A method for intelligent scheduling of computing power of edge computing equipment for power transmission lines
By constructing a three-level scheduling architecture for edge computing devices on power transmission lines, the problems of limited computing resources and uneven scheduling were solved, enabling rapid response to critical tasks and efficient system operation, thereby improving the real-time performance and reliability of power transmission line monitoring.
Patent Information
- Application Number
- CN202611324506.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-28
- Publication Date
- 2026-11-17
AI Technical Summary
The computing power resources of edge computing equipment for power transmission lines are limited, making it difficult to cope with sudden computing demands. There is a lack of collaborative scheduling mechanisms among multiple edge devices, resulting in uneven utilization of computing power resources, a lack of differentiated processing of task priorities, and insufficient optimization of computing power scheduling under the constraints of equipment energy consumption.
A three-level scheduling architecture is constructed, consisting of task priority evaluation, multi-device collaborative task distribution, and dynamic allocation of computing power within devices. By acquiring device computing power resources and task attribute information, task priority is evaluated, a scheduling optimization model is constructed, an improved particle swarm optimization algorithm is used to solve the task allocation scheme, and rescheduling is triggered when device load and power change.
It achieves differentiated and rapid response to critical alarm tasks, global balanced utilization of computing resources, and dynamic adjustment of equipment energy consumption, thereby improving the robustness and self-healing ability of the system and ensuring the real-time performance and reliability of power transmission line monitoring tasks.
Smart Images

Figure CN123114757A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of power system computing power scheduling technology, specifically to a method for intelligent scheduling of computing power of edge computing equipment for transmission lines. Background Technology
[0002] Transmission lines are scattered across mountains and rural areas, facing various safety hazards such as external damage, fires, foreign objects, and unauthorized operations. Traditional manual inspection methods are not only costly in terms of manpower and resources, but also fail to provide accurate and timely information about the lines. With the development of edge computing technology, deploying edge computing devices on the sides of transmission line towers to achieve on-site analysis and real-time alarms of monitoring data has become an important technological direction for the intelligent operation and maintenance of transmission lines.
[0003] However, edge computing devices for power transmission lines face several prominent problems in actual deployment and operation: limited computing resources make it difficult to cope with sudden computing demands; a lack of collaborative scheduling mechanisms among multiple edge devices leads to uneven utilization of computing resources; a lack of differentiated task priority processing means that critical alarm tasks may be blocked by non-critical tasks; and insufficient optimization of computing power scheduling under energy consumption constraints. To address these issues, this application proposes an intelligent computing power scheduling method for edge computing devices in power transmission lines. This method constructs a three-level scheduling architecture—task priority evaluation, multi-device collaborative task distribution, and dynamic allocation of computing power within devices—to achieve efficient task processing under limited computing resources. Summary of the Invention
[0004] To address the aforementioned technical problems, this application provides a method for intelligent scheduling of computing power for edge computing devices in power transmission lines, thereby resolving the existing issues.
[0005] The intelligent scheduling method for edge computing equipment of transmission lines proposed in this application adopts the following technical solution: One embodiment of this application provides a method for intelligent scheduling of computing power of edge computing equipment for power transmission lines, including the following steps: Step S1: Obtain the computing power resource information of each edge computing device in the online monitoring system of transmission lines, as well as the attribute information of the current computing tasks to be processed by each device. The computing power resource information includes available CPU computing power, available memory capacity and current remaining power. The task attribute information includes task type, data size and expected processing time limit. Step S2: Priority evaluation is performed on each computation task to be processed based on the task attribute information. The priority evaluation is calculated based on two dimensions: task urgency and task importance, to obtain a priority score for each task. Step S3: Based on the computing power resource information of each edge computing device and the priority score of each task, construct a scheduling optimization model with the goal of minimizing the processing latency of high-priority tasks and maximizing the system computing power utilization rate, and solve it to obtain the task allocation scheme. Step S4: Each edge computing device receives the assigned computing tasks according to the task allocation scheme, performs dynamic adjustment of computing power based on the current remaining power, determines the current available computing power limit, and executes the received computing tasks under the constraint of the computing power limit. Step S5: During the scheduling process, when it is detected that the computing power load rate of any device exceeds the preset load threshold or the remaining power is lower than the preset power threshold, the rescheduling process is triggered, and steps S2 to S4 are re-executed.
