Electricity calculation coupling system capacity configuration method considering load time shifting characteristic

Abstract

The invention provides a computing coupling system capacity configuration method considering load time shifting characteristics, and belongs to the technical field of integrated energy system planning and operation optimization. By constructing a wind and light output typical scene clustering and double-layer optimization configuration model, equipment locating and sizing and load time shifting scheduling are collaboratively optimized; the new energy utilization rate is effectively improved, the comprehensive cost and network loss of the system are reduced, the node voltage level is improved, and a systematic solution is provided for high-quality power supply of the data center under high-proportion new energy access.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of comprehensive energy system planning and operation optimization, and particularly relates to a wind-solar-storage system capacity collaborative optimization configuration method considering load time shift characteristics. BACKGROUND

[0002] The penetration rate of renewable energy such as wind power and photovoltaic power in power systems is continuously increasing, however, the volatility, intermittency and randomness of the output of the renewable energy also bring significant challenges to the stable operation of power grids and the efficient consumption of energy. Especially with the wide construction of distributed data centers, their high energy consumption and wide distribution characteristics further aggravate the load pressure of local power grids, leading to increased active power loss, decreased voltage quality, and limited local consumption capacity of new energy.

[0003] At present, for the optimal configuration of energy systems containing data centers, existing researches mostly focus on the operation and dispatch level, or do not fully consider the uncertainty and correlation of wind and light output in planning, and mostly use deterministic models or single scenario analysis, which is difficult to truly reflect the time sequence fluctuation and spatial correlation characteristics of wind and light resources in actual systems. In addition, although some researches have noticed the time shift scheduling potential of data center load, they often do not deeply coordinate the wind and light storage configuration in capacity planning, and do not build a clear cross-time scale coupling mechanism, resulting in that the system still relies on traditional power grid power supply in peak period, the capacity and location of energy storage configuration lack economic and reliability coordination, and the overall new energy utilization rate still has room for improvement. On the other hand, existing planning models usually ignore the influence of distribution network voltage constraints and network loss, or only take economy as a single target, and cannot guarantee power supply quality and operation economy at the same time.

[0004] The reasons for the above problems mainly include: first, the modeling of wind and light output and load multi-element uncertainty is complex and difficult to efficiently integrate into the planning model; second, there is a lack of system-level collaborative optimization framework between adjustable data center load and wind and light storage capacity configuration; third, traditional single-layer optimization model is difficult to coordinate long-term planning and short-term operation targets at the same time, resulting in insufficient adaptability and robustness of planning results in actual operation. Therefore, a wind-solar-storage collaborative planning method considering multiple uncertainties, source-load interaction characteristics and distribution network safety constraints needs to be constructed to realize economic and efficient, safe and reliable operation of the system. SUMMARY

[0005] This invention aims to address the problems of poor system economy, low renewable energy absorption rate, and poor voltage quality in existing integrated energy system planning that includes data centers. These problems stem from insufficient consideration of the uncertainties in wind and solar power output, the time-shifting characteristics of data center loads, and the constraints of power distribution network safety. Specifically, existing configuration methods typically do not simultaneously address the randomness and correlation of wind and solar resources, nor do they coordinate the optimization of data center schedulable computing power load with wind, solar, and energy storage capacity. This disconnect between planning and operational objectives prevents improvements in overall system economy and power quality.

[0006] To address this, the present invention provides a capacity configuration method for a computer-coupled system that takes into account load time-shifting characteristics. By constructing a clustering and two-layer optimization configuration model for typical wind and solar power output scenarios, the method collaboratively optimizes equipment location and capacity determination and load time-shifting scheduling, effectively improving the utilization rate of new energy sources, reducing overall system costs and network losses, and improving node voltage levels, thus providing a systematic solution for high-quality power supply to data centers with a high proportion of new energy access.

[0007] The technical solution adopted in this invention is as follows:

[0008] A method for configuring the capacity of a computer-coupled system that takes into account load time-shift characteristics, comprising the following steps:

[0009] The first step is to construct a coupled system architecture of wind turbine-solar photovoltaic-data center-energy storage.

