Method and system for operation optimization of intelligent computing center based on load regulation and multi-energy supply coordination

Abstract

This invention proposes a method and system for optimizing the operation of intelligent computing centers based on load adjustment and multi-power supply coordination. The method first divides the computing tasks to be executed in the intelligent computing center into rigid and flexible tasks based on the adjustability of task load in the time and power dimensions. Then, based on the computing load flexibility index of each flexible task, adjustable flexible tasks are selected. Finally, considering both rigid and flexible tasks, and combining the flexible adjustment mode of the adjustable flexible tasks with the multi-power supply mode of the intelligent computing center, an optimized operation model for the intelligent computing center is constructed with the goal of minimizing overall operating costs, carbon emissions, and load fluctuations. Solving this model generates the optimal low-carbon operation strategy for the intelligent computing center. This invention explicitly models the task timing and power adjustability as flexible computing load resources and coordinates them with a multi-mode power supply structure for optimization. This achieves adaptive matching of the computing load and multi-power supply structure of the intelligent computing center under low-carbon constraints, improving the low-carbon nature, economy, and power supply reliability of the intelligent computing center operation.

Description

Technical Field

[0001] This invention belongs to the field of power, specifically relating to a method and system for optimizing the operation of intelligent computing centers based on load regulation and multi-energy supply coordination. Background Technology

[0002] With the rapid development of next-generation information technologies such as artificial intelligence, big data, and cloud computing, intelligent computing centers, as crucial infrastructure supporting computing power services, have seen continuous growth in their construction scale and operational load, becoming a significant energy consumer in the power system. While providing high-performance computing services, intelligent computing centers are characterized by large-scale electricity consumption, concentrated load, and strong operational continuity, leading to increasingly prominent carbon emissions from electricity generation. Therefore, how to achieve low-carbon and efficient operation of intelligent computing centers while ensuring the quality of computing power services has become a key issue of concern in the industry.

[0003] In actual operation, intelligent computing centers typically connect to multiple energy sources simultaneously, including market electricity, green electricity, and self-generated renewable energy. However, these different energy sources differ significantly in terms of supply capacity, operating costs, and carbon emission characteristics, leading to complex energy selection processes and difficulties in balancing low-carbon and economic goals. Furthermore, the computing tasks within intelligent computing centers exhibit significant differences in operational characteristics, with varying adjustability in execution timing, power consumption, and continuity. Current operational strategies for intelligent computing centers often prioritize computing performance or single energy efficiency indicators, treating computing load as a rigid load for unified management. This approach fails to fully exploit the adjustment potential at the task level and lacks a systematic method for coordinating computing load adjustment capabilities with multiple energy sources. Consequently, it is difficult to achieve a synergistic reduction in operating costs and carbon emissions while maintaining service quality. Therefore, comprehensively considering the flexibility of the internal computing load and the differences in external energy sources to optimize the coordinated operation of intelligent computing centers has become a key research and practical issue. Summary of the Invention

[0004] The purpose of this invention is to address the aforementioned problems in the existing technology by providing a method and system for optimizing the operation of intelligent computing centers based on load regulation and multi-power supply coordination.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] In a first aspect, this invention proposes an intelligent computing center operation optimization method based on load regulation and multi-power supply coordination, including:

[0007] S1. Based on the adjustability of task load in the time and power dimensions, the computing power tasks to be executed in the intelligent computing center are divided into rigid tasks and flexible tasks.

[0008] S2. Select adjustable flexible tasks based on the computing load flexibility index of each flexible task.

[0009] S3. Simultaneously considering rigid and flexible tasks, and combining the flexible adjustment mode of adjustable flexible tasks with the multi-power supply mode of the intelligent computing center, with the goal of minimizing comprehensive operating costs, carbon emissions, and load fluctuations, construct an optimized operation model for the intelligent computing center, solve the model, and generate the low-carbon optimal operation strategy for the intelligent computing center. The strategy includes the execution sequence and execution power of adjustable flexible tasks, as well as the power supply adjustment mode of the intelligent computing center.

[0010] S2 includes:

[0011] S21. Calculate the computing load flexibility index for each flexible task using the following formula:

[0012] ;

[0013] ;

[0014] ;

[0015] ;

[0016] ;

[0017] In the above formula, For the task The computing load flexibility index, , , These are the weights for the time dimension, the power dimension, and the spatial dimension, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters , Tasks The earliest allowed start time and the latest allowed completion time, To complete the task Standard execution time For the task Execution continuity type variables, Indicates a task that is executed continuously. This indicates an interruptible task. For the task exist The total execution time has been accumulated. For the task The computational workload, For the task The computational workload coefficient corresponding to unit power For the task Rated operating power, , Tasks The minimum executable power and the maximum executable power, A rigid set of tasks. For a set of flexible tasks, For the task The flag indicating whether cross-node migration is allowed. For the task The set of computing power nodes allowed to execute. For computing power nodes exist The carbon emission intensity of electricity at any given moment;

[0018] S22. Determine whether each flexible task meets the requirements. If satisfied, then the task... Added to the set of adjustable flexible tasks as an adjustable flexible task. ,in, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively.

[0019] In S3, the objective function of the intelligent computing center's optimized operation model includes:

[0020] ;

[0021] ;

[0022] ;

[0023] In the above formula, The objective function of the intelligent computing center's operating strategy is... This is the cost weighting coefficient. for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This is the carbon emission weighting coefficient. For computing power nodes exist Carbon emission intensity of electricity at any given time For computing power nodes exist Electricity consumption at all times This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle, including both rigid and flexible tasks. For the task exist Power at any moment For time step, For the intelligent computing center New energy power that is constantly generated and used by the user. For the task computing nodes exist Operating power at any given time.

