Multi-microgrid data center planning method, system, equipment and medium

Abstract

The invention relates to the technical field of data center planning, and discloses a multi-micro-grid data center planning method, system, equipment and medium, and the method comprises the steps: analyzing the energy flow of a micro-grid according to the power operation characteristics and water network topology structures among a plurality of micro-grids, and obtaining an energy flow index; according to the load operation characteristics and the cooling water characteristics of the plurality of data centers, analyzing energy consumption of the data centers to obtain energy consumption indexes; analyzing the heat conduction process of the data center according to the environment temperature and humidity to obtain temperature and humidity indexes; establishing a data center planning model by taking the minimization of the total planning cost as an objective function and taking an energy flow index, an energy consumption index and a temperature and humidity index as constraint conditions; and solving the data center planning model to obtain a multi-microgrid data center planning scheme. According to the method, the influence of uncertain factors on the operation of the data center can be minimized, and the reliability and adaptability of a multi-microgrid data center planning result are enhanced.

Description

Technical Field

[0001] This invention relates to the field of data center planning technology, and in particular to a planning method, system, equipment and medium for a multi-micronet data center. Background Technology

[0002] Interconnected microgrid systems, due to their distributed energy integration capabilities, are one of the important platforms supporting the stable operation of data centers. However, with the rapid expansion of the number of water-cooled data centers, the "dual high" bottleneck problem of high water consumption of cooling systems and high power consumption of computing equipment has become increasingly prominent, becoming a key factor restricting sustainable development.

[0003] Current microgrid planning methods for data centers mostly focus on load modeling for individual data centers, neglecting the crucial operational characteristics of workload transfer between multiple data centers via information networks. Research on data center water usage models is limited, and those that exist rely heavily on water efficiency, mapping the water usage model to the data center's load model without considering the impact of temperature and humidity on heat transfer during cooling processes, resulting in low accuracy. Furthermore, current research on data center uncertainties in the planning phase largely focuses on the uncertainties of renewable energy output and data center electricity load, failing to consider the uncertainties of electricity purchase prices and environmental humidity within different microgrids. These limitations of existing planning methods directly restrict the accuracy and reliability of data center planning results. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a planning method, system, equipment, and medium for multi-micronet data centers, aiming to minimize the impact of uncertainties on data center operation and enhance the reliability and adaptability of multi-micronet data center planning results.

[0005] In a first aspect, the present invention provides a planning method for a multi-micronet data center, the method comprising: Based on the power operation characteristics and water network topology among multiple microgrids, the energy flow of the microgrids is analyzed to obtain energy flow indicators; Based on the load operation characteristics and cooling water characteristics of multiple data centers, the energy consumption of data centers is analyzed to obtain energy consumption indicators. The heat transfer process of the data center is analyzed based on the ambient temperature and humidity to obtain temperature and humidity indices; A data center planning model is established with the goal of minimizing the total planning cost and the constraints of the energy flow index, the energy consumption index, and the temperature and humidity index. The total planning cost includes construction cost and operating cost. The data center planning model is solved using a preset algorithm to obtain a planning scheme for a multi-micronet data center.

[0006] Furthermore, the step of analyzing the energy flow of the microgrid based on the power operation characteristics and water network topology among multiple microgrids to obtain energy flow indicators includes: Based on the voltage amplitude and active power of each microgrid, the power grid energy flow index is obtained; Based on the water network topology and water load flow balance among multiple microgrids, the water network energy flow index is obtained.

[0007] Furthermore, the step of analyzing the energy consumption of data centers based on the load operation characteristics and cooling water characteristics of multiple data centers to obtain energy consumption indicators includes: Based on the operating characteristics of various loads in the data center, the total energy consumption index of the data center is obtained. The total water consumption of a data center is calculated based on the server cooling water circulation and the power generation cooling water used for electricity purchases.

[0008] Furthermore, the step of obtaining the total energy consumption index of the data center based on the operating characteristics of various loads in the data center includes: Calculate offline load based on cross-network load transfer between various data centers; Calculate the online load based on the operating power consumption of servers in each data center; Calculate the equipment load based on the operating power consumption of the cooling equipment and auxiliary equipment in each data center; The total energy consumption index of the data center is obtained based on the offline load, the online load, and the device load.

[0009] Furthermore, the step of obtaining the total water consumption index of the data center based on the server cooling water circulation and the power generation cooling water caused by electricity purchase includes: Based on the water circulation process and cooling capacity of the data center's cooling system, calculate the water replenishment and daily operating water consumption of the cooling system. Calculate the water consumption of the fresh air system based on the humidity control of the data center's fresh air system. Calculate the power generation cooling water consumption based on the difference between the total energy consumption of the data center and the energy demand of renewable energy sources; The total water consumption of the data center is obtained based on the water replenishment of the cooling system, the water consumption of the daily operation, the water consumption of the fresh air system, and the water consumption for power generation cooling.

[0010] Furthermore, the step of analyzing the heat transfer process of the data center based on ambient temperature and humidity to obtain temperature and humidity indices includes: Based on the impact of server operating heat, external ambient temperature, and computer room humidity on server temperature cooling, server temperature indicators and ambient relative humidity indicators were obtained. Based on the influence of server temperature, external ambient temperature, and cooling capacity of the cooling system on the cooling water temperature, the cooling water temperature index of the cooling system is obtained.

