Space-time price-based computing power and power space-time collaborative flexible scheduling method and device

Abstract

The invention discloses a space-time price-based computing power and electric power space-time collaborative flexible scheduling method and device, and relates to the technical field of electric power systems, and the method comprises the steps: obtaining the architecture and operation characteristics of a data center, pre-establishing an energy consumption model and a load scheduling model of the data center, and pre-establishing a computing power-electric power collaborative regulation double-layer optimization model, the calculation power-electric power coordinated regulation double-layer optimization model is based on an energy consumption model and a load scheduling model of the data center, takes the initiative of the data center as an independent operator and the influence of the load of the data center on the electricity price of the system into consideration, and is constructed by taking the data center as a decision-making main body; a lower-layer power grid direct current optimal power flow problem is converted into a KKT condition to serve as a constraint to be substituted into an upper-layer problem, a strong dual theorem and a Big-M method are utilized to linearize and solve the double-layer optimization model of computing power-power coordinated regulation, and a computing power-power space-time coordinated flexible scheduling result is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power systems, in particular to a computing power and power space-time collaborative flexible scheduling method and device based on space-time price. BACKGROUND

[0002] With the rapid development of digital economy, the contradiction between the energy demand of computing power industry as a core component of new infrastructure and the low-carbon transformation of power system is increasingly prominent. As the core carrier of computing power resources, data centers have the characteristics of high energy consumption and continuous operation, and their electricity demand is growing exponentially. With the promotion of the "eastern data and western calculation" project, the migration of computing power resources to the western renewable energy-rich region helps to optimize the energy structure, but the time and space dynamic migration characteristics of large-scale computing power load pose new challenges to the safe and economic operation of the power system. On the one hand, the time and space two-dimensional adjustment capability of data center load (such as task delay processing and cross-regional computing power scheduling) has not been fully tapped, making it difficult to efficiently match the intermittent output of renewable energy with computing power demand. On the other hand, the power system and the computing power network belong to different operating entities, and the collaborative optimization of the two is subject to information barriers and interest games, and the existing control methods cannot meet the requirements of power grid operation constraints and computing power service quality.

[0003] In current research, the exploration of computing power-power collaborative optimization mainly focuses on static price guidance and fixed task allocation mode, which fails to effectively reflect the dynamic coupling characteristics of power-computing dual networks. Existing methods mostly assume that data centers directly participate in grid scheduling as passive response subjects, ignoring their privacy protection needs as independent operating entities. The centralized optimization framework is fundamentally contradictory to the distributed decision-making characteristics of power-computing dual networks, and the communication overhead brought by massive data interaction limits real-time scheduling performance. In addition, existing collaborative scheduling models generally adopt a one-way conduction mechanism of "given price-load response", ignoring the feedback impact of large-scale migration of computing power load on power flow. When multiple data centers respond to low-price regional electricity prices and migrate computing power, it may cause local line overload or saturation of renewable energy consumption capacity, forming a vicious cycle of "control lag-aggravated congestion". SUMMARY

[0004] To solve the problems mentioned in the background, the purpose of the present application is to provide a computing power and power space-time collaborative flexible scheduling method and device based on space-time price.

[0005] In the first aspect, the purpose of the present application can be achieved by the following technical solution: a computing power and power space-time collaborative flexible scheduling method based on space-time price, the method comprising the following steps:

[0006] obtain an architecture and operation characteristics of the data center, and pre-establish an energy consumption model and a load scheduling model of the data center, wherein the energy consumption model and the load scheduling model of the data center are constructed based on the architecture and operation characteristics of the data center, and a total operation cost of the data center is taken as an objective function;

[0007] pre-establish a double-layer optimization model of the computing power-electric power collaborative regulation, wherein the double-layer optimization model of the computing power-electric power collaborative regulation is based on the energy consumption model and the load scheduling model of the data center, and considers the initiative of the data center as an independent operator and the influence of the data center load on the system price, and is constructed by taking the data center as a decision subject;

[0008] convert a lower-layer power grid direct current optimal power flow problem into KKT conditions as constraints to be brought into an upper-layer problem, linearize and solve the double-layer optimization model of the computing power-electric power collaborative regulation by using a strong duality theorem and a Big-M method, and obtain a computing power-electric power space-time collaborative flexible scheduling result.