[0024] In S3, the constraints of the intelligent computing center's optimized operation model include continuous execution flexible task constraints, interruptible execution flexible task constraints, task completion constraints, intelligent computing center power limit constraints, budget cost constraints, and flexible adjustment task count constraints.

[0025] The constraints of the continuously executed flexible task include task power constraints, continuity constraints, and task workload constraints.

[0026] The task power constraint is:

[0027] ;

[0028] ;

[0029] In the above formula, For the task exist Power at any moment , Tasks The minimum executable power and the maximum executable power, For the task Rated operating power, For the task The source identifier variable for determination, This indicates that the mission is highly flexible overall, allowing for joint adjustments in time, power, and space dimensions. This indicates that the task has significant adjustment potential in its corresponding strength dimension, and adjustment is only permitted in that dimension, while other dimensions remain at their default state. For the task The computing load flexibility index, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters;

[0030] The continuity constraint is:

[0031] ;

[0032] In the above formula, , These are the start and end times of a certain time period. For the task exist The execution state variable at any given moment;

[0033] The workload constraint for the task is:

[0034] ;

[0035] In the above formula, For the task The computational workload coefficient corresponding to unit power For time step, For the task The computational workload;

[0036] The constraints of the interruptible flexible task include execution status and power constraints, and cumulative task workload constraints.

[0037] The execution state and power constraints are as follows:

[0038] ;

[0039] The cumulative task workload constraint is as follows:

[0040] ;

[0041] The task completion constraints are as follows:

[0042] ;

[0043] In the above formula, , Tasks The earliest allowed start time and the latest allowed completion time, , Tasks The actual start time and actual completion time;

[0044] The power limit constraint for the intelligent computing center is:

[0045] ;

[0046] In the above formula, This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle. This represents the maximum operating power of the intelligent computing center.

[0047] The budget cost constraint is:

[0048] ;

[0049] In the above formula, for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This represents the maximum electricity purchase budget for the intelligent computing center.

[0050] The flexible task count constraint applies to the adjustable flexible task set. The task in the text is:

[0051] ;

[0052] In the above formula, For the scheduling period, For the task The maximum number of adjustments.

[0053] In S1, the conditions for dividing tasks into rigid tasks and flexible tasks are as follows:

[0054] ;

[0055] In the above formula, A rigid set of tasks. For a set of flexible tasks, , Tasks The earliest allowed start time and the latest allowed completion time, For the task The computational workload, For the task The computational workload coefficient corresponding to unit power , Tasks The minimum executable power and the maximum executable power.

[0056] Secondly, this invention proposes an intelligent computing center operation optimization system based on load regulation and multi-power supply coordination, including a task division module, an adjustable and flexible task screening module, and an intelligent computing center operation optimization module.

[0057] The task partitioning module is used to divide the computing power tasks to be executed in the intelligent computing center into rigid tasks and flexible tasks based on the adjustability of the task load in the time and power dimensions.

[0058] The adjustable flexible task screening module is used to screen adjustable flexible tasks based on the computing power load flexibility index of each flexible task.

[0059] The intelligent computing center optimization operation module is used to simultaneously consider rigid and flexible tasks, and combine the flexible adjustment mode of adjustable flexible tasks with the multi-power supply mode of the intelligent computing center. With the goal of minimizing comprehensive operating costs, carbon emissions, and load fluctuations, it constructs an intelligent computing center optimization operation model, solves the model, and generates a low-carbon optimal operation strategy for the intelligent computing center. The strategy includes the execution sequence and execution power of adjustable flexible tasks, as well as the power supply adjustment mode of the intelligent computing center.

[0060] The adjustable flexible task screening module includes a computing load flexibility index calculation unit and a judgment and screening unit.

[0061] The computing load flexibility index calculation unit is used to calculate the computing load flexibility index of each flexible task using the following formula:

[0062] ;

[0063] ;

[0064] ;

[0065] ;

[0066] ;

[0067] In the above formula, For the task The computing load flexibility index, , , These are the weights for the time dimension, the power dimension, and the spatial dimension, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters , Tasks The earliest allowed start time and the latest allowed completion time, To complete the task Standard execution time For the task Execution continuity type variables, Indicates a task that is executed continuously. This indicates an interruptible task. For the task exist The total execution time has been accumulated. For the task The computational workload, For the task The computational workload coefficient corresponding to unit power For the task Rated operating power, , Tasks The minimum executable power and the maximum executable power, A rigid set of tasks. For a set of flexible tasks, For the task The flag indicating whether cross-node migration is allowed. For the task The set of computing power nodes allowed to execute. For computing power nodes exist The carbon emission intensity of electricity at any given moment;

[0068] The judgment and filtering unit is used to determine whether each flexible task is satisfied. If satisfied, then the task... Added to the set of adjustable flexible tasks as an adjustable flexible task. ,in, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively.

[0069] The intelligent computing center optimization operation module includes an objective function construction unit;

[0070] The objective function construction unit is used to construct the objective function of the following intelligent computing center optimization operation model:

[0071] ;

[0072] ;

[0073] ;

[0074] In the above formula, The objective function of the intelligent computing center's operating strategy is... This is the cost weighting coefficient. for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This is the carbon emission weighting coefficient. For computing power nodes exist Carbon emission intensity of electricity at any given time For computing power nodes exist Electricity consumption at all times This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle, including both rigid and flexible tasks. For the task exist Power at any moment For time step, For the intelligent computing center New energy power that is constantly generated and used by the user. For the task computing nodes exist Operating power at any given time.

[0075] The intelligent computing center optimization operation module also includes a constraint construction unit;

[0076] The constraint construction unit is used to construct the constraints of the intelligent computing center's optimized operation model, including continuous execution flexible task constraints, interruptible execution flexible task constraints, task completion constraints, intelligent computing center power limit constraints, budget cost constraints, and flexible adjustment task count constraints.

