Energy consumption elastic space-based calculation center and power grid cooperative scheduling optimization model construction and solving method

Abstract

The invention discloses a computing power center and power grid collaborative scheduling optimization model construction and solving method based on an energy consumption elastic space, and the method comprises the following steps: 1) constructing a computing power center energy consumption optimization model, converting the computing power center energy consumption optimization model into a linear programming model which takes an energy consumption feasible region as an optimization space, and calculating a computing power center energy consumption elastic space; 2) constructing a computing power center and power grid collaborative scheduling optimization model; 3) solving the computing power center and power grid collaborative scheduling optimization model to obtain a computing power center energy consumption result; and 4) substituting the energy consumption result of the computing power center into the energy consumption optimization model of the computing power center to obtain equipment control parameters such as the starting number of IT equipment of the computing power center and the air supply temperature of the air conditioning system. According to the method, the calculation center energy consumption optimization model is constructed, the solved elastic space is embedded into the power grid unit combination scheduling model, collaborative scheduling optimization is completed, the calculation center can be effectively guided to participate in power system demand response, and a flexible load resource configuration decision in power system planning is supported.

Description

Technical Field

[0001] This invention relates to the field of power system planning and operation, specifically to a method for constructing and solving a collaborative scheduling optimization model of a computing center and power grid based on energy consumption elasticity space. Background Technology

[0002] With the rapid development of technologies such as cloud computing, the Internet of Things, 5G communication, and virtual reality, the demand for computing centers and cloud computing services continues to grow, and their operating energy consumption accounts for a significant proportion of total social electricity consumption. There are approximately 8 million computing centers globally, consuming about 460 terawatt-hours of electricity, accounting for 1.4% of global energy consumption. Computing centers have gradually become a major consumer of electricity in the power system.

[0003] As a core infrastructure of the digital economy, computing centers are characterized by large load capacity, strong regulation capabilities, and stable energy demand. They possess rapid and flexible load control capabilities and are valuable demand-side response resources in the power system. Computing centers have the potential for flexible load control; their tasks are divisible, highly mobile in time and space, and have a high load elasticity coefficient, enabling them to respond to peak-valley regulation needs of the power grid. The participation of computing centers in grid collaborative dispatch not only effectively reduces the cost of computing centers but also improves the operational stability and resource utilization of the power grid, facilitating the integration of new power systems with the digital economy. The collaborative dispatch mechanism provides flexible regulation resources for power grid operation, effectively smoothing grid load fluctuations, reducing peak loads, and thus reducing power grid investment costs and operational losses.

[0004] Accurately constructing a collaborative scheduling optimization model between computing centers and the power grid has become a key research challenge. In the field of energy consumption optimization management, existing research focuses on hardware upgrades, cooling technology innovation, and algorithm scheduling to drive the transformation of computing centers from energy consumption reduction to low-carbon and high-efficiency operation. Research on computing-power collaborative planning mainly revolves around spatial layout planning and infrastructure collaborative planning, optimizing power grid resource allocation through targeted absorption of new energy sources to achieve collaborative scheduling between computing centers and the power grid. In summary, current research is mostly concentrated on energy consumption optimization management within computing centers, computing-power collaborative planning, and optimized operation of integrated energy systems in computing centers. There is relatively little research on collaborative scheduling models between computing centers and the power grid, and a lack of a systemic collaborative perspective: either focusing on economic scheduling on the power grid side or limited to energy efficiency optimization on the data center side, there is a lack of models that place both under a unified framework for joint optimization, and the value of computing load as a flexible resource for bidirectional regulation cannot be fully explored. Summary of the Invention

[0005] The purpose of this invention is to provide a method for constructing and solving an optimization model for the coordinated scheduling of computing centers and power grids based on energy consumption elasticity space, comprising the following steps:

[0006] Step 1) constructing a computing center energy consumption optimization model;

[0007] Step 2) transforming the computing center energy consumption optimization model into a linear programming model with an energy consumption feasible region as the optimization space, and solving the linear programming model to calculate the computing center energy consumption elasticity space;

[0008] Step 3) based on the computing center energy consumption elasticity space, constructing a computing center and power grid collaborative scheduling optimization model;

[0009] Step 4) solving the computing center and power grid collaborative scheduling optimization model to obtain the computing center energy consumption result;

[0010] Step 5) substituting the computing center energy consumption result into the computing center energy consumption optimization model to obtain the control parameters of each energy-consuming device of the computing center.

[0011] Further, in step 1), the step of constructing the computing center energy consumption optimization model includes:

[0012] Step 1.1) constructing a computing center IT device power consumption model, an air conditioning system power consumption model, and a heat exchange coupling model of the IT device and the air conditioning system;

[0013] The computing center IT device power consumption model is as follows:

[0014] (1)

[0015] In the formula, is the average power consumption when the server is idle, is the average power consumption when the server is running at full load; is a constant; are the utilization rate and the chip temperature of the IT device i at time period t, respectively; is the operating state of the IT device i at time period t, taking values 2, 1, and 0 to represent power-on, standby, and off, respectively; is the power consumption of the IT device i in the power-on state; P sleep is the power consumption when the key energy-consuming device enters the standby state;

[0016] The air conditioning system power consumption model is as follows:

[0017] (2)

[0018] In the formula, P aircon , T out are the power consumption and the supply air temperature of the air conditioning system at time period t, respectively; P IT is the total power consumption of all IT devices in a specified floor area at time period t; k T , S dc are the environmental temperature coefficient and the required floor area of the IT device;

[0019] The heat exchange coupling model of the IT equipment and the air conditioning system is as follows:

[0020] (3)

[0021] (4)

[0022] In the formula, is a heat transfer coefficient; is the inlet air temperature of the IT equipment i at the time period t; is the chip temperature and outlet air temperature of the IT equipment i at the time period t; is an equivalent thermal resistance;

[0023] Step 1.2) model the task load received by the computing power center, that is:

[0024] (5)

[0025] In the formula, are respectively the initial load demand of the delay-sensitive and delay-tolerant tasks of the computing power center m at the time period t; are respectively the processed load of the delay-sensitive and delay-tolerant tasks of the computing power center; , is the initial load demand and the processed load of the computing power center m at the time period t;

[0026] wherein the relationship between the initial load demand of the delay-sensitive tasks and the initial load demand and the processed load of the delay-tolerant tasks of the computing power center is as follows:

[0027] (6)

[0028] (7)

[0029] (8)

[0030] (9)

[0031] (10)

[0032] In the formula, are respectively the time dimension and the space dimension load variation of the computing power center m at the time period t; are respectively the start time period and the end time period; is the total number of computing power centers; is the load amount of the computing power center m from the time period t to the time period ; is the maximum time period number that the task can be delayed; for time period from the computing center migrate to the computing center the load amount; for all possible computing center ordered pairs, that is: ;

[0033] Step 1.3) Based on the task load received by the computing center, a computing center energy consumption optimization model is constructed to minimize the total energy cost of all computing centers in the entire optimization period.