[0010] The architecture topology and control structure of the wind-solar-computing-storage coupled system are as follows: Figure 1 As shown, the Figure 1 The system showcases the overall architecture and control structure of a coupled "wind turbine-photovoltaic-data center-energy storage" system. This system integrates wind turbines and photovoltaics as renewable energy generation units, with the data center as the core load. It includes latency-sensitive loads requiring real-time processing and latency-tolerant loads that can be flexibly scheduled, and is equipped with an energy storage system for peak shaving and valley filling. The system is connected to the upper-level power grid through a grid connection point and employs a two-layer optimization control strategy: the upper layer is responsible for the site selection and capacity determination of wind turbines, photovoltaics, and energy storage, aiming to minimize the overall cost throughout the entire lifecycle; the lower layer is responsible for real-time operation, optimizing scheduling to achieve optimal voltage quality and minimize network losses through methods such as adjusting energy storage charging and discharging and data center load shifting. As a new type of node integrating computing and energy consumption, the data center's high-reliability computing power requirements impose stringent requirements on power quality. Furthermore, its wide geographical distribution leads to long power supply distances and significant network losses. A two-tier optimization strategy that balances power supply and integration is proposed. During the planning stage, the volatility of renewable energy output and the flexible adjustability of computing load are fully considered. During peak renewable energy generation periods, computing load transfer is used to actively improve the absorption level, while during off-peak periods, the energy storage system is used for peak shaving. This achieves coordinated and optimized configuration of photovoltaic, wind turbine, energy storage and data center, disperses grid connection points and power flow distribution, reduces grid losses and improves overall energy efficiency.

[0011] Data center load type

[0012] The load of a data center (DC) includes delay-sensitive and delay-tolerant workloads. Delay-sensitive load needs to be processed immediately after the DC receives an instruction; while for delay-tolerant load, it can be completed before the specified time limit. By using the shiftable characteristics of delay-tolerant load, a scheduling mechanism is constructed for such tasks, so that delay-tolerant load can be flexibly allocated to subsequent periods within the maximum tolerance time range for processing, so as to improve the system's ability to consume new energy generation. In order to suppress invalid or excessive shifting behavior, an economic penalty mechanism is introduced to appropriately increase the electricity cost per unit time for delayed processing load.

[0013] Second step, data center energy consumption model construction

[0014] The main load of the data center includes the power consumption of the unit equipment, the power of the system refrigeration, the power of the network transmission and other system power,

[0015] The server unit load is the main part of the energy consumption of the data center, and the power of the IT equipment in the data center can be expressed as

[0016] (1)

[0017] (2)

[0018] In the formula: N dc is the amount of delay-tolerant load that needs to be calculated; μ is the service rate of the DC; T n is the maximum tolerance time of the nth task reported by the user; is the delay-sensitive load arriving at the DC, which needs to be processed immediately; is the batch processing work load arriving at the DC at time t; is the nth delay-tolerant load arriving at the DC at time t, which is allocated to the load component at time t+1, and the sum of the work load allocated to time t is the total server load of the DC at the current time t ,

[0019] The coupling relationship between the cooling power described by the thermodynamic process and the server computing power and server heat power of the data center is expressed as formulas (3) and (4)

[0020] (3)

[0021] (4)

[0022] In the formula: represents the heat power of the unit of the data center i at time t; represents the air conditioning refrigeration capacity of data center i at time t; represents the outdoor temperature of data center i at time t; represents the computer room temperature of data center i at time t; DC represents the thermal resistance of the data center computer room; DC represents the surface area of the data center computer room; represents the volume of the data center computer room; represents the heat generation efficiency of the data center unit; PUE represents the energy use efficiency coefficient,

[0023] The data center network includes transmission links and transmission equipment, and the dynamic range of the power of the two is small. The network transmission power of data center i is relatively stable, and the proportion of power of fresh air, fire fighting, lighting and other systems in the data center is small. Other energy consumption is usually considered as a fixed value of 2 MW.