[0077] The constraints of the continuously executed flexible task include task power constraints, continuity constraints, and task workload constraints.

[0078] The task power constraint is:

[0079] ;

[0080] ;

[0081] In the above formula, For the task exist Power at any moment , Tasks The minimum executable power and the maximum executable power, For the task Rated operating power, For the task The source identifier variable for determination, This indicates that the mission is highly flexible overall, allowing for joint adjustments in time, power, and space dimensions. This indicates that the task has significant adjustment potential in its corresponding strength dimension, and adjustment is only permitted in that dimension, while other dimensions remain at their default state. For the task The computing load flexibility index, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters;

[0082] The continuity constraint is:

[0083] ;

[0084] In the above formula, , These are the start and end times of a certain time period. For the task exist The execution state variable at any given moment;

[0085] The workload constraint for the task is:

[0086] ;

[0087] In the above formula, For the task The computational workload coefficient corresponding to unit power For time step, For the task The computational workload;

[0088] The constraints of the interruptible flexible task include execution status and power constraints, and cumulative task workload constraints.

[0089] The execution state and power constraints are as follows:

[0090] ;

[0091] The cumulative task workload constraint is as follows:

[0092] ;

[0093] The task completion constraints are as follows:

[0094] ;

[0095] In the above formula, , Tasks The earliest allowed start time and the latest allowed completion time, , Tasks The actual start time and actual completion time;

[0096] The power limit constraint for the intelligent computing center is:

[0097] ;

[0098] In the above formula, This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle. This represents the maximum operating power of the intelligent computing center.

[0099] The budget cost constraint is:

[0100] ;

[0101] In the above formula, for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This represents the maximum electricity purchase budget for the intelligent computing center.

[0102] The flexible task count constraint applies to the adjustable flexible task set. The task in the text is:

[0103] ;

[0104] In the above formula, For the scheduling period, For the task The maximum number of adjustments.

[0105] In the task partitioning module, the conditions for dividing tasks into rigid tasks and flexible tasks are as follows:

[0106] ;

[0107] In the above formula, A rigid set of tasks. For a set of flexible tasks, , Tasks The earliest allowed start time and the latest allowed completion time, For the task The computational workload, For the task The computational workload coefficient corresponding to unit power , Tasks The minimum executable power and the maximum executable power.

[0108] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0109] 1. This invention proposes an optimization method and system for the operation of intelligent computing centers based on load adjustment and multi-power supply coordination. The method first divides the computing tasks to be executed in the intelligent computing center into rigid tasks and flexible tasks based on the adjustability of the task load in the time and power dimensions. Then, based on the computing load flexibility index of each flexible task, adjustable flexible tasks are selected. Finally, considering both rigid and flexible tasks, and combining the flexible adjustment mode of the adjustable flexible tasks with the multi-power supply mode of the intelligent computing center, an optimized operation model for the intelligent computing center is constructed with the goal of minimizing overall operating costs, carbon emissions, and load fluctuations. The model is then solved to generate a low-carbon optimal operation strategy for the intelligent computing center. This strategy includes the execution sequence and power of the adjustable flexible tasks, as well as the power supply adjustment method of the intelligent computing center. This method comprehensively considers the flexibility of the computing load inside the intelligent computing center and the differences in multiple external power supply methods. It explicitly models the task timing and power adjustability, and performs collaborative optimization of task execution scheduling and power supply mode. This enables the intelligent computing center's computing load and multiple power supply structure to adaptively match under low-carbon constraints, effectively reducing the electricity purchase cost and carbon emissions of the intelligent computing center, and improving the overall economic efficiency, low-carbon operation and operational stability of the intelligent computing center.

[0110] 2. This invention proposes a method and system for optimizing the operation of intelligent computing centers based on load regulation and multi-power supply coordination. The method first divides the task load into rigid and flexible categories based on the adjustability of the task load in the time and power dimensions, and establishes a mapping relationship between the task execution status and the power load. Then, by constructing a unified flexible task computing power load flexibility index, the originally difficult-to-quantify flexible task adjustment potential is transformed into a calculable, sortable, and constrainable flexible adjustable parameter system, avoiding indiscriminate mobilization of all flexible tasks, reducing the system's scheduling scale and reconfiguration frequency, and improving the service quality of the intelligent computing center. Attached Figure Description

[0111] Figure 1 This is an overall flowchart of the method described in this invention.

[0112] Figure 2 This is a structural diagram of the system described in this invention. Detailed Implementation

[0113] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings.

[0114] This invention proposes a method and system for optimizing the operation of intelligent computing centers based on load regulation and multi-power supply coordination. It obtains the operational characteristic parameters of tasks to be executed within the intelligent computing center and classifies tasks based on whether they allow adjustment in the time or power dimensions. A computing load flexibility index is constructed to comprehensively quantify the adjustment potential of different tasks in the time, power, and spatial dimensions. Flexible tasks are classified into flexibility levels according to a preset computing load flexibility threshold, and a set of flexible and adjustable tasks is selected. Combining flexible adjustment modes and multi-power supply modes, with the goal of "minimizing overall operating cost, minimizing carbon emissions, and minimizing load fluctuations," the execution sequence, execution power of flexible tasks, and the power supply mode of the intelligent computing center are collaboratively optimized to generate a low-carbon optimal operation strategy for the intelligent computing center. This method explicitly models the task timing and power adjustability as computing load flexibility resources and collaboratively optimizes them with a multi-mode power supply structure. This achieves adaptive matching of the intelligent computing center's computing load and multi-power supply structure under low-carbon constraints, improving the low-carbon nature, economy, and power supply reliability of the intelligent computing center's operation.

