Quantitative characterization method for computing power-thermodynamic collaborative flexibility of data center

Abstract

The invention discloses a quantitative characterization method for computing power-thermodynamic collaborative flexibility of a data center. The quantitative characterization method comprises the following steps: acquiring real-time state data of the data center; on the basis of a thermodynamic dynamic equation, a coupling constraint model containing computing power adjusting power, a task schedulable time window and chip temperature change is established, the coupling constraint model reflects dynamic responses of different cooling modes by introducing thermal inertia parameters, and a temperature safety interval and a temperature change rate upper limit are set as hard constraints; based on the coupling constraint model, the maximum up-regulation flexibility and the maximum down-regulation flexibility which can be provided by the data center under the condition that the temperature safety constraint is met are calculated respectively, and adjustable IT load power ranges corresponding to different durations are obtained; and integrating the up-regulation flexibility and the down-regulation flexibility into a collaborative representation result containing a time-power-temperature three-dimensional relationship, and outputting the collaborative representation result in a structured parameter set form. According to the invention, the structured measurement of the flexible resources of the data center is realized.

Description

Technical Field

[0001] This invention relates to the field of data center energy management and intelligent optimization technology, and in particular to a quantitative characterization method for the synergistic flexibility of data center computing power and thermal power. Background Technology

[0002] With the rapid development of artificial intelligence, big data, and cloud computing technologies, data centers have become a critical infrastructure supporting the development of the digital economy. According to the International Energy Agency (IEA), data center energy consumption accounted for approximately 1.5% of global electricity consumption in 2024, and this proportion continues to rise. Data center power consumption mainly consists of information technology (IT) equipment, cooling equipment, and auxiliary equipment, with IT equipment and cooling equipment accounting for the vast majority of energy consumption. Therefore, fully exploring the potential for coordinated scheduling between IT loads and cooling systems is a key path to improving energy efficiency and reducing operating costs.

[0003] In existing technologies, energy efficiency optimization methods for data centers mainly focus on two independent dimensions: task load scheduling and thermal environment control. A quantitative mechanism for the synergistic flexibility between computing power time transfer capability and the dynamic response characteristics of the thermal system has not yet been established. For example, the patent "Data Center Energy Management Method Based on Load Flexibility Mining" (CN119209470A) proposes migrating latency-tolerant batch processing tasks to low-electricity-price periods based on differences in task latency sensitivity to reduce electricity procurement costs. This method constructs an economic scheduling model with the goal of minimizing electricity costs and introduces service level agreement constraints to ensure timely task completion. However, this scheme only uses the time schedulability of tasks as a degree of scheduling freedom, without explicitly modeling and outputting the synergistic flexibility capacity of "the maximum amount of computing load that can be migrated under the premise of thermal safety." Furthermore, it simplifies the cooling system as a fixed or static energy-consuming unit, failing to consider the actual impact of instantaneous heat load surges caused by concentrated task execution on cooling energy consumption, and also failing to model the dynamic evolution of room or server temperatures. Therefore, in actual operation, the migration of a large number of tasks to low-cost periods may lead to local overheating, forcing the cooling system to operate under overload. This not only partially offsets the benefits of power saving, but may also threaten the thermal safety of the chip, reflecting a lack of quantitative understanding of the synergistic boundary between computing power and thermal power.

[0004] On the other hand, the patent "A Data Center Energy Management Method with Computing Power-Thermal Flexibility Coordination" (CN116755336A) attempts to introduce dynamic constraints on the thermal environment into scheduling. By establishing a cold / hot aisle temperature evolution model and combining outdoor temperature and fresh air system parameters, it optimizes server load distribution and cooling equipment output. This method sets an upper limit for hot aisle temperature and a single-period temperature rise rate limit, and solves the problem with the goal of minimizing total operating costs. Although it has initially achieved coupled modeling of computing power and thermal forces, it still has significant shortcomings: First, although it introduces task tolerance time, it does not treat "adjustable load" as an independent output indicator, and flexibility is still implicitly embedded in the scheduling results, which cannot be directly invoked by external systems; second, although its thermal model includes temperature change rate constraints, it does not further define the maximum adjustable thermal state space under the temperature flexibility operating range, making it difficult to standardize the characterization of thermal flexibility; finally, this method is essentially still a scheduling execution layer technology and does not provide a universal collaborative flexibility quantification framework independent of specific optimization solvers, making it difficult to support flexibility evaluation and comparison across scenarios and architectures.

