Green data center scheduling optimization method under computing power and power coupling

Abstract

The invention discloses a green data center scheduling optimization method under computing power and electric power coupling, data load types of a data center comprise a transferable load, a reducible load and a rigid load, and the method comprises the following steps: constructing a double-target scheduling optimization model by taking the minimum power consumption cost of the data center and the maximum green power fitting degree as targets; the power consumption cost of the data center is used for representing the sum of expenses generated by the data center for purchasing power from a power grid in order to meet the power required for processing the data load in each time period, and the green power fitting degree is used for evaluating the matching degree between the power consumption demand of the data center and the renewable energy source supply; solving the dual-target scheduling optimization model by using a multi-target dogvessel swarm algorithm to obtain a Pareto frontier; and determining an optimal scheduling scheme based on the Pareto frontier, wherein the optimal scheduling scheme comprises a data center power consumption curve, an energy storage power station charging and discharging curve and a transferable load scheduling strategy. According to the invention, the power consumption cost of the data center can be reduced, and the renewable energy consumption capability can be improved.

Description

Technical Field

[0001] The present invention relates to the field of data center technology, and in particular to a green data center scheduling optimization method under computing power and electricity coupling. Background Art

[0002] Green data centers are a key component of building a new power system and play an important role in the development of the power industry. my country has clearly pointed out that the development of data centers must continue to carry out green construction, optimize load operation time periods, and improve the flexible regulation response capability of loads, energy utilization efficiency, and renewable energy utilization. However, current data centers face obstacles to green transformation due to high energy consumption and low clean energy utilization. How to optimize energy use and improve the absorption capacity of renewable energy has become a key bottleneck restricting its transformation. To this end, many scholars have conducted research from multiple aspects such as data center operation characteristics, computing power coupling, and scheduling optimization.

[0003] In terms of data center operation characteristics, existing research mainly focuses on server configuration characteristics and server energy consumption modeling. In terms of server configuration characteristics, optimizing server utilization is an effective way to reduce the total energy consumption of data centers. The number of server configurations is usually determined by the service quality of the peak workload. Therefore, during load flat and valley periods, servers are often not used. Studies have shown that the average utilization of data center servers is only about 15%. In response to this, scholars have adopted methods such as selectively shutting down idle or underutilized servers, dynamic voltage and frequency adjustment, and energy-aware task scheduling to improve energy efficiency. For server energy consumption modeling, the key issue is how to quantitatively describe the relationship between data load and energy consumption. Existing research mainly uses four models, including additive models, baseline load models, simple regression models, and multivariate regression models. The additive model is simple and intuitive, but it ignores the mutual influence between components and the nonlinear characteristics of energy consumption. The baseline load model takes into account the energy consumption characteristics to a certain extent, but the prediction accuracy for scenarios with dynamic changes in energy consumption is poor. The regression model has good prediction accuracy, but the accuracy is too dependent on historical data and is easily affected by overfitting and underfitting. In summary, the four models have their own strengths and should be selected according to the research context.

[0004] In terms of computing power coupling, since data load has adjustable characteristics and there is a linear relationship between the energy consumption of data centers and the amount of load to be processed, the time and space transfer of power demand can be achieved by scheduling data load. Data centers can divide loads into different types for processing according to the differences in data service needs. According to time adjustability, loads can be divided into rigid loads, transferable loads and reducible loads. According to the different characteristics of power grid electricity prices in different time periods, different types of loads can be scheduled in time, and ultimately the purpose of reducing electricity costs can be achieved. In addition to scheduling between different time periods, according to the different amounts of load to be processed in data centers in different regions during the same period, the data center cluster can also be used as the research object to further divide the above three types of data loads into the following four categories: rigid loads, time-transferable loads, spatially transferable loads and reducible loads. Scheduling data loads between different regions will be affected by limitations such as network bandwidth, service delay, and response quality. However, at the same time, according to the differences in time-of-use electricity prices, server operating conditions, etc. in different regions, the overall operation of the data center cluster will be optimized. In addition to utilizing the adjustability of data loads, using new power equipment to coordinate with the power grid to power data centers to optimize power demand is also a part of current research. Some existing technical literatures have studied the scenarios of wind energy, photovoltaics, geothermal energy, natural gas, combined heat and power systems and power grids working together to supply power to data centers, providing new ideas for improving the green electricity consumption capacity of data centers. As a resource with flexible scheduling characteristics, energy storage can effectively reduce the energy cost of data centers through the time-of-use electricity price peak-valley price difference arbitrage strategy, but energy storage equipment with different operating modes participate in power supply in different ways. The way independent energy storage participates in power supply is similar to that of virtual power plants. It is mainly used to reduce the peak and fill the valley of power demand in data centers, or as a backup power source. Shared energy storage provides energy storage capacity leasing services to data centers in the form of a third-party leasing entity independent of the data center and the power grid. This cooperative relationship is usually based on a cooperative game model to ensure the optimal benefits of all parties. Overall, there is great potential for optimizing the energy use strategy of data centers using new power equipment.

[0005] In terms of scheduling optimization, the data center scheduling problem involving computing power and power coupling can usually be decomposed into two sub-problems, namely, data load scheduling problem and power scheduling problem. In existing research, data load scheduling is divided into two types of models: static scheduling and dynamic scheduling. Static scheduling determines the processing volume of various data loads in advance based on historical data, and then schedules. The model solution is relatively simple, but it does not take into account dynamic factors such as the correlation between tasks and the operating status of the server. Dynamic scheduling is more comprehensive. It usually combines the server operation optimization process with the deep reinforcement learning algorithm, and optimizes indicators such as user satisfaction and server energy consumption based on the above dynamic factors. The power scheduling problem mainly studies the power supply architecture of the data center, which has been mentioned in the above part of the coordinated power supply of new power equipment and power grid. There are three main types of algorithms for solving data center scheduling problems: operations research planning algorithm, heuristic algorithm and reinforcement learning algorithm. The operations research method has high solution accuracy, but it is difficult and time-consuming to solve in large-scale optimization models involving coupling. For data centers, most studies use mixed integer programming. Heuristic algorithms can usually only obtain satisfactory solutions, but they are quick to solve complex multi-objective and nonlinear models, so they are more widely used than operations research methods. Common heuristic algorithms include particle swarm algorithms, genetic algorithms, ant colony algorithms, etc. Reinforcement learning algorithms are currently the most widely used type of algorithms. Compared with the first two, reinforcement learning algorithms are more suitable for solving dynamic programming problems with complexity, high uncertainty or incomplete models. For optimization problems in data centers, the model adaptability and solution objectives should be comprehensively considered to select the corresponding algorithm.

[0006] Existing technical literature provides a theoretical basis for the scheduling optimization of green data centers under the coupling of computing power and electricity, but it is insufficient in optimizing multiple objectives such as optimizing the electricity cost of data centers and the ability to absorb renewable energy. How to improve the level of renewable energy absorption while optimizing the electricity cost of data centers and strengthen the deep integration of computing power and electricity is a key issue in the optimization of data center operations.

[0007] Therefore, a green data center scheduling optimization method under computing power and electricity coupling is needed to solve the above technical problems. Summary of the invention

[0008] To this end, the present invention provides a green data center scheduling optimization method under computing power and power coupling to solve or at least alleviate the above problems.

[0009] According to one aspect of the present invention, a method for scheduling and optimizing a green data center under computing power and electricity coupling is provided, wherein the data center includes energy-consuming equipment and an energy storage power station, and is suitable for being jointly powered by a power grid, a distributed power station, and an energy storage power station, wherein the distributed power station includes distributed wind power and distributed photovoltaic power, and the data load types of the data center include transferable loads, reducible loads, and rigid loads, and the method includes: constructing a dual-objective scheduling optimization model with the goal of minimizing the power cost of the data center and maximizing the green electricity fitting degree, wherein the power cost of the data center is used to represent the sum of the costs incurred by the data center in purchasing electricity from the power grid in each time period to meet the power required for processing the data load, and the green electricity fitting degree is used to evaluate the degree of matching between the power demand of the data center and the supply of renewable energy; solving the dual-objective scheduling optimization model using a multi-objective Salp Unica swarm algorithm to obtain a Pareto frontier; determining an optimal scheduling scheme based on the Pareto frontier, wherein the optimal scheduling scheme includes a power curve of the data center, a charging and discharging curve of the energy storage power station, and a transferable load scheduling strategy.

[0010] Optionally, in the green data center scheduling optimization method under computing power and electricity coupling according to the present invention, it also includes: standardizing the data center time-sharing power consumption curve and the external green electricity output curve respectively to obtain corresponding standardized data center time-sharing power consumption curve and standardized external green electricity output curve; and using the cosine similarity between the standardized data center time-sharing power consumption curve and the standardized external green electricity output curve as the green electricity fitting degree.

[0011] Optionally, in the green data center scheduling optimization method under the computing power and electricity coupling according to the present invention, a dual-objective scheduling optimization model is constructed with the goals of minimizing the electricity cost of the data center and maximizing the green electricity fitting degree, including: constructing a first objective function with minimizing the electricity cost of the data center as the first objective; constructing a second objective function with maximizing the green electricity fitting degree as the second objective; and constructing a dual-objective scheduling optimization model based on the first objective function and the second objective function.

[0012] Optionally, in the green data center scheduling optimization method under computing power and electricity coupling according to the present invention, a dual-objective scheduling optimization model is constructed based on the first objective function and the second objective function, including: constructing a dual-objective scheduling optimization model based on the first objective function, the second objective function, data load constraints and power load constraints; wherein the data load constraints include transferable load constraints and reducible load constraints; the power load constraints include energy storage power station charging and discharging constraints, energy storage power station state change times constraints, energy storage power station capacity constraints, and power balance constraints.

