Heterogeneous data center-oriented low-carbon load directrix generation and decomposition method

Abstract

The invention discloses a heterogeneous data center-oriented low-carbon load directrix generation and decomposition method, and relates to the field of load regulation, and the method comprises the steps: constructing a nonlinear energy consumption model of IT equipment in a heterogeneous data center, and constructing a refrigeration power consumption model; constructing a virtual power node model about the computing power regulation quantity of the heterogeneous data center; constructing a cost minimization model of the heterogeneous data center based on the cost information of the power distribution network, and determining a low-carbon load directrix based on a target computing power adjusting quantity obtained by solving the cost minimization model and a virtual power node model; and constructing a directrix decomposition optimization model based on the individual execution directrix, and solving the directrix decomposition optimization model based on the low-carbon load directrix to obtain a target individual execution directrix, so that each heterogeneous data center executes a corresponding low-carbon load adjustment operation based on the target individual execution directrix. The method can guide the data center load to participate in the low-carbon scheduling of the power distribution network.

Description

Technical Field

[0001] This invention relates to the field of load regulation, and in particular to a method for generating and decomposing low-carbon load baselines for heterogeneous data centers. Background Technology

[0002] Data centers, as high-energy-consuming flexible loads, possess the spatiotemporal portability of computing power tasks, making them a crucial flexible resource for demand response in power distribution networks. Quasilinear demand response guides load tracking and regulation by publishing target load quasilinearities, and virtual power node modeling can abstract the relationship between computing power load and grid power regulation, thus becoming an important technical direction for data centers to participate in grid dispatch.

[0003] Existing data center demand response technologies still have significant shortcomings. Most solutions employ electricity price-driven or event-based response models, lacking a continuous and stable load baseline as a unified control benchmark, and failing to adequately characterize the intrinsic relationship between computing power and electricity consumption. Related research often simplifies data centers as conventional adjustable loads, using linear ratios or empirical coefficients to characterize power regulation characteristics, neglecting the coupled effects of multiple factors such as server utilization and cooling systems. This fails to accurately reflect nonlinear energy consumption patterns, easily leading to biased capacity estimation and mismatches between control commands and actual operation. Furthermore, existing load baseline formulations often prioritize economic benefits or system supply-demand balance, failing to embed tiered carbon trading costs as a core variable, thus lacking a clear low-carbon orientation. In baseline allocation, the heterogeneous differences in scale, energy efficiency, and control costs among multiple data centers are not fully considered, often employing uniform allocation or simple aggregation methods, making it difficult to achieve non-uniform and equitable decomposition of the overall baseline, thus reducing the accuracy and reliability of demand response execution.

[0004] Therefore, how to achieve accurate mapping and optimized allocation of system-level low-carbon target standards to the execution standards of each data center, so as to effectively guide the data center load to actively participate in the low-carbon scheduling of the power distribution network, is a technical problem that urgently needs to be solved. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method for generating and decomposing low-carbon load baselines for heterogeneous data centers. This method enables precise mapping and optimized allocation of system-level low-carbon target baselines to execution baselines of each data center, thereby effectively guiding data center loads to actively participate in the low-carbon scheduling of the power distribution network. The specific solution is as follows: Firstly, this application provides a method for generating and decomposing low-carbon load baselines for heterogeneous data centers, including: Based on the computing power adjustment of the heterogeneous data center, a nonlinear energy consumption model of the IT equipment in the heterogeneous data center is constructed. The target energy efficiency ratio of the IT equipment and the cooling system in the heterogeneous data center is determined. The energy efficiency conversion coefficient is determined based on the target energy efficiency ratio. A cooling power consumption model is constructed based on the nonlinear energy consumption model and the target energy efficiency ratio. A virtual power node model for the computing power adjustment is constructed based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model; the virtual power node model is a standardized equivalent model presented to the power distribution network after abstracting and encapsulating the internal characteristics of heterogeneous data centers. Based on the target cost of the power distribution network within a preset computing power scheduling cycle, a cost minimization model for the heterogeneous data center is constructed. Based on the target computing power adjustment amount obtained from solving the cost minimization model and the virtual power node model, the low-carbon load guideline for the heterogeneous data center is determined. Define individual execution guidelines for the heterogeneous data centers to achieve preset low-carbon load targets. Construct a guideline decomposition and optimization model based on the individual execution guidelines, and solve the guideline decomposition and optimization model based on the low-carbon load guidelines to obtain the target individual execution guidelines for each heterogeneous data center. Determine computing power task migration instructions based on the target individual execution guidelines so that each heterogeneous data center can perform corresponding low-carbon load adjustment operations based on the computing power task migration instructions.

