A Differentiated Demand Coordination Method Based on Data Center Mathematical Model

Abstract

This invention discloses a differentiated demand coordination method based on a mathematical model of a data center. The method includes constructing a spatiotemporal coordination data center mathematical model for cloud user loads that takes into account the building's thermal dynamics, based on the building's thermal dynamics and the differentiated real-time demands of cloud user loads. Furthermore, a two-stage robust optimization model for the data center that takes into account building thermal dynamics, with the goal of minimizing the data center's daily operating costs, is constructed to obtain an optimal differentiated demand coordination solution. This method considers the impact of uncertainties and establishes a two-stage robust optimization model for the data center's load arrival rate, air conditioning cooling power, indoor temperature, and operating energy consumption. This model exploits the synergistic potential of the data center's own building thermal inertia and spatiotemporal load regulation, maximizes the load and air conditioning system's regulatory flexibility, effectively reduces the energy consumption of the data center's IT equipment and air conditioning system, and improves the economic efficiency and risk mitigation capabilities of the data center's operations.

Description

Technical Field

[0001] The present invention relates to the technical field of data center mathematical models, and in particular to a differentiated demand coordination method based on a data center mathematical model. Background Art

[0002] With the rapid development of technologies such as big data and cloud computing, the demand for data communication and data computing is growing exponentially. The number and scale of data centers continue to expand, and their total energy consumption has increased dramatically. In the energy consumption structure of data centers, about 50% of the energy consumption comes from IT equipment, whose main function is to process cloud user data loads, and about 37% of the energy consumption comes from air-conditioning systems. Currently, the main goal of reducing the operating costs of data centers is to study the spatiotemporal load regulation capabilities of data centers.

[0003] However, existing research methods do not take into account the synergistic potential of the data center's own building thermal inertia and load spatiotemporal regulation, nor do they take into account the impact of the uncertainty of conventional circuit load and outdoor temperature on the load scheduling of the data center. Therefore, it is necessary to study a differentiated demand coordination method based on the mathematical model of the data center to tap the synergistic potential of the data center's own building thermal inertia and load spatiotemporal regulation, meet the differentiated real-time requirements of the data center building thermal dynamic characteristics and cloud user loads, and take into account the influence of uncertain factors to obtain the best solution to reduce the operating costs of the data center. Summary of the Invention

[0004] The purpose of the present invention is to provide a differentiated demand coordination method based on a data center mathematical model to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a differentiated demand coordination method based on a data center mathematical model, comprising the following steps:

[0006] S1. Based on the building thermal dynamic characteristics of the data center and the differentiated real-time requirements of cloud user loads, a mathematical model for the spatiotemporal coordination of cloud user loads is constructed, taking into account the ambient temperature and the building thermal dynamic characteristics.

[0007] S2. Setting constraints on the mathematical model of spatiotemporal coordination of cloud user loads in a data center taking into account the thermal dynamics of the building;

[0008] S3. Based on the mathematical model of the cloud user load spatiotemporal coordination data center taking into account the building thermal dynamic characteristics obtained in S1, with conventional load power and outdoor temperature as uncertain factors and the goal of minimizing the daily operating cost of the target data center, a two-stage robust optimization model of the data center taking into account the building thermal dynamic characteristics is constructed;

[0009] S4. The two-stage robust optimization model of the data center taking into account the thermal dynamic characteristics of the building is iteratively solved using the G&GG algorithm to obtain the optimal differential demand coordination solution for uncertain factors.

[0010] Preferably, the S1 further includes:

[0011] S11. Establish a data center data load model;

[0012] S12. Establishing a data center energy consumption model;

[0013] S13. Establish a mathematical model for the spatiotemporal coordination of cloud user load in a data center taking into account the thermal dynamics of buildings;

[0014] The establishment of the data center data load model in S11 also includes dividing the cloud user load in the data center into interactive data load and batch data load according to the real-time requirements of cloud users;

[0015] The data center data load model includes a response time model for interactive data loads. The response time model includes a transmission delay time model and an average waiting time calculation model. The transmission delay time model is as follows:

[0016] d itr =tc(1);

[0017] where d itr is the transmission time of interactive user load in period t, and tc is a constant;

[0018] The average waiting time calculation model is as follows:

[0019]

[0020] where t w,t is the average waiting time of interactive cloud user load in period t, M is the set of server types in the data center, i is a server type in set M, N is the set of server working states, j is a working state in set N, is the service rate of data center i in working state j processing interactive cloud user load in period t; is the interactive cloud user load rate during period t;

[0021] The data center data load model also includes a cloud user load processing time model for batch data loads, and the cloud user load processing time model is as follows:

[0022]

[0023] in, is the service rate of batch-type cloud user loads in data center i and in working state j during period t; is the load rate of batch cloud users in period t.

[0024] Preferably, the establishment of the data center energy consumption model includes the following steps:

[0025] S121. Establish a data center IT equipment power model;

[0026] S122. Establish a power model for the air conditioning system of the data center;

[0027] S123. Establish a total power model for the data center;

[0028] The establishment of the IT equipment power model for the data center specifically includes using the data center server power to represent the IT equipment power and modeling the server power based on the DVFS technology. The resulting IT equipment power model is as follows:

[0029]

[0030] P m,s,t =P m,st +P m,dy,t (5);

[0031]

[0032] Among them, P m,s,t P is the power of the m-type server in the s working state during the t period. m,st is the corresponding server static power, P m,dy,t is the corresponding dynamic energy consumption; k m is the dynamic energy consumption calculation coefficient of type m server; f m is the chip operating frequency;

[0033] The establishment of the air conditioning system power model specifically includes utilizing the building thermal dynamic characteristics of the data center to construct the air conditioning system power model as follows:

[0034]

[0035] Among them, C wall,ij is the wall heat capacity, which is divided into the heat capacity of the wall with window side and the heat capacity of the wall without window side; is the wall temperature; is the indoor temperature; is the temperature of the adjacent node; R wall,ij is the wall thermal resistance, which is divided into the wall thermal resistance with window side and the wall thermal resistance without window side; if the wall is exposed to external solar radiation, q ij Take 1, otherwise take 0; vij is the heat absorption rate of the wall; A wall ,ij is the wall area; is the light intensity in the corresponding direction of the wall; C romm,j Heat capacity for the room; is the outdoor temperature; N room Refers to the adjacent nodes of the region; is the window thermal resistance; ω ij is the window refractive index; is the light intensity in the direction corresponding to the window; It is an internal heat source; is the cooling capacity of the air-conditioning system during period t; is the internal heat dissipation coefficient of IT equipment in the data center;

[0036] The establishment of the total power model of the data center specifically includes constructing a total power model based on the IT equipment power model and the air conditioning system power model of the data center, and the total power model of the data center is as follows:

[0037]

[0038] in, is the total power of the data center in period t; is the power of IT equipment during period t; is the power of the air conditioning system during period t; is the total power of the conventional load during period t, which includes the power distribution system and lighting system of the data center, and is a constant;

[0039] The establishment of a data center mathematical model for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of buildings specifically includes: integrating the total power model of the data center to obtain the following mathematical model for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of the data center building and the differentiated real-time requirements of cloud user loads:

[0040]

[0041] in, is the total energy consumption of the data center in period t; is the energy consumption of IT equipment during period t; is the energy consumption of the air conditioning system during period t; is the total energy consumption of conventional load during period t; N T is the scheduling period;

[0042] Preferably, setting the constraint conditions of the mathematical model of the cloud user load spatiotemporal coordination data center taking into account the thermal dynamic characteristics of the building comprises the following steps:

[0043] S21. Construct cloud user data load rate constraints for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings;

[0044] S22. Construct a service rate constraint for a mathematical model of spatiotemporal coordination of cloud user loads in a data center taking into account the thermal dynamics of buildings.