[0024] Step 3: Wind-solar-storage model construction

[0025] The power generation output of a single photovoltaic panel at a certain moment can be expressed as

[0026] (5)

[0027] The output power of the wind power generation system is affected by the wind speed. The power generation output of the wind turbine at a certain moment can be expressed as

[0028] (6)

[0029] In the formula: V in , V out , V r are the cut-in, cut-out and rated wind speed of the wind turbine respectively; P wt (t) is the output power of the wind turbine at time t; P rwt is the rated power of a single wind turbine,

[0030] The energy storage realizes peak clipping and valley filling through the charging and discharging process, absorbs and stores the excess power of the wind-solar unit, and its operating state can be characterized by the state of charge at each moment, which can be expressed as

[0031] (7)

[0032] In the formula: is the self-discharge rate of the BESS; are the charging and discharging power of the BESS of node i in season s and time period t; are the state of charge of the BESS of node i in season s and time period t and t+1 respectively; η+ and η are the charging and discharging efficiencies respectively; Δt is the time interval, which is taken as 1 h.

[0033] The fourth step is to construct a dual-layer optimization configuration model for wind, solar, computing, and storage.

[0034] The two-layer optimization strategy model for integrated energy systems with DC power proposed in this invention is as follows: Figure 2 As shown. Figure 2 This paper demonstrates the two-layer optimization configuration model structure of the system. The upper planning layer aims to minimize the overall system cost, deciding on the installation location and capacity of wind turbines, photovoltaics, and energy storage, considering constraints such as equipment investment, operation and maintenance, and electricity purchase costs. The lower operation layer aims to ensure power quality and reduce network losses, optimizing the real-time operation of the system, including energy storage charging and discharging strategies, data center load scheduling, and grid power interaction. The two-layer model achieves collaborative optimization through parameter correlation, and finally, the original two-layer problem is transformed into a solvable single-layer optimization problem through a model transformation method.

[0035] Upper-level planning model construction

[0036] The upper-level planning layer aims to minimize overall cost, with decision variables being the location capacity of PV, WT, and BESS, and constraints being investment in each piece of equipment and site selection capacity constraints; the objective function F of the upper-level model is...

[0037] (8)

[0038] (1) Investment cost C I

[0039] (9)

[0040] In the formula: , where are the investment costs for PV, WT, and BESS, respectively; b is the discount rate. , These are the service lives of PV, WT, and BESS, respectively. The set consists of candidate installation locations for PV, WT, and BESS, respectively; These are the investment costs for a single PV, WT, and BESS unit, respectively. These represent the number of PV, WT, and BESS configurations on node i, respectively.

[0041] (2) Operation and maintenance cost C O

[0042] (10)

[0043] In the formula: These are the maintenance costs for PV, WT, and BESS, respectively. The cost of purchasing electricity from the higher-level power grid; Ds represents the number of days in the season (s); This is a collection consisting of the installation locations of PV, WT, and BESS, respectively. PV unit operation and maintenance cost; WT unit operation and maintenance cost; BESS unit charging and discharging power operation and maintenance cost; t period electricity purchase price; Active power injected by PV at node i in season s in period t; BESS charging and discharging power at node i in season s in period t, respectively; Δt is 1h; Set of terminal nodes of AC branch with generator node as head node; Active power of upper-level power grid flowing into the current-level power grid in season s in period t,

[0044] Upper-layer model constraint conditions include PV, WT and BESS location and quantity constraints, which are

[0045] (11)

[0046] (12)

[0047] In the formula: Maximum number of PVs allowed to be installed at node i; Maximum number of WTs allowed to be installed at node i; Maximum number of BESSs allowed to be installed at node i;

[0048] Lower-layer operation model construction

[0049] The lower layer is to ensure the power supply quality of the data center and solve the problem of large network loss, and the comprehensive minimum of voltage deviation and network loss is taken as the target, and the lower-layer model objective function f is

[0050] (13)

[0051] (1) ΔU all Voltage deviation

[0052] (14)

[0053] In the formula: Voltage of node i in season s in period t in per unit; Set of all nodes; Absolute value calculation formula

[0054] (2) P loss Total network loss

[0055] (15)

[0056] In the formula: Iij,s,t is the current value of branch ij at time t in season s; R ij is the resistance of branch ij ; is the set of all branches,