[0013] Optionally, in the green data center scheduling optimization method under computing power and power coupling according to the present invention, the transferable load constraint conditions include: the actual transfer load amount after the data center is scheduled in each time period is equal to the initial transferable load amount before the scheduling of the time period plus the transferable load amount scheduled from the previous time period to the time period minus the transferable load amount scheduled from the time period to the subsequent time period, and the transferable load amount scheduled from each time period to the subsequent time period does not exceed the initial transferable load amount of the time period; the reducible load constraint conditions include: the actual load reduction amount of the data center in each time period does not exceed the reducible load amount of the time period, and the total actual load reduction amount of all time periods of the data center cannot exceed the maximum limit of the reducible load, wherein the reducible load amount of each time period is equal to the initial total load amount of the time period multiplied by the reducible load proportion coefficient; the energy storage power station charging and discharging constraint conditions include: each energy storage power station The charging power of each time period shall not exceed the upper limit of the charging power allowed by the energy storage power station, and the discharging power of the energy storage power station in each time period shall not exceed the upper limit of the discharging power allowed by the energy storage power station; the energy storage power station state change number constraint conditions include: the sum of the charging and discharging times of the energy storage power station in each time period cannot exceed the maximum state change number of the energy storage power station, and the energy storage power station cannot charge and discharge at the same time in the same time period; the energy storage power station capacity constraint conditions include: the capacity of the energy storage power station in each time period is greater than or equal to the lower limit of the energy storage power station capacity, and less than or equal to the upper limit of the energy storage power station capacity; and, the initial capacity of the energy storage power station is equal to the capacity after charging and discharging in the last time period; the power balance constraint conditions include: the sum of the total energy consumption of the data center in each time period and the charging and discharging power of the energy storage power station is equal to the sum of the power supply power of the power grid, the power generation power of distributed photovoltaics, and the power generation power of distributed wind power, and the power supply power of the power grid in each time period is a non-negative value.

[0014] Optionally, in the green data center scheduling optimization method under the coupling of computing power and electricity according to the present invention, the dual-objective scheduling optimization model is solved by using a multi-objective salp swarm algorithm to obtain the Pareto frontier, including: using the Monte Carlo method to obtain an initial population, the initial population includes N individuals, including a leader and N-1 followers, N is a positive integer greater than or equal to 3; calculating the first fitness value of each individual in the initial population according to the first objective function, and determining the initial optimal individual in the initial population based on the first fitness value of each individual; adding the initial optimal individual to the food position set, and determining the initial food position corresponding to the initial optimal individual, wherein each food position in the food position set corresponds to a scheduling scheme; based on the initial food position corresponding to the initial optimal individual, updating the leader position and each follower position in the initial population; calculating the second fitness value of each individual in the initial population according to the second objective function, and using the subjective and objective integrated weighting method to determine a new optimal individual based on the first fitness value and the second fitness value of each individual; updating the food position based on the new optimal individual A food position set is configured, comprising: if the food position set is empty, adding the new optimal individual to the food position set; if the food position set is not empty, comparing the new optimal individual with the first fitness value and the second fitness value of each food position in the food position set, respectively, wherein if the first fitness value and the second fitness value of the new optimal individual are both better than the first fitness value and the second fitness value of any food position, replacing any food position with the new optimal individual; if the first fitness value or the second fitness value of the new optimal individual is better than the first fitness value or the second fitness value of each food position, adding the new optimal individual to the food position set; if the first fitness value and the second fitness value of any food position in the food position set are both better than the first fitness value and the second fitness value of the new optimal individual, discarding the new optimal individual; iteratively executing the above steps until the number of iterations reaches the maximum number of iterations or the fitness value of the optimal individual reaches the termination value, obtaining a final food position set, and taking the final food position set as the Pareto front.

[0015] Optionally, in the green data center scheduling optimization method under computing power and electricity coupling according to the present invention, a subjective and objective integrated weighting method is used to determine a new optimal individual based on the first fitness value and the second fitness value of each individual, including: using the subjective and objective integrated weighting method to determine the first comprehensive weight corresponding to the first objective function and the second comprehensive weight corresponding to the second objective function; based on the first fitness value and the second fitness value of each individual, as well as the first comprehensive weight and the second comprehensive weight, to determine the comprehensive fitness value of each individual; based on the comprehensive fitness value of each individual, determine the new optimal individual.

[0016] Optionally, in the green data center scheduling optimization method under the coupling of computing power and electricity according to the present invention, a multi-objective salp swarm algorithm is used to solve the dual-objective scheduling optimization model, including: inputting the initial transferable load of the data center in each time period, the charging status and charging power of the energy storage power station in each time period, the time-of-use electricity price curve, and the external green electricity output curve into the dual-objective scheduling optimization model, and using the multi-objective salp swarm algorithm to solve the dual-objective scheduling optimization model.

[0017] Optionally, in the green data center scheduling optimization method under computing power and power coupling according to the present invention, the energy-consuming equipment of the data center includes IT equipment, air-conditioning and refrigeration equipment, and auxiliary regulating equipment, wherein the sum of the energy consumption of the IT equipment and the air-conditioning and refrigeration equipment accounts for more than 90% of the total energy consumption of the data center.

[0018] According to one aspect of the present invention, there is provided a computing device, comprising: at least one processor; and a memory storing program instructions, wherein the program instructions are configured to be suitable for execution by the at least one processor, and the program instructions include instructions for executing the green data center scheduling optimization method under computing power and electricity coupling as described above.

[0019] According to one aspect of the present invention, a computer program product is provided, comprising a computer program / instruction, wherein the computer program / instruction implements the method as described above when executed by a processor.

[0020] According to one aspect of the present invention, a readable storage medium storing program instructions is provided. When the program instructions are read and executed by a computing device, the computing device executes the green data center scheduling optimization method under computing power and power coupling as described above.

[0021] According to the technical solution of the present invention, a green data center scheduling optimization method under the coupling of computing power and electricity is provided, focusing on the scheduling optimization problems at the two levels of computing power and electricity, and constructing a dual-objective scheduling optimization model with the goals of minimizing the electricity cost of the data center and maximizing the green electricity fitting degree. Then, the MSSA algorithm is used to solve the Pareto frontier, and the optimal scheduling plan of the data center is determined based on the Pareto frontier. In this way, the electricity cost of the data center can be reduced, the renewable energy consumption capacity of the data center can be improved, and the deep integration of computing power and electricity can be strengthened.

[0022] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to achieve the above and related purposes, the present invention describes certain illustrative aspects in conjunction with the following description and the accompanying drawings, which indicate various ways in which the principles disclosed in the present invention can be practiced, and all aspects and their equivalents are intended to fall within the scope of the claimed subject matter. By reading the following detailed description in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present disclosure will become more apparent. Throughout the present disclosure, the same reference numerals generally refer to the same parts or elements.

[0024] Figure 1 A schematic diagram of a data center computing power power coupling scheduling system 100 provided according to an embodiment of the present invention is shown;

[0025] Figure 2 A schematic diagram of a computing device 200 provided according to an embodiment of the present invention is shown;

[0026] Figure 3 A schematic flow chart of a green data center scheduling optimization method 300 under computing power and power coupling provided according to an embodiment of the present invention is shown;

[0027] Figure 4 A schematic diagram of the process of step 320 according to some embodiments of the present invention is shown;

[0028] Figure 5 A schematic diagram of a typical daily data load curve of a data center according to some embodiments of the present invention is shown;

[0029] Figure 6 A schematic diagram of a time-of-use electricity price curve of a power grid according to some embodiments of the present invention is shown;

[0030] Figure 7 A schematic diagram of distributed photovoltaic and distributed wind power generation according to some embodiments of the present invention is shown;

[0031] Figure 8 A schematic diagram of an unscheduled standardized data center power curve and a standardized external green power output curve according to some embodiments of the present invention is shown;

[0032] Fig. 9 A schematic diagram showing the relationship between the optimized data center electricity cost and green electricity fitting degree and algorithm iteration in four scenarios according to some embodiments of the present invention;

[0033] Fig.10 A schematic diagram showing the fitting of a data center power consumption curve and an external green power output curve in four scenarios according to some embodiments of the present invention is shown;

[0034] Fig.11A schematic diagram of the Pareto front obtained according to some embodiments of the present invention is shown;

[0035] Fig.12 A schematic diagram showing the scheduling of data load and power load of a data center throughout the day after scheduling according to some embodiments of the present invention is shown;

[0036] Fig.13 A schematic diagram showing the relationship between electricity cost, green electricity fitting degree and transferable load ratio coefficient;

[0037] Fig.14 A schematic diagram showing the relationship between electricity cost, green electricity fitting degree and energy storage capacity;

[0038] Fig.15 A schematic diagram showing the relationship between electricity cost, green electricity fitting degree and peak electricity price change ratio. DETAILED DESCRIPTION

[0039] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0040] In view of the problems existing in the prior art in optimizing multi-objective optimization such as optimizing the electricity cost of data centers and the renewable energy absorption capacity, an embodiment of the present invention proposes a green data center scheduling optimization method under the coupling of computing power and electricity, constructs a dual-objective scheduling optimization model with the goals of minimizing the electricity cost of data centers and maximizing the green electricity fitting degree, and uses the MSSA algorithm to solve the Pareto frontier to obtain the optimal scheduling plan of the data center. In this way, the optimal scheduling plan determined according to the present invention can not only reduce the electricity cost of the data center, but also improve the renewable energy absorption capacity of the data center.

[0041] The green data center scheduling optimization method under computing power and power coupling provided by the embodiment of the present invention can be used in a data center computing power and power coupling scheduling system.

[0042] Figure 1 A schematic diagram of a data center computing power power coupling scheduling system 100 provided according to an embodiment of the present invention is shown.

[0043] like Figure 1As shown, the data center computing power coupling scheduling system 100 includes a data center (i.e., a green data center), a distributed power station, and a power grid. Among them, the distributed power station includes distributed wind power and distributed photovoltaic. The data center includes energy-consuming equipment in the computer room and an energy storage power station. Specifically, the main energy-consuming equipment in the data center includes, for example, IT equipment, air-conditioning equipment (air-conditioning and refrigeration equipment), and other auxiliary regulating equipment.

