Photovoltaic electric field task scheduling method, system, equipment and medium

Abstract

The invention discloses a photovoltaic electric field task scheduling method, system and device and a medium. The method comprises the steps of collecting device position data of a photovoltaic electric field to construct a photovoltaic electric field device communication topology network; constructing a task scheduling model, a communication transmission model and a calculation processing model to construct a multi-objective optimization problem and convert the multi-objective optimization problem into a deterministic optimization problem; and obtaining a Pareto frontier optimal solution set, and obtaining a task unloading decision and a resource allocation strategy of the edge server. According to the method, edge calculation is introduced, so that the problem of high time delay of long-distance back transmission of cloud data is avoided; constructing a multi-objective optimization problem by taking a time delay performance index and a load balancing performance index as optimization objectives, guaranteeing queue stability through scalar function optimization, and then performing optimization through a multi-objective evolutionary algorithm to obtain a Pareto leading edge optimal solution set; and different operation and maintenance scene priorities can be flexibly adapted, the task time delay is effectively reduced, the resource utilization rate of an edge server is improved, and the real-time performance and stability of photovoltaic electric field scheduling are enhanced.

Description

Technical Field

[0001] This invention relates to the field of edge computing technology, and in particular to a photovoltaic power plant task scheduling method, system, device and medium. Background Technology

[0002] As a core infrastructure for renewable energy consumption and the low-carbon transformation of the power grid, photovoltaic (PV) power plants require the continuous collection of massive amounts of environmental and equipment data, such as solar radiation, temperature, equipment current and voltage, and inverter operating status, for short-term power output prediction, fault early warning and location, and operation and maintenance scheduling. The real-time processing of this data directly impacts grid connection safety and power generation revenue—delayed data processing not only reduces the accuracy of power output prediction and leads to deviations in power generation plans, but may also delay fault handling, causing economic losses and safety risks. With the gradual application of edge computing technology in the energy sector, deploying edge servers in PV power plants to achieve near-end data processing, shortening data backhaul links, and improving real-time response capabilities has become an important development direction for PV power plant task scheduling.

[0003] Currently, traditional centralized cloud processing requires long-distance backhaul of raw data to the cloud center, which is susceptible to factors such as uplink and downlink bandwidth limitations and link congestion. This results in high end-to-end latency and significant jitter, easily creating processing bottlenecks in scenarios with sudden alarms and high-concurrency data collection, making it difficult to meet the near real-time analysis and linkage control requirements of the photovoltaic power plant. Furthermore, some power plant-side scheduling solutions focus only on a single optimization objective, such as pursuing only the minimum latency or maximum throughput, without comprehensively considering the load balancing of multiple edge servers. This leads to excessive concentration of tasks on a few high-performance edge servers, causing problems such as long-term overload, queue backlog, and task timeout dropping, while the remaining servers are idle, resulting in a decrease in overall resource utilization and insufficient service stability, making it unable to adapt to the complex and ever-changing task processing needs of photovoltaic power plants. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a photovoltaic power plant task scheduling method, system, equipment, and medium to solve the problems of traditional cloud-based centralized processing requiring long-distance data transmission, being susceptible to bandwidth and congestion, having high latency and large jitter, making it difficult to meet near real-time requirements, and some site-side solutions focusing only on a single optimization objective without coordinating edge server load balancing, resulting in concentrated tasks, wasted resources, and poor service stability.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a photovoltaic power plant task scheduling method, comprising: Collect equipment location data of the photovoltaic power plant, and construct a communication topology network for the photovoltaic power plant equipment based on the equipment location data; Based on the communication topology network of the photovoltaic power plant equipment, a task scheduling model, a communication transmission model, and a computational processing model are constructed. Based on the task scheduling model, communication transmission model, and computation processing model, a multi-objective optimization problem is constructed. The multi-objective optimization problem is then subjected to queue stability constraints, transforming the constrained multi-objective optimization problem into a deterministic optimization problem. A multi-objective evolutionary algorithm is used to solve the deterministic optimization problem to obtain the Pareto front optimal solution set. Based on a comprehensive trade-off of the optimization objectives, task offloading decisions and resource allocation strategies for edge servers are obtained from the Pareto front optimal solution set.

[0007] As a preferred embodiment of the photovoltaic power plant task scheduling method described in this invention, the step of constructing the photovoltaic power plant equipment communication topology network includes: Collect device location data for various devices within the photovoltaic power plant and in the edge server; Based on the device location data, a communication topology network for photovoltaic power plant equipment is constructed.

[0008] The beneficial effects of this preferred technical solution are as follows: by collecting equipment location data of meteorological equipment, sensors, edge servers, etc. in the photovoltaic power plant, it provides basic data support for the construction of the photovoltaic power plant equipment communication topology network; the photovoltaic power plant equipment communication topology network constructed based on the equipment location data can truly reflect the spatial relationship between the equipment and the edge server, laying the foundation for the construction of subsequent task scheduling models and communication transmission models with a network architecture that fits the actual scenario.

[0009] As a preferred embodiment of the photovoltaic power plant task scheduling method described in this invention, the steps of constructing the task scheduling model, the communication transmission model, and the computational processing model include: The communication scheduling cycle in the photovoltaic power plant equipment communication topology network is divided into multiple time slots of equal length; task data is generated for each device in the time slot; an initial data cache queue is established for each device to cache the task data; and a task scheduling model is obtained by setting a queue upper limit constraint on the initial data cache queue. Based on the communication topology network of the photovoltaic power plant equipment, a communication transmission model for the task data from the equipment to the edge server is constructed; The task scheduling model is used to process the unloading task data on the edge server and construct a computational processing model.

