Data center demand response method, system and equipment considering electric power-computing power-thermal power collaboration

Abstract

The invention discloses a data center demand response method, system and equipment considering electric power-computing power-thermal power cooperation, and relates to the field of data center demand response. The method comprises the following steps: constructing an objective function and a constraint condition of a data center demand response considering electric power-computing power-thermal power collaboration, and forming a data center demand response model together; the target function takes the minimum total electricity price of the internet equipment and the cooling equipment in the scheduling period as a target; solving the demand response model of the data center according to the operation data of the scheduling period, and obtaining an electric power-computing power-heating power collaborative scheduling result of the scheduling period; the electric power-computing power-heating power collaborative scheduling result comprises the optimal load scheduling of each server in each scheduling time period, and the optimal on-off decision and the power demand of each server in each scheduling time period. According to the invention, collaborative optimization of the data center in the aspects of electric power, computing power and thermal power can be realized, the energy utilization efficiency of the data center is improved, and the operation cost is reduced.

Description

Technical Field

[0001] The present application relates to the field of data center demand response, and in particular to a data center demand response method, system and equipment that considers power-computing-heat synergy. Background Art

[0002] As the cornerstone of the informatization of modern society, data centers are becoming increasingly important. They not only carry the storage and processing tasks of massive data, but also serve as the physical support for Internet services and cloud computing applications. However, with the continuous expansion of the scale of data centers and the continuous growth of computing needs, their energy consumption problem has become increasingly serious. The power consumption of data centers not only includes the operation of computing equipment such as servers, but also covers the energy consumption of infrastructure such as cooling systems. These factors together determine the overall energy efficiency of data centers. Under the existing technical background, energy management in data centers faces a series of challenges. First, the power consumption pattern of data centers is often inflexible and difficult to adapt to real-time price changes in the electricity market. This means that during periods of high electricity costs, data centers may not be able to effectively reduce energy use, thereby increasing operating costs. Secondly, data centers also have deficiencies in computing resource allocation. Resource allocation fails to fully consider the characteristics and priorities of tasks, resulting in uneven resource utilization and inefficiency. In addition, the thermal management efficiency of data centers is also not high. The cooling system usually consumes a lot of energy and lacks coordinated optimization with the power and computing power systems, which limits the improvement of overall energy efficiency. Although some data centers have begun to try to reduce energy consumption by improving server hardware and adopting more efficient cooling systems, these measures often ignore the fact that data centers are complex energy systems with close interrelationships and synergies between electricity, computing power, and heat. Existing technologies have obvious deficiencies in the coordinated optimization of power-computing-heat in data centers, especially in the modeling and scheduling of demand response capabilities. Data centers lack effective demand response strategies and find it difficult to make dynamic adjustments based on power market price signals and internal operating status, which limits the participation of data centers in the power market and misses the opportunity to achieve cost savings through demand response. Fluctuations in power market prices provide data centers with opportunities to achieve cost savings through demand response. However, existing technologies often fail to effectively use these price signals to guide power consumption, computing resource allocation, and thermal management in data centers. The demand response capability of data centers is of great significance to the stable operation of the power grid and the efficient use of energy, but many data centers lack effective demand response strategies and cannot fully utilize power market price signals to optimize operations.

[0003] In order to improve the energy utilization efficiency of data centers and reduce operating costs, how to achieve comprehensive coordinated optimization of electricity, computing power and heat in data centers is of great significance to the stable operation of the power grid and the efficient use of energy. Summary of the invention

[0004] The purpose of this application is to provide a data center demand response method, system and equipment that considers the synergy of electricity, computing power and heat, which can achieve coordinated optimization of electricity, computing power and heat in the data center, improve the energy utilization efficiency of the data center and reduce operating costs.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In the first aspect, the present application provides a data center demand response method considering power-computing power-heat synergy, including: constructing an objective function for data center demand response considering power-computing power-heat synergy; the objective function takes the minimum total electricity price of Internet devices and cooling equipment during the scheduling period as the goal; establishing constraints for data center demand response considering power-computing power-heat synergy, and the objective function and the constraints together constitute a data center demand response model; obtaining operating data of the data center during the scheduling period; the operating data includes the electricity market price during the scheduling period; solving the data center demand response model based on the operating data during the scheduling period to obtain the power-computing power-heat synergy scheduling result during the scheduling period; the power-computing power-heat synergy scheduling result includes: the optimal load scheduling of each server in each scheduling time period, the optimal start and shut down decision and power demand of each server in each scheduling time period.