[0034] Further, the objective function of the computing center energy consumption optimization model is as follows:

[0035] (11)

[0036] In the formula, is the total power consumption of the computing center m in the time period t, is the total number of IT devices switched from the off state to the powered-on state, is the electricity price, is the start-up cost coefficient, is the time period.

[0037] Further, the constraints of the computing center energy consumption optimization model include computing center total power consumption constraints, computing center total computing task constraints, load migration limit constraints, air conditioning system supply air temperature constraints, IT device inlet air temperature constraints, and IT device chip temperature constraints.

[0038] Further, the computing center total power consumption constraints are as follows:

[0039] (12)

[0040] In the formula, are the total power consumption of all IT devices and the air conditioning system power consumption of the computing center m in the time period t, is the power consumption of IT device i of computing center m in time period t, is the number of IT devices of computing center m;

[0041] The computing center total computing task constraints are as follows:

[0042] (13)

[0043] In the formula, are the utilization rates of IT device i of computing center m in time period t, is the maximum processing load of IT device i of computing center m, is the number of IT devices of computing center m;

[0044] The load migration restriction constraint is shown as follows:

[0045] (14)

[0046] The air supply temperature constraint of the air conditioning system is shown as follows:

[0047] (15)

[0048] wherein, is the air supply temperature of the IT device i of the computing power center m at the time period t, are respectively the lower limit and the upper limit of the air supply temperature;

[0049] The IT device air inlet temperature constraint is shown as follows:

[0050] (16)

[0051] wherein, is the air inlet temperature of the IT device i of the computing power center m at the time period t, are respectively the lower limit and the upper limit of the air inlet temperature;

[0052] The IT device chip temperature constraint is shown as follows:

[0053] (17)

[0054] wherein, is the chip temperature of the IT device i of the computing power center m at the time period t, is the upper limit of the chip temperature.

[0055] Further, in step 2), the step of calculating the computing power center energy consumption flexibility space comprises:

[0056] Step 2.1) converts the representation of the energy consumption flexibility space into a feasibility analysis problem, that is:

[0057] (18)

[0058] (19)

[0059] wherein, w is an energy consumption variable vector, x is a decision variable vector, A and B are coefficient matrices, and C and D are coefficient matrices; the energy consumption variable w is a planning parameter;

[0060] Step 2.2) defines an optimal partition set and a critical region;

[0061] wherein, any subset of the optimal partition set is shown as follows:

[0062] (20)

[0063] where the active constraints are as follows:

[0064] (21)

[0065] The inactive constraints are as follows:

[0066] (22)

[0067] where: , and ; A, B and D represent the block sub-matrices of A, B and D corresponding to the constraint row number C;

[0068] The critical region CR with respect to a given parameter w0∈W is defined as follows: ξ0

[0069] (23)

[0070] Step 2.3) The space is divided according to the active constraint set and the inactive constraint set of the optimization problem, and the parameter space with the same effective constraint set is merged into a critical region;

[0071] where any critical region CR i is a convex space. Any critical region CR i is mutually exclusive with CR j , and there is no gap between the two adjacent critical regions; i, j = 1, 2, …, m, i≠j;

[0072] Step 2.4) The energy consumption elasticity space E is obtained by solving the optimal segmentation and the critical region.

[0073] Further, in step 3), the objective function of the computing power center and the power grid collaborative scheduling optimization model is as follows:

[0074] (24)

[0075] where the continuous variables represent the output, start-up and shut-down cost of the unit at time ; represents the cost of the unit output; and represent the set of units and scheduling periods. The physical meaning of the objective function is to minimize the unit output and start-up and shut-down cost coefficient.

[0076] ​​Furthermore, in step 3), the constraints of the computing center and power grid collaborative scheduling optimization model include the energy consumption elasticity space feasible region constraint and the power grid operation security constraint.

[0077] Power grid operation safety constraints include upper and lower limits of unit output, unit ramping constraints, minimum start-up and shutdown time limits of units, start-up and shutdown cost constraints, power balance constraints, and branch power flow constraints.

[0078] Furthermore, the feasible region constraint of the energy consumption elasticity space of the computing center is as follows:

[0079] (25)

[0080] In the formula, HA is the normal vector matrix of the hyperplane, and HB is the constant term vector of the hyperplane; P DC = {P DC,1 , P DC,2 ,…,P DC,m} represents the power consumption vector of each computing center;

[0081] The upper and lower limits of the unit's output are constrained as follows:

[0082] (26)

[0083] In the formula, For the unit At any moment The start / stop status, a value of 1 indicates the unit's start / stop status. At any moment When the unit is in the "on" state, a value of 0 indicates that the unit is in the "on" state. At any moment The device is currently powered off. and For the unit Upper and lower limits of output; These are 0-1 discrete variables in the unit combination model, which is modeled using a univariate approach.

[0084] The unit ramp-up constraints are as follows:

[0085] (27)

[0086] In the formula, and This is the unit's ramp rate. This constraint limits the output increment between adjacent time periods under different start-up and shutdown states of the unit.

[0087] The minimum start-up and shutdown time limits for the generating units are as follows:

[0088] (28)

[0089] In the formula, and denotes the minimum on or off time of the unit. and denotes the on or off time of the unit at time . This constraint limits the minimum time of the unit on or off.

[0090] The unit start-up and shut-down cost constraint is shown as follows:

[0091] (29)

[0092] wherein, and denote the on and off cost of the unit ;

[0093] The power balance constraint is shown as follows:

[0094] (30)

[0095] wherein, is the injection power offset of node i, is the active power output of the unit at time , is the active load demand of node at time ; is the node admittance matrix item; is the voltage angle of the jth node at time period t; is the set of power generation units located at node i; is the set of computing power centers located at node i;

[0096] The branch power flow constraint is shown as follows:

[0097] (31)

[0098] wherein, are the upper and lower limits of the branch power flow respectively, is the node associated admittance of the branch, is the flow direction admittance of line k and node i; is the additional active injection power of the branch, is the injection correction of line k.

[0099] Further, in step 4), the branch and bound method is used to solve the constructed computing power center and power grid collaborative scheduling optimization model, and the optimal solution of the energy consumption of each computing power center after the "electricity-computing" collaborative optimization can be obtained. ​​

[0100] Further, in step 5), the optimal solution of the energy consumption of each computing power center is substituted into the computing power center energy consumption optimization model constructed in step 1), and the score bounding method is used to solve the model to obtain the control parameters of each energy consumption device of the computing power center, including the number of IT device start-ups and the air conditioning system supply air temperature.

[0101] The technical effects of the present application are self-evident, and the beneficial effects of the present application are as follows:

[0102] 1) The present application proposes a computing power center and power grid collaborative scheduling optimization model. By constructing a computing power center energy consumption optimization model, the spatiotemporal migration characteristics of delay-tolerant loads are taken into account, and the energy consumption elasticity space of multiple computing power centers is calculated. The energy consumption elasticity space of multiple computing power centers is embedded in the power grid unit commitment model, and the power grid operation safety constraints are taken into account, including unit output upper and lower limit constraints, unit climbing constraints, unit minimum start-stop time limit constraints, unit start-stop cost constraints, power balance constraints, branch power flow constraints, and other power grid operation safety constraints. The computing power center power consumption operation results under the minimum target of power grid unit output and start-stop cost are optimized and solved. Thus, the computing power center and the power grid are effectively placed in a unified framework for joint optimization.

