource allocation scheme.

[0036] First, based on the edge-cloud resource requirements and task priorities of each electric field analysis task, a resource request list is compiled, sorted from high to low priority. Tasks of the same priority are then sorted from smallest to largest total resource requirement. The list clearly indicates the specific usage of computing, storage, and communication resources required for each task, along with its corresponding priority score. Next, a resource state space is constructed based on task priorities, dividing resource allocation priority intervals for different priority tasks. Higher priority tasks correspond to a wider resource allocation space and a better resource allocation sequence, while the resource allocation space for lower priority tasks is constrained by higher priority tasks. The boundary conditions of the resource state space are also defined, namely the maximum allocatable amount of each type of resource. Then, a latency calculation model is established based on the resource request list and the resource state space. This model is constructed using queuing theory combined with a linear regression algorithm, taking the resource requirements in the resource request list as input and the resource allocation constraints in the resource state space as boundary conditions. The model calculates the task latency of each electric field analysis task under the current resource allocation scenario, including resource scheduling latency and task execution latency. Then, based on the calculated task latency, a multi-objective priority reward function is constructed. The function aims to minimize task latency and maximize the reward of high-priority tasks, assigning higher reward weights to high-priority tasks and lower reward weights to low-priority tasks. The reward values ​​for different resource allocation schemes can be obtained through function calculation. Finally, based on the multi-objective priority reward function, the resource request list is matched item by item. Starting with the highest priority task, the optimal edge and cloud resources are matched in the resource state space based on its edge and cloud resource requirements. After allocation, the resource state space is updated, and subsequent priority tasks are processed in turn until all tasks have completed resource matching, finally obtaining the initial edge and cloud resource allocation strategy. The advantage of this approach is that it can initially achieve accurate matching between resources and tasks, ensure the resource requirements of high-priority tasks, and provide a reasonable basis for subsequent optimization. Queuing theory is a mathematical theory used to study queuing phenomena and resource allocation efficiency in systems, while linear regression algorithm is a statistical analysis algorithm used to establish linear relationships between variables, which can accurately fit the correlation between task resource requirements and task latency.

[0037] Furthermore, using resource scarcity as a constraint, a multi-objective optimization algorithm is employed to optimize the initial edge-cloud resource allocation strategy, resulting in an edge-cloud objective allocation strategy.

[0038] By using a multi-objective optimization algorithm to optimize the initial edge-cloud resource allocation strategy under the constraint of resource scarcity, a target edge-cloud allocation strategy is obtained. The core is to optimize the resource allocation scheme to better meet actual operational needs while ensuring that system resources do not exceed the carrying capacity. The target edge-cloud allocation strategy refers to the final edge-cloud resource allocation scheme after optimization by the multi-objective optimization algorithm, which takes into account resource utilization efficiency, task execution efficiency, and resource scarcity constraints. It can be directly used for subsequent resource allocation execution. The multi-objective optimization algorithm is an algorithm that can simultaneously optimize multiple mutually constraining objective functions. The non-dominated sorting genetic algorithm is selected, which can efficiently handle multi-objective optimization problems and obtain the optimal solution set.