[0009] In combination with the first aspect, in some implementations of the first aspect, the method further includes that the pre-establishment of the energy consumption model of the data center is as follows:

[0010]

[0011] wherein PUE is an index describing the energy efficiency level of the data center; n is a data center node number, and n is in N; t is a time section number, and t is in T; and T is a total processing period set; represents power consumed by the data center n at the time t; represents power consumed by the server of the data center n at the time t;

[0012] the power consumed by the server is represented by a linear model taking the CPU utilization rate of the server as a variable:

[0013]

[0014] wherein, represents the number of servers started by the data center n at the time t; p Ser-idle represents the no-load power of the server of the data center in an idle state; p Ser-peak represents the full-load power of the server of the data center in a full-load state; u n,t represents the CPU utilization rate of the server of the data center n at the time t; L n,t represents the data load request amount processed by the data center n at the time t; τ n represents the processing rate of a single server of the data center n.

[0015] With reference to the first aspect, in some implementations of the first aspect, the method further includes: the architecture and operational characteristics of the data center are modeled for different data load types by quantifying space-time flexibility, the load of the data center including: online load, offline load, reducible load, and rigid load.

[0016] The mathematical model of the online load is represented as follows:

[0017]

[0018] wherein n'∈N' and The set N' is a set of data center nodes excluding node n; is an initial data load request amount received by the data center n at time t; and ξ is a proportion of the online load request amount; is the online load request amount of the data center n at time t before spatial transfer; is the online load request amount of the data center n at time t after spatial transfer; is a data load request amount transferred from the data center n' to the data center n at time t; is a data load request amount transferred from the data center n to the data center n' at time t;

[0019] The mathematical model of the offline load is represented as follows:

[0020]

[0021] wherein is a proportion of the offline load request amount; is the offline load request amount of the data center n at time t before time migration; is the offline load request amount of the data center n at time t after time migration; is a data load request amount migrated from time t' to time t and received by the data center n; is a data load request amount migrated from time t to time t" and output by the data center n;

[0022] The mathematical model of the reducible load is represented as follows:

[0023]

[0024] wherein ζ is a proportion of the reducible load request amount; is a reducible data load amount of the data center n at time t; is an actually reduced data load amount;

[0025] The mathematical model of the rigid load is represented as follows:

[0026]

[0027] wherein, is the rigid data load of data center n at time t;

[0028] In combination with the first aspect, in some implementations of the first aspect, the method further comprises: the total operation cost in the objective function is the sum of the single-step operation cost of all time periods, as follows:

[0029]

[0030] wherein, λ n,t is the node electricity value of data center n at time t.

[0031] In combination with the first aspect, in some implementations of the first aspect, the method further comprises: the energy consumption constraint of the pre-established energy consumption model of the data center is as follows:

[0032]

[0033] 0≤u n,t ≤U max

[0034] wherein, is the total number of servers of data center n; U max is the maximum CPU utilization of a single server;

[0035] The data load processing delay constraint is as follows:

[0036]

[0037] wherein, is the data queuing time; is the data processing time; T Delay is the average delay required for data processing, and the processing delay constraint is expressed as follows:

[0038]

[0039] In combination with the first aspect, in some implementations of the first aspect, the method further comprises: the line DC model of the lower DC optimal power flow problem is expressed as follows:

[0040]

[0041] wherein, i∈I, I is the set of all nodes of the power system; j∈J, J is the set of all nodes connected to node i in the power system; p ij,t is the power value flowing through line ij at time t; θ i,tis the voltage phase angle value of node i at time t; x ij is the reactance value of line ij;

[0042] The constraints that the system DC optimal power flow problem needs to satisfy are as follows:

[0043]

[0044] θ 1,t = 0: μ 1,t

[0045] wherein s e S, S is the set of all nodes where the conventional load is located; is the conventional load demand power value of node s at time t; r e R, R is the set of all nodes where the renewable energy is located; k r,t is the renewable energy consumption proportion of node r at time t; is the output power value of the renewable energy of node r at time t; λ i,t is the Lagrange multiplier of the node power balance constraint, and the physical meaning is the electricity price of the system at node i at time t; is the Lagrange multiplier of the node voltage phase angle constraint; F max is the maximum value of the system line flow; is the Lagrange multiplier of the line power transmission constraint; is the maximum power of the segmented linear operation economic cost function of the conventional generator set m at time t; is the Lagrange multiplier of the conventional generator output constraint; is the Lagrange multiplier of the renewable energy output constraint; μ 1,t is the Lagrange multiplier of the reference node setting constraint.