[0057] The constraints are system security constraints, each device output constraints and batch processing load delay constraints,

[0058] 1) Power balance constraint

[0059] At any time, the power generated is equal to the power consumed, that is, the power of the wind turbine, photovoltaic, energy storage discharge and the power purchased from the grid is always consistent with the power of the user, data center and energy storage charging power,

[0060] (16)

[0061] 2) Power flow constraint

[0062] (17)

[0063] In the formula: represents the active and reactive power injection at node i at time t; represents the active and reactive power consumption at node i; U i , U j represents the voltage of node i and node j, θ ij represents the voltage phase angle difference between node i and node j; G ij and B ij represent the conductance and susceptance between node i and node j,

[0064] 3) Voltage constraint

[0065] (18)

[0066] In the formula: are the lower and upper limits of the voltage of node i, respectively, taking the values 0.93pu and 1.07pu,

[0067] 4) Power constraint of the upper grid flowing into the current grid

[0068] (19)

[0069] In the formula: is the reactive power of the upper grid flowing into the current grid at time t in season s; , are the lower and upper limits of the active (reactive) power of the upper grid flowing into the current grid,

[0070] 5) PV, WT constraint

[0071] (20)

[0072] In the formula: The power factor angle,

[0073] 6) BESS constraint

[0074] (twenty one)

[0075] In the formula: These represent the charging and discharging states of BESS, respectively. The rated charge and discharge power of BESS for node i; The maximum BESS power capacity for node i; These are the initial BESS charges at midnight on the current day and midnight on the next day for node i in season s, respectively. These represent the lower limit, upper limit, and initial value of the BESS charge capacity, respectively.

[0076] 7) Maximum delay constraint for batch processing load

[0077] (twenty two)

[0078] In the formula: d is the time consumed by batch processing load; D is the maximum tolerance time for batch processing load.

[0079] Step 5: Solving the model

[0080] The original mixed-integer nonlinear bilayer model was transformed into a single-layer mixed-integer linear programming model using parameter correlation and second-order cone transformation, thus unifying the parameter correlation between the upper and lower layers of "wind, solar, storage, and charging". The final transformed model is a single-layer multi-objective mixed-integer optimization problem. The NSGA-II intelligent optimization algorithm was used to solve the multi-objective programming problem, and the result is a Pareto optimal solution set. The Topsis method was used for selection analysis to obtain the preferred configuration scheme.

[0081] The solution process is as follows: Figure 3 As shown, the Figure 3 The complete solution process of the optimization model is demonstrated. First, the system's basic data and wind and solar resource characteristics are input, and typical operating scenarios are obtained through scene generation and reduction techniques. Then, the original mixed-integer nonlinear two-layer model is converted into a single-layer mixed-integer linear programming model. Next, the NSGA-II multi-objective optimization algorithm is used to obtain the Pareto optimal solution set. Finally, the TOPSIS method is used to select the configuration scheme with the best overall performance from the solution set, and the final location and capacity determination results of each device and system operation indicators are output.

[0082] The beneficial effects that can be achieved by adopting the above-mentioned technical solution in this invention are:

[0083] The method is suitable for a distributed energy system containing a data center, coordinates planning of capacities and layouts of photovoltaic (PV), wind turbine (WT) and battery energy storage system (BESS), and realizes source-load coordination optimization by using time shift characteristics of data center (DC) computing power load.

[0084] The present application constructs a double-layer collaborative optimization configuration model considering wind and light uncertainty and load time shift characteristics, integrates new energy such as wind and light, energy storage and data center into the overall planning of the distribution network, reduces the operation cost of the data center while optimizing the decision target of the power grid side, improves the safety and economy of the power grid operation, and can effectively solve the capacity configuration problem of the wind and light calculation storage system containing DC. Significantly improve the new energy consumption rate and the economic operation level of the system, effectively reduce the network loss and improve the voltage quality, and guarantee the power supply demand of high reliability loads such as data centers. At the same time, the method coordinates the planning and operation targets, enhances the adaptability of the system to different scenes, improves the rationality and practicability of the overall planning, and provides a more optimal capacity configuration solution for the distribution network containing high proportion of new energy and data center. BRIEF DESCRIPTION OF DRAWINGS

