Data center cluster comprehensive energy capacity configuration method under computing power and electric power coupling

Abstract

The invention discloses a data center cluster integrated energy capacity configuration method under computing power and electric power coupling, which comprises the following steps: taking the minimum annual total cost planned by a data center cluster integrated energy system as an upper layer target, constructing an upper layer target function, and constructing an upper layer planning model based on the upper layer target function; taking the minimum net load average absolute deviation of the data center cluster as a lower-layer target, constructing a lower-layer target function, and constructing a lower-layer operation model based on the lower-layer target function; based on the upper-layer planning model and the lower-layer operation model, obtaining a data center cluster comprehensive energy capacity configuration double-layer optimization model; and solving the data center cluster comprehensive energy capacity configuration double-layer optimization model by using a hierarchical sequence method to obtain a capacity configuration result. According to the invention, the system cost can be reduced while the comprehensive energy configuration is optimized, and the green electricity nearby consumption capability and the comprehensive energy utilization efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, and particularly relates to a data center cluster comprehensive energy capacity configuration method under power and computing power coupling. BACKGROUND

[0002] As the core computing power infrastructure of digital economy, data center clusters support massive data processing in key fields such as high-frequency trading in finance, optimization of industrial internet, and intelligent traffic scheduling. However, the development of data centers is facing the problems of rapid increase in energy consumption and carbon emission pressure. The International Energy Agency predicts that global energy consumption will double in 2026. In addition, the rapid development of technologies such as AI puts forward higher requirements for the scale and efficiency of computing power, and the traditional energy mode is difficult to adapt. Under the dual driving of global low-carbon transformation and the "eastern data and western computing" strategy, the scientific configuration of comprehensive energy of data center clusters has become the key to solving the contradiction between energy supply and demand and promoting green development.

[0003] In terms of dealing with wind and light output fluctuations and dynamic changes in data load, existing research has formed a diversified uncertainty processing system. Random optimization methods depict uncertainty through scenario generation and reduction techniques, such as using Frank-Copula functions and Monte Carlo methods to generate typical scenarios of wind power, photovoltaic power, and load, and combining NBI algorithms to convert multi-objective problems into single-objective optimization to provide probabilistic decision basis for capacity configuration. Robust optimization enhances system interference resistance by constructing uncertainty sets, such as describing the random characteristics of wind power based on ∞-Wasserstein fuzzy sets, which can balance economic and robustness without probability distribution, significantly reducing the complexity of solving. Distributed robust optimization further integrates the advantages of both, such as the chance constrained programming model based on Sinkhorn fuzzy sets, which ensures that more than 90% of scenarios meet operational constraints in the integration of data center multi-energy microgrids, and takes into account the stability under extreme conditions.

[0004] In terms of improving energy utilization efficiency, existing research generally focuses on the coordinated configuration of renewable energy, energy storage, and terminal energy equipment. In terms of renewable energy integration, by comparing the hybrid system of "photovoltaic + wind power + battery + supplemental energy", it is found that photovoltaic is the core to improve renewable energy penetration rate (RP), and when the capacity increases from 100kW to 800kW, RP can be improved by about 2500-3000 hours, and natural gas as a supplemental energy can reduce 15% of life cycle cost (LCC) and 50% of carbon emission compared with diesel. In terms of combined CCHP and energy storage, research shows that CCHP waste heat recovery combined with ice storage air conditioning (ISAC) can reduce cooling energy consumption by 20%, and battery energy storage and hydrogen energy storage can further smooth wind and light fluctuations, reducing data center operating costs by 10.48%.

[0005] In terms of data center load task processing, utilizing the spatiotemporal flexibility of load to improve green electricity consumption has become a research hotspot. In the time dimension, by distinguishing delay-sensitive and tolerant loads, a real-time response mechanism based on price incentives is constructed, which can increase the load transfer rate by 15%-20% and reduce peak electricity cost by 13.24%. In the spatial dimension, cross-data center load scheduling is regarded as a "virtual battery", which can reduce the loss of abandoned electricity by 23%-36% in the distribution network, especially in the renewable energy surplus scenario. In addition, combined with dynamic voltage frequency adjustment (DVFS) technology and thermal inertia, the matching degree of load and green power output can be further improved by adjusting the power consumption of servers and controlling temperature fluctuations, which can reduce electricity consumption by 28.8%.

[0006] Existing researches mostly use independent analysis methods to deal with uncertainties on the energy side or the load side, and fail to fully depict the coupling relationship between the two. Some researches focus on the flexibility of load scheduling, but simplify the characteristics of renewable energy output. Single-sided optimization is difficult to adapt to the strong coupling between source and load in the "green electricity direct connection" scenario of data center clusters, which may lead to the failure of capacity configuration schemes in extreme working conditions. In addition, the coordination mechanism of CCHP, energy storage and load scheduling is not perfect. CCHP selection and energy storage configuration are mostly optimized independently, without considering the timing matching of waste heat recovery and energy storage charging and discharging, ignoring the influence of CCHP efficiency change with load rate on energy storage scheduling; the communication delay of cross-data center load transfer is not considered in the start-stop scheduling of CCHP units. In terms of spatial layout and policy mechanism integration, existing researches have shortcomings. Although it is proposed to consume green electricity locally, the capacity matching model of wind and solar power stations and data center clusters is not established, which makes it difficult to support the layout strategy of "computing power following energy"; in terms of policy mechanism, the influence of market mechanisms such as carbon trading and green certificates on equipment configuration is simplified, the influence of green electricity consumption rate and carbon price linkage on CCHP investment return is not quantified, and the robustness parameters are mostly set statically, which cannot dynamically adapt to the wind and light resources and load characteristics of different regions.

[0007] Therefore, there is a need for a data center cluster comprehensive energy capacity configuration method under the coupling of computing power and electricity to solve the problems in the above scheme. SUMMARY

[0008] Therefore, the present application provides a data center cluster comprehensive energy capacity configuration method under the coupling of computing power and electricity to solve or at least alleviate the above problems.

[0009] According to one aspect of the present application, a method for configuring comprehensive energy capacity of data center cluster under power and computing power coupling is provided, which is executed in a computing device and includes: constructing an upper target function with a minimum total annual cost of a comprehensive energy system of the data center cluster as an upper target, constructing an upper planning model based on the upper target function, wherein the comprehensive energy system of the data center cluster includes a data center cluster, an energy storage system and a CCHP system, the total annual cost is composed of equivalent annual investment cost, annual operation and maintenance cost, annual energy cost, green certificate cost and energy storage system electricity sales revenue, and the equivalent annual investment cost is used to represent the full life cycle cost of the comprehensive energy system of the data center cluster planning converted into the equivalent annual internal cost; constructing a lower target function with a minimum average absolute deviation of net load of the data center cluster as a lower target, constructing a lower operation model based on the lower target function, wherein the average absolute deviation of the net load is the average value of the absolute deviation of the net load of the data center cluster in each period, and the absolute deviation of the net load is the difference between the total energy consumption of the data center cluster and the water source heat pump energy consumption of the CCHP system and the wind power, the photovoltaic power, the grid power, the CCHP system power and the energy storage system power; obtaining a double-layer optimization model for configuring comprehensive energy capacity of the data center cluster based on the upper planning model and the lower operation model; solving the double-layer optimization model for configuring comprehensive energy capacity of the data center cluster by using a hierarchical sequence method to obtain a capacity configuration result, and the capacity configuration result is used to indicate the number of IT equipment and the supporting energy storage capacity of each data center in the data center cluster, the number of PGU generator sets in the CCHP system, and the wind power, the photovoltaic power and the CCHP system power in each period.

[0010] Optionally, in the method for configuring comprehensive energy capacity of data center cluster under power and computing power coupling according to the present application, the data center cluster includes a plurality of data centers, the data center includes IT equipment and an air conditioning system, and is adapted to perform online space migration of load between different data centers through a virtual machine; the equivalent annual investment cost is adapted to be determined according to a full life cycle cost of the comprehensive energy system of the data center cluster, a discount rate and a design life of the data center cluster, wherein the full life cycle cost is the sum of construction investment fees of the energy storage system, the data center cluster, the CCHP system and the virtual machine.

[0011] Optionally, in the data center cluster comprehensive energy capacity configuration method under the computing power and electricity coupling according to the application, the annual operation and maintenance cost is related to the annual operation and maintenance cost of the energy storage system, the operation and maintenance cost of the PGU generator set of the data center cluster and the CCHP system in a typical day, and the number of typical days in a year; the annual energy consumption cost is related to the grid electricity fee generated by the data center cluster in a typical day, the renewable energy electricity fee and the natural gas fee, and the number of typical days in a year; the green certificate cost is related to the number of green certificates purchased by the data center cluster in a typical day and the price of a single green certificate; and the energy storage system electricity selling income is related to the grid electricity fee in each period in a typical day, the electricity selling power of each energy storage in the energy storage system, and the number of typical days in a year.