[0115] Example 1:

[0116] In this embodiment, a smart computing center is used as the research object. The operation optimization method of the smart computing center based on load adjustment and multi-power supply coordination is adopted to optimize its operation strategy when performing tasks.

[0117] like Figure 1 As shown, the intelligent computing center operation optimization method based on load regulation and multi-power supply coordination is carried out in the following steps:

[0118] 1. Based on the adjustability of task load in the time and power dimensions, the computing power tasks to be executed in the intelligent computing center are divided into rigid tasks and flexible tasks.

[0119] Obtain the runtime characteristic parameters of various tasks to be executed within the intelligent computing center, assuming the intelligent computing center is in a scheduling cycle. The set of tasks to be executed is For any task Construct its runtime characteristic parameter vector:

[0120] ;

[0121] In the above formula, For the task The vector of runtime characteristic parameters, For the task The computational workload, , Tasks The earliest allowed start time and the latest allowed completion time, For the task The interrupt flag indicates whether interruption is allowed (1 indicates interruption is allowed). For the task The flag indicating whether cross-node migration is allowed (1 indicates migration is allowed). For the task Rated operating power, , Tasks The minimum executable power and the maximum executable power, For the task The computational workload coefficient corresponding to unit power For the task The set of computing power nodes that are allowed to execute;

[0122] The intelligent computing center has three time segments in one scheduling cycle, each segment has a time step of 8 hours, and contains only one computing node. Its power supply is shown in Table 1.

[0123] Table 1 Power Supply Status of the Intelligent Computing Center

[0124] .

[0125] Based on task Whether adjustments are allowed in the time or power dimension, whether tasks are rigidly or flexibly divided, and whether a mapping relationship is established between task execution status and power load;

[0126] If task Maximum delay time satisfies This indicates that the task has room for adjustment in the time dimension; if the task The power adjustable range meets This indicates that the task is adjustable in the power dimension; therefore, the comprehensive judgment rule for the rigidity and flexibility of the task is:

[0127] ;

[0128] In the above formula, A rigid set of tasks. For a flexible set of tasks, satisfying , ;

[0129] In discrete time scale The above introduces task execution state variables. Construct the rigid load boundary and the flexible adjustable load boundary of the intelligent computing center in each time period:

[0130] The task execution status variable is:

[0131] ;

[0132] The rigid load boundary is:

[0133] ;

[0134] The adjustable boundary of the flexible load is:

[0135] ;

[0136] ;

[0137] In the above formula, for The rigid load power of the real-time intelligent computing center For the task exist The execution state variable at any given time. , They are respectively The flexible load of the Time-Based Intelligent Computing Center has an adjustable lower and upper boundary.

[0138] 2. Based on the computing load flexibility index of each flexible task, select adjustable flexible tasks;

[0139] By characterizing the adjustment potential of flexible tasks in the time, power, and spatial dimensions respectively, a unified computing load flexibility index is constructed, transforming the originally difficult-to-quantify task adjustment potential into a calculable, sortable, and constrainable flexible parameter system.

[0140] For any flexible task Construct its time-dimensional moderating potential parameters Describe the urgency of the task in terms of time: Based on the execution continuity type of flexible tasks, establish continuous execution tasks ( ) and interruptible tasks ( The time adjustment potential model is as follows: For continuously executed tasks, only the start time can be shifted as a whole; for interruptible tasks, the time adjustment potential is not only related to the task time window, but also closely related to the current execution progress, and the time adjustment potential parameters need to be dynamically updated according to the execution status.

[0141] The time adjustment potential parameter for the task is:

[0142] ;

[0143] ;

[0144] For any flexible task Construct its power dimension adjustment potential parameters Describe the impact of task adjustment on the overall power curve of the intelligent computing center:

[0145] ;

[0146] For any flexible task Construct its spatial dimension adjustment potential parameters The description describes the ability of the task to migrate to low-carbon electricity consumption nodes:

[0147] ;

[0148] The computational load flexibility index of each flexible task is calculated using the following formula to comprehensively quantify the adjustment potential of each flexible task in terms of time, power, and space. The spatial dimension adjustment potential is only activated when there are differentiated power supply channels within the intelligent computing center or when there are intelligent computing centers across different parks / sites; otherwise, its weight is not considered. Set to zero:

[0149] ;

[0150] In the above formula, For the task The computing load flexibility index, , , These are the weights for the time dimension, the power dimension, and the spatial dimension, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters , Tasks The earliest allowed start time and the latest allowed completion time, To complete the task Standard execution time For the task Execution continuity type variables, Indicates a task that is executed continuously. This indicates an interruptible task. For the task exist The total execution time has been accumulated. For the task The computational workload, For the task The computational workload coefficient corresponding to unit power For the task Rated operating power, , Tasks The minimum executable power and the maximum executable power, A rigid set of tasks. For a set of flexible tasks, For the task The flag indicating whether cross-node migration is allowed. For the task The set of computing power nodes allowed to execute. For computing power nodes exist The intensity of electricity carbon emissions at any given moment.

[0151] Because different flexible tasks have significantly different adjustment potentials in terms of time, power, and execution methods, simultaneously adjusting all flexible tasks during subsequent collaborative operation optimization can easily lead to excessive scheduling scale, frequent system reconfiguration, and service quality fluctuations. Therefore, flexible tasks are classified into flexibility levels based on a preset computing load flexibility threshold, and a set of adjustable flexible tasks that can participate in the subsequent intelligent computing center optimization operation is selected, specifically including:

[0152] Based on a combined criterion of comprehensive computing load flexibility index and single-dimensional adjustment potential, a set of adjustable flexible tasks in the intelligent computing center that can actually participate in collaborative operation optimization is selected to avoid indiscriminately mobilizing all flexible tasks:

[0153] For any flexible task The task is valid if and only if it satisfies any of the following conditions:

[0154] 1) ;

[0155] 2) ;

[0156] 3) ;

[0157] 4) ;

[0158] Then the task It was classified as an adjustable flexible task and added to the adjustable flexible task set. ,in, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively.