[0085] Figure 1 The wind and light calculation storage coupled system architecture topology in the present application is shown in the figure;

[0086] Figure 2 The double-layer optimization configuration model of "wind and light calculation storage" in the present application is shown in the figure;

[0087] Figure 3 The flow chart in the present application is shown in the figure;

[0088] Figure 4 The wind and light output scene clustering diagram in the present application is shown in the figure;

[0089] Figure 5 The load distribution diagram of each period in the present application is shown in the figure;

[0090] Figure 6 The three-dimensional distribution diagram of Pareto frontier solution in the present application is shown in the figure;

[0091] Figure 7 The DC load scheduling result diagram in the present application is shown in the figure;

[0092] Figure 8 Figures showing the output of each device in the present application. DETAILED DESCRIPTION

[0093] Example basic data

[0094] To verify and analyze the correctness of the model established by the present method, historical data of wind speed and solar radiation intensity in a certain region throughout the year are selected, and the wind and light output is calculated; then it is normalized, and 1000 scenarios considering correlation and uncertainty are generated based on kernel density estimation and Copula theory. The whole year is divided into seasons: summer, transition season, winter. To reduce the complexity of calculation, K-means is used for clustering, and 6 typical scenarios are obtained, and the specific output is as shown in Figure 4 .

[0095] To verify the effectiveness of the model, first, the IEEE 33-node distribution network is taken as the research object, and the performance of the model itself is analyzed; then, compared with other models. The main parameters of the wind turbine and photovoltaic renewable energy power generation equipment in the integrated energy system containing DC and the capacity configuration upper and lower limits of the decision variables are as shown in Table 1. In addition, the related parameter settings of the delay-tolerant load scheduling of the data center are as follows: the maximum delay-tolerant time is 2 hours, i.e. the load at time t can be allocated to the period from t+1 to t+2 for processing. In order to avoid invalid translation, an economic penalty factor is set, and the power cost of the load increases by 0.1 yuan / kW·h for each 1-hour backward translation.

[0096]

[0097] Taking the typical scenario in summer as an example, the 2 types of loads of DC and the user load are as shown in Figure 5 . In this paper, NSGA-II algorithm is used for solving: the size of the population is set to 200, the maximum number of generations is 100, the crossover probability is 0.9, and the mutation probability is 0.1. The Pareto frontier obtained by solving is as shown in Figure 6 . Each point in the figure represents a configuration scheme. Using Topsis method, the weights of the three objective functions are all set to 0.33, and the final configuration result preferred by the individual is obtained.

[0098] The planning results of the optimization configuration model established by the present method are as shown in Table 2, and the costs of each item are as shown in Table 3,

[0099] Table 3 Costs of each item of the optimization configuration model established

[0100]

[0101] Combining IEEE 33-node topology and Table 4, from the planning point of view: PV, WT installed nodes are mainly distributed in nodes 3, 6 and other heavy load nodes or nodes 18, 33 and other end nodes, which can realize "self-generation and self-use, and local consumption of the remaining amount", lift the power supply voltage, maximize the role of clean energy such as wind and light, and ensure the power supply quality of data center and residential power supply. BESS is distributed in nodes 3, 6 where PV and WT coincide or at the end of node 18 load. The reason for the partial overlap of BESS installation location and PV, WT is that when the wind and light output is too large, BESS can consume part of the excess electricity to reduce the wind and light abandonment rate. The wind-light-storage joint action makes the equipment utilization rate reach the maximum. The reason for installing at the end of the load is that the voltage drop degree is large at the end of the load when PV output is 0 at night and the load peak period, and the distance from the power supply point is very far, and the active loss is high. Configuring BESS at this position can act as a power supply point at night to discharge and consume the load locally to lift the voltage and reduce the loss. At the same time, it realizes the "peak shaving" effect relative to the entire distribution network. The reason for the relatively small number of WT relative to PV is that the investment cost of a single WT is higher than that of PV. In order to reduce the investment cost, the installed capacity of PV is greater than that of WT. The reason for the small number of BESS installations is that even if PV output is 0 at night, with the flexible configuration strategy of DC load, BESS does not need to discharge too much to lift the voltage, so the number of BESS configurations is small.