[0007] In the second aspect, the present application provides a data center demand response system considering power-computing power-heat coordination, including: a monitoring system and a management system. The management system is used to construct the objective function and constraints of the data center demand response considering power-computing power-heat coordination, and the objective function and the constraints together constitute a data center demand response model; the objective function takes the minimum total electricity price of Internet equipment and cooling equipment during the scheduling period as the goal; the monitoring system is used to collect the operation data of the data center other than the electricity market price during the scheduling period, and transmit it to the management system; the monitoring system is also used to obtain the electricity market price during the scheduling period from the system operator and transmit it to the management system; the management system is also used to solve the data center demand response model according to the operation data during the scheduling period, and obtain the power-computing power-heat coordination scheduling result during the scheduling period; the power-computing power-heat coordination scheduling result includes: the optimal load scheduling of each server in each scheduling time period, the optimal start and close decision and power demand of each server in each scheduling time period.

[0008] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-mentioned data center demand response methods considering power-computing power-heat synergy.

[0009] According to the specific embodiments provided in this application, this application has the following technical effects:

[0010] The present application provides a data center demand response method, system and equipment that considers power-computing power-heat synergy. By establishing an objective function and constraint conditions for data center demand response that considers power-computing power-heat synergy, the objective function takes the minimum total electricity price of Internet devices and cooling equipment during the scheduling period as the goal, and solves to obtain the power-computing power-heat synergy scheduling result during the scheduling period. Since the objective function comprehensively considers the power consumption, computing task scheduling and thermal management requirements of the data center, it realizes comprehensive coordinated optimization management of power, computing power and thermal resources. This collaborative optimization method improves energy utilization efficiency, reduces energy consumption, and reduces operating costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0012] Figure 1 A flowchart of a data center demand response method considering power-computing-heat synergy provided in an embodiment of the present application;

[0013] Figure 2 A flow chart for constructing a data center demand response model provided by another embodiment of the present application;

[0014] Figure 3 A schematic diagram of the definition of data center demand response capability provided by another embodiment of the present application;

[0015] Figure 4 A schematic diagram of a general flow chart of a data center demand response method considering power-computing-heat synergy provided in another embodiment of the present application;

[0016] Figure 5 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0018] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0019] The comprehensive coordinated optimization of data centers in terms of electricity, computing power and heat includes establishing a demand response capability model for data centers, evaluating and optimizing the power consumption, computing resource allocation and thermal management of data centers under different electricity prices. At the same time, it is necessary to develop a dynamic demand response scheduling strategy so that data centers can make real-time adjustments based on electricity market price signals and internal operating status. In addition, a comprehensive evaluation of the operating benefits of data centers after the implementation of demand response strategies and the design of incentive mechanisms to encourage demand-side resources to participate in the market are also key steps to achieve the coordinated optimization of power, computing power and heat in data centers. This can not only improve the energy efficiency of data centers, but also promote the interaction between data centers and power grids, contributing to the stable operation of power grids and the efficient use of energy.

[0020] Specifically, the problems faced by data centers in terms of power-computing-heat synergy include inflexible power consumption patterns, unreasonable allocation of computing resources, low thermal management efficiency, and lack of demand response strategies. In order to meet these challenges, new technologies and methods need to be developed to achieve comprehensive synergistic optimization of power, computing power, and heat in data centers, thereby improving energy utilization efficiency, reducing operating costs, and promoting effective interaction between data centers and power grids.

[0021] A Chinese patent with publication number CN117667639A discloses a data center energy regulation and evaluation method based on task scheduling. According to the resource allocation of each task at each time step, multiple constraint equations are established between the electric power and the energy consumption of the server, thereby obtaining an energy consumption schedulable domain model for each task; a Chinese patent application with publication number CN116755336A provides a data center energy management method with flexible coordination of computing power and heat. The system structure and parameters of the data center are input to establish a data center optimization scheduling model with flexible coordination of computing power and heat; the first method above realizes the joint scheduling of computing power and electricity in the data center, and the second method realizes the joint scheduling of computing power and heat in the data center, but the energy consumption of computing power equipment and cooling equipment in the data center is very high, and both methods fail to achieve the coordination of power, computing power and heat in the data center.

[0022] In order to solve the above problems, in an exemplary embodiment, Figure 1 As shown, a data center demand response method considering power-computing-heat synergy is provided, including the following steps 101 to 104. Among them:

[0023] Step 101: construct an objective function for data center demand response that considers power-computing-heat synergy; the objective function aims to minimize the total electricity price of Internet devices and cooling devices during the scheduling period;

[0024] Step 102: Establish constraint conditions for data center demand response considering power-computing-heat synergy, and use the objective function and the constraint conditions together to form a data center demand response model;

[0025] Step 103: Acquire the operation data of the data center during the dispatch period; the operation data includes the power market price during the dispatch period;

[0026] Step 104: Solve the data center demand response model based on the operating data during the scheduling period to obtain the power-computing-heat coordinated scheduling result during the scheduling period; the power-computing-heat coordinated scheduling result includes: the optimal load scheduling of each server in each scheduling time period, the optimal start-up and shutdown decision and power demand of each server in each scheduling time period.