[0103] 2) The present application proposes a computing power center power consumption instruction analysis and control execution method. For the computing power center power consumption operation results obtained by solving the computing power center and power grid collaborative scheduling optimization model, the power grid side power instruction can be decomposed and fed back to the computing power center device control layer through the computing power center energy consumption optimization model constructed by the present application. This method can generate a set of refined control instructions, including IT device cluster start-stop strategy, refrigeration system supply air temperature, and load scheduling scheme, thereby achieving accurate tracking of power trajectory and collaborative optimization of system energy efficiency under the premise of ensuring service quality and equipment safety.

[0104] 3) The method proposed by the present application can achieve the dual benefits of computing power center energy efficiency improvement and power system flexibility enhancement, and has wide engineering application value. Through simulation verification of real operation data of large Internet companies, the method can accurately incorporate the multi-time scale energy consumption elasticity space of the computing power center as a boundary condition into the power grid safety constraint unit commitment model, effectively promote the source-load collaborative optimization under the premise of ensuring the quality of computing power services and the safe and stable operation of the power grid.

[0105] The present application provides a computing power center and power grid collaborative optimization scheduling method based on energy consumption elasticity space, which can be widely applied to power dispatching institution day-ahead optimization decision, computing power center smart energy management, and new type of power system flexibility resource allocation evaluation, etc. It provides a quantifiable and feasible collaborative scheduling solution for safe and economic operation of power system and green and low-carbon transformation of computing power industry, thereby supporting the collaborative promotion of "double carbon" target and energy and power safety at the technical level. Attached Figure Description

[0106] Figure 1 The initial workload demand curves for the three computing centers;

[0107] Figure 2 The load curves for some nodes (nodes 3, 4, 8, 20, 39) of the power grid over a 12-hour period are shown.

[0108] Figure 3 The power consumption command results are calculated by the three computing centers through a collaborative scheduling model.

[0109] Figure 4 The number of IT devices in the computing center that are activated after the power consumption command of the computing center is fed back to the device level.

[0110] Figure 5 The air supply temperature of the air conditioning system after the power consumption command of the computing center is fed back to the device level. Detailed Implementation

[0111] The present invention will be further described below with reference to embodiments, but it should not be construed that the scope of the present invention is limited to the following embodiments. Various substitutions and modifications made based on ordinary technical knowledge and common practices in the art without departing from the above-described technical concept of the present invention should be included within the scope of protection of the present invention.

[0112] Example 1:

[0113] See Figures 1 to 5 The method for constructing and solving a collaborative scheduling optimization model between computing centers and power grids based on energy consumption elasticity space includes the following steps:

[0114] Step 1) Construct an energy consumption optimization model for the computing center;

[0115] Step 2) Transform the computing center energy consumption optimization model into a linear programming model with the energy consumption feasible region as the optimization space, solve the linear programming model, and calculate the computing center energy consumption elastic space.

[0116] Step 3) Based on the energy consumption elasticity space of the computing center, construct an optimization model for the coordinated scheduling of the computing center and the power grid;

[0117] Step 4) Solve the optimization model for the coordinated scheduling of the computing center and the power grid to obtain the energy consumption results of the computing center;

[0118] Step 5) Substitute the energy consumption results of the computing center into the energy consumption optimization model of the computing center to obtain the control parameters of each energy-consuming device in the computing center.

[0119] Example 2:

[0120] The method for constructing and solving the computing power center and power grid collaborative scheduling optimization model based on energy consumption elastic space, the technical content is the same as embodiment 1, further, in step 1), the step of constructing the energy consumption optimization model of the computing power center includes:

[0121] Step 1.1) Constructing an IT equipment power consumption model, an air conditioning system power consumption model, and a heat exchange coupling model of the IT equipment and the air conditioning system;

[0122] The IT equipment power consumption model of the computing power center is as follows:

[0123] (1)

[0124] In the formula, is the average power consumption when the server is idle, is the average power consumption when the server is running at full load; is a constant; are the utilization rate and the chip temperature of the IT equipment i at period t, respectively; is the running state of the IT equipment i at period t, and the values 2, 1, and 0 represent power-on, standby, and off, respectively; is the power consumption of the IT equipment i in the power-on state; P sleep is the power consumption when the key energy-consuming equipment enters the standby state; is the power consumption of the key energy-consuming equipment;

[0125] The air conditioning system power consumption model is as follows:

[0126] (2)

[0127] In the formula, P aircon , T out are the power consumption and the supply air temperature of the air conditioning system at period t, respectively; P IT is the total power consumption of all IT equipment in a specified floor area at period t; k T , S dc are the environmental temperature coefficient and the required floor area of the IT equipment;

[0128] The heat exchange coupling model of the IT equipment and the air conditioning system is as follows:

[0129] (3)

[0130] (4)

[0131] In the formula, is the heat transfer coefficient; is the inlet air temperature of the IT equipment i at period t; is the chip temperature and the outlet air temperature of the IT equipment i at period t; is the equivalent thermal resistance;

[0132] Step 1.2) Model the task load received by the computing power center, that is:

[0133] (5)

[0134] wherein, respectively, are the initial load requirements of the delay-sensitive and delay-tolerant tasks of the computing power center m at time period t; respectively, are the processed loads of the delay-sensitive and delay-tolerant tasks of the computing power center; , is the initial load requirement and the processed load of the computing power center m at time period t;

[0135] wherein, the relationship between the initial load requirement of the delay-sensitive tasks and the initial load requirement and the processed load of the delay-tolerant tasks of the computing power center is as follows:

[0136] (6)

[0137] (7)

[0138] (8)

[0139] (9)

[0140] (10)

[0141] wherein, respectively, are the time dimension and space dimension load changes of the computing power center m at time period t; respectively, are the start time period and the end time period; is the total number of computing power centers; is the load amount of the computing power center m from time period t to time period ; is the maximum time period number that the task can be delayed; is the time period from the computing power center to the computing power center ; is the set of all possible ordered pairs of computing power centers, that is: ;

[0142] Step 1.3) Based on the task load received by the computing power center, an energy consumption optimization model of the computing power center is constructed, with the minimum total energy cost of all computing power centers in the entire optimization period as the target.

[0143] Embodiment 3:

[0144] The power center and power grid collaborative scheduling optimization model construction and solution method based on energy consumption elasticity space, the technical content is the same as any one of embodiments 1-2, further, the objective function of the power center energy consumption optimization model is as follows:

[0145] (11)

[0146] In the formula, is the total power consumption of the power center m in the time period t, is the total number of IT devices switched from the off state to the powered-on state, is the electricity price, is the start-up cost coefficient, is the time period length.