[0039] First, the specific requirements of resource scarcity as a constraint are clarified. Based on the previously obtained resource scarcity levels, corresponding constraint thresholds are set. If the resource scarcity level is severe or extremely severe, the total resource consumption of each task is strictly limited to not exceeding the total available resources of the system. If it is moderate or lenient, the constraints are appropriately relaxed, but resource consumption is still ensured to remain within a reasonable range. Next, the core objective function of the multi-objective optimization is determined, mainly including three mutually constraining objectives: minimizing total task delay, maximizing resource utilization, and maximizing the execution satisfaction of high-priority tasks. The total task delay is the sum of the delays of all electric field analysis tasks. Resource utilization is the ratio of the actual allocated resources to the total available resources of the system. The execution satisfaction of high-priority tasks is related to the resource allocation sufficiency rate and execution delay of high-priority tasks. Then, the relevant parameters of the algorithm are initialized, setting the population size to 100, the number of iterations to 50, the crossover probability to 0.8, and the mutation probability to 0.05. The initial edge cloud resource allocation strategy is used as the initial population individuals, with each population individual corresponding to one resource allocation scheme. Then, the algorithm iterative calculation begins. In each iteration, the population individuals are first sorted using a non-dominated method. The algorithm distinguishes the optimization levels of different individuals and calculates the crowding distance for each individual to ensure the diversity of the solution set. Then, it generates new individuals through crossover and mutation operations. The crossover operation uses a single-point crossover method, exchanging some resource allocation parameters between two individuals. The mutation operation randomly adjusts the edge or cloud-side resource allocation for some tasks within the individuals, while simultaneously checking whether the newly generated individuals meet resource stress constraints and removing those that do not. The optimal individual is continuously retained during the iteration process until a preset number of iterations is reached. Finally, from the optimal solution set obtained through iteration, the individual with the best overall performance is selected based on actual business needs. The resource allocation scheme corresponding to this individual is the edge-cloud target allocation strategy. The advantage of this approach is that it achieves optimal resource allocation within the system's resource capacity, avoiding resource waste and excessive task latency while ensuring the smooth execution of high-priority tasks. Non-dominated sorting is the core step of the algorithm, used to divide the individuals into different levels according to their optimization level. Crowding distance is used to measure the dispersion of individuals in the solution set, ensuring the diversity of the optimal solution set. Crossover and mutation operations are used to generate new resource allocation schemes, improving the algorithm's search capability.

[0040] Finally, based on the edge-cloud target allocation strategy, corresponding edge-side resources and cloud-side resources are allocated to each electric field analysis task.

[0041] The allocation of corresponding edge-side and cloud-side resources to each electric field analysis task based on the edge-cloud target allocation strategy is the core execution step for the electric field analysis system to implement resource allocation and ensure the smooth operation of various tasks. Edge-side resources are the hardware and software resources deployed at the edge nodes of the insulator monitoring site, mainly including the CPU computing power, memory capacity, local storage resources of the edge server, and the bandwidth resources of the edge communication module. Cloud-side resources are the data center resources deployed in the remote cloud, mainly including the CPU computing power, memory capacity, massive storage resources of the cloud server, and cloud communication bandwidth resources.

[0042] First, the edge-cloud target allocation strategy is analyzed, extracting the edge-side resource allocation details and cloud-side resource allocation details corresponding to each electric field analysis task. This clarifies the required number of edge-side CPU cores, memory size, local storage capacity, and communication bandwidth for each task, as well as the number of cloud-side CPU cores, memory size, storage capacity, and communication bandwidth. Simultaneously, the current available resource status of edge nodes and cloud servers is reviewed to ensure that the allocated resources do not exceed the actual available range. Next, the system resource scheduling module is activated. This module uses remote procedure call technology and edge node management protocols to establish communication connections with each edge node and cloud server, enabling real-time transmission and execution feedback of resource allocation commands. Then, according to task priority from high to low, the corresponding edge-side and cloud-side resources are allocated to each electric field analysis task sequentially. During the allocation process, the target resources of the corresponding edge nodes and cloud servers are locked first to avoid resource conflicts. Then, the resource scheduling module sends resource allocation commands to the edge nodes and cloud servers. After receiving the commands, the edge nodes and cloud servers utilize resource virtualization... The technology allocates resources of corresponding specifications and binds them to the target electric field analysis task. Simultaneously, it updates the system resource status database, recording the allocation status and available reserves of each resource in real time. After allocation, the resource scheduling module verifies the resource allocation results for each task, checking whether the allocated resource specifications and usage are consistent with the edge-cloud target allocation strategy, and whether resource binding is successful. If an allocation anomaly occurs, a re-allocation process is immediately initiated until all tasks have correctly allocated edge-side and cloud-side resources. This approach ensures accurate implementation of the edge-cloud target allocation strategy, guarantees sufficient and suitable resources for each electric field analysis task, ensures efficient and stable task execution, and improves overall system operating efficiency and resource utilization. The resource scheduling module is the core module responsible for coordinating edge-side and cloud-side resources, executing resource allocation instructions, and verifying allocation results. Resource virtualization technology abstracts physical resources into virtual resources, enabling flexible resource allocation and efficient utilization. Remote procedure call technology is used to achieve remote communication between different nodes and transmit resource allocation instructions.

