A cross-domain computing power spatiotemporal coordination optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential

Abstract

The application discloses a cross-domain computing power space-time collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential, relates to the cross field of power system operation optimization and computing power collaborative scheduling, and calculates the carbon emission increment and the dynamic marginal carbon potential of the system by applying unit load disturbance to the bus of the data center. The dynamic marginal carbon potential is used as a low-carbon guiding signal for the space-time migration of computing power tasks. The cross-domain mapping relationship between the data center, the computing power task, the power grid bus, the generator unit and the carbon emission factor is established. The unified mapping relationship between the computing power network and the power system is realized. The comprehensive cost constraints including the power cost, the marginal carbon cost, the cross-domain transmission cost and the service quality penalty cost are considered for the cross-domain computing power optimization scheduling. The method has the effects of realizing the collaborative migration of the computing power task in the time and space dimensions, reducing the system operation cost and carbon emission, and meeting the task time delay, data center capacity and power grid safe operation constraints.

Description

Technical Field

[0001] This invention relates to the technical field of power system operation optimization and computing power collaborative scheduling, and in particular to a cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential. Background Technology

[0002] Currently, with the rapid development of cloud computing, artificial intelligence, big data, and the digital economy, the electricity load of data centers and computing power hubs continues to grow. Large data centers are usually distributed in different geographical regions, and there are significant differences in the power supply structure, electricity price level, new energy output characteristics, and grid operation status of different regions. For example, some regions have abundant wind power, photovoltaic, or hydropower resources and low carbon emission intensity of electricity, while some regions with concentrated power loads have large demands for electricity and computing power, especially with a high proportion of fossil energy units and relatively high carbon emission intensity of electricity. Therefore, scheduling computing power in different regions to balance the electricity and computing power demands of different data centers has become the main direction of current grid operation optimization.

[0003] Existing data center computing power scheduling methods typically use regional average carbon emission factors or static carbon emission coefficients, which cannot reflect changes in electricity consumption and carbon emissions over different time periods. Green computing power scheduling methods only focus on the proportion of local renewable energy sources or local electricity prices, ignoring the impact of data center load changes on the entire power grid, resulting in biased carbon emission analysis. Data center electricity load and computing power demand are mutually influential and dynamically changing. Separating the calculation of power grid carbon emissions or the optimization of computing power scheduling ignores the influence of the other, leading to a discrepancy between the final computing power scheduling results and the actual operating status of the power grid. Furthermore, existing computing power scheduling methods do not consider the impact of business types, making it difficult to accurately reflect the marginal impact of data center load changes on power grid carbon emissions over different time periods. Summary of the Invention

[0004] To address the problem that existing technologies do not comprehensively consider the influencing factors of data center computing power scheduling and cannot accurately reflect the marginal impact of data center load changes on power grid carbon emissions at different time periods, this application provides a cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential. This method can establish a mapping relationship between data center computing power load and power grid nodes, calculate the dynamic marginal carbon potential of different time periods and different nodes using optimal power flow disturbance, and use it as a low-carbon guiding signal for cross-domain computing power task scheduling. This enables collaborative migration of computing power tasks in time and space dimensions, reduces system operating costs and carbon emissions, and simultaneously meets the constraints of task latency, data center capacity, and power grid safe operation.

[0005] Firstly, the above-mentioned inventive objective of this application is achieved through the following technical solution: A cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential, the method comprising: Obtain data processing task information from data centers in various regions of the power grid, classify the data processing task information into tasks, establish time delay constraints based on the classification results, and calculate the computing power task execution volume under the time delay constraints. Obtain data center operation information for each data center, establish a mapping relationship between the data center and the power grid topology nodes in the region based on the power grid topology, and convert the computing power task execution volume into total power consumption; Based on the total electricity consumption, the system's basic carbon emissions and the system's carbon emissions after disturbance are calculated. The power grid carbon flow source tracing and dynamic marginal carbon potential calculation are performed, and the source carbon emissions of the carbon flow source tracing path are calculated according to the dynamic marginal carbon potential. The marginal carbon cost of the corresponding data processing task is calculated based on the carbon emission traceability, and the cross-domain computing power scheduling of the data processing task is optimized based on the marginal carbon cost to obtain the cross-domain computing power optimization scheduling decision.

[0006] In a preferred embodiment, this application can be further configured as follows: obtaining data processing task information from data centers in various regions of the power grid, classifying the data processing task information into tasks and establishing latency constraints based on the classification results, and calculating the computing power task execution volume under the latency constraints, specifically includes: Obtain task request information from data centers in various regions of the power grid, and classify and process the tasks of each data center based on the task request information. The task types include sensitive tasks and tolerant tasks. Based on the task classification results, latency constraints are applied to the processing tasks of each data center and cached to the corresponding message queue to obtain the task processing priority order of each data processing center. Calculate the computing power task execution volume of the data center processing tasks, and generate the computing power task execution order of the data processing center according to the task processing priority order.

[0007] In a preferred embodiment, this application may be further configured as follows: the step of calculating the computing power task execution volume of the data center processing tasks and generating the computing power task execution order of the data processing center according to the task processing priority order further includes: To build a dynamic task balance relationship, we obtain the task reception volume, task execution volume, task migration volume, and task backlog volume of each data center in each time period. The data center computing resources are dynamically allocated according to the execution order of the computing tasks and the dynamic balance relationship of the tasks for the corresponding time period.