[0027] Implement the above steps 101 to 104, by establishing the objective function and constraints of the data center demand response considering the power-computing power-heat coordination, the objective function takes the minimum total electricity price of Internet devices and cooling equipment during the scheduling period as the goal, and obtains the power-computing power-heat coordination scheduling result during the scheduling period. Since the objective function comprehensively considers the power consumption, computing task scheduling and thermal management requirements of the data center, it realizes the comprehensive coordinated optimization management of power, computing power and thermal resources. This coordinated optimization method improves energy utilization efficiency, reduces energy consumption, and reduces operating costs.

[0028] By establishing a comprehensive demand response model, the dynamic power consumption, computing resource allocation and thermal management optimization of data centers under different electricity market prices are realized. This method uses electricity market price signals to guide the real-time scheduling of data centers, including server consolidation, virtual machine migration and cooling system optimization, to improve energy efficiency, reduce costs, and support grid stability. In addition, through resource incentive mechanisms, the efficient use of data center resources and market participation are further promoted.

[0029] Server consolidation means that a server can be divided into multiple virtual servers to improve operating efficiency; virtual machine migration refers to the virtual machine technology that allows a single physical server to run multiple operating systems and applications, each operating system and application running in its own virtual machine. The above two technologies are implemented by software simulating hardware, allowing multiple virtual machines to share the resources of the same physical server.

[0030] Resource incentives refer to the ability of data centers to schedule workloads based on price signals from the electricity market to take advantage of lower-priced electricity.

[0031] In another exemplary embodiment of the present application, in order to construct the objective function of the data center demand response, the present application determines the load factor through experiments, and the load factor is the ratio of the energy consumption of the Internet device to the energy consumption of the cooling device. When determining the load factor, the collection and preprocessing of power-computing-thermal data is a crucial first step. As the execution subject, the data center monitoring system is responsible for real-time collection of operating data of servers, storage systems, network devices and cooling systems. This includes power consumption data, such as input power and energy consumption of each device; computing power-related data, such as the CPU and memory usage of the server, the load intensity of job processing; and thermal-related data, such as the operating status of the cooling system, the temperature of the air outlet and the air inlet, and the heat discharged by the server. After the data collection is completed, the data center management system will preprocess these raw data, including data cleaning, denoising, standardization and other steps to ensure the accuracy and availability of the data. In addition, the system will extract the typical operating characteristics of the data center based on historical data and operating modes, and provide high-quality input data for the subsequent demand response model establishment. This step is the basis for realizing the coordinated optimization of power-computing-thermal and provides data support for the refined management of the data center. The load factor for each time period is obtained through the above data. The method for determining the load factor can be summarized as follows: Steps 201 to 204 are replaced by:

[0032] Step 201: Obtain the original data of power consumption, computing power and thermal power of the data center in each historical time period; the power consumption data includes: input power and energy consumption of each device in the data center; the computing power data includes: CPU and memory usage of the server, and load intensity of job processing; the thermal data includes: operating status of the cooling system, temperature of the cooling system outlet, temperature of the cooling system inlet and heat discharged by the server.

[0033] Step 202: Preprocess the original data of power consumption, computing power and thermal power for each historical time period to obtain the power consumption data, computing power data and thermal power data for each historical time period; the preprocessing includes cleaning, denoising and standardization.

[0034] Step 203: Extract the operation characteristics of the data center based on the power consumption data, computing power data, thermal data and operation mode of each historical time period.

[0035] Step 204: Based on the operating characteristics of the data center, as well as the power consumption data, computing power data, and thermal data of each historical time period, the ratio of the energy consumption of the Internet equipment to the energy consumption of the cooling equipment is determined through experiments as a load factor.

[0036] The load factor is designed as a known input rather than a variable, and is mainly used to quantify the cooling efficiency performance of a specific data center over a period of time. It reflects the ratio between the heat removed by the cooling system (assumed to be equal to the electrical power consumed by the computing equipment) and the energy consumption of the air conditioning system (CRAC). The load factor is affected by a combination of factors, including external environmental conditions (such as temperature, humidity, etc.), the configuration of the cooling system (such as equipment efficiency and design characteristics), and the operating status of the data center (such as server density and task load). In order to determine the load factor value, based on long-term monitoring data, the actual load factors of different time periods are recorded, and the average value is taken through historical data to obtain a stable load factor input used in the model.