[0012] Optionally, in the data center cluster comprehensive energy capacity configuration method under the computing power and electricity coupling according to the application, an upper-layer planning model is constructed based on the upper-layer objective function, including: constructing an upper-layer planning model based on the upper-layer objective function and upper-layer constraint conditions, the upper-layer constraint conditions including investment capability constraint conditions, configuration capacity constraint conditions, energy output constraint conditions, and renewable energy consumption constraint conditions; wherein, the investment capability constraint conditions represent that the equivalent annual investment cost does not exceed the maximum planning investment limit; the configuration capacity constraint conditions represent that: the electricity capacity of the energy storage equipped by the data center in each period in a typical day does not exceed the supporting energy storage capacity of the data center; the configuration quantity of the IT equipment per unit area of the data center is the ratio of the IT equipment configuration quantity of the data center to the planning area of the data center, which does not exceed the maximum configurable IT equipment quantity per unit area of the data center; and the number of PGU generator sets in the CCHP system does not exceed the maximum configuration quantity of the PGU generator sets; the energy output constraint conditions represent that: the grid transmission power does not exceed the maximum grid transmission power; the CCHP system power generation power does not exceed the product of the maximum power generation power of a single PGU generator set and the number of PGU generator sets in the CCHP system; and the actual transmission capacity of the wind power does not exceed the wind power generation power, and the actual transmission capacity of the photovoltaic power does not exceed the photovoltaic power generation power; and the renewable energy consumption constraint conditions represent that: the renewable energy consumption rate is within the consumption rate range, wherein, the renewable energy consumption rate is the proportion of the actually consumed renewable energy power in the system total power consumption, and the renewable energy power is the sum of the wind power generation power, the photovoltaic power generation power and the green certificate supplemented renewable energy power.

[0013] Optionally, in the data center cluster integrated energy capacity configuration method under the computing power and electricity coupling according to the application, a lower layer operation model is constructed based on the lower layer objective function, including: constructing an upper layer planning model based on the lower layer objective function and lower layer constraint conditions, the lower layer constraint conditions including energy conservation constraint conditions, data center cluster service quality constraint conditions, energy utilization efficiency constraint conditions, and data center room temperature constraint conditions; wherein the energy conservation constraint conditions include an electric energy balance constraint condition and a thermal energy balance constraint condition; the electric energy balance constraint condition indicates that the sum of wind power, photovoltaic power, grid power, CCHP system power, and energy storage system power in each period is equal to the sum of total data center cluster energy consumption and CCHP system water source heat pump energy consumption in each period; the thermal energy balance constraint condition indicates that the sum of heat generated by IT equipment in the data center cluster and waste heat generated by the PGU generator set in the CCHP system in each period is equal to the sum of heat energy consumption of the absorption chiller in the CCHP system, refrigerating capacity of the air conditioning system in the data center cluster, and heat recovered and transported by the water source heat pump in the CCHP system in each period; the data center cluster service quality constraint condition indicates that the total amount of data load received by the data center cluster in each period is greater than or equal to the total amount of data load processed by the data center cluster in each period, the total amount of data load processed by the data center cluster in each period is greater than or equal to the total amount of data load that must be processed by the data center cluster in each period, and the data center cluster service quality index is less than or equal to 2000, wherein the data center cluster service quality index is the average value of the waiting queue length of the data center cluster in each period, and the waiting queue length of the data center cluster in each period is the difference between the total amount of data load received by the data center cluster in each period and the total amount of data load processed by the data center cluster; the energy utilization efficiency constraint condition indicates that the air conditioning system energy consumption of the data center cluster is instantaneously matched with the renewable energy power generation; and the data center room temperature constraint condition indicates that the indoor temperature of each data center room in the data center cluster in each period needs to be maintained within a predetermined temperature range.

[0014] Optionally, in the data center cluster integrated energy capacity configuration method under the computing power and electricity coupling according to the application, the data load received by the data center cluster includes interactive load and batch processing load, the interactive load includes spatially transferable load and rigid load, and the batch processing load includes temporally transferable load.

[0015] Optionally, in the data center cluster integrated energy capacity configuration method under the computing power and power coupling according to the application, the hierarchical sequence method is used to solve the data center cluster integrated energy capacity configuration double-layer optimization model to obtain the capacity configuration result, including the steps that initial input parameters and a selected robust method are obtained, the initial input parameters including initial wind power, initial photovoltaic power and initial data load; initial CCHP system power is predicted based on the initial input parameters; the upper-layer planning model is solved based on the initial input parameters and the initial CCHP system power to obtain an initial upper-layer solving result and a lower bound of the objective function; the wind and light output uncertainty is described by using the robust method based on the initial upper-layer solving result to solve the lower-layer operation model to obtain an initial lower-layer solving result and an upper bound of the objective function; it is judged whether a convergence condition is reached, the convergence condition including that a difference between the upper bound of the objective function and the lower bound of the objective function is less than an absolute difference threshold, or a relative error between the upper bound of the objective function and the lower bound of the objective function is less than a relative difference threshold, or an iteration number reaches a maximum iteration number; if the convergence condition is reached, the capacity configuration result is obtained based on the initial upper-layer solving result and the initial lower-layer solving result; if the convergence condition is not reached, the initial input parameters are smoothed and updated based on a relaxation factor to obtain new input parameters, and the above steps are iteratively executed until the capacity configuration result is obtained when the convergence condition is reached.

[0016] Optionally, in the data center cluster integrated energy capacity configuration method under the computing power and power coupling according to the application, the wind and light output uncertainty is described by using the robust method, including that a fuzzy set is constructed based on a mean value and a variance of the wind and light power by using a distribution robust method, and the wind and light output uncertainty is described based on the fuzzy set.

[0017] Optionally, in the data center cluster integrated energy capacity configuration method under the computing power and power coupling according to the application, the hierarchical sequence method is used to solve the data center cluster integrated energy capacity configuration double-layer optimization model, including that when the hierarchical sequence method is used to solve the data center cluster integrated energy capacity configuration double-layer optimization model, a subjective and objective integrated weighting method based on the analytic hierarchy process and the entropy weight method is used to set upper-layer target weights and lower-layer target weights.

[0018] According to an aspect of the application, a computing device is provided, including at least one processor, a memory storing program instructions, wherein the program instructions are configured to be suitable for being executed by the at least one processor, and the program instructions include instructions for executing the data center cluster integrated energy capacity configuration method under the computing power and power coupling.

[0019] According to an aspect of the present application, a computer program product is provided, comprising computer programs / instructions, wherein the computer programs / instructions, when executed by a processor, implement the method as described above.

[0020] According to an aspect of the present application, a readable storage medium storing program instructions is provided, which, when read and executed by a computing device, causes the computing device to perform the data center cluster integrated energy capacity configuration method under the coupling of computing power and electricity as described above.

[0021] According to the technical solution of the present application, a data center cluster integrated energy capacity configuration method under the coupling of computing power and electricity is provided. An upper planning model is constructed to optimize the scale and capacity configuration of the data center cluster with the minimum annual total cost of the data center cluster integrated energy system planning as the target. A lower operation model is constructed to dynamically adjust the energy output and load scheduling with the minimum average absolute deviation of the net load of the data center cluster as the target. Thus, a double-layer optimization model of the data center cluster integrated energy capacity configuration is obtained, and the optimized capacity configuration result is obtained by solving. Based on this, the multi-energy complementation and the deep coupling of computing power and electricity within the data center cluster can be realized, the system cost can be reduced while optimizing the integrated energy configuration, and the local consumption capacity of green electricity and the comprehensive energy utilization efficiency can be improved.

[0022] In addition, the present application adopts a distributed robust optimization method, which can more efficiently adapt to the bilateral uncertainty of wind and light generation and data load, further optimize resource configuration, and reduce system cost.

[0023] The above description is only a summary of the technical solutions of the present application. In order to enable a clearer understanding of the technical means of the present application, the present application can be implemented according to the content of the specification, and in order to enable the above and other purposes, features and advantages of the present application to be more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to achieve the above and related purposes, the present application describes certain illustrative aspects in combination with the following description and drawings, which indicate various ways in which the principles disclosed by the present application can be practiced, and all aspects and their equivalents are intended to fall within the scope of the claimed subject matter. The above and other objects, features and advantages of the present disclosure will become more apparent through reading the following detailed description in conjunction with the accompanying drawings. Throughout the disclosure, the same reference numerals generally refer to the same components or elements.

[0025] Figure 1 A structural schematic diagram of a data center cluster integrated energy system 100 according to an embodiment of the present application is shown; Figure 2 A schematic diagram of a computing device 200 according to an embodiment of the present application is shown. Figure 3 Fig. 1 shows a flow diagram of a method for configuring comprehensive energy capacity of a data center cluster under coupling of computing power and electricity according to an embodiment of the present application; Figure 4 Fig. 2 shows a diagram for comparing cost and benefit of two robust methods according to some embodiments of the present application; Figure 5 Fig. 3 shows a diagram for comparing target values in various scenarios according to some embodiments of the present application; Figure 6 Fig. 4 shows a diagram for comparing capacity configurations in various scenarios according to some embodiments of the present application; Figure 7 Fig. 5 shows a diagram for comparing auxiliary capacity configurations in various scenarios according to some embodiments of the present application; Figure 8 Fig. 6 shows a diagram for comparing cost and benefit in various scenarios according to some embodiments of the present application. DETAILED DESCRIPTION

[0027] Exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0028] To optimize the comprehensive energy configuration of the data center cluster, solve the contradiction between energy supply and demand, and promote green development, the embodiments of the present application first break through the traditional single data center energy configuration mode, build a green electricity direct connection data center cluster comprehensive energy system to realize multi-energy complementation and deep coupling of computing power and electricity within the data center cluster. Further, on this basis, a method for configuring comprehensive energy capacity of a data center cluster under coupling of computing power and electricity is proposed, which can optimize the comprehensive energy configuration while reducing system cost, and improve the ability of local consumption of green electricity and the efficiency of comprehensive energy utilization.

[0029] First, a data center cluster comprehensive energy system according to an embodiment of the present application will be introduced.