[0006] Preferably, the computing power resource information of the edge computing device in step S1 specifically includes: the attribute information of each computing task to be processed specifically includes: the task type is divided into wildfire identification, external damage detection, ice accumulation monitoring and routine inspection, and the expected processing time limit is the time interval from the time the task arrives to the time required to complete.
[0007] Preferably, the calculation method for the urgency of the task in step S2 is as follows: obtain the time difference between the current time and the expected processing time limit of the task, map the time difference to an urgency score, and the mapping adopts a piecewise linear function. When the time difference is less than or equal to the first time limit threshold, the urgency score is the highest value. When the time difference is greater than the first time limit threshold and less than or equal to the second time limit threshold, the urgency score decreases linearly as the time difference increases. When the time difference is greater than the second time limit threshold, the urgency score is the lowest value. The first time limit threshold is less than the second time limit threshold.
[0008] Preferably, the method for calculating the importance of a task in step S2 is as follows: a baseline value for importance is preset based on the task type, and the baseline value for the task type is dynamically adjusted based on the triggering frequency of each type of task within a preset duration. Specifically, when the task triggering frequency is higher than a preset frequency threshold, the baseline value for importance is reduced by a preset value to obtain an importance score; when the task triggering frequency is lower than the preset frequency threshold, the baseline value for importance is increased by a preset value to obtain an importance score; otherwise, the baseline value for importance of the task type is used as its importance score.
[0009] Preferably, the priority score in step S2 is calculated as follows: the task urgency score and the task importance score are normalized respectively, and the weighted sum of the normalized urgency score and importance score is taken as the task priority score.
[0010] Preferably, the scheduling optimization model in step S3 is constructed as follows: minimizing the total processing latency of all pending tasks is the first optimization objective, minimizing the variance of the computing power load rate of each edge device is the second optimization objective, and the first and second optimization objectives are weighted and merged into a comprehensive optimization objective; the constraints of the scheduling optimization model include: each task can only be assigned to one edge device for processing, the total computing power required by tasks assigned to any device does not exceed the upper limit of the available CPU computing power of that device, and the total memory required by tasks assigned to any device does not exceed the upper limit of the available memory capacity of that device.
[0011] Preferably, the specific method for solving the scheduling optimization model in step S3 is as follows: an improved particle swarm optimization algorithm is used to solve the problem, the task allocation scheme is encoded as the position vector of the particle, the comprehensive optimization objective is used as the fitness function, and the task allocation scheme corresponding to the optimal particle is output after iteration. In this process, a linear decreasing strategy of inertia weight is introduced during the particle swarm iteration.
[0012] Preferably, the specific method for dynamically adjusting computing power based on the current remaining power in step S4 is as follows: obtain the current remaining power percentage of the device; when the remaining power is higher than a first power threshold, the upper limit of computing power is the maximum computing power of the device; when the remaining power is lower than the first power threshold but higher than a second power threshold, the upper limit of computing power decreases linearly according to the remaining power percentage; when the remaining power is lower than the second power threshold, the upper limit of computing power is reduced to a preset percentage of the maximum computing power of the device, wherein the first power threshold is greater than the second power threshold.
[0013] Preferably, in step S4, when executing the received computing task, the execution order of each task in the internal queue of the device is further determined according to the task priority score: the task with the higher priority score is placed at the front of the queue and executed first, and the task with the lower priority score is placed at the back of the queue and executed later; when the computing power limit of the device is exceeded, the task with the lowest priority in the queue is marked as a task to be migrated and the unloading operation is triggered.
[0014] Preferably, the load threshold for triggering the rescheduling process in step S5 is when the device CPU computing power load rate exceeds the load rate threshold or the memory occupancy rate exceeds the occupancy rate threshold.
[0015] This application has at least the following beneficial effects: (i) A task priority evaluation mechanism was constructed to achieve differentiated and rapid response for critical alarm tasks. This application comprehensively evaluates priorities from two dimensions: urgency and importance. Urgency focuses on the timeliness of the task deadline, while importance focuses on the security impact weight of the task type. Through weighted synthesis of these two dimensions, it ensures that high real-time tasks such as wildfire identification and external damage detection receive priority access to computing resources, effectively avoiding the problem of critical tasks being blocked by non-critical tasks under the existing first-come, first-served strategy.