[0006] Optionally, the construction of a nonlinear energy consumption model for IT equipment in the heterogeneous data center based on the computing power adjustment of the heterogeneous data center includes: The computing load of the heterogeneous data center is determined based on the computing power adjustment amount of the heterogeneous data center and the preset baseline load of the heterogeneous data center. A nonlinear energy consumption model for IT equipment in the heterogeneous data center is constructed based on the idle power consumption of the server cluster in the heterogeneous data center, the preset linear growth coefficient, the preset quadratic coefficient, and the computing load; the preset quadratic coefficient is a coefficient characterizing the degree of energy efficiency degradation of the server cluster under a preset high load.

[0007] Optionally, determining the target energy efficiency ratio of the IT equipment and cooling system in the heterogeneous data center, determining the energy efficiency conversion coefficient based on the target energy efficiency ratio, and constructing a cooling power consumption model based on the nonlinear energy consumption model and the target energy efficiency ratio includes: The target energy efficiency ratio of the IT equipment and cooling system in the heterogeneous data center is determined by a linear temperature-sensitive model based on a preset reference temperature, the energy efficiency ratio corresponding to the preset reference temperature, a preset temperature sensitivity coefficient, and the current ambient temperature. The energy efficiency conversion coefficient is determined based on the target energy efficiency ratio, and a cooling power consumption model is constructed based on the nonlinear energy consumption model and the target energy efficiency ratio.

[0008] Optionally, the computing power adjustment amount is a variable that satisfies the constraints of computing power adjustment capability, the constraints of total task conservation, and the constraints of physical capacity security.

[0009] Optionally, the step of constructing a cost minimization model for the heterogeneous data center based on the target cost of the power distribution network within a preset computing power scheduling period includes: A cost minimization model for the heterogeneous data center is constructed based on the net load cost, power generation cost, computing power compensation cost, power curtailment penalty cost, and tiered carbon trading cost of the distribution network within a preset computing power scheduling cycle.

[0010] Optionally, the process for determining the computing power compensation cost includes: The computing power compensation cost of the heterogeneous data center is determined based on the unit computing power compensation cost of the power distribution network for the heterogeneous data center under a preset computing power task migration unit and the computing power adjustment amount. Furthermore, the process for determining the tiered carbon trading costs includes: The tiered carbon trading cost is determined based on total carbon emissions, carbon allowances, a pre-set reward coefficient for issuing surplus carbon allowances, a pre-set basic carbon trading cost, a punitive carbon trading cost for meeting pre-set excess carbon emission conditions, and the length of the tiered carbon trading range.

[0011] Optionally, determining the low-carbon load baseline of the heterogeneous data center based on the target computing power adjustment amount obtained from solving the cost minimization model and the virtual power node model includes: The conditions under which the computing power adjustment amount satisfies the computing power adjustment capability constraint, the total task conservation constraint, and the physical capacity security constraint are determined as the data center operation constraints. The cost minimization model is solved based on the power balance constraints between the heterogeneous data center and the power distribution network, the unit constraints of the coal-fired or gas-fired units, the operation constraints of the data center, and the line transmission capacity constraints of the power distribution network to obtain the target computing power adjustment amount. The low-carbon load baseline of the heterogeneous data center is then determined using the target computing power adjustment amount and the virtual power node model.

[0012] Optionally, the step of solving the optimization model based on the low-carbon load baseline to obtain the target individual execution baseline for each heterogeneous data center includes: The condition that the sum of the individual execution guidelines of each heterogeneous data center is equal to the low-carbon load guideline is determined as the overall guideline conservation constraint condition. Based on the preset maximum load capacity, the preset maximum load capacity, and the nonlinear energy consumption model, the computing power adjustment range is determined, and the conditions of the individual execution guideline within the computing power adjustment range are determined as individual physical boundary constraints. The guideline decomposition optimization model is solved based on the overall guideline conservation constraint and the individual physical boundary constraint to obtain the target individual execution guideline for each heterogeneous data center.

[0013] Optionally, the constraint on the computing power adjustment capability is to limit the amount of computing power adjustment to be no less than the preset maximum amount of computing power tasks that can be moved into the heterogeneous data center and no more than the preset maximum amount of computing power tasks that can be moved out of the heterogeneous data center.

