CN-122026528-A - Double-layer scheduling method, system, equipment and medium for collaborative optimization of data center load and energy storage

CN122026528ACN 122026528 ACN122026528 ACN 122026528ACN-122026528-A

Abstract

The invention relates to the technical field of data center energy management and power dispatching, and discloses a double-layer dispatching method, a double-layer dispatching system, double-layer dispatching equipment and double-layer dispatching media for collaborative optimization of data center load and energy storage. The method comprises the steps of collecting operation data of a data center, preprocessing the operation data, constructing a prediction model to obtain load and electricity price prediction data, formulating a day-ahead scheduling scheme based on the prediction data, solving a first optimization model to generate a first scheduling instruction, executing the first scheduling instruction, monitoring actual deviation, starting a second optimization model to correct the actual deviation in real time when the actual deviation exceeds a threshold value to obtain a second scheduling instruction, carrying out safety inspection on energy storage charging and discharging power, feeding back a safety state to the optimization model to form closed loop control, and starting and stopping a scheduling server and distributing calculation tasks based on a safety inspection result. The invention adopts a double-layer dispatching framework combining daily optimization and real-time correction to realize electric-calculation-storage cooperative optimization, reduce the operation cost, improve the renewable energy source digestion capability and enhance the real-time response and safe operation level of the system.

Inventors

HU JIANHUI
Meng Jingrong
LI YILIANG
HE CHUNXIAO
SUN LIN
HE GUOXIN
ZHANG JIANWEN
HU TONG
HE LINJUAN

Assignees

上海交通大学四川研究院

Dates

Publication Date: 20260512
Application Date: 20260413

Claims (10)

1. The double-layer dispatching method for the collaborative optimization of the load and the energy storage of the data center is characterized by comprising the following steps of, Responding to the result of preprocessing data by the data center, constructing a first prediction model, wherein the first prediction model is used for generating prediction data required by subsequent scheduling; based on the prediction data, a first optimization model targeting an optimal scheduling scheme is established, and the first optimization model is used for generating a first scheduling instruction; Responding to the deviation value of the first scheduling instruction, judging whether to start a second optimization model, wherein the second optimization model is used for generating a second scheduling instruction; Based on the second scheduling instruction, performing safety constraint verification on the stored energy charging and discharging power to obtain a corresponding safety check result; and executing final scheduling operation of the start-stop state and the calculation task allocation of the data center server according to the security check result.
2. The method for dual-layer scheduling of data center load and energy storage co-optimization of claim 1, wherein the specific step of generating the first scheduling instructions by the first optimization model comprises, And a global day-ahead scheduling scheme is formulated based on the prediction data, and is solved through a first optimization model, wherein the first optimization model takes the minimum total running cost as an objective function, and a first scheduling instruction is generated.
3. The method for dual layer scheduling of data center load and energy storage co-optimization of claim 2, wherein determining whether to initiate the second optimization model in response to the deviation value of the first scheduling command comprises, And optimizing according to the deviation value of the actual running data and the first scheduling instruction, and adjusting the day-ahead scheduling scheme to obtain a real-time scheduling scheme in response to the deviation value of the first scheduling instruction exceeding a first threshold value, wherein the second optimization model is based on the real-time scheduling scheme to minimize the correction cost and the penalty of deviating from the day-ahead scheduling scheme and generate a second scheduling instruction.
4. The method for dual-layer scheduling of data center load and energy storage co-optimization of claim 3, wherein the performing a security constraint check on the energy storage charging and discharging power to obtain a corresponding security check result comprises, Based on the second scheduling instruction, the energy storage converter is controlled to execute actual charging and discharging, whether the charging and discharging power exceeds the rated power is checked, whether the battery temperature is normal is checked, and the battery temperature is fed back to the first optimization model, so that a closed loop feedback mechanism of energy storage control is formed.
5. The method for dual-layer scheduling of data center load and energy storage co-optimization of claim 4, wherein performing a final scheduling operation of data center server start-stop status and computing task allocation comprises, And controlling the IT management system switch server through the Ethernet interface according to the second scheduling instruction, supporting the dynamic migration of the virtual machine and the switch of the server by adopting a virtualization technology, selecting a corresponding task from the task queue, and distributing the task to the server for execution.
6. The method for dual layer scheduling of data center load and energy storage co-optimization of claim 5, wherein said responding to data center pre-processing results comprises, And collecting data of the data center in real time through a communication interface, performing checksum pretreatment, and storing the pretreated data, wherein the pretreated data is used for daily prediction of a first prediction model and periodic monitoring of real-time correction of a second optimization model.
7. The method for dual-layer scheduling of data center load and energy storage co-optimization of claim 6, wherein constructing a first predictive model for generating predictive data for subsequent scheduling comprises, And processing the time sequence data through a control mechanism, carrying out network training on the data of the data center by adopting an optimizer, updating the first prediction model to obtain prediction data, wherein the prediction data is used for formulating a global day-ahead scheduling scheme, checking and adjusting the prediction model, and supporting the scheduling correction of the second optimization model.
8. The double-layer scheduling system for collaborative optimization of data center load and energy storage, which applies the double-layer scheduling method for collaborative optimization of data center load and energy storage according to any one of claims 1-7, is characterized by comprising the following steps: the data acquisition and preprocessing module acquires the operation data of the data center in real time through the communication interface, performs data verification and preprocessing, stores the processed data and provides the processed data for the first prediction model; The prediction module is used for constructing and training a first prediction model based on the first preprocessed time sequence data, outputting prediction data of a future period and providing input for optimal scheduling; The optimization module is used for establishing a first optimization model aiming at optimizing system scheduling, solving an optimal scheduling scheme based on predicted data of a future period and generating a first scheduling instruction; the execution and judgment module executes the first scheduling instruction, monitors the deviation value between the actual operation data and the instruction, triggers the second optimization model and generates a second scheduling instruction; The safety check module is used for controlling the energy storage converter to execute charging and discharging operation according to the second scheduling instruction, monitoring whether the charging and discharging power exceeds the limit and whether the battery temperature is normal or not in real time, and feeding back the safety state to the first optimization model to form closed-loop control; and the server and task allocation module is used for issuing a server start-stop instruction to the IT management system through the Ethernet interface, supporting virtual machine dynamic migration, selecting tasks from the task queue according to the optimization instruction and allocating the tasks to corresponding servers for execution.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the data center load and energy storage co-optimized two-tier scheduling method of any one of claims 1 to 7.
10. A computer readable storage medium having stored thereon a computer program, characterized in that the computer program when executed by a processor implements the steps of the data center load and energy storage co-optimized dual layer scheduling method of any of claims 1 to 7.