[0016] (ii) A multi-edge device collaborative scheduling optimization model was established to achieve global balanced utilization of computing resources. This application takes minimizing the total processing latency and minimizing the device load rate variance as the comprehensive optimization objectives, and adopts an improved particle swarm optimization algorithm to solve the optimal task allocation scheme, so that the surplus computing power of devices with sufficient computing power can be utilized and the load of devices with limited computing power can be distributed, thus solving the problem of uneven computing power utilization caused by the independent operation of existing devices.
[0017] (iii) A dynamic computing power adjustment mechanism based on equipment energy consumption perception is introduced. This application dynamically adjusts the upper limit of computing power based on the current remaining power of the equipment, giving full play to the performance of the equipment when the power is sufficient, and actively reducing the frequency to maintain power when the power is insufficient. This avoids the solar power supply equipment from shutting down due to excessive discharge during continuous cloudy and rainy weather, and ensures the continuous operation capability of the system under adverse weather conditions.
[0018] (iv) A rescheduling triggering and execution mechanism based on status monitoring was designed. This application monitors the device load rate and remaining power in real time. When any device is overloaded or has low power, rescheduling is automatically triggered, migrating the affected tasks to other healthy devices for execution. This dynamic rescheduling mechanism enables the system to adaptively cope with sudden task impacts and power fluctuations, improving the system's robustness and self-healing ability. Attached Figure Description
[0019] Figure 1 A flowchart illustrating the steps of an intelligent scheduling method for edge computing equipment in power transmission lines provided in this application; Figure 2 A schematic diagram illustrating the specific priority scoring process for each task provided in this application. Detailed Implementation
[0020] The following description, in conjunction with the accompanying drawings, details a specific scheme for an intelligent scheduling method for edge computing equipment of transmission lines provided in this application.
[0021] This application provides an embodiment of a method for intelligent scheduling of computing power for edge computing devices in power transmission lines. For details, please refer to [link to specific documentation]. Figure 1 This includes the following steps: Preferably, this embodiment uses a transmission line online monitoring system comprising five edge computing devices as an example for detailed description. The five devices are evenly deployed along a 50-kilometer transmission line, positioned on the sides of the towers, each powered by a solar power system. Device A is deployed at the beginning of the line, responsible for monitoring four towers; devices B and C are deployed in the middle section of the line, responsible for monitoring three and five towers respectively; devices D and E are deployed near the end of the line, responsible for monitoring four and three towers respectively. Each device uses a quad-core ARM Cortex-A55 processor with a clock speed of 2.0 GHz. The current remaining power of each device is as follows: Device A 85%, Device B 92%, Device C 45%, Device D 78%, and Device E 63%.
[0022] Step S1: Obtain the computing power resource information of each edge computing device in the online monitoring system of the transmission line, as well as the attribute information of the computing tasks to be processed by each device. The computing power resource information includes the available computing power of the CPU, the available memory capacity, and the current remaining power. The task attribute information includes the task type, the data size, and the expected processing time limit.
[0023] The purpose of this step is to obtain the current computing resource status of each edge computing device and the attribute information of the computing tasks to be processed, so as to provide a data basis for subsequent scheduling decisions.
[0024] Obtaining device computing power resource information: Each edge computing device collects its own status information in real time through a local resource monitoring module and reports it to the dispatch center via the inter-tower communication network (in this embodiment, a 5G private network is used, with a communication latency of less than 10ms). Specific data collected includes: • Available CPU computing power: Measured in floating-point operations per second (FLOPS). Device A currently has a CPU load of 40%, and its available computing power is approximately 60% of its maximum computing power.
[0025] • Available memory capacity: Device A has a total memory of 2GB, of which 680MB is currently used, leaving approximately 1.3GB of available memory.
[0026] • Current remaining battery power: Device A has 85% remaining battery power, Device B has 92%, Device C has 45%, Device D has 78%, and Device E has 63%.
[0027] Task attribute information retrieval: The current pending computing tasks for each device are as follows: Device A currently has 3 pending tasks: Task A1 is a wildfire identification task with a data size of 150MB and an expected processing time of 60 seconds from the current moment; Task A2 is an external damage detection task with a data size of 200MB and an expected processing time of 45 seconds; Task A3 is a routine inspection task with a data size of 100MB and an expected processing time of 120 seconds.
[0028] Device C currently has two pending tasks: Task C1 is an icing monitoring task with a data size of 80MB and an expected processing time of 90 seconds; Task C2 is an external damage detection task with a data size of 180MB and an expected processing time of 50 seconds.
[0029] The remaining devices currently have no pending tasks.