[0008] In a preferred embodiment, this application can be further configured as follows: obtaining the data center operation information of each data center, establishing a mapping relationship between the data center and the power grid topology nodes in the region based on the power grid topology, and converting the computing power task execution volume into total power consumption, specifically includes: Obtain data center operation information for each data center and establish a mapping relationship between the data center and the power grid topology nodes in the region, based on the power grid topology structure. Calculate the computing power task execution volume of the data center, and superimpose it onto the power grid bus according to the mapping relationship to generate the total power consumption of the data center; The formula for calculating the total electricity consumption is as follows: in, Indicates the first Data centers during the time period Total electricity consumption Indicates the amount of task execution. Indicates the first Data centers during the time period The power consumption coefficient per unit task, Indicates the first Energy efficiency of a data center.

[0009] In a preferred embodiment, this application can be further configured as follows: the process of calculating the system's basic carbon emissions based on the total electricity consumption and the system's carbon emissions after disturbance, and calculating the source carbon emissions of the carbon flow tracing path according to the dynamic marginal carbon potential, specifically includes: Using a pre-constructed optimal power flow model, the system's basic carbon emissions for the corresponding grid nodes are calculated based on the total electricity consumption. The expression for the system's basic carbon emissions is as follows: in, express System baseline carbon emissions for a given period Indicates the first Data center units During the period Total electricity consumption Indicates the number of data center units. Indicates the time of electricity consumption. Indicates the first Data center units The average carbon potential; The data center access bus is subjected to load disturbance, and the optimal power flow model is solved again based on the load disturbance results to redistribute the unit output, thereby obtaining the total carbon emissions of the system after the disturbance. The calculation expression for the total carbon emissions of the system after the disturbance is as follows: in, This represents the total carbon emissions of the system after the disturbance. This indicates the result of the redistribution of unit output; Based on the optimal power flow model, the unit output for the basic carbon emissions of the system and the unit output for the total carbon emissions of the system after the disturbance are calculated to determine the change in unit output after the disturbance, and the contribution ratio of the unit output of the data center is analyzed. Based on the mapping relationship and the contribution ratio of unit output, the carbon flow path of unit output change after the disturbance is traced in reverse to construct a complete carbon flow path between the target load node and the data center access node. Calculate the dynamic marginal carbon potential of the complete carbon flow path after disturbance, and calculate the source carbon emissions based on the dynamic marginal carbon potential; The calculation expression for the dynamic marginal carbon potential is as follows: in, Indicates the first Data centers during the time period The dynamic marginal carbon potential, Indicates the amount of load disturbance; The calculation expression for the traceable carbon emissions is as follows: in, Indicates the carbon emissions traced back to their source. This indicates the percentage of power output contributed by the generating unit. Indicates changes in unit output. This indicates the change in the marginal carbon potential of the unit. This represents the power loss or energy dissipation along the complete carbon flow path. This represents the path-average dynamic marginal carbon potential.

[0010] In a preferred embodiment, this application can be further configured as follows: calculating the marginal carbon cost of the corresponding data processing task based on the traced carbon emissions, and optimizing the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost to obtain a cross-domain computing power optimization scheduling decision, specifically including: The formula for calculating the marginal carbon cost is as follows: in, Indicates marginal carbon cost, Indicates the first Data centers during the time period The source of carbon emissions, Indicates the first Data centers during the time period Total electricity consumption Indicates the time period Carbon cost weighting; The power cost, cross-domain transmission cost, and service quality penalty cost of cross-domain computing power scheduling for the data processing task are obtained, and a comprehensive cost constraint is applied in combination with the marginal carbon cost. The expression for the comprehensive cost constraint is as follows: in, Indicates the target total cost. Indicates electricity cost, Indicates the cost of cross-domain transmission. This indicates the cost of penalties for poor service quality. Under the constraint of dynamic task balancing, the computing power is dynamically adjusted according to the comprehensive cost constraint results for the task execution volume, cross-domain migration task volume, and task backlog in the data center, resulting in cross-domain computing power optimization scheduling decisions.

[0011] In a preferred embodiment, this application can be further configured as follows: the step of obtaining the power cost, cross-domain transmission cost, and service quality penalty cost for cross-domain computing power scheduling of the data processing task, combined with the marginal carbon cost for comprehensive cost constraint, also includes: The formula for calculating the electricity cost is as follows: in, Indicates the first Data centers during the time period Electricity price, Indicates the first Data centers during the time period Total electricity consumption.

[0012] In a preferred embodiment, this application can be further configured as follows: the step of calculating the marginal carbon cost of the corresponding data processing task based on the dynamic marginal carbon potential, and optimizing the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost to obtain a cross-domain computing power optimization scheduling decision, further includes: The updated power load after the data center executes the cross-domain computing power optimization scheduling decision is obtained, and the dynamic marginal carbon potential for the next time period is calculated based on the pre-built optimal power flow model and dynamically updated to obtain the updated dynamic marginal carbon potential. Based on the updated dynamic carbon potential, the cross-domain computing power optimization scheduling decision is iteratively optimized. When the preset iterative convergence condition is met, the computing power scheduling of the data center and the closed-loop feedback data of the power grid carbon potential are obtained.