[0043] This invention provides a real-time analysis method for the electric field of insulators based on an edge-cloud collaborative architecture for intelligent operation and maintenance of power equipment. The overall process includes the following core steps: First, the complete insulator electric field analysis task is decomposed into several sub-tasks with clear functional boundaries and execution logic. Secondly, a task priority quantification model is constructed based on the scenario-specific attributes of each sub-task; Furthermore, based on this task priority quantification model, predictive unloading decisions are generated for each subtask. Subsequently, predictive offloading decisions drive the dynamic allocation of edge cloud resources; Finally, after resource allocation is completed, the edge server and cloud server work together to execute the insulator electric field analysis subtask, achieving end-to-end real-time electric field analysis.

[0044] The insulator electric field analysis subtask specifically includes: data preprocessing, geometric modeling, electric field numerical calculation, anomaly detection, and visualization.

[0045] In particular, the geometric modeling task involves constructing three-dimensional geometric models of insulators under three typical operating conditions: normal, polluted, and iced, in order to support subsequent high-fidelity electric field simulations.

[0046] Furthermore, constructing a task priority quantification model based on the scenario-specific attributes of the insulator electric field analysis subtask includes the following operations: Based on the scenario-specific attributes of the insulator electric field analysis subtask and their associated digital twin states of the insulator, a dynamically adjustable scenario dynamic weight factor is generated. By combining the basic attributes of subtasks (such as computational complexity and input data size) with dynamic weight factors of the scenario, a task priority quantization model that integrates multi-dimensional semantics is constructed.

[0047] Accordingly, the predictive offloading decision for the insulator electric field analysis subtask, set according to the task priority quantification model, specifically includes: Calculate the task priority coefficient of each subtask using a task priority quantification model; Define the corresponding decision action space for each subtask; Based on the decision action space and a four-dimensional decision state space, a task delay calculation model is established to estimate the end-to-end task delay under different unloading options. Taking into account the performance requirements of subtasks in terms of accuracy, energy consumption, and task latency, a multi-objective priority reward function is defined. By employing reinforcement learning algorithms and combining them with a multi-objective priority reward function, predictive unloading decisions for future states are achieved.

[0048] Edge-cloud resource allocation for insulator electric field analysis tasks through predictive offloading decisions includes: Based on the predictive offloading decision results, specific insulator electric field analysis sub-tasks to be performed on the edge side and the cloud side are specified respectively; After execution, the edge computing capability assessment value is dynamically updated by utilizing the actual computing performance and task feedback on the edge side, and the scene dynamic weight factor is adjusted in parallel to form a closed-loop optimization mechanism.

[0049] Among them, the scenario-specific attributes include four key dimensions: real-time sensitivity, accuracy sensitivity, data dependency, and task importance, which are used to characterize the differentiated needs of different subtasks in specific operation and maintenance scenarios.

[0050] The scene dynamic weighting factor is output in real time by the digital twin model of the insulator, and can be adaptively adjusted according to the equipment status, environmental conditions and historical operating data.

[0051] The four-dimensional decision state space specifically includes the following five elements (Note: Although it is called "four-dimensional", it actually includes five coupled states, reflecting the abstraction of high-dimensional states): real-time computing power load of edge servers, real-time computing power load of cloud servers, network transmission bandwidth, digital twin state of insulators, and historical anomaly frequency in the past hour, which together constitute the context-aware basis for offloading decisions.

[0052] The decision action space defines three optional offloading actions for each insulator electric field analysis subtask. For example, for the data preprocessing task, the offloading actions include: execution at the edge, execution in the cloud, or execution using an edge-cloud collaborative approach. Other subtasks follow the same pattern, supporting fine-grained resource scheduling.