[0044] like Figure 1 As shown, in an embodiment of the present invention, the data center can be powered by a power grid, a distributed power station (including distributed wind power, distributed photovoltaic power), and an energy storage power station. Among them, the power grid is the main source of electricity for the data center. The distributed power station can supply a small amount of power to the data center, mainly used to reduce the cost of purchasing electricity from the power grid during the peak and peak periods of power grid demand. On the one hand, the energy storage power station is used to charge during the flat and valley periods of power grid demand, and discharge during the peak and peak periods of power grid demand, so as to achieve the purpose of reducing the electricity cost of the data center. On the other hand, different charging and discharging strategies can be selected to affect the renewable energy absorption capacity of the data center.

[0045] In the embodiment of the present invention, according to the different response times of the data center in providing demand response services to customers, the data load of the data center can be divided into transferable load, reducible load and rigid load. That is, the data load types of the data center include transferable load, reducible load and rigid load.

[0046] In an embodiment of the present invention, the green data center scheduling optimization method 300 under the coupling of computing power and electricity can be executed in a computing device to reduce the electricity cost of the data center while improving the renewable energy consumption capacity. The green data center scheduling optimization method 300 under the coupling of computing power and electricity of the present invention will be described below.

[0047] A computing device 200 provided by an embodiment of the present invention is introduced below.

[0048] Figure 2 FIG. 2 shows a schematic diagram of a computing device 200 provided according to an embodiment of the present invention. Figure 2 As shown, in a basic configuration, the computing device 200 includes at least one processing unit 202 and a system memory 204. According to one aspect, depending on the configuration and type of the computing device, the processing unit 202 can be implemented as a processor. The system memory 204 includes, but is not limited to, volatile storage (e.g., random access memory), non-volatile storage (e.g., read-only memory), flash memory, or any combination of such memories. According to one aspect, the system memory 204 includes an operating system 205.

[0049] According to one aspect, operating system 205 is suitable for controlling the operation of computing device 200, for example. Furthermore, examples are practiced in conjunction with graphics libraries, other operating systems, or any other application programs, and are not limited to any particular application or system. Figure 2 The basic configuration is shown in FIG. 2 by those components within the dashed lines. According to one aspect, the computing device 200 has additional features or functionality. For example, according to one aspect, the computing device 200 includes additional data storage devices (removable and / or non-removable), such as magnetic disks, optical disks, or tapes. Such additional storage Figure 2 2 is illustrated by a removable storage device 209 and a non-removable storage device 210.

[0050] As stated above, according to one aspect, a program module 203 is stored in the system memory 204. According to one aspect, the program module 203 may include one or more application programs, and the present invention does not limit the type of application program, for example, the application program may include: email and contact application programs, word processing application programs, spreadsheet application programs, database application programs, slide show application programs, drawing or computer-aided application programs, web browser application programs, etc.

[0051] In an embodiment according to the present invention, the program module 203 includes a plurality of program instructions for executing the green data center scheduling optimization method 300 under computing power and electricity coupling of the present invention.

[0052] According to one aspect, the examples may be practiced on a circuit comprising discrete electronic components, a packaged or integrated electronic chip containing logic gates, a circuit utilizing a microprocessor, or a single chip containing electronic components or a microprocessor. Figure 2 Each or many components shown in can be integrated into a system on a chip (SOC) on a single integrated circuit to practice examples. According to one aspect, such a SOC device may include one or more processing units, a graphics unit, a communication unit, a system virtualization unit, and various application functions, all of which are integrated (or "burned") into a chip substrate as a single integrated circuit. When operated via SOC, the functions described in the present invention can be operated via a dedicated logic integrated with other components of the computing device 200 on a single integrated circuit (chip). Other technologies capable of performing logical operations (such as AND, OR, and NOT) can also be used to practice embodiments of the present invention, including but not limited to mechanical, optical, fluid, and quantum technologies. In addition, embodiments of the present invention can be practiced in a general-purpose computer or in any other circuit or system.

[0053] According to one aspect, the computing device 200 may also have one or more input devices 212, such as a keyboard, a mouse, a pen, a voice input device, a touch input device, etc. Output devices 214 may also be included, such as a display, a speaker, a printer, etc. The aforementioned devices are examples and other devices may also be used. The computing device 200 may include one or more communication connections 216 that allow communication with other computing devices 218. Examples of suitable communication connections 216 include, but are not limited to: RF transmitters, receivers and / or transceiver circuits; Universal Serial Bus (USB), parallel and / or serial ports.

[0054] The term computer-readable medium as used in the present invention includes computer storage media. Computer storage media may include volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (e.g., computer-readable instructions, data structures, or program modules). System memory 204, removable storage device 209, and non-removable storage device 210 are all examples of computer storage media (i.e., memory storage). Computer storage media may include random access memory (RAM), read-only memory (ROM), electrically erasable read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical storage, cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other products that can be used to store information and can be accessed by computing device 200. According to one aspect, any such computer storage medium may be part of computing device 200. Computer storage media do not include carrier waves or other propagated data signals.

[0055] According to one aspect, communication media is implemented by computer readable instructions, data structures, program modules, or other data in a modulated data signal (e.g., a carrier wave or other transport mechanism), and includes any information delivery media. According to one aspect, the term "modulated data signal" describes a signal that has one or more characteristics set or changed in a manner that encodes information in the signal. By way of example and not limitation, communication media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, radio frequency (RF), infrared, and other wireless media.

[0056] In an embodiment according to the present invention, the computing device 200 is configured to execute the green data center scheduling optimization method 300 under computing power power coupling. The computing device 200 includes one or more processors and one or more readable storage media storing program instructions. When the program instructions are configured to be executed by one or more processors, the computing device executes the green data center scheduling optimization method 300 under computing power power coupling in the embodiment of the present invention.

[0057] In some embodiments, the computing device 200 that executes the green data center scheduling optimization method 300 under computing power and power coupling in the embodiment of the present invention may be a terminal or a server.

[0058] The following is a detailed description of the green data center scheduling optimization method 300 under computing power and power coupling in an embodiment of the present invention.

[0059] Figure 3 FIG. 3 is a flow chart showing a method 300 for optimizing scheduling of a green data center under computing power and power coupling according to an embodiment of the present invention. Figure 3 As shown, the green data center scheduling optimization method 300 under computing power and power coupling includes the following steps 310 to 330.

[0060] Step 310: The computing device 200 constructs a dual-objective scheduling optimization model with the goal of minimizing the electricity cost of the data center and maximizing the green electricity fit.

[0061] In the embodiment of the present invention, the data center can be powered by the power grid, distributed power stations, and energy storage power stations. Distributed power stations include distributed wind power and distributed photovoltaic power. The data load types of the data center include transferable loads, curtailable loads, and rigid loads.

[0062] It should be noted that the electricity cost of a data center is the sum of the costs incurred by the data center in purchasing electricity from the power grid in each period to meet the power required to process data loads. The electricity cost of a data center is an important economic indicator for measuring the operating efficiency of a data center, and minimizing the electricity cost is one of the important goals of optimizing data center operations. The electricity cost of a data center is not only related to electricity demand, but also affected by fluctuations in time-of-use electricity prices. By optimizing the data load scheduling strategy and flexibly adjusting the daily electricity strategy, the data center's dependence on peak power from the power grid can be reduced, thereby reducing the cost of purchasing electricity. Minimizing electricity costs not only helps to improve the economic benefits of data centers, but also relieves pressure on the power grid and promotes more efficient and sustainable energy use.

[0063] Green electricity fit is used to evaluate the degree of match between the electricity demand of data centers and the supply of renewable energy (external green electricity output), and is a key indicator for measuring the renewable energy utilization rate of data centers. By calculating the cosine similarity between the electricity demand of data centers and the supply of renewable energy (external green electricity output), the fit between the two in time and trend can be quantified. Among them, when the green electricity fit reaches 1, it means that the electricity demand of the data center is completely synchronized with the supply of renewable energy, and the data center can maximize the use of renewable energy. This high level of matching is of great significance to promoting the development of green data centers. Therefore, maximizing the green electricity fit is an important goal of green data center operation optimization and a key measure for achieving sustainable development of data centers.

[0064] Step 320 , the computing device 200 may use a multi-objective salp swarm algorithm (MSSA algorithm) to solve the dual-objective scheduling optimization model to obtain a Pareto front.

[0065] Step 330: The computing device 200 may determine an optimal scheduling solution based on the Pareto front. Specifically, after solving the Pareto front, an optimal scheduling solution on the Pareto front may be selected. The optimal scheduling solution includes a data center power consumption curve, an energy storage power station charge and discharge curve, and a transferable load scheduling strategy.

[0066] In the embodiment of the present invention, according to the different response times of the data center in providing demand response services to customers, the data load of the data center can be divided into transferable load, reducible load and rigid load. That is, the data load types of the data center include transferable load, reducible load and rigid load.

[0067] It should be noted that the transferable load (SDL), also known as batch processing load, refers to a data load with a fixed task volume but flexible processing time within a certain time range. The processing method is batch processing, that is, the transferable load in each time period can only be scheduled to the subsequent time period to adjust the electricity demand curve. The inherent transferable load (initial transferable load) of each time period of the data center can be obtained by multiplying the total load of the time period by the transferable load ratio coefficient, and can be used as the maximum transfer amount of the time period. Since the transferable load can only be scheduled to the subsequent time period, the actual transfer load after each time period is equal to: the initial transferable load before the scheduling of the time period plus the transferable load scheduled from the previous time period to the time period minus the transferable load scheduled from the time period to the subsequent time period.

[0068] In some embodiments, the final transferable load of the data center in each time period (the actual transfer load after scheduling) can be calculated by the following formula.

[0069]

[0070] In the formula, represents the initial transferable load in period t; β SDL represents the transferable load proportional coefficient; Represents the initial total load during period t. represents the final transferable load in period t (i.e., the actual transfer load after scheduling in period t); represents the load that can be transferred from the t' period to the t period; It represents the load that can be transferred from period t to period t”.

[0071] Reducible load (RDL) refers to the data load that can be reduced by reducing the quality of demand response service. The data center can adjust the reducible load according to its electricity demand to reduce the data load to be processed. The reducible load in each period can be obtained by multiplying the initial total load of the period by the reducible load ratio coefficient. Due to the limitation of the minimum service quality, the data center cannot completely reduce the reducible load. Therefore, the actual load reduction amount of the data center in each period cannot exceed the reducible load upper limit (the reducible load amount of the period), and the total actual load reduction amount of all periods of the data center in a scheduling cycle cannot exceed the maximum reducible load limit of the data center.