[0013] Secondly, the above-mentioned inventive objective of this application is achieved through the following technical solutions: A cross-domain computing power spatiotemporal collaborative optimization scheduling system based on power grid carbon flow tracing and dynamic marginal carbon potential, the system comprising: The data acquisition and task classification module is used to acquire data processing task information from data centers in various regions of the power grid, classify the data processing task information into tasks, establish time delay constraints based on the classification results, and calculate the computing power task execution volume under the time delay constraints. The computing power and power grid node mapping module is used to obtain the data center operation information of each data center, establish the mapping relationship between the data center and the power grid topology nodes in the area in combination with the power grid topology structure, and convert the computing power task execution volume into total power consumption. The dynamic marginal carbon potential calculation module is used to calculate the system's basic carbon emissions and the system's carbon emissions after disturbance based on the total electricity consumption, and to perform grid carbon flow tracing and dynamic marginal carbon potential calculation processing, and to calculate the source carbon emissions of the carbon flow tracing path according to the dynamic marginal carbon potential. The computing power collaborative optimization module is used to calculate the marginal carbon cost of the corresponding data processing task based on the traceability carbon emission, and to perform cost constraint optimization on the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost, so as to obtain the cross-domain computing power optimization scheduling decision.

[0014] Thirdly, the above-mentioned objectives of this application are achieved through the following technical solutions: A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential.

[0015] In summary, this application includes at least one of the following beneficial technical effects: 1. This application calculates the incremental carbon emissions of the system by applying a unit load disturbance to the data center access bus, and then analyzes the dynamic marginal carbon potential. The dynamic marginal carbon potential is used as a low-carbon guidance signal for the spatiotemporal migration of computing power tasks. Compared with the traditional average carbon emission factor or static carbon factor, it can more accurately reflect the actual marginal impact of computing power load changes on grid carbon emissions, and improve the accuracy of carbon emission assessment. Furthermore, by establishing a cross-domain mapping relationship between data centers, computing power tasks, grid buses, generator sets and carbon emission factors, a unified mapping relationship between computing power networks and power systems is realized. By simultaneously considering the comprehensive cost constraints including electricity costs, marginal carbon costs, cross-domain transmission costs and service quality penalty costs, it is possible to reduce carbon emissions while taking into account economic efficiency and service quality, and realize cross-domain computing power optimization scheduling. 2. This application incorporates business types into the scope of computing power scheduling by classifying computing power tasks and imposing latency constraints on each task. Based on the spatiotemporal changes in the output of new energy sources such as wind power, photovoltaics, and hydropower, it guides tolerable tasks to migrate between different regions and time periods, thereby improving the level of new energy consumption and reducing system carbon emissions. 3. This application uses the execution results of cross-domain computing power optimization scheduling decisions for iterative optimization, avoiding the problem of carbon potential information lag after computing power load adjustment, so that the optimized scheduling results are consistent with the grid operation status, and improving the stability and reliability of the scheduling results; 4. This application improves the robustness of cross-domain computing power scheduling in complex power grid operation scenarios by setting up mechanisms such as OPF degradation calculation, outlier limiting and relaxation iteration. It is applicable to low-carbon scheduling scenarios in multi-regional data centers, computing power hubs, cloud computing platforms, edge computing nodes and power market environments, and has good scalability. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. The elements or parts in the drawings are not necessarily drawn to scale.

[0017] Figure 1 This is a flowchart illustrating the implementation of the cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential in this embodiment.

[0018] Figure 2 This is a comparison diagram showing the effect of computing power optimization scheduling before and after simulation in the Beijing computing power hub in this embodiment.

[0019] Figure 3 This is a comparison diagram showing the effect of computing power optimization scheduling before and after simulation in the Yangtze River Delta computing power hub in this embodiment.

[0020] Figure 4 This is a comparison diagram of the power consumption of each hub simulation before and after optimization in this embodiment.

[0021] Figure 5 This is a graph showing the change in the computing power scheduling objective function of this embodiment.

[0022] Figure 6 This is a schematic diagram of the marginal carbon potential iterative convergence curve in this embodiment.

[0023] Figure 7 This is a dynamic marginal carbon potential comparison curve of the power grid nodes where data centers in different regions are located in this embodiment.

[0024] Figure 8 This is a comparison diagram of the global marginal carbon potential emission reduction and economic benefits in this embodiment.

[0025] Figure 9 This is a schematic diagram of the framework of the cross-domain computing power spatiotemporal collaborative optimization scheduling system in this embodiment.

[0026] Figure 10 This is a schematic diagram of the internal structure of a computer device used to implement a cross-domain computing power spatiotemporal collaborative optimization scheduling method. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] It should be understood that, when used in this specification, the terms “comprising” and “including” indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0029] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification, unless the context clearly indicates otherwise, the singular forms “a,” “an,” and “the” are intended to include the plural forms.

[0030] It should also be further understood that the term "and / or" as used in this specification refers to any combination of one or more of the associated listed items, as well as all possible combinations, and includes such combinations.

[0031] In one embodiment, such as Figure 1 As shown, this application discloses a cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential, which specifically includes the following steps: S1: Obtain data processing task information from data centers in various regions of the power grid, classify the data processing task information into tasks, establish latency constraints based on the classification results, and calculate the amount of computing power task execution under the latency constraints.

[0032] Specifically, step S1 includes: S11: Obtain task request information from data centers in various regions of the power grid, and classify and process the tasks of each data center based on the task request information. The task types include sensitive tasks and tolerant tasks.

[0033] Specifically, within each scheduling cycle, task request information, i.e., data processing tasks, is obtained from data processing centers in various regions of the power grid. The task request information includes task type and task quantity. Based on the task request information, the processing tasks received by each data center are classified. Tasks that are sensitive to latency and have high real-time requirements are classified as sensitive tasks, while tasks that can be delayed or migrated across regions are marked as tolerable tasks. If the task request is online interaction, real-time inference, financial trading, industrial control, etc., which need to be executed in the local or nearby data center first, it is classified as a sensitive task. If the task request is offline training, batch processing, data backup, scientific computing, video transcoding, etc., which allow buffering, have low real-time requirements, or can be scheduled across regions and time periods based on dynamic marginal carbon potential, electricity price, and renewable energy output, it is classified as a tolerable task.

