Data center microgrid planning method considering water resource constraint and risk measurement

Abstract

The invention discloses a data center micro-grid planning method considering water resource constraint and risk measurement, and relates to the field of computer technology application, and the method not only improves the accuracy and reliability of data center planning through data-driven modeling and a DRO-CVaR algorithm under the condition of considering water resource constraint and multiple uncertainties, but also improves the accuracy and reliability of data center planning through data-driven modeling and a DRO-CVaR algorithm. In addition, balance between economical efficiency and robustness is achieved, and the method has remarkable practical value and theoretical significance.

Description

Technical Field

[0001] The present invention relates to the field of computer technology applications, and in particular to a data center microgrid planning method taking into account water resource constraints and risk metrics. Background Art

[0002] With the rapid development of cloud computing services and the exponential growth in the number of data centers, the economic and environmental performance of data center energy use has become one of the research hotspots of the power system. The huge energy consumption of data centers has led operators to equip photovoltaic units to reduce the high cost of purchasing electricity and the power usage efficiency (PUE) parameters of data centers. On the one hand, they will form a data center microgrid and reduce their own energy costs. On the other hand, among the large-scale data center construction projects coordinated by the "East Data West Computing" policy, 90% of the new ones are large and super-large data centers. More and more data center operators are beginning to choose water-cooled equipment units with higher efficiency and lower power consumption to replace the original air-cooled equipment units. However, the resulting water use problem has also attracted the attention of data center operators and the government. Taking a single data center with a power capacity of 20MW and an open cooling tower for heat dissipation as an example, the theoretical annual evaporation water consumption is about 200,000 m 3 , which is equivalent to the water consumption required by 3,000 urban residents. This poses a major challenge to the economic and environmental protection considerations of data center planning.

[0003] Defects and shortcomings of the existing technology:

[0004] 1. Insufficient water consumption modeling: Currently, domestic and foreign scholars mostly construct a mapping relationship between data center water consumption and power consumption based on the traditional data center planning model, so as to realize the construction and solution of data center planning model taking into account water footprint. However, existing studies use the method of water usage effectiveness (WUE) to map the water consumption of data centers into a linear function of power consumption, while there are few studies that take into account the fine-grained modeling of water consumption by outdoor temperature, which makes it difficult to quantitatively evaluate the impact of temperature on data centers, and thus leads to insufficient or excessive planning capacity of renewable energy units such as photovoltaics; on the other hand, few studies have considered the water resource constraints of the system or region where the water-cooled data center is located, which makes the data center planning scheme calculated by traditional methods possibly infeasible due to constraint violations.

[0005] 2. Insufficient uncertainty handling: Compared with the power load of traditional microgrids, in addition to the output uncertainty of the photovoltaic units equipped in the data center microgrid, the data and computing service requests that the data center servers need to process also have greater randomness, which makes the data center server load have higher uncertainty. This uncertainty, together with the outdoor temperature, makes the power and water consumption of the data center cooling system also uncertain. In existing studies, although DRO has been more widely used than RO and stochastic optimization (SO) algorithms due to its economic and robustness, the DRO model still considers all elements and risk scenarios in the data set, and cannot solve the system planning optimization problem while ignoring some extreme scenarios. Therefore, for some data center operators with strong risk tolerance, it is particularly necessary to develop a DRO model that is compatible with the risk management model. In addition, the DRO models in existing studies mostly use CCG, Benders decomposition and other methods to convert and split the min-max-min model in the microgrid planning and operation, but when the objective function is the expectation and integral form in risk control models such as CVaR instead of the linear form, the existing methods will find it difficult to calculate its optimal solution.