[0010] The beneficial effects of this preferred technical solution are as follows: By dividing time slots to achieve time-domain management of task scheduling, combined with the initial data cache queue and queue upper limit constraints, task data overflow can be avoided and the orderly storage of task data can be guaranteed, providing a stable data management foundation for the task scheduling model; the communication transmission model constructed based on the communication topology network of photovoltaic power plant equipment can fit the actual communication link between the equipment and the edge server, ensuring a reasonable data transmission path; the computational processing model constructed based on the task scheduling model can adapt to the task processing scenario of the edge server. The three work together to provide model support that fits the actual needs of the photovoltaic power plant for the subsequent construction and solution of multi-objective optimization problems, improving the feasibility of the scheduling strategy.

[0011] As a preferred embodiment of the photovoltaic power plant task scheduling method described in this invention, the step of constructing a multi-objective optimization problem includes: The latency performance index is obtained through the communication transmission model and the computational processing model, the load balancing performance index is obtained through the computational processing model, and the optimization target is obtained by combining the latency performance index and the load balancing performance index. Optimization decisions are obtained through task offloading in the task scheduling model, transmission links in the communication transmission model, and resource allocation in the computing processing model. The constraints are obtained by combining the queue upper limit constraint in the task scheduling model, the access constraint in the communication transmission model, and the capacity constraint in the computation processing model. Based on the optimization objective, optimization decisions, and constraints, a multi-objective optimization problem is constructed.

[0012] The beneficial effects of this preferred technical solution are as follows: supported by communication transmission model and computing processing model, the optimization objectives of latency performance index and load balancing performance index are extracted, which not only ensures the real-time performance of task processing, but also avoids load imbalance of edge servers; at the same time, the optimization decision and constraints are clarified, so that the construction of multi-objective optimization problem takes into account both optimization direction and actual boundary, which not only meets the core requirements of photovoltaic power plant task scheduling, but also avoids the problem of decision exceeding the communication and computing capabilities of equipment, laying a rigorous problem framework for solving feasible and efficient scheduling strategies.

[0013] As a preferred embodiment of the photovoltaic power plant task scheduling method described in this invention, the step of transforming the problem into a deterministic optimization problem includes: Based on the queue upper limit constraint in the task scheduling model, and combined with the task data, the initial data cache queue is modified to obtain the corrected data cache queue. Based on the scalar function defined in the modified data cache queue, and combined with the optimization objective of the multi-objective optimization problem, a drift penalty function is constructed and its upper limit is derived to form the drift penalty analysis results; By minimizing the drift penalty analysis results, the multi-objective optimization problem is transformed into a deterministic optimization problem without long-term queue constraints.

[0014] The beneficial effects of this preferred technical solution are as follows: by modifying the initial data buffer queue, it further adapts to changes in task data and provides a data foundation for queue stability constraints; by constructing a drift penalty function in conjunction with the optimization objective and clarifying the upper limit of the drift penalty function, it effectively balances the needs of queue stability and optimization objectives; finally, it transforms the complex problem with long-term queue constraints into a deterministic optimization problem, significantly reducing the difficulty of solving the problem, while ensuring the effectiveness of the constraints, and providing a concise and rigorous problem form for subsequent solving and outputting feasible scheduling strategies.

[0015] As a preferred embodiment of the photovoltaic power plant task scheduling method described in this invention, the step of obtaining the Pareto front optimal solution set includes: Based on the deterministic optimization problem, the initial solution of the multi-objective evolutionary algorithm is obtained by encoding the task offloading, transmission link and resource allocation in the optimization decision. The running parameters of the multi-objective evolutionary algorithm are configured according to the initial solution to generate an initial population. The initial population is selected by combining the constraints to obtain a compliant population; the compliant population is iteratively updated according to the optimization objective to obtain an iterative population; A maximum number of iterations is preset. When the number of iterations of the iterative population reaches the maximum number of iterations, the Pareto front optimal solution set is output.

[0016] The beneficial effects of this preferred technical solution are as follows: Initial solutions are generated by encoding based on optimization decisions; the initial population generated is ensured to fit the actual problem by combining the configuration of operating parameters; and a compliant population is obtained by screening under constraints, avoiding invalid solutions from occupying computational resources. The population is iteratively updated according to the optimization objective, enabling directional approximation of the optimal solution domain. Combined with a preset maximum number of iterations, the solution efficiency and accuracy are balanced. The final Pareto front optimal solution set covers multi-objective trade-off schemes, providing a rich and reliable solution foundation for subsequent selection of suitable scheduling strategies.

[0017] As a preferred embodiment of the photovoltaic power plant task scheduling method described in this invention, the steps of obtaining task offloading decisions and edge server resource allocation strategies include: The weights of latency performance indicators and load balancing performance indicators in the optimization objective are preset, and a comprehensive cost evaluation is performed on the Pareto front optimal solution set to calculate the comprehensive cost value. The solution with the optimal overall cost value is selected as the target solution, and the task unloading decision and the resource allocation strategy for the edge server are obtained from the target solution.