[0045] S23. Constructing a maximum response time constraint for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings;

[0046] S24. Constructing server operating environment temperature constraints for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings;

[0047] S25. Construct a maximum data load rate constraint for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings;

[0048] S26. Constructing server operating frequency and service rate constraints for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings;

[0049] S27. Construct a mathematical model of the spatiotemporal coordination of cloud user loads in a data center with server quantity constraints taking into account the thermal dynamics of buildings.

[0050] S28. Construct the air conditioning system operation constraints of the mathematical model of the data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of the building.

[0051] Preferably, the setting of the constraint conditions specifically includes the following steps:

[0052] The step S21 of constructing the cloud user data load rate constraint specifically includes the following steps:

[0053] Calculate the total cloud user data load of the data center using the following formula:

[0054]

[0055] Among them, λ t is the sum of the data load rates of E data centers during period t, where the data load rate is the cloud user load allocated to the data center per unit time;

[0056] Based on the calculation formula (15), the data load rate of the e-th data center is λ e,t , and λ e,t The calculation formula is as follows:

[0057]

[0058] in, is the sum of the interactive data load rates that need to be processed by E data centers within a unit time period t; is the sum of the batch data load rates that need to be processed by E data centers within a unit time period t, is the interactive data load rate that the e-th data center needs to process in unit time period t, is the batch data load rate that needs to be processed by the e-th data center in unit time period t;

[0059] The construction of the maximum response time constraint in S23 specifically includes the following steps:

[0060] The maximum response time of interactive cloud user loads must meet the following constraint:

[0061] t w,t ≤D itr -d itr (17);

[0062] The maximum response time constraint for batch cloud user loads is as follows:

[0063]

[0064] Among them, D itr is the maximum response time of interactive cloud user load; d itr is the load transmission delay time; T batch The maximum response time for batch cloud user loads;

[0065] The step S28 of constructing the air conditioning system operation constraints specifically includes the following steps:

[0066] Based on the air conditioning system meeting the cooling requirements of the servers in the data center and the operating indoor temperature requirements, the operating power of the air conditioning system must meet the following constraint relationship:

[0067]

[0068] in, Sets a power cap for air conditioning system operation.

[0069] Preferably, the setting of the constraint conditions further includes the following steps:

[0070] The construction of the service rate sum constraint in S22 specifically includes the following steps:

[0071] From the perspective of server types, the sum of the service rates of e data centers satisfies the following relationship:

[0072]

[0073] From the perspective of cloud users' real-time requirements, the sum of the service rates of e data centers satisfies the following relationship:

[0074]

[0075] Among them, μ t is the sum of the service rates of the servers in E data centers during period t; is the service rate of the m-type servers in the s-working state in the data center i during period t; is the service rate of E data centers for processing interactive cloud user loads during period t; is the service rate of E data centers for processing batch cloud user loads during period t;

[0076] The step of establishing the server operating environment temperature constraint in S24 specifically includes the following steps:

[0077] The constraint relationship between the temperature range of the indoor temperature in the data center server during normal operation is as follows:

[0078]

[0079] in, They are the lower and upper limits of the indoor temperature when the data center servers are operating normally;

[0080] The construction of the maximum data load rate constraint in S25 specifically includes the following steps:

[0081] The maximum data load that a data center server can handle when processing data satisfies the following constraint:

[0082] 0≤λ e,t ≤λ e,max (twenty three);

[0083] Among them, λ e,max is the maximum data load that the e-th data center can process per second;

[0084] The step of constructing the server operating frequency and service rate constraints in S26 specifically includes the following steps:

[0085] The server processor has discrete CPU operating frequency and service rate ladders, and the two correspond to each other. The CPU operating frequency and service rate satisfy the following relationship:

[0086] f t ∈{f1,f2,…f h …,f H} (twenty four);

[0087] μ t ∈{μ1,μ2,…μh …,μ H} (25);

[0088] Among them, H is the optional operating frequency type of the server, f t is the CPU operating frequency of the server during the period t, μ t is the server service rate during period t;

[0089] In the CPU operating frequency and service rate adjustment of the data center, the CPU operating frequency and service rate based on the real-time network load are selected from the sets (24) and (25), and the CPU operating frequency and service rate based on the real-time network load are obtained. The CPU operating frequency and service rate based on the real-time network load satisfy the following relationship:

[0090]

[0091]

[0092] in, The selection variables of the working frequency and service rate during the period. Since the CPU can only work at one frequency during a certain period, The following constraints should be satisfied:

[0093]

[0094] The step of establishing the server quantity constraint in S27 specifically includes the following steps:

[0095] The constraint relationship satisfied by the number of servers participating in the adjustment in each data center is as follows:

[0096]

[0097] Among them, n sev The number of servers involved in the regulation in each data center, The upper limit of the number of servers that can participate in the adjustment in each data center.

[0098] Preferably, S3 includes the following steps:

[0099] S31. Select a target data center and obtain the target data center energy consumption based on a cloud user load spatiotemporal coordination data center mathematical model that takes into account the thermal dynamic characteristics of the building. Minimize the target data center's daily operating cost as the objective function, and establish the objective function expression as follows:

[0100]

[0101] Among them, M dcA,t is the total energy consumption of the selected target data center A in period t, M dcB,tis the total energy consumption of the selected target data center B during period t, c t The mathematical model of spatiotemporal coordination of cloud user loads of Data Center A and Data Center B taking into account the thermal dynamic characteristics of buildings satisfies the constraints in S21-S28.