[0147] Embodiment 4:

[0148] The power center and power grid collaborative scheduling optimization model construction and solution method based on energy consumption elasticity space, the technical content is the same as any one of embodiments 1-3, further, the constraint conditions of the power center energy consumption optimization model include the total power consumption constraint of the power center, the total computing task constraint of the power center, the load migration limit constraint, the air conditioning system supply air temperature constraint, the IT device inlet air temperature constraint, and the IT device chip temperature constraint.

[0149] Embodiment 5:

[0150] The power center and power grid collaborative scheduling optimization model construction and solution method based on energy consumption elasticity space, the technical content is the same as any one of embodiments 1-4, further, the total power consumption constraint of the power center is as follows:

[0151] (12)

[0152] In the formula, are respectively the total power consumption of all IT devices of the power center m in the time period t and the air conditioning system power consumption, is the power consumption of the IT device i of the power center m in the time period t, is the number of IT devices of the power center m;

[0153] The total computing task constraint of the power center is as follows:

[0154] (13)

[0155] In the formula, are respectively the utilization rate of the IT device i of the power center m in the time period t, is the maximum processing load of the IT device i of the power center m, is the number of IT devices of the power center m;

[0156] The load migration limit constraint is as follows:

[0157] (14)

[0158] wherein, is the time-delay tolerant task load amount;

[0159] The air supply temperature constraint of the air conditioning system is as follows:

[0160] (15)

[0161] wherein, is the air supply temperature of the IT device i of the computing power center m at the time period t, are respectively the lower limit and the upper limit of the air supply temperature;

[0162] The IT device air inlet temperature constraint is as follows:

[0163] (16)

[0164] wherein, is the air inlet temperature of the IT device i of the computing power center m at the time period t, are respectively the lower limit and the upper limit of the air inlet temperature;

[0165] The IT device chip temperature constraint is as follows:

[0166] (17)

[0167] wherein, is the chip temperature of the IT device i of the computing power center m at the time period t, is the upper limit of the chip temperature.

[0168] Embodiment 6:

[0169] The method for constructing and solving the computing power center and power grid collaborative scheduling optimization model based on the energy consumption elasticity space, the technical content of which is the same as any one of embodiments 1-5, further, in step 2), the step of calculating the energy consumption elasticity space of the computing power center comprises:

[0170] Step 2.1) transform the representation of the energy consumption elasticity space into a feasibility analysis problem, that is:

[0171] (18)

[0172] (19)

[0173] wherein, w is the energy consumption variable vector, x is the decision variable vector, A, B are coefficient matrices, C, D are coefficient matrices; the energy consumption variable w is a planning parameter;

[0174] Step 2.2) defining the optimal partition set and critical region;

[0175] wherein the optimal partition set of any subset is shown as follows:

[0176] (20)

[0177] wherein the active constraints are shown as follows:

[0178] (21)

[0179] The inactive constraints are shown as follows:

[0180] (22)

[0181] In the formula: , and ; A, B and D represent the block sub-matrix of A, B and D corresponding to the constraint row number C;

[0182] The critical region CR ξ0 with respect to a given parameter w0∈W is defined as follows:

[0183] (23)

[0184] Step 2.3) dividing the space according to the active constraint set and the inactive constraint set of the optimization problem, and merging the parameter space with the same effective constraint set into a critical region;

[0185] wherein any critical region CR i is a convex space. Any critical region CR i is mutually exclusive with CR j , and there is no gap between the two adjacent critical regions; i, j = 1, 2, …, m, i≠j;

[0186] Step 2.4) solving by optimal partition and critical region to obtain the energy elasticity space .

[0187] Example 7:

[0188] The technical content of the computing power center and power grid collaborative scheduling optimization model construction and solving method based on the energy elasticity space is the same as any one of embodiments 1-6, and further, in step 3), the objective function of the computing power center and power grid collaborative scheduling optimization model is shown as follows:

[0189] (24)

[0190] In the formula: the continuous variables respectively represent the unit at time output, start-up and shut-down costs; representing the unit output cost; and representing the set of units and scheduling periods. The objective function has the physical meaning of minimizing the unit output and start-up and shut-down costs coefficients.

[0191] Embodiment 8:

[0192] The method for constructing and solving the computing power center and power grid collaborative scheduling optimization model based on energy consumption flexibility space, the technical content of which is the same as any one of embodiments 1-7, further, in step 3), the constraint conditions of the computing power center and power grid collaborative scheduling optimization model include energy consumption flexibility space feasible region constraints and power grid operation safety constraints.

[0193] The power grid operation safety constraints include unit output upper and lower limit constraints, unit ramping constraints, unit minimum start-up and shut-down time limit constraints, unit start-up and shut-down cost constraints, power balance constraints, branch power flow constraints.

[0194] Embodiment 9:

[0195] The method for constructing and solving the computing power center and power grid collaborative scheduling optimization model based on energy consumption flexibility space, the technical content of which is the same as any one of embodiments 1-8, further, the energy consumption flexibility space feasible region constraints of the computing power center are as follows:

[0196] (25)

[0197] In the formula, HA is the normal vector matrix of the hyperplane, and HB is the constant item vector of the hyperplane; P DC = {P DC,1 , P DC,2 ,…,P DC,m} is the power consumption vector of each computing power center;

[0198] The unit output upper and lower limit constraints are as follows:

[0199] (26)

[0200] In the formula, is the start-up and shut-down state of the unit at time , and the value of 1 indicates that the unit is in the start-up state at time , and the value of 0 indicates that the unit is in the shut-down state at time ; and are the upper and lower limits of the unit upper and lower limits of the output; 0-1 discrete variables in the unit commitment model which is modeled in a univariate way;

[0201] The unit ramping constraint is shown as follows:

[0202] (27)

[0203] where, and is the unit ramping rate. This constraint limits the output increment between adjacent time periods for different start-stop states of the unit. is the start-stop state change of the unit

[0204] The unit minimum start-stop time limit constraint is shown as follows:

[0205] (28)

[0206] where, and is the minimum continuous start-up or shutdown time of the unit and is the continuous start-up or shutdown time of the unit at time . This constraint limits the minimum time of continuous start-up or shutdown of the unit.

[0207] The unit start-stop cost constraint is shown as follows:

[0208] (29)

[0209] where, and is the start-up and shutdown cost of the unit

[0210] The power balance constraint is shown as follows:

[0211] (30)

[0212] where, is the injection power offset of node i, is the active power output of the unit at time , is the active load demand of node at time ; is the node admittance matrix item; is the voltage angle of the jth node at time period t; is the set of generators located at node i;​​​ a set of computing power centers located at node i;

[0213] The branch power flow constraint is shown as follows:

[0214] (31)

[0215] wherein, are the upper and lower limits of the branch power flow, respectively, is the node associated susceptance of the branch, is the flow direction susceptance of line k and node i; is the additional active power injection of the branch, is the injection correction of line k.

[0216] Embodiment 10:

[0217] The method for constructing and solving the computing power center and power grid collaborative scheduling optimization model based on the energy consumption flexibility space, the technical content of which is the same as any one of embodiments 1-9, further, the branch and bound method is used to solve the constructed computing power center and power grid collaborative scheduling optimization model, and the optimal solution of the energy consumption of each computing power center after the "electricity-computing" collaborative optimization can be obtained.