[0053] For example, this application takes scenario adaptation, proactive decision prediction, edge-cloud collaboration, and closed-loop iteration as its core design concepts. Through four interconnected and mutually reinforcing technical links—task decoupling and attribute modeling, priority quantification, predictive offloading decision, and dynamic resource allocation and execution—it achieves a leap from a general edge-cloud task offloading mechanism to a dedicated intelligent analysis system for insulator electric fields.

[0054] As the first step, namely "Insulator Electric Field Analysis Task Decomposition and Sub-task Attribute Definition," this method systematically decomposes the complete analysis process into the aforementioned five core sub-tasks, and clearly defines the scenario-specific attributes for each sub-task, as shown in Table 1. This step provides a structured input and semantic foundation for subsequent priority modeling and intelligent offloading.

[0055] Table 1. Attributes of the Insulator Electric Field Analysis Subtask After structural decoupling of the insulator electric field analysis task, the system enters the second stage: constructing a task priority quantification mechanism that integrates static attributes and dynamic environmental perception. This mechanism no longer relies on fixed rules but generates a dynamic scene weighting factor that can adaptively adjust according to operating conditions by introducing the real-time physical state output from the insulator digital twin model. Specifically, the system comprehensively considers three key environmental variables: pollution level, operating voltage level, and ambient temperature, assigning them different influence weights, with operating voltage having the highest weight, followed by pollution level, and then ambient temperature. Each variable is mapped to a corresponding coefficient value according to a preset engineering threshold: for example, when the salt density on the insulator surface is below 0.03 mg / cm², the pollution influence coefficient takes a baseline value of 1.0; once the salt density exceeds 0.1 mg / cm², the coefficient increases to 1.8, significantly enhancing the urgency of task scheduling. Similarly, under high-risk conditions such as low-temperature icing risk (ambient temperature below -10℃) or overvoltage (reaching 1.2 times the rated voltage), the corresponding coefficients are also adjusted upwards. Ultimately, this dynamic weighting factor is weighted and integrated with the four inherent basic attributes of the task—sensitivity to response timeliness, the task's importance in the overall analysis process, the degree of dependence on the integrity of the original data, and the requirement for the accuracy of the calculation results—to form the final priority coefficient for each subtask. This design enables the task scheduling strategy to accurately respond to the full spectrum of device state changes from "normal operation" to "high-risk alarm," achieving true scenario-driven resource pre-allocation.

[0056] Building upon this foundation, the system further constructs a predictive offloading decision engine, upgrading the traditional passive task distribution to proactive intelligent scheduling. This engine, centered on reinforcement learning, constructs a high-dimensional decision state space encompassing five dimensions: not only covering traditional resource indicators such as real-time computing load at the edge and cloud, and current network bandwidth, but also innovatively incorporating the comprehensive risk level of devices output by a digital twin model and the frequency of historical anomalies within the past hour. This multi-source state fusion enables the decision system to proactively perceive potential faults. For the five types of electric field analysis subtasks, the system defines three feasible execution modes for each type: pure edge execution, pure cloud execution, or edge-cloud collaborative execution, resulting in fifteen action options. For each action, the system accurately estimates its end-to-end execution latency, which is composed of local computation time and cross-node data transmission time. Furthermore, it combines the task's performance in terms of accuracy and edge energy consumption to construct a multi-objective optimization-oriented reward function. Crucially, the optimization focus of this reward function can be dynamically adjusted according to the system's risk level: when equipment is in a high-risk state, the system automatically suppresses energy consumption optimization objectives and instead focuses on ensuring the timeliness and accuracy of analysis results; while under normal operating conditions, it maintains a balanced optimization among the three. By solving this complex decision-making problem using a deep Q-network algorithm, the system ultimately learns an offloading strategy that maximizes long-term cumulative benefits, achieving a paradigm shift from "execution only when resources are available" to "predictive demand and advance scheduling."