[0102] The costs of the model built by the method are shown in Table 3, the time-of-use electricity price and the DC load dispatching result are shown in Figure 7 , and the output of each device is shown in Figure 8 . From the operation point of view, the comprehensive cost of the built configuration model is 685.53 million yuan, of which the total investment cost is 73.45 million yuan, the total operation cost is 612.08 million yuan, which is 2.29% lower than the original distribution network. From this dimension, the power supply is more economical. The annual voltage deviation is 11074.5pu, which is 2.79% lower than the original distribution network. The minimum voltage is increased from 0.9224pu to 0.9472pu. The active loss is 303.33MWh, which is 4.15% lower than the original distribution network. From this dimension, the power supply is more high-quality.

[0103] The total amount of wind and light power generation is 8.4454MW, and the consumption amount after dispatching is increased from 8.1160MW to 8.2173MW, and the consumption rate is increased from 96.1% to 97.3%. The dispatching optimization improves the utilization rate of power generation. At the same time, the comprehensive cost is reduced by 160.8 thousand yuan. Figure 7It can be seen that: in the peak of 5:00 and 6:00 and 20:00 and 21:00, the DC load shows a downward trend; while the price decreases at 22:00, the DC load shows an upward trend. This is because the batch processing load in these two time periods is shifted to the time period with low price, reducing the operation cost of DC. From 11:00, the wind and light output power generation is surplus, and the batch processing load at 9:00 and 10:00 is shifted to the time period with more new energy surplus power and low price in the afternoon. This makes the system fully utilize the surplus renewable energy generation, reduces the operation cost, and improves the renewable energy consumption rate. In other time periods, the load scheduling strategy of DC remains unchanged because the load shifting cannot significantly reduce the cost or improve the benefit.

[0104] Model comparison and analysis: From the perspective of the participating subject, the following comparison schemes are set to analyze the configuration results under different conditions and verify the advantages of the established model.

[0105] Scheme 1: The configuration model in this paper;

[0106] Scheme 2: without considering the DC flexibility configuration strategy;

[0107] Scheme 3: replace the lower layer objective function with only considering the lowest operation cost based on model 1.

[0108]

[0109] Table 5 Model configuration objective function results

[0110]

[0111] As shown in Table 5, comparing model 1 and 2, the comprehensive cost of model 2 increases by 63,600 yuan, the minimum voltage is 0.9303pu, and the minimum voltage of model 1 decreases by 0.0169pu. This is because when the flexibility configuration strategy of DC load is not considered, the load peak period at night can only rely on BESS discharge and power supply from the upper grid, and the long line length makes the voltage decrease to a large extent, and only these two subjects cannot meet the stable and reliable power supply to the data center and users; the voltage deviation and network loss increase by 443.6pu and 12.49MWh respectively, which verifies the advantages of considering the flexibility configuration strategy of DC load for improving the distribution of batch processing load period.

[0112] Comparing model 1 and model 3, model 3 only considers the lowest operating cost, which makes the comprehensive cost decrease by 706,000 yuan, and the voltage is 0.9317pu under the cooperation of wind, light and energy storage. The lowest voltage is only reduced by 0.0155pu compared with model 1. From the perspective of the lowest voltage, it indeed has advantages. However, if combined with the social responsibility of the power grid company, the annual voltage deviation and the line loss rate, it will be found that the annual voltage deviation of model 5 increases by 783.3pu, and the annual network total loss increases by 32.37MWh. It shows that the voltage is not good enough throughout the year, and the resistance to external disturbance is poor, which weakens the economy and makes it difficult to meet the high reliability requirements of data center power supply quality.

[0113] From the above comparison results, it can be seen that the "wind light algorithm storage" double-layer optimization configuration model considering the DC flexibility configuration strategy has better comprehensive performance, enhances voltage quality, reduces network loss, optimizes the decision target of the grid side, and reduces the operating cost of the data center, meeting the requirements of grid and data center cooperation.