[0046] In combination with the first aspect, in some implementations of the first aspect, the method further includes that the upper problem of the computing power-electricity collaborative regulation and double-layer optimization model is a data center economic scheduling problem, the data center economic scheduling problem is solved in dependence on the system node electricity value obtained by the lower flow optimization, and the scheduling strategy of the data center is optimized based on the endogenous electricity price to obtain the power consumed by the data center operation, and the lower problem needs to receive the load plan from the upper data center to perform the flow optimization and solve to obtain the system node electricity value; the upper problem and the lower problem are tightly coupled.

[0047] In combination with the first aspect, in some implementations of the first aspect, the method further includes that the Lagrange function of the lower power grid DC optimal power flow problem is as follows:

[0048]

[0049] Subsequently, the Lagrangian function is taken partial derivative with respect to the decision variable of the lower layer problem to obtain the stability condition in the KKT condition:

[0050]

[0051] The inequality condition in the lower layer DC optimal power flow problem constitutes the complementary relaxation condition in the KKT condition:

[0052]

[0053]

[0054] The above stability condition, original equation condition and complementary relaxation condition constitute the KKT condition of the lower layer problem and are substituted into the upper layer problem as constraints to constitute the equilibrium constrained programming problem.

[0055] In combination with the first aspect, in some implementations of the first aspect, the method further comprises: linearizing and solving the double-layer optimization model of computing power and power collaborative regulation by using the strong duality theorem and the Big-M method, including:

[0056] The linearization of the objective function and the complementary relaxation condition of the converted equilibrium constrained programming problem, the linearization of the objective function adopts the strong duality theorem p * = d * :

[0057]

[0058] Wherein, p * is the optimal value of the lower layer original problem; d * is the optimal value of the lower layer dual problem;

[0059] After linearization, the linearized objective function is obtained:

[0060]

[0061] Subsequently, the Big-M method is used to linearize the complementary relaxation condition, and each complementary condition is recorded as 0≤d⊥p≥0, the Big-M method is to replace the condition d≥0, p≥0, d≤wM D , p≤(1-w)M P , where w is an auxiliary binary variable, M D and M P are normal numbers, and each complementary relaxation expression in the KKT condition is linearized.

[0062] Secondly, in order to achieve the above object, the application discloses a computing power and power space-time collaborative flexible scheduling device based on space-time price, comprising:

[0063] A model pre-establishing module acquires the architecture and operation characteristics of the data center, and pre-establishes an energy consumption model and a load scheduling model of the data center, wherein the energy consumption model and the load scheduling model of the data center are constructed based on the architecture and operation characteristics of the data center, and the total operation cost of the data center is taken as an objective function;

[0064] A model processing module is configured to pre-establish a double-layer optimization model of the computing power-electric power collaborative regulation, wherein the double-layer optimization model of the computing power-electric power collaborative regulation is based on the energy consumption model and the load scheduling model of the data center, and considers the initiative of the data center as an independent operator and the influence of the data center load on the system price, and is constructed by taking the data center as a decision subject;

[0065] A model solving module is configured to convert the lower-layer power grid direct current optimal power flow problem into KKT conditions as constraints, and bring the KKT conditions into the upper-layer problem, linearize and solve the double-layer optimization model of the computing power-electric power collaborative regulation by using the strong duality theorem and the Big-M method, and obtain a computing power-electric power space-time collaborative flexible scheduling result.