[0034] S12: Based on the task classification results, apply latency constraints to the processing tasks of each data center and cache them to the corresponding message queue to obtain the task processing priority order of each data processing center.

[0035] Specifically, based on the task classification results, tasks processed by each data center are cached in their corresponding message queues and matched with corresponding latency constraints. For example, sensitive tasks are subject to maximum migration ratio or maximum transmission latency constraints, while tolerant tasks are subject to maximum waiting time constraints. Task processing priority is generated according to the matched latency constraints and the message queue in which they reside. This embodiment also sets a period-end queue clearing constraint for each message queue. That is, when the waiting time of tasks at the end of the queue exceeds the current scheduling period, the excess tasks are migrated to other data centers across domains, ensuring that all pending tasks are cleared from the message queue at the end of the scheduling period, thus avoiding indefinite task delays.

[0036] S13: Calculate the computing power task execution volume of the data center processing tasks, and generate the computing power task execution order of the data processing center according to the task processing priority.

[0037] Specifically, the computing power required for data center tasks is the amount of computing resources needed to process a task. The computing power is calculated using OPF (Optimal Power Flow) and the computing power is then degraded according to the priority of task processing to obtain the execution order of computing power tasks.

[0038] In this embodiment, an optimal power flow model (OPF model) is constructed with the goal of minimizing the total power generation cost or the total carbon emissions of the system. The expression of the optimal power flow model is as follows: in, This represents the target total operating cost of the optimal power flow model. Indicates the active power output of the generator Traditional power generation costs This indicates the electricity cost of the data center. For time period The corresponding time-of-use electricity price, This indicates the period of load affected by the degradation strategy. Indicates the degradation factor, used to refer to the time period. The amount of tasks executed.

[0039] S14: Obtain the task reception volume, task execution volume, task migration volume, and task backlog volume of each data center in each time period to build a dynamic task balance relationship.

[0040] Specifically, the task reception volume, task execution volume, task migration volume, and task backlog volume of each data center in each time period, i.e., a complete scheduling cycle, are obtained by counting the number of tasks in the corresponding message queues. A computing power allocation weight is set for each message queue, and the sum of the product of the computing power required for each task in each message queue and the corresponding computing power allocation weight is equal to the real-time available computing power resources of the data center. Thus, the dynamic balance relationship of tasks in all message queues is obtained under the constraint of the real-time available computing power resources of the data center.

[0041] The joint recursive relationship between the task reception volume, task execution volume, task migration volume, and task backlog volume in this embodiment is as follows: in, Indicates the current time period The backlog of tasks corresponds to the length of the message queue. Indicates the previous time period The backlog of tasks Indicates the current time period The number of tasks received. , These represent the number of tasks migrated in from other data centers or nodes and the number of tasks migrated out, respectively. This indicates the amount of task execution, where, and , This indicates the maximum cache capacity. Exceeding the maximum cache capacity can easily lead to packet loss. This indicates the maximum computing power of the data center. This represents the backlog threshold that triggers migration; in this embodiment, it is set to the maximum computing power. 80%, Indicates the migration ratio coefficient. This indicates the proportion of overflow tasks that are reselected for migration.

[0042] S15: Dynamically allocate data center computing resources for the corresponding time period according to the execution order of computing tasks and the dynamic balance relationship of tasks.

[0043] Specifically, the computing resources of the data center are dynamically allocated within the corresponding scheduling cycle according to the execution order of computing tasks and the dynamic balance of tasks, and the available computing resources of the data center are matched to the corresponding tasks proportionally and sequentially.

[0044] S2: Obtain the data center operation information of each data center, establish the mapping relationship between the data center and the power grid topology nodes in the area, and convert the computing power task execution volume into total power consumption.

[0045] Specifically, step S2 includes: S21: Obtain the data center operation information of each data center and establish a mapping relationship between the data center and the power grid topology nodes in the region, based on the power grid topology structure.

[0046] Specifically, the system acquires operational information for each data center, including maximum computing capacity, current load rate, PUE, and unit task power consumption parameters, as well as the region where the data center is located and the grid bus number it connects to. This information also includes grid topology, line parameters, node load, generator output limits, ramp constraints, generation cost parameters, carbon emission factors for each generator unit, renewable energy output forecast data, time-of-use pricing, carbon prices, cross-domain data transmission costs, network latency parameters, and scheduling cycle duration. A mapping relationship is then established between the data center and the grid topology nodes in the region where it is located, based on the grid topology of the area.

[0047] S22: Calculate the computing power task execution volume of the data center and superimpose it onto the power grid bus according to the mapping relationship to generate the total power consumption of the data center.

[0048] The formula for calculating total electricity consumption is as follows: in, Indicates the first Data centers during the time period Total electricity consumption Indicates the amount of task execution. Indicates the first Data centers during the time period The power consumption coefficient per unit task, Indicates the first Energy efficiency of a data center.

[0049] In this embodiment, the amount of computing power required for task execution refers to the computing power resources that the data center needs to schedule to process tasks.

[0050] S3: Based on the total electricity consumption, calculate the system's basic carbon emissions and the system's carbon emissions after disturbance, perform grid carbon flow tracing and dynamic marginal carbon potential calculation, and calculate the source carbon emissions of the carbon flow tracing path according to the dynamic marginal carbon potential.