[0005] Furthermore, recent research, such as "Optimization Strategy for Collaborative Optimization of Computing Power, Power, and Heat in Edge Data Center Clusters," has expanded to multi-dimensional collaboration of power, computing power, and heat, introducing mechanisms such as UPS energy storage and waste heat recovery, and constructing a refined model that includes spatiotemporal migration of data loads and heat pump regulation. However, its core objective remains generating specific sequences of operating instructions, rather than parsing or outputting the overall flexibility capacity that the cluster can provide under given constraints. All existing works treat flexibility as an intermediate variable in the scheduling process, rather than an independently measurable and reusable resource attribute.

[0006] In summary, current mainstream technical solutions generally prioritize scheduling over metering. They either focus on the time dimension of task scheduling while neglecting the actual dynamic response and temperature flexibility boundaries of the thermal system, or they focus on thermal environment control without transforming task migration potential into quantifiable collaborative adjustment capabilities. Both lack explicit definitions, modeling, and outputs of computing power-thermal synergy flexibility. This results in data centers lacking a unified, comparable, and reusable flexibility measurement benchmark when participating in cloud-edge collaboration, energy efficiency assessment, or multi-level optimization, severely restricting the efficient release of internal resource potential and intelligent collaboration with external systems. Summary of the Invention

[0007] To address the technical problems existing in the prior art, this invention proposes a quantitative characterization method for the computing power-thermal synergy flexibility of data centers. By constructing a coupled model that integrates task time window constraints and thermodynamic dynamic evolution, and under the premise of strictly meeting service level agreements and equipment thermal safety, it provides standardized and reusable quantitative inputs for upper-layer scheduling, demand response, or resource trading platforms, thereby releasing the synergistic value of diverse flexible resources within the data center.

[0008] To achieve the above objectives, the present invention provides a method for quantitatively characterizing the computing power-thermal coordination flexibility of data centers, comprising:

[0009] Collect real-time status data of the data center, including computing load information, server chip temperature, cooling system type, ambient temperature, and maximum tolerable latency time for latency-tolerant tasks.

[0010] Based on the thermodynamic dynamic equation, a coupled constraint model is established that includes the power adjustment of computing power, the schedulable time window of the task and the chip temperature change. The coupled constraint model reflects the dynamic response of different cooling methods by introducing thermal inertia parameters, and sets a safe temperature range and an upper limit of temperature change rate as hard constraints.

[0011] Based on the coupling constraint model, the maximum upward and downward adjustment flexibility that the data center can provide under the temperature safety constraint are calculated respectively, and the adjustable IT load power range corresponding to different durations is obtained.

[0012] The up-adjustment flexibility and down-adjustment flexibility are integrated into a co-characterization result that includes the three-dimensional relationship of time-power-temperature, and output in the form of a structured parameter set.

[0013] Preferably, the computing load information includes real-time interactive tasks and latency-tolerant tasks, wherein the latency-tolerant tasks participate in the quantitative calculation of computing power flexibility, and the implementation time is adjusted within its maximum tolerable latency time.

[0014] Preferably, the thermal inertia parameters in the coupled constraint model are dynamically set according to the cooling system type, wherein the cooling system type includes at least one of air cooling, liquid cooling, and hybrid cooling.

[0015] Preferably, the maximum upward adjustment flexibility and the maximum downward adjustment flexibility are represented in the form of a bidirectional flexibility curve, where the horizontal axis is the adjustment duration and the vertical axis is the adjustable power, with the upper region of the curve corresponding to the upward adjustment capability and the lower region corresponding to the downward adjustment capability.