[0057] The task execution phase employs a closed-loop collaborative architecture, fully leveraging the complementary advantages of low-latency edge sensing and high-precision cloud computing. Edge nodes first utilize lightweight models to perform data cleaning, feature extraction, and initial anomaly screening, uploading only highly compressed key features (such as electric field gradient abrupt changes and partial discharge preliminary judgment tags) to the cloud. This reduces data transmission volume by approximately 80% compared to the original data acquisition, effectively alleviating bandwidth pressure. The cloud, based on a high-fidelity finite element model, performs refined electric field simulation and cross-device correlation analysis on feature data from single or multiple insulators, generating a structured diagnostic report containing anomaly location, cause mechanism, and risk level. More importantly, this process simultaneously produces a lightweight detection model optimized with global data, which is distributed to edge nodes via a secure channel, enabling online evolution of the local model. Experimental results show that this mechanism can improve the edge anomaly identification accuracy from the initial 80% to over 95%. Meanwhile, the real-time operational data continuously transmitted from the edge (such as newly measured pollution levels, temperature, humidity, and electric field strength) drives the digital twin model to update, thereby correcting the dynamic weight factors required for the next round of task scheduling, forming a complete closed loop of "perception—decision—execution—feedback—evolution." This two-way empowerment mechanism not only ensures the high efficiency and accuracy of a single analysis but also achieves a continuous improvement in the overall intelligence level of the system.

[0058] To verify the adaptability and superiority of this method in complex power scenarios, the system underwent empirical tests under three typical operating conditions. In a high-risk icing scenario, facing -15℃ low temperatures and frequent abnormal alarms, the system automatically anchored the anomaly detection task to a millisecond-level response at the edge, while simultaneously scheduling cloud resources for high-precision electric field reconstruction, ultimately achieving an early warning within 0.1 seconds, an 80% speedup compared to traditional solutions, and accurately locating electric field distortions caused by 10mm-level icing. In a highly polluted coastal, battery-powered, energy-sensitive scenario, the system, through an edge-cloud collaborative computing mode, reduced edge power consumption by 33% while maintaining an analysis accuracy of ±0.02kV / m, significantly extending the continuous operating time of power stations without mains power. In a 220kV substation cluster batch analysis scenario, the system, through a collaborative strategy of edge parallel preprocessing and cloud centralized precision calculation, reduced the time for full-station electric field analysis of 20 sets of insulators to 0.35 seconds, improving efficiency by 77%, and successfully identified early-aging insulator units. These examples fully demonstrate that this method can not only flexibly adapt to diverse needs ranging from single-point high-risk early warning to station-level batch operation and maintenance, but also achieve a dynamic optimal balance between timeliness, accuracy and energy efficiency, providing reliable technical support for smart grid state perception.

[0059] Another embodiment of the present invention provides an edge-cloud resource allocation system. For details, please refer to [link to relevant documentation]. Figure 2 , Figure 2 The diagram shown illustrates the structure of an edge-cloud resource allocation system according to one embodiment of the present invention, comprising: Module 11 is used to acquire real-time operating data of the insulator; The determination module 12 is used to determine each electric field analysis task and the task priority of each electric field analysis task based on the abnormal analysis results of the real-time running data. Module 13 is used to determine the edge-cloud resource requirements of each electric field analysis task based on the simulation processing results of each electric field analysis task. The resource stress module 14 is used to determine the resource stress level of the electric field analysis system. Initial module 15 is used to determine the initial edge cloud resource allocation strategy based on the edge cloud resource requirements and task priority corresponding to each electric field analysis task. Target module 16 is used to optimize the initial edge cloud resource allocation strategy using a multi-objective optimization algorithm under the constraint of resource scarcity, so as to obtain the edge cloud target allocation strategy. The allocation module 17 is used to allocate corresponding edge-side resources and cloud-side resources to each electric field analysis task based on the edge-cloud target allocation strategy.

[0060] Preferably, the determining module 12 includes: The anomaly analysis unit is used to determine each electric field analysis task based on the anomaly analysis results; The influence factor unit is used to analyze various electric field analysis tasks and determine the influence factors; Priority quantization model unit, used to construct task priority quantization model based on each electric field analysis task and corresponding influencing factors; Priority units are used to obtain the task priority of each electric field analysis task based on the task priority quantization model.