[0051] Specifically, step S3 includes: S31: Using a pre-built optimal power flow model, calculate the system's basic carbon emissions for the corresponding grid nodes based on the total electricity consumption. The expression for the system's basic carbon emissions is as follows: in, express System baseline carbon emissions for a given period Indicates the first Data center units During the period Total electricity consumption Indicates the number of data center units. Indicates the time of electricity consumption. Indicates the first Data center units The average carbon potential; S32: Perform load disturbance on the data center access bus and, based on the load disturbance results, solve the optimal power flow model again to redistribute unit output, obtaining the total carbon emissions of the system after the disturbance. The calculation expression for the total carbon emissions of the system after the disturbance is as follows: in, This represents the total carbon emissions of the system after the disturbance. This indicates the result of the redistribution of unit output; In this embodiment, a preset unit load disturbance is applied to the data center access bus, and the unit output is redistributed based on the load change after the unit load disturbance using the optimal power flow model. The total carbon emissions of the system after the disturbance are calculated based on the redistribution results of the output of each unit after the disturbance.

[0052] S33: Based on the optimal power flow model, calculate the unit output for the basic carbon emissions of the system and the unit output for the total carbon emissions of the system after the disturbance to calculate the change in unit output after the disturbance, and analyze the contribution ratio of the data center's unit output. The expression for the change in unit output in this embodiment is as follows: in, This refers to the change in unit output after the disturbance. This represents the unit output corresponding to the total carbon emissions of the system after the disturbance. The unit output corresponding to the system's basic carbon emissions is obtained by solving the optimal power flow model. The expression for the unit's output contribution ratio in this embodiment is as follows: in, This indicates the percentage of power output contributed by the generating unit. This represents the K generating units with the largest positive output increase after the disturbance. In this embodiment, K is taken as 3-5 units. Indicates the first Changes in unit output after disturbance.

[0053] S34: According to the mapping relationship and the proportion of unit output contribution, the carbon flow path of the unit output change after the disturbance is traced in reverse to construct a complete carbon flow path between the target load node and the data center access node. Specifically, in this embodiment, a directed graph of the power grid is pre-constructed based on the power grid topology and current transmission direction. The power flow distribution of each branch on the directed graph of the power grid is calculated, and the back-tracking of the previous load node is performed according to the proportion of the unit output contribution as the outgoing line admittance ratio. The carbon emission of each confirmed back-tracking load node is calculated, and the carbon emission difference between the basic carbon emission of the system and the total carbon emission of the system after the disturbance is calculated. When the preset source tracing end condition is met, the target load node is confirmed, and the complete carbon flow path between the target load node and the data center access node is obtained.

[0054] The preset source tracing termination condition in this embodiment is: In this embodiment This represents the difference in carbon emissions between the system's basic carbon emissions and the system's total carbon emissions after the disturbance. This indicates the carbon emissions traced between the load nodes and the data center access nodes during the tracing process.

[0055] S35: Calculate the dynamic marginal carbon potential of the complete carbon flow path after disturbance, and calculate the source carbon emissions based on the dynamic marginal carbon potential; The calculation expression for the dynamic marginal carbon potential is as follows: in, Indicates the first Data centers during the time period The dynamic marginal carbon potential, Indicates the amount of load disturbance; The calculation expression for the traceable carbon emissions is as follows: in, Indicates the carbon emissions traced back to their source. This indicates the percentage of power output contributed by the generating unit. Indicates changes in unit output. This indicates the change in the marginal carbon potential of the unit. This represents the power loss or energy dissipation along the complete carbon flow path. This represents the path-average dynamic marginal carbon potential.

[0056] S4: Calculate the marginal carbon cost of the corresponding data processing task based on the traceable carbon emissions, and optimize the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost to obtain the cross-domain computing power optimization scheduling decision.

[0057] Specifically, step S4 includes: S41: Calculate the marginal carbon cost of the corresponding data processing task based on the dynamic marginal carbon potential. The calculation expression for the marginal carbon cost is as follows: in, Indicates marginal carbon cost, Indicates the first Data centers during the time period The source of carbon emissions, Indicates the first Data centers during the time period Total electricity consumption Indicates the time period Carbon cost weighting.

[0058] S42: Obtain the power cost, cross-domain transmission cost, and service quality penalty cost for cross-domain computing power scheduling of data processing tasks, and combine them with marginal carbon cost for comprehensive cost constraints.

[0059] The expression for the comprehensive cost constraint is as follows: in, This indicates that the overall objective cost is minimized. Indicates electricity cost, This refers to the cost of cross-domain transmission, which includes data transfer fees and network resource usage costs incurred during task migration between different regions. This refers to the service quality penalty cost, which is the penalty incurred for sensitive tasks being migrated across regions, for tolerable tasks being delayed, or for tasks not being completed on time.

[0060] The formula for calculating electricity costs is as follows: in, Indicates the first Data centers during the time period Electricity price, Indicates the first Data centers during the time period Total electricity consumption.

[0061] S43: Under the constraint of dynamic task balance, the computing power is dynamically adjusted according to the comprehensive cost constraint results for the task execution volume, cross-domain migration task volume, and task backlog of the data center, so as to obtain the cross-domain computing power optimization scheduling decision.