[0066] The present application has the following advantages:

[0067] The present application enables the data center to make efficient decisions under its own initiative, and realizes the collaborative interaction of the two networks; the present application replaces the traditional static price guide mode with the endogenous node price of the system, considers the influence of the data center load on the power grid power flow feedback, and promotes the collaborative efficiency of the computing network and the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description, and obviously, other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings;

[0069] Figure 1 The present application is a method flowchart;

[0070] Figure 2 The present application is an upper and lower problem coupling relationship diagram;

[0071] Figure 3 The present application is a specific embodiment of the interconnection line topology structure;

[0072] Figure 4 The present application is a specific embodiment of the system part node price when congestion occurs;

[0073] Figure 5 The present application is a specific embodiment of the load transfer situation and load situation of the data center 39;

[0074] Figure 6 The load transfer and load conditions of the data center 29 in the specific embodiment of the present invention;

[0075] Figure 7 This is a diagram showing the total load situation of regions W and E in a specific embodiment of the present invention;

[0076] Figure 8 The power flow of the transmission lines 14-24 in the specific embodiment of the present invention;

[0077] Figure 9 The power flow condition of the transmission line 25-26 in the specific embodiment of the present invention;

[0078] Figure 10 The renewable energy consumption situation in the specific implementation of the present invention;

[0079] Figure 11 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0080] 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 any creative efforts shall fall within the scope of protection of the present invention.

[0081] Example 1:

[0082] like Figure 1 As shown, a method for space-time coordinated flexible scheduling of computing power and electricity based on space-time prices includes the following steps:

[0083] S101: Obtain the architecture and operating characteristics of the data center, and pre-establish an energy consumption model and a load scheduling model for the data center. The energy consumption model and the load scheduling model for the data center are constructed based on the architecture and operating characteristics of the data center and with minimizing the total operating cost of the data center as an objective function.

[0084] The energy consumption model of the data center is pre-established as follows:

[0085]

[0086] Where PUE is an indicator that describes the energy efficiency level of a data center; n∈N, n is the data center node number, N is the set of all data center nodes; t∈T, t is the time section number, T is the total processing period set; represents the power consumed by data center n at time t; represents the power consumed by the server in data center n at time t;

[0087] The power consumed by the server is characterized by a linear model with the server CPU utilization as the variable:

[0088]

[0089] in, represents the number of servers turned on in data center n at time t; p Ser-idle Indicates the no-load power of the server in the data center in idle state; p Ser-peak Indicates the full load power of the server in the data center under full load; u n,t represents the CPU utilization of the server in data center n at time t; L n,t represents the data load request volume processed by data center n at time t; τ n Indicates the processing rate of a single server in data center n.

[0090] The architecture and operational characteristics of the data center model different data load types by quantifying spatiotemporal flexibility. The loads of the data center include online load, offline load, curtailable load, and rigid load.

[0091] The mathematical model of the online load is expressed as follows:

[0092]

[0093] Where n'∈N' and Set N' is the set of data center nodes that does not include node n; is the initial data load request volume received by data center n at time t; ξ is the proportion of online load requests; is the online load request volume of data center n at time t before space transfer; is the online load request volume of data center n at time t after space transfer; is the data load request amount transferred from data center n' to data center n at time t; is the data load request amount transferred from data center n to data center n' at time t;

[0094] The mathematical model of the offline load is expressed as follows:

[0095]

[0096] in, The proportion of offline load requests; is the offline load request volume of data center n at time t before time migration; is the offline load request amount of data center n at time t after time migration; is the data load request amount of data center n received from time t' to time t migration; is the data load request amount of data center n migrated from time t to time t'';

[0097] The mathematical model of the reducible load is represented as follows:

[0098]

[0099] wherein ζ is the proportion of the reducible load request amount; is the reducible data load amount of data center n at time t; is the actual reduced data load amount;

[0100] The mathematical model of the rigid load is represented as follows:

[0101]

[0102] wherein, is the rigid data load amount of data center n at time t.

[0103] The total running cost in the objective function with the minimum total running cost of the data center is the sum of single-step running costs of all time periods, as follows:

[0104]

[0105] wherein λ n,t is the node electricity value of data center n at time t.