[0006] 3. Insufficient model validity: Existing studies have the following three problems: (1) Only the impact of load or photovoltaic output uncertainty on system operation planning is considered. Few studies consider the impact of temperature on data center operation planning, which makes the model insufficiently valid; (2) For the few studies that consider multiple uncertainties for planning problems, existing methods mostly describe the joint distribution of multiple uncertainties through ideal probability distribution functions rather than data-driven methods, resulting in insufficient reliability of planning schemes; (3) The existing DRO algorithm only optimizes decisions for the worst probability distribution in the fuzzy set, which makes the formulation of planning strategies take into account the most extreme cases. Although the robustness of the model is guaranteed, when decision makers have the willingness and ability to avoid risks, the resulting model has the problem of insufficient economic efficiency. Summary of the invention

[0007] In order to solve the above problems, the present invention provides a data center microgrid planning method taking into account water resource constraints and risk metrics, comprising the following steps:

[0008] According to the workload of the water-cooled data center and the heat transfer method of the water-cooled data center, the water footprint of the water-cooled data center during operation is quantified, and the total cost of the water-cooled data center during the planning period is taken as the target to build a data center planning model that considers water resource constraints;

[0009] Based on the data center planning model, according to the uncertainty of data center load and photovoltaic unit power, and taking into account the uncertainty of outdoor temperature of the data center, the three uncertainties are converted into conditional risk values. Through the DRO algorithm, risks are actively managed under multiple uncertainties.

[0010] Preferably, in the process of constructing a data center planning model that takes water resource constraints into consideration, based on the workload of the water-cooled data center, the power consumption of the data center is obtained according to the total power of the data center in time period t, the equivalent power consumption of processing batch loads, interactive loads, and the power consumption of the cooling system, and the water footprint is quantified based on the power consumption.

[0011] Preferably, in the process of quantifying the water footprint, the water footprint is quantified based on the electricity consumption, according to the heat transfer mode of water circulation-evaporation, by setting the server temperature constraints of the data center and the temperature change and evaporation rate constraints of the cooling water.

[0012] Preferably, in the process of obtaining the total cost of the data center during the planning period, the total cost of the data center during the planning period is obtained based on minimizing the annual equivalent planning cost, the annual equivalent operating cost and the water cost of the data center.

[0013] Preferably, in the process of obtaining the total cost of the data center during the planning period, the total cost of the data center during the planning period is expressed as:

[0014] C=C inv +365·C ope

[0015]

[0016] Where C is the total cost of the data center microgrid during the planning period; C inv is the investment cost of data center equipment; r is the discount rate; y s is the equipment life; S is the equipment set, including photovoltaic units, chillers and energy storage units; c s is the installation cost per unit capacity of the equipment; Q s Install capacity for equipment; C ope is the total operating cost of the data center; C buy is the electricity purchase cost; C carbon is the carbon emission cost, C water is the water purchase cost; C w Penalize data centers for excessive water use; C s The operating cost of energy storage; is the time-of-use electricity price during period t; p W is the penalty price when the water purchased by the data center exceeds the total water consumption limit; maxis the city’s restriction on the total water consumption of data centers, and WUE is the city’s policy restriction on the WUE of data centers; is the charging and discharging loss cost coefficient of the energy storage equipment.

[0017] Preferably, in the process of converting the three uncertainties into conditional risk value, the three uncertainties include the equivalent electrical load P of the data center: t DC , the output of photovoltaic units installed in data centers P t PV and the temperature T of the data center’s location t O .

[0018] Preferably, in the process of converting the three uncertainties into conditional risk values, the historical data of the three uncertainties are processed and merged by normalizing and adding multipliers, the original three uncertainties are converted into a one-dimensional uncertainty problem, and the multiplier combination that can optimize the planning economy is calculated.

[0019] The present invention also discloses a data center microgrid planning system taking into account water resource constraints and risk metrics for implementing the method, comprising:

[0020] The model building module is used to quantify the water footprint of the water-cooled data center during operation according to the workload of the water-cooled data center and the heat transfer method of the water-cooled data center, and to build a data center planning model that takes into account water resource constraints with the total cost during the planning period of the water-cooled data center as the target;

[0021] The active risk management module is used to convert the three uncertainties into conditional risk values ​​based on the data center planning model, according to the uncertainty of data center load and photovoltaic unit power, and taking into account the uncertainty of outdoor temperature of the data center. Through the DRO algorithm, risks are actively managed in a multiple uncertainty environment.