[0062] Specifically, based on the comprehensive cost constraints, the computing power is dynamically adjusted for the task execution volume, cross-domain migration task volume, and task backlog of each data center. This includes, but is not limited to, increasing the cross-domain migration task volume and task backlog when computing power is insufficient, and increasing the task execution volume and decreasing the cross-domain migration task volume and task backlog when computing power is sufficient, thus obtaining cross-domain computing power optimization scheduling decisions. These decisions include, but are not limited to, sensitive task scheduling volume, tolerant task scheduling volume, task migration volume between different data centers, queuing and latency processing volume of tolerant tasks, target power load of the data center, dynamic marginal carbon potential signals of each node, and low-carbon load response suggestions. The cross-domain computing power optimization scheduling decisions are used to obtain the task execution plan and cross-domain task migration plan for each data center in each time period.

[0063] This embodiment also includes: S5: Obtain the updated power load after the data center executes cross-domain computing power optimization scheduling decision, and dynamically update the dynamic marginal carbon potential for the next time period based on the pre-built optimal power flow model to obtain the updated dynamic marginal carbon potential.

[0064] Specifically, the data center executes cross-domain computing power optimization scheduling decisions within the current scheduling cycle, obtains the updated power load after the data center executes the cross-domain computing power optimization scheduling decisions, and substitutes it into the calculation formula of dynamic marginal carbon potential to obtain the updated dynamic marginal carbon potential.

[0065] S6: Based on the updated dynamic carbon potential, the cross-domain computing power optimization scheduling decision is iteratively optimized. When the preset iterative convergence condition is met, the computing power scheduling of the data center and the closed-loop feedback data of the power grid carbon potential are obtained.

[0066] Specifically, the cross-domain computing power optimization scheduling decision is iteratively optimized based on the updated dynamic carbon potential. In this embodiment, three levels of convergence conditions are set. When all three levels of convergence conditions are met simultaneously, it is determined that the optimal state of carbon potential and computing power synergy has been reached, and closed-loop feedback data is output. The first level of convergence condition is that the computing power scheduling decision variable reaches a steady state, and the expression is as follows: in, Data centers representing adjacent iteration rounds The weighted root mean square offset of the allocated computing load. Indicates the number of data centers. , They represent data centers In the Round iteration and the first In the iterative process, the unit output based on cross-domain computing power optimization scheduling decisions is equivalent to the computing power load. Indicates data center Basic unit output, The offset threshold is set to 1% in this embodiment. That is, when the weighted root mean square offset of the computing power load in adjacent iterations is greater than 1%, it is determined that the scheduling strategy is still in an oscillating state and has not found a steady-state optimal solution, and it needs to continue iterating. When it is less than or equal to the offset threshold, it is determined that the scheduling strategy is in a steady-state optimal solution.

[0067] It should be noted that if the change in computing load of the data center is less than the start-stop threshold of the output of a single unit, it is directly determined to be in a steady state and no further iteration is needed.

[0068] The second-level convergence condition is that the global cost reaches a preset economic benefit, as shown in the following expression: in, , They represent the first Round iteration and the first The total cost during the iteration includes the electricity costs, carbon emission costs, and network bandwidth costs of all data centers. This indicates the minimum economic benefit that is pre-set.

[0069] The third-level convergence condition is that the cross-domain carbon potential marginal elastic convergence meets the marginal sensitivity of computing power scheduling and is within the allowable rebound range, as shown in the following expression: in, , They represent the first Round iteration and the first Carbon potential sensitivity corresponding to cross-domain scheduling of computing power during iteration. The preset carbon potential sensitivity threshold is set to 0.02 in this embodiment. , They represent data centers receiver node In the The dynamic marginal carbon potential of the cycle iteration and the initial state. This represents the preset allowable rebound amplitude. The carbon potential sensitivity expression in this embodiment is as follows: in, For data centers In the The derivative of the dynamic marginal carbon potential in the round iteration, For data centers In the The output of the unit in the round of iteration is the derivative of the computing load.

[0070] In this embodiment, the absolute cutoff condition is considered met when the maximum number of iterations is reached and the grid carbon potential update time is exceeded. The iteration stops, and the computing power scheduling and grid carbon potential closed-loop feedback data of the data center from the previous iteration round are output. In this embodiment, the maximum number of iteration rounds is set to 20. When the maximum number of iteration rounds is reached, iteration stops, and the optimal solution with the lowest cost is selected as the computing power scheduling and grid carbon potential closed-loop feedback data.

[0071] In this embodiment, closed-loop feedback enables computing tasks to continuously migrate to regions and time periods with low marginal carbon potential, low electricity prices, and high renewable energy absorption capacity, i.e., the scheduling cycle, thereby achieving low-carbon operation of the power grid and efficient utilization of computing resources.

[0072] The computing power optimization scheduling in this embodiment also satisfies the following constraints: data center computing power capacity constraint, task traffic conservation constraint, sensitive task latency constraint, sensitive task maximum migration ratio constraint, tolerant task queue update constraint, queue clearing constraint at the end of the scheduling cycle constraint, cross-domain network bandwidth constraint, power grid optimal power flow safe operation constraint, new energy output constraint, and generator unit output upper and lower limit constraint.

[0073] In one simulation embodiment, such as Figure 2 , Figure 3 The image shows a comparison of the optimization results of computing power hubs in two different regions before and after optimization. Figure 4 The image shows a comparison of the power consumption of each computing hub before and after optimization. Figure 5 The figure shown is a graph illustrating the change in the objective function of computing power scheduling, namely, cost minimization. Figure 6 The figure shown is the iterative convergence curve of the marginal carbon potential, as follows: Figure 7 The figure shown is a curve comparing the dynamic marginal carbon potential of power grid nodes where data centers are located in different regions. Figure 8 The diagram shows a comparison of global marginal carbon potential emission reduction and economic benefits.