[0102] S32. Establish a data center deterministic mathematical model that does not consider uncertain factors, and the compact form of the data center deterministic mathematical model is as follows:

[0103]

[0104] Among them, x and y are variables, and x and y satisfy the following relationship:

[0105]

[0106] Where C is the number of servers in the data center, and is the coefficient column vector of formula (31); D, K, F, G and I u is the coefficient matrix under the corresponding constraints, d and k are constant column vectors, where the inequality constraints are the second row in (32) and satisfy the constraint relationships (17)-(19), (22)-(23), and (30). The equality constraints are the third row in (32) and satisfy the constraint relationships (15)-(16), (18), (20)-(21), and (28)-(29). The relationship between variables x and y is the fourth row and satisfies the constraint relationships (26)-(27). The fifth row indicates that in the data center deterministic mathematical model, the uncertain factor variables are the predicted values ​​of each time period, and the predicted values ​​satisfy the following relationship:

[0107]

[0108] in, represents the predicted value of conventional load power in period t, represents the predicted value of outdoor temperature in period t;

[0109] S33. Construct a box type uncertainty set U to which the uncertain factors conventional load and outdoor temperature belong, and U satisfies the following relationship:

[0110]

[0111] Among them, u P,t and u T,t are uncertain variables, namely, normal load power and outdoor temperature; and is the maximum fluctuation deviation allowed for the uncertain variable;

[0112] S34. Determine the mathematical model based on the data center. Using the box-shaped uncertainty set U of the uncertain factors as a constraint, a two-stage robust optimization method is used to construct a mathematical model of the data center that takes into account the thermal dynamic characteristics of the building as follows:

[0113]

[0114] The first-stage problem is the outermost min problem, with the optimization variable x; the second-stage problem is the inner max-min problem, with the optimization variables u and y. The min problem is equivalent to the objective function of Equation (31), which represents the minimization of the daily operating cost of the data center. Ω(x,u) means the feasible domain of the optimization variable y when the value of (x,u) is given, and the expression of Ω(x,u) is as follows:

[0115]

[0116] Among them, γ, λ, ν, and π are dual variables used to deal with the minimization problem in the second stage.

[0117] Preferably, the step S4 further includes the following steps:

[0118] S41. Decompose the mathematical model of the data center taking into account the thermal dynamic characteristics of the building, which is constructed using the two-stage robust optimization method in S34, to obtain a main problem model and sub-problem models. The main problem model is as follows:

[0119]

[0120] Where k and l are the number of iterations, and l∈k, y l is the solution of the data center subproblem after the lth iteration; is the value of the uncertain factors normal load power and outdoor temperature for the first iteration;

[0121] The sub-problem model is as follows:

[0122]

[0123] For the subproblem model (38) of the NP-hard problem, given (x,u), the inner minimization in (38) is transformed into a linear problem, and according to the pair theory, the inner min in (38) is transformed into the max form and merged with the outer max. The transformed subproblem model is as follows:

[0124]

[0125] For the bilinear term u in (40) Tπ, when the uncertainty factors of the data center, the conventional load and the outdoor temperature, are taken as the right boundary in equation (35), the required operating cost of the data center is high. Therefore, the uncertainty U to which the uncertainty factors, the conventional load and the outdoor temperature belong, is optimized as follows:

[0126]

[0127] Where B=[B P,t B T,t ] T is a binary variable, and the value 1 indicates that the normal load power and outdoor temperature take the maximum value of the uncertainty set; Γ p,t and Γ T,t are the uncertainty adjustment parameters corresponding to conventional load power and outdoor temperature, respectively. Smaller Γ values ​​indicate a more risky model, so Γ is used to adjust the conservatism of the scheme. Substituting the uncertainty set of the uncertainties in Equation (41) into Equation (40) yields the product of a continuous variable and a binary variable. Therefore, linearizing Equation (40) using the large-M method yields the following expression:

[0128]

[0129] Where, Δu=(Δu P,t ,Δu T,t ) T ,B'=(B' P,t ,B' T,t ) T is a continuous auxiliary variable, is a positive real number;

[0130] Preferably, the step S4 further includes the following steps:

[0131] S42. For the main problem model and sub-problem model obtained in S41, the G&GG algorithm is used to iteratively solve the problem to obtain the optimal differential demand coordination solution for the uncertain factors. The solution method includes the following steps:

[0132] S421: Given a set of conventional load power and outdoor temperature as the worst scenario u, set the lower bound of the data center daily operating cost provided by the main problem LB = +∞, the upper bound provided by the subproblem UB = -∞, and the convergence threshold ε;

[0133] S422: The worst scenario obtained based on the first iteration of the data center Solve the main problem equation (38) and get Update the Nether

[0134] S423: Will be solved Substitute it into Equation (41) as a fixed value to obtain the objective function value of the sub-problem and the value of the uncertain variable u, and use the obtained objective function value as the new upper bound to update the upper bound

[0135] S424: Determine whether the upper and lower bounds meet the convergence conditions. When UB-LB≤ε, the iteration ends and the optimal solution is obtained. Otherwise, the variable y is increased. k+1 And the following constraints:

[0136]

[0137] Let , re-solve the main and sub-problems until the algorithm converges or reaches the maximum number of iterations.

[0138] Compared with the prior art, the present invention has the following beneficial effects:

[0139] 1. Based on the synergistic effect of the spatiotemporal regulation characteristics of interactive data loads and batch data loads in the data center and the thermal dynamic characteristics of the data center's building, the present invention establishes a spatiotemporal coordinated data center data model for user loads that takes into account the thermal inertia of the data center's own building. A quantitative mathematical relationship is established between the data center's load arrival rate, air conditioning cooling power, indoor temperature, and operating energy consumption. This fully taps the synergistic potential of the data center's own building thermal inertia and spatiotemporal regulation of loads, and leverages the regulatory flexibility of cloud user loads and the air conditioning system in the data center. While ensuring the normal operating ambient temperature of the servers, this method achieves economical and efficient operation of the data center, effectively reduces the energy consumption of the data center's IT equipment and air conditioning system, and improves the economic efficiency of data center operation.

[0140] 2. The present invention takes into account the uncertainty of conventional load power and outdoor temperature, and proposes a data center mathematical model that takes into account the thermal dynamic characteristics of the building and is constructed using a two-stage robust optimization method. On the basis of ensuring the normal operating temperature of the data center server, the data center's ability to resist fluctuations in conventional load power and outdoor temperature is significantly improved, and the reduction in the system's operating economy due to excessive conservatism is avoided. This enables the data center to meet the usage needs of cloud users under the condition of uncertainty in conventional load and outdoor temperature, improve the data center's ability to resist risks, and reduce operating costs.

[0141] 3. The present invention proposes a C&CG algorithm that efficiently solves the mathematical model of a data center taking into account the thermal dynamic characteristics of the building, which is constructed using a two-stage robust optimization method. By decomposing the original problem into a main problem and sub-problems and performing alternating calculations, the optimal differential demand coordination solution for uncertain factors is obtained. By constructing an uncertainty set of uncertain factors, the NP-hard problem in the sub-problem is converted into a linear problem, which reduces the computational difficulty and can adjust the conservatism of the model, which is conducive to reducing the redundancy of the scheduling results.