[0022] The present invention discloses the following technical effects:

[0023] Taking into account water resource constraints and multiple uncertainties, the present invention not only improves the accuracy and reliability of data center planning, but also achieves a balance between economy and robustness through data-driven modeling and DRO-CVaR algorithm, which has significant practical value and theoretical significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 This is the data center system structure described in the present invention;

[0026] Figure 2 The data center evaporative cooling chiller system structure of the present invention;

[0027] Figure 3 It is a flow chart of the DRO-CVaR algorithm described in the present invention. DETAILED DESCRIPTION

[0028] To make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application.

[0029] like Figure 1-3 As shown, the present invention provides a data center microgrid planning method taking into account water resource constraints and risk metrics, which specifically includes the following contents:

[0030] 1. Data center microgrid model:

[0031] 1.1 Water-cooled data center operation model:

[0032] A typical data center system consists of servers, energy storage units, photovoltaic units, cooling systems, etc. The power supply of the data center is first supplied by photovoltaic units, and the insufficient part is supplemented by energy storage units and power grid purchases. On the one hand, the energy storage unit can reduce the operating cost of the data center through peak and valley arbitrage of time-of-use electricity prices, and on the other hand, it can absorb the excessive output of the photovoltaic unit to avoid the phenomenon of abandoned light. The data center system structure studied in this invention is as follows Figure 1 shown.

[0033] Considering the business characteristics of data centers, the workload of data centers consists of batch workloads that can be processed late and interactive workloads that cannot be processed late

[15] . The power consumption of data centers can be expressed as:

[0034] P t DC =P t DC.bat +P t DC.int +P t DCW (1)

[0035] Where: P t DC , P t DC.bat , P t DC.int , P t DCW They are the total power of the data center in period t, the equivalent power consumption of processing batch loads, interactive loads, and the power consumption of the cooling system. The batch workload can be delayed to another period before the processing deadline. The operation mode of the batch workload: Formula (2a) indicates that the total load of the data center remains unchanged before and after the batch workload is delayed, and formula (2b) indicates the transfer of the batch workload within the allowed delay:

[0036]

[0037]

[0038] Where: P t DC,ad is the power of the data center after the transfer during period t; T max is the upper limit of the number of delay periods. The power balance of the data center is shown in formula (3a). Formula (3b) indicates that the maximum charge and discharge power of the energy storage unit is limited by its planned capacity. Formula (3c) indicates that the energy storage unit is balanced at the beginning and end of each regulation cycle. Formula (3d) indicates the energy storage unit's power constraint, which prevents the energy storage device from exceeding the power limit at any time.

[0039]

[0040] Where: P t buy , P t PV , P t s They are the power purchase amount, photovoltaic power generation and energy storage unit power of the data center during period t respectively; is the planned capacity of the energy storage unit; Δt is the scheduling time interval; E0 is the initial value of the energy storage in the 0 period; Emax 、E min are the upper and lower limits of the energy storage device’s power; t′ is any period from 1 to 24.

[0041] The equivalent carbon emissions of data centers come from the carbon emissions of power plants during the electricity purchase process, which can be described by formula (4):

[0042] F t C =η C [P t DC -P t PV ] + (4)

[0043] Where: F t C is the carbon footprint of the data center in period t; η C is the ratio coefficient between the carbon emissions of the power grid and the power generation.

[0044] 1.2 Data Center Water Footprint Model:

[0045] Water-cooled data centers transfer the heat generated by data center servers through water circulation and evaporation to reduce the cooling power consumption of the data center, thereby reducing the power consumption outside the server, that is, the power of the cooling system, and thus reducing the PUE of the data center. Figure 2 It is a typical evaporative cooling chiller system structure.

[0046] Figure 2 In the process, the data center transfers the hot air generated by the server heat dissipation to the evaporator, and the hot air is cooled by the chilled water and turned into cold air and returned to the data center. In this process, the heated chilled water will pass through the condenser, absorb the heat released by the liquefied refrigerant and turn back into chilled water with a lower temperature for a new round of cooling in the data center. The cooling of the condenser is completed by the cooling tower, which inputs low-temperature cooling water into the condenser and recycles the high-temperature cooling water output by the condenser. After recovery, it is cooled by evaporation and other methods. The evaporation amount at this time is the water consumption of the data center.