Description

Double-layer scheduling method, system, equipment and medium for collaborative optimization of data center load and energy storage Technical Field The invention relates to the technical field of data center energy management and power dispatching, in particular to a double-layer dispatching method, system, equipment and medium for collaborative optimization of data center load and energy storage. Background Along with the rapid development of digital economy in China, the energy consumption of a data center is continuously increased, the fluctuation problem of high-proportion renewable energy access to a power grid is prominent, and the intelligent energy scheduling and collaborative optimization technology is focused. However, the current energy management method is difficult to cope with the challenges of computing load rigidity, extensive energy storage utilization, difficult energy consumption of renewable energy sources and the like. The computing load of the data center has schedulable potential, but the existing scheduling strategy and the energy storage system are not in cooperation, an effective combined optimization mechanism is not formed, the flexibility of the system is limited, and the running cost is difficult to optimize. And most of optimization models adopt a single-layer architecture, so that the daily schedule and the real-time schedule are confused, the scheduling efficiency is influenced by different optimization cycle lengths, and long-term economy and short-term real-time response capability are difficult to be considered. In coping with renewable energy fluctuation, the existing research regards the renewable energy fluctuation as uncontrollable external input, relies on passive balance of power exchange of a power grid, increases the burden of the power grid, possibly causes energy waste and lacks a cross-system cooperative mechanism. At present, the field of data center electricity-calculation-storage collaborative optimization lacks a systematic scheme, and has obvious defects in the aspects of economy, renewable energy consumption rate, real-time response capability, overall intelligent scheduling level and the like. Disclosure of Invention In view of the existing problems, the invention provides a double-layer scheduling method, system, equipment and medium for collaborative optimization of data center load and energy storage. Therefore, the invention solves the technical problems of how to combine the calculation load space-time scheduling characteristic of the data center, the charge and discharge flexibility of the energy storage system and the renewable energy source fluctuation stabilization requirement to construct an electric-calculation-storage cooperative optimization system and a day-ahead-real-time double-layer progressive optimization framework so as to realize multi-objective cooperative optimization of data center operation cost minimization, renewable energy source absorption maximization and power grid interaction initiative. In order to solve the technical problems, the invention provides a double-layer scheduling method for collaborative optimization of data center load and energy storage, which comprises the following steps, Responding to the result of preprocessing data by the data center, constructing a first prediction model, wherein the first prediction model is used for generating prediction data required by subsequent scheduling; based on the prediction data, a first optimization model targeting an optimal scheduling scheme is established, and the first optimization model is used for generating a first scheduling instruction; Responding to the deviation value of the first scheduling instruction, judging whether to start a second optimization model, wherein the second optimization model is used for generating a second scheduling instruction; Based on the second scheduling instruction, performing safety constraint verification on the stored energy charging and discharging power to obtain a corresponding safety check result; and executing final scheduling operation of the start-stop state and the calculation task allocation of the data center server according to the security check result. As a preferable scheme of the double-layer scheduling method for collaborative optimization of data center load and energy storage, the method comprises the specific steps of generating a first scheduling instruction by a first optimization model, And a global day-ahead scheduling scheme is formulated based on the prediction data, and is solved through a first optimization model, wherein the first optimization model takes the minimum total running cost as an objective function, and a first scheduling instruction is generated. As a preferable scheme of the double-layer scheduling method for collaborative optimization of data center load and energy storage, the method comprises the steps of responding to the deviation value of a first scheduling instruction to judge whether