[0011] Furthermore, the step of solving the data center planning model using a preset algorithm to obtain the planning scheme for the multi-micronet data center includes: Historical data of preset uncertainties are processed based on the first and second derivatives in the time dimension to obtain an uncertainty data sequence. Then, the K-means clustering algorithm based on time distance is used to perform cluster analysis on the uncertainty data sequence to obtain typical uncertainty scenarios. Based on the uncertainty data sequence, an uncertainty fuzzy set of the data center is constructed. The objective function is transformed into a two-stage distributed robust optimization model, wherein the first stage model is based on minimizing the construction cost of typical uncertain scenarios, and the second stage model is based on minimizing the operating cost of uncertain fuzzy sets. The distributed robust optimization model is linearized using a linearization method, and the linearized model is optimized and solved using a solver to obtain a planning scheme for a multi-micronet data center.

[0012] Secondly, the present invention provides a planning system for a multi-micronet data center, the system comprising: The data analysis module is used to analyze the energy flow of the microgrid based on the power operation characteristics and water network topology among multiple microgrids, and to obtain energy flow indicators. Based on the load operation characteristics and cooling water characteristics of multiple data centers, the energy consumption of data centers is analyzed to obtain energy consumption indicators. The heat transfer process of the data center is analyzed based on the ambient temperature and humidity to obtain temperature and humidity indices; The planning model construction module is used to establish a data center planning model with the goal of minimizing the total planning cost and with the energy flow index, the energy consumption index, and the temperature and humidity index as constraints. The total planning cost includes construction cost and operating cost. The model solving module is used to solve the data center planning model using a preset algorithm to obtain the planning scheme of the multi-micronet data center.

[0013] Thirdly, embodiments of the present invention also provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0014] Fourthly, embodiments of the present invention also provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0015] This invention provides a planning method, system, equipment, and medium for multi-micronet data centers. By analyzing the operational characteristics of various loads in multi-micronet data centers, this invention can accurately describe the energy consumption process at each stage of the multi-micronet data center. By analyzing the impact of temperature and humidity on heat conduction in the data center, it can accurately describe the water consumption process. Based on a comprehensive consideration of multiple uncertainties, it uses a clustering algorithm improved based on time-series characteristics to generate typical uncertainties and utilizes dynamic uncertainties fuzzy sets to represent the vertex scenarios of uncertainty. Typical scenarios are used to optimize the daily operating costs of the data center, while vertex scenarios are used to minimize losses in the worst-case scenario of the distributed robust optimization problem. While minimizing the impact of uncertainties on data center operation, this invention also enhances the reliability and adaptability of the multi-micronet data center planning results. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the planning method for a multi-micronet data center in an embodiment of the present invention; Figure 2 This is a schematic diagram of the planning system for a multi-micronet data center in an embodiment of the present invention; Figure 3 This is an internal structural diagram of the computer device in an embodiment of the present invention.

[0017] Figure label: 10. Data Analysis Module; 20. Planning Model Construction Module; 30. Model Solving Module. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 The first embodiment of the present invention proposes a planning method for a multi-micronet data center, including steps S10 to S50: Step S10: Based on the power operation characteristics and water network topology among multiple microgrids, analyze the energy flow of the microgrids to obtain energy flow indicators; Step S20: Analyze the energy consumption of data centers based on the load operation characteristics and cooling water characteristics of multiple data centers to obtain energy consumption indicators. Step S30: Analyze the heat transfer process of the data center based on the ambient temperature and humidity to obtain temperature and humidity indices; Step S40: With minimizing the total planning cost as the objective function and the energy flow index, energy consumption index, and temperature and humidity index as constraints, establish a data center planning model. The total planning cost includes construction cost and operating cost. Step S50: Solve the data center planning model using a preset algorithm to obtain a planning scheme for the multi-micronet data center.

[0020] This invention provides a planning method for multi-microgrid data centers. It is applied to scenarios involving interconnected microgrids, where each microgrid contains at least one data center for energy supply. A typical data center system consists of servers, energy storage units, photovoltaic (PV) units, and cooling systems. Power to the data center is primarily supplied by the PV units, with any shortfall supplemented by energy storage units and grid power purchases. The planning method provided by this invention enables collaborative optimization among multiple data centers.

[0021] When planning, the energy flow and energy consumption of the multi-micronet data center are first analyzed based on its topology and operating equipment. The specific steps of the energy flow analysis include: Based on the voltage amplitude and active power of each microgrid, the power grid energy flow index is obtained; Based on the water network topology and water load flow balance among multiple microgrids, the water network energy flow index is obtained.

[0022] In this embodiment, it is assumed that the interconnection between multiple microgrids is via DC lines, and the water network within the multiple microgrid scenario adopts a ring topology. Therefore, the energy flow analysis for multiple microgrids includes grid energy flow analysis and water network energy flow analysis. When performing grid energy flow analysis on multiple microgrids (i.e., multiple microgrids), only the impact of active power on the power system is considered. According to the active power transmission principle of DC circuits, the square difference of the voltage amplitude between two interconnected microgrids is linearly related to the active power transmitted through the transmission lines between them, and the active power injected into the microgrids needs to be kept balanced. Therefore, the grid model between multiple microgrids can be expressed as: (1) In the formula, For the i-th micronet exist Voltage amplitude at time 10:00 , and Each connects to the i-th micronet. and the jth micro-network The Types of transmission lines in Resistance, active power, and active power loss during a given time period. In order to be in Injecting the i-th microgrid at any time The active power.

[0023] This embodiment characterizes the voltage change caused by power flow based on the linear relationship between voltage amplitude difference and active power, thereby establishing a simplified power grid model. This simplified model is adapted to the collaborative optimization of multiple microgrids. The aforementioned power grid model essentially restricts the flow of power grid energy between multiple microgrids. Through this model, physical constraints on the power grid's energy flow can be obtained. It can be understood that the indicators in this embodiment and subsequent embodiments can be represented in model form.