[0030] The scheduling center summarizes the attribute information of all pending tasks to form a task list, including fields such as task number, source device, task type, data volume, and expected processing time.
[0031] Step S2: Priority evaluation is performed on each computational task to be processed based on the task attribute information. The priority evaluation is calculated based on two dimensions: task urgency and task importance, to obtain a priority score for each task.
[0032] This step is one of the core components of this embodiment for achieving differentiated task processing. Its technical basis lies in the fact that different types of transmission line monitoring tasks have significantly different real-time requirements—wildfires and external damage incidents are related to line safety and public safety, requiring identification and alarm completion in the shortest possible time; icing monitoring has moderate timeliness requirements; and routine inspections have lower real-time requirements. By quantitatively assessing the urgency and importance of tasks, critical tasks can be prioritized. Specifically, the priority scoring process for each task is illustrated in the diagram below. Figure 2 As shown.
[0033] (a) Calculation of task urgency: For each task to be processed, obtain the time difference ΔT between the current time and the expected processing time limit of the task. In this embodiment: • Task A2 (External Damage Detection, Expected Time Limit 45 seconds): 5 seconds have passed since the task was completed, with 40 seconds remaining, ΔT=40 seconds.
[0034] • Task C2 (External Damage Detection, Expected Time Limit 50 seconds): 8 seconds have passed since the task was completed, with 42 seconds remaining, ΔT=42 seconds.
[0035] Task A1 (Wildfire Identification, Expected Time Limit 60 seconds): 2 seconds have passed since the task was completed, with 58 seconds remaining, ΔT=58 seconds.
[0036] • Task C1 (Icing Monitoring, Expected Time Limit 90 seconds): 10 seconds have passed since the task was completed, with 80 seconds remaining, ΔT=80 seconds.
[0037] • Task A3 (Routine Inspection, Expected Time Limit 120 seconds): 5 seconds have passed since the task was completed, with 115 seconds remaining, ΔT=115 seconds.
[0038] The time difference ΔT is mapped to an urgency score. The mapping rule uses a piecewise linear function. Preferably, in this embodiment, the specific mapping process is as follows: • When ΔT ≤ 30 seconds, the urgency score is set to the highest value of 1.0. This indicates that the task is nearing its deadline and must be addressed immediately.
[0039] • When 30 seconds < ΔT ≤ 90 seconds, the urgency score decreases linearly from 1.0 to 0.3. The calculation formula is: Score = 1.0 - 0.7 × (ΔT - 30) / 60.
[0040] • When ΔT > 90 seconds, the urgency score is set to the lowest value of 0.3, indicating that the task time limit is relatively lenient.
[0041] According to the above rules, the urgency scores for each task are as follows: Task A2 is 1.0, Task C2 is 1.0, Task A1 is 0.673, Task C1 is 0.417, and Task A3 is 0.3.
[0042] (ii) Calculation of task importance: The importance baseline values are preset according to the task type: the importance baseline value for wildfire identification and external damage detection tasks is 1.0, for ice accumulation monitoring tasks it is 0.7, and for routine inspection tasks it is 0.4.
[0043] Based on this, dynamic adjustments are made according to the task triggering frequency. The triggering frequency of each type of task in the past 24 hours is statistically analyzed. If the triggering frequency of a certain task type is higher than the preset frequency threshold (set to 2 times per hour in this embodiment), its importance score is reduced by 0.1; if the triggering frequency is lower than the preset frequency threshold (set to 0.5 times per hour in this embodiment), it is increased by 0.1.
[0044] The importance scores for each task were calculated as follows: Task A2 (external damage detection) 1.0, Task C2 (external damage detection) 1.0, Task A1 (wildfire identification) 1.0, Task C1 (ice accumulation monitoring) 0.7, and Task A3 (routine inspection) 0.4.
[0045] (III) Priority Score Synthesis: After normalizing the urgency and importance scores, a weighted sum is used to calculate the priority score. In this embodiment, the urgency weight is 0.6, and the importance weight is 0.4. Therefore, task A2 has a priority score of 1.0; task C2 has a priority score of 1.0; task A1 has a priority score of 0.804; task C1 has a priority score of 0.530; and task A3 has a priority score of 0.34.
[0046] Based on priority scores, tasks A2 and C2 are tied for the highest (1.0), followed by task A1 (0.804), then task C1 (0.530), and task A3 is the lowest (0.34).