[0018] The beneficial effects of this preferred technical solution are as follows: by pre-setting the weights of latency performance indicators and load balancing performance indicators, it can adapt to the priority requirements of different operation and maintenance scenarios of photovoltaic power plants, such as emphasizing latency performance during fault handling and emphasizing load balancing performance during daily operation and maintenance; by performing a comprehensive cost evaluation on the Pareto front optimal solution set, it can select the target solution that best fits the actual needs from the multi-objective trade-off solutions; and the task offloading decision and edge server resource allocation strategy obtained by the final analysis effectively balance the real-time performance of tasks and the resource utilization rate, and improve the actual adaptability of the scheduling strategy.

[0019] Secondly, the present invention provides a photovoltaic power plant task scheduling system, comprising: The data acquisition module is used to collect device location data of equipment and edge servers within the photovoltaic power plant; The network construction module is used to construct a communication topology network for photovoltaic power plant equipment based on the device location data. The model building module is used to build a task scheduling model, a communication transmission model, and a computational processing model based on the communication topology network of the photovoltaic power plant equipment. The optimization problem construction module is used to construct a multi-objective optimization problem by combining the task scheduling model, communication transmission model and computation processing model. The problem transformation module is used to introduce queue stability constraints into the multi-objective optimization problem and transform it into a deterministic optimization problem; The solution module is used to solve the deterministic optimization problem using a multi-objective evolutionary algorithm and output the Pareto front optimal solution set; The decision generation module is used to generate task unloading decisions and resource allocation strategies for edge servers based on the Pareto front optimal solution set and the weight evaluation of the optimization objective.

[0020] Thirdly, the present invention provides an electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the photovoltaic power plant task scheduling method.

[0021] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the photovoltaic power plant task scheduling method.

[0022] Compared with existing technologies, the beneficial effects of this invention are as follows: by introducing edge computing and combining the location data of photovoltaic power plant equipment to construct a communication topology network and task scheduling model, communication transmission model, and computing processing model for photovoltaic power plant equipment, the problem of high latency in long-distance data backhaul from the cloud is avoided; furthermore, a multi-objective optimization problem is constructed with latency performance indicators and load balancing performance indicators as optimization objectives, and the stability of the queue is ensured by scalar function optimization and transformed into a deterministic problem, and then the Pareto front optimal solution set is obtained by multi-objective evolutionary algorithm; it can also flexibly adapt to the priorities of different operation and maintenance scenarios, effectively reduce task latency, improve the utilization rate of edge server resources, and enhance the real-time performance and stability of photovoltaic power plant scheduling. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a schematic diagram of the overall process of a photovoltaic power plant task scheduling method according to an embodiment of the present invention. Detailed Implementation

[0025] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0026] Example 1, referring to Figure 1 As an embodiment of the present invention, a photovoltaic power plant task scheduling method is provided, comprising: S100. Collect equipment location data of the photovoltaic power plant, and construct a photovoltaic power plant equipment communication topology network based on the equipment location data.

[0027] S200. Based on the communication topology network of the photovoltaic power plant equipment, construct a task scheduling model, a communication transmission model, and a computational processing model.

[0028] S300. Based on the task scheduling model, communication transmission model, and computation processing model, construct a multi-objective optimization problem.

[0029] S400. Apply queue stability constraints to the multi-objective optimization problem, and transform the constrained multi-objective optimization problem into a deterministic optimization problem.

[0030] S500. The deterministic optimization problem is solved using a multi-objective evolutionary algorithm to obtain the Pareto front optimal solution set. Based on a comprehensive trade-off of the optimization objectives, the task offloading decision and the resource allocation strategy for the edge server are obtained from the Pareto front optimal solution set.

[0031] It should be noted that photovoltaic power plants, as core infrastructure of new energy, require the continuous collection of massive amounts of data, such as solar radiation, equipment current and voltage, and inverter status, to support output prediction and fault handling. The real-time performance of data processing directly affects grid connection safety and power generation revenue. Traditional centralized cloud processing requires long-distance data transmission, which is susceptible to bandwidth limitations and link congestion, resulting in high end-to-end latency and large jitter, easily creating processing bottlenecks in sudden alarm scenarios. Some site-side scheduling solutions focus only on a single optimization objective, such as minimizing latency, without coordinating edge server load balancing, leading to excessive task concentration causing overload and backlog, while other servers remain idle, resulting in low resource utilization.

[0032] Therefore, to address the aforementioned issues of high data processing latency, singular scheduling objectives, and unbalanced resource allocation in photovoltaic power plants, the following steps (S100-S500) are employed: First, a communication topology network for photovoltaic power plant equipment is constructed using equipment location data, laying a scheduling foundation that aligns with the actual scenario. Next, a task scheduling model, a communication transmission model, and a computational processing model are built, thereby constructing a multi-objective optimization problem. By applying queue stability constraints, the problem is transformed into a deterministic optimization problem, reducing solution complexity and ensuring stable data storage. Finally, a multi-objective evolutionary algorithm is used to optimize the task offloading decision and the resource allocation strategy for edge servers, providing data support and execution basis for near-real-time scheduling of photovoltaic power plants, efficient utilization of edge resources, and improved operation and maintenance efficiency.

[0033] Example 2, refer to Figure 1 As an embodiment of the present invention, a photovoltaic power plant task scheduling method is provided based on the above embodiment.

[0034] In this embodiment, a medium-sized photovoltaic power plant, covering approximately 50 acres, including 3 edge server deployment areas, 6 photovoltaic array areas, and 2 meteorological monitoring areas, is used as the application scenario. The specific implementation steps for constructing the communication topology network of the photovoltaic power plant equipment in S100 A1~A2 are as follows: A1. Collect device location data for various devices in the photovoltaic power plant and edge server.