[0074] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0075] In one embodiment, a cross-domain computing power spatiotemporal collaborative optimization scheduling system based on grid carbon flow tracing and dynamic marginal carbon potential is provided. This system corresponds one-to-one with the cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential in the above embodiments. Figure 9 As shown, this cross-domain computing power spatiotemporal collaborative optimization scheduling system based on grid carbon flow tracing and dynamic marginal carbon potential includes a data acquisition and task classification module, a computing power and grid node mapping module, a dynamic marginal carbon potential calculation module, and a computing power collaborative optimization module. Detailed descriptions of each functional module are as follows: The data acquisition and task classification module is used to acquire data processing task information from data centers in various regions of the power grid, classify the data processing task information, establish time delay constraints based on the classification results, and calculate the computing power task execution volume under the time delay constraints.

[0076] The computing power and power grid node mapping module is used to obtain the data center operation information of each data center, establish the mapping relationship between the data center and the power grid topology nodes in the area, and convert the computing power task execution volume into total power consumption.

[0077] The dynamic marginal carbon potential calculation module is used to calculate the system's basic carbon emissions and the system's carbon emissions after disturbance based on the total electricity consumption, and to perform grid carbon flow tracing and dynamic marginal carbon potential calculation processing, and to calculate the source carbon emissions of the carbon flow tracing path according to the dynamic marginal carbon potential.

[0078] The computing power collaborative optimization module is used to calculate the marginal carbon cost of the corresponding data processing task based on the traceability carbon emission, and to perform cost constraint optimization on the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost, so as to obtain the cross-domain computing power optimization scheduling decision.

[0079] Specific limitations regarding the cross-domain computing power spatiotemporal collaborative optimization scheduling system based on grid carbon flow tracing and dynamic marginal carbon potential can be found in the limitations of the cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential mentioned above, and will not be repeated here. Each module in the aforementioned cross-domain computing power spatiotemporal collaborative optimization scheduling system based on grid carbon flow tracing and dynamic marginal carbon potential can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0080] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The database stores relevant data for cross-domain computing power spatiotemporal collaborative optimization. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it implements a cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential.

[0081] Those skilled in the art will recognize that the units of the various examples described in connection with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application of the technical solution and the constraints involved. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0082] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical functional division. In actual implementation, there may be other division methods, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored.

[0083] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0084] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0085] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention 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 or all of the technical features therein. 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 the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential, characterized in that, The method includes: Obtain data processing task information from data centers in various regions of the power grid, classify the data processing task information into tasks, establish time delay constraints based on the classification results, and calculate the computing power task execution volume under the time delay constraints. Obtain data center operation information for each data center, establish a mapping relationship between the data center and the power grid topology nodes in the region based on the power grid topology, and convert the computing power task execution volume into total power consumption; Based on the total electricity consumption, the system's basic carbon emissions and the system's carbon emissions after disturbance are calculated. The power grid carbon flow source tracing and dynamic marginal carbon potential calculation are performed, and the source carbon emissions of the carbon flow source tracing path are calculated according to the dynamic marginal carbon potential. The marginal carbon cost of the corresponding data processing task is calculated based on the carbon emission traceability, and the cross-domain computing power scheduling of the data processing task is optimized based on the marginal carbon cost to obtain the cross-domain computing power optimization scheduling decision.

2. The cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential as described in claim 1, is characterized in that, The process of acquiring data processing task information from data centers in various regions of the power grid, classifying the data processing task information into tasks, establishing latency constraints based on the classification results, and calculating the computing power task execution volume under the latency constraints specifically includes: Obtain task request information from data centers in various regions of the power grid, and classify and process the tasks of each data center based on the task request information. The task types include sensitive tasks and tolerant tasks. Based on the task classification results, latency constraints are applied to the processing tasks of each data center and cached to the corresponding message queue to obtain the task processing priority order of each data processing center. Calculate the computing power task execution volume of the data center processing tasks, and generate the computing power task execution order of the data processing center according to the task processing priority order.

3. The cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential as described in claim 2, is characterized in that, The step of calculating the computing power task execution volume of the data center processing tasks and generating the computing power task execution order of the data processing center according to the task processing priority order further includes: To build a dynamic task balance relationship, we obtain the task reception volume, task execution volume, task migration volume, and task backlog volume of each data center in each time period. The data center computing resources are dynamically allocated according to the execution order of the computing tasks and the dynamic balance relationship of the tasks for the corresponding time period.

4. The cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential as described in claim 1, characterized in that, The process of acquiring data center operation information for each data center, establishing a mapping relationship between the data center and the power grid topology nodes in the region based on the power grid topology, and converting the computing power task execution volume into total electricity consumption specifically includes: Obtain data center operation information for each data center and establish a mapping relationship between the data center and the power grid topology nodes in the region, based on the power grid topology structure. Calculate the computing power task execution volume of the data center, and superimpose it onto the power grid bus according to the mapping relationship to generate the total power consumption of the data center; The formula for calculating the total electricity consumption is as follows: in, Indicates the first Data centers during the time period Total electricity consumption Indicates the amount of task execution. Indicates the first Data centers during the time period The power consumption coefficient per unit task, Indicates the first Energy efficiency of a data center.