[0159] The intelligent computing center includes four types of tasks within the scheduling cycle, and the corresponding task sets are shown in Table 2.

[0160] Table 2 Task Set of the Intelligent Computing Center

[0161] .

[0162] 3. Simultaneously considering rigid and flexible tasks, and combining the flexible adjustment mode of adjustable flexible tasks with the multi-power supply mode of the intelligent computing center, with the goal of minimizing comprehensive operating costs, carbon emissions, and load fluctuations, an optimized operation model for the intelligent computing center is constructed, and the model is solved to generate the low-carbon optimal operation strategy for the intelligent computing center. The strategy includes the execution sequence and execution power of adjustable flexible tasks, as well as the power supply adjustment mode of the intelligent computing center.

[0163] By incorporating all pending tasks within the intelligent computing center into a unified scheduling framework, and combining a flexible adjustment mode for adjustable task sets with a multi-power supply mode, an optimized operation model for the intelligent computing center is constructed with the goals of "minimizing overall operating costs, minimizing carbon emissions, and minimizing load fluctuations." The final execution sequence of each task and its actual power level during execution are collaboratively decided, naturally deriving the power consumption demand of the intelligent computing center for each time period within the scheduling cycle. Based on this, multiple power supply methods are introduced, including market electricity, green electricity, and self-consumed renewable energy. While meeting the power needs of the intelligent computing center, the procurement or usage ratio of each power supply method at different times is adjusted to jointly optimize operating costs and carbon emission levels, forming a low-carbon optimal operation strategy for the intelligent computing center.

[0164] The objective function of the intelligent computing center's optimized operating model includes:

[0165] ;

[0166] ;

[0167] ;

[0168] In the above formula, The objective function of the intelligent computing center's operating strategy is... This is the cost weighting coefficient. for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This is the carbon emission weighting coefficient. For computing power nodes exist Carbon emission intensity of electricity at any given time For computing power nodes exist Electricity consumption at all times This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle. For the task exist Power at any moment For time step, For the intelligent computing center New energy power that is constantly generated and used by the user. For the task computing nodes exist Operating power at any given time.

[0169] The constraints of the intelligent computing center's optimized operation model include constraints on continuously executed flexible tasks, constraints on interruptible flexible tasks, task completion constraints, intelligent computing center power limit constraints, budget cost constraints, and constraints on the number of flexible adjustment tasks.

[0170] To support differentiated modeling, a source identification variable is introduced. Based on the characteristics of task execution continuity, and for different flexible adjustment modes of adjustable flexible tasks, we construct differentiated load adjustment models for continuous execution flexible task constraints and interruptible execution flexible task constraints, and clarify their adjustment methods.

[0171] The source identifier variable for determination is:

[0172] ;

[0173] In the above formula, For the task The source identifier variable for determination, where, This indicates that the overall mission is highly flexible, allowing for joint adjustments in time, power, and space dimensions; When this occurs, it indicates that the task has significant adjustment potential in the corresponding strength dimension, and adjustment is only allowed in its corresponding strength dimension, while the other dimensions remain at their fixed state;

[0174] For continuously executing flexible tasks, interruptions are not allowed during execution, and the load adjustment method is determined by the source identifier variable. The constraints of the continuously executed flexible task include task power constraints, continuity constraints, and task workload constraints.

[0175] The task power constraint is:

[0176] ;

[0177] In the above formula, For the task exist Power at any moment , Tasks The minimum executable power and the maximum executable power, For the task Rated operating power;

[0178] The continuity constraint is:

[0179] ;

[0180] In the above formula, , These are the start and end times of a certain time period. For the task exist The execution state variable at any given moment;

[0181] The workload constraint for the task is:

[0182] ;

[0183] In the above formula, For the task The computational workload coefficient corresponding to unit power For time step, For the task The computational workload;

[0184] For interruptible flexible tasks, which can be split into multiple execution segments, the load adjustment method is also subject to the determination source identifier variable. The constraints include execution state and power constraints, and cumulative task workload constraints.

[0185] The execution state and power constraints are as follows:

[0186] ;

[0187] The cumulative task workload constraint is as follows:

[0188] .

[0189] All tasks must be completed within the allowed time window, and the task completion constraints are as follows:

[0190] ;

[0191] In the above formula, , Tasks The earliest allowed start time and the latest allowed completion time, , Tasks The actual start time and actual completion time;

[0192] The power limit constraint for the intelligent computing center is:

[0193] ;

[0194] In the above formula, This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle. This represents the maximum operating power of the intelligent computing center.

[0195] The budget cost constraint is:

[0196] ;

[0197] In the above formula, for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This represents the maximum electricity purchase budget for the intelligent computing center.

[0198] To avoid excessively frequent start-stop, splitting, or rescheduling of adjustable flexible tasks during optimization, and to ensure the engineering feasibility and operational stability of the scheduling scheme, the number of times a task participating in flexible adjustment changes its execution state within the scheduling cycle is limited to no more than a preset upper limit. This constraint on the number of flexible adjustment tasks applies to the set of adjustable flexible tasks. The task in the text is:

[0199] ;

[0200] In the above formula, For the scheduling period, For the task The maximum number of adjustments.

[0201] To verify the effectiveness of the present invention, the intelligent computing center was modified without changing the total computing load and service constraints. Differentiated flexible adjustments were made to continuous flexible tasks and interruptible flexible tasks. The adjustment strategies and the application power consumption are shown in Table 3.