[0047] Step S3: Based on the computing power resource information of each edge computing device and the priority score of each task, construct a scheduling optimization model with the goal of minimizing the processing latency of high-priority tasks and maximizing the system computing power utilization rate, and solve it to obtain the task allocation scheme.
[0048] This step involves constructing and solving a scheduling optimization model based on the computing power resource status of each device and the priority score of each task, to obtain the optimal task allocation scheme.
[0049] (I) Construction of the scheduling optimization model: The scheduling optimization model includes two optimization objectives: The primary optimization objective is to minimize the total processing latency of all pending tasks. The total processing latency is the sum of the waiting time and execution time of each task. The processing latency of each task is related to the current task queue length and processing rate of the device to which it is assigned.
[0050] The second optimization objective is to minimize the variance of the computing power load rate of each edge device. The computing power load rate is defined as the total computing power required for the currently assigned tasks of a device divided by the available computing power of the device's CPU. This objective is introduced to avoid the imbalance of computing power, where some devices are overloaded while others are idle.
[0051] The two optimization objectives are weighted and combined into a comprehensive optimization objective: Comprehensive Optimization Objective = 0.7 × Normalized Total Processing Latency + 0.3 × Normalized Load Rate Variance. The weighting coefficients are set based on the fact that total processing latency is directly related to the timeliness of task alarms and is crucial for high-priority tasks to complete on time, therefore it is given a higher weight.
[0052] The constraints include: Each task can only be assigned to one edge device for processing.
[0053] • The total computing power required for tasks assigned to any device shall not exceed the upper limit of the available computing power of the device's CPU.
[0054] • The total memory required by tasks assigned to any device shall not exceed the maximum available memory capacity of that device.
[0055] (II) Solution Method: An improved particle swarm optimization algorithm is used to solve the problem. The task allocation scheme is encoded as particle position vectors. In this embodiment, there are 5 tasks, and the position vector contains 5 elements. The value of each element is between 1 and 5, representing the allocation to the corresponding 5 devices.
[0056] The overall optimization objective is used as the fitness function; a smaller fitness value indicates a better allocation scheme. The particle swarm size is set to 50, and the maximum number of iterations is set to 200. During the iteration process, a linear decreasing inertia weight strategy is introduced into the particle swarm iteration. Preferably, in this embodiment, the inertia weight decreases linearly from 0.9 to 0.4. There are many linear decreasing functions; this embodiment uses a linear function to achieve the linear decreasing process. Specific implementations are not particularly limited in this regard. By introducing the inertia weight decreasing operation, the algorithm has a strong global search capability in the early stages to avoid getting trapped in local optima, and a strong local search capability in the later stages to achieve fine convergence.
[0057] After iteration, the task allocation scheme corresponding to the optimal particle is output. In this embodiment, the allocation scheme obtained after solving is as follows: Task A2 and Task A1 are assigned to device A for local execution; Task C2 is assigned to device B for execution (device B has sufficient computing power and is close to device C in communication distance); Task C1 is assigned to device D for execution; Task A3 is assigned to device E for execution (as a low-priority task, utilizing the surplus computing power of device E).
[0058] Step S4: Each edge computing device receives the assigned computing tasks according to the task allocation scheme, performs dynamic adjustment of computing power based on the current remaining power, determines the upper limit of the currently available computing power, and executes the received computing tasks under the constraint of the upper limit of computing power.
[0059] In this step, each edge device receives tasks according to the allocation scheme and dynamically adjusts computing power based on the current remaining power.
[0060] (a) Dynamic adjustment of computing power: First, for each device, obtain the current remaining power percentage. When the remaining power is higher than the first power threshold, the computing power limit is the device's maximum computing power. When the remaining power is lower than the first power threshold but higher than the second power threshold, the computing power limit decreases linearly according to the remaining power. When the remaining power is lower than the second power threshold, the computing power limit is reduced to 30% of the device's maximum computing power. The first power threshold is greater than the second power threshold.
[0061] Taking device C as an example, its current remaining battery power is 45%. The computing power adjustment rule is as follows: • When the remaining power is higher than the first power threshold (70% in this embodiment), the computing power limit is set to 100% of the device's maximum computing power.
[0062] • When the remaining battery power is below 70% but above the second battery power threshold (30% in this embodiment), the computing power limit decreases linearly according to the remaining battery power. The calculation formula is: Computing power limit percentage = 30% + (remaining battery power - 30%) / (70% - 30%) × 70%.