[0016] Preferably, the temperature evolution equation in the coupled constraint model is:

[0017] ;

[0018] In the formula, T t P represents the chip or hot channel temperature during time period t. base ΔP is the reference IT load power. t Let α, β, and γ be the adjustable load power increment to be determined, and let α, β, and γ be the thermodynamic parameters determined by the cooling architecture and environmental conditions, respectively. This refers to the chip or hot channel temperature during time period t+1.

[0019] Preferably, the collaborative characterization results are used to support at least one application among data center energy efficiency assessment, power demand response bidding, cloud-edge collaborative task scheduling, and virtual power plant aggregation optimization.

[0020] Preferably, the structured parameter set includes the maximum up-adjustment power, the maximum down-adjustment power, the corresponding duration window, the implicit temperature evolution path, and the time resources occupied by the delay-tolerant task.

[0021] The present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the quantitative characterization method for the data center computing power-thermal synergy flexibility.

[0022] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the quantitative characterization method for the data center computing power-thermal synergy flexibility.

[0023] Compared with the prior art, the present invention has the following advantages and technical effects:

[0024] (1) This invention proposes to tightly couple the schedulable time window of the delay-tolerant task with the thermal dynamic safety boundary of the server chip on the same time axis, and to characterize the set of feasible load adjustment actions under the dual constraints of service level agreement and device thermal safety, thereby providing a physically realizable decision space for flexible solution and avoiding the risk of over-adjustment or under-adjustment caused by ignoring thermal response characteristics.

[0025] (2) This invention solves the bidirectional flexibility boundary of maximum upscaling and maximum downscaling respectively, and generates a structured flexibility capability curve with duration as independent variable and adjustable IT load power as dependent variable. This makes the adjustment potential of the data center no longer dependent on specific scheduling targets or algorithms, but output independently in the form of a standardized parameter set (including power amplitude, duration and task resource occupation), which significantly improves the reusability and cross-scenario comparability of flexibility resources.

[0026] (3) The characterization results are naturally compatible with the thermal inertia differences under different cooling architectures (such as air cooling and liquid cooling), and by setting the junction temperature limit and temperature change rate as dual hard constraints, the dynamic adjustable boundary of the thermal system is truly reflected, thereby ensuring that the quantitative results are both engineering feasible and physical accurate, and providing a highly reliable input interface for upper-level applications such as energy efficiency assessment, demand response bidding, and cloud-edge collaborative scheduling.

[0027] (4) The “time-power-temperature” three-dimensional collaborative characterization framework formed by the present invention is not only conducive to the unified measurement of diverse and flexible resources within the data center, but also lays a standardized technical foundation for future participation in the power market, virtual power plant aggregation or multi-level energy-information collaborative optimization, and has outstanding practical value and promotion prospects. Attached Figure Description

[0028] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0029] Figure 1 This is a flowchart illustrating a method for quantitatively characterizing the collaborative flexibility of data center computing power and thermal power, according to an embodiment of the present invention.

[0030] Figure 2 This is a schematic diagram illustrating the coupling constraints between server task scheduling and chip temperature and power consumption in an embodiment of the present invention.

[0031] Figure 3 This is a schematic diagram of the data center time-power-temperature three-dimensional collaborative characterization flexibility curve output by an embodiment of the present invention. Detailed Implementation

[0032] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0033] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0034] Existing data center scheduling methods face two main technical challenges in terms of flexibility:

[0035] (1) Lack of a quantitative mechanism for the synergistic flexibility of computing power and thermal management. Existing technologies typically treat task time transfer capability or thermal state adjustment capability as implicit decision variables in the scheduling optimization process. They do not construct an independent characterization model for the maximum computing load adjustment capability that a data center can provide under the conditions of meeting service level agreements and equipment thermal safety constraints, nor do they output corresponding synergistic flexibility capacity indicators. Neither migration strategies based on task latency tolerance characteristics nor load allocation methods combined with temperature dynamic response characteristics can provide reusable and comparable flexibility measurement results, making it difficult for upper-level management systems to effectively assess or utilize the actual adjustment potential of the data center.