[0024] The energy flow analysis of multi-microgrid water networks is based on the topology of the multi-microgrid water network. According to the principle of water load flow balance between water networks, the hydraulic characteristics of the water network are analyzed, and the energy flow indicators of the water network are quantified by establishing a water network energy flow model. Taking a multi-microgrid scenario where the water network consists of a water distribution pipeline system, pumping stations, a water storage system, and a data center's water-cooled water supply system (i.e., cooling system) as an example, in this embodiment, the water network energy flow model is established by comprehensively considering the hydraulic characteristics of the ring network, pumping station energy consumption, water storage regulation capacity, and the flow demand of the water-cooled system. Specifically, for the water distribution pipeline network, considering the ring topology and pipe roughness, the relationship between pipe section pressure drop and flow rate is described based on hydraulic calculation formulas; for the pumping stations, based on the working principle of centrifugal pumps, their hydraulic characteristics and power consumption are described through parameters such as head, number of units, and flow rate; for the water storage system, the head is dynamically updated through the water load flow balance of inflow and outflow and storage capacity. Based on the above principles, the water network energy flow model can be expressed as: (2) In the formula, , , and Representing the i-th node of the water network and the j-th node Pressure drop, friction coefficient, flow rate, and resistance index of the pipe section. , , , , These represent the head, inherent head, loss coefficient, number of units, and flow rate of the centrifugal pumps in the pumping station, respectively. 、 、 、 These represent the power of the pumping station, the density of water, the acceleration due to gravity, and the efficiency of the pumping station, respectively. , , Represent The i-th node of the time-based water network The water head, inflow-outflow difference, and area of ​​the water storage system. , and Represent The i-th node of the time-based water network The total flow rate, the flow rate supplied to the water cooling system of the data center, and other flows, where Δt represents the time difference.

[0025] This embodiment uses the above-mentioned water network energy flow model to quantify the balance relationship between pipe section pressure drop, pump station power consumption, water storage head and total flow in the water network. Therefore, the above model can be used to characterize the physical constraints of water network energy flow, that is, to obtain water network energy flow index.

[0026] In addition to physical constraint analysis of energy flow from hydroelectric coupling, this embodiment also considers the energy consumption of the power grid and water network. By analyzing the hydroelectric energy consumption of multiple data centers, energy consumption indicators based on physical constraints of hydroelectric energy consumption are determined. These energy consumption indicators can be divided into power grid energy consumption indicators and water network energy consumption indicators. The power grid energy consumption indicators are determined by constructing a power grid load model based on the operating characteristics of various loads in the data centers. The water network energy consumption indicators are determined by constructing a water network energy consumption model based on the server cooling water circulation and renewable energy power generation cooling processes in the data centers. The analysis process for these two indicators is explained below.

[0027] In a preferred embodiment, the analysis steps for the energy consumption indicators of the power grid include: Calculate offline load based on cross-network load transfer between various data centers; Calculate the online load based on the operating power consumption of servers in each data center; Calculate the equipment load based on the operating power consumption of the cooling equipment and auxiliary equipment in each data center; The total energy consumption index of the data center is obtained based on the offline load, the online load, and the device load.

[0028] In this embodiment, based on the operating characteristics of various load devices in the data center, their power consumption can be divided into offline load, online load, and equipment operation load. For offline load, current research on data center offline load only focuses on the time-shiftable load of traditional data centers, essentially shifting less important loads while keeping the total load handled by the data center unchanged. It fails to consider the crucial characteristic of optimizing energy consumption by changing the spatial distribution of load through cross-data center load transfer. Furthermore, because different microgrids have different resource endowments—namely, varying real-time electricity prices, renewable energy processing capabilities, and the carrying capacity of data centers within each microgrid—cross-data center load transfer cannot be simply done by directly migrating offline loads. Instead, it requires comprehensive consideration of multiple factors and coordinated efforts. Therefore, in this embodiment, cross-network load transfer is used to achieve load balancing optimization. During the load transfer process, based on the load migration characteristics, the offline load is divided into several parts: offline load transferred from other time periods to be processed in this microgrid data center, offline load transferred to other time periods, offline load processed in other microgrids in this time period, offline load transferred to other microgrids in this time period, and offline load transferred to other microgrids in other time periods in this time period. When situations such as excessively high real-time electricity prices in the microgrid, data center server running tasks exceeding limits, or insufficient renewable energy output occur, the load is transferred according to its importance. Based on the transfer of different types of offline loads, the offline load that the microgrid data center needs to process consists of two parts: offline load transferred from other time periods to be processed in this microgrid data center and offline load processed in other microgrids in this time period. The transferred offline load consists of three parts: offline load transferred to other time periods, offline load transferred to other microgrids in this time period, and offline load transferred to other microgrids in other time periods in this time period. The actual offline load of the data center is the difference between the processed offline load and the transferred offline load. Therefore, this embodiment uses 0-1 variables to control the transfer conditions and describes the spatial transfer mechanism of offline load between multiple microgrids through an offline load model. The model can be expressed as: In the formula, They represent the i-th micronet at time t, respectively. The data center's offline load, the offline load transferred to other time periods, the offline load transferred to other time periods, the offline load processed by other micronets in this time period, the offline load transferred to other micronets in this time period, and the offline load transferred to other micronets in this time period for processing at other times. , , and All represent 0-1 variables, used to control various offline loads. and These represent the i-th microgrid in time period t. The purchase price of electricity and the upper limit of electricity price. , These represent the i-th microgrid in time period t. Internal renewable energy power and renewable energy power cap. and Represents the i-th microgrid in time period t. Internal data center server operating load and capacity limits. This represents the transmission compensation coefficient.

[0029] This embodiment of offline load computing overcomes the limitations of traditional load shifting that relies solely on the time dimension. It introduces cross-data center load transfer in the spatial dimension. Considering the differences in electricity prices, renewable energy output, and server capacity between microgrids, it fully utilizes the ability of data centers to achieve load transfer in the spatial dimension through information networks. By optimizing the spatial distribution of load, it reduces energy consumption, thereby enabling more efficient use of green power resources, differentiated electricity prices, and available computing power from different microgrids. This reduces overall operating costs, promotes renewable energy consumption, and enhances the flexibility and stability of the system.