[0030] Figure 1 Fig. 1 shows a structure diagram of a data center cluster comprehensive energy system 100 according to an embodiment of the present application.

[0031] As shown in Fig. 1, the data center cluster comprehensive energy system 100 includes a plurality of data centers 101, a plurality of green power plants 102, a plurality of power transmission and transformation devices 103, a plurality of energy storage devices 104, a plurality of energy transmission and transformation devices 105, and a plurality of energy distribution devices 106. Figure 1As shown, the data center cluster integrated energy system 100 includes a data center cluster, an energy storage system, a CCHP system, and introduces wind power, photovoltaic power, and a power grid. The data center cluster includes multiple data centers, and each data center includes IT equipment, an air conditioning system, a lighting system, and a power distribution system.

[0032] According to the data center cluster integrated energy system 100 in the embodiment of the present application, a data center cluster including multiple data centers is established around a wind-solar power station, and each data center is equipped with corresponding energy storage to form an energy storage system. Considering the uncertainty of renewable energy output and the amount of data load arriving at each moment, a power grid is introduced and a certain capacity of CCHP system (including one or more PGU generator sets) is configured to supply power to the data center cluster when the wind-solar output is insufficient. At the same time, the CCHP system can also convert heat energy into cold energy to relieve the refrigeration pressure of the data center air conditioning system.

[0033] In the embodiment of the present application, the load types of the data center cluster are divided into time-shiftable load, space-shiftable load, and rigid load according to different response modes, and are received via a front-end portal server. The time and space scheduling of the load is realized within the data center cluster to avoid the case that the computing power allocation between each data center is greatly different at a certain moment, and to improve the energy utilization efficiency of the data center cluster.

[0034] In the embodiment of the present application, the data center cluster adopts a power supply mode of wind-solar-grid cooperation. On a typical day, the power generation power of a t period, the photovoltaic power generation power are respectively denoted as 、 , and the power grid transmission power is denoted as . In order to meet the energy consumption demand of the data center cluster and cope with the instability of renewable energy affected by seasons and climate, the power grid transmission needs to be used as a guarantee for smooth operation. The clean energy can be preferentially consumed by limiting the power grid transmission power. Specifically, the power grid transmission power does not exceed the maximum power grid transmission power, as shown in the following formula: (1) The energy storage system adopts distributed energy storage. According to the load conditions of each data center, a corresponding capacity of energy storage can be configured for each data center. For a single energy storage, only the current data center can be powered, and charging or discharging cannot be performed at the same time. However, for an energy storage cluster, each energy storage does not affect each other, and charging and discharging can be performed at the same time within the energy storage cluster.

[0035] To reduce the electricity cost of data center, the energy storage system in the embodiment of the present application adopts the power distribution scheme of charging in the peak period of renewable energy output and discharging in the trough period of renewable energy output, and allows surplus power to be sold to the power grid on the premise of priority to meet the load demand. For a single data center, the power of the energy storage in period t of a certain typical day is defined as: (2) (3) wherein, , are the charging power and discharging power of the energy storage in period t of a certain typical day respectively; , are the maximum charging power and maximum discharging power of the energy storage respectively.

[0036] For a single data center, the power of the energy storage in period t of a certain typical day is defined as: (4) (5) wherein, is the self-discharge rate of the energy storage system; , are the charging and discharging efficiencies of the energy storage system respectively; is the unit scheduling duration; is the supporting energy storage capacity of a single data center (i.e., the energy storage planning capacity), and it is assumed that the initial power of the energy storage is 0 every day.

[0037] In the embodiment of the present application, according to different response modes, the data load is divided into time-shiftable load, space-shiftable load and rigid load. Since the load task amount of each typical day is relatively constant, the data load arrival condition of the whole year is replaced by typical data here. For a single data center, the total amount of data load arriving at a data center in period t is defined as: (6) wherein, , , are the time-shiftable load amount, space-shiftable load amount and rigid load amount of the data center arriving in period t respectively.

[0038] Time-shiftable load (TSL): The time-shiftable load arriving at a data center can be delayed to the next period for processing, but cannot be transferred to other data centers in the cluster. The time-shiftable load amount of the data center arriving in period t is defined as: (7) (8) In the formula, represents the percentage of time-transferable load in the total load; represents the actual time-transferable load processed during time period t. This represents the amount of workload that can be transferred from the previous time period to time period t. The initial value for each day is 0. ; The amount of load that can be transferred from time period t to the next time period.

[0039] (9) (10) Spatially Transferable Load (STL): Spatially transferable load arriving at a data center can be transferred to other data centers in the cluster for processing, but cannot be delayed to the next time period. Define the amount of spatially transferable load arriving at the data center in time period t as: (11) (12) In the formula, The percentage of space-transferable load in the total load; The actual transferable load processed during time period t; The space transferable load that is moved from other data centers in the cluster to this data center during time period t; The space-transferable load that is moved from this data center to other data centers in the cluster during time period t.

[0040] (13) Virtualization technology is a key technology for enabling online spatial migration of workloads between different data centers. It can virtualize server hardware resources, and the total number of virtual machines in a data center cluster must satisfy the following relationship.

[0041] (14) (15) In the formula, M represents the supply capacity of a single virtual machine; The number of virtual machines required to complete the transferable load scheduling of the data center cluster space during time period t; This represents the total number of virtual machines configured in the data center cluster.

[0042] Rigid Load (RL): Rigid loads arriving at a data center cannot be transferred to other data centers in the cluster, nor can they be delayed to the next time period; they must be processed immediately by that data center. Define the rigid load arriving at the data center during time period t as... . Data center mainly includes IT equipment, air conditioning system, lighting system and power distribution system. IT equipment energy consumption accounts for about 50%, air conditioning system energy consumption accounts for about 37%, and since the lighting system energy consumption is small, in order to simplify the model, only the power distribution system energy consumption is considered in the lighting system and power distribution system. The energy consumption of air conditioning system and power distribution system has strong correlation with the energy consumption of IT equipment.

[0043] It is assumed that the relevant parameters and configurations of each data center in the data center cluster are completely the same except for the construction scale and load task source. For a single data center, the total energy consumption of the data center at time t on a typical day is defined as follows: (16) In the formula, is the energy consumption of IT equipment at time t on a typical day; is the energy consumption of air conditioning system at time t on a typical day (EC electric refrigeration); is the energy consumption of power distribution system at time t on a typical day.

[0044] For a single data center, the energy consumption of IT equipment at time t on a typical day is defined as follows: (17) (18) (19) In the formula, , are the energy consumptions of IT equipment in idle and full load states respectively; is the average utilization rate of IT equipment CPU at time t on a typical day; is the number of IT equipment in working state at time t on a typical day; is the actual processing load at time t; is the processing rate of a single IT equipment.

[0045] (20) (21) (22) In the formula, is the upper limit of the utilization rate of data center IT equipment CPU; is the configuration number of data center IT equipment; k is the maximum number of IT equipment per unit area that can be configured in the data center; is the planning area of the data center.

[0046] The energy consumption model for the air conditioning system adopts a first-order equivalent thermal parameter model: (twenty three) In the formula, C and R are the equivalent heat capacity and thermal resistance of the data center, respectively; , These represent the indoor and outdoor temperatures of the data center during time period t. , These represent the heat dissipation of IT equipment and the cooling capacity of the air conditioning system during time period t.

[0047] The relationship between heat dissipation and energy consumption of IT equipment, and between cooling capacity and energy consumption of air conditioning systems, is as follows: (twenty four) (25) In the formula, The conversion coefficient between electrical power and heat; This refers to the performance coefficient of the air conditioning system.

[0048] The normal operation of a data center requires a sufficient and stable power supply. Power distribution units, also known as "rack rooms," typically operate independently of the server room and are usually located in different parts of the server room. They can simultaneously meet the power requirements of IT equipment and server racks. They generally consist of generators, transformers, distribution cabinets, and uninterruptible power supplies (UPS). The energy consumption of the power distribution system during time period t on a typical day is defined as: (26) In the formula, This is the energy consumption coefficient of the power distribution system.

[0049] like Figure 1 As shown, the CCHP system includes a PGU generator set, an absorption chiller, and a water source heat pump. Specifically, the CCHP system mainly consists of a power supply system and a refrigeration subsystem. The power supply system primarily generates electricity by burning natural gas or other primary energy sources through the PGU generator set. The refrigeration subsystem primarily recovers waste heat from the PGU generator set and cooling water through an absorption chiller for cyclic refrigeration.

[0050] In this embodiment of the invention, natural gas is selected as the input energy source, and the power generation of the PGU generator set during time period t on a typical day is defined as: (27) In the formula, This represents the natural gas consumption during time period t on a typical day. This refers to the calorific value of natural gas. The power conversion efficiency of the PGU; The time interval of the time slot.

[0051] The power generation of the PGU generator set of the CCHP system is also limited by the number of PGU generator sets and the capacity of the generator set: (28) In the formula, is the number of PGU generator sets of the CCHP system; is the maximum power generation of a single PGU generator set.

[0052] The energy consumption of the absorption chiller can be divided into thermal energy consumption and electrical energy consumption, of which the electrical energy consumption is relatively small and can be ignored, and the thermal energy consumption can be expressed as follows: (29) (30) In the formula, is the cooling capacity of the absorption chiller in the t period; is the coefficient of performance of the absorption chiller; is the constant-pressure specific heat capacity of the cooling water; is the mass flow rate of the cooling water of the absorption chiller; , are the return water temperature and the outlet water temperature of the cooling water of the absorption chiller, respectively.