[0072] In some embodiments, the amount of load that can be reduced in each time period of the data center can be calculated using the following formula.

[0073]

[0074]

[0075] In the formula, represents the load that can be reduced in period t; β RDL Indicates the load reduction ratio factor; Represents the initial total load during period t. represents the actual load reduction during period t; Indicates the maximum limit to which load can be reduced.

[0076] Rigid load (RGDL) refers to data load that has high requirements on processing quality and time, must be processed in the current period, and cannot be transferred or reduced.

[0077] In some embodiments, the rigid load amount in each time period can be calculated by the following formula (7).

[0078]

[0079] In the formula, Indicates the rigid load during period t; β SDLrepresents the transferable load ratio coefficient; β RDL Indicates the load reduction ratio factor; Represents the initial total load during period t.

[0080] In some embodiments, by dividing the data load types of each time period of the data center, the final total load that the data center needs to handle in each time period is equal to: the initial total load plus the load scheduled from the previous time period to the time period minus the load scheduled from the time period to the subsequent time period minus the actual load reduction in the time period. That is, the final total load that needs to be handled in each time period can be calculated by the following formula (8).

[0081]

[0082] Where D t It represents the final total load that the data center needs to handle during period t. represents the initial total load during period t; represents the load that can be transferred from the t' period to the t period; represents the load that can be transferred from period t to period t”; Indicates the actual load reduction during period t.

[0083] In some embodiments, the energy consumption equipment of the data center includes IT equipment, air conditioning and refrigeration equipment, and auxiliary regulating equipment, wherein the ratio of the sum of the energy consumption of the IT equipment and the air conditioning and refrigeration equipment to the total energy consumption of the data center (i.e., the energy consumption ratio of the data center) is greater than 90%. Considering that the energy consumption of the auxiliary regulating equipment accounts for a small proportion, the total energy consumption of the data center can be expressed in an additive form as the following formula (9), i.e., the total energy consumption model of the data center, so as to calculate the total energy consumption of the data center in each period according to the total energy consumption model of the data center.

[0084]

[0085] In the formula, represents the total energy consumption of the data center during period t (kW); represents the energy consumption of IT equipment during period t (kW); represents the energy consumption of air conditioning and refrigeration equipment during period t (kW); η DC It indicates the energy consumption ratio of the data center, that is, the ratio of the sum of the energy consumption of IT equipment and air-conditioning and refrigeration equipment to the total energy consumption of the data center.

[0086] In some embodiments of the present invention, the relationship between IT equipment energy consumption and data load is analyzed, wherein the IT equipment power (energy consumption) is proportional to the cube of the central processing unit operating frequency and is proportional to the final total load that the data center needs to handle in time period t. Based on this, an IT equipment energy consumption model can be established, as shown in the following formula (10), so as to calculate the electrical power (energy consumption) of the IT equipment in each time period according to the IT equipment energy consumption model.

[0087]

[0088] In the formula, represents the energy consumption of IT equipment in time period t (kW); k represents the dynamic energy consumption coefficient; f represents the CPU operating frequency (GHz); μ represents the average service rate of each IT equipment in the data center; D t It represents the final total load that the data center needs to handle during period t.

[0089] In addition, to ensure that IT equipment is in normal operation, it is necessary to keep the operating temperature of IT equipment no higher than a certain temperature. Therefore, the data center must be equipped with sufficient air-conditioning and refrigeration equipment to balance the indoor temperature. The air-conditioning and refrigeration equipment converts electrical energy into low-temperature internal energy through a compressor. The energy consumption model of the air-conditioning and refrigeration equipment is established based on the linear relationship between the electrical power of IT equipment and the cooling power of the air-conditioning and refrigeration equipment, as shown in the following formula (11). The energy consumption model of the air-conditioning and refrigeration equipment can then be used to calculate the electrical power (energy consumption) of the air-conditioning and refrigeration equipment in each period.

[0090]

[0091] In the formula, represents the energy consumption of air conditioning and refrigeration equipment during period t (kW); represents the energy consumption of IT equipment during period t (kW); ε represents the conversion coefficient, which is the ratio of the cooling power of air-conditioning equipment to the electrical power of IT equipment; η ACS It represents the energy efficiency coefficient of air-conditioning and refrigeration equipment, that is, the ratio of the cooling power of air-conditioning and refrigeration equipment to the electrical power (energy consumption) of air-conditioning and refrigeration equipment.

[0092] An energy storage power station is a type of equipment used to store energy. It can convert electrical energy into other forms of energy (such as internal energy, chemical energy) through energy storage batteries and store them. When needed, other forms of energy can be converted into electrical energy and released to facilitate regulating electricity demand. The energy storage power station model established according to some embodiments of the present invention is shown in the following formulas (12) and (13), where the charging and discharging power of the energy storage power station in each time period can be calculated according to formula (12), and the capacity (electricity) of the energy storage power station in each time period can be calculated according to formula (13).

[0093]

[0094] In the formula, Indicates the charging and discharging power of the energy storage power station during period t (kW); represents the charging power of the energy storage power station during period t (kW); E represents the discharge power of the energy storage power station during period t (kW). t represents the capacity of the energy storage power station during period t (kWh); η ES / c Represents the charging efficiency of the energy storage power station; η ES / d Indicates the discharge efficiency of the energy storage power station.

[0095] Distributed photovoltaic and distributed wind power can generate a small amount of electricity in each period. This part of electricity has the characteristics of being used immediately and difficult to store. By using distributed photovoltaic and distributed wind power to supply a small amount of electricity to the data center, the power purchase of the data center from the power grid can be slightly reduced, thereby reducing the electricity cost. The power generation of distributed power stations is affected by environmental factors. The power generation of distributed photovoltaic in each period can be calculated according to the following formula (14), and the power generation of distributed wind power in each period can be calculated according to the following formula (15).

[0096]

[0097] In the formula, represents the power generation of distributed photovoltaic power in period t (kW); N S represents the number of photovoltaic panels included in distributed photovoltaics; A represents the surface area of ​​a single photovoltaic panel (m 2 );η S Represents the energy conversion efficiency of photovoltaic cells; G t Indicates the solar radiation intensity during period t (W / m 2 );PR represents the performance ratio of distributed photovoltaic system.

[0098] represents the power generation of distributed wind power in period t (kW); N W Represents the number of wind turbines included in distributed wind power; η W represents the energy conversion efficiency of the wind turbine; ρ represents the air density (kg / m 3 );R represents the radius of the wind turbine blade (m); v t Indicates the wind speed perpendicular to the fan blades (m / s).

[0099] It is worth noting that in order to improve the data center's ability to absorb renewable energy, existing research usually selects minimizing carbon emissions as the optimization goal. This method aims to indirectly increase the utilization rate of green electricity by reducing the use of thermal power. However, since carbon emissions can only reflect the total amount of thermal power used by the data center in a scheduling cycle, and does not take into account the changing trend of the total green power generation in a scheduling cycle, there are certain defects in using carbon emissions as a goal to improve the green power absorption capacity.

[0100] According to some embodiments of the present invention, taking into account the changing trend of the total green electricity generation over time, the green electricity fitting degree is defined as the cosine similarity between the standardized data center time-sharing electricity consumption curve after scheduling and the standardized external green electricity output curve. Based on this, the green electricity fitting degree can be calculated according to the following formula (16). It should be noted that cosine similarity is a commonly used measurement method used to calculate the similarity of the directions of two vectors in a high-dimensional space. For two curves, cosine similarity can measure the consistency of the trends of the two curves and is not affected by the absolute distance between the two curves. The present invention uses cosine similarity to characterize the fitting degree. The closer the cosine similarity is to 1, the closer the two curves are, that is, the higher the green electricity fitting degree is, which can reflect to a certain extent that the proportion of data center power grid electricity consumption in renewable energy output is larger.

[0101]

[0102] In the formula, represents the standardized power consumption of the data center grid during period t; Represents the standardized external wind and solar power generation power (external green power) during period t.

[0103] It is worth noting that before calculating the cosine similarity, the data center time-sharing power consumption curve (the data center time-sharing power consumption curve after scheduling) and the external green power output curve need to be standardized to obtain the corresponding standardized data center time-sharing power consumption curve and standardized external green power output curve. Furthermore, the cosine similarity of the standardized data center time-sharing power consumption curve and the standardized external green power output curve can be calculated, and the cosine similarity can be used as the green power fitting degree. The green power fitting degree obtained in this way can reflect the proportion of power consumption of the data center power grid in the output of renewable energy. Among them, the higher the green power fitting degree, to a certain extent, it can reflect the greater proportion of power consumption of the data center power grid in the output of renewable energy.

[0104] In some embodiments, the normalization process of the data center time-of-use power consumption curve and the external green power output curve is shown in formulas (17) and (18). The normalization process can eliminate the influence of the dimension on the one hand, and ensure that the calculation result is not affected by the absolute value or offset of the curve on the other hand, so that the calculation result can better reflect the similarity of the trends of the two curves.

[0105]

[0106] In the formula, represents the standardized power consumption of the data center grid during period t; It represents the power consumption of the data center power grid during period t; Indicates the minimum value of the power consumption of the data center power grid; Indicates the maximum power consumption of the data center power grid.

[0107] represents the standardized external wind and solar power generation power during period t (standardized external green power during period t); Represents the external wind and solar power generation power in period t (external green power in period t); Indicates the minimum value of external wind and solar power generation (minimum value of external green power); Indicates the maximum value of external wind and solar power generation (maximum value of external green power).

[0108] In some embodiments, in step 310, a dual-objective scheduling optimization model can be constructed in the following manner: taking the minimum electricity cost of the data center as the first objective, constructing a first objective function; taking the maximum green electricity fitting degree as the second objective, constructing a second objective function; and then, the dual-objective scheduling optimization model can be constructed based on the first objective function and the second objective function. Based on this, the constructed dual-objective scheduling optimization model includes the first objective function and the second objective function.