[0014] Optionally, the total task conservation constraint is a condition that limits the total computing power task quantity related to the computing power adjustment amount to meet the preset task quantity balance condition within a preset computing power scheduling period.

[0015] Optionally, the physical capacity safety constraint is a condition that limits the difference between the computing load and the computing power adjustment amount to not less than zero and not greater than the preset maximum physical processing capacity of the server cluster.

[0016] Secondly, this application provides a low-carbon load baseline generation and decomposition device for heterogeneous data centers, comprising: The first model construction module is used to construct a nonlinear energy consumption model of IT equipment in the heterogeneous data center based on the computing power adjustment amount of the heterogeneous data center, determine the target energy efficiency ratio of the IT equipment and cooling system in the heterogeneous data center, determine the energy efficiency conversion coefficient based on the target energy efficiency ratio, and construct a cooling power consumption model based on the nonlinear energy consumption model and the target energy efficiency ratio. The second model construction module is used to construct a virtual power node model for the computing power adjustment based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model; the virtual power node model is a standardized equivalent model presented to the power distribution network after abstracting and encapsulating the internal characteristics of heterogeneous data centers. The baseline determination module is used to construct a cost minimization model of the heterogeneous data center based on the target cost of the distribution network within a preset computing power scheduling cycle, and to determine the low-carbon load baseline of the heterogeneous data center based on the target computing power adjustment amount obtained by solving the cost minimization model and the virtual power node model. The guideline decomposition module is used to define individual execution guidelines for the heterogeneous data centers to achieve preset low-carbon load targets, construct a guideline decomposition optimization model based on the individual execution guidelines, and solve the guideline decomposition optimization model based on the low-carbon load guidelines to obtain the target individual execution guidelines for each heterogeneous data center. The module then determines computing power task migration instructions based on the target individual execution guidelines, so that each heterogeneous data center can perform corresponding low-carbon load adjustment operations based on the computing power task migration instructions.

[0017] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor is used to execute the computer program to implement the aforementioned method for generating and decomposing low-carbon load baselines for heterogeneous data centers.

[0018] Fourthly, this application provides a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned method for generating and decomposing low-carbon load baselines for heterogeneous data centers.

[0019] In this application, a nonlinear energy consumption model of IT equipment in a heterogeneous data center is constructed based on the computing power adjustment amount. The target energy efficiency ratio (EER) of the IT equipment and cooling system in the heterogeneous data center is determined, and an energy efficiency conversion coefficient is determined based on the target EER. A cooling power consumption model is constructed based on the nonlinear energy consumption model and the target EER. A virtual power node model is constructed based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model, relating to the computing power adjustment amount. The virtual power node model is a standardized equivalent model presented to the power distribution network after abstracting and encapsulating the internal characteristics of the heterogeneous data center. The model is based on the power distribution network within a preset computing power scheduling period. A cost minimization model for the heterogeneous data center is constructed based on the target cost. The low-carbon load baseline for the heterogeneous data center is determined based on the target computing power adjustment amount obtained from solving the cost minimization model and the virtual power node model. Individual execution baselines for achieving the preset low-carbon load target are defined for the heterogeneous data center. A baseline decomposition optimization model is constructed based on the individual execution baselines. The baseline decomposition optimization model is solved based on the low-carbon load baselines to obtain the target individual execution baselines for each heterogeneous data center. Computing power task migration instructions are determined based on the target individual execution baselines so that each heterogeneous data center can perform corresponding low-carbon load adjustment operations based on the computing power task migration instructions. As can be seen from the above, this application constructs a nonlinear energy consumption model for IT equipment based on the computing power adjustment amount of heterogeneous data centers, determines the target energy efficiency ratio and energy efficiency conversion coefficient of IT equipment and cooling systems, and constructs a cooling power consumption model; based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model, a virtual power node model for heterogeneous data centers regarding computing power adjustment amount is established; a cost minimization model for heterogeneous data centers is constructed using the target cost within the preset computing power scheduling cycle of the distribution network, and the target computing power adjustment amount is obtained by solving the model, and the low-carbon load baseline of heterogeneous data centers is determined by combining the virtual power node model; individual execution baselines for heterogeneous data centers to achieve preset low-carbon load targets are defined, a baseline decomposition optimization model is constructed based on the individual execution baselines, and the target individual execution baselines of each data center are obtained by solving the low-carbon load baselines, thereby generating computing power task migration instructions to enable each heterogeneous data center to perform corresponding low-carbon load adjustment operations. In this way, this application can achieve accurate mapping and optimized allocation of system-level low-carbon target baselines to execution baselines of each data center, thereby effectively guiding data center loads to actively participate in the low-carbon scheduling of the distribution network. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is a flowchart of a low-carbon load baseline generation and decomposition method for heterogeneous data centers disclosed in this application; Figure 2 This is a schematic diagram of a low-carbon load baseline generation and decomposition device for heterogeneous data centers disclosed in this application. Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Currently, existing data center demand response technologies still have significant shortcomings. Most solutions adopt electricity price signal-driven or event-based response models, lacking a continuous and stable load baseline as a unified control benchmark, and failing to adequately characterize the intrinsic relationship between computing power and power consumption. Related research often simplifies data centers as conventional adjustable loads, using linear proportions or empirical coefficients to characterize power regulation characteristics, ignoring the coupled effects of multiple factors such as server utilization and cooling systems. This fails to accurately reflect nonlinear energy consumption patterns, easily leading to biased capacity estimation and mismatch between control commands and actual operation. Furthermore, existing load baseline formulations are mostly aimed at economic benefits or system supply-demand balance, failing to embed tiered carbon trading costs as a core variable, and lacking a clear low-carbon orientation. In baseline allocation, the heterogeneous differences in scale, energy efficiency, and control costs among multiple data centers are not fully considered, often employing uniform allocation or simple aggregation methods, making it difficult to achieve non-uniform and fair decomposition of the overall baseline, thus reducing the accuracy and reliability of demand response execution. To this end, this application provides a method for generating and decomposing low-carbon load baselines for heterogeneous data centers, which can achieve accurate mapping and optimized allocation of system-level low-carbon target baselines to execution baselines of each data center, thereby effectively guiding data center loads to actively participate in the low-carbon scheduling of the power distribution network.