[0035] Specifically, the types of equipment to be collected include, but are not limited to, meteorological monitoring equipment, such as wind speed sensors, temperature sensors, and solar radiation sensors, totaling 8 units, deployed at the highest point of the site; photovoltaic operation equipment, such as intelligent photovoltaic panels, inverter monitoring modules, and string current sensors, totaling 120 units, deployed in 6 photovoltaic array zones; edge computing equipment, including 3 edge servers, deployed next to the central control boxes in array zones 1, 3, and 5 respectively; and auxiliary equipment, such as security cameras and environmental temperature and humidity sensors, totaling 15 units, arranged along the site fence and passageways.

[0036] The collected device location data is in the form of three-dimensional coordinates (x, y, z), where x is the east-west distance, y is the north-south distance, and z is the device installation elevation. For example, the coordinates of the northwest meteorological station are (80.2, 45.6, 2.8); the coordinates of edge server 1 are (200.5, 120.3, 3.0); the smart photovoltaic panels in array area 2 are divided into strings, and the coordinate range of strings A1-A10 is (150.0-180.0, 90.0-110.0, 1.8); the coordinates of the security camera are (300.1, 250.5, 2.5).

[0037] A2. Based on the device location data, construct a communication topology network for the photovoltaic power plant equipment.

[0038] Specifically, the south entrance of the site is uniformly adopted as the coordinate origin (0, 0, 0), and the three-dimensional coordinates of all equipment are converted into the site's local coordinate system to avoid coordinate deviations between different positioning points. Based on the equipment communication method, such as wireless communication between edge servers and core equipment with a coverage radius ≤ 50 meters, and wired communication between edge servers, the straight-line distance between devices is calculated using the formula: distance... ; For devices within ≤50 meters and without obstructions, the ray casting method is used to determine whether tall equipment such as weather station towers obstruct the wireless signal; if so, a communication link can be established. A node link graph structure is used, where the set of device nodes V = {meteorological equipment nodes}. Photovoltaic equipment nodes Edge server nodes Auxiliary equipment nodes }, Link set E={ ( (), (), The link relationships are recorded using an adjacency matrix, and the matrix elements are... This indicates that there is a link between nodes i and j. Indicates no link, such as in the adjacency matrix. and The matrix element is 1. and With a distance of 62 meters, exceeding the coverage area, and matrix elements of 0, a communication topology network for photovoltaic power plant equipment is constructed.

[0039] In an optional implementation, step S100 can also incorporate device communication bandwidth parameters to optimize the communication transmission model. The steps are as follows: Collect the communication interface bandwidth of each device, such as a weather sensor using a LoRa module with a bandwidth of 125kHz; an edge server Ethernet interface with a bandwidth of 100Mbps; and a smart photovoltaic panel using an RS485 interface with a bandwidth of 9600bps. Supplement the link attributes of the communication topology network with bandwidth threshold parameters, such as link... ( The bandwidth threshold is 125kHz. When constructing the communication transmission model, the maximum data transmission rate of the link can be calculated based on the bandwidth threshold parameter. For example, link rate = bandwidth × modulation efficiency. If the LoRa modulation efficiency is taken as 0.8, then the link... The link rate is 125kHz × 0.8 = 100kbps, which avoids transmission congestion caused by insufficient bandwidth and improves the model's adaptability to actual communication capabilities.

[0040] In another optional implementation, step S100 can further optimize the task scheduling model by combining equipment task priorities. The steps are as follows: Prioritize tasks according to the functional importance of the photovoltaic power plant equipment, such as fault monitoring equipment as first-level priority, real-time power generation data acquisition equipment as second-level priority, and environmental monitoring equipment as third-level priority; when constructing the task scheduling model, assign link occupancy weights to devices of different priorities, such as first-level devices with a weight of 0.4, second-level devices with a weight of 0.3, and third-level devices with a weight of 0.3; when multiple devices compete for the same communication link, allocate link resources according to their weights, such as link allocation... ( Simultaneously, two Level 1 devices and three Level 3 devices are connected. The Level 1 devices occupy a total bandwidth of 0.8, and the Level 3 devices occupy 0.2 bandwidth, ensuring that high-priority tasks are transmitted first and reducing task latency in emergency scenarios.

[0041] In this embodiment of the application, step S200, which involves constructing the task scheduling model, the communication transmission model, and the computation processing model, includes steps B1 to B3: B1. Divide the communication scheduling cycle in the photovoltaic power plant equipment communication topology network into multiple time slots of equal length; generate task data for each device in the time slot; establish an initial data cache queue for each device to cache the task data; and obtain the task scheduling model by setting a queue upper limit constraint on the initial data cache queue.

[0042] Specifically, the daily communication scheduling cycle of the photovoltaic power plant, which is 1440 minutes, is divided into 144 equal-length time slots, each time slot lasting 10 minutes. For each time slot, the device node Generate task data triples 'i' represents the equipment number, and the equipment location data is taken from the equipment's functional characteristics, such as meteorological equipment. wind speed sensor For example, , , Photovoltaic equipment string current sensors For example, , , Auxiliary equipment security cameras For example, , , Finally, an initial data cache queue is established for each device. And set queue upper limit constraints to prevent overflow, such as in meteorological equipment. Photovoltaic equipment Auxiliary equipment Thus, the task scheduling model is obtained.

[0043] B2. Based on the communication topology network of the photovoltaic power plant equipment, construct a communication transmission model for the task data from the equipment to the edge server.