[0109] According to an embodiment of the present invention, the electricity cost of a data center is the sum of the costs incurred by the data center in purchasing electricity from the power grid in each period to meet the electricity required for processing data loads. The electricity cost of a data center is an important economic indicator for measuring the operating efficiency of a data center, and minimizing the electricity cost is one of the important goals of optimizing the operation of a data center. The electricity cost of a data center is not only related to electricity demand, but also affected by fluctuations in time-of-use electricity prices. By optimizing the data load scheduling strategy and flexibly adjusting the daily electricity strategy, the data center's dependence on peak power from the power grid can be reduced, thereby reducing the cost of purchasing electricity. Minimizing electricity costs not only helps to improve the economic benefits of the data center, but also relieves pressure on the power grid and promotes more efficient and sustainable energy use.

[0110] Based on this, the first objective function constructed with the minimum electricity cost of the data center as the first goal is shown in the following formula (19):

[0111]

[0112] In the formula, represents the power consumption of the data center power grid during period t; δ tIt represents the electricity purchase price of the power grid during period t (yuan / kWh).

[0113] According to an embodiment of the present invention, the green electricity fitting degree is used to evaluate the degree of match between the electricity demand of the data center and the renewable energy supply, and is a key indicator for measuring the renewable energy utilization rate of the data center. By calculating the cosine similarity between the electricity demand of the data center and the renewable energy supply, the fit between the two in time and trend can be quantified. Among them, when the green electricity fitting degree reaches 1, it means that the electricity demand of the data center is completely synchronized with the renewable energy supply, and the data center can maximize the use of renewable energy. This high level of matching is of great significance to promoting the development of green data centers. Therefore, maximizing the green electricity fitting degree is an important goal of green data center operation optimization, and it is also a key measure for achieving sustainable development of data centers.

[0114] The green electricity fitting degree can be calculated according to the above formula (16). The second objective function constructed with the maximum green electricity fitting degree as the second objective is shown in the following formula (20):

[0115] Fitness=max(ρ) (20)

[0116] In some embodiments, the dual-objective scheduling optimization model also includes data load constraints and power load constraints. When constructing a dual-objective scheduling optimization model, on the basis of minimizing the electricity cost of the data center and maximizing the green electricity fit (minimizing the electricity cost of the data center as the first goal, constructing the first objective function; maximizing the green electricity fit as the second goal, constructing the second objective function), data load constraints and power load constraints can also be considered. Specifically, the dual-objective scheduling optimization model can be constructed based on the first objective function, the second objective function, the data load constraints and the power load constraints.

[0117] The data load constraints include transferable load constraints and reducible load constraints.

[0118] It should be noted that the transferable load (SDL), also known as batch processing load, refers to a data load with a fixed task volume but flexible processing time within a certain time range. The processing method is batch processing, that is, the transferable load of each time period can only be scheduled to the subsequent time period to adjust the electricity demand curve. The transferable load constraint conditions are as follows: the actual transfer load amount of each time period after the data center is scheduled should be equal to the initial transferable load amount before the scheduling of the time period plus the transferable load amount scheduled from the previous time period to the time period minus the transferable load amount scheduled from the time period to the subsequent time period, and the transferable load amount scheduled from each time period to the subsequent time period cannot exceed the initial transferable load amount of the time period (that is, the upper limit of the transferable load amount of the time period). In some embodiments, the transferable load constraint conditions are shown in the above formulas (1)-(3).

[0119] The reducible load is a data load that does not have strict service quality requirements. Reducing such data load within a certain limit will not significantly affect the normal operation of the core business of the green data center. The reducible load of the data center in each time period is equal to the initial total load of the time period multiplied by the reducible load ratio coefficient. The reducible load constraint conditions are as follows: the actual load reduction of the data center in each time period cannot exceed the reducible load of the time period (the upper limit of the reducible load), and in order to ensure that the service quality is not too low, the total actual load reduction of the data center in all time periods within a scheduling cycle cannot exceed its maximum reducible load limit. In some embodiments, the reducible load constraint conditions are specifically shown in the above formulas (4)-(6).

[0120] The power load constraints specifically include energy storage power station charging and discharging constraints, energy storage power station state change times constraints, energy storage power station capacity constraints, and power balance constraints.

[0121] It should be noted that the charging and discharging constraints of the energy storage station are as follows: the charging power of the energy storage station in each period cannot exceed the upper limit of the charging power allowed by the energy storage station, and the discharging power of the energy storage station in each period cannot exceed the upper limit of the discharging power allowed by the energy storage station. The specific charging and discharging constraints of the energy storage station are shown in formulas (21)-(22).

[0122]

[0123]

[0124] In the formula, represents the charging power of the energy storage power station during period t (kW); Indicates the charging status of the energy storage power station during period t; Indicates the maximum charging power of the energy storage station (kW).

[0125] Indicates the discharge power of the energy storage power station during period t (kW); Indicates the discharging state of the energy storage power station during period t; Indicates the maximum discharge power of the energy storage power station (kW).

[0126] The constraints on the number of state changes of the energy storage power station are as follows: The sum of the number of charging and discharging times of the energy storage power station in each period cannot exceed the given upper limit N max (maximum number of state changes of the energy storage station), as shown in formula (23), this constraint can prevent the energy storage station from losing performance and life due to frequent charging and discharging in one day. In addition, the energy storage station cannot charge and discharge at the same time in the same period, that is, the energy storage station can only choose one of the three states of charging, discharging, and neither charging nor discharging in each period, as shown in formula (24).

[0127]

[0128] In the formula, Indicates the charging status of the energy storage power station during period t; Indicates the discharging state of the energy storage power station during period t; N max Indicates the maximum number of state changes of the energy storage power station.

[0129] in, Taking 1 means that the energy storage power station is charged during period t. Taking 0 means that the energy storage power station does not charge during period t; Taking 1 means that the energy storage power station discharges during period t. Taking 0 means that the energy storage power station does not discharge during period t.

[0130] The capacity constraint of the energy storage power station is as follows: the capacity of the energy storage power station after charging and discharging in each period cannot be higher than the maximum capacity (the upper limit of the energy storage power station capacity) nor lower than the minimum specified capacity (the lower limit of the energy storage power station capacity), that is, the capacity of the energy storage power station in each period is greater than or equal to the lower limit of the energy storage power station capacity and less than or equal to the upper limit of the energy storage power station capacity, as shown in formula (25). This constraint can ensure the utilization rate of the energy storage power station, so that it can fully participate in the dispatching at the power level and play the role of "peak shaving and valley filling". In addition, the capacity constraint of the energy storage power station also includes the initial and final capacity state constraint of the energy storage power station, that is, the initial capacity of the energy storage power station must be equal to the capacity after charging and discharging in the last period, as shown in formula (26). This constraint can ensure that the energy storage power station can operate continuously and stably under this dispatching scheme.

[0131] E min ≤E t ≤E max (25)

[0132] E0=E T(26)

[0133] In the formula, E t represents the capacity of the energy storage power station during period t (kWh); E min Indicates the lower limit of energy storage power station capacity (kWh); E max Indicates the upper limit of the energy storage power station capacity (kWh); E0 indicates the initial capacity of the energy storage power station (kWh); E T Indicates the capacity (kWh) of the energy storage station after charging and discharging in the last period.

[0134] The power balance constraint is an important constraint that links the power consumption of the green data center, the power supply power of the grid, the power of the energy storage power station and the power supply power of the distributed power station. The power balance constraint conditions are as follows: the power demand and supply of each time period should satisfy the constraint of formula (27), that is, the sum of the total energy consumption of the data center and the charging and discharging power of the energy storage power station in each time period should be equal to the sum of the power supply power of the grid (that is, the power consumption of the data center grid) and the power generation power of wind and solar power (including the power generation power of distributed photovoltaic power stations and the power generation power of distributed wind power). In addition, since the power generated by the distributed power station during the period of high wind and solar power generation may fully meet the power consumption of the data center and the charging of the energy storage power station, the data center does not need to purchase electricity from the grid during these periods. Therefore, formula (28) can be set to ensure that the power supply power of the grid in each period is non-negative.

[0135]

[0136] In the formula, represents the total energy consumption of the data center during period t (kW); Indicates the charging and discharging power of the energy storage power station during period t (kW); represents the power consumption of the data center power grid during period t (that is, the power supplied by the power grid to the data center); represents the power generation of distributed photovoltaic in period t (kW); It represents the power generation of distributed wind power in period t (kW).

[0137] In some embodiments, after constructing the dual-objective scheduling optimization model, in step 320, the initial transferable load of each time period of the data center, the charging state and charging power of each time period of the energy storage power station, the time-of-use electricity price curve, the external green electricity output curve and the energy storage power station parameters can be input into the dual-objective scheduling optimization model, and the multi-objective salp swarm (MSSA) algorithm can be used to solve the dual-objective scheduling optimization model. The dual-objective scheduling optimization model is solved using the multi-objective salp swarm algorithm to obtain the Pareto frontier of the dual-objective scheduling optimization model. Furthermore, in step 330, the optimal scheduling plan can be determined based on the Pareto frontier (an optimal scheduling plan is selected from the Pareto frontier), and the optimal scheduling plan can specifically include the power consumption curve of the data center, the charging and discharging curve of the energy storage power station, and the transferable load scheduling strategy.

[0138] It should be noted that the salp group (including multiple salp individuals) forages in the form of a chain, with the leading individual (leader) at the front end, and the individuals behind following the leader. The present invention refers to the individuals in this following behavior as followers. In practical applications, the number of leaders of the MSSA algorithm can be set according to actual needs. The more leaders there are, the stronger the randomness of the algorithm, but the stability of the algorithm will also be reduced. In an embodiment of the present invention, since there are many variables in the dual-objective scheduling optimization model, in order to obtain a more stable optimal solution, only one individual in the salp group can be selected as the leader.