[0061] Preferably, requirement module 13 includes: Clustering units are used to perform cluster analysis on the simulation results to obtain the resource requirement feature vector for each electric field analysis task; The resource requirement unit is used to perform heterogeneous resource adaptation processing on the resource requirement feature vector using the edge-cloud collaborative resource mapping model to obtain the edge-cloud resource requirements for each electric field analysis task.

[0062] Preferably, the tension level module 14 includes: The status data unit is used to acquire the current edge cloud resource usage status data of the electric field analysis system. Occupancy vector unit, used to obtain resource occupancy vector based on edge cloud resource usage status data; The mapping unit is used to perform stress mapping on the resource occupancy vector to obtain the resource stress level.

[0063] Preferably, the initial module 15 includes: The resource request unit is used to obtain a list of resource requests based on edge cloud resource requirements and task priorities; State space units are used to construct resource state spaces based on task priorities. The task delay unit is used to establish a delay calculation model based on the resource request list and resource state space to obtain the task delay of each electric field analysis task. The reward function unit is used to construct a multi-objective priority reward function based on task latency; The allocation strategy unit is used to perform item-by-item resource matching on the resource request list based on a multi-objective priority reward function to obtain the initial edge-cloud resource allocation strategy.

[0064] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0065] Accordingly, embodiments of the present invention provide a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform steps in the edge-cloud resource allocation method of the above embodiments, for example... Figure 1 Steps S1 to S7 as described above.

[0066] This invention first acquires real-time operating data of insulators and dynamically generates electric field analysis tasks and their priorities based on anomaly analysis results. This ensures that the system only triggers calculations for data with diagnostic value, avoiding ineffective full-scale processing. Then, based on the simulation results of each task, its edge-cloud resource requirements are inferred, mapping abstract tasks to schedulable resource requests. Simultaneously, the current resource scarcity of the electric field analysis system is determined in real time, serving as a global feedback signal reflecting the load status of the edge and cloud sides. Based on this, an initial allocation strategy is generated by combining task priorities and resource requirements. Using resource scarcity as a hard constraint, a multi-objective optimization algorithm iteratively solves the edge-cloud objective allocation strategy, ensuring that high-priority tasks are still guaranteed even when resources are limited, while low-priority or lightweight tasks are guided to idle edge nodes for execution. Finally, according to this optimized strategy, corresponding edge-side or cloud-side resources are allocated to each task, achieving on-demand distribution and load balancing of computing tasks between the edge and cloud, fundamentally improving the overall utilization efficiency of edge-cloud resources.

[0067] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for allocating edge-cloud resources, applied in an electric field analysis system for insulators, characterized in that, include: Obtain the real-time operating data of the insulator; Based on the anomaly analysis results of the real-time operating data, each electric field analysis task and the task priority of each electric field analysis task are determined. Based on the simulation results of each electric field analysis task, the edge-cloud resource requirements of each electric field analysis task are determined. Determine the resource scarcity level of the electric field analysis system; Based on the edge cloud resource requirements and task priority corresponding to each electric field analysis task, an initial edge cloud resource allocation strategy is determined. Using the resource scarcity level as a constraint, the initial edge cloud resource allocation strategy is optimized using a multi-objective optimization algorithm to obtain the edge cloud objective allocation strategy; Based on the edge-cloud target allocation strategy, corresponding edge-side resources and cloud-side resources are allocated to each of the electric field analysis tasks.

2. The edge-cloud resource allocation method as described in claim 1, characterized in that, The step of determining each electric field analysis task and its priority based on the anomaly analysis results of the real-time operating data includes: Based on the anomaly analysis results, each electric field analysis task is determined; Each of the electric field analysis tasks was analyzed to determine the influencing factors; A task priority quantification model is constructed based on each of the electric field analysis tasks and the corresponding influencing factors. The task priority of each electric field analysis task is obtained according to the task priority quantification model.