[0044] Specifically, based on the communication topology of the photovoltaic power plant equipment, orthogonal frequency division multiple access (OFDMA) technology is used to divide the communication interface bandwidth into eight orthogonal subcarriers. Each subcarrier has a bandwidth of 15kHz; Distributed to photovoltaic equipment , Assigned to meteorological equipment , Assigned to auxiliary equipment Based on the equipment's three-dimensional coordinate data, use the Euclidean distance formula. The straight-line distance between the computing device and the edge server, combined with the channel gain formula: in, Let s be the unit channel power gain and s be the edge server number. Substitute these values ​​into the rate formula: Among them, the equipment has a unified transmission power. Gaussian white noise , For subcarrier bandwidth (15kHz). Coordinates (160.3, 100.5, 1.8) and Distance at coordinates (200.5, 120.3, 3.0) Calculated ; Configure each device to offload to only one edge server within a single time slot, such as... Access only Matching adjacency matrix Each subcarrier serves only one device, such as Only assigned to To avoid transmission conflicts, a communication transmission model is obtained.

[0045] B3. The task scheduling model is used to process the unloading task data on the edge server to construct a computational processing model.

[0046] Specifically, based on 3 edge servers Set the maximum CPU computing frequency for the edge server. , With consistent configuration, substitute the values ​​into the processing delay formula: in, CPU frequency allocated to server s for device i .like of , Uninstall to If the allocation ,but ; Set the total allocated CPU frequency for a single edge server to not exceed ,Right now , For the set of device nodes connected to server s, ; Measured using the Jain Fairness Index The load balancing performance index is obtained by determining the load balancing properties, and the formula is: Where J is the load balancing performance metric. Let M be the total computational workload of server s, and M be 3. For example, time slots. hour, of , of , of Calculated It has excellent load balancing performance, approaching 1. Transmission delay in conjunction with communication transmission models The total task delay is obtained. This yields the computational processing model.

[0047] In an optional implementation, step S200 may further involve real-time channel quality adjustment of subcarrier parameters, the steps of which include: using meteorological equipment such as... Wind speed sensor and Temperature sensors collect real-time environmental data; for example, when wind speed > 5 m / s, the channel signal-to-noise ratio is calculated. ,like This reduces the LoRa modulation efficiency from the base 0.8 to 0.6, while increasing the subcarrier bandwidth from 15kHz to 20kHz to avoid data packet loss; when At that time, the LoRa modulation efficiency was increased to 0.9, the subcarrier bandwidth was restored to 15kHz, saving spectrum resources and making the communication transmission model more suitable for the outdoor environment of photovoltaic power plants.

[0048] In another optional implementation, step S200 can also adjust the cache queue strategy according to task priority, the steps being: continuing with fault monitoring devices such as inverter fault sensors from S100. Priority is assigned as follows: Level 1 for real-time power generation data equipment, Level 2 for environmental monitoring equipment, and Level 3 for other equipment. The upper limit of the buffer queue is adjusted accordingly. For example, for Level 1 equipment… Increase from 20MB to 30MB to prevent fault data overflow, Level 3 equipment. Reduce the size from 10MB to 8MB to free up resources; simultaneously, set a priority uninstallation mechanism when cache usage exceeds 80%. of If the occupancy rate is 83%, which is greater than 80%, then it will be given priority in the next time slot. distribute Link CPU resources This further ensures the real-time performance of high-priority tasks.

[0049] In this embodiment of the application, step S300, the step of constructing a multi-objective optimization problem, includes C1 to C4: C1. The latency performance index is obtained through the communication transmission model and the computational processing model, the load balancing performance index is obtained through the computational processing model, and the optimization target is obtained by combining the latency performance index and the load balancing performance index.

[0050] Specifically, based on the transmission delay of the communication transmission model Processing latency of the computational processing model The formula for calculating the latency performance index is: in, For latency performance indicators, For the total number of devices, For task offloading decision variables, 1 indicates that device i is offloaded to server s in time slot t. For example, time slot hour, of , , Uninstall to ,but , , =80.4s; The load balancing performance index J can be calculated using the computational processing model: when The L values ​​are respectively , , hour, Excellent balance; if of , of ,but We need to optimize load balancing to obtain the optimization objective of the multi-objective optimization problem.

[0051] C2. Optimization decisions are obtained through task unloading in the task scheduling model, transmission links in the communication transmission model, and resource allocation in the computing processing model.

[0052] Specifically, task unloading decision ,when This indicates that device i will offload the task to edge server s in time slot t, and the single device, single server condition is met. For example, V9 in time slots Only selectable ,Right now , Transmission link allocation ,when This indicates that device i uses subcarrier f to transmit data in time slot t, satisfying the single subcarrier single device requirement. For example, in S100 Using subcarriers , Resource allocation This indicates that server s is the CPU frequency allocated to device i in time slot t, which must satisfy... For example, in S200 distribute .

[0053] C3. Constraints are obtained by combining the queue upper limit constraint in the task scheduling model, the access constraint in the communication transmission model, and the capacity constraint in the computing processing model.

[0054] Specifically, the upper limit constraint of the task scheduling model is the buffer queue of device i in time slot t. Must not exceed the upper limit ,Right now , And the queue update satisfies The access constraint of the communication transmission model is that devices can only connect to servers with communication links. ,Right now and If there is no link, then Subcarrier allocation matches the device type, such as photovoltaic equipment using only : , The set of subcarriers allowed for device i; the capability constraint of the computational processing model is that the total allocated CPU frequency of a single server does not exceed the upper limit: , .