[0037] In another exemplary embodiment of the present application, the objective function takes into account the energy consumption costs of Internet devices (referred to as IT devices) and cooling devices at the same time, wherein the energy consumption of IT devices can be accurately calculated by calculating the energy consumption of the CPU and the cost of starting and shutting down the server; and the energy consumption of cooling devices is calculated by measuring data to obtain the load factor of each time period: that is, the ratio of the energy consumption of IT devices and cooling devices, so that the total energy consumption of IT devices and cooling devices can be calculated more accurately. Then the objective function in the above step 101 is:

[0038]

[0039]

[0040] In the formula, F IT,t is the total electricity price of IT equipment in period t, F CL,t is the total electricity price of the cooling equipment in period t, NT is the total time period of the scheduling period, LMP t is the node marginal electricity price (electricity market price) in period t, E IT,t is the total power consumption of IT equipment in period t, E CL,t is the total power consumption of the cooling equipment during period t, LF t is the load factor during period t, E IT,t,s is the IT equipment energy consumption of server s in period t, NS is the total number of servers; DT is the total time period, P I,sis the idle cost of server s, I t,s is whether server s is running during period t, P c,s is the full load cost of the server, U t,s is the CPU utilization of server s during period t, E SU,s is the startup cost of server s, Y t,s Server s is turned on for period t, E SD,s is the shutdown cost of server s, Z t,s Server s is shut down for period t, It is a universal quantifier.

[0041] The construction process of the above objective function can be summarized as Figure 2 The process shown.

[0042] In another exemplary embodiment of the present application, the constraints in the above step 102 include: CPU frequency constraint, Internet data center capacity constraint, server minimum power-on constraint, server power-on state constraint, server power-off state constraint, server minimum power-off time constraint, server state control start-stop constraint, task load execution state control start-stop constraint, workload distribution constraint, workload execution time constraint and continuous execution workload constraint.

[0043] The CPU frequency constraints are:

[0044]

[0045] Where U t,s is the CPU utilization of server s during period t, NI is the total number of workloads, and f i is the CPU frequency requirement of workload i, I t,i,s is the indicator variable of whether workload i is assigned to server s in period t (i.e., whether the workload is assigned to server s), f s is the CPU frequency requirement of server s, U max,s The upper limit of the CPU utilization of server s.

[0046] Formula (7) means that in each time period t and server s, the CPU utilization of server s is defined as the sum of the CPU requirements of all workloads i assigned to server s divided by the CPU frequency of the server, and the CPU utilization must be less than or equal to the upper limit of the CPU utilization of server s.

[0047] The capacity constraints of Internet data centers are:

[0048]

[0049] Where NS is the total number of servers, I t,sis the indicator variable of server s in period t, I t,i is the indicator variable of workload i in period t. CM is the total capacity margin of the data center, that is, the additional computing capacity that needs to be reserved in addition to the currently required computing resources.

[0050] Formula (8) means that in each time period t, the sum of the CPU frequencies provided by all servers minus the sum of the CPU frequencies required by all workloads must be greater than or equal to the total capacity margin of the data center.

[0051] The minimum server startup constraints are:

[0052]

[0053] In the formula, MU s is the minimum startup time of server s, TU s is the time that server s has been powered on before the scheduling period, I 0,s is the indicator variable of the server s in the initial state. If the server is already powered on, then I 0,s is 1.

[0054] Formula (9) means that within the time range of the formula, if server s is powered on at the beginning of the scheduling period, then during this period, server s must remain powered on, regardless of whether there is workload assigned to it. This formula is used to ensure that the minimum power-on constraint of the data center server is met.

[0055] The server remains powered on with the following constraints:

[0056]

[0057] In the formula, I n,s is the indicator variable of server s in period n, Y t,s The server s is turned on during period t, and NT is the total time period of the scheduling period.

[0058] Formula (10) means that for each server s, if the server is turned on in time period t, then the MU starting from time t is s During the time period, server s must remain powered on. This condition applies to the time period described in the above formula. Formula (11) means that for each server s, if the server is powered on at time t, then during the time period from time t to the end of the scheduling period, server s must remain powered on.

[0059] The server remains closed with the following constraints:

[0060]

[0061] Where, MD s The minimum OFF time of server s, which indicates the shortest time period that the server must remain in the off state after being shut down. s is the initial OFFperiods of server s, which indicates the time period during which the server has been shut down before scheduling begins. t,s Server s is shut down for period t.

[0062] Formula (12) means that if the server is turned off at the beginning of the scheduling, and the current time t is still within the minimum off time, then the server s must remain turned off at time t. This constraint ensures that the server can remain off for a sufficient time after being turned off to meet the minimum off time requirement, while also avoiding frequent on and off of the server, thereby improving the reliability and efficiency of the server. Formula (13) means that for server s, the MD starting at time t s In a continuous time period, that is, when time t is close to the end of the scheduling period NT, it is ensured that if server s is shut down at time t, then from t to the end of the scheduling period, the server must remain shut down. This constraint ensures that the server will not be reopened before the minimum shutdown time MDs is reached, thereby maintaining the stability and efficiency of the server.

[0063] The minimum server shutdown time constraint is:

[0064]

[0065] Formula (14) means that if server s is shut down at time (given by Z t,s denoted), then the MD starting from t s The server must remain closed for a continuous period of time. This constraint ensures that the server meets the minimum shutdown time requirement when it needs to be shut down, and also avoids frequently starting and shutting down the server in the last stage of the scheduling cycle.