3. The edge-cloud resource allocation method as described in claim 1, characterized in that, The determination of the edge-cloud resource requirements for each electric field analysis task based on the simulation results of each electric field analysis task includes: Cluster analysis is performed on the simulation results to obtain the resource requirement feature vector for each electric field analysis task; By using the edge-cloud collaborative resource mapping model, the resource demand feature vector is subjected to heterogeneous resource adaptation processing to obtain the edge-cloud resource requirements of each of the electric field analysis tasks.

4. The edge-cloud resource allocation method as described in claim 1, characterized in that, Determining the resource scarcity level of the electric field analysis system includes: Acquire the current edge-cloud resource usage status data of the electric field analysis system; Based on the edge cloud resource usage status data, a resource occupancy vector is obtained; The resource occupancy vector is subjected to a stress mapping process to obtain the resource stress level.

5. The edge-cloud resource allocation method as described in claim 1, characterized in that, The determination of the initial edge-cloud resource allocation strategy based on the edge-cloud resource requirements and task priorities corresponding to each electric field analysis task includes: Based on the edge cloud resource requirements and the task priorities, a resource request list is obtained; Construct a resource state space based on the task priorities: A delay calculation model is established based on the resource request list and resource state space to obtain the task delay of each electric field analysis task. Construct a multi-objective priority reward function based on the task delay; The resource request list is matched item by item based on the multi-objective priority reward function to obtain the initial edge cloud resource allocation strategy.

6. An edge-cloud resource allocation system, characterized in that, include: The acquisition module is used to acquire the real-time operating data of the insulator; The determination module is used to determine each electric field analysis task and the task priority of each electric field analysis task based on the anomaly analysis results of the real-time running data. The requirement module is used to determine the edge-cloud resource requirements of each electric field analysis task based on the simulation processing results of each electric field analysis task. The resource stress level module is used to determine the resource stress level of the electric field analysis system; The initial module is used to determine the initial edge cloud resource allocation strategy based on the edge cloud resource requirements and task priority corresponding to each electric field analysis task. The target module is used to optimize the initial edge cloud resource allocation strategy using a multi-objective optimization algorithm, with the resource scarcity level as a constraint, to obtain the edge cloud target allocation strategy. The allocation module is used to allocate corresponding edge-side resources and cloud-side resources to each of the electric field analysis tasks based on the edge-cloud target allocation strategy.

7. The edge-cloud resource allocation system as described in claim 6, characterized in that, The determining module includes: An anomaly analysis unit is used to determine each electric field analysis task based on the anomaly analysis results; The influence factor unit is used to analyze each of the electric field analysis tasks and determine the influence factors; The priority quantization model unit is used to construct a task priority quantization model based on each of the electric field analysis tasks and the corresponding influencing factors. The priority unit is used to obtain the task priority of each electric field analysis task according to the task priority quantization model.

8. The edge-cloud resource allocation system as described in claim 6, characterized in that, The requirement module includes: Clustering unit, used to perform cluster analysis on the simulation processing results to obtain the resource requirement feature vector of each electric field analysis task; The resource requirement unit is used to perform heterogeneous resource adaptation processing on the resource requirement feature vector using the edge-cloud collaborative resource mapping model to obtain the edge-cloud resource requirements of each of the electric field analysis tasks.

9. The edge-cloud resource allocation system as described in claim 6, characterized in that, The tension level module includes: The status data unit is used to acquire the current edge cloud resource usage status data of the electric field analysis system. The occupancy vector unit is used to obtain a resource occupancy vector based on the edge cloud resource usage status data; The mapping unit is used to perform stress mapping processing on the resource occupancy vector to obtain the resource stress level.

10. The edge-cloud resource allocation system as described in claim 6, characterized in that, The initial module includes: The resource request unit is used to obtain a resource request list based on the edge cloud resource requirements and the task priority; State space unit, used to construct resource state space according to the task priority: The task delay unit is used to establish a delay calculation model based on the resource request list and the resource state space to obtain the task delay of each of the electric field analysis tasks. The reward function unit is used to construct a multi-objective priority reward function based on the task delay. The allocation strategy unit is used to perform item-by-item resource matching on the resource request list based on the multi-objective priority reward function to obtain the initial edge cloud resource allocation strategy.

Images