[0036] (2) The flexibility of the thermal side is difficult to support precise coordinated control. Most existing solutions only set static high temperature thresholds or simple temperature change rate limits, which do not fully characterize the temperature flexibility of the server within the safe operating range, nor do they consider the differences in thermal inertial response under different cooling architectures such as air cooling and liquid cooling. This simplified thermal model cannot accurately reflect the adjustable boundary of the thermal system under dynamic load, which in turn distorts the quantitative results of computing power-thermal coordinated flexibility and limits its practical application value in energy efficiency assessment, cloud-edge collaboration or multi-level optimization.

[0037] This embodiment aims to address the aforementioned issues by proposing a quantitative characterization method for data center flexibility oriented towards computing power-thermal synergy. This method constructs a coupled model that integrates task time window constraints with thermodynamic dynamic evolution. Under the premise of strictly meeting service level agreements and equipment thermal safety, it provides standardized and reusable quantitative inputs for upper-layer scheduling, demand response, or resource trading platforms, thereby unlocking the synergistic value of diverse flexibility resources within the data center.

[0038] A quantitative characterization method for the computing power-thermal energy synergy flexibility of data centers includes:

[0039] Collect real-time status data of the data center, including computing load information, server chip temperature, cooling system type, ambient temperature, and maximum tolerable latency time for latency-tolerant tasks.

[0040] Based on the thermodynamic dynamic equation, a coupled constraint model is established that includes the power adjustment of computing power, the schedulable time window of the task and the chip temperature change. The coupled constraint model reflects the dynamic response of different cooling methods by introducing thermal inertia parameters, and sets a safe temperature range and an upper limit of temperature change rate as hard constraints.

[0041] Based on the coupling constraint model, the maximum upward and downward adjustment flexibility that the data center can provide under the temperature safety constraint are calculated respectively, and the adjustable IT load power range corresponding to different durations is obtained.

[0042] The up-adjustment flexibility and down-adjustment flexibility are integrated into a co-characterization result that includes the three-dimensional relationship of time-power-temperature, and output in the form of a structured parameter set.

[0043] Specifically, the core of this embodiment lies in constructing a three-dimensional collaborative characterization framework of time-power-temperature to achieve structured measurement of data center flexibility resources.

[0044] Furthermore, the computing load information includes real-time interactive tasks and latency-tolerant tasks, wherein the latency-tolerant tasks participate in the quantitative calculation of computing power flexibility, and the implementation time is adjusted within its maximum tolerable latency time.

[0045] Specifically, this embodiment obtains the current operating status parameters of the data center, including the task queue composition, real-time temperature of the server or chip, cooling system architecture type, outdoor ambient temperature, and the maximum tolerable latency time for latency-tolerant tasks as specified in the service level agreement.

[0046] The cooling system architecture types include air cooling, liquid cooling, or air-liquid hybrid cooling modes. The coupling constraint model introduces a corresponding thermal inertia coefficient according to the cooling architecture type to reflect the temperature response characteristics under different cooling methods.

[0047] The task queue consists of real-time interactive tasks and latency-tolerant tasks, with only latency-tolerant tasks being assigned a schedulable time window and participating in the construction of the coupling constraint model as a flexibility boundary on the computing power side.

[0048] Furthermore, the maximum upward adjustment flexibility and the maximum downward adjustment flexibility are represented by a bidirectional flexibility curve, where the horizontal axis represents the adjustment duration and the vertical axis represents the adjustable power. The area above the curve corresponds to the upward adjustment capability, and the area below the curve corresponds to the downward adjustment capability.

