CN-121998332-A - Regional intelligent computing center-based power system computing power collaborative scheduling method, system, equipment and medium

CN121998332ACN 121998332 ACN121998332 ACN 121998332ACN-121998332-A

Abstract

The invention discloses a regional intelligent computing center-based power system calculation force cooperative scheduling method and system, which belong to the technical field of computer data processing and comprise the steps of obtaining real-time operation data and prediction data of a regional power grid, calculating scheduling priority factors of all nodes of the power grid, obtaining real-time task queue data and calculation resource state data of all intelligent computing centers, calculating calculation force load factors of all intelligent computing centers based on the real-time task queue data and the calculation resource state data, obtaining a cooperative strategy library containing symbiotic strategies, migration strategies and buffering strategies, inputting the scheduling priority factors and the calculation force load factors into the cooperative strategy library for matching to generate a task scheduling strategy set, and generating and issuing power scheduling instructions and calculation force scheduling instructions according to the task scheduling strategy set. The invention realizes the deep coordination and global optimization configuration of the electric power resource and the computing power resource.

Inventors

XIE YIPENG
ZHANG FUXIN
ZHANG YUHUI
MA WEI
XIE XIN
WANG GUANFU
ZHOU BOWEN
ZHANG JUAN
Fang Yongji
SONG LI
WANG LIPENG
LIU YULONG
WANG WEI

Assignees

国网辽宁省电力有限公司辽阳供电公司
国网辽宁省电力有限公司
东北大学

Dates

Publication Date: 20260508
Application Date: 20260122

Claims (12)