[0202] Table 3 Differentiated Flexible Adjustment Strategies for Intelligent Computing Centers

[0203] ;

[0204] During the scheduling cycle, all electricity consumption of the intelligent computing center is met by purchasing electricity from the market. According to Tables 1 and 3, the electricity purchase costs for Strategy 1, Strategy 2, and Strategy 3 are RMB 649,000, RMB 641,000, and RMB 635,000, respectively, with corresponding carbon emissions of 488 tons, 484 tons, and 481 tons. Therefore, Strategy 3 can effectively reduce the electricity purchase cost and carbon emissions of the intelligent computing center, improving its economic and low-carbon operation level.

[0205] Example 2:

[0206] like Figure 2 As shown, the intelligent computing center operation optimization system based on load regulation and multi-power supply coordination includes a task partitioning module, an adjustable and flexible task selection module, and an intelligent computing center operation optimization module.

[0207] The task partitioning module is used to divide the computing power tasks to be executed in the intelligent computing center into rigid tasks and flexible tasks based on the adjustability of the task load in the time and power dimensions.

[0208] The adjustable flexible task screening module is used to screen adjustable flexible tasks based on the computing power load flexibility index of each flexible task.

[0209] The intelligent computing center optimization operation module is used to simultaneously consider rigid and flexible tasks, and combine the flexible adjustment mode of adjustable flexible tasks with the multi-power supply mode of the intelligent computing center. With the goal of minimizing comprehensive operating costs, carbon emissions, and load fluctuations, it constructs an intelligent computing center optimization operation model, solves the model, and generates a low-carbon optimal operation strategy for the intelligent computing center. The strategy includes the execution sequence and execution power of adjustable flexible tasks, as well as the power supply adjustment mode of the intelligent computing center.

[0210] In the task partitioning module, the conditions for dividing tasks into rigid tasks and flexible tasks are as follows:

[0211] ;

[0212] In the above formula, A rigid set of tasks. For a set of flexible tasks, , Tasks The earliest allowed start time and the latest allowed completion time, For the task The computational workload, For the task The computational workload coefficient corresponding to unit power , Tasks The minimum executable power and the maximum executable power.

[0213] The adjustable flexible task screening module includes a computing load flexibility index calculation unit and a judgment and screening unit.

[0214] The computing load flexibility index calculation unit is used to calculate the computing load flexibility index of each flexible task using the following formula:

[0215] ;

[0216] ;

[0217] ;

[0218] ;

[0219] ;

[0220] In the above formula, For the task The computing load flexibility index, , , These are the weights for the time dimension, the power dimension, and the spatial dimension, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters , Tasks The earliest allowed start time and the latest allowed completion time, To complete the task Standard execution time For the task Execution continuity type variables, Indicates a task that is executed continuously. This indicates an interruptible task. For the task exist The total execution time has been accumulated. For the task The computational workload, For the task The computational workload coefficient corresponding to unit power For the task Rated operating power, , Tasks The minimum executable power and the maximum executable power, A rigid set of tasks. For a set of flexible tasks, For the task The flag indicating whether cross-node migration is allowed. For the task The set of computing power nodes allowed to execute. For computing power nodes exist The carbon emission intensity of electricity at any given moment;

[0221] The judgment and filtering unit is used to determine whether each flexible task is satisfied. If satisfied, then the task... Added to the set of adjustable flexible tasks as an adjustable flexible task. ,in, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively.

[0222] The intelligent computing center optimization operation module includes an objective function construction unit;

[0223] The objective function construction unit is used to construct the objective function of the following intelligent computing center optimization operation model:

[0224] ;

[0225] ;

[0226] ;

[0227] In the above formula, The objective function of the intelligent computing center's operating strategy is... This is the cost weighting coefficient. for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This is the carbon emission weighting coefficient. For computing power nodes exist Carbon emission intensity of electricity at any given time For computing power nodes exist Electricity consumption at all times This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle, including both rigid and flexible tasks. For the task exist Power at any moment For time step, For the intelligent computing center New energy power that is constantly generated and used by the user. For the task computing nodes exist Operating power at any given time.

[0228] The intelligent computing center optimization operation module also includes a constraint construction unit;

[0229] The constraint construction unit is used to construct the constraints of the intelligent computing center's optimized operation model, including continuous execution flexible task constraints, interruptible execution flexible task constraints, task completion constraints, intelligent computing center power limit constraints, budget cost constraints, and flexible adjustment task count constraints.

[0230] The constraints of the continuously executed flexible task include task power constraints, continuity constraints, and task workload constraints.

[0231] The task power constraint is:

[0232] ;

[0233] ;

[0234] In the above formula, For the task exist Power at any moment , Tasks The minimum executable power and the maximum executable power, For the task Rated operating power, For the task The source identifier variable for determination, This indicates that the mission is highly flexible overall, allowing for joint adjustments in time, power, and space dimensions. This indicates that the task has significant adjustment potential in its corresponding strength dimension, and adjustment is only permitted in that dimension, while other dimensions remain at their default state. For the task The computing load flexibility index, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters;

[0235] The continuity constraint is:

[0236] ;

[0237] In the above formula, , These are the start and end times of a certain time period. For the task exist The execution state variable at any given moment;

[0238] The workload constraint for the task is:

[0239] ;

[0240] In the above formula, For the task The computational workload coefficient corresponding to unit power For time step, For the task The computational workload;

[0241] The constraints of the interruptible flexible task include execution status and power constraints, and cumulative task workload constraints.

[0242] The execution state and power constraints are as follows:

[0243] ;

[0244] The cumulative task workload constraint is as follows:

[0245] ;

[0246] The task completion constraints are as follows:

[0247] ;

[0248] In the above formula, , Tasks The earliest allowed start time and the latest allowed completion time, , Tasks The actual start time and actual completion time;

[0249] The power limit constraint for the intelligent computing center is:

[0250] ;

[0251] In the above formula, This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle. This represents the maximum operating power of the intelligent computing center.