[0106] The energy consumption constraint of the pre-established energy consumption model of the data center is as follows:

[0107]

[0108] 0≤u n,t ≤U max

[0109] wherein, is the total number of servers of data center n; U max is the maximum CPU utilization of a single server;

[0110] The data load processing delay constraint is as follows:

[0111]

[0112] wherein, is the data queuing time; is the data processing time; T DelayThe average delay required for data processing is expressed as follows by the above formula:

[0113]

[0114] S102: A dual-layer optimization model for computing power and power collaborative regulation is established in advance, wherein the dual-layer optimization model for computing power and power collaborative regulation is based on an energy consumption model and a load scheduling model of the data center, and considers the initiative of the data center as an independent operator and the influence of the data center load on the system electricity price, and is constructed by taking the data center as a decision subject;

[0115] The upper problem of the dual-layer optimization model for computing power and power collaborative regulation is a data center economic scheduling problem, the solution of the data center economic scheduling problem depends on the system node electricity value obtained by the lower flow optimization, and the scheduling strategy of the data center is optimized based on the endogenous electricity price to obtain the power consumed by the data center operation, the lower problem needs to receive the load plan from the upper data center to perform flow optimization to obtain the system node electricity value; the upper and lower problems are tightly coupled.

[0116] S103: The lower power grid DC optimal power flow problem is converted into KKT conditions as constraints and brought into the upper problem, the dual-layer optimization model for computing power and power collaborative regulation is linearized and solved by using the strong duality theorem and Big-M method, and a computing power and power space-time collaborative flexible scheduling result is obtained.

[0117] The line DC model of the lower power grid DC optimal power flow problem is expressed as follows:

[0118]

[0119] Wherein, i∈I, I is the set of all nodes of the power system; j∈J, J is the set of all nodes connected to node i in the power system; p ij,t is the power value flowing through the line ij at t; θ i,t is the voltage phase angle value of node i at t; x ij is the reactance value of the line ij;

[0120] The constraints that the system DC optimal power flow problem needs to satisfy are as follows:

[0121]

[0122] θ 1,t =0: μ 1,t

[0123] Wherein, s∈S, S is the set of all nodes where the regular load is located; is the regular load demand power value of node s at time t; r e R, R is the set of all nodes where renewable energy is located; k r,t is the renewable energy consumption proportion of node r at time t; is the output power value of renewable energy of node r at time t; λ i,t is the Lagrange multiplier of node power balance constraint, the physical meaning is the electricity price of system at node i at time t; is the Lagrange multiplier of node voltage phase angle constraint; F max is the maximum value of system line flow; is the Lagrange multiplier of line power transmission constraint; is the maximum power of segmented linear operation economic cost function of regular generator set m at time t; is the Lagrange multiplier of regular generator output constraint; is the Lagrange multiplier of renewable energy output constraint; μ 1,t is the Lagrange multiplier of reference node setting constraint.

[0124] The Lagrange function of the lower-layer power grid DC optimal power flow problem is:

[0125]

[0126] Subsequently, the partial derivative of the Lagrange function with respect to the decision variable of the lower-layer problem is obtained, and the stability condition in the KKT condition is obtained:

[0127]

[0128] The inequality condition in the lower-layer DC optimal power flow problem constitutes the complementary relaxation condition in the KKT condition:

[0129]

[0130]

[0131] The above stability condition, original equation condition and complementary relaxation condition constitute the KKT condition of the lower-layer problem, and are substituted into the upper-layer problem as constraints to constitute the equilibrium constrained programming problem.

[0132] The strong duality theorem and Big-M method are used to linearize and solve the computing power-electricity collaborative regulation double-layer optimization model, including:

[0133] The linearization processing of the objective function and the complementary relaxation condition of the converted equilibrium constrained programming problem, the linearization of the objective function adopts the strong duality theorem p * =d * :

[0134]

[0135] Among them, p * is the optimal value of the original problem at the lower level; d * is the optimal value of the lower-level dual problem;

[0136] After linearization, the linearized objective function is obtained:

[0137]

[0138] Subsequently, the complementary slack conditions are linearized using the Big-M method, where each complementary condition is written as 0≤d⊥p≥0. The Big-M method uses the conditions d≥0, p≥0, d≤wM D , p≤(1-w)M P Instead, where w is an auxiliary binary variable, M D and M P For a sufficiently large positive constant, each complementary relaxation expression in the KKT condition is linearized.