[0139] Figure 4 FIG. 4 is a flow chart showing step 320 according to some embodiments of the present invention. Figure 4 As shown, the process of solving the dual-objective scheduling optimization model using the multi-objective salp swarm algorithm (MSSA) in step 320 further includes the following steps 321 to 327:

[0140] Step 321: Obtain an initial population using the Monte Carlo method.

[0141] It should be noted that, usually, the initial population is based on the upper and lower bounds of each dimension, which can be obtained by randomization using the following formula (29). However, it is difficult to obtain a feasible initial solution using this initialization method in a model with many constraints.

[0142] Therefore, the present invention uses the Monte Carlo method to obtain the initial population (initial solution) in step 320. Specifically, the Monte Carlo method is first used to randomly select a part of the variables as 0, and then the remaining variables are randomized according to formula (29). This method can obtain multiple feasible solutions through multiple screenings, which are respectively used as the positions of multiple individuals in the initial population.

[0143] X N×D =rand(N,D)×(ub-lb)+lb (29)

[0144] Where, X 1d is the state of each dimension of the leader, X md is the state of each dimension of the follower. Where d=1,2,···,D, represents the dimension of the leader. m=2,3,···,N, represents the ordinal number of the follower.

[0145] It should be understood that the initial population includes N individuals, among which there is a leader and N-1 followers. N can be a positive integer greater than or equal to 3.

[0146] Step 322, calculate the first fitness value of each individual (a total of N individuals) in the initial population, and determine the initial optimal individual in the initial population based on the first fitness value of each individual. Specifically, the first fitness value of each individual in the initial population can be calculated according to the first objective function constructed according to the present invention, that is, the above formula (19).

[0147] It should be noted that the first objective function and the second objective function in the dual-objective scheduling optimization model correspond to a fitness respectively, for example, the first objective function corresponds to the first fitness, and the second objective function corresponds to the second fitness. Since the population participating in the dual-objective optimization has two fitnesses, the fitness corresponding to one of the objective functions can be selected first, for example, the first fitness corresponding to the first objective function can be selected, and the first fitness value of each individual can be calculated. Then, each individual in the initial population can be sorted according to the first fitness value of each individual, and the individual with the best first fitness ranked first can be determined as the initial optimal individual.

[0148] Step 323: An initial food position set may be established, the initial optimal individual may be added to the food position set, and the initial food position corresponding to the initial optimal individual may be determined.

[0149] It should be noted that the food location set may include one or more food locations, wherein each food location in the food location set corresponds to a scheduling scheme (data center scheduling scheme).

[0150] Step 324: Based on the initial food position corresponding to the initial optimal individual, the leader position and each follower position in the initial population can be determined, and the leader position and each follower position can be updated accordingly.

[0151] In some embodiments, the leader position may be calculated using the following formula (30), and the follower position may be calculated using the following formula (31).

[0152]

[0153] Where, X m,dis the position of the d-dimensional feature of the m-th follower before updating, X m',d is the position of the d-dimensional feature of the m-th follower after update; F d is the food location; c1, c2, c3 are three control factors, among which c2 and c3 are random numbers between 0 and 1, which are used to enhance the randomness of the leader's movement, strengthen the global search capability and the diversity of individuals, and avoid falling into the local optimum. c1 is the most important control factor in the formula.

[0154] In the embodiment of the present invention, the expression of the convergence factor c1 taken by the MSSA algorithm is shown in the following formula (32):

[0155]

[0156] Where t is the current iteration number and T is the maximum iteration number.

[0157] Step 325, calculate the second fitness value of each individual (a total of N individuals) in the initial population according to the second objective function (Formula 20), and then use the subjective and objective integrated weighting method to determine the new optimal individual based on the first fitness value and the second fitness value of each individual.

[0158] Specifically, the subjective and objective integrated weighting method can be used to determine the first comprehensive weight corresponding to the first objective function and the second comprehensive weight corresponding to the second objective function. Subsequently, the comprehensive fitness value of each individual can be determined based on the first fitness value and the second fitness value of each individual, as well as the first comprehensive weight and the second comprehensive weight. Furthermore, a new optimal individual can be determined based on the comprehensive fitness value of each individual.

[0159] In a specific embodiment, the specific method of determining a new optimal individual based on the first fitness value and the second fitness value of each individual using the subjective and objective integrated weighting method is shown in the following formulas (33)-(36). The comprehensive weights corresponding to the two objective functions (the first objective function and the second objective function) can be calculated according to formula (33).

[0160]

[0161] x * =argmax{x|ω1·A i +ω2·B i ,x∈P} (36)

[0162] In the formula, ω i is the comprehensive weight corresponding to the i-th objective function obtained by the subjective and objective integrated weighting method, μ i and ν iare the weights obtained by using the hierarchical analysis method and the entropy weight method respectively; r is the preference coefficient of the decision maker for subjective and objective factors, which ranges from 0 to 1; y1 and y2 are the set consisting of the electricity cost (first fitness value) corresponding to all individuals in the Pareto solution set (population) and the set consisting of the green electricity fitness (second fitness value) corresponding to all individuals respectively; A i and B i are the normalized electricity cost (first fitness value) and green electricity fitness (second fitness value) corresponding to the i-th individual in the Pareto solution set (population); y i is the target value (fitness value) corresponding to the i-th individual; x * Represents the newly determined optimal individual; P is the Pareto solution set (population).

[0163] Step 326, update the food position set based on the new optimal individual. Specifically, if the current food position set is empty, the new optimal individual determined in step 325 can be added to the food position set, thereby updating the new food position set. If the current food position set is not empty, the new optimal individual determined in step 325 can be compared with the first fitness value and the second fitness value of each food position in the food position set, respectively. If the first fitness value and the second fitness value of the new optimal individual are better than the first fitness value and the second fitness value of any food position in the food position set, then the any food position is replaced with the new optimal individual; if the first fitness value or the second fitness value (one of the fitness values) of the new optimal individual is better than the first fitness value or the second fitness value of each food position in the food position set, then the new optimal individual is added to the food position set; if the first fitness value and the second fitness value of any food position in the food position set are better than the first fitness value and the second fitness value of the new optimal individual, then the new optimal individual is discarded, and the food position set does not change.

[0164] Step 327, iteratively execute the above steps 324 to 326 until the number of iterations reaches the maximum number of iterations or the fitness value of the optimal individual reaches the termination value (for example, the first fitness value reaches the first termination value, and the second fitness value reaches the second termination value), the termination condition is met, and the algorithm can be terminated at this time to obtain the final food location set, which is the Pareto frontier of the dual-objective scheduling optimization model, based on which the optimal scheduling plan can be determined.

[0165] In addition, the dual-objective scheduling optimization model constructed according to the embodiment of the present invention can be combined with calculation examples to verify the effectiveness of the dual-objective scheduling optimization model.

[0166] First, we can set up the example based on actual cases. For example, the case selects a data center in a certain area in northwest my country as the research object. The scheduling period of this data center is 1 hour, and the unit scheduling cycle is 1 day. The relevant parameter settings of the example are shown in Table 1:

[0167] Table 1. Parameters of the example

[0168]

[0169]

[0170] Here, the data load of 24 periods of a typical day in the data center is selected. Based on this data load, the data load characteristics and energy consumption characteristics of the data center are first analyzed.

[0171] Figure 5 A schematic diagram of a typical daily data load curve of a data center according to some embodiments of the present invention is shown. Figure 5 It can be seen that the total data load of the data center is relatively small between 0:00 and 8:00. The main reason is that the user data processing demand is reduced at night. The data load type is mainly composed of rigid load and a small amount of transferable load, so the data load is relatively small at night. From 9:00 to 24:00 in the daytime, the data load is large and stable. This is mainly because the user data load demand is large during the daytime, and the data load calculation is relatively stable. There will be a small data load peak from 19:00 to 21:00 in the evening, and the overall trend is stable and fluctuating from 9:00 to 22:00.

[0172] Figure 6 The following is a schematic diagram of the time-of-use electricity price curve of the power grid according to some embodiments of the present invention. The electricity purchase price of a data center within a day can be referred to as Figure 6 The time-of-use electricity price curve of the power grid is shown. Figure 6 As shown, in one day, 0:00 to 7:00 and 23:00 to 24:00 are the low electricity consumption periods, and the electricity price is 0.3766 yuan / kWh; 7:00 to 10:00, 15:00 to 18:00, and 21:00 to 23:00 are the peak electricity consumption periods, and the electricity price is 0.677 yuan / kWh; 10:00 to 15:00 and 18:00 to 21:00 are the peak electricity consumption periods, and the electricity price is 0.9864 yuan / kWh.

[0173] Figure 7 Schematic diagram of distributed photovoltaic and distributed wind power generation in some embodiments of the present invention is shown. In some embodiments, distributed photovoltaic and distributed wind power stations are set up in the data center park. According to the local environmental parameters obtained from the CloudPSS platform, the photovoltaic power generation and wind power generation in each time period can be calculated according to the above formulas (14)-(15), see Figure 7The photovoltaic power generation power is affected by the light intensity. It gradually increases from 8 am to 12 pm, reaches a peak between 12 pm and 2 pm, and gradually decreases after 2 pm, and the power generation power drops to 0 at night. Figure 5 It can be seen that the trend of photovoltaic power generation curve is similar to the total load curve of data center, which is mainly used to reduce the demand for power purchase from the power grid during the daytime of data center. Wind power generation is affected by wind speed. The power generation reaches its peak between 0:00 and 6:00 at night, and the power generation is lower between 9:00 and 18:00 during the day, which is mainly used to handle the data load at night. This model adopts the method of generating and using the power generation of distributed power stations. The power generation in any period is only consumed by that period, and the amount not consumed is regarded as the amount of abandoned wind and solar power.

[0174] Figure 8 Schematic diagram of standardized data center power consumption curve and standardized external green power output curve without scheduling according to some embodiments of the present invention. Figure 8 As shown in Figure 2, by calculating the cosine similarity between the standardized data center power consumption curve and the standardized external green power output curve, the green power fitting degree is 68.53%. Figure 6 The time-of-use electricity price can be used to calculate that the electricity cost of the data center at this time is 10,688 yuan.