[0047] During operation, the water used in the data center comes from the evaporation of cooling water, and the evaporation rate increases with the increase of cooling water temperature. The server temperature constraint of the data center: Formula (5a) updates the server temperature of the data center according to the outdoor temperature and the heat generated by the data center. Formula (5b) represents the upper and lower limit constraints of the data center server temperature. Formula (5c) represents the heat generated during the operation of the data center.

[0048]

[0049] Where: Tt DC is the temperature of the data center server in period t; U wt , A wt , C wt They are the heat loss coefficient, effective heat loss area, and heat capacity of the server respectively; They are the power generation and heat production and the expected heat demand of the data center’s photovoltaic power generation during period t; The heat generated by the data center during period t; T t O is the outdoor temperature during period t; and T DC are the upper and lower limits of the data center server temperature; κ DC is the ratio coefficient of heat generation to power in the data center.

[0050] Cooling water temperature change and evaporation rate constraints: Formula (6a) updates the data center cooling water temperature according to the outdoor temperature and the heat generated by the data center. Formula (6b) represents the upper and lower limit constraints of the data center cooling water temperature. Formula (6c) represents the data center heat dissipation water consumption. Formula (6d) represents the total power consumption of the chiller in the data center cooling system. Formula (6e) indicates that the power of the data center chiller will not exceed its planned capacity.

[0051]

[0052] Where: T t DCW is the temperature of the cooling water in the data center during period t; R o is the equivalent thermal resistance of the cooling unit room in the data center; C r The heat capacity of the data center cooling unit room; U w , A w They are the heat loss coefficient and heat dissipation area of ​​the cooling system room; is the cooling capacity of the cooling unit in the data center during period t; and T DCW W is the upper and lower limits of the cooling water temperature in the data center; t DCW is the cooling water consumption of the data center during period t; η DC is the proportional coefficient between water evaporation rate and water temperature; η DCW It is the ratio coefficient between the power of the data center cooling system and the cooling capacity of the cooling unit; Plan capacity for data center chillers.

[0053] Based on the above content, formula (7) can be further used to quantify the water footprint of the data center during operation. The water footprint of the data center includes the water used for cooling the server itself and the water used for power generation and cooling in the power plant caused by purchasing electricity.

[0054] W t =W t DCW +η E [P t DC -P t PV ] + (7)

[0055] Where: W t is the water footprint of the data center during period t; η E is the ratio coefficient of power grid cooling water and power generation, where [·] + =max{·,0}.

[0056] 2. Data center microgrid planning considering water resource constraints:

[0057] Based on the data center energy consumption characteristic model constructed above, we take the total cost of the data center planning period as the target and construct a data center planning model that considers water resource constraints. The data center planning cost includes minimizing the annual equivalent planning cost, annual equivalent operating cost and the water cost of the data center. The corresponding data center planning objective function can be expressed as: Formula (8a) is the total cost of the data center microgrid during the planning period, Formula (8b) is the equipment investment cost of the data center, and Formula (8c) is the operating cost of the data center.

[0058] C=C inv +365·C ope (8a)

[0059]

[0060] Where: C is the total cost of the data center microgrid during the planning period; C inv is the investment cost of data center equipment; r is the discount rate; y s is the equipment life; S is the equipment set, including photovoltaic units, chillers and energy storage units; c s is the installation cost per unit capacity of the equipment; Q s Install capacity for equipment; C ope is the total operating cost of the data center; C buy is the electricity purchase cost; C carbon is the carbon emission cost, C water is the water purchase cost; C w Penalize data centers for excessive water use; C s The operating cost of energy storage; is the time-of-use electricity price during period t; p W is the penalty price when the water purchased by the data center exceeds the total water consumption limit; max is the city’s restriction on the total water consumption of data centers, and WUE is the city’s policy restriction on the WUE of data centers; is the charging and discharging loss cost coefficient of the energy storage equipment.

[0061] Since (8c) contains nonlinear terms, an auxiliary variable P is introduced. t s,+ =max{0,P t s} and P t s,- =max{0,-P t s}Convert the energy storage operating cost in equation (8c) into:

[0062]

[0063] The carbon emission price and water purchase price of the data center adopt the step-by-step price form of formula (10) and formula (11).