[0055] C4. Based on the optimization objective, optimization decision and constraints, a multi-objective optimization problem is constructed.

[0056] Specifically, by integrating the optimization objective, decision variables, and constraints into a mathematical model, a multi-objective optimization problem is obtained, as follows: In an optional implementation, step S300 may further introduce an energy consumption index as a third optimization objective, the step being: defining the edge server energy consumption. The optimization objective is expanded to , , Furthermore, by using a weighting method, such as a weight of 0.4 for latency performance indicators, 0.3 for load balancing performance indicators, and 0.3 for edge server energy consumption, a comprehensive target is generated to meet the green and low-carbon operation and maintenance requirements of photovoltaic power plants.

[0057] In another alternative implementation, step S300 may further adjust the constraint relaxation coefficient, as follows: for high-priority tasks, such as fault monitoring equipment... Allow temporary relaxation of the queue upper limit constraint. Increase from 20MB to 30MB and introduce slack variables. Change the constraint to At the same time, an optimization penalty term is added to the optimization objective. ,in, To optimize the target penalty coefficient, we need to ensure that high-priority tasks are not lost while avoiding excessive relaxation that could affect system stability.

[0058] In this embodiment of the application, step S400, the steps of transforming the problem into a deterministic optimization problem, include D1 to D3: D1. Based on the queue upper limit constraint in the task scheduling model, modify the initial data cache queue according to the task data to obtain the corrected data cache queue.

[0059] Specifically, the initial data cache queue Update rules: Based on this, and combined with queue upper limit constraints and task data, correction parameters are defined: Let... The buffer offset adjustment amount for device i in time slot t, when Approaching the upper limit hour, If positive, release the cache; when... When too low, If the value is negative, then a cache is reserved; Corrected queue formula: Corrected data cache queue satisfy: And it satisfies the condition that the adjusted value does not exceed the upper limit. Such as photovoltaic equipment ; Such as time gap At that time, photovoltaic equipment , , , ,because Approaching the upper limit, take ,but .

[0060] D2. Based on the scalar function defined in the modified data cache queue, and in conjunction with the optimization objective of the multi-objective optimization problem, construct the drift penalty function and derive the upper limit of the drift penalty function to form the drift penalty analysis result.

[0061] Specifically, based on the Lyapunov scalar function defined by the modified data cache queue, and combined with the optimization objective of the multi-objective problem, a drift penalty function is constructed, as shown in the following formula: in Set a set of correction data cache queues for all devices; construct a drift penalty function: Combined with latency performance indicators And load balancing performance metric J, introduce latency performance weights. Load balancing weights ,get ,in, This is because the optimization objective is to maximize the load balancing performance metric J, which is equivalent to minimizing... ; Deriving the upper limit and forming the drift penalty analysis results: using , such as the largest , We obtain the upper bound of the drift penalty function: where constant , , The constant is calculated. .

[0062] D3. By minimizing the drift penalty analysis results, the multi-objective optimization problem is transformed into a deterministic optimization problem without long-term queue constraints.

[0063] Specifically, based on the drift penalty analysis results, since C is a constant, minimizing the upper bound of the drift penalty function is equivalent to minimizing... The objective function is transformed from a bi-objective to a single objective, while the constraints are simplified by using Lyapunov scalar function optimization to transform them into on-the-fly corrected queue constraints. Since there is no need to consider long-term trends, the problem is ultimately transformed into a deterministic optimization problem without long-term queue constraints. The mathematical model is as follows: In an alternative implementation, the Lyapunov weights can also be adjusted in step S400. The steps are as follows: Based on the equipment task priority, when a primary device, such as a fault sensor... , When the latency performance exceeds the 75% limit, the latency performance weight will be adjusted. The performance was increased from 0.6 to 0.8, prioritizing the reduction of fault data processing latency and load balancing weights. Dropped to 0.2; when all devices hour, It dropped to 0.4. The performance has been improved to 0.6, prioritizing load balancing optimization and making it more suitable for different photovoltaic power plant scenarios.

[0064] In another alternative implementation, step S400 may also introduce a queue fluctuation penalty term, the steps of which are: defining the fluctuation coefficient. Set fluctuation threshold Add a drift penalty term to the drift penalty function. ,in The drift penalty coefficient is used to define the final objective function, which then becomes: This helps avoid data loss caused by drastic fluctuations in the queue.

[0065] In this embodiment of the application, step S500, the step of obtaining the task offloading decision and the resource allocation strategy of the edge server, includes E1~E2: E1. Preset the weights of latency performance indicators and load balancing performance indicators in the optimization objective, perform a comprehensive cost evaluation on the Pareto front optimal solution set, and calculate the comprehensive cost value.

[0066] Specifically, preset latency performance weights =0.6, load balancing weight For each solution in the Pareto front optimal solution set, calculate the comprehensive cost value: For example, the overall cost of the Pareto solution set A. .

[0067] The steps for obtaining the optimal solution set at the Pareto front include E1.1 to E1.3: E1.1. Based on the deterministic optimization problem, the initial solution of the multi-objective evolutionary algorithm is obtained by encoding the task unloading, transmission link and resource allocation in the optimization decision. The running parameters of the multi-objective evolutionary algorithm are configured according to the initial solution to generate the initial population.