[0175] In some embodiments, in order to verify the effectiveness of the dual-objective scheduling optimization model, the following four scenarios can be set for comparison. Table 3 shows the settings of different optimization objectives under each scenario. Among them, scenario 4 is a comparison of scenario 3. The cosine similarity between the data center time-of-use power curve and the external green power and thermal power total output curve is selected as the calculation formula for the green power fit to verify the reliability of the green power fit target.

[0176] Table 2 Optimization target settings for each scenario

[0177]

[0178] Furthermore, the effectiveness of the dual-objective scheduling optimization model can be analyzed.

[0179] Fig. 9 The figure shows the relationship between the optimized data center electricity cost and green electricity fitting degree and algorithm iteration in four scenarios according to some embodiments of the present invention. Fig. 9 As shown in the figure, after 50 iterations, the target value of each scene converges to a different optimal value.

[0180] Scenarios 1 to 3 set in the present invention correspond to the scenarios of dual-objective optimization and single optimization of electricity cost or green electricity fitting degree, respectively. By comparing the target values, it can be seen that the electricity cost of 6453 yuan obtained by optimizing scenario 1 is between 4500 yuan in scenario 2 and 7189 yuan in scenario 3, and the green electricity fitting degree of 52.74% is between 9.17% in scenario 2 and 78.19% in scenario 3. This shows that the dual-objective scheduling optimization model constructed by the present invention actually makes a compromise between the optimization degrees of the two objectives, balances the weights of economy and greenness, and can reduce the electricity cost of the data center and improve the ability of the data center to absorb renewable energy. Compared with scenario 2 that optimizes electricity cost alone, scenario 1 effectively avoids the situation where the data center wastes renewable energy due to excessive use of thermal power. Compared with scenario 3 that optimizes green electricity fitting parameters alone, scenario 1 ensures that renewable energy is absorbed while guaranteeing the economic benefits of the data center to a certain extent, making it more competitive in the market system under the new power system.

[0181] Fig.10 The schematic diagram of the fitting of the data center power consumption curve and the external green power output curve in four scenarios according to some embodiments of the present invention is shown. By comparing scenario 1 and scenario 3, it can be seen that since both scenarios optimize the green power fitting degree, the trends of the data center power consumption curve and the external green power output curve are relatively similar. By comparing scenario 1 and scenario 2, it can be seen that since scenario 2 only optimizes the power cost of the data center, its l green fitting effect is relatively poor, which is different from scenario 1. However, since both scenarios optimize the power cost, the power consumption curves of the two scenarios have a commonality, that is, a large amount of data loads in the daytime high power price period are dispatched to the last low power price period for processing, resulting in the accumulation of data loads from 23:00 to 24:00. The more serious the accumulation phenomenon, the lower the power cost, but the lower the green power fitting degree. In scenario 4, the fitting degree of the data center power consumption curve and the external total output curve reaches 97.62%, but the green power fitting degree is only 59.72%, and it is also inferior to the scenario of dual-objective optimization in terms of power cost. In general, the dual-objective scheduling optimization model constructed by the present invention compromises the advantages of optimizing the two objectives separately, and the target selected to measure carbon emission reduction is reasonable, which is of substantial help to data centers in improving their renewable energy absorption capacity.

[0182] Fig.11A schematic diagram of the Pareto front obtained according to some embodiments of the present invention is shown. In an embodiment of the present invention, a dual-objective scheduling optimization model is constructed with the minimum electricity cost of the data center as the first goal and the maximum green electricity fit as the second goal, and the MSSA algorithm is used to solve the dual-objective scheduling optimization model to approximate the Pareto front. Among them, the two objectives are used as the two fitness of the MSSA algorithm, and the charging state, discharging state, charging power, discharging power of the energy storage power station in 24 time periods and the transfer load from each time period to the subsequent time period, a total of 372 variables, are used as the characteristic dimension of the sea squirt. The maximum number of iterations T is 3000 times, and the population contains 20 individuals. Since the dual-objective scheduling optimization model has many constraints, in order to make the algorithm easier to find a feasible solution, in some embodiments of the present invention, the Monte Carlo method is used to obtain 20 relatively dispersed feasible solutions as 20 individuals of the initial population for iteration. The final Pareto front is as follows Fig.11 shown.

[0183] The optimal scheduling scheme on the Pareto frontier was selected for analysis using the subjective and objective integrated weighting method. Under the optimal scheduling scheme, the electricity cost of the data center was 6,453 yuan, and the green electricity fitting degree was 52.74%. Compared with the electricity cost of 10,688 yuan and the green electricity fitting parameter of 68.53% corresponding to the scheme before scheduling, the electricity cost was significantly reduced. However, due to the conflict between the two optimization objectives, the green electricity fitting degree was also slightly reduced.

[0184] Fig.12 The schematic diagram of the scheduling of the data load and power load of the data center throughout the day after scheduling according to some embodiments of the present invention is shown. The above figure corresponds to the distribution of data loads in each time period after scheduling. Under this scheme, after scheduling, compared with before scheduling, the power consumption from 0 to 3 and from 6 to 23 has different degrees of reduction, while the power consumption from 3 to 6 has increased significantly. This is because the scheme includes the goal of optimizing the green electricity fitting degree, and the external green electricity curve reaches the peak in this period, so the data load of the previous period is scheduled to this period for processing, which is conducive to the improvement of green electricity fitting degree. The power consumption from 23 to 24 has increased significantly. This is due to the three reasons that the scheme includes the goal of optimizing the power cost of the data center, the transferable load has the characteristic that it can only be scheduled from the previous period to the subsequent period, and the power price from 23 to 24 is in the valley period. By scheduling the transferable load, the data load during the peak power price period is transferred to the valley period for processing, which effectively reduces the power cost of the data center.

[0185] Fig.12The output of distributed power stations and the charging and discharging of energy storage stations are clearly shown below. Wind power and photovoltaic power are generated in large quantities in the early morning and noon, respectively, which effectively reduces the amount of electricity purchased by the data center from the power grid during these periods. In addition, the charging strategy of the energy storage station can be used to effectively store excess wind power from 0:00 to 7:00, reducing the abandonment rate of electricity. The energy storage station charges from 0:00 to 7:00 and discharges from 10:00 to 11:00 and 18:00 to 21:00. The charging and discharging periods include both normal and peak-valley periods. The charging and discharging of the energy storage station during normal periods indicates that the strategy adopted by the energy storage station is to improve the green electricity fit by scheduling the electricity demand of the data center at the power level, rather than reducing the electricity cost through peak-valley price arbitrage. When the energy storage station charges during the low electricity price period and discharges during the peak period, it means that it adopts the peak-valley price arbitrage strategy, which effectively reduces the electricity cost of the data center. Although this solution is obtained by solving the Pareto frontier, because on the Pareto frontier composed of conflicting goals, different points are the optimal solutions obtained by assigning different weights to the two goals, so a better goal will inevitably lead to the degradation of the other goal. Under this solution, the optimization weight of the data center electricity cost goal is higher than the green electricity fitting parameter, so the energy storage power station will be more inclined to choose the peak-valley price difference arbitrage strategy when choosing the charging and discharging strategy, which also leads to the total amount of electricity dispatched from the peak period of electricity price to the valley period in the figure is more than the dispatched amount at normal electricity prices.

[0186] It can be seen that the present invention proves the effectiveness of the dual-objective scheduling optimization model constructed by the present invention by setting different scenarios for comparative analysis.

[0187] In addition, the present invention also conducts sensitivity analysis on the proportion of transferable load, the upper limit of energy storage power station capacity, and the peak-valley price difference of time-of-use electricity prices.

[0188] Fig.13 The figure shows the relationship between electricity cost, green electricity fitting degree and transferable load proportion coefficient. For data centers, the dispatching amount of transferable load is the key factor in reducing electricity cost and improving green electricity fitting degree. Therefore, reasonable adjustment of the proportion of various data loads plays a vital role in optimizing the two goals of data centers. The above dual-objective optimization scheme is selected as the control scheme. Without changing the reducible load proportion coefficient, that is, without changing the sum of the transferable load proportion coefficient and the rigid load proportion coefficient, the transferable load proportion coefficient is taken as 0.4, 0.45, 0.5, 0.55, and 0.6 respectively. Other parameters remain unchanged. The target values ​​(electricity cost, green electricity fitting degree) of the adjusted scheme are shown as follows: Fig.13 As shown. Fig.13It can be seen that with the increase of the transferable load ratio coefficient and the decrease of the rigid load ratio coefficient, the electricity cost of the data center gradually decreases and the green electricity fitting degree gradually increases. This is because as more rigid loads become transferable loads, the flexibility of data loads becomes stronger and stronger, and data centers can more flexibly schedule batch processing loads to improve their own economic benefits and green electricity consumption rate.

[0189] Fig.14 A schematic diagram showing the relationship between electricity cost, green electricity fit and energy storage capacity is shown. For the overall scheduling plan, since the scheduling volume of the energy storage power station is relatively small, it plays a small role in reducing the electricity cost of the green data center, but it has great potential in improving the green electricity fit. The above dual-objective optimization scheme is selected as the control scheme. Under the condition that other parameters remain unchanged, the energy storage power station models are adjusted to 200kW / 500kWh, 400kW / 1000kWh, 600kW / 1500kWh, 800kW / 2000kWh, and 1000kW / 2500kWh, respectively. The target values ​​(electricity cost, green electricity fit) after the adjustment scheme are as follows: Fig.14 As shown. Fig.14 It can be seen that with the increase in the capacity of energy storage power stations, the electricity cost of data centers is slightly reduced, and the green electricity fitting degree is greatly improved. This shows that although the energy storage power station participates in scheduling to a lesser extent than the transferable load, it does not cause the data load to accumulate in the last period, and the effect of improving the green electricity fitting degree is better than load scheduling.