[0049] Specifically, for increasing flexibility, the maximum IT load power that can be superimposed under different durations is calculated, provided that the temperature does not exceed the safe upper limit and the temperature rise rate is limited; for decreasing flexibility, the maximum IT load power that can be reduced under different durations is calculated, provided that the temperature does not fall below the safe lower limit.

[0050] Finally, the above results are integrated into a standardized two-way flexibility capability curve, forming a complete three-dimensional collaborative characterization result of time-power-temperature, and output in the form of a parameter set, including the maximum up-adjustment power, the maximum down-adjustment power, the corresponding duration window, the implicit temperature evolution path, and the task delay tolerance resources occupied.

[0051] Furthermore, the temperature evolution equation in the coupled constraint model is:

[0052] ;

[0053] In the formula, T t P represents the chip or hot channel temperature during time period t. base ΔP is the reference IT load power. t Let α, β, and γ be the adjustable load power increment to be determined, and let α, β, and γ be the thermodynamic parameters determined by the cooling architecture and environmental conditions, respectively. This refers to the chip or hot channel temperature during time period t+1.

[0054] Furthermore, the collaborative characterization results are used to support at least one application in data center energy efficiency assessment, power demand response bidding, cloud-edge collaborative task scheduling, and virtual power plant aggregation optimization.

[0055] Specifically, the collaborative quantitative characterization results are independent of specific scheduling algorithms and can be directly used as a flexible product interface for upper-layer cloud resource scheduling platforms, energy efficiency assessment systems, or multi-level collaborative optimization frameworks to achieve unified measurement, efficient reuse, and cross-scenario comparison of the computing power-thermal synergy potential of data centers.

[0056] Furthermore, the structured parameter set includes the maximum up-adjustment power, the maximum down-adjustment power, the corresponding duration window, the implicit temperature evolution path, and the time resources occupied by the delay-tolerant task.

[0057] This embodiment also provides an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it realizes a method for quantitatively characterizing the computing power-thermal synergy flexibility of a data center.

[0058] This embodiment also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, realizes a method for quantitatively characterizing the computing power-thermal synergy flexibility of a data center.

[0059] This embodiment first obtains the current operating status of the data center, including the real-time temperature of the server or key chip, the queue of tasks to be processed (distinguishing between real-time interactive tasks and latency-tolerant tasks), the cooling system architecture type (such as air cooling, liquid cooling or hybrid mode), the outdoor ambient temperature, and the maximum tolerable latency time of the task as specified in the service level agreement.

[0060] Secondly, based on the principle of thermodynamic dynamic evolution, a coupling constraint model is established among computing power adjustment power, adjustable time window and thermal state temperature. The temperature state equation introduces corresponding thermal inertia parameters according to the differences in cooling architecture, and sets the safe operating range of chip junction temperature and the maximum temperature change rate in a single time period as hard boundaries.

[0061] Based on this, we solve for the upward and downward flexibility that the data center can achieve in the future scheduling cycle: For upward flexibility, under the premise that the temperature does not exceed the safe upper limit and the temperature rise rate is limited, we calculate the maximum IT load power that can be superimposed under different durations; For downward flexibility, under the premise that the temperature does not fall below the safe lower limit, we calculate the maximum IT load power that can be reduced under different durations.

[0062] Finally, the above results are integrated into a standardized two-way flexibility capability curve, forming a complete three-dimensional collaborative characterization result of time-power-temperature, and output in the form of a parameter set, including the maximum up-adjustment power, the maximum down-adjustment power, the corresponding duration window, the implicit temperature evolution path, and the task delay tolerance resources occupied.

[0063] The collaborative quantitative characterization results are independent of specific scheduling algorithms and can be directly used as a flexible product interface for upper-layer cloud resource scheduling platforms, energy efficiency assessment systems, or multi-level collaborative optimization frameworks to achieve unified measurement, efficient reuse, and cross-scenario comparison of the computing power-thermal synergy potential of data centers.