[0064]

[0065]

[0066] Where: μ c,0 is the benchmark carbon emission rights price; υ is the electricity carbon emission factor of the coal-fired power unit; C c,p is the carbon quota; η is the carbon quota increment; σ is the carbon price step increment; m is the number of carbon price steps; μ w,0 is the base water purchase price; C c,w is the water quota; ψ is the water quota increment; ζ is the water price step increment; n is the number of water price steps.

[0067] In summary, the planning model of the data center can be expressed in a compact form as described in formula (12).

[0068]

[0069] 3. DRO-CVaR algorithm framework for data centers based on uncertainty:

[0070] 3.1 Data-driven fuzzy sets of uncertain quantities:

[0071] During the operation of the data center, data center operators will face the economic risk of excessively high "operation + planning" costs. From the existing research, due to the randomness of data user service demand and sunshine intensity, data center server load, photovoltaic units are the two most common uncertain factors. On this basis, we added temperature uncertainty as a research object. Therefore, the uncertainty economic risk influencing factors of the data center can be expressed by the following three uncertainties:

[0072] ① Equivalent electrical load P of the data center t DC ;

[0073] ② Output P of photovoltaic units installed in data centers t PV ;

[0074] ③ Temperature T of the data center’s location t O .

[0075] In traditional risk measurement models, decision makers need to know the probability distribution function of the risk quantity, but in actual situations, due to the limited size of the data set, it is impossible to obtain an accurate probability distribution. Therefore, we use a data-driven approach to process the data set of the model. Considering that the units and numerical values ​​of the three variables of data center load, photovoltaic output and external temperature are somewhat different, and the CVaR fuzzy set that relies on the mean and variance of uncertainties is difficult to handle physical quantities with different dimensions and large numerical differences, the present invention uses normalization and adding multipliers to process and merge the historical data of the three uncertainties, converting the original three uncertainties into a one-dimensional uncertainty problem, and calculating the multiplier combination that can optimize the planning economy. The conversion method is shown in Equation (13)-Equation (15).

[0076]

[0077] Where: DC , PV , TP is the per unit value of the equivalent electrical load, photovoltaic output and outdoor temperature of the data center; DC,0 , PV,0 , TP,0 is the historical data of three sets of variables; DC,base , PV,base , TP,base is the baseline value of the three groups of variables; k DC , k PV , k TP are the multipliers corresponding to the three groups of variables.

[0078] Let T be the number of time periods in a typical day, and the three sets of 1×T dimensional data obtained by equations (12) to (14) are transformed into DC λ PV λ TP ], a set of 1×3T-dimensional uncertainties λ is used to represent the original multidimensional uncertainties. Based on this, the ellipsoidal fuzzy set of the data center uncertainty λ can be constructed.

[0079]

[0080] Where: is a closed convex set containing P, express The set of all probability distributions on . To simplify the expression, the present invention will use D M1 replace γ1≥0 and γ2≥0 are the characteristic parameters of the fuzzy set; M is the number of samples; and are the mean and variance of the random vector λ, respectively, and their expressions are:

[0081]

[0082] 3.2 Characterization of Data Center Uncertainty Based on DRO-CVaR

[0083] The fuzzy set D driven by the uncertain data of load, photovoltaic and temperature constructed in Section 3.1 is M1 We need to construct the DRO-CVaR form of the model corresponding to formula (12) for data center planning decision makers to make decisions based on D M1 decision.

[0084] First, we construct the CVaR form of the objective function through equations (19)-(20) to quantify the risk cost in extreme scenarios. Equation (19) represents the minimum planning cost that does not exceed the expected threshold α, and equation (20) represents the objective function (8a) in extreme scenarios that exceeds α. β (x) is the expected value of the objective function φ β (x).