[0139] Specifically, the present invention will be further described below through examples:

[0140] This embodiment uses the improved IEEE39 node for testing. The topology of the line between power grids is as shown in the attached figure. Figure 3 As shown in the figure, the simulation is divided into two regions. Region W's power supply mainly comes from renewable energy sources such as wind power and photovoltaics. It has only one data center at node 39 and has very few data requests. Region E's power supply mainly comes from conventional generators and has four data centers at nodes 21, 24, 28, and 29. The data request volume of each data center is much larger than that of the data center in region W, and there is a certain regularity in the time scale.

[0141] A 24-hour working day with a time resolution of 1 hour is selected, and two comparison scenarios are designed: (1) The data center has no flexibility and is regarded as a rigid load; (2) The data center has spatiotemporal flexibility and realizes spatiotemporal coordinated flexible scheduling with the power grid. In the case of scenario 1, the load demand of area E is much greater than the load demand of area W, which will bring huge transmission pressure to the transmission line. In order to meet the high load demand of area E, the transmission line transmits electricity at the capacity limit, and system congestion is inevitable. When system congestion occurs, the node electricity price of the system will show differences in the spatial dimension, as shown in the attached figure. Figure 4 As shown, the electricity price in area E is significantly higher than that in area W.

[0142] In the case of scenario 2, the data center load transfer capability in both time and space dimensions should be fully considered, as shown in the attached Figure 5 、 6As shown in Figure 3, data center 39 is located in area W, where renewable energy output is abundant, and mainly receives computing loads transferred from data centers in area E. Data center 29 is located in area E, where renewable energy output is scarce, and mainly transfers computing loads to area W. In addition, during periods of high electricity prices and scarce renewable energy output, peak computing loads are transferred to periods of low electricity prices and low computing loads. This transfer of computing power replaces the transmission of electricity, balancing computing resources as much as possible. The load conditions of the two areas after the time-space transfer are shown in Figure 3. Figure 7 The transmission lines no longer have to transmit electricity in multiple periods according to the capacity limit, which successfully alleviates the congestion problem of the system (as shown in the attached Figure 8 、 9 As shown in the figure, the safety of power transmission and system operation is guaranteed, and the value of the system electricity no longer has spatial differences. In addition, the spatial flexibility of the data center can transfer the data load of area E to the data center in area W for processing, and can directly consume local redundant renewable energy in area W, as shown in the attached figure. Figure 10 The test shows that the absorption rate of renewable energy has increased by about 15%, which has reduced the proportion of wind and solar power curtailment to a certain extent.

[0143] Example 2: In order to achieve the above purpose, Figure 11 As shown, based on the first embodiment, the present invention discloses a computing power and power spatiotemporal coordinated flexible scheduling device based on spatiotemporal prices, including:

[0144] A model pre-building module 11 obtains the architecture and operating characteristics of the data center and pre-establishes an energy consumption model and a load scheduling model for the data center. The energy consumption model and the load scheduling model for the data center are constructed based on the architecture and operating characteristics of the data center and with minimizing the total operating cost of the data center as the objective function.

[0145] Model processing module 12 is used to pre-establish a two-tier optimization model for computing power and power coordinated regulation. The two-tier optimization model for computing power and power coordinated regulation is based on the data center's energy consumption model and load scheduling model, and takes into account the initiative of the data center as an independent operator and the impact of the data center load on system electricity prices. The model is constructed with the data center as the decision-making entity.

[0146] The model solving module 13 is used to convert the DC optimal power flow problem of the lower-level power grid into KKT conditions as constraints and introduce them into the upper-level problem. The strong duality theorem and the Big-M method are used to linearize and solve the two-level optimization model of computing power and power coordinated control, and obtain the results of flexible scheduling of computing power and power in spatiotemporal coordination.

[0147] Based on the same inventive concept, the present application further provides a computer device, comprising: one or more processors, and a memory for storing one or more computer programs; the program comprises program instructions, and the processor is configured to execute the program instructions stored in the memory. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are configured to implement one or more instructions, and are specifically configured to load and execute one or more instructions in the computer storage medium to implement the above method.

[0148] It needs to be further explained that, based on the same inventive concept, the present application further provides a computer storage medium, which stores a computer program, and the computer program is executed by the processor to perform the above method. The storage medium can adopt any combination of one or more computer readable media. The computer readable medium can be a computer readable signal medium or a computer readable storage medium. The computer readable storage medium may, for example, be but is not limited to an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or component, or any combination of the above. More specific examples (non-exhaustive list) of the computer readable storage medium include: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or component.