[0142] 4. The present invention sets constraints on the mathematical model of the data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of buildings, which is conducive to improving the reliability of the mathematical model of the data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of buildings. By solving the problem, the optimal solution to the original problem is obtained, that is, the scheduling plan with the optimal operating cost of the data center when the conventional load power of the data center is the largest and the outdoor temperature is the highest, so as to meet the differentiated coordination needs of the mathematical model of the data center. BRIEF DESCRIPTION OF THE DRAWINGS

[0143] Figure 1 is a flow chart of the overall method of the present invention;

[0144] Figure 2 This is the flow chart of S1 in the present invention;

[0145] Figure 3 This is a flow chart of S2 in the present invention;

[0146] Figure 4 This is the flow chart of S4 in the present invention;

[0147] Figure 5 This is a flow chart of S4 in the present invention. DETAILED DESCRIPTION

[0148] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0149] In the description of the present invention, it should be noted that the terms "upper," "lower," "inner," "outer," "front end," "rear end," "both ends," "one end," "the other end," and the like, indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0150] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "provided with," "connected," etc., should be understood in a broad sense. For example, "connected" may refer to a fixed connection, a detachable connection, or an integral connection; it may refer to a mechanical connection or an electrical connection; it may refer to a direct connection or an indirect connection through an intermediate medium; it may refer to internal communication between two components. Those skilled in the art will be able to understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0151] See also Figure 1 and Figure 2 The present invention provides an embodiment of a method for coordinating differentiated demands based on a data center mathematical model, which specifically includes the following steps:

[0152] S1. Based on the building thermal dynamic characteristics of the data center and the differentiated real-time requirements of cloud user loads, a mathematical model for the spatiotemporal coordination of cloud user loads is constructed, taking into account the ambient temperature and the building thermal dynamic characteristics.

[0153] S1 also includes:

[0154] S11. Establish a data center data load model;

[0155] S12. Establishing a data center energy consumption model;

[0156] S13. Establishing a mathematical model for spatiotemporal coordination of cloud user loads in a data center taking into account the thermal dynamic characteristics of the building;

[0157] Specifically, the establishment of the data center data load model in S11 also includes dividing the cloud user load in the data center into interactive data load and batch data load according to the real-time requirements of cloud users;

[0158] The data center data load model includes a response time model for interactive data loads. The response time model includes a transmission delay time model and an average waiting time calculation model. The transmission delay time model is as follows:

[0159] d itr =tc(1);

[0160] where d itr is the transmission time of interactive user load in period t, tc is a constant;

[0161] The average waiting time calculation model is as follows:

[0162]

[0163] where t w,tis the average waiting time of interactive cloud user load in period t, M is the set of server types in the data center, i is a server type in set M, N is the set of server working states, j is a working state in set N, is the service rate of data center i in working state j processing interactive cloud user load in period t; is the interactive cloud user load rate during period t;

[0164] The data center data load model also includes a cloud user load processing time model for batch data loads, and the cloud user load processing time model is as follows:

[0165]

[0166] in, is the service rate of batch-type cloud user loads in data center i and in working state j during period t; is the load rate of batch processing cloud users in period t;

[0167] S12. Establishing a data center energy consumption model includes the following steps:

[0168] S121. Establish a data center IT equipment power model;

[0169] S122. Establish a power model for the air conditioning system of the data center;

[0170] S123. Establish a total power model for the data center;

[0171] The establishment of the IT equipment power model in the data center specifically involves using the data center server power to represent the IT equipment power and modeling the server power based on the DVFS technology. The resulting IT equipment power model is as follows:

[0172]

[0173] P m,s,t =P m,st +P m,dy,t (5);

[0174]

[0175] Among them, P m,s,t P is the power of the m-type server in the s working state during the t period. m,st is the corresponding server static power, P m,dy,t is the corresponding dynamic energy consumption; k m is the dynamic energy consumption calculation coefficient of type m server; f m is the chip operating frequency;

[0176] The establishment of the air conditioning system power model specifically includes using the building thermal dynamic characteristics of the data center to construct the air conditioning system power model as follows:

[0177]

[0178] Among them, C wall,ij is the wall heat capacity, which is divided into the heat capacity of the wall with window side and the heat capacity of the wall without window side; is the wall temperature; is the indoor temperature; is the temperature of the adjacent node; R wall,ij is the wall thermal resistance, which is divided into the wall thermal resistance with window side and the wall thermal resistance without window side; if the wall is exposed to external solar radiation, q ij Take 1, otherwise take 0; v ij is the heat absorption rate of the wall; A wall ,ij is the wall area; is the light intensity in the corresponding direction of the wall; C romm,j Heat capacity for the room; is the outdoor temperature; N room Refers to the adjacent nodes of the region; is the window thermal resistance; ω ij is the window refractive index; is the light intensity in the direction corresponding to the window; It is an internal heat source; is the cooling capacity of the air-conditioning system during period t; is the internal heat dissipation coefficient of IT equipment in the data center;

[0179] Establishing the total power model of a data center specifically includes constructing a total power model based on the IT equipment power model and the air conditioning system power model of the data center. The total power model of the data center is as follows:

[0180]

[0181] in, is the total power of the data center in period t; is the power of IT equipment during period t; is the power of the air conditioning system during period t; is the total power of the conventional load during period t, which includes the power distribution system and lighting system of the data center, and is a constant;

[0182] The establishment of a data center mathematical model for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of the building in S13 specifically includes: integrating the total power model of the data center to obtain the following mathematical model for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of the data center building and the differentiated real-time requirements of cloud user loads:

[0183]

[0184] in, is the total energy consumption of the data center in period t; is the energy consumption of IT equipment during period t; is the energy consumption of the air conditioning system during period t; is the total energy consumption of conventional load during period t; N T The scheduling period.

[0185] Furthermore, by adopting the above steps, based on the synergistic effect of the spatiotemporal regulation characteristics of interactive data loads and batch data loads in the data center and the building thermal dynamic characteristics of the data center, a user load spatiotemporal coordination data center data model taking into account the thermal inertia of the data center's own building is established, and a quantitative mathematical relationship between the data center's load arrival rate, air-conditioning cooling power, indoor temperature and operating energy consumption is established. The synergistic potential of the data center's own building thermal inertia and load spatiotemporal regulation is fully tapped, and the adjustment flexibility of the cloud user load and the air-conditioning system in the data center is brought into play, thereby effectively reducing the energy consumption of the data center's IT equipment and air-conditioning system and improving the economic efficiency of the data center's operation.

[0186] See also Figure 1 and Figure 3 The present invention provides an embodiment of a method for coordinating differentiated demands based on a data center mathematical model, comprising the following steps:

[0187] S2. Setting constraints on a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of the building, specifically comprising the following steps:

[0188] S21. Constructing a cloud user data load rate constraint for a mathematical model of a cloud user load spatiotemporal coordinated data center taking into account the thermal dynamic characteristics of a building, specifically comprising the following steps:

[0189] Calculate the total cloud user data load of the data center using the following formula:

[0190]

[0191] Among them, λ t is the sum of the data load rates of E data centers during period t, where the data load rate is the cloud user load allocated to the data center per unit time;

[0192] Based on the calculation formula (15), the data load rate of the e-th data center is λ e,t , and λ e,t The calculation formula is as follows:

[0193]

[0194] in, is the sum of the interactive data load rates that need to be processed by E data centers within a unit time period t; is the sum of the batch data load rates that need to be processed by E data centers within a unit time period t, is the interactive data load rate that the e-th data center needs to process in unit time period t, is the batch data load rate that needs to be processed by the e-th data center in unit time period t;

[0195] S22. Constructing a service rate sum constraint for a mathematical model of a cloud user load spatiotemporal coordinated data center taking into account the thermal dynamic characteristics of the building, specifically comprising the following steps:

[0196] From the perspective of server types, the sum of the service rates of e data centers satisfies the following relationship:

[0197]

[0198] From the perspective of cloud users' real-time requirements, the sum of the service rates of e data centers satisfies the following relationship:

[0199]

[0200] Among them, μ t is the sum of the service rates of the servers in E data centers during period t; is the service rate of the m-type servers in the s-working state in the data center i during period t; is the service rate of E data centers for processing interactive cloud user loads during period t; is the service rate of E data centers for processing batch cloud user loads during period t;

[0201] S23. Constructing a maximum response time constraint for a mathematical model of a cloud user load spatiotemporal coordinated data center taking into account the thermal dynamic characteristics of the building, specifically comprising the following steps:

[0202] The maximum response time of interactive cloud user loads must meet the following constraint:

[0203] t w,t ≤D itr -d itr (17);

[0204] The maximum response time constraint for batch cloud user loads is as follows:

[0205]

[0206] Among them, D itr is the maximum response time of interactive cloud user load; d itr is the load transmission delay time; T batch The maximum response time for batch cloud user loads;

[0207] S24. Constructing a server operating environment temperature constraint for a mathematical model of a cloud user load spatiotemporal coordinated data center taking into account the dynamic thermal characteristics of the building, specifically comprising the following steps:

[0208] The constraint relationship between the temperature range of the indoor temperature in the data center server during normal operation is as follows:

[0209]

[0210] in, They are the lower and upper limits of the indoor temperature when the data center servers are operating normally;

[0211] S25. Constructing a maximum data load rate constraint for a mathematical model of a cloud user load spatiotemporal coordinated data center taking into account the thermal dynamic characteristics of the building, specifically comprising the following steps:

[0212] The maximum data load that a data center server can handle when processing data satisfies the following constraint:

[0213] 0≤λ e,t ≤λ e,max (twenty three);

[0214] Among them, λ e,max is the maximum data load that the e-th data center can process per second;

[0215] S26. Constructing server operating frequency and service rate constraints for a mathematical model of a cloud user load spatiotemporal coordinated data center taking into account the thermal dynamic characteristics of the building, specifically comprising the following steps:

[0216] The server processor has discrete CPU operating frequency and service rate ladders, and the two correspond to each other. The CPU operating frequency and service rate satisfy the following relationship:

[0217] f t ∈{f1,f2,…f h …,f H} (twenty four);

[0218] μ t∈{μ1,μ2,…μ h …,μ H} (25);

[0219] Among them, H is the optional operating frequency type of the server, f t is the CPU operating frequency of the server during the period t, μ t is the server service rate during period t;

[0220] In the CPU operating frequency and service rate adjustment of the data center, the CPU operating frequency and service rate based on the real-time network load are selected from the sets (24) and (25), and the CPU operating frequency and service rate based on the real-time network load are obtained. The CPU operating frequency and service rate based on the real-time network load satisfy the following relationship:

[0221]

[0222] in, The selection variables of the working frequency and service rate during the period. Since the CPU can only work at one frequency during a certain period, The following constraints should be satisfied:

[0223]

[0224] S27. Constructing a server quantity constraint for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the dynamic thermal characteristics of buildings, specifically comprising the following steps:

[0225] The constraint relationship satisfied by the number of servers participating in the adjustment in each data center is as follows:

[0226]

[0227] Among them, n sev The number of servers involved in the regulation in each data center, The upper limit of the number of servers that can participate in the regulation in each data center;

[0228] S28. Constructing air conditioning system operation constraints for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of the building, specifically comprising the following steps:

[0229] Based on the air conditioning system meeting the cooling requirements of the servers in the data center and the operating indoor temperature requirements, the operating power of the air conditioning system must meet the following constraint relationship:

[0230]

[0231] in, Sets a power cap for air conditioning system operation.

[0232] Furthermore, by setting constraints on the mathematical model of the spatiotemporal coordination of cloud user loads in data centers taking into account the thermal dynamic characteristics of buildings, it is beneficial to improve the reliability of the mathematical model of the spatiotemporal coordination of cloud user loads in data centers taking into account the thermal dynamic characteristics of buildings.

[0233] See also Figure 1 and Figure 4 The present invention provides an embodiment: a differentiated demand coordination method based on a data center mathematical model. In S3, a two-stage robust optimization model for a data center taking into account the thermal dynamic characteristics of a building is constructed based on the cloud user load spatiotemporal coordination data center mathematical model obtained in S1, with conventional load power and outdoor temperature as uncertain factors and the goal of minimizing the daily operating cost of the target data center. The method specifically includes the following steps:

[0234] S31. Select a target data center and obtain the target data center energy consumption based on a cloud user load spatiotemporal coordination data center mathematical model that takes into account the thermal dynamic characteristics of the building. Minimize the target data center's daily operating cost as the objective function, and establish the objective function expression as follows:

[0235]

[0236] Among them, M dcA,t is the total energy consumption of the selected target data center A in period t, M dcB,t is the total energy consumption of the selected target data center B in period t, c t The mathematical model of spatiotemporal coordination of cloud user loads of Data Center A and Data Center B taking into account the thermal dynamic characteristics of buildings satisfies the constraints in S21-S28.

[0237] S32. Establish a data center deterministic mathematical model that does not consider uncertain factors, and the compact form of the data center deterministic mathematical model is as follows:

[0238]

[0239] Among them, x and y are variables, and x and y satisfy the following relationship:

[0240]

[0241] Where C is the number of servers in the data center, and is the coefficient column vector of formula (31); D, K, F, G and I uis the coefficient matrix under the corresponding constraints, d and k are constant column vectors, where the inequality constraints are the second row in (32) and satisfy the constraint relationships (17)-(19), (22)-(23), and (30). The equality constraints are the third row in (32) and satisfy the constraint relationships (15)-(16), (18), (20)-(21), and (28)-(29). The relationship between variables x and y is the fourth row and satisfies the constraint relationships (26)-(27). The fifth row indicates that in the data center deterministic mathematical model, the uncertain factor variables are the predicted values ​​of each time period, and the predicted values ​​satisfy the following relationship:

[0242]

[0243] in, represents the predicted value of conventional load power in period t, represents the predicted value of outdoor temperature in period t;

[0244] S33. Construct a box type uncertainty set U to which the uncertain factors conventional load and outdoor temperature belong, and U satisfies the following relationship:

[0245]

[0246] Among them, u P,t and u T,t are uncertain variables, namely, normal load power and outdoor temperature; and is the maximum fluctuation deviation allowed for the uncertain variable;

[0247] S34. Determine the mathematical model based on the data center. Using the box-shaped uncertainty set U of the uncertain factors as a constraint, use the two-stage robust optimization method to construct a two-stage robust optimization model of the data center that takes into account the thermal dynamic characteristics of the building as follows:

[0248]

[0249] The first-stage problem is the outermost min problem, with the optimization variable x; the second-stage problem is the inner max-min problem, with the optimization variables u and y. The min problem is equivalent to the objective function of Equation (31), which represents the minimization of the daily operating cost of the data center. Ω(x,u) means the feasible domain of the optimization variable y when the value of (x,u) is given, and the expression of Ω(x,u) is as follows:

[0250]

[0251] Among them, γ, λ, ν, and π are dual variables used to deal with the minimization problem in the second stage.