[0030] For online load analysis in data centers, GPUs are gradually replacing traditional CPU servers as the mainstream computing platform, demonstrating superior efficiency in large model training and parallel computing. However, current research on online load in data centers remains focused on CPU servers. Research on the energy consumption characteristics of GPU servers is relatively lacking. Therefore, this embodiment fully considers the operating characteristics of GPU and CPU servers in data centers when performing online load analysis. By combining the static idle power consumption and dynamic operating power consumption of GPUs and CPUs, the energy consumption characteristics of GPUs are supplemented to improve the modeling accuracy of online load models. Static idle power consumption refers to the basic power consumption of the server during idle operation, while dynamic operating power consumption refers to the impact of CPU and GPU operating voltage and frequency on power consumption. This embodiment uses dynamic voltage frequency scaling (DVFS) technology to represent the power consumption through the dynamic operating voltage, frequency, service rate, and reach rate. Therefore, the online load of a data center can be represented as: (4) In the formula, They represent the i-th micronet at time t, respectively. Online load, static load of GPU servers, dynamic load of GPU servers, static load of CPU servers, and dynamic load of CPU servers within the internal data center. The representatives represent the i-th microgrid. The internal data center displays the GPU voltage, CPU voltage, GPU frequency, and CPU frequency under operating mode s. C1 and C2 represent the operating state coefficients of the GPU server and CPU server, respectively. The operating state coefficient is a constant fitted to the computational efficiency under the server's operating state. This represents the total number of working modes. Different server frequency levels can be used to represent different working modes. For the i-th micronet at time t The dynamic load factor of a server in an internal data center under operating mode s is the ratio of load arrival rate to service rate.

[0031] The equipment load of a data center is determined by the various types of equipment it contains, primarily including cooling equipment and auxiliary equipment. Cooling equipment refers to the equipment used in the cooling process of a water-cooled data center, including cooling water pumps, chillers, cooling towers, precision air conditioners, etc. The sum of the loads of these devices constitutes the cooling equipment load, expressed as: (5) in, , , , , They represent the i-th micronet at time t, respectively. The total cooling load, cooling tower load, cooling water pump load, chiller unit load, and precision air conditioning load of the internal data center.

[0032] In addition, data centers also include auxiliary equipment, which also requires load calculation. The power consumption of data center auxiliary equipment is mainly determined by the power consumption of the servers and the power consumption of the communication load, which can be expressed as: (6) in, , , , , They represent the i-th micronet at time t, respectively. Internal auxiliary equipment load, server power consumption factor, server power distribution equipment load, communication transmission load, and communication power consumption factor.

[0033] Adding up all the above load types together gives the total energy consumption of the data center: (7) In the formula, Represents the i-th micronet at time t Total energy consumption of the data center within the facility.

[0034] This embodiment introduces spatial-dimensional cross-data center load transfer and optimizes the spatial distribution of offline load through coupled analysis of load transfer amount and control variables. By integrating the static idle power consumption and dynamic operating power consumption of servers, it supplements the power consumption characteristics of GPU servers, adapting to the new trend of online load in data centers and improving the accuracy of online load modeling. Based on offline and online loads, and combined with the equipment load of cooling equipment and auxiliary equipment, it can achieve accurate modeling of the total data center load, thereby improving the accuracy of total energy consumption index calculation.

[0035] In addition to the various loads of the data center, it is also necessary to analyze the water consumption during the data center's cooling process to determine the total water consumption index. The specific analysis steps include: Based on the water circulation process and cooling capacity of the data center's cooling system, calculate the water replenishment and daily operating water consumption of the cooling system. Calculate the water consumption of the fresh air system based on the humidity control of the data center's fresh air system. Calculate the power generation cooling water consumption based on the difference between the total energy consumption of the data center and the energy demand of renewable energy sources; The total water consumption of the data center is obtained based on the water replenishment of the cooling system, the water consumption of the daily operation, the water consumption of the fresh air system, and the water consumption for power generation cooling.

[0036] In this embodiment, the main equipment used in the cooling system of the data center during the cooling process includes a cooling tower and a chiller unit for cooling and cooling, as well as a fresh air system for humidity control.

[0037] For cooling towers, based on the water replenishment process of an open cooling tower, the amount of water replenished can be expressed by the amount of cooling water participating in the water circulation and the inlet and outlet water temperatures of the cooling tower: (8) In the formula, b Represents the water loss coefficient of the cooling tower. Represents the i-th micronet at time t The amount of water to make up the internal cooling tower. and Representing the i-th micronet The inlet and outlet water temperatures of the m-th cooling tower. Represents the i-th micronet The amount of cooling water involved in the water circulation of the m-th cooling tower, where M represents the amount of cooling water in the i-th microgrid. Total number of internal cooling towers.

[0038] The water consumption of a fresh air system can be expressed by the fresh air volume and humidity control efficiency, based on the system's control of relative humidity in the data center server room. (9) In the formula, , They represent the i-th micronet at time t, respectively. The water consumption and fresh air volume for humidity control in the internal server room. and Representing the i-th micronet Setting humidity levels in internal data center server rooms and The actual moisture content at any given time. Represents the i-th micronet The humidity control efficiency of the indoor fresh air system.

[0039] This embodiment uses the deviation between the target value and the actual value of moisture content to drive humidity control decisions, overcoming the shortcomings of traditional modeling methods that ignore the influence of humidity, and improving the modeling accuracy of water models in data centers.