1. The power system calculation force cooperative scheduling method based on the regional level intelligent calculation center is characterized by comprising the following steps of: Acquiring real-time operation data and prediction data of a regional power grid, and calculating a scheduling priority factor of each node of the power grid based on the real-time operation data and the prediction data; Acquiring real-time task queue data and computing resource state data of each intelligent computing center, and computing the computing power load factor of each intelligent computing center based on the real-time task queue data and the computing resource state data; Acquiring a collaborative strategy library comprising a symbiotic strategy, a migration strategy and a buffer strategy, and inputting the scheduling priority factor and the computational load factor into the collaborative strategy library for matching to generate a task scheduling strategy set; and generating and issuing a power generation dispatching instruction and a power calculation dispatching instruction to a corresponding power grid control system and intelligent computation center task management system according to the task dispatching strategy set.
2. The regional intelligent computing center-based power system power collaborative scheduling method according to claim 1, wherein the obtaining real-time operation data and prediction data of a regional power grid, calculating a scheduling priority factor of each node of the power grid based on the real-time operation data and prediction data, comprises: Extracting power supply margin parameters, line transmission congestion parameters and real-time electricity price parameters of each node from the real-time operation data, and extracting renewable energy output prediction parameters from the prediction data to form a multi-dimensional power grid state parameter set; acquiring a first weight configuration for representing the power grid stability, transmission economy, electricity cost and green energy consumption importance; And carrying out weighted fusion on the multidimensional power grid state parameter set by adopting the first weight configuration to generate the dispatching priority factor.
3. The regional intelligent computing center-based power system power collaborative scheduling method according to claim 1, wherein the acquiring real-time task queue data and computing resource status data of each intelligent computing center, calculating a power load factor of each intelligent computing center based on the real-time task queue data and computing resource status data, comprises: Analyzing calculated quantity parameters, task delay time parameters and data migration cost parameters of a task to be processed from the real-time task queue data to form a task flexibility parameter set; Obtaining a unit calculation energy consumption coefficient for calibrating unit calculation amount power consumption, and calculating a predicted power load increment according to the calculation amount parameter and the unit calculation energy consumption coefficient; and correcting the predicted power load increment by combining the task flexibility parameter set to generate the calculated power load factor.
4. The regional intelligent computing center-based power system power collaborative scheduling method according to claim 1, wherein the obtaining a collaborative policy base including a symbiotic policy, a migration policy and a buffering policy, inputting the scheduling priority factor and the power load factor into the collaborative policy base for matching, generating a task scheduling policy set, includes: According to the numerical combination of the scheduling priority factor and the computational power load factor, matching a local collaborative strategy with highest priority from the collaborative strategy library; Acquiring rule parameters for converting the local collaborative policy into executable instructions; And processing the tasks in the real-time task queue data based on the local collaborative strategy and the rule parameters, generating a specific scheduling strategy comprising a task execution position, task execution time and a task processing mode, and converging to form the task scheduling strategy set.
5. The regional intelligent computing center-based power system power coordination scheduling method according to claim 1, wherein the generating and issuing power scheduling instructions and power scheduling instructions to corresponding power grid control systems and intelligent computing center task management systems according to the task scheduling policy set specifically comprises: receiving instruction execution result data fed back by the power grid control system and the intelligent computation center task management system; Acquiring a system optimization target for evaluating the running economy and efficiency of a system, and generating an adjustment signal according to the performance deviation between the instruction execution result data and the system optimization target; And adjusting the first weight configuration and the rule parameters according to the adjustment signal.
6. The regional intelligent computing center-based power system power coordination scheduling method according to claim 5, wherein the generating and issuing power scheduling instructions and power scheduling instructions to the corresponding power grid control system and intelligent computing center task management system according to the task scheduling policy set further comprises: When the matched local collaborative policy is the migration policy, Screening out target nodes from all power grid nodes according to the dispatching priority factor, and sending a task migration negotiation request to an intelligent computing center associated with the target nodes; receiving a task migration negotiation response containing real-time load parameters returned by the intelligent computing center; And selecting an optimal migration target according to the real-time load parameter in the task migration negotiation response, determining a final task migration path and a migration task copy according to the optimal migration target, and writing the task migration path and the migration task copy into the specific scheduling strategy.
7. The regional intelligent computing center-based power system power collaborative scheduling method according to claim 6, wherein the selecting an optimal migration target according to a real-time load parameter in the task migration negotiation response, determining a final task migration path and a migration task copy according to the optimal migration target, and writing the task migration path and the migration task copy into the specific scheduling policy, further comprises: generating a clone copy of a task to be migrated in a source intelligent computing center, and marking the clone copy as the migration task copy; Transmitting the migration task copy to an intelligent computing center corresponding to the optimal migration target; Before the migration instruction takes effect, the clone copies are calculated in parallel in the source intelligent computing center and the intelligent computing center corresponding to the optimal migration target, so that seamless connection of calculation tasks is realized.
8. The regional intelligent computing center-based power system power collaborative scheduling method according to claim 1, further comprising: periodically converging the dispatching priority factors of all the power grid nodes and the calculation power load factors of all intelligent calculation centers to form a global system state snapshot; Identifying system-level conflict and collaboration opportunities based on the global system state snapshot, traversing all intelligent computation centers for planning to execute migration strategies, aggregating predicted load increments contained in computation load factors to target nodes, and marking migration congestion conflicts when the aggregate load increment of a certain target node is identified to exceed the admission capacity represented by a scheduling priority factor; And dynamically generating and broadcasting global evolution parameters for adjusting local decision logic according to the identification result, dynamically generating and broadcasting global evolution parameters when conflict or opportunity is identified, correcting the decision logic of the local agent, reducing the aggregation of load increment to a target node, broadcasting global evolution parameter parameters if a large-scale renewable energy grid-connected coordination opportunity is identified, improving the weight of renewable energy output prediction parameters of the area, and guiding more calculation load to be aggregated.
9. The regional intelligent computing center-based power system power collaborative scheduling method according to claim 8, wherein if a collaborative opportunity for large-scale renewable energy grid connection is identified, further comprising: initiating a reload suggestion of the computational flow comprising the reorganization time period and the target area; receiving feedback data which is fed back by each intelligent computing center to the advice and contains a participatable task list; and arranging deferrable tasks according to the feedback data, and generating a centralized scheduling instruction which is directed to the target area.
10. The regional level intelligent computing center-based power system power calculation collaborative scheduling system, which is applied to the regional level intelligent computing center-based power system power calculation collaborative scheduling method according to any one of claims 1-9, is characterized by comprising the following steps: The power grid situation awareness module is used for acquiring real-time operation data and prediction data of the regional power grid and calculating scheduling priority factors of all nodes of the power grid; The computing force situation sensing module is used for acquiring real-time task queue data and computing resource state data of each intelligent computing center and computing the computing force load factor of each intelligent computing center; the collaborative decision engine module is used for acquiring a collaborative strategy library, inputting the scheduling priority factor and the computational power load factor into the collaborative strategy library for matching, and generating a task scheduling strategy set; The instruction generation and distribution module is used for generating and issuing a power generation dispatching instruction and a power calculation dispatching instruction according to the task dispatching strategy set; The global evolution coordination module is used for periodically converging the system states, identifying system-level conflict and coordination opportunities and generating global evolution parameters for adjusting the local decision logic.
11. An electronic device comprising a processor and a memory, the memory for storing a computer program which, when executed by the processor, causes the electronic device to perform the regional level intelligent computing center-based power system power collaborative scheduling method of any one of claims 1-9.
12. A computer readable storage medium comprising a computer program or instructions which, when run on a computer, cause the computer to perform the regional level intelligent computing center based power system power co-scheduling method of any one of claims 1-9.

