CN-121979665-A - Electric calculation collaborative integrated management method and system based on virtual power plant

CN121979665ACN 121979665 ACN121979665 ACN 121979665ACN-121979665-A

Abstract

The invention provides an electric calculation collaborative integrated management method and system based on a virtual power plant, wherein the method comprises the steps of obtaining a batch calculation task set to be scheduled, wherein each task in the batch calculation task set to be scheduled is marked with a time window constraint, a resource demand spectrum, a mobility identification and a task priority; based on prospective information provided by a preset digital twin collaborative optimization layer, a power calculation task scheduling optimization model crossing space-time scales is constructed and solved by taking multi-objective optimization of total economic cost and total carbon emission of a system as a guide, a global scheduling blueprint comprising optimal execution time, execution place and resource allocation of each task is obtained, the global scheduling blueprint is decomposed into a real-time scheduling instruction sequence, and the real-time scheduling instruction sequence is issued to a corresponding physical power calculation node for execution, so that power calculation scheduling work is completed. The method completes the cooperative scheduling of the power and the calculation power, and realizes the cooperative idea that the calculation power follows the power or the power follows the calculation power.

Inventors

LI YAJIE
LI YU
CHEN SHUTING
LIANG GANG
LI LUMIN
YANG DAWEI
MING TAO
ZHANG WEI
BAI XIAOJUN
MA BIN

Assignees

国网新疆电力有限公司信息通信公司

Dates

Publication Date: 20260505
Application Date: 20251219

Claims (10)

1. The utility model provides an electricity calculation collaborative integration management method based on virtual power plant, which is characterized in that the method comprises the following steps: Acquiring a batch computing power task set to be scheduled, wherein each task in the batch computing power task set to be scheduled is marked with a time window constraint, a resource demand spectrum, a mobility identification and a task priority; Based on prospective information provided by a preset digital twin collaborative optimization layer, constructing and solving a computational power task scheduling optimization model crossing space-time scales by taking multi-objective optimization of total economic cost and total carbon emission of a system as a guide, and obtaining a global scheduling blueprint comprising optimal execution time, execution place and resource allocation of each task; and decomposing the global scheduling blueprint into a real-time scheduling instruction sequence, and issuing the real-time scheduling instruction sequence to a corresponding physical computing node for execution to complete computing scheduling work.
2. The virtual power plant-based electricity-computing collaborative integration management method of claim 1, further comprising: Constructing and maintaining a digital twin collaborative optimization layer for connecting the power system and the computing center, wherein the digital twin collaborative optimization layer synchronizes and predicts node marginal electricity price, renewable energy output curve, load curve and heterogeneous resource state, task queue and network topology of the power system in real time; And acquiring actual execution feedback data of a physical system, comparing the actual execution feedback data with a predicted track in a preset digital twin collaborative optimization layer, and carrying out on-line calibration on model parameters of the collaborative optimization layer according to deviation to form closed loop optimization.
3. The virtual power plant-based electric computing collaborative integration management method according to any one of claims 1 or 2, wherein the constructing and solving a power computing task scheduling optimization model across a space-time scale specifically comprises: Setting a long-period look-ahead optimization time window and a short-period rolling execution time window; In the prospective optimization time window, taking the weighted sum of the minimum system total operation cost and the carbon emission as a target, applying power grid safe operation constraint, task completion constraint, calculation power resource capacity constraint, task time window and mobility constraint, constructing a mixed integer programming model, and solving to obtain a theoretical optimal global scheduling blueprint: And after entering the rolling execution time window, carrying out re-optimization and fine adjustment on the global scheduling blueprint based on the latest system state information to generate an accurate scheduling instruction of the current execution period.
4. The virtual power plant-based electricity calculation collaborative integration management method according to claim 3, wherein the total system operation cost comprises power cost generated by migration and scheduling of an electricity task, network transmission cost and service quality penalty cost generated by deviation of task completion time from an optimal time window, and the carbon emission is quantified by the product of electric energy consumed by the electricity task and an average carbon emission factor of a regional power grid corresponding to an electric energy source.
5. The virtual power plant-based electricity-computing collaborative integration management method of claim 1, further comprising: Forming an experience sample library according to the historical scheduling decision and the multi-target benefit result generated by the historical scheduling decision; training a deep reinforcement learning module based on the experience sample library, wherein the deep reinforcement learning module directly outputs optimal scheduling strategy suggestions or key parameter configuration of an optimization model under different system state characteristics; and introducing the output of the deep reinforcement learning module as an initial solution or decision guide when solving the computational power task scheduling optimization model crossing the space-time scale each time so as to accelerate the solving process and improve the strategy quality.
6. The virtual power plant-based electric calculation collaborative integration management method according to claim 1, wherein before the digital twin collaborative optimization layer synchronizes and predicts in real time, the method further comprises the step of collecting real-time asynchronous data streams, and performing data fusion and calibration, and the specific process is as follows: acquiring characteristic parameters of a real-time asynchronous data stream, wherein the characteristic parameters comprise data types, data acquisition and transmission delay and data packet size; according to the data type and the corresponding time delay, combining the physical or business change sensitivity of the data type, dynamically determining a time delay influence coefficient, and according to the time delay influence coefficient and the data acquisition and transmission time delay, calculating the state representation deviation amount caused by data asynchronization; the real-time asynchronous data stream is subjected to priority ranking according to the calculated deviation amount, and is dynamically divided into n data batch processing groups, wherein n is more than or equal to 1, and the total data amount of each group is ensured to be within a preset memory processing threshold; According to the historical processing log, time consumption of calibration processing of data packets with different types and sizes is learned and estimated, and then a processing time window is estimated for each newly generated packet to serve as a reference for monitoring the subsequent process.
7. The virtual power plant-based electric calculation collaboration integrated management method as claimed in claim 6, wherein when the calibration processing is performed on the nth group of data, the method specifically comprises: dividing the grouping data into a plurality of data blocks according to a preset size, and sequentially processing the data blocks; monitoring the actual processing time consumption of each data block, and comparing the actual processing time consumption with the estimated time consumption in real time; If the difference value between the actual time consumption and the estimated time consumption is within a reasonable threshold value, judging that the packet data processing is normal, and continuing; If the difference exceeds a reasonable threshold, the anomaly diagnosis is immediately triggered.
8. The virtual power plant-based electricity calculation collaborative integration management method according to claim 7, wherein the specific process of abnormality diagnosis is as follows: once the abnormality is determined, immediately starting a repair sub-flow, and repairing the system; After the repair is completed, the calibration process is restarted automatically from the interrupted packet data block until all the data packets are processed, and the final consistency of the data stream processing is ensured.
9. An electric calculation collaboration integrated management system based on a virtual power plant for implementing the method of any one of claims 1-8, wherein the platform comprises: The computing power task acquisition module is used for acquiring a batch computing power task set to be scheduled, wherein each task in the batch computing power task set to be scheduled is marked with a time window constraint, a resource demand spectrum, a mobility identification and a task priority; The scheduling blueprint generation module is used for constructing and solving a space-time-scale-crossing computational power task scheduling optimization model based on prospective information provided by a preset digital twin collaborative optimization layer and with multi-objective optimization of total economic cost and total carbon emission of the system as a guide to obtain a global scheduling blueprint comprising optimal execution time, execution place and resource allocation of each task; and the execution module is used for decomposing the global scheduling blueprint into a real-time scheduling instruction sequence, and issuing the real-time scheduling instruction sequence to the corresponding physical computing node for execution to complete computing scheduling work.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 8 when executing the computer program.