[0040] The water consumption for daily operation of a data center can be expressed by the cooling capacity of the chiller unit: (10) In the formula, Represents the i-th micronet at time t Water consumption during daily operation of the internal data center The proportionality coefficient representing the rate of water evaporation and the water temperature. Represents the i-th micronet at time t Cooling capacity of the internal cooling water unit.

[0041] In addition to considering the water consumption for cooling towers, daily operation, and fresh air systems, this embodiment also takes into account the water demand for power generation and cooling from renewable energy sources such as photovoltaics and wind turbines. Therefore, the total water consumption of the data center can be expressed as: (11) In the formula, Represents the i-th micronet at time t The total water consumption of the data center The power generation cooling water coefficient represents the amount of water used for renewable energy generation. Represents the i-th micronet Energy efficiency and They represent the i-th micronet at time t, respectively. Power of internal photovoltaic and wind turbines. represent The larger value compared to 0. As you can see, The value represents the difference between renewable energy and total energy consumption demand, expressed in terms of power difference. If renewable energy cannot meet the total energy consumption demand, electricity needs to be purchased. The power generation cooling water coefficient can be used to quantify the water consumption of the power plant's cooling process caused by electricity purchase.

[0042] As can be seen, the total water consumption index mentioned above actually maps the data center cooling process to water consumption through water use efficiency. However, in actual application scenarios, temperature and humidity also have a significant impact on server operation, and temperature and humidity are closely coupled and mutually influential physical quantities. Although this embodiment considers the impact of humidity on server water consumption in the calculation of the total water consumption, outdoor temperature, server operating heat, and data center humidity will also affect server temperature cooling. Therefore, this embodiment analyzes the impact of environmental temperature and humidity data on the heat conduction process of the data center to construct a temperature and humidity index. The specific steps include: Based on the impact of server operating heat, external ambient temperature, and computer room humidity on server temperature cooling, server temperature indicators and ambient relative humidity indicators were obtained. Based on the influence of server temperature, external ambient temperature, and cooling capacity of the cooling system on the cooling water temperature, the cooling water temperature index of the cooling system is obtained.

[0043] In this embodiment, the effects of outdoor air temperature, server operating heat, and data center humidity on server temperature cooling are comprehensively considered. Based on the thermodynamic laws of air regarding temperature and humidity, physical constraints on server temperature and the relative humidity of the server operating environment are established. Specifically, server temperature can be represented by the server temperature, the outdoor ambient temperature, and the heat generated by server operation. The relative humidity of the server operating environment is quantified according to thermodynamic laws to determine the relationship between data center humidity and server temperature and relative humidity. Therefore, server temperature and the relative humidity of the server operating environment can be expressed as: In the formula, , , Represents the i-th micronet at time t Internal server temperature, server lower temperature limit, server upper temperature limit. , Representing the i-th micronet The heat loss coefficient and air thermal conductivity of the internal server. Represents the i-th micronet at time t The heat production and demand difference of renewable energy sources is the difference between the heat generated by renewable energy (such as photovoltaic and wind turbines) and the expected heat demand. Represents the i-th micronet at time t Internal server operating heat. and Representing the i-th micronet The thermal conductivity of dry air and water vapor inside the container. , These represent the binding coefficients of dry air and water vapor, respectively. This represents the ratio of the molecular weight of dry air to that of water vapor. , , These represent atmospheric pressure, humid air density, and saturated water vapor pressure, respectively. , , They represent the i-th micronet at time t, respectively. The relative humidity, lower humidity limit, and upper humidity limit of the internal server environment. This represents the ratio of server operating heat to load. Represents the i-th micronet at time t Total energy consumption of the data center within the facility This represents the ambient temperature at time t.

[0044] Furthermore, this embodiment also considers the impact of outdoor temperature in the data center on server cooling, describing the temperature change of the chiller's cooling water using the external ambient temperature and server temperature. The chiller's return water temperature can be expressed as: (13) In the formula, , , and They represent the i-th micronet at time t, respectively. The return water temperature, upper limit of return water temperature, lower limit of return water temperature, and cooling capacity of the internal chiller unit. and These are the i-th microgrids. Equivalent thermal resistance and heat capacity of internal cooling units. and Represent the ambient temperature at time t and the i-th microgrid, respectively. Internal server temperature. For the i-th micronet The heat loss coefficient of the internal cooling space.

[0045] Through the above embodiments, energy flow and energy consumption in data centers are modeled and analyzed from the perspective of hydroelectric coupling. The impact of temperature and humidity on the data center cooling process is also comprehensively considered, adding physical constraints related to temperature and humidity. Based on the above modeling and the various indicators in the above formula, this embodiment establishes a data center planning model with the goal of minimizing the total planning cost of multi-micronet and multi-data center interconnection. The total planning cost includes construction costs and operating costs. (14) in, Total cost of planning data centers within multiple micronets. For construction costs, Operating costs.

[0046] Construction costs include the cost of expanding the multi-microgrid power grid, the cost of expanding the multi-microgrid water network, and the construction cost of the multi-microgrid data center. Operating costs include electricity purchase costs, water purchase costs, carbon emission costs, curtailment costs, and equipment maintenance costs. Electricity purchase costs and carbon emission costs can be represented by the total energy consumption index of the data center, water purchase costs by the total water consumption index, curtailment costs by the amount of curtailed electricity, and equipment maintenance costs refer to the operating and maintenance costs of various equipment in the data center. These costs can be expressed as follows: (15) In the formula, To increase the construction cost of multi-microgrid power grids, To increase the construction cost of multi-micro-network water supply, For data center construction costs, For electricity purchase costs, For water purchase costs, For carbon emission costs, For the cost of curtailing electricity, For equipment operation and maintenance costs. This represents a collection of micronets that exist within a data center. , , , and They represent the i-th micronet at time t, respectively. The electricity price, water price, carbon emission cost coefficient, curtailment cost coefficient, and curtailment volume for externally purchased electricity, with T representing the total number of moments.