[0149] In the description of the present application, the description of the terms "one embodiment", "an example", "a specific example" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present disclosure. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any suitable manner in one or more embodiments or examples.

[0150] The foregoing presents and describes the basic principles, main features and advantages of the present disclosure. It should be understood by those skilled in the art that the present disclosure is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only to illustrate the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, various changes and improvements can be made to the present disclosure, and all these changes and improvements fall within the scope of the present disclosure.

Claims

1. A flexible scheduling method for computing power and electricity spatiotemporal coordination based on spatiotemporal prices, characterized by: The method comprises the following steps: Obtain the architecture and operating characteristics of the data center and pre-establish the data center's energy consumption model and load scheduling model. The energy consumption model and load scheduling model are based on the data center's architecture and operating characteristics and are constructed with minimizing the total operating cost of the data center as the objective function. A two-tier optimization model for computing power and power coordinated regulation is pre-established. This model is based on the data center's energy consumption model and load scheduling model, and takes into account the initiative of the data center as an independent operator and the impact of the data center load on system electricity prices. The model is constructed with the data center as the decision-making entity. The optimal DC power flow problem of the lower-level power grid is converted into KKT conditions and introduced into the upper-level problem as constraints. The strong duality theorem and Big-M method are used to linearize and solve the two-level optimization model of computing power and power coordinated regulation, and the results of spatiotemporal coordinated flexible scheduling of computing power and power are obtained.

2. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 1 is characterized in that: The energy consumption model of the data center is pre-established as follows: Where PUE is an indicator that describes the energy efficiency level of a data center; n∈N, n is the data center node number, N is the set of all data center nodes; t∈T, t is the time section number, T is the total processing period set; represents the power consumed by data center n at time t; represents the power consumed by the server in data center n at time t; The power consumed by the server is characterized by a linear model with the server CPU utilization as the variable: in, represents the number of servers opened in data center n at time t; p Ser-idle Indicates the no-load power of the server in the data center in idle state; p Ser-peak Indicates the full load power of the server in the data center under full load; u n,t represents the CPU utilization of the server in data center n at time t; L n,t represents the data load request volume processed by data center n at time t; τ n Indicates the processing rate of a single server in data center n.

3. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 1 is characterized in that: The architecture and operational characteristics of the data center model different data load types by quantifying spatiotemporal flexibility. The loads of the data center include online load, offline load, curtailable load, and rigid load. The mathematical model of the online load is expressed as follows: Where n'∈N' and Set N' is the set of data center nodes that does not include node n; is the initial data load request volume received by data center n at time t; ξ is the proportion of online load requests; is the online load request volume of data center n at time t before space transfer; is the online load request volume of data center n at time t after space transfer; is the data load request amount transferred from data center n' to data center n at time t; is the data load request amount transferred from data center n to data center n' at time t; The mathematical model of the offline load is expressed as follows: in, The proportion of offline load requests; is the offline load request volume of data center n at time t before time migration; is the offline load request volume of data center n at time t after time migration; is the data load request amount received by data center n from time t' to time t; is the data load request amount that migrates from data center n from time t to time t”; The mathematical model of the load reduction is expressed as follows: Among them, ζ is the proportion of load requests that can be reduced; is the amount of data load that can be reduced in data center n at time t; The actual amount of data load reduced; The mathematical model of the rigid load is expressed as follows: in, is the rigid data load of data center n at time t.

4. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 1 is characterized in that: The total operating cost within the objective function of minimizing the total operating cost of the data center is the sum of the single-step operating costs of all time periods, as follows: Among them, λ n,t is the electricity value of the node where data center n is located at time t.

5. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 2 is characterized in that: The energy consumption constraints of the pre-established data center energy consumption model are as follows: 0≤u n,t ≤U max in, is the total number of servers in data center n; U max The maximum CPU utilization of a single server; The data payload processing delay constraints are as follows: in, Queue time for data; is the data processing time; T Delay is the average delay required for data processing. The processing delay constraint is comprehensively expressed as follows using the above formula:

6. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 1 is characterized in that: The line DC model of the DC optimal power flow problem of the lower grid is expressed as follows: Where i∈I, I is the set of all nodes in the power system; j∈J, J is the set of all nodes connected to node i in the power system; p ij,t is the power value flowing through line ij at time t; θ i,t is the voltage phase angle value of node i at time t; x ij is the reactance value of line ij; The constraints that need to be satisfied by the system DC optimal power flow problem are as follows: i 1,t =0:μ 1,t Where s∈S, S is the set of nodes where all conventional loads are located; is the conventional load power demand value of node s at time t; r∈R, R is the set of all nodes where renewable energy is located; k r,t The proportion of renewable energy consumption at node r at time t; is the output power value of renewable energy at node r at time; i,t is the Lagrange multiplier of the node power balance constraint, and its physical meaning is the electricity price of node i in the system at time t; is the Lagrange multiplier of the node voltage phase angle constraint; F max is the maximum value of the system line power flow; is the Lagrange multiplier for the line power transfer constraint; is the maximum power produced by the piecewise linear economic cost function of conventional generator set m at time t; is the Lagrange multiplier for the output constraint of conventional generator sets; is the Lagrange multiplier for renewable energy output constraint; μ 1,t Sets the Lagrange multiplier for the constraint at the reference node.

7. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 1 is characterized in that: The upper-level problem of the two-level optimization model of computing power-electricity coordinated control is the data center economic scheduling problem. The solution to the data center economic scheduling problem depends on the system node electricity value obtained by the lower-level flow optimization, and optimizes the data center scheduling strategy based on the endogenous electricity price to obtain the power consumed by the data center operation. The lower-level problem needs to receive the load plan from the upper-level data center to perform flow optimization to obtain the system node electricity value; the upper and lower-level problems are tightly coupled.

8. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 6 is characterized in that: The Lagrangian function of the DC optimal power flow problem of the lower grid is: Subsequently, the Lagrangian function is partially derived with respect to the decision variables of the underlying problem, and the stability condition in the KKT condition is obtained: The inequality conditions in the lower-level DC optimal power flow problem constitute the complementary relaxation conditions in the KKT conditions: The above stability conditions, original equality conditions, and complementary relaxation conditions constitute the KKT conditions of the lower-level problem and are substituted into the upper-level problem as constraints to form an equilibrium constrained programming problem.

9. The method for flexible scheduling of computing power and electricity in time and space based on time and space prices according to claim 1 is characterized in that: The strong duality theorem and Big-M method are used to linearize and solve the two-level optimization model for computing power and power coordinated control, including: The objective function and complementary relaxation conditions of the converted equilibrium constraint programming problem are linearized. The linearization of the objective function adopts the strong duality theorem p * =d * : Among them, p * is the optimal value of the original problem at the lower level; d * is the optimal value of the lower-level dual problem; After linearization, the linearized objective function is obtained: Subsequently, the complementary slack conditions are linearized using the Big-M method, where each complementary condition is written as 0≤d⊥p≥0. The Big-M method uses the conditions d≥0, p≥0, d≤wM D , p≤(1-w)M P Instead, where w is an auxiliary binary variable, M D and M P is a positive constant, and each complementary relaxation expression in the KKT condition is linearized.

10. A computing power and electricity spatiotemporal coordinated flexible scheduling device based on spatiotemporal prices, characterized in that: include: The model pre-building module obtains the data center's architecture and operating characteristics and pre-establishes the data center's energy consumption model and load scheduling model. The data center's energy consumption model and load scheduling model are based on the data center's architecture and operating characteristics, with the minimum total operating cost of the data center as the objective function. The model processing module is used to pre-establish a two-layer optimization model for computing power and power coordinated control, where: The two-tier optimization model for computing power and power coordinated regulation is based on the data center's energy consumption model and load scheduling model. It considers the initiative of the data center as an independent operator and the impact of the data center load on system electricity prices. It is constructed with the data center as the decision-making entity. The model solving module is used to convert the DC optimal power flow problem of the lower-level power grid into KKT conditions as constraints and introduce them into the upper-level problem. It uses the strong duality theorem and the Big-M method to linearize and solve the two-level optimization model of computing power and power coordinated control, and obtains the results of flexible scheduling of computing power and power coordinated in time and space.

Images