[0064] To more clearly illustrate the technical solution of the present invention, specific embodiments are provided below for description:

[0065] See Figure 1 The diagram shows an overall flowchart of a method for quantitatively characterizing the flexibility of data center computing power-thermal coordination. The method includes the following steps:

[0066] First, obtain the current operating status parameters of the data center, including task queue composition, real-time server temperature, cooling architecture type, outdoor ambient temperature, and the maximum tolerable latency time of tasks as specified in the service level agreement.

[0067] Secondly, a coupled model of computing power adjustment and thermal state evolution is constructed based on thermodynamic principles, and corresponding constraint spaces are defined for key parameters.

[0068] Then, solve the bidirectional flexibility boundaries for upward and downward adjustment respectively, and determine whether it is the upward direction of flexibility. If it is the upward direction, calculate the maximum upward flexibility; otherwise, calculate the maximum downward flexibility.

[0069] Finally, the above parameters are integrated into a standardized set of flexibility parameters, and the flexibility-related parameters and parameter interfaces that are easy for external systems to call are output.

[0070] Further, see Figure 2 The diagram shows the coupling constraints between server task scheduling and chip temperature and power consumption. Figure 2The horizontal axis represents time, and the upper rectangular bar chart represents the workload the server needs to handle. The shaded diagonal lines within the rectangles represent interactive workloads, while the shaded horizontal lines represent batch workloads. The portion between the task arrival time and the task deadline represents the schedulable time window for the task. Batch workloads are latency-tolerant tasks, assigned an executable interval from the current moment until the end of the maximum tolerable latency time, within which the execution time can be freely chosen. Interactive workloads, on the other hand, must be processed promptly and cannot be delayed. Simultaneously, the server temperature must always be maintained within a preset safe operating range, and the rate of temperature change between adjacent time periods must not exceed physical limits. The vertical axis shows the distribution of server chip core temperature and power consumption. The dark line at the top represents chip power consumption, the light line represents chip core temperature, and the two horizontal dashed lines represent the upper and lower limits of chip temperature, respectively. When a latency-tolerant task arrives at the server at a certain moment, if the server chooses to delay processing, the chip core temperature and power consumption will not fluctuate significantly and will remain unchanged. As time progresses, when a task load is delayed and then resumed, the chip power consumption will increase dramatically. Due to thermal inertia, the chip core temperature will rise slowly, and the delayed rise in core temperature will lead to a gradual increase in energy consumption. As time continues, some workloads are processed, allowing the core temperature to decrease simultaneously, and energy consumption will also decrease accordingly. If an interactive load suddenly arrives at this time, the chip power consumption will again increase dramatically, and due to thermal inertia, the chip core temperature will rise slowly. As time progresses, interactive and batch processing loads will be processed less and less, and the chip power consumption and core temperature will slowly decrease. Therefore, the adjustable range of chip temperature between its upper and lower limits, achievable through the cooling system, represents the theoretical thermal flexibility of the server, while the flexibly adjustable batch processing load represents computing power flexibility. This embodiment couples the task scheduling degree of freedom with the dynamic temperature response path on the same time axis, ensuring that any load adjustment decision simultaneously meets business timing requirements and thermal safety boundaries, thereby providing a physically feasible domain for flexibility quantification.

[0071] Based on this, see Figure 3 The diagram shows a three-dimensional coordinated flexibility capability curve based on time, power, and temperature. Figure 3 In the diagram, the horizontal axis represents the duration, the vertical axis represents the adjustable IT load power, and the vertical axis represents the server chip temperature. Figure 3The upper and lower boundary lines represent the maximum upward and downward flexibility achievable by the data center under current operating conditions. Each point on each boundary line represents a feasible adjustment scheme, corresponding to a temperature evolution path that satisfies thermal safety constraints. To more intuitively illustrate the possible range of these adjustment schemes, grayscale shading is used to fill the gap between the upward and downward flexibility paths, forming a closed adjustable region. This region not only reflects the potential adjustment space of the data center's computing power-thermal synergy but also clearly indicates the temperature change trend with time and load power adjustments, thus providing a structured and easily understood three-dimensional measurement framework. This time-power-temperature three-dimensional synergy characterization result, as the output of this embodiment, can be called by upper-layer systems in a standardized interface for resource assessment, market bidding, or multi-level collaborative optimization, significantly improving the accuracy and reliability of data center flexibility resource utilization.