[0066] The server status control start and stop constraints are:

[0067]

[0068] In the formula, I t-1,s is the indicator variable of server s in period t-1.

[0069] Formula (15) and Formula (16) are the start-stop constraints for server state control, which ensure that the server state change matches the number of start-up and shutdown operations, and are important constraints for the logical consistency of server state transitions. For example, in Formula (15), if it is turned on at time t, then Yt,s is 1, otherwise Z t,s =1; Formula (16) means that only one of the operations of opening or closing can be performed in the same time period.

[0070] The workload execution state control start and stop constraints are:

[0071]

[0072] Where Y t,i For workload i in period t, Z t,i For workload i closed during period t, I t-1,i is the indicator variable of workload i in period t-1. The start and stop constraints of workload execution state control are similar to the constraints of server state control.

[0073] The workload distribution constraints are:

[0074]

[0075] Formula (19) means that for any time period t and any workload i, the sum of the number of times workload i is assigned to all servers s in time period t must be equal to the number of times workload i is assigned to at least one server in time period t. Formula (19) ensures that each workload is either assigned to a server or not assigned in each time period. Formula (20) means that for any time period t and any workload i, the number of workloads assigned to server s cannot exceed the product of the total number of workloads NI and the number of server s on states in time period t. This ensures that no workload is assigned to any single server in excess of what the server can handle.

[0076] The task load execution time constraints are:

[0077]

[0078] In the formula, DL i is the deadline time of workload i, TP i is the time period that workload i needs to execute, TO i is the time period that workload i has been executed in the previous scheduling cycle.

[0079] Formula (21) means that for any workload i, the total time allocated to the workload from its execution time to its deadline must be equal to the task execution time it needs. Formula (22) means that for each workload i, after its deadline, the workload should no longer be allocated server resources.

[0080] The workload constraints for continuous execution are:

[0081]

[0082] In the formula, S C A set of workloads that must be executed continuously and cannot be executed in different time periods. n,i is the indicator variable of workload i in period n.

[0083] Formula (23) means that for a set of workloads that are executed continuously, if the server is turned on at the beginning, then workload i needs to be allocated server resources from the first time period to the time period required to complete the task. Formula (24) means that for a set of workloads that are executed continuously, if one of the workloads starts executing at a certain time point, then within a certain time period starting from this time point, sufficient resources must be allocated to this workload to ensure that it can run continuously until completion, and it must be ensured that workload i can only start executing when the startup conditions are met. Formula (25) means that for a set of workloads that are executed continuously, if they start executing near the end of the planned time window, they can still get enough resources to meet their execution time requirements and meet the requirements of the service level agreement (SLA).

[0084] In another exemplary embodiment of the present application, in order to build and solve the model of data center power-computing-heat coordinated optimization, it is necessary to first obtain a series of known data, including the execution status, scheduling status, CPU frequency requirements, deadlines and execution time requirements of the workload. At the same time, it is also necessary to collect the CPU frequency of the server, the power consumption when idle and fully loaded, the upper limit of CPU utilization, and the energy consumption when starting and shutting down. In addition, the capacity margin and electricity market price information of each location in the data center are also necessary. The load factor and energy consumption data of the cooling system are equally important for the accuracy of the model. The operating status of the server, the CPU utilization rate, and the indication signals for startup and shutdown are also factors that need to be considered in the model. These data will be used as input parameters of the model to describe the operating characteristics, power demand, computing task scheduling and thermal management model of the data center in order to build an optimization model with the goal of minimizing the power cost of the data center while ensuring the quality of service and system reliability. The operating data in the above step 103 includes, in addition to the electricity market price during the scheduling period, the following: the execution status of the workload, the scheduling status, the CPU frequency requirement, the deadline, the execution time requirement, the CPU frequency of the server, the power consumption when the server is idle, the power consumption when the server is fully loaded, the upper limit of the CPU utilization of the server, the energy consumption when the server is started, the energy consumption when the server is shut down, the capacity margin of each location in the data center, the load factor of the cooling system, the energy consumption data pair of the cooling system, the operating status of the server, the CPU utilization of the server, the indication signal of the server startup and the indication signal of the server shutdown.