Claims

1. A method for capacity configuration of an electrically coupled system considering load time shift characteristics, characterized in that, The steps are as follows: The first step is to construct the coupled system architecture of wind turbine-solar photovoltaic-data center-energy storage. The system integrates wind turbines and photovoltaics as renewable energy power generation units, with the data center as the core load. It includes latency-sensitive loads requiring real-time processing and latency-tolerant loads that can be flexibly scheduled. An energy storage system is also configured for peak shaving and valley filling. The system is connected to the upper-level power grid through a grid connection point and adopts a two-layer optimization control strategy: the upper layer is responsible for the site selection and capacity determination of wind turbines, photovoltaics, and energy storage, with the goal of minimizing the overall cost over the entire life cycle; the lower layer is responsible for real-time operation, and optimizes scheduling with the goal of optimal voltage quality and minimum network loss by adjusting the charging and discharging of energy storage and the shifting of data center loads. The second step is to build a data center energy consumption model. The main loads of a data center include the power consumption of server equipment, system cooling power, network transmission power, and other system power. Server load is a major component of data center energy consumption. The power of IT equipment in a data center can be expressed as... （1） （2） Where: N dc The latency-tolerant load to be calculated is given by μ; the DC service rate is given by T. n The maximum tolerance time for the nth task reported by the user; For delay-sensitive loads to reach the DC, the DC needs to process them immediately; The batch processing workload arriving at DC at time t; Let the nth delay-tolerant load arriving at DC at time t be allocated to the load component at time t+1. The sum of the workload allocated to time t is the total server load of the DC at time t. , The coupling relationship between cooling power and data center server computing power and server heat generation power, described by thermodynamic processes, is represented by equations (3) and (4). （3） （4） In the formula: represents the unit heat generation power of the data center i at time t; represents the air conditioning refrigerating capacity of the data center i at time t; represents the outdoor temperature of the data center i at time t; represents the room temperature of the data center i at time t;R DC represents the data center room thermal resistance;S DC represents the data center room surface area; represents the data center room volume; represents the data center unit heat generation efficiency;PUE represents the energy use efficiency coefficient, The data center network includes transmission links and transmission devices, both of which have a small dynamic range of power, and the network transmission power of the data center i For more stable, fresh air, fire, lighting and other system power in the data center accounts for a smaller proportion, other energy consumption Usually regarded as a fixed value of 2MW; The third step is to construct a wind-solar-storage model. The power output of a single photovoltaic panel at a given moment can be expressed as: （5） The output power of a wind power generation system is affected by wind speed. The power output of a wind turbine at a certain moment can be expressed as: （6） wherein: V in , V out , V r are the cut-in, cut-out and rated wind speed of the wind turbine, respectively; P wt (t) is the output power of the wind turbine at time t; P rwt is the rated power of the single wind turbine, Energy storage achieves peak shaving and valley filling through charging and discharging processes, absorbing and storing excess electrical energy from wind and solar power units. Its operating state can be characterized by the state of charge at various times, and can be expressed as follows: （7） wherein: is the self-discharge rate of the BESS; are the charging and discharging power of the BESS at node i in season s time period t, respectively; are the state of charge of the BESS at node i in season s time period t and t+1, respectively; η+and η are the charging and discharging efficiency, respectively; Δt is the time interval, taken as 1 h; The fourth step is to construct a two-layer optimization configuration model for wind, solar, computing, and storage. The upper planning layer aims to minimize the overall system cost, deciding on the installation location and capacity of wind turbines, photovoltaics, and energy storage, taking into account constraints on equipment investment, operation and maintenance, and electricity purchase costs. The lower operation layer aims to ensure power quality and reduce network losses, optimizing the real-time operation of the system, including energy storage charging and discharging strategies, data center load scheduling, and grid power interaction. The two-layer model achieves collaborative optimization through parameter correlation, and finally, the original two-layer problem is transformed into a solvable single-layer optimization problem through model transformation methods. Step 5: Solve the model. First, input the system's basic data and wind and solar resource characteristics, and obtain typical operating scenarios through scene generation and reduction technology; then, convert the original mixed-integer nonlinear two-layer model into a single-layer mixed-integer linear programming model; then, use the NSGA-II multi-objective optimization algorithm to obtain the Pareto optimal solution set; finally, use the TOPSIS method to select the configuration scheme with the best overall performance from the solution set, and output the final location and capacity determination results of each device and system operation indicators. 2.The method of claim 1, wherein, In the fourth step above, Upper-level planning model construction, The upper-level planning layer aims to minimize overall cost. The decision variables are PV, WT, and BESS location capacity, with constraints including equipment investment and site selection capacity constraints. The objective function F of the upper-level model is... （8） （1） ， （9） wherein: respectively the investment cost of PV, WT, BESS; b is the discount rate; , respectively the lifetime of PV, WT, BESS; respectively the set of candidate installation locations for PV, WT, BESS; respectively the investment cost of a single PV, WT, BESS; respectively the number of configurations of PV, WT, BESS at node i, （2） ， （10） In the formula: respectively, are the operation and maintenance costs of PV, WT, and BESS; is the electricity purchase cost from the upper-level power grid; Ds is the number of days in the s season; respectively, are the sets of installation locations of PV, WT, and BESS; is the unit operation and maintenance cost of PV; is the unit operation and maintenance cost of WT; is the unit charging and discharging power operation and maintenance cost of BESS; is the electricity price of the t period; is the active power injected by the PV of the i node in the s season and t period; respectively, are the charging power and discharging power of the BESS of the i node in the s season and t period; Δt is 1 h; is the set of terminal nodes of the alternating branch with the generator node as the first end node; is the active power of the upper-level power grid flowing into the current-level power grid in the s season and t period, The upper-level model constraints include the position and quantity constraints of PV, WT, and BESS. （11） （12） In the formula: The maximum number of PVs allowed to be installed on node i; The maximum number of WTs allowed to be installed on node i; The maximum number of BESS that can be installed on node i.