[0053] (31) The heat power recovered in the t period during the power generation of the PGU generator set is as follows: (32) In the formula, is the PGU heat recovery efficiency.

[0054] The water source heat pump recovers waste heat from the cooling water and improves its quality, and delivers the waste heat to the absorption chiller for recooling, and the recovered heat power is as follows: (33) The energy consumption of the water source heat pump is: (34) In the formula, is the mass flow rate of the cooling water of the water source heat pump; , are the return water temperature and the outlet water temperature of the cooling water of the water source heat pump, respectively; is the water source heat pump heat recovery efficiency; is the coefficient of performance of the water source heat pump.

[0055] In embodiments of the present application, the computing device can be configured to perform the data center cluster integrated energy capacity configuration method 300 under the computing power and power coupling. The data center cluster integrated energy capacity configuration method 300 under the computing power and power coupling of the present application will be described below.

[0056] A computing device 200 according to an embodiment of the present application is introduced below.

[0057] Figure 2 A schematic diagram of a computing device 200 according to an embodiment of the present application is shown. As shown in the 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 computing device, the processing unit 202 can be implemented as a processor. The system memory 204 includes, but is not limited to, volatile (e.g., random access memory (RAM)), non-volatile (e.g., read-only memory (ROM)), flash memory, or any combination thereof. According to one aspect, the system memory 204 includes an operating system 205. Figure 2 According to one aspect, the operating system 205 is suitable for controlling the operation of the computing device 200, for example. In addition, the example is practiced in conjunction with a graphics library, other operating systems, or any other application programs, and is not limited to any particular application or system. In

[0058] The basic configuration is shown by those components within the dashed line. 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, for example, magnetic disks, optical disks, or tape. Such additional storage is illustrated in Figure 2 by removable storage 209 and non-removable storage 210. Figure 2 As stated above, according to one aspect, program modules 203 are stored in the system memory 204. According to one aspect, the program modules 203 can include one or more application programs, and the present application does not limit the type of application programs, for example, the application programs can 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.

[0059] In embodiments according to the present application, the program modules 203 include a plurality of program instructions for performing the data center cluster integrated energy capacity configuration method 300 under the computing power and power coupling of the present application.

[0060]

[0061] ​According to one aspect, the examples can be practiced with circuitry integrated on a single integrated circuit chip, with separate electronic components, in a package, or with other components arranged on a single printed circuit board, with other examples implemented using two or more printed circuit boards. For example, the examples can be implemented using a combination of one or more of the following technologies: quantum computing, optical computing, and / or superconductor-based computing. According to one aspect, the examples can be practiced in a system-on-a-chip (SOC) that includes discrete electronic elements integrated on a single integrated circuit (IC) or within multiple ICs. According to another aspect, the examples can be practiced using other technologies capable of reconfiguring themselves while operating. For example, by using artificial intelligence based on a neural network, expert system, or the like. Figure 2 According to one aspect, the examples can be practiced with circuitry integrated on a single integrated circuit chip, with separate electronic components, in a package, or with other components arranged on a single printed circuit board, with other examples implemented using two or more printed circuit boards. For example, the examples can be implemented using a combination of one or more of the following technologies: quantum computing, optical computing, and / or superconductor-based computing. According to one aspect, the examples can be practiced in a system-on-a-chip (SOC) that includes discrete electronic elements integrated on a single integrated circuit (IC) or within multiple ICs. According to another aspect, the examples can be practiced using other technologies capable of reconfiguring themselves while operating. For example, by using artificial intelligence based on a neural network, expert system, or the like.

[0062] According to one aspect, the computing device 200 can also have one or more input device(s) 212 such as a keyboard, a mouse, a pen, a microphone, a touch input device, etc. One or more output device(s) 214 such as a display, speakers, a printer, etc. can also be included. The aforementioned devices are examples and others can be used. The computing device 200 can include one or more communication connections 216 allowing communications with other computing devices 218. Examples of suitable communication connections 216 include, but are not limited to: RF transmitter, receiver, and / or transceiver circuitry; universal serial bus (USB), parallel, and / or serial ports.

[0063] The term computer readable media as used herein includes computer storage media. Computer storage media can include volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, or program modules. The system memory 204, the removable storage device 209, and the non-removable storage device 210 are all computer storage media examples (i.e., memory storage.) Computer storage media can include Random Access Memory (RAM), Read Only Memory (ROM), Electronically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium which can be used to store information and which can be accessed by computing device 200. According to an aspect, any such computer storage media can be part of computing device 200. Computer storage media does not include a modulated data signal or other propagated data signal.

[0064] According to an aspect, communication media is embodied by computer readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave or other transport mechanism, and includes any information delivery media. According to an aspect, the term "modulated data signal" describes a signal that has one or more characteristics set or changed in such a manner as to encode 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.

[0065] In embodiments according to the present application, the computing device 200 is configured to perform the data center cluster integrated energy capacity configuration method 300 under the computing power and power coupling. The computing device 200 includes one or more processors, and one or more readable storage media storing program instructions, which when configured to be executed by the one or more processors, cause the computing device to perform the data center cluster integrated energy capacity configuration method 300 under the computing power and power coupling in embodiments of the present application.

[0066] In some embodiments, the computing device 200 performing the data center cluster integrated energy capacity configuration method 300 under the computing power and power coupling in embodiments of the present application can be a terminal or a server.

[0067] The data center cluster integrated energy capacity configuration method 300 under the computing power and power coupling in embodiments of the present application is described in detail as follows.

[0068] Figure 3A flowchart of a method 300 for configuring a comprehensive energy capacity of a data center cluster under a power-computing coupling according to an embodiment of the present application is shown. As shown in Figure 3 The method 300 for configuring the comprehensive energy capacity of the data center cluster under the power-computing coupling includes the following steps 310-340.

[0069] At step 310, the computing device 200 constructs an upper target function with a minimum total annual cost of the comprehensive energy system planning of the data center cluster as an upper target, and then constructs an upper planning model based on the upper target function.

[0070] As described above, in the embodiment of the present application, the comprehensive energy system of the data center cluster includes the data center cluster (including a plurality of data centers), an energy storage system, and a CCHP system.

[0071] The total annual cost of the comprehensive energy system planning of the data center cluster is composed of an equivalent annual investment cost, an annual operation and maintenance cost, an annual energy cost, a green certificate cost, and a power sale income of the energy storage system, wherein the equivalent annual investment cost is used to represent the full life cycle cost of the comprehensive energy system planning of the data center cluster converted into an internal cost.

[0072] At step 320, the computing device 200 constructs a lower target function with a minimum average absolute deviation of the net load of the data center cluster as a lower target, and then constructs a lower running model based on the lower target function. Here, the average absolute deviation of the net load of the data center cluster is the average value of the absolute deviation of the net load of the data center cluster in each period. The absolute deviation of the net load is the difference between the total energy consumption of the data center cluster and the water source heat pump energy consumption of the CCHP system and the wind power, the photovoltaic power, the grid power, the CCHP system power, and the energy storage system power.

[0073] At step 330, the computing device 200 can obtain a double-layer optimization model for the comprehensive energy capacity configuration of the data center cluster based on the upper planning model and the lower running model.

[0074] At step 340, the computing device 200 can use a hierarchical sequence method to solve the double-layer optimization model for the comprehensive energy capacity configuration of the data center cluster to obtain a capacity configuration result. The capacity configuration result is used to indicate the number of IT equipment configurations and the supporting energy storage capacity of each data center in the data center cluster, the number of PGU generator sets in the CCHP system, and the wind power, the photovoltaic power, and the CCHP system power in each period.

[0075] It should be noted that the hierarchical sequence method can convert the double-layer optimization problem into a single-layer solving problem to improve the solving efficiency of the double-layer optimization model for the comprehensive energy capacity configuration of the data center cluster.

[0076] In some embodiments, the total annual cost of the data center cluster comprehensive energy system planning (composed of equivalent annual investment cost, annual operation and maintenance cost, annual energy cost, green certificate cost, and energy storage system electricity sales income) is minimized as the upper goal, and the upper goal function is constructed as shown in the following formula.

[0077] (35) Wherein, the equivalent annual investment cost is used to represent the full life cycle cost of the data center cluster comprehensive energy system planning converted into the equivalent annual internal cost. Specifically, the full life cycle cost of the data center cluster comprehensive energy system planning can be converted into the equivalent annual investment cost by the following formula.

[0078] (36) In the formula, r is the discount rate; y is the design life of the data center cluster; 、 、 、 The construction investment cost of the energy storage system, the data center cluster, the CCHP system, and the virtual machine, respectively.

[0079] It should be understood that the equivalent annual investment cost can be determined according to the full life cycle cost of the data center cluster comprehensive energy system planning and the discount rate, the design life of the data center cluster. The full life cycle cost of the data center cluster comprehensive energy system planning is the sum of the construction investment cost of the energy storage system, the data center cluster, the CCHP system, and the virtual machine.

[0080] The construction investment cost of the energy storage system, the data center cluster, the CCHP system, and the virtual machine is shown in the following formulas (37)-(40), respectively.

[0081] (37) (38) (39) (40) In the formula, 、 、 、 The configuration cost of the unit capacity energy storage, the unit area of the data center, the single PGU generator set in the CCHP system, and the single virtual machine, respectively.

[0082] It should be understood that the construction investment cost of the energy storage system, the data center cluster, the CCHP system, and the virtual machine is related to the configuration cost of the unit capacity energy storage, the unit area of the data center, the single PGU generator set in the CCHP system, and the single virtual machine, respectively.