Description

Electric calculation collaborative integrated management method and system based on virtual power plant Technical Field The invention relates to the technical field of energy and information intersection, in particular to an electric calculation collaborative integrated management method and system based on a virtual power plant. Background With the explosive growth of digital economics, the energy consumption of computing infrastructure such as data centers and intelligent computing centers has become a critical load affecting the balance of power systems. Meanwhile, the new energy power mainly comprising wind power and photovoltaic has volatility and intermittence, and the large-scale grid connection of the new energy power provides challenges for the stable operation of a power grid. Under the background, the deep cooperation of the power grid and the power grid (electric calculation cooperation) is considered as a key path for improving the energy utilization efficiency, guaranteeing the green low carbon of power supply and promoting new energy consumption. Currently, the operating schedules of power and power systems are basically in a "self-administration" state: On the power system side, although the demand side response and virtual power plant technology are promoted and aim at aggregating adjustable industrial, commercial and residential loads, the adjustment objects of the power system side mainly comprise traditional loads such as temperature control and illumination, and the adjustment potential, response accuracy and flexibility are limited. On the computing power system side, the resource scheduling in the data center mainly focuses on the computing efficiency, service level agreements and internal energy consumption, and the electricity consumption is regarded as an unregulated rigid load. The scheduling strategy is completely unhooked from the real-time electricity price, the carbon strength and the new energy output condition of the power grid. This results in a possible concentration of the computational load in time and area operations with high electricity prices and high carbon emissions, both pushing up the operating costs and not facilitating the consumption of renewable energy sources by the grid. Thus, the deep fusion and co-scheduling (i.e., "electric computing synergy") that motivates "electric power grid" and "computing power grid" is considered critical to the breaking of offices. Disclosure of Invention In order to solve the technical problems, the invention provides an electric calculation collaborative integrated management method and system based on a virtual power plant. To solve the above problems, in a first aspect, the present invention provides an electric calculation collaboration integrated management method based on a virtual power plant, where the method includes: Acquiring a batch computing power task set to be scheduled, wherein each task in the batch computing power task set to be scheduled is marked with a time window constraint, a resource demand spectrum, a mobility identification and a task priority; Based on prospective information provided by a preset digital twin collaborative optimization layer, constructing and solving a computational power task scheduling optimization model crossing space-time scales by taking multi-objective optimization of total economic cost and total carbon emission of a system as a guide, and obtaining a global scheduling blueprint comprising optimal execution time, execution place and resource allocation of each task; and decomposing the global scheduling blueprint into a real-time scheduling instruction sequence, and issuing the real-time scheduling instruction sequence to a corresponding physical computing node for execution to complete computing scheduling work. Preferably, the method further comprises: Constructing and maintaining a digital twin collaborative optimization layer for connecting the power system and the computing center, wherein the digital twin collaborative optimization layer synchronizes and predicts node marginal electricity price, renewable energy output curve, load curve and heterogeneous resource state, task queue and network topology of the power system in real time; And acquiring actual execution feedback data of a physical system, comparing the actual execution feedback data with a predicted track in a preset digital twin collaborative optimization layer, and carrying out on-line calibration on model parameters of the collaborative optimization layer according to deviation to form closed loop optimization. Preferably, the method is characterized in that the construction and the solution of the computational power task scheduling optimization model crossing the space-time scale specifically comprise the following steps: Setting a long-period look-ahead optimization time window and a short-period rolling execution time window; In the prospective optimization time window, taking the weighted sum of the min