[0068] Specifically, the optimization decision variables are encoded as chromosomes to adapt to different types of task offloading, transmission links, and resource allocation, such as binary-encoded task offloading decisions. ,length That is, 143 devices × 3 edge servers, where each "1" indicates that device i is offloaded to server s in time slot t; binary encoded transmission link allocation. ,length That is, 143 devices × 8 subcarriers, where each "1" indicates that device i uses subcarrier f in time slot t; real number coding resource allocation. ,scope It has 143 dimensions, which are mapped to CPU frequency values; Configure parameters for multi-objective evolutionary algorithms, such as NSGA-II algorithm parameters: population size. Crossover probability Probability of mutation This generates the initial population. For example, in a certain chromosome segment, the device... The binary bit is "1", and the subcarrier is... The binary bit is "1". The real value is 0.4, which means 0.4 GHz is allocated.

[0069] E1.2. Based on the constraints, the initial population is selected to obtain a compliant population; according to the optimization objective, the compliant population is iteratively updated to obtain an iterative population.

[0070] Specifically, the initial population is traversed, and individuals that violate the constraints are removed, including: adjacency matrix constraints: if device i has no link with server s, such as... But in chromosomes If the total allocation frequency of edge servers s in an individual is high, then the individual is filtered out; Then, individuals are filtered out to obtain a compliant population; For a compliant population, simulated binary crossover and polynomial mutation are used to generate offspring. Fitness is calculated based on the optimization objective, and superior individuals are selected for the next generation through fast non-dominated sorting and crowding distance to obtain an iterative population. For example, the fitness of a certain offspring individual... , Because of its high non-dominated level and high crowding, it is retained in the iterative population.

[0071] E1.3. Preset maximum number of iterations. When the number of iterations of the iterative population reaches the maximum number of iterations, output the Pareto front optimal solution set.

[0072] Specifically, the maximum number of iterations is preset. When the iterative population reaches 200 generations, evolution stops, and the Pareto front optimal solution set is output.

[0073] E2. Select the solution with the optimal comprehensive cost value as the target solution, and parse the task unloading decision and the resource allocation strategy of the edge server from the target solution.

[0074] Specifically, the solution with the lowest overall cost is selected as the objective solution, and the task offloading decision and the resource allocation strategy for the edge server are obtained analytically: such as in binary bits. , , Bits equal to "1" correspond to devices offloading to designated edge servers; as in real number encoding. , This refers to the CPU frequency allocated by the server to the device. Ultimately, this forms an executable scheduling policy, such as: Array Zone 1... Uninstall all , for Allocate 0.4GHz, 0.3GHz was allocated to achieve a balance between task real-time performance and resource utilization.

[0075] In an optional implementation, step S500 can further adjust the weights according to the scenario. The steps are as follows: the scenario is identified by the task priority in S200, such as the proportion of first-level faulty devices. If the proportion of first-level faulty devices is >30%, it is a faulty scenario, and the weights are automatically adjusted. Increased to 0.8 Reduce to 0.2; if the proportion is less than 10% and it is a daily scenario, then Reduced to 0.4 The value has been increased to 0.6, making the overall cost assessment more aligned with operational and maintenance needs.

[0076] In another optional implementation, step S500 can also incorporate energy consumption optimization, which involves adding an energy consumption weight to the overall cost. The formula has been updated to , This represents the total power consumption of the edge server. When analyzing resource allocation, the solution with low energy consumption and high balance is selected first, which is suitable for the green operation and maintenance goals of photovoltaic power plants.

[0077] In summary, this invention takes a medium-sized photovoltaic power plant as the scenario. First, it collects the three-dimensional coordinates of the equipment to construct a communication topology network for the photovoltaic power plant equipment. Then, it builds a task scheduling model, a communication transmission model, and a computational processing model. Next, it constructs a multi-objective optimization problem, which is transformed into a deterministic optimization problem by constraining queue stability. Finally, it obtains the Pareto front optimal solution set through a multi-objective evolutionary algorithm. After weighing the weights, it analyzes the task offloading decision and the resource allocation strategy of the edge server, achieving a balance between scheduling real-time performance and resource utilization.

[0078] Example 3 illustrates a photovoltaic power plant task scheduling method. It should be noted that the technical solution of this photovoltaic power plant task scheduling system belongs to the same concept as the aforementioned photovoltaic power plant task scheduling method. Details not described in detail in this example can be found in the description of the aforementioned photovoltaic power plant task scheduling method.

[0079] This embodiment also provides a photovoltaic power plant task scheduling system, including: The data acquisition module is used to collect device location data of equipment and edge servers within the photovoltaic power plant; The network construction module is used to construct a communication topology network for photovoltaic power plant equipment based on the device location data. The model building module is used to build a task scheduling model, a communication transmission model, and a computational processing model based on the communication topology network of the photovoltaic power plant equipment. The optimization problem construction module is used to construct a multi-objective optimization problem by combining the task scheduling model, communication transmission model and computation processing model. The problem transformation module is used to introduce queue stability constraints into the multi-objective optimization problem and transform it into a deterministic optimization problem; The solution module is used to solve the deterministic optimization problem using a multi-objective evolutionary algorithm and output the Pareto front optimal solution set; The decision generation module is used to generate task unloading decisions and resource allocation strategies for edge servers based on the Pareto front optimal solution set and the weight evaluation of the optimization objective.

[0080] This embodiment also provides an electronic device suitable for photovoltaic power plant task scheduling, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the photovoltaic power plant task scheduling method proposed in the above embodiment.