3. The capacity configuration method for a computer-coupled system considering load time-shift characteristics according to claim 1, characterized in that, In the fourth step above, Lower-level operational model construction, In the lower operational layer, to ensure the power quality of the data center and solve the problem of high network loss, the objective function f of the lower-layer model is to minimize both voltage deviation and network loss. （13） (1) ΔU all is a voltage offset, （14） In the formula: Let be the per-unit value of the voltage at node i during the time period t in season s; The set of all nodes; For absolute value calculation formulas, (2) P loss Total network loss, （15） In the formula: I ij,s,t R represents the current value of branch ij during the seasonal period t in season s; ij branch road ij The resistance; For the set of all branches, The constraints include system safety constraints, output constraints for each device, and batch processing load delay constraints. 1) Power balance constraint, At any given moment, the power generation capacity is equal to the power consumption capacity. That is, the power dissipation capacity of wind turbines, photovoltaic systems, and energy storage, as well as the power purchased from the grid, are always consistent with the power consumption of users, data centers, and energy storage charging capacity. （16） 2) Trend constraints, （17） In the formula: This represents the active and reactive power injected into node i at time t; U represents the active and reactive power consumption of node i; i U j θ represents the voltage at nodes i and j. ij This represents the voltage phase angle difference between node i and node j; G ij and B ij denotes the conductance and susceptance of the line between i-node and j-node, 3) Voltage constraint, （18） In the formula: These are the lower and upper voltage limits for node i, respectively, with values ​​of 0.93 pu and 1.07 pu. 4) Power constraints on the flow of power from the upstream power grid into the local power grid. （19） In the formula: The reactive power flowing from the upstream power grid into the local power grid during the t-period of season s; , These are the lower and upper limits of active (reactive) power flowing from the upper-level power grid into the local power grid, respectively. 5) PV and WT constraints, （20） In the formula: , 6) BESS constraint, （21） In the formula: These represent the charging and discharging states of BESS, respectively. The rated charge and discharge power of BESS for node i; The maximum BESS power capacity for node i; These are the initial BESS charges at midnight on the current day and midnight on the next day for node i in season s, respectively. These represent the lower limit, upper limit, and initial value of the BESS charge capacity, respectively. 7) Maximum delay constraint for batch processing load. （22） In the formula: D represents the maximum tolerance time for batch processing load.

Images