[0085] In another exemplary embodiment of the present application, the data center demand response model can be solved using a commercial mixed integer programming solver such as CPLEX and a branch-cut algorithm. The optimal solution of the data center demand response model provides: 1) the optimal load scheduling from the front-end balancer to the back-end servers at different IDC (Internet Data Center) locations in each time period; 2) the optimal load scheduling from the front-end balancer to the back-end servers at different IDC (Internet Data Center) locations in each time period; 3) the optimal load scheduling from the front-end balancer to the back-end servers at different IDC (Internet Data Center) locations in each time period; 4) the optimal load scheduling from the front-end balancer to the back-end servers at different IDC (Internet Data Center) locations in each time period; 5) the optimal load scheduling from the front-end balancer to the back-end servers at different IDC (Internet Data Center) locations in each time period; 6) the optimal load scheduling from the front-end balancer to the back-end servers at different IDC (Internet Data Center) locations t,i,s ; 2) Each backend server is in Y t,s and Z t,s The optimal wake-up / shutdown decision for each time period (the variables involved in formulas (8)-(14)); 3) Power demand (the result of formula (1)), which is the sum of the energy consumption of local IT computing equipment and cooling energy consumption.

[0086] refer to Figure 3 , the data center demand response capabilities are as follows:

[0087] This application defines the demand response capability of a data center as its time-varying power demand, which is essentially similar to a demand curve that responds to electricity prices at different times. After receiving batch computing jobs from customers, the data center analyzes the computing resources and service quality requirements of the virtual machine instances assigned to each task, coordinates the market electricity price according to the scheduling task, executes the demand management plan, and constructs the data center demand bid to submit to the electricity market. The system operator (ISO) is responsible for clearing the market and sending the electricity price back to the data center. Here, the demand management solution is essentially a complex bidding model equipped with a data center, similar to the price-based unit commitment (PBUC) tool of the Generation Company (GENCO). The above figure shows the role of data center demand management in the market compared to GENCO. In a perfectly competitive market, each consumer can submit his own demand-price-bid curve to the market, just as power generators usually do. However, in practice, only those consumers with self-generation and / or storage devices, not most people, have this capability. Data centers represent a large portion of electricity consumers who can reschedule their consumption through demand management capabilities and have flexibility in when and where to use electricity. By accepting demand-side bids from data centers, the electricity market gives data centers the opportunity to adjust their electricity consumption activities through workload scheduling rather than acting as pure "price takers". Bids from both the supply and demand sides are submitted to the system operator. That is, data center operators will submit demand-side bids to the electricity market using the demand response capabilities of data centers. The operational constraints proposed by both parties will be taken into account in the system operator's market clearing mechanism. Demand-side bidding and supply-side bidding can be used for market clearing in any wholesale electricity forward market, such as day-ahead, balancing, and even real-time markets, with the goal of minimizing operating costs. The system operator will execute a network constraint economic dispatch model to determine the optimal generation dispatch and electricity price.

[0088] Reference Figure 4The goal of this application is to solve the problem that the existing technology fails to achieve the coordination of power, computing power and heat in data centers, and the data center participates in the power market through the demand response mechanism and plays a significant role. The demand response capability of the data center refers to its ability to adjust its power consumption pattern according to the power market price signal to achieve cost savings and grid support. The contributions of this application are summarized as follows: (1) In view of the highly dispersed operating indicators of the data center, such as the power consumption pattern, computing load intensity, cooling system energy consumption and server CPU utilization, this paper establishes a power consumption characteristic model of the data center (such as formula (3)-formula (4)) based on the demand response capability evaluation theory. (2) The present invention establishes a power demand model (such as formula (1)), a computing task scheduling model (such as formula (5)-formula (25)) and a thermal management model (such as formula (2)) of the data center, and optimizes the power consumption and computing resource allocation of the data center through dynamic server integration and virtual machine migration technology. These models characterize the power load characteristics of the data center, which helps to break through the limitations of a single data center in demand response, give play to the complementary advantages of data center clusters, and improve the degree of coupling and correlation between various devices in the data center. (3) The power consumption characteristics of the data center in this application construct a multi-time scale demand response model, including day-ahead and intraday demand response strategies. These models take into account the price signals of the electricity market, optimize the power consumption, computing resource allocation and thermal management of the data center under different electricity prices, and provide important technical support for the creation of a flexible data center demand response system. According to the final demand response strategy, the benefits of data center operators, power grid companies and users are comprehensively evaluated, and reasonable compensation or incentives are given to demand response resources.

[0089] The technical effects of this application are as follows:

[0090] (1) Virtual Machine Live Migration Technology

[0091] Innovation effect: Through virtual machine migration technology, data centers can flexibly move workloads between servers without affecting service quality. This enables data centers to start more servers when electricity prices are low and reduce the number of servers when prices are high, thereby achieving cost savings.

[0092] (2) Electricity-computing-heat synergistic optimization

[0093] Innovative effect: Comprehensive energy management is achieved by taking into account the power consumption, computing task scheduling and thermal management requirements of the data center. This collaborative optimization method improves energy utilization efficiency, reduces the energy consumption of the cooling system, and improves the overall performance of the data center.

[0094] (3) Price-sensitive scheduling

[0095] Innovation effect: Data center workload scheduling not only takes into account the execution requirements of tasks, but also the fluctuations in electricity prices. This price-sensitive scheduling strategy enables data centers to execute more tasks when electricity prices are lower, thereby reducing overall electricity costs.