[0081] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the photovoltaic power plant task scheduling method proposed in the above embodiments.

[0082] The storage medium proposed in this embodiment and the photovoltaic electric field task scheduling method proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0083] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A photovoltaic power plant task scheduling method, characterized in that, include: Collect equipment location data of the photovoltaic power plant, and construct a communication topology network for the photovoltaic power plant equipment based on the equipment location data; Based on the communication topology network of the photovoltaic power plant equipment, a task scheduling model, a communication transmission model, and a computational processing model are constructed. Based on the task scheduling model, communication transmission model, and computation processing model, a multi-objective optimization problem is constructed. The multi-objective optimization problem is then subjected to queue stability constraints, transforming the constrained multi-objective optimization problem into a deterministic optimization problem. A multi-objective evolutionary algorithm is used to solve the deterministic optimization problem to obtain the Pareto front optimal solution set. Based on a comprehensive trade-off of the optimization objectives, task offloading decisions and resource allocation strategies for edge servers are obtained from the Pareto front optimal solution set.

2. The photovoltaic power plant task scheduling method as described in claim 1, characterized in that, The steps for constructing a communication topology network for photovoltaic power plant equipment include: Collect device location data for various devices within the photovoltaic power plant and in the edge server; Based on the device location data, a communication topology network for photovoltaic power plant equipment is constructed.

3. The photovoltaic power plant task scheduling method as described in claim 2, characterized in that, The steps for constructing the task scheduling model, communication transmission model, and computation processing model include: The communication scheduling cycle in the photovoltaic power plant equipment communication topology network is divided into multiple time slots of equal length; task data is generated for each device in the time slot; an initial data cache queue is established for each device to cache the task data; and a task scheduling model is obtained by setting a queue upper limit constraint on the initial data cache queue. Based on the communication topology network of the photovoltaic power plant equipment, a communication transmission model for the task data from the equipment to the edge server is constructed; The task scheduling model is used to process the unloading task data on the edge server and construct a computational processing model.

4. The photovoltaic power plant task scheduling method as described in claim 3, characterized in that, The steps to construct a multi-objective optimization problem include: The latency performance index is obtained through the communication transmission model and the computational processing model, the load balancing performance index is obtained through the computational processing model, and the optimization target is obtained by combining the latency performance index and the load balancing performance index. Optimization decisions are obtained through task offloading in the task scheduling model, transmission links in the communication transmission model, and resource allocation in the computing processing model. The constraints are obtained by combining the queue upper limit constraint in the task scheduling model, the access constraint in the communication transmission model, and the capacity constraint in the computation processing model. Based on the optimization objective, optimization decisions, and constraints, a multi-objective optimization problem is constructed.

5. The photovoltaic power plant task scheduling method as described in claim 4, characterized in that, The steps to transform a problem into a deterministic optimization problem include: Based on the queue upper limit constraint in the task scheduling model, and combined with the task data, the initial data cache queue is modified to obtain the corrected data cache queue. Based on the scalar function defined in the modified data cache queue, and combined with the optimization objective of the multi-objective optimization problem, a drift penalty function is constructed and its upper limit is derived to form the drift penalty analysis results; By minimizing the drift penalty analysis results, the multi-objective optimization problem is transformed into a deterministic optimization problem without long-term queue constraints.

6. The photovoltaic power plant task scheduling method as described in claim 5, characterized in that, The steps to obtain the optimal solution set of the Pareto front include: Based on the deterministic optimization problem, the initial solution of the multi-objective evolutionary algorithm is obtained by encoding the task offloading, transmission link and resource allocation in the optimization decision. The running parameters of the multi-objective evolutionary algorithm are configured according to the initial solution to generate an initial population. The initial population is selected by combining the constraints to obtain a compliant population; the compliant population is iteratively updated according to the optimization objective to obtain an iterative population; A maximum number of iterations is preset. When the number of iterations of the iterative population reaches the maximum number of iterations, the Pareto front optimal solution set is output.

7. The photovoltaic power plant task scheduling method as described in claim 6, characterized in that, The steps to obtain task offloading decisions and edge server resource allocation policies include: The weights of latency performance indicators and load balancing performance indicators in the optimization objective are preset, and a comprehensive cost evaluation is performed on the Pareto front optimal solution set to calculate the comprehensive cost value. The solution with the optimal overall cost value is selected as the target solution, and the task unloading decision and the resource allocation strategy for the edge server are obtained from the target solution.

8. A photovoltaic power plant task scheduling system, using the method described in any one of claims 1-7, characterized in that, include: The data acquisition module is used to collect device location data of equipment and edge servers within the photovoltaic power plant; The network construction module is used to construct a communication topology network for photovoltaic power plant equipment based on the device location data. The model building module is used to build a task scheduling model, a communication transmission model, and a computational processing model based on the communication topology network of the photovoltaic power plant equipment. The optimization problem construction module is used to construct a multi-objective optimization problem by combining the task scheduling model, communication transmission model and computation processing model. The problem transformation module is used to introduce queue stability constraints into the multi-objective optimization problem and transform it into a deterministic optimization problem; The solution module is used to solve the deterministic optimization problem using a multi-objective evolutionary algorithm and output the Pareto front optimal solution set; The decision generation module is used to generate task unloading decisions and resource allocation strategies for edge servers based on the Pareto front optimal solution set and the weight evaluation of the optimization objective.

9. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the photovoltaic power plant task scheduling method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the photovoltaic power plant task scheduling method according to any one of claims 1 to 7.

Images