Description

Regional intelligent computing center-based power system computing power collaborative scheduling method, system, equipment and medium Technical Field The invention belongs to the technical field of computer data processing, and in particular to a regional intelligent computing center-based power system computing power collaborative scheduling method, system, equipment and medium. Background With the rapid development of compute-intensive applications such as artificial intelligence and big data analysis, regional intelligent computing centers have been deployed in large scale on various places as core infrastructure for bearing massive computing power demands, and the intelligent computing centers aggregate a large number of high-performance computing servers, and the operation of the servers has extremely high energy density, and the influence on an electric power system is increasingly remarkable, so that the servers become important loads in a power grid. Meanwhile, the computing tasks carried in the system, especially the non-real-time batch processing tasks, have time and space scheduling flexibility. In the prior art, the dispatching management of the power system and the intelligent computing center is usually carried out separately, the power dispatching system mainly manages large users such as a data center through a demand response mechanism, the adopted technical means are mostly indirect guidance based on price signals, such as time-of-use electricity price, interruptible load compensation and the like, or a forced power reduction instruction is sent out in an emergency, and the power system regards the intelligent computing center as an integral load unit with relatively slow response. On the other hand, the task scheduling system in the intelligent computing center is focused on optimizing the utilization rate of computing resources and guaranteeing the service quality of tasks, and the decision is based on the priority, the deadline, the data dependency relationship and the load balancing of an internal server of the tasks, and the real-time running state of the power grid is simply considered as an external cost factor. The prior art means described above have the following inherent drawbacks: 1. the information interaction between the electric power system and the intelligent computing center is unidirectional and shallow, depth perception of the internal running states of the electric power system and the intelligent computing center is lacked, the electric power network cannot acquire the specific flexibility of computing tasks, the computing power scheduling cannot predict the accurate influence of decisions on the local electric power network, and the independent optimization targets of the two parties often cause behavior conflicts and cannot form system-level resultant force; 2. Multiple computing centers simultaneously respond to low electricity price signals to migrate tasks to the same area, new power grid congestion may be locally caused, a cooperative scheduling mode is lacked, flexibility value of intelligent computing center loads cannot be fully mined, and capacity of the power grid for accommodating renewable energy sources and economy of overall operation are limited. Therefore, a power system calculation force cooperative scheduling method based on regional intelligent computing centers is needed, so that the power grid system and the power system are deeply integrated. Disclosure of Invention In view of the above or the shortcomings of the prior art, the invention provides a regional intelligent computing center-based power system power computing collaborative scheduling method and a regional intelligent computing center-based power system power computing collaborative scheduling system, which are used for realizing bidirectional perception through calculating a scheduling priority factor and a power computing load factor and realizing deep collaboration and global optimization configuration of power resources and power computing resources based on a collaborative policy library to make a unified decision. In order to solve the technical problems, the invention provides the following technical scheme: In a first aspect, the present invention provides a power system computing power collaborative scheduling method based on a regional level computing center, where the method is applied to a power system computing power collaborative scheduling system based on a regional level computing center, the system includes a regional power grid object for forming a power grid, a computing center, a power grid control system for scheduling, and a regulation terminal, and the method includes: Acquiring real-time operation data and prediction data of a regional power grid, and calculating a scheduling priority factor of each node of the power grid based on the real-time operation data and the prediction data; Acquiring real-time task queue data and computing resource state data of each