5. The cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential as described in claim 1, characterized in that, The process of calculating the system's basic carbon emissions based on the total electricity consumption and the system's carbon emissions after disturbance, and then performing grid carbon flow tracing and dynamic marginal carbon potential calculation, specifically includes calculating the source carbon emissions of the carbon flow tracing path according to the dynamic marginal carbon potential: Using a pre-constructed optimal power flow model, the system's basic carbon emissions for the corresponding grid nodes are calculated based on the total electricity consumption. The expression for the system's basic carbon emissions is as follows: in, express System baseline carbon emissions for a given period Indicates the first Data center units During the period Total electricity consumption Indicates the number of data center units. Indicates the time of electricity consumption. Indicates the first Data center units The average carbon potential; The data center access bus is subjected to load disturbance, and the optimal power flow model is solved again based on the load disturbance results to redistribute the unit output, thereby obtaining the total carbon emissions of the system after the disturbance. The calculation expression for the total carbon emissions of the system after the disturbance is as follows: in, This represents the total carbon emissions of the system after the disturbance. This indicates the result of the redistribution of unit output; Based on the optimal power flow model, the unit output for the basic carbon emissions of the system and the unit output for the total carbon emissions of the system after the disturbance are calculated to determine the change in unit output after the disturbance, and the contribution ratio of the unit output of the data center is analyzed. Based on the mapping relationship and the contribution ratio of unit output, the carbon flow path of unit output change after the disturbance is traced in reverse to construct a complete carbon flow path between the target load node and the data center access node. Calculate the dynamic marginal carbon potential of the complete carbon flow path after disturbance, and calculate the source carbon emissions based on the dynamic marginal carbon potential; The calculation expression for the dynamic marginal carbon potential is as follows: in, Indicates the first Data centers during the time period The dynamic marginal carbon potential, Indicates the amount of load disturbance; The calculation expression for the traceable carbon emissions is as follows: in, Indicates the carbon emissions traced back to their source. This indicates the percentage of power output contributed by the generating unit. Indicates changes in unit output. This indicates the change in the marginal carbon potential of the unit. This represents the power loss or energy dissipation along the complete carbon flow path. This represents the path-average dynamic marginal carbon potential.

6. The cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential as described in claim 1, characterized in that, The step of calculating the marginal carbon cost of the corresponding data processing task based on the traced carbon emissions, and optimizing the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost to obtain a cross-domain computing power optimization scheduling decision, specifically includes: The formula for calculating the marginal carbon cost is as follows: in, Indicates marginal carbon cost, Indicates the first Data centers during the time period The source of carbon emissions, Indicates the first Data centers during the time period Total electricity consumption Indicates the time period Carbon cost weighting; The power cost, cross-domain transmission cost, and service quality penalty cost of cross-domain computing power scheduling for the data processing task are obtained, and a comprehensive cost constraint is applied in combination with the marginal carbon cost. The expression for the comprehensive cost constraint is as follows: in, Indicates the target total cost. Indicates electricity cost, Indicates the cost of cross-domain transmission. This indicates the cost of penalties for poor service quality. Under the constraint of dynamic task balancing, the computing power is dynamically adjusted according to the comprehensive cost constraint results for the task execution volume, cross-domain migration task volume, and task backlog in the data center, resulting in cross-domain computing power optimization scheduling decisions.

7. The cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential as described in claim 6, is characterized in that, The step of obtaining the power cost, cross-domain transmission cost, and service quality penalty cost for cross-domain computing power scheduling of the data processing task, combined with the marginal carbon cost for comprehensive cost constraint, also includes: The formula for calculating the electricity cost is as follows: in, Indicates the first Data centers during the time period Electricity price, Indicates the first Data centers during the time period Total electricity consumption.

8. The cross-domain computing power spatiotemporal collaborative optimization scheduling method based on power grid carbon flow tracing and dynamic marginal carbon potential as described in claim 1, characterized in that, The step of calculating the marginal carbon cost of the corresponding data processing task based on the dynamic marginal carbon potential, and optimizing the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost to obtain a cross-domain computing power optimization scheduling decision, further includes: The updated power load after the data center executes the cross-domain computing power optimization scheduling decision is obtained, and the dynamic marginal carbon potential for the next time period is calculated based on the pre-built optimal power flow model and dynamically updated to obtain the updated dynamic marginal carbon potential. Based on the updated dynamic carbon potential, the cross-domain computing power optimization scheduling decision is iteratively optimized. When the preset iterative convergence condition is met, the computing power scheduling of the data center and the closed-loop feedback data of the power grid carbon potential are obtained.

9. A cross-domain computing power spatiotemporal collaborative optimization scheduling system based on power grid carbon flow tracing and dynamic marginal carbon potential, characterized in that, The system includes: The data acquisition and task classification module is used to acquire data processing task information from data centers in various regions of the power grid, classify the data processing task information into tasks, establish time delay constraints based on the classification results, and calculate the computing power task execution volume under the time delay constraints. The computing power and power grid node mapping module is used to obtain the data center operation information of each data center, establish the mapping relationship between the data center and the power grid topology nodes in the area in combination with the power grid topology structure, and convert the computing power task execution volume into total power consumption. The dynamic marginal carbon potential calculation module is used to calculate the system's basic carbon emissions and the system's carbon emissions after disturbance based on the total electricity consumption, and to perform grid carbon flow tracing and dynamic marginal carbon potential calculation processing, and to calculate the source carbon emissions of the carbon flow tracing path according to the dynamic marginal carbon potential. The computing power collaborative optimization module is used to calculate the marginal carbon cost of the corresponding data processing task based on the traceability carbon emission, and to perform cost constraint optimization on the cross-domain computing power scheduling of the data processing task based on the marginal carbon cost, so as to obtain the cross-domain computing power optimization scheduling decision.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the cross-domain computing power spatiotemporal collaborative optimization scheduling method based on grid carbon flow tracing and dynamic marginal carbon potential as described in any one of claims 1 to 8.

Images