[0252] The budget cost constraint is:

[0253] ;

[0254] In the above formula, for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This represents the maximum electricity purchase budget for the intelligent computing center.

[0255] The flexible task count constraint applies to the adjustable flexible task set. The task in the text is:

[0256] ;

[0257] In the above formula, For the scheduling period, For the task The maximum number of adjustments.

Claims

1. A method for optimizing the operation of intelligent computing centers based on load regulation and multi-power supply coordination, characterized in that, The method includes: S1. Based on the adjustability of task load in the time and power dimensions, the computing power tasks to be executed in the intelligent computing center are divided into rigid tasks and flexible tasks. S2. Select adjustable flexible tasks based on the computing load flexibility index of each flexible task. S3. Simultaneously considering rigid and flexible tasks, and combining the flexible adjustment mode of adjustable flexible tasks with the multi-power supply mode of the intelligent computing center, with the goal of minimizing comprehensive operating costs, carbon emissions, and load fluctuations, construct an optimized operation model for the intelligent computing center, solve the model, and generate the low-carbon optimal operation strategy for the intelligent computing center. The strategy includes the execution sequence and execution power of adjustable flexible tasks, as well as the power supply adjustment mode of the intelligent computing center.

2. The intelligent computing center operation optimization method based on load regulation and multi-power supply coordination according to claim 1, characterized in that, S2 includes: S21. Calculate the computing load flexibility index for each flexible task using the following formula: ； ； ； ； ； In the above formula, For the task The computing load flexibility index, , , These are the weights for the time dimension, the power dimension, and the spatial dimension, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters , Tasks The earliest allowed start time and the latest allowed completion time, To complete the task Standard execution time For the task Execution continuity type variables, Indicates a task that is executed continuously. This indicates an interruptible task. For the task exist The total execution time has been accumulated. For the task The computational workload, For the task The computational workload coefficient corresponding to unit power For the task Rated operating power, , Tasks The minimum executable power and the maximum executable power, A rigid set of tasks. For a set of flexible tasks, For the task The flag indicating whether cross-node migration is allowed. For the task The set of computing power nodes allowed to execute. For computing power nodes exist The carbon emission intensity of electricity at any given moment; S22. Determine whether each flexible task meets the requirements. If satisfied, then the task... Added to the set of adjustable flexible tasks as an adjustable flexible task. ,in, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively.

3. The intelligent computing center operation optimization method based on load regulation and multi-power supply coordination according to claim 1, characterized in that, In S3, the objective function of the intelligent computing center's optimized operation model includes: ； ； ； In the above formula, The objective function of the intelligent computing center's operating strategy is... This is the cost weighting coefficient. for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This is the carbon emission weighting coefficient. For computing power nodes exist Carbon emission intensity of electricity at any given time For computing power nodes exist Electricity consumption at all times This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle, including both rigid and flexible tasks. For the task exist Power at any moment For time step, For the intelligent computing center New energy power that is constantly generated and used by the user. For the task computing nodes exist Operating power at any given time.

4. The intelligent computing center operation optimization method based on load regulation and multi-power supply coordination according to claim 1, characterized in that, In S3, the constraints of the intelligent computing center's optimized operation model include continuous execution flexible task constraints, interruptible execution flexible task constraints, task completion constraints, intelligent computing center power limit constraints, budget cost constraints, and flexible adjustment task count constraints. The constraints of the continuously executed flexible task include task power constraints, continuity constraints, and task workload constraints. The task power constraint is: ； ； In the above formula, For the task exist Power at any moment , Tasks The minimum executable power and the maximum executable power, For the task Rated operating power, For the task The source identifier variable for determination, This indicates that the mission is highly flexible overall, allowing for joint adjustments in time, power, and space dimensions. This indicates that the task has significant adjustment potential in its corresponding strength dimension, and adjustment is only permitted in that dimension, while other dimensions remain at their default state. For the task The computing load flexibility index, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters; The continuity constraint is: ； In the above formula, , These are the start and end times of a certain time period. For the task exist The execution state variable at any given moment; The workload constraint for the task is: ； In the above formula, For the task The computational workload coefficient corresponding to unit power For time step, For the task The computational workload; The constraints of the interruptible flexible task include execution status and power constraints, and cumulative task workload constraints. The execution state and power constraints are as follows: ； The cumulative task workload constraint is as follows: ； The task completion constraints are as follows: ； In the above formula, , Tasks The earliest allowed start time and the latest allowed completion time, , Tasks The actual start time and actual completion time; The power limit constraint for the intelligent computing center is: ； In the above formula, This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle. This represents the maximum operating power of the intelligent computing center. The budget cost constraint is: ； In the above formula, for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This represents the maximum electricity purchase budget for the intelligent computing center. The flexible task count constraint applies to the adjustable flexible task set. The task in the text is: ； In the above formula, For the scheduling period, For the task The maximum number of adjustments.

5. The intelligent computing center operation optimization method based on load regulation and multi-power supply coordination according to claim 1, characterized in that, In S1, the conditions for dividing tasks into rigid tasks and flexible tasks are as follows: ； In the above formula, A rigid set of tasks. For a set of flexible tasks, , Tasks The earliest allowed start time and the latest allowed completion time, For the task The computational workload, For the task The computational workload coefficient corresponding to unit power , Tasks The minimum executable power and the maximum executable power.