[0047] Since the data center planning model is based on the various indicators of the data center mentioned above, the planning model also needs to meet the various physical constraints represented by these indicators. Therefore, the data center planning model can be expressed in the following compact form: (16) Equation (16) above indicates that the data center planning model takes the minimization of formula (14) as the objective function and the above formulas (1) to (13) and formula (15) as constraints.

[0048] After constructing the above data center planning model, considering the data center load, renewable energy power generation, and the relative humidity of the external environment and the server operating environment, the model can be solved using conventional model solving algorithms to obtain the parameter combination with the highest system economy. The model solving algorithm can be the legacy algorithm, particle swarm optimization algorithm, linear programming algorithm, multi-objective optimization algorithm, or distributed robust optimization algorithm, etc. There is no limitation on the solving algorithm here.

[0049] Taking the use of a distributed robust optimization algorithm for model solving as an example, this algorithm primarily targets uncertain problems. During the optimization process, uncertainties are taken into account to find a more robust solution. The core idea is to divide the decision-making problem into two stages. The first stage is to make an initial decision, and the second stage is to formulate a better decision-making strategy based on the results of the first stage to cope with the impact of data uncertainty. This method can reduce conservatism and improve robustness. Due to the randomness of data user service demand, solar radiation intensity, and wind speed, data center server load and photovoltaic / wind turbine units are two of the most common uncertainties. In addition, external temperature also presents uncertainty. When solving the model, the distributed robust optimization algorithm can be used to solve the data center planning model under these three uncertainties. For example, in the solution process, the historical data of the three uncertainties are first processed and merged to transform the original three uncertainties into a one-dimensional uncertainty problem, and an ellipsoidal fuzzy set for constructing the uncertainties of the data center is constructed. Then, the planning model is transformed into the form of DRO-CVaR (Distributed Bar Optimization-Conditional Risk Value), where the first stage is the minimum planning cost not exceeding the threshold, and the second stage is the expected value of the objective function when the objective function exceeds the threshold in extreme scenarios. Finally, the model is linearized using linearization methods such as duality theory, and then optimized and solved using a conventional solver (such as the Gurobi solver) to obtain the final planning scheme.

[0050] As can be seen, when using conventional distributed robust optimization algorithms to solve the model, the uncertainty research focuses on the uncertainty of renewable energy output, data center load demand, and external temperature, without considering the uncertainty of parameters such as electricity purchase price, ambient humidity, and energy consumption. In multi-microgrid scenarios containing multiple data centers, each microgrid can buy and sell electricity to other microgrids. The bidirectional energy flow and frequent interaction characteristics of multiple microgrids, along with the resource endowments within each microgrid under different spatiotemporal conditions, exacerbate the uncertainty of offline load transfer in microgrids containing data centers due to the uncertainty of electricity purchase price, posing challenges to robust data center planning. Furthermore, current clustering algorithms for generating typical scenarios (such as extreme scenarios) are mostly conventional algorithms such as K-means and K-medoids. The random selection of initial cluster centers leads to a dense distribution of cluster centers, making it difficult to effectively capture the true distribution characteristics of the data. The clustering results can fluctuate significantly due to random selection, easily causing the algorithm to get trapped in local optima. Furthermore, since clustering algorithms use Euclidean distance as a similarity measure, this method is insensitive to the correlation structure of the data and struggles to identify non-convex data structures with complex correlations, resulting in poor performance on complex datasets. In addition, current typical scenario generation methods only consider static feature associations, neglecting the dynamic temporal evolution of random variables such as data center power output and load demand. This lack of consideration for the dynamic nature of the data generation process may lead to constructed scenarios that fail to accurately reflect the operational risks of the actual system, thus resulting in insufficient robustness of subsequent planning schemes.

[0051] To improve the robustness of the planning scheme, in a preferred embodiment, the present invention uses time-series features to classify typical uncertainty scenarios and constructs an uncertainty fuzzy set. Based on the typical uncertainty scenarios and the uncertainty fuzzy set, a distributed robust optimization algorithm is used to solve the model. The specific steps include: Historical data of preset uncertainties are processed based on the first and second derivatives in the time dimension to obtain an uncertainty data sequence. Then, the K-means clustering algorithm based on time distance is used to perform cluster analysis on the uncertainty data sequence to obtain typical uncertainty scenarios. Based on the uncertainty data sequence, an uncertainty fuzzy set of the data center is constructed. The objective function is transformed into a two-stage distributed robust optimization model, wherein the first stage model is based on minimizing the construction cost of typical uncertain scenarios, and the second stage model is based on minimizing the operating cost of uncertain fuzzy sets. The distributed robust optimization model is linearized using a linearization method, and the linearized model is optimized and solved using a solver to obtain a planning scheme for a multi-micronet data center.

[0052] In this embodiment, the total energy consumption of the data center within the micronet is taken into account. Photovoltaic power output Fan output Ambient temperature relative humidity Electricity price Total water consumption To address the uncertainty, the aforementioned parameters are treated as uncertainties. When processing the normalized historical data of these uncertainties, the data is divided by time scale. The first and second derivatives of each type of data in the time dimension are calculated and used as time-series features to obtain an uncertain data sequence. Then, the elbow method is used to determine the number of clusters. A set of time sample sequences with a preset time scale is randomly selected as the initial cluster centers. Dynamic time warping is used to calculate the similarity between samples, and the time sample sequence with the smallest similarity is selected as the next cluster center until the number of clusters is met.