[0085]

[0086] Where: Ψ(x,α) is the probability that the planning cost does not exceed the expected threshold α, and β is the confidence parameter with a value between 0 and 1. Assuming h(x,α,λ)=max{α,C(x,λ) / (1-β)-βα / (1-β)}, we can get:

[0087]

[0088] Combining the objective functions of RO in (22) and SO in (23), based on the fuzzy set D M1 The data center planning cost under DRO-CVaR can be expressed as shown in formula (24):

[0089]

[0090] By using the duality theory and Schur complement model semi-positive programming transformation method, the data center planning cost calculation problem of formula (24) can be equivalently transformed into the semi-positive programming problem shown in formula (25):

[0091]

[0092] In summary, the optimization problem expressed in formula (24) is transformed into a semi-positive programming form that can be directly solved by a solver. The process of the algorithm proposed in the present invention is as follows: Figure 3 shown.

[0093] In order to more accurately describe the water consumption characteristics of the data center, the present invention establishes a water consumption model of the data center considering the outdoor temperature, the indoor and outdoor heat conduction process and the water resource constraints of the data center, and accordingly proposes a data center microgrid planning model taking into account the water consumption characteristics of the data center;

[0094] Based on the uncertainty of traditional data center load and photovoltaic unit power, the present invention takes into account the uncertainty of outdoor temperature of the data center and converts the three uncertainties into a combination of uncertainties that can be directly used by the conditional value at risk (CVaR). The data-driven method is used to characterize the multiple uncertainties in data center planning, and the parameter combination that can maximize the system economy is obtained through calculation.

[0095] Based on the DRO algorithm, the present invention constructs the DRO-CVaR algorithm by integrating the CVaR model, and proposes a data center planning model for data center operators that can actively manage risks under multiple uncertain environments. The effectiveness of the method proposed in the present invention is verified through examples.

[0096] The present invention has the following advantages:

[0097] 1. More accurate water usage modeling:

[0098] Considering the influence of outdoor temperature: The water footprint model of the data center established in this paper fully considers the influence of outdoor temperature on the water consumption of the data center. Through the heat conduction equation, a model of the cooling water consumption rate of the data center is established, and the outdoor temperature, indoor and outdoor heat conduction process and water resource constraints are considered, so that the model can more accurately reflect the water consumption in actual operation.

[0099] Penalty price constraint: By setting an excess penalty price, the water resource utilization of the data center is restricted, effectively reducing excess water consumption and improving the reliability of data center planning results. For example, taking a single data center with a power capacity of 20MW as an example, the theoretical annual evaporation water consumption is about 200,000 m 3 ,Through penalty price constraints, water consumption can be controlled within a reasonable range, avoiding resource waste and cost increase caused by excessive water use.

[0100] 2. Multiple uncertainty handling:

[0101] Data-driven uncertainty modeling: This paper adopts a data-driven approach to transform the three uncertainties of data center load, photovoltaic output and outdoor temperature into a one-dimensional uncertainty problem, and calculates the multiplier combination that can optimize the planning economy. This method not only considers the joint distribution of multiple uncertainties, but also improves the reliability and scientificity of the model through data-driven methods.

[0102] DRO-CVaR algorithm: Based on the DRO algorithm, the DRO-CVaR algorithm is constructed by integrating the CVaR model. This algorithm can effectively measure the potential risks of planning by adjusting the confidence parameters while balancing robustness and economy. Compared with the traditional DRO algorithm, the DRO-CVaR algorithm can ignore some extreme scenarios, avoid overly conservative planning strategies, and improve the economy of the model. For example, under different risk parameters, the DRO-CVaR algorithm can provide a more economical planning solution while maintaining a certain degree of robustness.

[0103] 3. Improve model effectiveness and economy:

[0104] Comprehensive consideration of multiple uncertainties: This invention not only considers the uncertainty of data center load and photovoltaic output, but also adds temperature uncertainty as a research object. By comprehensively considering these three uncertainties, the model can more comprehensively reflect the uncertainty in the operation of the data center and improve the effectiveness of the model.

[0105] Balance between economy and robustness: The DRO-CVaR algorithm can provide the best planning solution under different risk preferences by adjusting parameters. Decision makers can choose different confidence levels of planning results according to their own risk tolerance to balance the preferences for robustness and economy. For example, by adjusting the characteristic parameters and confidence parameters of the fuzzy set, it is possible to reduce planning costs and improve economy while ensuring a certain degree of robustness.