[0083] In some embodiments, the annual operation and maintenance cost is related to the annual operation and maintenance cost of the energy storage system, the operation and maintenance cost of the data center cluster and the PGU generator set (the PGU generator set of the CCHP system) in a typical day, and the number of typical days in a year. Specifically, the annual operation and maintenance cost can be calculated according to the following formula.

[0084] (41) In the formula, is the annual operation and maintenance cost of the energy storage system; , is the operation and maintenance cost of the data center cluster and the PGU generator set in a typical day, respectively; is the number of typical days in a year.

[0085] The calculation formula of the annual operation and maintenance cost of the energy storage system, the operation and maintenance cost of the data center cluster in a typical day, and the operation and maintenance cost of the PGU generator set in a typical day is as follows: (42) (43) (44) In the formula, , , is the operation and maintenance cost of the unit capacity energy storage, the unit energy consumption of the data center cluster, and the unit power generation of the PGU generator set, respectively.

[0086] As can be seen from the above formula, the annual operation and maintenance cost of the energy storage system, the operation and maintenance cost of the data center cluster in a typical day, and the operation and maintenance cost of the PGU generator set in a typical day are related to the operation and maintenance cost of the unit capacity energy storage, the unit energy consumption of the data center cluster, and the unit power generation of the PGU generator set, respectively.

[0087] In some embodiments, the annual energy consumption cost is related to the grid electricity fee, the renewable energy electricity fee, and the natural gas fee of the data center cluster to meet the energy demand in a typical day, and the number of typical days in a year. Specifically, the annual energy consumption cost can be calculated according to the following formula.

[0088] (45) In the formula, , , is the grid electricity fee, the renewable energy electricity fee, and the natural gas fee of the data center cluster to meet the energy demand in a typical day, respectively.

[0089] The specific calculation formula of the grid electricity fee, the renewable energy electricity fee, and the natural gas fee of the data center cluster to meet the energy demand in a typical day is as follows: (46) (47) (48) wherein, , , , are the grid electricity cost, the wind power cost, the photovoltaic power cost, and the natural gas cost of the data center cluster in the t-th time period of a typical day, respectively; is the external gas purchase amount of the data center cluster in the t-th time period of a typical day.

[0090] It should be understood that the grid electricity cost generated by the data center cluster in a typical day is related to the grid electricity cost of each time period of a typical day and the grid power transmitted by the grid accessed by the data center cluster. The renewable energy electricity cost generated by the data center cluster in a typical day is related to the wind power cost and the wind power generation power of each time period of a typical day, and the photovoltaic power cost and the photovoltaic power generation power. The natural gas cost generated by the data center cluster in a typical day is related to the natural gas cost of each time period of a typical day and the external gas purchase amount.

[0091] In some embodiments, the green certificate cost is related to the number of green certificates purchased by the data center cluster to meet the renewable energy consumption requirement (renewable energy consumption constraint condition) and the price of a single green certificate in a typical day. Specifically, the green certificate cost can be calculated by the following formula.

[0092] (49) wherein, is the number of green certificates purchased by the data center cluster to meet the renewable energy consumption requirement in a typical day; is the price of a single green certificate.

[0093] In some embodiments, the energy storage system electricity sale revenue is related to the grid electricity cost of each time period of a typical day, the electricity sale power of each energy storage in the energy storage system, and the number of typical days in a year. Specifically, the calculation formula of the energy storage system electricity sale revenue is as follows.

[0094] (50) In some embodiments, the upper-layer constraint condition needs to be considered when constructing the upper-layer planning model. That is, in step 310, the upper-layer planning model can be constructed based on the upper-layer objective function and the upper-layer constraint condition.

[0095] Specifically, the upper-layer constraint condition includes the investment capability constraint condition, the configuration capacity constraint condition, the energy output constraint condition, and the renewable energy consumption constraint condition.

[0096] Wherein, the investment capacity constraint condition indicates that the equivalent annual investment cost should not exceed the maximum planning investment limit, and is specifically shown in the following formula: (51) The configuration capacity constraint condition indicates that, for a single data center (assuming that the initial power of energy storage per day is 0), the power of energy storage equipped by the data center in each period of a typical day does not exceed the supporting energy storage capacity of the data center, as shown in (5) above; the configuration quantity of IT equipment per unit area of the data center (i.e., the ratio of the configuration quantity of IT equipment of the data center to the planning area of the data center) does not exceed the maximum configurable IT equipment quantity per unit area of the data center, as shown in formula (22) above; and the number of PGU generator sets in the CCHP system does not exceed the maximum configuration quantity of PGU generator sets, as shown in formula (28) above.

[0097] It should be noted that, considering the power generation efficiency of energy equipment and maximizing the proportion of renewable energy supply, the energy output constraint condition is set.

[0098] The energy output constraint condition specifically indicates that the power transmission power of the power grid does not exceed the maximum power transmission power of the power grid, as shown in formula (1) above; the power generation power of the CCHP system (i.e., the power generation power of the PGU generator set of the CCHP system) does not exceed the product of the maximum power generation power of a single PGU generator set and the number of PGU generator sets in the CCHP system, as shown in formula (28) above; due to the uncertainty of supply-demand on both sides, there may be a short-term mismatch between supply and demand, and in the period when wind and light resources are sufficient and data load is flat, renewable energy will be abandoned, therefore, the actual power transmission power of wind power does not exceed the upper limit of wind power output (i.e., the actual power transmission power of wind power does not exceed the wind power generation power), the actual power transmission power of photovoltaic power does not exceed the upper limit of photovoltaic power output (i.e., the actual power transmission power of photovoltaic power does not exceed the photovoltaic power generation power), and is specifically shown in the following formulas (53) and (54).

[0099] (52) (53) The renewable energy consumption constraint condition indicates that the renewable energy consumption rate needs to be controlled within a reasonable range, and here, the renewable energy consumption rate is defined as the proportion of the actual consumed renewable energy power in the total power consumption of the system, as shown in formula (54) below. In addition, when the physical consumption amount (the actual consumed renewable energy power) is insufficient, the renewable energy consumption rate is allowed to be supplemented within a certain limit range by purchasing green certificates, in other words, the actual consumed renewable energy power is the sum of the wind power generation power, the photovoltaic power generation power and the green certificate supplemented renewable energy power, as shown in formula (55) below.

[0100] (54) (55) In the embodiments of the present application, for the lower layer operation model, the typical day 24h of the data center cluster comprehensive energy system is selected as the scheduling scale to realize the economic stability scheme of the daily operation of the comprehensive energy system.

[0101] In some embodiments, in step 320, the average absolute deviation of the net load of the data center cluster can be expressed as the following formula.

[0102] (56) (57) (58) As can be seen from formula (57), the absolute deviation of the net load of the data center cluster is the difference between the total energy consumption of the data center cluster and the water source heat pump energy consumption of the CCHP system and the wind power, the photovoltaic power, the power grid transmission power, the CCHP system power and the energy storage system power (i.e. the energy storage system charging and discharging power, wherein the discharging is positive).

[0103] As can be seen from formula (58), the average absolute deviation of the net load of the data center cluster is the average value of the absolute deviation of the net load of the data center cluster in each period.

[0104] It should be noted that using the average absolute deviation of the net load can effectively realize peak clipping and valley filling, and bring significant effect for realizing stable operation of the system.

[0105] In some embodiments, when constructing the lower layer operation model, the lower layer constraint condition also needs to be considered. That is, in step 320, the lower layer operation model can be constructed based on the lower layer objective function and the lower layer constraint condition.

[0106] Specifically, the lower layer constraint condition includes the energy conservation constraint condition, the data center cluster service quality constraint condition, the energy utilization efficiency constraint condition and the data center machine room temperature constraint condition.

[0107] Among them, the energy conservation constraint condition includes the electric energy balance constraint condition and the thermal energy balance constraint condition.

[0108] The electrical energy balance constraint condition is represented as: the sum of the power generated by various types of power sources in the data center cluster integrated energy system (including wind power, photovoltaic power, grid power, CCHP system power, and energy storage system power) is equal to the power consumed by the data center cluster (including the total energy consumption of the data center cluster and the water source heat pump energy consumption of the CCHP system). That is, the sum of the wind power, photovoltaic power, grid power, CCHP system power (i.e., the power generated by the PGU generator set of the CCHP system), and energy storage system power in each period is equal to the sum of the total energy consumption of the data center cluster and the water source heat pump energy consumption of the CCHP system.

[0109] Specifically, the electrical energy balance constraint condition can be represented as follows: (59) In the formula, the left side of the equation is the sum of the power generated by various types of power sources in the data center cluster integrated energy system, including wind power, photovoltaic power, grid power, CCHP system power (i.e., the power generated by the PGU generator set of the CCHP system), and energy storage system power. The right side of the equation is the power consumed by the data center cluster, i.e., the sum of the total energy consumption of the data center cluster and the water source heat pump energy consumption.

[0110] It can be understood that the electrical energy balance constraint condition reflects the balance relationship of the data center cluster integrated energy system in the electrical energy level, which can ensure that the energy supply in the system can meet the power demand of the data center cluster.

[0111] The thermal energy balance constraint condition is represented as: the sum of the heat generated by the IT equipment in the data center cluster and the waste heat generated by the PGU generator set of the CCHP system in each period is equal to the sum of the thermal energy consumption of the absorption chiller of the CCHP system, the refrigeration capacity of the air conditioning system of the data center cluster, and the heat recovered and transported by the water source heat pump of the CCHP system. That is, in the thermal energy transfer process of the data center cluster integrated energy system, the thermal energy in each period has the following balance relationship: (60) (61) (62) In formula (60), the left side of the equation is the heat generated by the IT equipment of the data center cluster and the waste heat generated by the PGU generator set of the CCHP system; the right side of the equation is the thermal energy consumption of the absorption chiller of the CCHP system, the refrigeration capacity of the air conditioning system of the data center cluster, and the heat recovered and transported by the water source heat pump of the CCHP system.