[0096] (4) Cooling efficiency integration

[0097] Innovation effect: By integrating the efficiency of the cooling system into the demand response model, data centers are able to more effectively manage cooling-related energy consumption. This not only reduces cooling costs, but also improves the data center's ability to adapt to environmental changes.

[0098] (5) Quality of Service (QoS) Guarantee

[0099] Innovation effect: Through the constraints in the model, it is ensured that the data center will not sacrifice the user's service quality when managing power demand. This ensures the user satisfaction of the data center and maintains the market competitiveness of the data center.

[0100] (6) Real-time market participation

[0101] Innovation effect: Data centers can adjust their power demand according to real-time power market price signals, making data centers not only consumers of power but also active participants in the power market. This real-time market participation capability improves the economic efficiency of data centers and may have a positive impact on the stability of the power market.

[0102] The performance indicators of this application are as follows:

[0103] 1. High efficiency: Compared with the traditional data center operation mode, the proposed model can utilize power resources more efficiently and reduce energy waste.

[0104] 2. High precision: Through precise forecasting and scheduling algorithms, the forecast accuracy of electricity demand and supply is improved.

[0105] 3. Strong market adaptability: Data centers can flexibly respond to changes in the electricity market, improving market adaptability and competitiveness.

[0106] Based on the same inventive concept, the embodiment of the present application also provides a data center demand response system that considers power-computing power-heat coordination for implementing the data center demand response method that considers power-computing power-heat coordination involved above. The implementation scheme for solving the problem provided by the system is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more embodiments of the data center demand response system that considers power-computing power-heat coordination provided below can refer to the limitations of the data center demand response method that considers power-computing power-heat coordination above, and will not be repeated here.

[0107] In an exemplary embodiment, a data center demand response system considering power-computing power-heat coordination is provided, including: a monitoring system and a management system. The management system is used to construct an objective function and constraint conditions for data center demand response considering power-computing power-heat coordination, and the objective function and the constraint conditions together constitute a data center demand response model; the objective function takes the minimum total electricity price of Internet devices and cooling equipment during the scheduling period as the goal. The monitoring system is used to collect operating data of the data center other than the power market price during the scheduling period, and transmit it to the management system. The monitoring system is also used to obtain the power market price during the scheduling period from the system operator and transmit it to the management system. The management system is also used to solve the data center demand response model according to the operating data during the scheduling period, and obtain the power-computing power-heat coordination scheduling result during the scheduling period; the power-computing power-heat coordination scheduling result includes: the optimal load scheduling of each server in each scheduling time period, the optimal start-up and shutdown decision and power demand of each server in each scheduling time period.

[0108] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 5As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. Among them, the processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the power-computing power-heat coordination scheduling results during the scheduling period. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a data center demand response method considering power-computing power-heat coordination is implemented.

[0109] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components. In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0110] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0111] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A data center demand response method considering power-computing-heat synergy, characterized in that: include: Constructing an objective function for data center demand response that considers power-computing-heat synergy; the objective function aims to minimize the total electricity price of Internet devices and cooling equipment during the scheduling period; Establishing constraint conditions for data center demand response considering power-computing-heat synergy, and forming a data center demand response model with the objective function and the constraint conditions; Acquiring operation data of the data center during the dispatch period; the operation data includes the electricity market price during the dispatch period; According to the operating data during the scheduling period, the data center demand response model is solved to obtain the power-computing power-heat coordinated scheduling result during the scheduling period; the power-computing power-heat coordinated scheduling result includes: the optimal load scheduling of each server in each scheduling time period, the optimal start and shut down decision and power demand of each server in each scheduling time period.

2. The data center demand response method considering power-computing-heat synergy according to claim 1 is characterized in that: The objective function is: In the formula, F IT,t is the total electricity price of IT equipment in period t, F CL,t is the total electricity price of the cooling equipment in period t, NT is the total time period of the scheduling period, LMP t is the node marginal electricity price in period t, E IT,t is the total power consumption of IT equipment in period t, E CL,t is the total power consumption of the cooling equipment during period t, LF t is the load factor during period t, E IT,t,s is the IT equipment energy consumption of server s in period t, NS is the total number of servers; DT is the total time period, P I,s is the idle cost of server s, I t,s is whether server s is running during period t, P c,s is the full load cost of the server, U t,s is the CPU utilization of server s during period t, E SU,s is the startup cost of server s, Y t,s Server s is turned on for period t, E SD,s is the shutdown cost of server s, Z t,s Server s is shut down for period t, It is a universal quantifier.