[0218] Embodiment 11:

[0219] The method for constructing and solving the computing power center and power grid collaborative scheduling optimization model based on the energy consumption flexibility space, the technical content of which is the same as any one of embodiments 1-10, further, the optimal solution of the energy consumption of each computing power center is substituted into the computing power center energy consumption optimization model constructed in step 1), and the score bound method is used to solve the model to obtain the control parameters of each energy consumption device of the computing power center, including the number of IT equipment turned on and the air conditioning system supply air temperature.

[0220] Embodiment 12:

[0221] The method for constructing and solving the computing power center and power grid collaborative scheduling optimization model based on the energy consumption flexibility space, comprising the following steps:

[0222] 1) Construct a computing power center energy consumption optimization model;

[0223] 2) Convert the computing power center energy consumption optimization model into a linear programming model with the energy consumption feasible region as the optimization space, calculate the computing power center energy consumption flexibility space based on the multi-parameter programming theory;

[0224] 3) Based on the computing power center energy consumption flexibility space in 2), construct a computing power center and power grid collaborative scheduling optimization model;

[0225] 4) Solve the computing power center and power grid collaborative scheduling optimization model to obtain the computing power center energy consumption result;

[0226] 5) Substitute the computing center energy consumption result of 4) into the computing center energy consumption optimization model to obtain the computing center IT device start-up quantity, air conditioning system supply air temperature and other device control parameters.

[0227] In step 1), first, the key energy-consuming equipment model is constructed:

[0228] The IT device power consumption model is constructed. The power consumption of IT device i at time period t can be expressed as:

[0229] (1)

[0230] In the formula, is the average power consumption when the server is idle, is the average power consumption when the server is running at full load, when the IT device enters the standby state, it will automatically reduce the working voltage and frequency to save energy, at this time the power consumption is usually a small constant value, denoted as P sleep ; are the utilization rate and chip temperature of IT device i at time period t, respectively; is the running state of IT device i at time period t, taking values 2, 1, 0 respectively representing power-on, standby, and off; is the power consumption of IT device i in the power-on state; is a constant.

[0231] The air conditioning system power consumption model is constructed. The power consumption of the air conditioning system at time period t can be expressed as:

[0232] (2)

[0233] In the formula, k T is the environmental temperature coefficient, S dc is the required floor area of the IT device; T out is the power consumption and supply air temperature of the air conditioning system at time period t; T IT is the total power consumption of all IT devices in a specified floor area at time period t.

[0234] The heat exchange coupling model of IT device and air conditioning system is constructed. The heat exchange coupling model of IT device and air conditioning system is constructed. In the forced air cooling link inside the IT device, the heat generated by the electronic components inside the IT device is carried away by air convection heat transfer:

[0235] (3)

[0236] In the formula, is the inlet air temperature of IT device i at time period t; is the chip temperature and outlet air temperature of IT device i at time period t; is the equivalent thermal resistance, which is related to air flow density, volume flow and specific heat capacity.

[0237] In the hot air recycling and cooling regeneration link, the hot air discharged by the IT equipment is recycled by the air conditioning system, and after cooling, it is converted into cold air for recycling:

[0238] (4)

[0239] In the formula, is the heat transfer coefficient, which depends on the equipment layout, and can be approximately constant in the case of uniform equipment deployment.

[0240] Secondly, under the constraints of ensuring the quality of computing services and the safety of IT equipment operation, the flexibility modeling of the computing center is carried out:

[0241] The initial load demand of computing center m at time period t And the processed load Is:

[0242] (5)

[0243] In the formula, are the initial load demand of delay-sensitive and delay-tolerant tasks of computing center m at time period t respectively. are the processed load of delay-sensitive and delay-tolerant tasks of computing center respectively.

[0244] The initial load demand of delay-sensitive tasks needs to be completely processed within the same time period:

[0245] (6)

[0246] The relationship between the initial load demand of delay-tolerant tasks and the processed load of the same time period:

[0247] (7)

[0248] In the formula, are the time dimension and space dimension load changes of computing center m at time period t respectively.

[0249] The total initial load demand before space-time migration must be equal to the total processed load after migration:

[0250] (8)

[0251] In the formula, are the starting time period and the ending time period respectively; is the total number of computing centers.

[0252] Delay-tolerant tasks can use time dimension flexibility to delay execution from earlier time periods to later time periods:

[0253] (9)

[0254] where, is the load amount of the computing center m migrated from time period t to time period . is the maximum time period that the task can be delayed, determined by the task level protocol and scheduling requirements.

[0255] In the spatial dimension, assuming that the load can be migrated between computing centers instantaneously, the network transmission time is ignored:

[0256] (10)

[0257] where, is the load amount of the computing center migrated from computing center to computing center ; is the set of all possible ordered pairs of computing centers, i.e.: .

[0258] An energy consumption optimization model is constructed, which aims to minimize the total operating cost of all computing centers in the entire optimization period. The objective function is composed of operating cost and startup cost:

[0259] (11)

[0260] where, is the total power consumption of computing center m at time period t, is the total number of IT devices switched from the off state to the powered-on state, is the electricity price, is the startup cost coefficient, is the time period length.

[0261] The constraints include computing center operation constraints and regional power grid load migration constraints. The total power consumption constraint of the computing center is as follows:

[0262] (12)

[0263] where, are the total power consumption of all IT devices and air conditioning system of computing center m at time period t, respectively, is the power consumption of IT device i of computing center m at time period t, is the number of IT devices of computing center m.

[0264] The total computing task constraint of the computing center is as follows:

[0265] (13)

[0266] wherein, is the utilization rate of IT equipment i of the computing center m at time period t, is the maximum processing load of IT equipment i of the computing center m at time period t, is the number of IT equipment of the computing center m.

[0267] The load migration restriction constraint is as follows:

[0268] (14)

[0269] The air conditioning system supply air temperature constraint is as follows:

[0270] (15)

[0271] wherein, is the supply air temperature of IT equipment i of the computing center m at time period t, are the lower limit and upper limit of the supply air temperature, respectively.

[0272] The IT equipment inlet air temperature constraint is as follows:

[0273] (16)

[0274] wherein, is the inlet air temperature of IT equipment i of the computing center m at time period t, are the lower limit and upper limit of the inlet air temperature, respectively.

[0275] The IT equipment chip temperature constraint is as follows:

[0276] (17)

[0277] wherein, is the chip temperature of IT equipment i of the computing center m at time period t, is the upper limit of the chip temperature.

[0278] To ensure the computing efficiency, and are linearized from the original quadratic expression, and finally form a mixed integer linear programming model.

[0279] In step 2), for the energy consumption optimization model, it is converted into a linear programming model with the energy consumption feasible region as the optimization space. Based on the multi-parameter programming theory, the energy consumption elasticity space capable of representing the power supply boundary required by the regional computing center is calculated, and the steps are as follows:

[0280] First, the energy consumption optimization model is converted into a feasibility analysis problem in the following form to determine the feasible energy consumption value:

[0281] (18)

[0282] (19)

[0283] where, is the energy consumption variable vector, is the decision variable vector, is the coefficient matrix, is the coefficient matrix. The energy consumption variable is considered as a planning parameter, the model is transformed into a typical multi-parameter linear programming problem. A, B, C, D are determined by the computing power center energy consumption model.