[0112] It can be understood that the thermal energy balance constraint condition reflects the generation, consumption and transfer of thermal energy in the data center cluster comprehensive energy system, ensures that the heat generated by the data center cluster can be reasonably utilized and processed, thereby maintaining the stable temperature of the data center room and ensuring the normal operation of the equipment, while realizing the cascade utilization of energy and improving the comprehensive utilization rate of energy.

[0113] The data center cluster service quality constraint condition is that: the total amount of data load received by the data center cluster in each period is greater than or equal to the total amount of data load processed by the data center cluster in each period, the total amount of data load processed by the data center cluster in each period is greater than or equal to the total amount of data load that must be processed by the data center cluster in each period; and the data center cluster service quality index is less than or equal to 2000, wherein the data center cluster service quality index is the average value of the queue length of the data center cluster in each period, and the queue length of the data center cluster in each period is the difference between the total amount of data load received by the data center cluster in each period and the total amount of data load processed by the data center cluster.

[0114] It should be noted that the data load received by the data center cluster (the data load arriving at the data center cluster) includes interactive load and batch processing load, wherein the interactive load includes space transferable load and rigid load, and the batch processing load includes time transferable load. Specifically, the interactive load belongs to non-delayable tasks, including space transferable load and rigid load, which need to be processed immediately; while the batch processing load belongs to delayable tasks, which are generally time transferable load and can be delayed for processing in the next period.

[0115] In the data center cluster, data load can be uniformly received through the front-end portal server of the data center cluster, and then the data center cluster plans and allocates computing power according to the actual requirements of the received data load. The scheduling model of the data center cluster is as follows: (63) (64) (65) In the formula, is the total amount of data load received by the data center cluster in period t; is the total amount of data load processed by the data center cluster in period t; is the total amount of data load that must be processed by the data center cluster in period t.

[0116] Wherein, the total amount of data load received by the data center cluster in period t is greater than or equal to the total amount of data load processed by the data center cluster in period t, and the total amount of data load processed by the data center cluster in period t is greater than or equal to the total amount of data load that must be processed by the data center cluster in period t. That is, the arrival curve , the leaving curve and the minimum leaving curve satisfy the following relationship: (66) The data center cluster service quality index defined by the waiting queue length can be expressed as follows: (67) (68) As can be seen from the above formula, the data center cluster service quality index is the average value of the waiting queue length of the data center cluster in each period, and the waiting queue length of the data center cluster in each period is the difference between the total amount of data load received by the data center cluster and the total amount of data load processed by the data center cluster.

[0117] And, to ensure the processing quality of the data center cluster to the data load, the data center cluster service quality index is defined.

[0118] The energy utilization efficiency constraint condition represents that the air conditioning system energy consumption of the data center cluster is instantaneously matched with the renewable energy power generation power.

[0119] It is worth noting that in the prior art, PUE is taken as the core index for measuring the energy efficiency of the data center, which is defined as the ratio of the total energy consumption of the data center cluster to the IT equipment energy consumption. However, the traditional PUE index focuses on energy efficiency evaluation, and it is difficult to quantify the renewable energy consumption level. Therefore, in some embodiments of the present application, the renewable energy utilization rate is represented by the improved index PUE-R of the variable temperature control, as shown in the following formula: (69) (70) By dynamically adjusting the air conditioning refrigeration power and the computer room temperature threshold, wherein when the renewable energy output is more, the refrigeration power of the air conditioning system is increased, and the data center computer room temperature is appropriately reduced within the safety range; when the renewable energy output is less, the refrigeration power of the air conditioning system is reduced, and the data center computer room temperature is allowed to be appropriately increased within the safety range. The air conditioning system energy consumption of the data center cluster is instantaneously matched with the renewable energy power generation power , thereby improving the green electricity penetration rate and reducing the dependence on grid power supply.

[0120] The data center computer room temperature constraint condition represents that the indoor temperature of each data center computer room in the data center cluster in each period needs to be maintained within a predetermined temperature range. Specifically, it is shown in the following formula: (71) It should be understood that by setting the data center machine room temperature constraint condition, it can be ensured that the data center machine room temperature is in a safe and reasonable range, avoiding the influence of high or low temperature on the normal operation of IT equipment and increasing the equipment failure rate, and also being associated with the design of the refrigeration and heat dissipation system of the data center.

[0121] In some embodiments, the hierarchical sequence method can be used to solve the data center cluster comprehensive energy capacity configuration bi-level optimization model in step 340 by the Gurobi solver. Among them, the upper layer objective function and the lower layer objective function can be defined in the Gurobi solver in advance.

[0122] Specifically, the specific process of solving the data center cluster comprehensive energy capacity configuration bi-level optimization model by the hierarchical sequence method includes the following steps D1~D5: D1, obtain the initial input parameters and the user-selected robust method (for handling the uncertainty problem of data load arrival amount), the initial input parameters including the initial wind power, the initial photovoltaic power, and the initial data load amount.

[0123] D2, based on the initial input parameters, predict the initial CCHP system power generation power (i.e., the PGU generator set power generation power of the CCHP system).

[0124] D3, based on the initial input parameters and the initial CCHP system power generation power, solve the upper layer planning model to obtain the initial upper layer solving result and the lower bound value of the objective function. The initial upper layer solving result includes the initial value of the IT equipment configuration quantity, the supporting energy storage capacity, and the PGU generator set quantity of the CCHP system of each data center.

[0125] D4, based on the initial upper layer solving result, use the robust method to describe the wind and light output uncertainty to solve the lower layer running model to obtain the initial lower layer solving result and the upper bound value of the objective function. The initial lower layer solving result includes the optimized wind power, photovoltaic power, CCHP system power generation power, and data load amount.

[0126] D5, judge whether the convergence condition is reached. Specifically, the convergence condition includes: the difference between the upper bound value of the objective function and the lower bound value of the objective function is less than the absolute difference threshold, or the relative error between the upper bound value of the objective function and the lower bound value of the objective function is less than the relative difference threshold, or the iteration number reaches the maximum iteration number.

[0127] If the convergence condition is reached, the capacity configuration result can be obtained based on the initial upper layer solving result and the initial lower layer solving result.

[0128] If the convergence condition is not reached, the initial input parameters are updated based on a relaxation factor (to avoid too drastic update process) to obtain new input parameters, and the steps D2-D5 are iteratively performed (based on the new input parameters, a new CCHP system power generation is predicted; based on the new input parameters and the new CCHP system power generation, the upper planning model is solved to obtain a new upper solving result and a lower bound of the objective function; based on the new upper solving result, a robust method is used to describe the wind and light output uncertainty to solve the lower running model to obtain a new lower solving result and an upper bound of the objective function; it is judged whether the convergence condition is reached) until the convergence condition is reached, and the capacity configuration result is obtained based on the latest upper solving result and the latest lower solving result.

[0129] Therefore, by using the hierarchical sequence method, the double-layer optimization problem can be converted into a single-layer solving problem to improve the solving efficiency.

[0130] In some embodiments, the robust method can be a distribution robust method.

[0131] It should be noted that in the data center energy configuration scenario, the uncertainty of wind and light output makes it difficult to accurately determine the probability distribution. According to the distribution robust method, the uncertainty of wind and light output is described by defining a fuzzy set containing multiple possible distributions.

[0132] In some embodiments of the present application, a fuzzy set based on moments is used to define the range of distribution by using known statistical moment information. Specifically, in the foregoing step D4, the distribution robust method can be used to construct a fuzzy set (i.e., a fuzzy set based on moments) based on the mean and variance of wind and light power generation, and the uncertainty of wind and light output is described based on the constructed fuzzy set. The fuzzy set constructed based on the mean and variance of wind and light power generation can cover all probability distributions that satisfy the moment condition, as shown in the following formula: (72) (73) In the formula, is the wind and light power generation at t period; , are the estimated values of the mean and variance of wind and light power generation, respectively; , are parameters set according to the uncertainty degree of data, used to adjust the size of the fuzzy set.

[0133] In some embodiments, the lower running model can be simplified to the following form based on the column and constraint generation algorithm (C&CG): (74) Wherein, C is a coefficient matrix of the objective function; A, B, F, D, are coefficient matrices of respective constraint conditions; x, y are variables involved, including output variables of respective devices in the lower-layer operation model; a, b, d are constant column vectors.

[0134] Further, the C&CG algorithm can be used for solving, and the specific steps are as follows: Set the upper limit of the objective function as and the lower limit as Solve the main problem according to the initial scenario to obtain an optimal solution, and take the objective function value of the main problem as a new lower limit.

[0135] Take the obtained main problem solution into the sub-problem, solve the sub-problem to obtain the objective function value of the sub-problem and the value of the uncertain variable of the corresponding scenario, and take the objective function value of the sub-problem as a new upper limit.

[0136] The convergence precision of the given algorithm is If , a final robust optimization operation scheduling scheme is obtained, and the operation cost in the scheme obtained by the lower layer is transmitted to the upper-layer planning model as one of the components of the upper-layer target.