[0053] Then, the temporal distance between each time sample sequence and the cluster center is calculated. Based on the temporal distance, each time sample sequence is assigned to a different cluster center, and the time sample sequence with the smallest total temporal distance is selected as the new cluster center. The above steps are repeated until the cluster centers no longer change, thus obtaining typical scenarios of uncertainty in the number of clusters, and the probability of each typical scenario in the total dataset is calculated.

[0054] This embodiment constructs a time-series dataset with various uncertainties in a data center, uses a dynamic time warping method to select initial cluster centers, and improves the K-means clustering algorithm by replacing the original Euclidean distance with time-series distance, which can effectively improve the fidelity of dynamic processes in typical data center scenarios.

[0055] Currently, descriptions of uncertain vertex scenarios mostly rely on static fuzzy sets and the introduction of interval uncertainties. This approach considers the uncertainties of all time periods within the planning scope as isolated and static features, ignoring the coupled temporal characteristics brought about by renewable energy output, data center load, electricity price, temperature and humidity, and water consumption at different time periods within the multi-microgrid scenario. This approach cannot effectively characterize the complex uncertainties of the dynamic and spatiotemporal relationships of multi-microgrid data centers, resulting in the disadvantages of low uncertainty model representation and poor adaptability of subsequent planning results.

[0056] Based on the above determinations, this embodiment, based on the aforementioned uncertain data sequence, constructs a first-order feature using the mean of the uncertainties and a second-order feature using the variance of the uncertainties according to the empirical distribution of the previous moment, thereby constructing a dynamic uncertain fuzzy set for the current moment. This leverages the advantage of dynamically and adaptively updating the fuzzy set, where the uncertain fuzzy set can be represented as: (17) In the formula, for A fuzzy set with uncertainty at time. The data sequence of uncertainties at time t-1 Let be the data sequence of uncertain quantities at time t. Known Down The conditional probability distribution, For an uncertain data sequence at time t numerical closed convex sets, for The true mean Known Down The first-order characteristic moments, Let be the error variable of the first-order characteristic moment. for The estimated mean, for The second-order characteristic moments.

[0057] This embodiment uses dynamic fuzzy sets to characterize the uncertainties of multi-microgrid data center load, electricity purchase price, renewable energy output, and temperature and humidity peak scenarios. This can eliminate overreaction to extremely low probability events of uncertainty, improve the precision and accuracy of the uncertainty characterization of data centers at different times, and make data center planning more in line with actual needs, thereby effectively avoiding over-allocation of resources.

[0058] Based on the aforementioned typical scenarios and fuzzy sets, a distributed robust optimization algorithm is used to solve the planning model, transforming the objective function into a two-stage DRO (Distributed Robust Optimization) model. The first stage model minimizes the data center construction cost under the probabilities of each typical scenario, while the second stage model minimizes the operating cost of each microgrid under extreme scenarios (i.e., the worst-case uncertainty) under the constraints of uncertain fuzzy sets. Then, the DRO model is linearized using duality theory and Big M linearization methods. Finally, the model is optimized and solved using Gurobi in Python, resulting in a data center planning scheme with high robustness.

[0059] This embodiment provides a planning method for multi-micronet data centers. Based on the operational characteristics of various loads in the multi-micronet data center, this embodiment performs energy consumption modeling, achieving an accurate description of the energy consumption process at each stage of the multi-micronet data center. It also performs water consumption modeling based on the impact of temperature and humidity on heat conduction in the data center, achieving an accurate description of the water consumption process. Furthermore, considering multiple uncertainties, it employs a clustering algorithm based on improved time-series features to generate typical uncertainties, and utilizes dynamic uncertainties fuzzy sets to represent the vertex scenarios of uncertainty. Typical scenarios are used to optimize the daily operating costs of the data center, while vertex scenarios are used to minimize losses in the worst-case scenario of the distributed robust optimization problem, thereby improving the reliability of the planning scheme.

[0060] Please see Figure 2 Based on the same inventive concept, the second embodiment of this invention proposes a planning system for a multi-micronet data center, comprising: Data analysis module 10 is used to analyze the energy flow of the microgrid based on the power operation characteristics and water network topology among multiple microgrids, and obtain energy flow indicators. Based on the load operation characteristics and cooling water characteristics of multiple data centers, the energy consumption of data centers is analyzed to obtain energy consumption indicators. The heat transfer process of the data center is analyzed based on the ambient temperature and humidity to obtain temperature and humidity indices; The planning model construction module 20 is used to establish a data center planning model with the goal of minimizing the total planning cost and with the energy flow index, the energy consumption index and the temperature and humidity index as constraints. The total planning cost includes construction cost and operating cost. The model solving module 30 is used to solve the data center planning model using a preset algorithm to obtain the planning scheme of the multi-micronet data center.

[0061] The technical features and effects of the multi-micronet data center planning system proposed in this invention are the same as those of the method proposed in this invention, and will not be repeated here. Each module in the above-mentioned multi-micronet data center planning system can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or it can be stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0062] Furthermore, embodiments of the present invention also propose a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0063] Please see Figure 3 The diagram illustrates the internal structure of a computer device in one embodiment. This computer device can specifically be a terminal or a server. The computer device includes a processor, memory, network interface, display, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a multi-microgrid data center planning method. The display screen of the computer device can be a liquid crystal display (LCD) or an e-ink display. The input devices of the computer device can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse, etc.

[0064] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computing devices may include more or fewer components than those shown in the figure, or combine certain components, or have the same component arrangement.