[0284] Secondly, the theoretical knowledge of optimal partition and critical region is clarified.

[0285] Concept 1: Optimal Partition

[0286] Let L be the set of constraint numbers of formula (19). For any subset , A, B and D represent the block sub-matrix of A, B and D corresponding to the constraint row number C. The active constraints and inactive constraints in formula (19) can be expressed as:

[0287] Active constraints:

[0288] (20)

[0289] Inactive constraints:

[0290] (21)

[0291] That is: , and . For the parameter w in the feasible space W, an optimal partition (ξ(w), ξ c (w)) about the number set C is defined as:

[0292] (22)

[0293] Concept 2: Critical Region

[0294] Given the parameter w0∈W, define , then the critical region CR ξ0 about w0 is defined as:

[0295] (23)

[0296] The concept of optimal partitioning provides a standard for optimally partitioning the parameter space: dividing the space according to the effective and ineffective constraint sets of the optimization problem. The concept of critical region merges parameter spaces with the same effective constraint set into a critical region. Therefore, by solving through optimal partitioning and critical regions, the parameter feasible space W, i.e., the energy consumption elasticity space, can be obtained.

[0297] In step 3), the energy consumption elasticity space calculated based on multi-parameter programming theory is effectively embedded into the grid unit combination scheduling model. The goal is to minimize the output and start-up / shutdown costs of the grid units, with the feasible domain equation of the energy consumption elasticity space and the grid operation safety constraints as constraints, thereby realizing the collaborative scheduling between the computing center and the grid.

[0298] The optimization model for collaborative scheduling between computing centers and power grids takes minimizing the output and start-up / shutdown costs of power grid units as its objective function.

[0299] (twenty four)

[0300] Where: continuous variables They represent the generating units. At any moment The costs of power output, startup, and shutdown; Indicates the unit The cost of labor; and This represents the set of generating units and scheduling periods. The objective function is... The physical meaning is: to minimize the unit output and start-up / shutdown cost coefficient.

[0301] The feasible domain of energy consumption elasticity space and grid operation safety constraints are determined as constraints. The grid operation safety constraints include upper and lower limits of unit output, unit ramping constraints, minimum start-up and shutdown time limits of units, start-up and shutdown cost constraints, power balance constraints, and branch power flow constraints.

[0302] Energy consumption elasticity space feasible region constraint for computing centers:

[0303] The energy consumption elastic space of the computing center, as a convex polyhedron, can be completely characterized by a set of boundary hyperplanes. The hyperplane equation is expressed as:

[0304] (25)

[0305] In the formula, HA is the normal vector matrix of the hyperplane, and HB is the constant term vector of the hyperplane; together, they describe the boundary of the feasible region; P DC = {P DC,1 , P DC,2 ,…,P DC,m} represents the power consumption vector of each computing center.

[0306] Unit output upper and lower limit constraints:

[0307] (26)

[0308] In the formula, For the unit At any moment The start / stop status, with a value of 1 indicating the unit's start / stop status. At any moment When the unit is in the "on" state, a value of 0 indicates that the unit is in the "on" state. At any moment The device is currently powered off. and For the unit The upper and lower limits of power output. This constraint limits the unit's output; for thermal power units, The output space is discontinuous, which leads to the discreteness of the unit combination problem. It is the only discrete variable in the unit combination model modeled in a single-variable manner. Its properties are 0-1 variables, and it can be called a binary variable.

[0309] Unit ramp-up constraints:

[0310] (27)

[0311] In the formula, and This is the unit's ramp rate. This constraint limits the output increment between adjacent time periods under different start-up and shutdown states of the unit.

[0312] Minimum start-up and shutdown time constraints for generating units:

[0313] (28)

[0314] In the formula, and Indicates the unit The minimum continuous on or off time. and Indicates the unit At any moment The constraint limits the minimum continuous start-up or shutdown time of the unit.

[0315] Unit start-up and shutdown cost constraints:

[0316] (29)

[0317] In the formula, and Indicates the unit The start-up and shutdown costs. This constraint defines the unit start-up and shutdown costs, which can be linearly represented by the unit's start-up and shutdown states.

[0318] Power balance constraints:

[0319] (30)

[0320] Equation (35) represents the power balance equation for node i, where: The injected power offset for node i. For the unit At any moment Contributing to the cause at the time For nodes At any moment The active power load demand; For nodal admittance matrix terms; Let be the voltage angle of the j-th node during time period t; Let i be the set of generators located at node i; The set of computing power centers located at node i.

[0321] Branch flow constraints:

[0322] (31)

[0323] In the formula, These are the upper and lower limits of the branch power flow, respectively. Associating susceptance with the nodes of the branch circuit. Let k be the flow admittance between line k and node i; To inject additional active power into the branch circuit, This is an injection correction for line k. This is the voltage angle.

[0324] In step 4), the optimization model for coordinated scheduling between the computing center and the power grid is solved. Since this model is a mixed-integer linear programming problem, the branch and bound method is used for solving it.

[0325] Branch and bound is a core algorithm in commercial solvers such as GUROBI and CPLEX. After model preprocessing, the optimal solution to a mixed-integer linear programming problem is searched using the branch and bound process. Essentially, the branch and bound process is a strategic enumeration of combinations of discrete variables. Taking a mixed-integer linear programming problem with a minimum objective function as an example, in the branch and bound search tree, each node represents a different combination of discrete variable values. When a node in the search tree is assigned a new discrete variable with a discrete value, a new node will branch out. The branch and bound process solves a relaxation model at each node, that is, it solves a relaxation model that relaxes the unassigned discrete variables into continuous variables. If all discrete variables in the obtained solution are integer values, then the solution is a feasible solution to the optimization problem, and its corresponding objective function value is an upper bound. If there are discrete variables with non-integer values ​​in the obtained solution, then the solution is called a relaxation solution, and its corresponding objective function value is called a lower bound. The branch and bound process updates the obtained upper and lower bounds and reduces the gap between them by continuously branching and relaxing the solution. When the solution gap converges to 0, that is, when the upper and lower bounds coincide, it will be impossible to find a feasible solution with the target value better than the current lower bound. The feasible solution corresponding to the current upper bound is the optimal solution.

[0326] By employing the branch-and-bound method to solve the collaborative scheduling optimization model between computing centers and the power grid, the optimal solution P for the power consumption of each computing center after collaborative optimization can be obtained. DC = {P DC,1 , P DC,2 ,…,P DC,m}

[0327] In step 5), the power consumption of each computing center obtained from solving the collaborative scheduling optimization model is substituted into the computing center energy consumption optimization model constructed in step 1). That is, in the computing center power consumption P DC = {P DC,1 , P DC,2 ,…,P DC,m Given a fixed set of conditions, by jointly optimizing the operating status of key energy-consuming equipment (IT equipment, air conditioning system) and the migration strategy of computing tasks, we can optimize the equipment control parameters such as the number of IT equipment to be turned on in the computing center and the air supply temperature of the air conditioning system under the condition of the lowest IT equipment startup cost.