[0063] • When the remaining power is below 30%, the computing power limit is reduced to a preset percentage of the device's maximum computing power (30% in this embodiment).
[0064] Device C currently has 45% remaining battery power, falling between 30% and 70%. Its maximum computing power is calculated as 30% + (45% - 30%) / (70% - 30%) × 70% = 56.25%. In other words, the current usable maximum computing power of device C is approximately 56% of its maximum computing power.
[0065] Device A has 85% remaining battery power (above 70%), and its computing power limit remains at 100%. Device E has 63% remaining battery power (between 30% and 70%), and its computing power limit is approximately 57.75%.
[0066] (ii) Internal task queue sorting: When executing received computing tasks, the execution order of each task in the device's internal queue is further determined based on task priority scores: tasks with higher priority scores are placed at the front of the queue and executed first, while tasks with lower priority scores are placed at the back of the queue and executed later. When the device's local computing power is insufficient, the lowest priority task in the queue is marked as a task to be migrated, triggering an unloading operation. Each device executes tasks under the constraint of the computing power limit, and the execution order is determined internally by task priority scores, specifically: In this embodiment, device A is assigned tasks A2 (priority 1.0) and A1 (priority 0.804), with task A2 executed first, followed by task A1; device B is assigned task C2 (priority 1.0) and executed directly; device D is assigned task C1 (priority 0.530) and executed directly; device E is assigned task A3 (priority 0.34) and executed directly. During task execution, the devices continuously monitor their own computing power usage to ensure that it does not exceed the computing power limit.
[0067] Step S5: During the scheduling process, when it is detected that the computing power load rate of any device exceeds the preset load threshold or the remaining power is lower than the preset power threshold, the rescheduling process is triggered, and steps S2 to S4 are re-executed.
[0068] This step monitors the operating status of each edge device in real time during the scheduling process, and automatically performs rescheduling when the triggering conditions are met.
[0069] Triggered monitoring: The dispatch center monitors the computing load rate and remaining power of each device every second.
[0070] The load threshold that triggers the rescheduling process is when the device's CPU computing power load rate exceeds the load rate threshold or the memory occupancy rate exceeds the occupancy rate threshold. Preferably, in this embodiment, the load threshold is set to when the CPU computing power load rate exceeds 85% or the memory occupancy rate exceeds 85%, and the preset power threshold is when the device's current remaining power is less than 20%.
[0071] During the execution of task C2, device B suddenly triggered a new external damage detection task B1 (high priority), causing the CPU load rate to instantly rise to 92%, triggering a rescheduling process. At the same time, the remaining battery power of device C further decreased from 45% to 18%, also triggering the battery threshold condition.
[0072] Rescheduling execution: The rescheduling process re-executes steps S2 to S4, and re-prioritizes and reassigns all pending tasks (including existing unfinished tasks and newly arrived tasks).
[0073] Preferably, in this embodiment, during rescheduling, some tasks on device B that triggered the rescheduling are migrated to other edge devices with a computing load rate below 60% and remaining power above 40% for execution. Device D currently has a load rate of 45% and a remaining power of 75%, meeting the migration target conditions. The newly added task B1 on device B is migrated to device D for execution, reducing the load rate of device B to 70% and restoring a safe operating state.
[0074] After the rescheduling is completed, the scheduling center will distribute the updated task allocation plan to each edge device, and each device will continue to execute according to the new plan.
[0075] Technical effects of the scheduling strategy: Through a three-level scheduling architecture, the average processing latency of high-priority tasks is reduced by about 45% compared to the first-come-first-served strategy, the variance of equipment computing power load rate is reduced by about 60%, and the overall task completion rate of the system is increased to over 95%, effectively ensuring the real-time performance and reliability of online monitoring tasks for power transmission lines.
[0076] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.