[0065] Furthermore, embodiments of the present invention also propose a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0066] In summary, the present invention proposes a planning method, system, equipment, and medium for a multi-microgrid data center. The method analyzes the energy flow of the microgrids based on their power operation characteristics and water network topology, obtaining energy flow indicators; analyzes the energy consumption of the data centers based on their load operation characteristics and cooling water characteristics, obtaining energy consumption indicators; analyzes the heat transfer process of the data centers based on ambient temperature and humidity, obtaining temperature and humidity indicators; establishes a data center planning model with the goal of minimizing the total planning cost, and using the energy flow indicators, energy consumption indicators, and temperature and humidity indicators as constraints. The total planning cost includes construction costs and operating costs. A preset algorithm is used to solve the data center planning model to obtain a planning scheme for the multi-microgrid data center. This invention can minimize the impact of uncertainties on data center operation and enhance the reliability and adaptability of the multi-microgrid data center planning results.

[0067] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on its differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0068] The embodiments described above are merely preferred embodiments of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the technical principles of this invention, and these improvements and substitutions should also be considered within the scope of protection of this application. Therefore, the scope of protection of this patent application should be determined by the scope of the claims.

Claims

1. A planning method for a multi-micronet data center, characterized in that, The method is applied to multiple microgrid interconnection scenarios, and each microgrid contains at least one data center, including: Based on the power operation characteristics and water network topology among multiple microgrids, the energy flow of the microgrids is analyzed to obtain energy flow indicators; Based on the load operation characteristics and cooling water characteristics of multiple data centers, the energy consumption of data centers is analyzed to obtain energy consumption indicators. The heat transfer process of the data center is analyzed based on the ambient temperature and humidity to obtain temperature and humidity indices; A data center planning model is established with the goal of minimizing the total planning cost and the constraints of the energy flow index, the energy consumption index, and the temperature and humidity index. The total planning cost includes construction cost and operating cost. The data center planning model is solved using a preset algorithm to obtain a planning scheme for a multi-micronet data center.

2. The planning method for multi-micronet data centers according to claim 1, characterized in that, The steps for analyzing the energy flow of a microgrid based on the power operation characteristics and water network topology among multiple microgrids to obtain energy flow indices include: Based on the voltage amplitude and active power of each microgrid, the power grid energy flow index is obtained; Based on the water network topology and water load flow balance among multiple microgrids, the water network energy flow index is obtained.

3. The planning method for a multi-micronet data center according to claim 1, characterized in that, The step of analyzing the energy consumption of data centers based on the load operation characteristics and cooling water characteristics of multiple data centers to obtain energy consumption indicators includes: Based on the operating characteristics of various loads in the data center, the total energy consumption index of the data center is obtained. The total water consumption of a data center is calculated based on the server cooling water circulation and the power generation cooling water used for electricity purchases.

4. The planning method for a multi-micronet data center according to claim 3, characterized in that, The steps for obtaining the total energy consumption index of the data center based on the operating characteristics of various loads in the data center include: Calculate offline load based on cross-network load transfer between various data centers; Calculate the online load based on the operating power consumption of servers in each data center; Calculate the equipment load based on the operating power consumption of the cooling equipment and auxiliary equipment in each data center; The total energy consumption index of the data center is obtained based on the offline load, the online load, and the device load.

5. The planning method for a multi-micronet data center according to claim 3, characterized in that, The step of obtaining the total water consumption index of the data center based on the server cooling water circulation and the power generation cooling water caused by electricity purchase includes: Based on the water circulation process and cooling capacity of the data center's cooling system, calculate the water replenishment and daily operating water consumption of the cooling system. Calculate the water consumption of the fresh air system based on the humidity control of the data center's fresh air system. Calculate the power generation cooling water consumption based on the difference between the total energy consumption of the data center and the energy demand of renewable energy sources; The total water consumption of the data center is obtained based on the water replenishment of the cooling system, the water consumption of the daily operation, the water consumption of the fresh air system, and the water consumption for power generation cooling.

6. The planning method for a multi-micronet data center according to claim 1, characterized in that, The steps for analyzing the heat transfer process of the data center based on ambient temperature and humidity to obtain temperature and humidity indices include: Based on the impact of server operating heat, external ambient temperature, and computer room humidity on server temperature cooling, server temperature indicators and ambient relative humidity indicators were obtained. Based on the influence of server temperature, external ambient temperature, and cooling capacity of the cooling system on the cooling water temperature, the cooling water temperature index of the cooling system is obtained.

7. The planning method for a multi-micronet data center according to claim 1, characterized in that, The steps of solving the data center planning model using a preset algorithm to obtain the planning scheme for the multi-micronet data center include: Historical data of preset uncertainties are processed based on the first and second derivatives in the time dimension to obtain an uncertainty data sequence. Then, the K-means clustering algorithm based on time distance is used to perform cluster analysis on the uncertainty data sequence to obtain typical uncertainty scenarios. Based on the uncertainty data sequence, an uncertainty fuzzy set of the data center is constructed. The objective function is transformed into a two-stage distributed robust optimization model, wherein the first stage model is based on minimizing the construction cost of typical uncertain scenarios, and the second stage model is based on minimizing the operating cost of uncertain fuzzy sets. The distributed robust optimization model is linearized using a linearization method, and the linearized model is optimized and solved using a solver to obtain a planning scheme for a multi-micronet data center.

8. A planning system for a multi-micronet data center, characterized in that, include: The data analysis module is used to analyze the energy flow of the microgrid based on the power operation characteristics and water network topology among multiple microgrids, and to obtain energy flow indicators. Based on the load operation characteristics and cooling water characteristics of multiple data centers, the energy consumption of data centers is analyzed to obtain energy consumption indicators. The heat transfer process of the data center is analyzed based on the ambient temperature and humidity to obtain temperature and humidity indices; The planning model construction module is used to establish a data center planning model with the goal of minimizing the total planning cost and with the energy flow index, the energy consumption index, and the temperature and humidity index as constraints. The total planning cost includes construction cost and operating cost. The model solving module is used to solve the data center planning model using a preset algorithm to obtain the planning scheme of the multi-micronet data center.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images