[0328] Example 13:

[0329] The verification of the optimization model for collaborative scheduling between computing centers and power grids based on energy consumption elasticity space is as follows:

[0330] (1) Construct an energy consumption optimization model for the computing center and solve for the energy consumption elasticity space.

[0331] First, an energy consumption optimization model for computing centers is constructed, with the goal of minimizing the total operating cost of all computing centers throughout the entire optimization cycle (Equation (11)). The constraints are the total power consumption of computing centers (Equation (12)), the total computing tasks of computing centers (Equation (13)), the load migration limit constraint (Equation (14)), and the temperature constraint (Equations (15)–(17)). Second, it is transformed into a linear programming model with the energy consumption feasible region as the optimization space. Based on the multi-parameter programming theory, the energy consumption elasticity space that can characterize the power supply boundary required by the regional computing centers is calculated.

[0332] (2) Construct a collaborative scheduling model between computing centers and power grids

[0333] The energy consumption elastic space calculated based on the multi-parameter programming theory is effectively embedded into the grid unit combination scheduling model. The goal is to minimize the power output and start-up and shutdown costs of the grid units (Equation (24)). The feasible domain equation of the energy consumption elastic space (Equation (25)) and the grid operation safety constraints are used as constraints (Equations (26)-(31)), thereby realizing the collaborative scheduling between the computing center and the grid.

[0334] (3) Solve the collaborative scheduling model between computing center and power grid

[0335] The constructed optimization model for the coordinated scheduling of computing centers and power grids is a mixed-integer linear programming model, which is solved using the branch and bound method.

[0336] (4) Power consumption instruction parsing and control execution method of computing center

[0337] The power consumption of the computing center obtained from solving the collaborative scheduling optimization model is substituted into the power consumption optimization model of the computing center. That is, when the power consumption P of the computing center... DC = {P DC,1 , P DC,2 ,…,P DC,m Given a fixed set of conditions, by jointly optimizing the operating status of key energy-consuming equipment (IT equipment, air conditioning system) and the migration strategy of computing tasks, we can optimize the equipment control parameters such as the number of IT equipment to be turned on in the computing center and the air supply temperature of the air conditioning system under the condition of the lowest IT equipment startup cost.

[0338] Specific simulation results

[0339] To verify the effectiveness of the proposed optimization for coordinated scheduling between computing centers and the power grid, this study uses real operational data from three computing centers of a large internet company for simulation. The power grid selected is the IEEE 39-node standard network. The initial workload requirements and operating parameters of the three computing centers are as follows: Figure 3 As shown in Table 1.

[0340] Table 1 Equipment Parameters of the Computing Center

[0341]

[0342] The parameters of the IEEE 39-node standard network unit are shown in Table 2.

[0343] Table 2 Parameters of Power Grid Units

[0344]

[0345] The optimization uses a 5-minute time step and a 12-hour optimization cycle, totaling 144 optimization periods. The initial load demand curves for the three computing centers over the 12-hour period are shown below. Figure 1 As shown. The load curves of some nodes (nodes 3, 4, 8, 20, 39) in the power grid over 12 hours are as follows. Figure 2 As shown.

[0346] The energy consumption optimization model for computing centers is transformed into a linear programming model with the energy consumption feasible region as the optimization space. Based on multi-parameter programming theory, the energy consumption elasticity space that can characterize the power supply boundary required by the regional computing centers is calculated. Based on this model, the energy consumption elasticity space of multiple computing centers in multiple time periods can be calculated. With three data centers and a coupling time period of 2, the energy consumption elasticity space is 6-dimensional. Table 3 below shows some of the elasticity space equations formed by the power consumption of the three computing centers in time period 1 and time period 2:

[0347] Table 3. Partial elastic space equations formed by the power consumption of the three computing centers during time period 1 and time period 2.

[0348]

[0349] The energy consumption elasticity space calculated based on multi-parameter programming theory is effectively embedded into the power grid unit combination scheduling model. The branch and bound method is used to solve the model to obtain the optimal power consumption solution for the computing centers. The power consumption command results calculated by the three computing centers through the collaborative scheduling model are as follows: Figure 3 As shown.

[0350] The power consumption of the computing center obtained from the collaborative scheduling optimization model is substituted into the computing center energy consumption optimization model to achieve the parsing of computing center power consumption instructions at the device level. By jointly optimizing the operating status of key energy-consuming equipment (IT equipment, air conditioning system) and the migration strategy of computing tasks, the optimal number of IT equipment to be activated in the computing center is obtained under the condition of lowest IT equipment startup cost. Figure 4 As shown, the air supply temperature of the air conditioning system is as follows: Figure 5 As shown.

Claims

1. A method for constructing and solving an optimization model for collaborative scheduling between computing centers and power grids based on energy consumption elasticity space, characterized in that, Includes the following steps: Step 1) Construct an energy consumption optimization model for the computing center; Step 2) Transform the computing center energy consumption optimization model into a linear programming model with the energy consumption feasible region as the optimization space, solve the linear programming model, and calculate the computing center energy consumption elastic space. Step 3) Based on the energy consumption elasticity space of the computing center, construct an optimization model for the coordinated scheduling of the computing center and the power grid; Step 4) Solve the optimization model for the coordinated scheduling of the computing center and the power grid to obtain the energy consumption results of the computing center; Step 5) Substitute the energy consumption results of the computing center into the energy consumption optimization model of the computing center to obtain the control parameters of each energy-consuming device in the computing center.

2. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, Step 1) involves constructing an energy consumption optimization model for the computing center, including the following steps: Step 1.1) Construct the power consumption model of the IT equipment in the computing center, the power consumption model of the air conditioning system, and the heat exchange coupling model between the IT equipment and the air conditioning system; The power consumption model for IT equipment in the computing center is shown below: （1） In the formula, This represents the average power consumption of the server when it is idle. This represents the average power consumption of the server when it is running at full load. It is a constant; These represent the utilization rate of IT device i and the chip temperature during time period t, respectively. This represents the operating status of IT device i during time period t, with values ​​of 2, 1, and 0 representing power-on, standby, and power-off, respectively. P represents the power consumption of IT device i when it is powered on. sleep This refers to the power consumption of key energy-consuming equipment when it enters standby mode. Power consumption of key energy-consuming equipment; The power consumption model of the air conditioning system is shown below: （2） In the formula, P aircon T out P represents the power consumption and supply air temperature of the air conditioning system during time period t; IT k represents the total power consumption of all IT devices within a specified area during time period t. T S dc Considering the ambient temperature coefficient and the floor space required for IT equipment; The heat exchange coupling model of IT equipment and air conditioning system is shown below: （3） （4） In the formula, The heat transfer coefficient; The air intake temperature of IT device i during time period t; For IT equipment i, the chip temperature and exhaust temperature during time period t; Equivalent thermal resistance; Step 1.2) Model the task load received by the computing center, that is: （5） In the formula, These represent the initial load requirements of latency-sensitive and latency-tolerant tasks for computing center m during time period t. These are the latency-sensitive and latency-tolerant task loads already processed by the computing center, respectively. , The initial load demand and processed load of computing center m in time period t; The relationship between the initial load requirements for latency-sensitive tasks and the initial load requirements for latency-tolerant tasks in the computing center and the already processed load is as follows: （6） （7） （8） （9） （10） In the formula, These represent the changes in load of computing center m in the time dimension and spatial dimension during time period t, respectively. These are the start and end time periods, respectively. This represents the total number of computing centers. The computing center m is delayed from time period t to time period m. The load capacity; This represents the maximum number of time periods during which the task can be delayed. For time period From the computing center Migration to computing center The load capacity; For the ordered set of all possible computing centers, i.e.: ; Step 1.3) Based on the task load received by the computing center, construct a computing center energy consumption optimization model with the goal of minimizing the total energy cost of all computing centers throughout the entire optimization cycle.

3. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, The objective function of the computing center energy consumption optimization model is shown below: （11） In the formula, Let m be the total power consumption of computing center m during time period t. This represents the total number of IT devices that switched from a powered-on state to a powered-on state. For electricity price, This is the startup cost coefficient. This refers to the duration of the time period.

4. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, The constraints of the computing center energy consumption optimization model include total power consumption constraints of the computing center, total computing tasks constraints of the computing center, load migration restriction constraints, air supply temperature constraints of the air conditioning system, air intake temperature constraints of IT equipment, and chip temperature constraints of IT equipment. The total power consumption constraints of the computing center are as follows: （12） In the formula, These represent the total power consumption of all IT equipment in computing center m during time period t, and the power consumption of the air conditioning system, respectively. Let be the power consumption of IT device i in computing center m during time period t. Let m be the number of IT devices in the computing center; The total computing task constraints for the computing center are as follows: （13） In the formula, These represent the utilization rates of IT equipment i in computing center m during time period t. The maximum processing load of IT equipment i in computing center m. Let m be the number of IT devices in the computing center; The load migration constraints are as follows: （14） In the formula, For latency-tolerant task load; The supply air temperature constraints of the air conditioning system are as follows: （15） In the formula, Let be the air supply temperature of IT equipment i in computing center m during time period t. These are the lower and upper limits of the supply air temperature, respectively. The intake air temperature constraints for IT equipment are as follows: （16） In the formula, Let be the air intake temperature of IT equipment i in computing center m during time period t. These are the lower and upper limits of the inlet air temperature, respectively. The temperature constraints for IT device chips are shown below: （17） In the formula, The chip temperature of IT equipment i in computing center m during time period t. This is the upper limit of the chip's temperature.

5. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, Step 2) involves calculating the energy consumption elasticity space of the computing center, including: Step 2.1) Transform the characterization of the energy consumption elasticity space into a feasibility analysis problem, namely: （18） （19） In the formula, w is the energy consumption variable vector, x is the decision variable vector, A and B are coefficient matrices, C and D are coefficient matrices; the energy consumption variable w is the planning parameter. Step 2.2) Define the optimal partition set and the critical region; Among them, any subset The optimal splitting set is shown below: （20） The effective constraints are as follows: （21） The following constraints are ineffective: （22） In the formula: , and A, B, and D represent the block submatrices A, B, and D corresponding to constraint row number C; The critical region CR with respect to the given parameter w0∈W ξ0 The definition is as follows: （23） Step 2.3) Divide the space according to the effective constraint set and the ineffective constraint set of the optimization problem, and merge the parameter spaces with the same effective constraint set into the critical domain; Wherein, any critical region CR i All are convex spaces; any critical region CR i With CR j Mutually exclusive; there are no gaps between two adjacent critical domains; i,j =1,2,…,m, i≠j; Step 2.4) Solve by optimal partitioning and critical domain to obtain the energy consumption elasticity space. .

6. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, In step 3), the objective function of the collaborative scheduling optimization model between the computing center and the power grid is as follows: （24） Where: continuous variables They represent the generating units. At any moment The costs of power output, startup, and shutdown; Indicates the unit The cost of labor; and This represents the set of generating units and scheduling periods.

7. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, In step 3), the constraints of the computing center and power grid collaborative scheduling optimization model include the energy consumption elasticity space feasible region constraint and the power grid operation security constraint. Power grid operation safety constraints include upper and lower limits of unit output, unit ramping constraints, minimum start-up and shutdown time limits of units, start-up and shutdown cost constraints, power balance constraints, and branch power flow constraints.

8. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 7, characterized in that, The feasible region constraints for the energy consumption elasticity space of the computing center are as follows: （25） In the formula, HA is the normal vector matrix of the hyperplane, and HB is the constant term vector of the hyperplane; P DC = {P DC,1 , P DC,2 ,…,P DC,m } represents the power consumption vector of each computing center; The upper and lower limits of the unit's output are constrained as follows: （26） In the formula, For the unit At any moment The start / stop status, a value of 1 indicates the unit's start / stop status. At any moment When the unit is in the "on" state, a value of 0 indicates that the unit is in the "on" state. At any moment The device is currently powered off. and For the unit Upper and lower limits of output; These are 0-1 discrete variables in the unit combination model, which is modeled using a univariate approach. The unit ramp-up constraints are as follows: （27） In the formula, and The unit's ramp rate; For the unit Changes in start / stop status; The minimum start-up and shutdown time limits for the generating units are as follows: （28） In the formula, and Indicates the unit The minimum continuous power-on or power-off time; and Indicates the unit At any moment The continuous power-on and power-off time; The unit start-up and shutdown cost constraints are as follows: （29） In the formula, and Indicates the unit The cost of powering on and off; The power balance constraints are as follows: （30） in, The injected power offset for node i. For the unit At any moment Contributing to the cause at the time For nodes At any moment The active power load demand; For nodal admittance matrix terms; Let be the voltage angle of the j-th node during time period t; Let i be the set of generators located at node i; The set of computing power centers located at node i; The total power consumption of computing center m during time period t; Branch flow constraints are shown below: （31） In the formula, These are the upper and lower limits of the branch power flow, respectively. Associating susceptance with the nodes of the branch circuit. Let k be the flow admittance between line k and node i; To inject additional active power into the branch circuit, Injection correction for line k; This is the voltage angle.

9. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, In step 4), the branch and bound method is used to solve the constructed computing center and power grid collaborative scheduling optimization model to obtain the optimal solution of energy consumption of each computing center after "power-computing" collaborative optimization.

10. The method for constructing and solving the optimization model of computing center and power grid collaborative scheduling based on energy consumption elastic space according to claim 1, characterized in that, In step 5), the optimal solution of energy consumption of each computing center is substituted into the computing center energy consumption optimization model constructed in step 1), and the model is solved by the score bounding method to obtain the control parameters of each energy-consuming device in the computing center, including the number of IT devices turned on and the air supply temperature of the air conditioning system.

Images