[0106] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] In the description of the present invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0108] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A data center microgrid planning method considering water resource constraints and risk measurement, characterized in that: The following steps are involved: According to the workload of the water-cooled data center and the heat transfer mode of the water-cooled data center, the water footprint of the water-cooled data center during operation is quantified, and the total cost of the water-cooled data center during the planning period is taken as the target to construct a data center planning model considering water resource constraints; Based on the data center planning model, according to the uncertainty of data center load and photovoltaic unit power, and taking into account the uncertainty of outdoor temperature of the data center, the three uncertainties are converted into conditional risk values, and the risks are actively managed under multiple uncertainty environments through the DRO algorithm.

2. The data center microgrid planning method taking into account water resource constraints and risk metrics according to claim 1 is characterized by: In the process of constructing a data center planning model that considers water resource constraints, based on the workload of the water-cooled data center, the power consumption of the data center is obtained according to the total power of the data center in time period t, the equivalent power consumption of processing batch loads, interactive loads, and the power consumption of the cooling system, and the water footprint is quantified based on the power consumption.

3. The data center microgrid planning method taking into account water resource constraints and risk metrics according to claim 2 is characterized by: In the process of quantifying the water footprint, based on the electricity consumption and according to the heat transfer method in the form of water circulation-evaporation, the water footprint is quantified by setting the server temperature constraints of the data center and the temperature change and evaporation rate constraints of the cooling water.

4. The data center microgrid planning method taking into account water resource constraints and risk metrics according to claim 3 is characterized by: In the process of obtaining the total cost of the data center during the planning period, the total cost of the data center during the planning period is obtained by minimizing the annual equivalent planning cost, the annual equivalent operating cost and the water cost of the data center.

5. The data center microgrid planning method taking into account water resource constraints and risk metrics according to claim 4 is characterized by: In the process of obtaining the total cost of the data center during the planning period, the total cost of the data center during the planning period is expressed as: C=C inv +365·C ope Where C is the total cost of the data center microgrid during the planning period; C inv is the investment cost of data center equipment; r is the discount rate; y s is the equipment life; S is the equipment set, including photovoltaic units, chillers and energy storage units; c s is the installation cost per unit capacity of the equipment; Q s Install capacity for equipment; C ope is the total operating cost of the data center; C buy is the electricity purchase cost; C carbon is the carbon emission cost, C water is the water purchase cost; C w Penalize data centers for excessive water use; C s The operating cost of energy storage; is the time-of-use electricity price during period t; p W is the penalty price when the water purchased by the data center exceeds the total water consumption limit; max is the city’s restriction on the total water consumption of data centers, and WUE is the city’s policy restriction on the WUE of data centers; is the charging and discharging loss cost coefficient of the energy storage equipment.

6. The data center microgrid planning method taking into account water resource constraints and risk metrics according to claim 5 is characterized by: In the process of converting the three uncertainties into conditional risk value, the three uncertainties include the equivalent electrical load of the data center. P t DC , Output of photovoltaic units installed in data centers P t PV and the temperature of the data center's geographic location T t O .

7. The data center microgrid planning method taking into account water resource constraints and risk metrics according to claim 6 is characterized by: In the process of converting the three uncertainties into conditional risk values, the historical data of the three uncertainties are processed and merged by normalizing and adding multipliers, the original three uncertainties are converted into one-dimensional uncertainty problems, and the multiplier combination that can make the planning economically optimal is calculated.

8. The data center microgrid planning method considering water resource constraints and risk metrics according to any one of claims 1 to 7, characterized in that: A data center microgrid planning system considering water resource constraints and risk metrics for implementing the method includes: A model building module is used to quantify the water footprint of the water-cooled data center during operation according to the workload of the water-cooled data center and the heat transfer mode of the water-cooled data center, and to build a data center planning model that considers water resource constraints with the total cost during the planning period of the water-cooled data center as the target; The active risk management module is used to convert the three uncertainties into conditional risk values ​​based on the data center planning model, according to the uncertainty of the data center load and the power of the photovoltaic unit, and taking into account the uncertainty of the outdoor temperature of the data center, and actively manage the risks in a multiple uncertainty environment through the DRO algorithm.

Images