[0252] Furthermore, considering the uncertainty of conventional load power and outdoor temperature, a mathematical model of the data center taking into account the thermal dynamic characteristics of the building is proposed, which is constructed using a two-stage robust optimization method. On the basis of ensuring the normal operating temperature of the data center server, the data center's ability to withstand fluctuations in conventional load power and outdoor temperature is significantly improved, and the reduction in the system's operating economy due to excessive conservatism is avoided. This enables the data center to meet the usage needs of cloud users under the condition of uncertainty in conventional load and outdoor temperature, while reducing operating costs.

[0253] See also Figure 1 and Figure 5 , an embodiment provided by the present invention: a differentiated demand coordination method based on a mathematical model of a data center, including the two-stage robust optimization model for a data center taking into account the thermal dynamic characteristics of the building proposed in S4, using the G&GG algorithm for iterative solution to obtain the optimal differentiated demand coordination solution for uncertain factors;

[0254] Specifically, S4 also includes the following steps:

[0255] S41. Decompose the mathematical model of the data center taking into account the thermal dynamic characteristics of the building, which is constructed using the two-stage robust optimization method in S34, to obtain a main problem model and sub-problem models. The main problem model is as follows:

[0256]

[0257] Where k and l are the number of iterations, and l∈k, y l is the solution of the data center subproblem after the lth iteration; is the value of the uncertain factors normal load power and outdoor temperature for the first iteration;

[0258] The sub-problem model is as follows:

[0259]

[0260] For the subproblem model (38) of the NP-hard problem, given (x,u), the inner minimization in (38) is transformed into a linear problem, and according to the pair theory, the inner min in (38) is transformed into the max form and merged with the outer max. The transformed subproblem model is as follows:

[0261]

[0262] For the bilinear term u in (40) Tπ, when the uncertainty factors of the data center, the conventional load and the outdoor temperature, are taken as the right boundary in equation (35), the required operating cost of the data center is high. Therefore, the uncertainty U to which the uncertainty factors, the conventional load and the outdoor temperature belong, is optimized as follows:

[0263]

[0264] Where B=[B P,t B T,t ] T is a binary variable, and the value 1 indicates that the normal load power and outdoor temperature take the maximum value of the uncertainty set; Γ p,t and Γ T,t are the uncertainty adjustment parameters corresponding to conventional load power and outdoor temperature, respectively. Smaller Γ values ​​indicate a more risky model, so Γ is used to adjust the conservatism of the scheme. Substituting the uncertainty set of the uncertainties in Equation (41) into Equation (40) yields the product of a continuous variable and a binary variable. Therefore, linearizing Equation (40) using the large-M method yields the following expression:

[0265]

[0266] Where, Δu=(Δu P,t ,Δu T,t ) T ,B'=(B' P,t ,B' T,t ) T is a continuous auxiliary variable, is a large positive real number.

[0267] Furthermore, a C&CG algorithm is proposed to efficiently solve the mathematical model of a data center taking into account the thermal dynamic characteristics of the building, which is constructed using a two-stage robust optimization method. By decomposing the original problem into a main problem and sub-problems for alternating calculations, the optimal differential demand coordination solution for uncertain factors is obtained. By constructing an uncertainty set of uncertain factors, the NP-hard problem in the sub-problem is transformed into a linear problem, reducing the computational difficulty and being able to adjust the conservatism of the model, which is conducive to reducing the redundancy of the scheduling results.

[0268] See also Figure 1 and Figure 5 , an embodiment provided by the present invention: a differentiated demand coordination method based on a mathematical model of a data center, including the two-stage robust optimization model for a data center taking into account the thermal dynamic characteristics of the building proposed in S4, using the G&GG algorithm for iterative solution to obtain the optimal differentiated demand coordination solution for uncertain factors;

[0269] Specifically, S4 also includes the following steps:

[0270] S42. For the main problem model and sub-problem model obtained in S41, the G&GG algorithm is used to iteratively solve the problem to obtain the optimal differential demand coordination solution for the uncertain factors. The solution method includes the following steps:

[0271] S421: Given a set of conventional load power and outdoor temperature as the worst scenario u, set the lower bound of the data center daily operating cost provided by the main problem LB = +∞, the upper bound provided by the subproblem UB = -∞, and the convergence threshold ε;

[0272] S422: The worst scenario obtained based on the first iteration of the data center Solve the main problem equation (38) and get Update the Nether

[0273] S423: Will be solved Substitute it into Equation (41) as a fixed value to obtain the objective function value of the subproblem and the value of the uncertain variable u, and use the obtained objective function value as the new upper bound to update the upper bound

[0274] S424: Determine whether the upper and lower bounds meet the convergence conditions. When UB-LB≤ε, the iteration ends and the optimal solution is obtained. Otherwise, the variable y is increased. k+1 And the following constraints:

[0275]

[0276] Let , re-solve the main and sub-problems until the algorithm converges or reaches the maximum number of iterations.

[0277] Furthermore, the optimal solution to the original problem is obtained by solving it, that is, the scheduling plan with the optimal operating cost of the data center when the conventional load power of the data center is the largest and the outdoor temperature is the highest, so as to meet the differentiated coordination needs of the mathematical model of the data center.

[0278] Working principle: A differentiated demand coordination method based on a mathematical model of a data center, including constructing a spatiotemporal coordination data center mathematical model of cloud user load taking into account the thermal dynamic characteristics of the building according to the thermal dynamic characteristics of the data center and the differentiated real-time requirements of the cloud user load, setting constraints on the spatiotemporal coordination data center mathematical model of cloud user load taking into account the thermal dynamic characteristics of the building, and constructing a two-stage robust optimization model of the data center taking into account the thermal dynamic characteristics of the building based on the obtained spatiotemporal coordination data center mathematical model of cloud user load taking into account the thermal dynamic characteristics of the building, with conventional load power and outdoor temperature as uncertain factors and the goal of minimizing the daily operating cost of the target data center. The two-stage robust optimization model of the data center taking into account the thermal dynamic characteristics of the building is iteratively solved using the G&GG algorithm to obtain the optimal differentiated demand coordination solution for uncertain factors.