Claims
1. A method for intelligent scheduling of computing power of edge computing equipment for power transmission lines, characterized in that, Includes the following steps: Step S1: Obtain the computing power resource information of each edge computing device in the online monitoring system of transmission lines, as well as the attribute information of the current computing tasks to be processed by each device. The computing power resource information includes available CPU computing power, available memory capacity and current remaining power. The task attribute information includes task type, data size and expected processing time limit. Step S2: Priority evaluation is performed on each computation task to be processed based on the task attribute information. The priority evaluation is calculated based on two dimensions: task urgency and task importance, to obtain a priority score for each task. Step S3: Based on the computing power resource information of each edge computing device and the priority score of each task, construct a scheduling optimization model with the goal of minimizing the processing latency of high-priority tasks and maximizing the system computing power utilization rate, and solve it to obtain the task allocation scheme. Step S4: Each edge computing device receives the assigned computing tasks according to the task allocation scheme, performs dynamic adjustment of computing power based on the current remaining power, determines the current available computing power limit, and executes the received computing tasks under the constraint of the computing power limit. Step S5: During the scheduling process, when it is detected that the computing power load rate of any device exceeds the preset load threshold or the remaining power is lower than the preset power threshold, the rescheduling process is triggered, and steps S2 to S4 are re-executed.
2. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, The computing power resource information of the edge computing device in step S1 specifically includes: the attribute information of each computing task to be processed specifically includes: the task type is divided into wildfire identification, external damage detection, ice accumulation monitoring and routine inspection, and the expected processing time limit is the time interval from the time the task arrives to the time required to complete.
3. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, The calculation method for the urgency of the task in step S2 is as follows: obtain the time difference between the current time and the expected processing time limit of the task, and map the time difference to an urgency score. The mapping adopts a piecewise linear function. When the time difference is less than or equal to the first time limit threshold, the urgency score is the highest value. When the time difference is greater than the first time limit threshold and less than or equal to the second time limit threshold, the urgency score decreases linearly as the time difference increases. When the time difference is greater than the second time limit threshold, the urgency score is the lowest value. The first time limit threshold is less than the second time limit threshold.
4. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 2, characterized in that, The method for calculating the importance of a task in step S2 is as follows: a baseline value for importance is preset based on the task type, and the baseline value for the task type is dynamically adjusted based on the triggering frequency of each type of task within a preset duration. Specifically, when the task triggering frequency is higher than a preset frequency threshold, the baseline value for importance is reduced by a preset value to obtain an importance score; when the task triggering frequency is lower than the preset frequency threshold, the baseline value for importance is increased by a preset value to obtain an importance score; otherwise, the baseline value for importance of the task type is used as its importance score.
5. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, The specific calculation method for the priority score in step S2 is as follows: the task urgency score and the task importance score are normalized respectively, and the weighted sum of the normalized urgency score and importance score is taken as the task priority score.
6. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, The scheduling optimization model in step S3 is constructed as follows: the first optimization objective is to minimize the total processing latency of all tasks to be processed, and the second optimization objective is to minimize the variance of the computing power load rate of each edge device. The first and second optimization objectives are weighted and combined into a comprehensive optimization objective. The constraints of the scheduling optimization model include: each task can only be assigned to one edge device for processing; the total computing power required by tasks assigned to any device does not exceed the upper limit of the available CPU computing power of that device; and the total memory required by tasks assigned to any device does not exceed the upper limit of the available memory capacity of that device.
7. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, The specific method for solving the scheduling optimization model in step S3 is as follows: an improved particle swarm optimization algorithm is used to solve the problem, the task allocation scheme is encoded as the position vector of the particle, the comprehensive optimization objective is used as the fitness function, and the task allocation scheme corresponding to the optimal particle is output after the iteration is completed. In the process of particle swarm iteration, a linear decreasing strategy of inertia weight is introduced.
8. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, The specific method for dynamically adjusting computing power based on the current remaining power in step S4 is as follows: obtain the current remaining power percentage of the device; when the remaining power is higher than the first power threshold, the upper limit of computing power is the maximum computing power of the device; when the remaining power is lower than the first power threshold but higher than the second power threshold, the upper limit of computing power decreases linearly according to the remaining power ratio; when the remaining power is lower than the second power threshold, the upper limit of computing power is reduced to a preset percentage of the maximum computing power of the device, wherein the first power threshold is greater than the second power threshold.
9. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, In step S4, when executing the received computing task, the execution order of each task in the internal queue of the device is further determined according to the task priority score: the task with the higher priority score is placed at the front of the queue and executed first, and the task with the lower priority score is placed at the back of the queue and executed later; when the computing power limit of the device is exceeded, the task with the lowest priority in the queue is marked as a task to be migrated and the unloading operation is triggered.
10. The intelligent scheduling method for computing power of edge computing equipment in transmission lines as described in claim 1, characterized in that, The load threshold for triggering the rescheduling process in step S5 is when the device's CPU computing power load rate exceeds the load rate threshold or the memory occupancy rate exceeds the occupancy rate threshold.