[0137] It should be noted that in the hierarchical sequence method, different weights need to be assigned to the two targets and priorities need to be set, and the weight assignment method includes subjective weight assignment and objective weight assignment. The result of subjective weight assignment is seriously dependent on the subjective understanding of experts, and the result of objective weight assignment may not reflect the actual importance of the index.

[0138] Therefore, in some embodiments of the present application, when the hierarchical sequence method is used to solve the data center cluster comprehensive energy capacity configuration double-layer optimization model, the subjective and objective integrated weight assignment method based on the analytic hierarchy process and the entropy weight method can be used to assign weights to the upper-layer target and the lower-layer target, respectively, wherein the upper-layer target weight can be set as , the lower-layer target weight can be set as , and the specific calculation formula is as follows: (75) Wherein, , respectively are the weights obtained by the analytic hierarchy process and the entropy weight method; r is a preference coefficient of the decision maker for subjective and objective factors, and the value is between 0 and 1.

[0139] In some embodiments, a data center cluster is to be built around a certain wind power station, with a construction period of 30 years and a discount rate of 8%. Through k-means clustering analysis method, the wind speed and illumination data of each season are clustered to obtain the historical data of four seasonal typical days. Then, a long short-term memory network (LSTM) is used to build a prediction model to predict the wind power and photovoltaic output based on the meteorological data of the seasonal typical days. Each seasonal typical day lasts for 91 days, and the initial value of the uncertainty conservatism of wind and light output is set to 6.

[0140] To verify the adaptability of different robust optimization methods to the source-load bilateral uncertainty of the data center cluster (wind and light output fluctuation and data load dynamic change), taking the initial value of the uncertainty conservatism of wind and light output and data load as 6 and the confidence level as 0.9 as the benchmark conditions, the two-stage robust method and the distribution robust method are compared and analyzed from three core dimensions of algorithm solving performance, device configuration rationality and system cost benefit. The relevant results are shown in Tables 1 and 2.

[0141] Table 1 Comparison of robust method algorithm results From the algorithm solving efficiency and the optimization effect of the objective function, the distribution robust method shows better adaptability. In terms of convergence characteristics, the two-stage robust method needs to go through 7 iterations to converge, and the final solution gap reaches 627758, with a convergence rate of only 61.36%; while the distribution robust method only needs 5 iterations to converge, and the final gap is reduced to 395024, with a convergence rate of 83.62%, which significantly improves the calculation efficiency and also improves the solving accuracy. In terms of the objective function, the total annual cost of the system under the two-stage robust method is about 7.8×10 7 6 yuan, and the average absolute deviation of the net load is about 700; the total annual cost of the system under the distribution robust method is reduced to 6.8×10 7 4 yuan, and the average absolute deviation of the net load is reduced to 450, which not only effectively controls the economic cost, but also improves the interactive friendliness of the data center cluster and the power grid by stabilizing the net load fluctuation, and reduces the pressure on the power grid peak regulation.

[0142] Table 2 Comparison of robust method capacity configuration results The difference in device configuration of the two robust methods essentially reflects the impact of optimization conservatism on resource configuration redundancy. In terms of core device configuration, the two-stage robust method needs to cover all the boundaries of the uncertainty set, and configures 3 CCHP units to cope with extreme scenarios, while the distribution robust method only needs 2 CCHP units to meet the operation demand by dynamically adapting the uncertainty through the probability distribution fuzzy set, reducing the investment of 1 redundant unit; The total amount of virtual machines is maintained at 26 sets under both methods, indicating that the computing power demand under the load scheduling mechanism is relatively stable, but the green certificate purchase amount is significantly different, and the two-stage robust method needs to purchase 270 green certificates to meet the compliance requirements, while the distribution robust method only needs 190 green certificates, and the green electricity consumption compliance cost is reduced by 33.3%. In terms of operation indicators, the two-stage robust method has a wind curtailment rate of 2.95%, slightly better than the distribution robust method of 3.83%, but there is still a 0.008% light curtailment rate, while the distribution robust method completely eliminates the light curtailment phenomenon, and the utilization rate of photovoltaic resources is more sufficient; The quality of service index of the two is similar, and the device utilization efficiency is maintained at 100%, indicating that the distribution robust method effectively avoids the problem of idle resources caused by excessive conservative optimization while ensuring the quality of data processing services and efficient operation of devices. From the analysis of fine-grained configuration balance, the IT equipment and energy storage capacity of some data centers (such as IDC3 and IDC6) under the two-stage robust method fluctuate greatly, and there is an over-configuration phenomenon; While the energy storage capacity and IDC construction scale of the distribution robust method are more in line with the actual operation demand, and the configuration balance is significantly improved.

[0143] The difference in system cost and benefit is mainly due to the change in investment and operation cost caused by the configuration conservatism of the two methods. Figure 4 A schematic diagram of the cost and benefit comparison of the two robust methods according to some embodiments of the present application is shown. As Figure 4 shown, the total cost of the two-stage robust method reaches 7.8×10 7 6 per year, while the distribution robust method reduces to 6.9×10 7 4 per year, with a cost difference of nearly 13%. From the cost composition analysis, the two-stage robust method has higher equivalent annual investment cost and annual energy cost due to redundant configuration; At the same time, its more green certificate purchase demand further increases the compliance cost, intensifying the total cost pressure. In terms of benefits, the energy storage electricity selling income of the two-stage robust method does not achieve significant improvement, but instead increases the system operation burden due to excessive reliance on energy storage for peak shaving and frequency modulation; The distribution robust method optimizes the matching of load scheduling and device output without relying on large-scale energy storage arbitrage to balance the cost and benefit, and has stronger operation robustness.

[0144] In summary, the two-stage robust method tends to over-conservative configuration in the worst-case scenario due to the pursuit of strict coverage of the boundary of the uncertainty set, which easily leads to investment redundancy, high operating cost and low convergence efficiency of the algorithm. The distribution robust method adjusts the response strategy to uncertainty through dynamic adjustment of the fuzzy set of probability distribution, and optimizes resource allocation in high-probability scenarios first. On the premise of ensuring the quality of service and equipment utilization of the data center cluster, the algorithm efficiency, configuration rationality and economy are synergistically improved, which is more suitable for the optimization and solution of large-scale data center cluster integrated energy system in the uncertain environment of source-load bilateral.

[0145] In some embodiments, to verify the effectiveness of the data center cluster integrated energy capacity configuration double-layer optimization model proposed in the application, five differentiated planning scenarios as shown in Table 3 are set, and comparison is made from three aspects of objective function optimization, configuration scheme rationality and cost-benefit balance.

[0146] Table 3 Scenario setting The advantages of the standard scenario considering bilateral uncertainty, CCHP configuration, load scheduling and C&CG algorithm are verified, and the results are shown in Figures 5 to 8

[0147] Figure 5 The target value comparison diagram under each scenario according to some embodiments of the application is shown. As shown in Figure 5 From the optimization objective, the standard scenario shows better Pareto equilibrium characteristics. The total annual cost is about 6.92×10 7 million, which is lower than that of the scenario without using the C&CG algorithm and the scenario without setting the CCHP. Although the total cost of scenario three is lower, it needs to bear double the equivalent annual investment cost. At the same time, the average absolute deviation of net load of the standard scenario is about 590, which is significantly better than that of the scenario without considering bilateral uncertainty and the scenario without considering load scheduling, indicating that it effectively reduces the pressure of power grid peak regulation while ensuring economy.

[0148] Figure 6 The capacity configuration comparison diagram under each scenario according to some embodiments of the application is shown. Figure 7 The auxiliary capacity configuration comparison diagram under each scenario according to some embodiments of the application is shown.

[0149] As shown in Figure 6 and Figure 7 ​As shown, in terms of configuration scheme, the resource matching degree of the standard scenario is more in line with the actual demand. In terms of core equipment, the standard scenario configures 2 CCHP units, which not only meets the cold, heat and electricity step utilization demand, but also avoids single energy dependence. Scenario three does not set CCHP, and the green certificate purchase amount reaches 1096, which increases by 477% compared with 190 of the standard scenario, and the compliance cost increases sharply; the number of virtual machines is maintained at 26 sets, and the computing power is flexibly allocated through load scheduling, and the equipment utilization efficiency is maintained at 100%. In terms of operation index, the standard scenario has a wind curtailment rate of 3.83%, which is lower than that of scenario two and scenario four, and can reduce the wind and light curtailment phenomenon by 11.95% at most, and the green electricity consumption capacity is better; the service quality index is 11.67%, which is in the relatively optimal interval of each scenario, and guarantees business continuity.

[0150] Figure 8 The cost-benefit comparison diagram under each scenario according to some embodiments of the application is shown. As shown Figure 8 In terms of cost-benefit structure, the standard scenario realizes full-cycle cost balance, and in all the scenarios set, it can reduce the cost input by 5.43% at most. The equivalent annual investment cost is 7439500 yuan, which is much lower than 13613400 yuan of scenario three, and no CCHP needs to rely on traditional energy facility expansion; the annual operation cost is 57685340 yuan, which is higher than 43097400 yuan of scenario three, but through CCHP waste heat recovery and green electricity consumption, the green certificate cost is only 1477 yuan. In terms of income, the standard scenario has a storage electricity sales income of 84800 yuan, which is lower than 483200 yuan of scenario three, but the latter relies on large-scale energy arbitrage to maintain economy, and the operation stability is insufficient, while the standard scenario realizes cost structure smoothing through device configuration, load scheduling and energy cooperation, which is more suitable for long-term operation demand.

[0151] In summary, the standard scenario realizes the balance of economy, stability and compliance through multi-element cooperation, and verifies the effectiveness and applicability of the model in the green electricity nearby consumption scenario.