[0190] Fig.15 A schematic diagram of the relationship between electricity cost, green electricity fitting degree and peak electricity price change ratio is shown. Time-of-use electricity price is a direct factor affecting the electricity cost of green data centers. As the peak-valley price difference increases, whether it is by dispatching transferable loads to transfer the load during peak electricity price periods to low electricity price periods for processing, or by using energy storage power stations to charge during valley hours and discharge during peak hours, a larger electricity price difference can be taken, thereby reducing the electricity cost of green data centers. This subsection selects the above-mentioned dual-objective optimization scheme as the control scheme, and separately processes the peak electricity price and valley electricity price, and studies the changes in the difference in electricity cost of the green data center before and after scheduling at different change ratios. The base of the change ratio is the difference between the peak electricity price and the valley electricity price in scheme one. From Fig.15It can be seen that with the increase of peak electricity prices or the decrease of valley electricity prices, the difference in electricity costs of green data centers before and after scheduling shows an upward trend, which means that the greater the peak-valley price difference, the more beneficial the scheduling scheme of the present invention is to the reduction of electricity costs. However, the rates of increase of the two are different, indicating that the difference in electricity costs is more sensitive to changes in peak electricity prices than to valley electricity prices. The reason is that under this scheme, the green data center purchases more electricity during peak hours than during valley hours. Therefore, when the peak electricity price is changed, the change in the difference in electricity costs is more significant. Since changes in peak-valley electricity prices will not cause changes in the dispatching state of transferable loads and the charging and discharging state of energy storage, nor will they cause changes in the dispatching amount, the green electricity fitting degree will not change with the peak-valley electricity price.

[0191] In summary, according to the green data center scheduling optimization method 300 under the coupling of computing power and electricity of the present invention, the scheduling optimization problems at the two levels of computing power and electricity are focused on, and a dual-objective scheduling optimization model is constructed with the goals of minimizing the electricity cost of the data center and maximizing the green electricity fitting degree. Then, the MSSA algorithm is used to solve the Pareto frontier, and the optimal scheduling plan of the data center is determined based on the Pareto frontier. In this way, the electricity cost of the data center can be reduced, the renewable energy absorption capacity of the data center can be improved, and the deep integration of computing power and electricity can be strengthened.

[0192] By way of example and not limitation, readable media include readable storage media and communication media. Readable storage media stores information such as computer readable instructions, data structures, program modules or other data. Communication media generally embody computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and include any information transfer medium. Any combination of the above is also included within the scope of readable media.

[0193] In the description provided herein, algorithms and displays are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the examples of the present invention. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to implement the content of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of the present invention.

[0194] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0195] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof.

[0196] Those skilled in the art will appreciate that the modules or units or components of the devices in the examples disclosed herein may be arranged in the devices described in the embodiments, or alternatively may be located in one or more devices different from the devices in the examples. The modules in the foregoing examples may be combined into one module or may be divided into a plurality of submodules.

[0197] Unless otherwise specified, the use of ordinal numbers "first", "second", "third", etc. to describe common objects merely indicates that different instances of similar objects are involved and is not intended to imply that the objects so described must have a given order in time, space, order or in any other manner.

Claims

1. A method for optimizing the scheduling of a green data center under computing power and power coupling, wherein the data center includes energy-consuming equipment and energy storage power stations, and is suitable for joint power supply through a power grid, a distributed power station, and an energy storage power station, wherein the distributed power station includes distributed wind power and distributed photovoltaic power, and the data load types of the data center include transferable loads, curtailable loads, and rigid loads, and the method includes: A dual-objective scheduling optimization model is constructed with the goal of minimizing the electricity cost of the data center and maximizing the green electricity fit. The electricity cost of the data center is used to represent the sum of the costs incurred by the data center in purchasing electricity from the power grid in each period to meet the electricity required for processing data loads, and the green electricity fit is used to evaluate the matching degree between the electricity demand of the data center and the renewable energy supply. The dual-objective scheduling optimization model is solved by using a multi-objective salp swarm algorithm to obtain a Pareto frontier; An optimal scheduling scheme is determined based on the Pareto front, wherein the optimal scheduling scheme includes a data center power consumption curve, an energy storage power station charging and discharging curve, and a transferable load scheduling strategy.

2. The method of claim 1, wherein: Also includes: The data center time-sharing power consumption curve and the external green power output curve are standardized respectively to obtain the corresponding standardized data center time-sharing power consumption curve and the standardized external green power output curve; The cosine similarity between the standardized data center time-of-use electricity consumption curve and the standardized external green electricity output curve is used as the green electricity fitting degree.

3. The method according to claim 1 or 2, wherein: With the goal of minimizing the electricity cost of the data center and maximizing the green electricity fitting degree, a dual-objective scheduling optimization model is constructed, including: Taking the minimum electricity cost of the data center as the first goal, the first objective function is constructed; Taking the maximum green electricity fitting degree as the second goal, the second objective function is constructed; A dual-objective scheduling optimization model is constructed based on the first objective function and the second objective function.

4. The method of claim 3, wherein: Based on the first objective function and the second objective function, a dual-objective scheduling optimization model is constructed, including: Based on the first objective function, the second objective function, the data load constraint condition and the power load constraint condition, a dual-objective scheduling optimization model is constructed; Among them, the data load constraints include transferable load constraints and reducible load constraints; the power load constraints include energy storage power station charging and discharging constraints, energy storage power station state change times constraints, energy storage power station capacity constraints, and power balance constraints.

5. The method of claim 4, wherein: The transferable load constraint conditions include: the actual transfer load amount of the data center after scheduling in each time period is equal to the initial transferable load amount before scheduling in the time period plus the transferable load amount scheduled from the previous time period to the time period minus the transferable load amount scheduled from the time period to the subsequent time period, and the transferable load amount scheduled from each time period to the subsequent time period does not exceed the initial transferable load amount of the time period; The load reduction constraint conditions include: the actual load reduction amount of the data center in each period does not exceed the load reduction amount of the period, and the total actual load reduction amount of all periods of the data center cannot exceed the maximum limit of the load reduction amount, wherein the load reduction amount of each period is equal to the initial total load amount of the period multiplied by the load reduction ratio coefficient; The energy storage power station charging and discharging constraint conditions include: the charging power of the energy storage power station in each period does not exceed the upper limit of the charging power allowed by the energy storage power station, and the discharging power of the energy storage power station in each period does not exceed the upper limit of the discharging power allowed by the energy storage power station; The energy storage power station state change times constraint condition includes: the sum of the charging times and discharging times of the energy storage power station in each time period cannot exceed the maximum state change times of the energy storage power station, and the energy storage power station cannot be charged and discharged at the same time in the same time period; The energy storage power station capacity constraint conditions include: the capacity of the energy storage power station in each period is greater than or equal to the lower limit of the energy storage power station capacity, and less than or equal to the upper limit of the energy storage power station capacity; and the initial capacity of the energy storage power station is equal to the capacity after charging and discharging in the last period; The power balance constraint conditions include: the sum of the total energy consumption of the data center and the charging and discharging power of the energy storage power station in each time period is equal to the sum of the power supply power of the power grid, the power generation power of distributed photovoltaics, and the power generation power of distributed wind power, and the power supply power of the power grid in each time period is a non-negative value.

6. The method according to any one of claims 1 to 5, wherein: The dual-objective scheduling optimization model is solved using a multi-objective salp swarm algorithm to obtain the Pareto frontier, including: Using the Monte Carlo method to obtain an initial population, the initial population includes N individuals, including a leader and N-1 followers, where N is a positive integer greater than or equal to 3; Calculating a first fitness value of each individual in the initial population according to the first objective function, and determining an initial optimal individual in the initial population based on the first fitness value of each individual; Adding the initial optimal individual to a food position set, and determining an initial food position corresponding to the initial optimal individual, wherein each food position in the food position set corresponds to a scheduling scheme; Based on the initial food position corresponding to the initial optimal individual, updating the leader position and each follower position in the initial population; Calculating the second fitness value of each individual in the initial population according to the second objective function, and determining a new optimal individual based on the first fitness value and the second fitness value of each individual by using a subjective and objective integrated weighting method; Updating the food position set based on the new optimal individual, including: if the food position set is empty, adding the new optimal individual to the food position set; if the food position set is not empty, comparing the new optimal individual with the first fitness value and the second fitness value of each food position in the food position set, respectively, wherein if the first fitness value and the second fitness value of the new optimal individual are both better than the first fitness value and the second fitness value of any food position, replacing any food position with the new optimal individual; if the first fitness value or the second fitness value of the new optimal individual is better than the first fitness value or the second fitness value of each food position, adding the new optimal individual to the food position set; if the first fitness value and the second fitness value of any food position in the food position set are both better than the first fitness value and the second fitness value of the new optimal individual, discarding the new optimal individual; The above steps are iterated until the number of iterations reaches the maximum number of iterations or the fitness value of the optimal individual reaches the termination value, and the final food position set is obtained, and the final food position set is used as the Pareto front.

7. The method of claim 6, wherein: The subjective and objective integrated weighting method is used to determine the new optimal individual based on the first fitness value and the second fitness value of each individual, including: Using the subjective and objective integrated weighting method, determine the first comprehensive weight corresponding to the first objective function and the second comprehensive weight corresponding to the second objective function; Determine a comprehensive fitness value of each individual based on the first fitness value and the second fitness value of each individual, and the first comprehensive weight and the second comprehensive weight; The new optimal individual is determined based on the comprehensive fitness value of each individual.

8. The method according to any one of claims 1 to 7, wherein: The dual-objective scheduling optimization model is solved by using a multi-objective salp swarm algorithm, including: The initial transferable load of the data center in each time period, the charging status and charging power of the energy storage power station in each time period, the time-of-use electricity price curve, and the external green electricity output curve are input into the dual-objective scheduling optimization model, and the multi-objective Salp Unicinus swarm algorithm is used to solve the dual-objective scheduling optimization model.

9. The method according to any one of claims 1 to 8, wherein: The energy-consuming equipment of the data center includes IT equipment, air-conditioning and refrigeration equipment, and auxiliary regulating equipment, wherein the sum of the energy consumption of the IT equipment and the air-conditioning and refrigeration equipment accounts for more than 90% of the total energy consumption of the data center.

10. A computing device comprising: at least one processor; and A memory storing program instructions, wherein the program instructions are configured to be processed by the at least one processor, and the program instructions include instructions for processing the method according to any one of claims 1 to 9.

Images