3. The data center demand response method considering power-computing-heat synergy according to claim 2 is characterized in that: The method for determining the load factor includes: Obtain the original data of power consumption, computing power and thermal power of the data center in each historical time period; the power consumption data includes: input power and energy consumption of each device in the data center; the computing power data includes: CPU and memory usage of the server, and the load intensity of job processing; the thermal power data includes: operating status of the cooling system, temperature of the cooling system outlet, temperature of the cooling system inlet and heat discharged by the server; Preprocessing the original data of power consumption, computing power and thermal power in each historical time period to obtain the power consumption data, computing power data and thermal power data in each historical time period; the preprocessing includes cleaning, denoising and standardization; Extract the operation characteristics of the data center based on the power consumption data, computing power data, thermal data and operation mode of each historical time period; Based on the operating characteristics of the data center, as well as the power consumption data, computing power data, and thermal data of each historical time period, the ratio of the energy consumption of Internet equipment to the energy consumption of cooling equipment is determined through experiments as the load factor.

4. The data center demand response method considering power-computing-heat synergy according to claim 1 is characterized in that: The constraints include: CPU frequency constraint, Internet data center capacity constraint, server minimum power-on constraint, server power-on state constraint, server power-off state constraint, server minimum power-off time constraint, server state control start-stop constraint, task load execution state control start-stop constraint, workload distribution constraint, workload execution time constraint and continuous execution workload constraint.

5. The data center demand response method considering power-computing-heat synergy according to claim 4 is characterized in that: The CPU frequency constraints are: Where U t,s is the CPU utilization of server s during period t, NI is the total number of workloads, and f i is the CPU frequency requirement of workload i, I t,i,s is the indicator variable of workload i being assigned to server s in period t, f s is the CPU frequency requirement of server s, U max,s is the upper limit of CPU utilization of server s; The Internet data center capacity constraint is: Where NS is the total number of servers, I t,s is the indicator variable of server s in period t, I t,i is the indicator variable of workload i in period t, and CM is the marginal value of the total capacity of the data center.

6. The data center demand response method considering power-computing-heat synergy according to claim 5 is characterized in that: The minimum startup constraint of the server is: In the formula, MU s is the minimum startup time of server s, TU s is the time that server s has been powered on before the scheduling period, I 0,s is the indicator variable of server s in the initial state; The server remains powered on with the following constraints: In the formula, I n,s is the indicator variable of server s in period n, Y t,s The server s is turned on during period t, and NT is the total time period of the scheduling period; The server remains closed with the following constraints: Where, MD s is the minimum shutdown time of server s, TD s is the initial shutdown period of server s, Z t,s Server s is shut down for period t; The minimum server shutdown time constraint is: The server status control start and stop constraints are: In the formula, I t-1,s is the indicator variable of server s in period t-1.

7. The data center demand response method considering power-computing-heat synergy according to claim 6 is characterized in that: The workload execution state control start and stop constraints are: Where Y t,i For workload i in period t, Z t,i For workload i closed during period t, I t-1,i is the indicator variable of workload i in period t-1; The workload distribution constraints are: The task load execution time constraint is: In the formula, DL i is the deadline of workload i, TP i is the time period that workload i needs to execute, TO i is the time period that workload i has been executed in the previous scheduling cycle; The workload constraints for the continuous execution are: In the formula, S C is a set of workloads that must be executed continuously, I n,i is the indicator variable of workload i in period n.

8. The data center demand response method considering power-computing-heat synergy according to claim 1 is characterized in that: The operating data also includes: the execution status of the workload, the scheduling status, the CPU frequency requirement, the deadline, the execution time requirement, the CPU frequency of the server, the power consumption when the server is idle, the power consumption when the server is fully loaded, the CPU utilization upper limit of the server, the energy consumption when the server is started, the energy consumption when the server is shut down, the capacity margin of each location in the data center, the load factor of the cooling system, the energy consumption data pair of the cooling system, the operating status of the server, the CPU utilization of the server, the indication signal of the server startup and the indication signal of the server shutdown.

9. A data center demand response system considering power-computing-heat synergy, characterized in that: include: Monitoring systems and management systems; The management system is used to construct an objective function and constraint conditions for data center demand response that considers power-computing-heat synergy, and the objective function and the constraint conditions together constitute a data center demand response model; the objective function takes the minimum total electricity price of Internet devices and cooling equipment during the scheduling period as the goal; The monitoring system is used to collect the operation data of the data center other than the power market price during the dispatch period and transmit it to the management system; The monitoring system is also used to obtain the electricity market price during the dispatch period from the system operator and transmit it to the management system; The management system is also used to solve the data center demand response model according to the operating data during the scheduling period, and obtain the power-computing power-heat coordinated scheduling results during the scheduling period; the power-computing power-heat coordinated scheduling results include: the optimal load scheduling of each server in each scheduling time period, the optimal start and shut down decision and power demand of each server in each scheduling time period.

10. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the data center demand response method considering power-computing power-heat synergy as described in any one of claims 1-8.

Images