[0152] In summary, the data center cluster comprehensive energy capacity configuration method 300 according to the power coupling of the computing power and electricity of the application takes the minimum annual total cost of the data center cluster comprehensive energy system planning as the target to build an upper planning model to optimize the data center cluster size and capacity configuration, takes the minimum average absolute deviation of the net load of the data center cluster as the target to build a lower running model to dynamically adjust the energy output and load scheduling, thereby obtaining a data center cluster comprehensive energy capacity configuration double-layer optimization model, and solving to obtain the optimized capacity configuration result. Based on this, the multi-energy complementation and computing power-electricity deep coupling in the data center cluster can be realized, the green electricity nearby consumption capacity and the comprehensive energy utilization efficiency are improved, the system cost is reduced while the comprehensive energy configuration is optimized.

[0153] Furthermore, the application adopts a distributed robust optimization method, can more efficiently adapt to the bilateral uncertainty of wind and light power generation and data load, further optimizes resource allocation and reduces system cost.

Claims

1. A method for configuring the integrated energy capacity of a data center cluster under computing power-electricity coupling, executed in a computing device, comprising: The upper-level objective function is to minimize the total annual cost of the integrated energy system planning for the data center cluster. Based on the upper-level objective function, an upper-level planning model is constructed. The integrated energy system for the data center cluster includes a data center cluster, an energy storage system, and a CCHP system. The total annual cost consists of the equivalent annual investment cost, annual operation and maintenance cost, annual energy consumption cost, green certificate cost, and electricity sales revenue from the energy storage system. The equivalent annual investment cost is used to represent the cost of converting the total life cycle cost of the integrated energy system planning for the data center cluster to the cost within the equivalent annual period. The lower-level objective function is constructed with the minimum net load average absolute deviation of the data center cluster as the lower-level objective. A lower-level operation model is then constructed based on the lower-level objective function. The net load average absolute deviation is the average net load absolute deviation of the data center cluster in each time period. The net load absolute deviation is the difference between the total energy consumption of the data center cluster and the energy consumption of the water source heat pump of the CCHP system and the power of wind power generation, photovoltaic power generation, power grid transmission, CCHP system power generation, and energy storage system power. Based on the upper-level planning model and the lower-level operation model, a two-layer optimization model for the comprehensive energy capacity configuration of the data center cluster is obtained. Using the hierarchical sequence method, the two-level optimization model for the comprehensive energy capacity configuration of the data center cluster is solved to obtain the capacity configuration results. The capacity configuration results are used to indicate: the number of IT equipment configurations and supporting energy storage capacity of each data center in the data center cluster, the number of PGU generator sets in the CCHP system, and the wind power generation, photovoltaic power generation and CCHP system power generation at each time period.

2. The method as described in claim 1, wherein, The data center cluster includes multiple data centers, each containing IT equipment and air conditioning systems, and is adapted to perform online spatial migration of loads between different data centers via virtual machines; The equivalent annual investment cost is suitable for determination based on the full life cycle cost, discount rate, and design life of the data center cluster integrated energy system plan. The full life cycle cost is the sum of the construction investment costs of the energy storage system, data center cluster, CCHP system, and virtual machine.

3. The method as described in claim 1 or 2, wherein, The annual operation and maintenance cost is related to the annual operation and maintenance cost of the energy storage system, the operation and maintenance cost of the PGU generator set of the data center cluster and CCHP system on a typical day, and the number of typical days in a year. The annual energy cost is related to the grid electricity cost, renewable energy electricity cost, and natural gas cost generated by the data center cluster on a typical day, as well as the number of typical days in a year; The cost of the green certificate is related to the number of green certificates purchased in a typical Sunrise data center cluster and the price of a single green certificate. The electricity sales revenue of the energy storage system is related to the grid electricity price at each time period on a typical day, the electricity sales power of each energy storage unit in the energy storage system, and the number of typical days in a year.

4. The method according to any one of claims 1-3, wherein, Based on the aforementioned objective function, a higher-level planning model is constructed, including: A higher-level planning model is constructed based on the aforementioned objective function and constraints. The constraints include investment capacity constraints, configuration capacity constraints, energy output constraints, and renewable energy consumption constraints. The investment capacity constraint means that the equivalent annual investment cost does not exceed the maximum planned investment limit; The configuration capacity constraints are as follows: the amount of electricity stored in the data center during each typical daytime period does not exceed the supporting energy storage capacity of the data center; the number of IT equipment configured per unit area of ​​the data center is the ratio of the number of IT equipment configured in the data center to the planned area of ​​the data center, which does not exceed the maximum number of IT equipment that can be configured per unit area of ​​the data center; and the number of PGU generator sets in the CCHP system does not exceed the maximum number of PGU generator sets. The energy output constraints are as follows: the grid transmission power does not exceed the grid's maximum transmission power; the CCHP system's power generation does not exceed the product of the maximum power generation of a single PGU generator unit and the number of PGU generator units in the CCHP system; and the actual wind power transmission does not exceed the wind power generation power, and the actual photovoltaic power transmission does not exceed the photovoltaic power generation power. The renewable energy consumption constraint means that the renewable energy consumption rate is within the consumption rate range, where the renewable energy consumption rate is the proportion of the actual renewable energy power consumed in the total power consumption of the system, and the renewable energy power is the sum of wind power generation, photovoltaic power generation and renewable energy power supplemented by green certificates.

5. The method according to any one of claims 1-4, wherein, Based on the aforementioned lower-level objective function, a lower-level operational model is constructed, including: The upper-level planning model is constructed based on the lower-level objective function and lower-level constraints. The lower-level constraints include energy conservation constraints, data center cluster service quality constraints, energy utilization efficiency constraints, and data center room temperature constraints. The energy conservation constraints include electrical energy balance constraints and thermal energy balance constraints. The electrical energy balance constraint states that the sum of wind power generation, photovoltaic power generation, grid transmission power, CCHP system power generation, and energy storage system power in each time period is equal to the sum of the total energy consumption of the data center cluster and the water source heat pump energy consumption of the CCHP system in each time period. The thermal energy balance constraint states that the sum of the heat generated by IT equipment in the data center cluster and the waste heat generated by the PGU generator set of the CCHP system in each time period is equal to the sum of the heat energy consumption of the absorption chiller of the CCHP system, the cooling capacity of the air conditioning system of the data center cluster, and the heat recovered and transported by the water source heat pump of the CCHP system in each time period. The data center cluster service quality constraints are as follows: the total data load received by the data center cluster in each time period is greater than or equal to the total data load processed by the data center cluster in each time period; the total data load processed by the data center cluster in each time period is greater than or equal to the total data load that the data center cluster must process in each time period; and the data center cluster service quality index is less than or equal to 2000, wherein the data center cluster service quality index is the average waiting queue length of the data center cluster in each time period, and the waiting queue length of the data center cluster in each time period is the difference between the total data load received by the data center cluster in each time period and the total data load processed by the data center cluster in each time period. The energy efficiency constraint means that the energy consumption of the air conditioning system of the data center cluster is matched with the power generation of renewable energy in real time. The temperature constraint condition for the data center computer room means that the indoor temperature of each data center computer room in the data center cluster must be maintained within a predetermined temperature range at any time.

6. The method of claim 5, wherein, The data load received by the data center cluster includes interactive load and batch processing load. The interactive load includes spatially transferable load and rigid load, and the batch processing load includes time-transferable load.

7. The method according to any one of claims 1-6, wherein, The hierarchical sequence method is used to solve the two-level optimization model for the comprehensive energy capacity configuration of the data center cluster, and the capacity configuration results are obtained. The steps include: The initial input parameters and the selected robust method are obtained, wherein the initial input parameters include initial wind power generation, initial photovoltaic power generation, and initial data load. Based on the initial input parameters, predict the initial CCHP system power generation. Based on the initial input parameters and the initial CCHP system power generation, the upper-level planning model is solved to obtain the initial upper-level solution results and the lower bound of the objective function. Based on the initial upper-level solution results, a robust method is used to describe the uncertainty of wind and solar power output in order to solve the lower-level operation model and obtain the initial lower-level solution results and the upper bound of the objective function. Determine whether the convergence condition has been met. The convergence condition includes: the difference between the upper bound and the lower bound of the objective function is less than the absolute difference threshold; or the relative error between the upper bound and the lower bound of the objective function is less than the relative difference threshold; or the number of iterations reaches the maximum number of iterations. If the convergence condition is met, the capacity configuration result is obtained based on the initial upper-layer solution result and the initial lower-layer solution result; If the convergence condition is not met, the initial input parameters are smoothly updated based on the relaxation factor to obtain new input parameters, and the above steps are iteratively executed until the convergence condition is met to obtain the capacity configuration result.

8. The method of claim 7, wherein, Robust methods are used to describe uncertainties in wind and solar power output, including: Using the sparse bar method, a fuzzy set is constructed based on the mean and variance of wind and solar power generation, and the uncertainty of wind and solar power output is described based on the fuzzy set.

9. The method according to any one of claims 1-8, wherein, The hierarchical sequence method is used to solve the two-level optimization model for the comprehensive energy capacity configuration of the data center cluster, including: When solving the two-layer optimization model for the comprehensive energy capacity configuration of the data center cluster using the hierarchical sequence method, the subjective and objective integrated weighting method based on the analytic hierarchy process and the entropy weight method is used to set the upper-layer target weight and the lower-layer target weight.

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, the program instructions including instructions for processing the method as described in any one of claims 1-9.

Images