[0279] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A differentiated demand coordination method based on a data center mathematical model, characterized in that: The steps include: S1. Based on the building thermal dynamic characteristics of the data center and the differentiated real-time requirements of cloud user loads, a mathematical model for the spatiotemporal coordination of cloud user loads is constructed, taking into account the ambient temperature and the building thermal dynamic characteristics. S2. Setting constraints on the mathematical model of spatiotemporal coordination of cloud user loads in a data center taking into account the thermal dynamics of the building; S3. Based on the mathematical model of the cloud user load spatiotemporal coordination data center taking into account the building thermal dynamic characteristics obtained in S1, with conventional load power and outdoor temperature as uncertain factors and the goal of minimizing the daily operating cost of the target data center, a two-stage robust optimization model of the data center taking into account the building thermal dynamic characteristics is constructed; S4. A two-stage robust optimization model for data centers that takes into account the thermal dynamic characteristics of buildings is iteratively solved using the G&GG algorithm to obtain the optimal differential demand coordination solution for uncertain factors. Said S1 also includes: S11. Establish a data center data load model; S12. Establishing a data center energy consumption model; S13. Establish a mathematical model for the spatiotemporal coordination of cloud user loads in a data center taking into account the thermal dynamics of buildings; The establishment of the data center data load model in S11 also includes dividing the cloud user load in the data center into interactive data load and batch data load according to the real-time requirements of cloud users; The data center data load model includes a response time model for interactive data loads. The response time model includes a transmission delay time model and an average waiting time calculation model. The transmission delay time model is as follows: d itr =tc(1); where d itr is the transmission time of interactive user load in period t, tc is a constant; The average waiting time calculation model is as follows: where t w,t is the average waiting time of interactive cloud user load in period t, M is the set of server types in the data center, i is a server type in set M, N is the set of server working states, j is a working state in set N, is the service rate of data center i that processes interactive cloud user loads in working state j during period t; is the interactive cloud user load rate during period t; The data center data load model also includes a cloud user load processing time model for batch data loads, and the cloud user load processing time model is as follows: in, is the service rate of batch-type cloud user loads in data center i and in working state j during period t; is the load rate of batch processing cloud users in period t; The establishment of the data center energy consumption model includes the following steps: S121. Establish a data center IT equipment power model; S122. Establish a power model for the air conditioning system of the data center; S123. Establish a total power model for the data center; The establishment of the IT equipment power model for the data center specifically includes using the data center server power to represent the IT equipment power and modeling the server power based on the DVFS technology. The resulting IT equipment power model is as follows: P m,s,t =P m,st +P m,dy,t (5); Among them, P m,s,t is the power of the m-type server in the s working state during the t period, P m,st is the corresponding server static power, P m,dy,t is the corresponding dynamic energy consumption; k m is the dynamic energy consumption calculation coefficient of type m server; f m is the chip operating frequency; The establishment of the air conditioning system power model specifically includes using the building thermal dynamic characteristics of the data center to construct the air conditioning system power model as follows: Among them, C wall,ij is the wall heat capacity, which is divided into the heat capacity of the wall with window side and the heat capacity of the wall without window side; is the wall temperature; is the indoor temperature; is the temperature of the adjacent node; R wall,ij is the wall thermal resistance, which is divided into the wall thermal resistance with window side and the wall thermal resistance without window side; if the wall is exposed to external solar radiation, q ij Take 1, otherwise take 0; v ij is the wall heat absorption rate; A wall,ij is the wall area; is the light intensity in the corresponding direction of the wall; C romm,j Heat capacity for the room; is the outdoor temperature; N room Refers to the adjacent nodes of the region; is the window thermal resistance; ω ij is the window refractive index; is the light intensity in the direction corresponding to the window; It is an internal heat source; is the cooling capacity of the air-conditioning system during period t; is the internal heat dissipation coefficient of IT equipment in the data center; The establishment of the total power model of the data center specifically includes constructing a total power model based on the IT equipment power model and the air conditioning system power model of the data center, and the total power model of the data center is as follows: in, is the total power of the data center in period t; is the power of IT equipment during period t; is the power of the air conditioning system during period t; is the total power of the conventional load during period t, which includes the power distribution system and lighting system of the data center, and is a constant; The establishment of a data center mathematical model for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of buildings specifically includes: integrating the total power model of the data center to obtain the following mathematical model for spatiotemporal coordination of cloud user loads taking into account the thermal dynamic characteristics of the data center building and the differentiated real-time requirements of cloud user loads: in, is the total energy consumption of the data center in period t; is the energy consumption of IT equipment during period t; is the energy consumption of the air conditioning system during period t; is the total energy consumption of conventional load during period t; N T is the scheduling period; The setting of constraint conditions for a mathematical model of a cloud user load spatiotemporal coordinated data center taking into account the thermal dynamic characteristics of a building comprises the following steps: S21. Construct cloud user data load rate constraints for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings; S22. Construct a service rate constraint for a mathematical model of spatiotemporal coordination of cloud user loads in a data center taking into account the thermal dynamics of buildings. S23. Constructing a maximum response time constraint for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings; S24. Constructing server operating environment temperature constraints for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings; S25. Construct a maximum data load rate constraint for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings; S26. Constructing server operating frequency and service rate constraints for a mathematical model of a data center for spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of buildings; S27. Construct a mathematical model of the spatiotemporal coordination of cloud user loads in a data center with server quantity constraints taking into account the thermal dynamics of buildings. S28. Constructing air conditioning system operation constraints for a mathematical model of a data center with spatiotemporal coordination of cloud user loads taking into account the thermal dynamics of the building; The setting of the constraint conditions specifically includes the following steps: The step S21 of constructing the cloud user data load rate constraint specifically includes the following steps: Calculate the total cloud user data load of the data center using the following formula: Among them, λ t is the sum of the data load rates of E data centers during period t, where the data load rate is the cloud user load allocated to the data center per unit time; Based on the calculation formula (15), the data load rate of the e-th data center is λ e,t , and λ e,t The calculation formula is as follows: in, is the sum of the interactive data load rates that need to be processed by E data centers within a unit time period t; is the sum of the batch data load rates that need to be processed by E data centers within a unit time period t, is the interactive data load rate that the e-th data center needs to process in unit time period t, is the batch data load rate that needs to be processed by the e-th data center in unit time period t; The construction of the maximum response time constraint in S23 specifically includes the following steps: The maximum response time of interactive cloud user loads must meet the following constraint: t w,t ≤D itr -d itr (17); The maximum response time constraint for batch cloud user loads is as follows: Among them, D itr is the maximum response time of interactive cloud user load; d itr is the load transmission delay time; T batch The maximum response time for batch cloud user loads; The step S28 of constructing the air conditioning system operation constraints specifically includes the following steps: Based on the air conditioning system meeting the cooling requirements of the servers in the data center and the operating indoor temperature requirements, the operating power of the air conditioning system must meet the following constraint relationship: in, Set a power cap for the air conditioning system; The setting of the constraint conditions also includes the following steps: The construction of the service rate sum constraint in S22 specifically includes the following steps: From the perspective of server types, the sum of the service rates of e data centers satisfies the following relationship: From the perspective of cloud users' real-time requirements, the sum of the service rates of e data centers satisfies the following relationship: Among them, μ t is the sum of the service rates of the servers in E data centers during period t; is the service rate of the m-type servers in the s-working state in the data center i during period t; is the service rate of E data centers for processing interactive cloud user loads during period t; is the service rate of E data centers for processing batch cloud user loads during period t; The step of establishing the server operating environment temperature constraint in S24 specifically includes the following steps: The constraint relationship between the temperature range of the indoor temperature in the data center server during normal operation is as follows: in, They are the lower and upper limits of the indoor temperature when the data center servers are operating normally; The construction of the maximum data load rate constraint in S25 specifically includes the following steps: The maximum data load that a data center server can handle when processing data satisfies the following constraint: 0≤λ e,t ≤λ e,max (23); Among them, λ e,max is the maximum data load that the e-th data center can process per second; The step of constructing the server operating frequency and service rate constraints in S26 specifically includes the following steps: The server processor has discrete CPU operating frequency and service rate ladders, and the two correspond to each other. The CPU operating frequency and service rate satisfy the following relationship: f t ∈{f1,f2,…f h …,f H } (24); m t ∈{μ1,μ2,…μ h …,m H } (25); Among them, H is the optional operating frequency type of the server, f t is the CPU operating frequency of the server during the period t, μ t is the server service rate during period t; In the CPU operating frequency and service rate adjustment of the data center, the CPU operating frequency and service rate based on the real-time network load are selected from the sets (24) and (25), and the CPU operating frequency and service rate based on the real-time network load are obtained. The CPU operating frequency and service rate based on the real-time network load satisfy the following relationship: in, The selection variables of the working frequency and service rate during the period. Since the CPU can only work at one frequency during a certain period, The following constraints should be satisfied: The step of establishing the server quantity constraint in S27 specifically includes the following steps: The constraint relationship satisfied by the number of servers participating in the adjustment in each data center is as follows: Among them, n sev The number of servers involved in the regulation in each data center, The upper limit of the number of servers that can participate in the regulation in each data center; The S3 includes the following steps: S31. Select a target data center and obtain the target data center energy consumption based on a cloud user load spatiotemporal coordination data center mathematical model that takes into account the thermal dynamic characteristics of the building. Minimize the target data center's daily operating cost as the objective function, and establish the objective function expression as follows: Among them, M dcA,t is the total energy consumption of the selected target data center A in period t, M dcB,t is the total energy consumption of the selected target data center B in period t, c t The mathematical model of spatiotemporal coordination of cloud user loads of Data Center A and Data Center B taking into account the thermal dynamic characteristics of buildings satisfies the constraints in S21-S28. S32. Establish a data center deterministic mathematical model that does not consider uncertain factors, and the compact form of the data center deterministic mathematical model is as follows: Among them, x and y are variables, and x and y satisfy the following relationship: Where C is the number of servers in the data center, and is the coefficient column vector of formula (31); D, K, F, G and I u is the coefficient matrix under the corresponding constraints, d and k are constant column vectors, where the inequality constraints are the second row in (32) and satisfy the constraint relationships (17)-(19), (22)-(23), and (30), the equality constraints are the third row in (32) and satisfy the constraint relationships (15)-(16), (18), (20)-(21), and (28)-(29), the relationship between variables x and y is the fourth row and satisfies the constraint relationships (26)-(27), and the fifth row indicates that in the data center deterministic mathematical model, the uncertain factor variables are the predicted values ​​of each time period, and the predicted values ​​satisfy the following relationship: in, represents the predicted value of conventional load power in period t, represents the predicted value of outdoor temperature in period t; S33. Construct a box type uncertainty set U to which the uncertain factors conventional load and outdoor temperature belong, and U satisfies the following relationship: Among them, u P,t and u T,t are uncertain variables, namely, normal load power and outdoor temperature; and is the maximum fluctuation deviation allowed for the uncertain variable; S34. Determine the mathematical model based on the data center. Using the box-shaped uncertainty set U of the uncertain factors as a constraint, a two-stage robust optimization method is used to construct a mathematical model of the data center that takes into account the thermal dynamic characteristics of the building as follows: The first-stage problem is the outermost min problem, with the optimization variable x; the second-stage problem is the inner max-min problem, with the optimization variables u and y. The min problem is equivalent to the objective function of Equation (31), which represents the minimization of the daily operating cost of the data center. Ω(x,u) means the feasible domain of the optimization variable y when the value of (x,u) is given, and the expression of Ω(x,u) is as follows: Among them, γ, λ, v, and π are dual variables used to deal with the minimization problem in the second stage.