6. A smart computing center operation optimization system based on load regulation and multi-power supply coordination, characterized in that, The system includes a task partitioning module, an adjustable and flexible task filtering module, and an intelligent computing center optimization operation module. The task partitioning module is used to divide the computing power tasks to be executed in the intelligent computing center into rigid tasks and flexible tasks based on the adjustability of the task load in the time and power dimensions. The adjustable flexible task screening module is used to screen adjustable flexible tasks based on the computing power load flexibility index of each flexible task. The intelligent computing center optimization operation module is used to simultaneously consider rigid and flexible tasks, and combine the flexible adjustment mode of adjustable flexible tasks with the multi-power supply mode of the intelligent computing center. With the goal of minimizing comprehensive operating costs, carbon emissions, and load fluctuations, it constructs an intelligent computing center optimization operation model, solves the model, and generates a low-carbon optimal operation strategy for the intelligent computing center. The strategy includes the execution sequence and execution power of adjustable flexible tasks, as well as the power supply adjustment mode of the intelligent computing center.

7. The intelligent computing center operation optimization system based on load regulation and multi-power supply coordination according to claim 6, characterized in that, The adjustable flexible task screening module includes a computing load flexibility index calculation unit and a judgment and screening unit. The computing load flexibility index calculation unit is used to calculate the computing load flexibility index of each flexible task using the following formula: ； ； ； ； ； In the above formula, For the task The computing load flexibility index, , , These are the weights for the time dimension, the power dimension, and the spatial dimension, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters , Tasks The earliest allowed start time and the latest allowed completion time, To complete the task Standard execution time For the task Execution continuity type variables, Indicates a task that is executed continuously. This indicates an interruptible task. For the task exist The total execution time has been accumulated. For the task The computational workload, For the task The computational workload coefficient corresponding to unit power For the task Rated operating power, , Tasks The minimum executable power and the maximum executable power, A rigid set of tasks. For a set of flexible tasks, For the task The flag indicating whether cross-node migration is allowed. For the task The set of computing power nodes allowed to execute. For computing power nodes exist The carbon emission intensity of electricity at any given moment; The judgment and filtering unit is used to determine whether each flexible task is satisfied. If satisfied, then the task... Added to the set of adjustable flexible tasks as an adjustable flexible task. ,in, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively.

8. The intelligent computing center operation optimization system based on load regulation and multi-power supply coordination according to claim 6, characterized in that, The intelligent computing center optimization operation module includes an objective function construction unit; The objective function construction unit is used to construct the objective function of the following intelligent computing center optimization operation model: ； ； ； In the above formula, The objective function of the intelligent computing center's operating strategy is... This is the cost weighting coefficient. for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This is the carbon emission weighting coefficient. For computing power nodes exist Carbon emission intensity of electricity at any given time For computing power nodes exist Electricity consumption at all times This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle, including both rigid and flexible tasks. For the task exist Power at any moment For time step, For the intelligent computing center New energy power that is constantly generated and used by the user. For the task computing nodes exist Operating power at any given time.

9. The intelligent computing center operation optimization system based on load regulation and multi-power supply coordination according to claim 6, characterized in that, The intelligent computing center optimization operation module also includes a constraint construction unit; The constraint construction unit is used to construct the constraints of the intelligent computing center's optimized operation model, including continuous execution flexible task constraints, interruptible execution flexible task constraints, task completion constraints, intelligent computing center power limit constraints, budget cost constraints, and flexible adjustment task count constraints. The constraints of the continuously executed flexible task include task power constraints, continuity constraints, and task workload constraints. The task power constraint is: ； ； In the above formula, For the task exist Power at any moment , Tasks The minimum executable power and the maximum executable power, For the task Rated operating power, For the task The source identifier variable for determination, This indicates that the mission is highly flexible overall, allowing for joint adjustments in time, power, and space dimensions. This indicates that the task has significant adjustment potential in its corresponding strength dimension, and adjustment is only permitted in that dimension, while other dimensions remain at their default state. For the task The computing load flexibility index, The threshold for the computing power load flexibility index. , , These are the regulation potential thresholds for time, power, and space dimensions, respectively. For the task Time adjustment potential parameter For the task Power regulation potential parameters For the task Spatial regulation potential parameters; The continuity constraint is: ； In the above formula, , These are the start and end times of a certain time period. For the task exist The execution state variable at any given moment; The workload constraint for the task is: ； In the above formula, For the task The computational workload coefficient corresponding to unit power For time step, For the task The computational workload; The constraints of the interruptible flexible task include execution status and power constraints, and cumulative task workload constraints. The execution state and power constraints are as follows: ； The cumulative task workload constraint is as follows: ； The task completion constraints are as follows: ； In the above formula, , Tasks The earliest allowed start time and the latest allowed completion time, , Tasks The actual start time and actual completion time; The power limit constraint for the intelligent computing center is: ； In the above formula, This refers to the set of tasks to be executed by the intelligent computing center within its scheduling cycle. This represents the maximum operating power of the intelligent computing center. The budget cost constraint is: ； In the above formula, for The market price of electricity at any given time For the intelligent computing center Electricity purchased from the market at all times for Green electricity prices at any time For the intelligent computing center Green electricity that is constantly being purchased This represents the maximum electricity purchase budget for the intelligent computing center. The flexible task count constraint applies to the adjustable flexible task set. The task in the text is: ； In the above formula, For the scheduling period, For the task The maximum number of adjustments.

10. The intelligent computing center operation optimization system based on load regulation and multi-power supply coordination according to claim 6, characterized in that, In the task partitioning module, the conditions for dividing tasks into rigid tasks and flexible tasks are as follows: ； In the above formula, A rigid set of tasks. For a set of flexible tasks, , Tasks The earliest allowed start time and the latest allowed completion time, For the task The computational workload, For the task The computational workload coefficient corresponding to unit power , Tasks The minimum executable power and the maximum executable power.

Images