[0072] In summary, this embodiment, through the above process, achieves an explicit, structured, and reusable quantitative characterization of the computing power-thermal power synergy flexibility of data centers. It effectively fills the technical gap in the prior art where flexibility resources cannot be independently measured, and provides key technical support for data centers to participate in intelligent collaborative operation efficiently.

[0073] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for quantitatively characterizing the computing power-thermal synergy flexibility of data centers, characterized in that, include: Collect real-time status data of the data center, including computing load information, server chip temperature, cooling system type, ambient temperature, and maximum tolerable latency time for latency-tolerant tasks. Based on the thermodynamic dynamic equation, a coupled constraint model is established that includes the power adjustment of computing power, the schedulable time window of the task and the chip temperature change. The coupled constraint model reflects the dynamic response of different cooling methods by introducing thermal inertia parameters, and sets a safe temperature range and an upper limit of temperature change rate as hard constraints. Based on the coupling constraint model, the maximum upward and downward adjustment flexibility that the data center can provide under the temperature safety constraint are calculated respectively, and the adjustable IT load power range corresponding to different durations is obtained. The up-adjustment flexibility and down-adjustment flexibility are integrated into a co-characterization result that includes the three-dimensional relationship of time-power-temperature, and output in the form of a structured parameter set.

2. The method for quantitatively characterizing the computing power-thermal synergy flexibility of data centers according to claim 1, characterized in that, The computational load information includes real-time interactive tasks and latency-tolerant tasks, wherein the latency-tolerant tasks participate in the quantitative calculation of computing power flexibility, and the implementation time is adjusted within its maximum tolerable latency time.

3. The method for quantitatively characterizing the computing power-thermal synergy flexibility of data centers according to claim 1, characterized in that, The thermal inertia parameters in the coupled constraint model are dynamically set according to the cooling system type, wherein the cooling system type includes at least one of air cooling, liquid cooling, and hybrid cooling.

4. The method for quantitatively characterizing the computing power-thermal synergy flexibility of data centers according to claim 1, characterized in that, The maximum upward and downward adjustment flexibility are represented by a bidirectional flexibility curve, where the horizontal axis represents the adjustment duration and the vertical axis represents the adjustable power. The area above the curve corresponds to the upward adjustment capability, and the area below the curve corresponds to the downward adjustment capability.

5. The method for quantitatively characterizing the computing power-thermal synergy flexibility of data centers according to claim 1, characterized in that, The temperature evolution equation in the coupled constraint model is: ； In the formula, T t P represents the chip or hot channel temperature during time period t. base ΔP is the reference IT load power. t Let α, β, and γ be the adjustable load power increment to be determined, and let α, β, and γ be the thermodynamic parameters determined by the cooling architecture and environmental conditions, respectively. This refers to the chip or hot channel temperature during time period t+1.

6. The method for quantitatively characterizing the computing power-thermal synergy flexibility of data centers according to claim 1, characterized in that, The collaborative characterization results are used to support at least one application in data center energy efficiency assessment, power demand response bidding, cloud-edge collaborative task scheduling, and virtual power plant aggregation optimization.

7. The method for quantitatively characterizing the computing power-thermal synergy flexibility of data centers according to claim 1, characterized in that, The structured parameter set includes the maximum up-adjustment power, the maximum down-adjustment power, the corresponding duration window, the implicit temperature evolution path, and the time resources occupied by the delay-tolerant task.

8. An electronic device, characterized in that, It includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement a quantitative characterization method for the data center computing power-thermal synergy flexibility as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The system stores a computer program, which, when executed by a processor, implements a method for quantitatively characterizing the data center computing power-thermal synergy flexibility as described in any one of claims 1-7.

Images