2. The differentiated demand coordination method based on a data center mathematical model according to claim 1, characterized in that: The S4 further includes the following steps: S41. Decompose the mathematical model of the data center taking into account the thermal dynamic characteristics of the building, which is constructed using the two-stage robust optimization method in S34, to obtain a main problem model and sub-problem models. The main problem model is as follows: Where k and l are the number of iterations, and l∈k, y l is the solution of the data center subproblem after the lth iteration; is the value of the uncertain factors normal load power and outdoor temperature for the first iteration; The sub-problem model is as follows: For the subproblem model (39) of the NP-hard problem, given (x,u), the inner minimization in the subproblem model (39) is transformed into a linear problem, and according to the duality theory, the inner min in the subproblem model (39) is transformed into the max form and merged with the outer max. The transformed subproblem model is as follows: For the bilinear term u in (40) T π, when the uncertainty factors of the data center, the conventional load and the outdoor temperature, are taken as the right boundary in formula (35), the required operating cost of the data center is high. Therefore, the uncertainty set U (35) to which the uncertainty factors, the conventional load and the outdoor temperature belong, is optimized as follows: Where B=[B P,t B T,t ] T is a binary variable, and the value 1 indicates that the normal load power and outdoor temperature take the maximum value of the uncertainty set; Γ p,t and Γ T,t are the uncertainty adjustment parameters corresponding to conventional load power and outdoor temperature, respectively. The smaller the Γ value, the more risky the model is. Therefore, Γ is used to adjust the conservatism of the scheme. When the uncertainty set of the uncertain factors in formula (41) is substituted into formula (40), the product of continuous variables and binary variables appears. Therefore, the large M method is used to linearize formula (40) to obtain the following expression: Where, Δu=(Δu P,t ,Δu T,t ) T ,B'=(B' P,t ,B' T,t ) T is a continuous auxiliary variable, is a positive real number.

3. The differentiated demand coordination method based on a data center mathematical model according to claim 2, characterized in that: The S4 further includes the following steps: S42. For the main problem model and sub-problem model obtained in S41, the G&GG algorithm is used to iteratively solve the problem to obtain the optimal differential demand coordination solution for the uncertain factors. The solution method includes the following steps: S421: Given a set of conventional load power and outdoor temperature as the worst scenario u, set the lower bound of the data center daily operating cost provided by the main problem LB = +∞, the upper bound provided by the subproblem UB = -∞, and the convergence threshold ε; S422: The worst scenario obtained based on the first iteration of the data center Solve the main problem equation (38) and get Update the Nether S423: Will be solved Substitute it into Equation (41) as a fixed value to obtain the objective function value of the sub-problem and the value of the uncertain variable u, and use the obtained objective function value as the new upper bound to update the upper bound S424: Determine whether the upper and lower bounds meet the convergence conditions. When UB-LB≤ε, the iteration ends and the optimal solution is obtained. Otherwise, the variable y is increased. k+1 And the following constraints: Let k = k + 1, and re-solve the main and sub-problems until the algorithm converges or reaches the maximum number of iterations.

Images