[0024] See Figure 1As shown, this invention discloses a method for generating and decomposing low-carbon load baselines for heterogeneous data centers, including: Step S11: Construct a nonlinear energy consumption model for IT equipment in the heterogeneous data center based on the computing power adjustment amount of the heterogeneous data center, determine the target energy efficiency ratio of the IT equipment and cooling system in the heterogeneous data center, determine the energy efficiency conversion coefficient based on the target energy efficiency ratio, and construct a cooling power consumption model based on the nonlinear energy consumption model and the target energy efficiency ratio.

[0025] In this embodiment, a computing power adjustment amount for a heterogeneous data center is defined, and the computing power load of the heterogeneous data center is determined based on the computing power adjustment amount and a preset baseline load of the heterogeneous data center. A nonlinear energy consumption model of the IT equipment in the heterogeneous data center is constructed based on the idle power consumption of the server cluster in the heterogeneous data center, a preset linear growth coefficient, a preset quadratic coefficient, and the computing power load. The preset quadratic coefficient is a coefficient characterizing the degree of energy efficiency degradation of the server cluster under a preset high load. By introducing the quadratic coefficient, the energy consumption characteristics of IT equipment during load changes can be described more accurately.

[0026] Regarding the computing power adjustment amount, it should be noted that the computing power adjustment amount reflects the data center's ability to participate in low-carbon load regulation. To ensure the practical operability of the model subsequently constructed, the computing power adjustment amount in this embodiment is defined as a variable that satisfies multiple constraints.

[0027] Specifically, the computing power adjustment amount is a variable that satisfies the constraints of computing power adjustment capability, total task conservation, and physical capacity safety. The computing power adjustment capability constraint limits the computing power adjustment amount to be no less than the preset maximum number of computing power tasks that can be moved into the heterogeneous data center and no more than the preset maximum number of computing power tasks that can be moved out of the heterogeneous data center. The total task conservation constraint limits the total number of computing power tasks related to the computing power adjustment amount to meet a preset task balance condition within a preset computing power scheduling period. The physical capacity safety constraint limits the difference between the computing power load and the computing power adjustment amount to be no less than zero and no more than the preset maximum physical processing capacity of the server cluster.

[0028] In this embodiment, it is also necessary to determine the target energy efficiency ratio (EER) of the IT equipment and cooling system in the heterogeneous data center. In one specific implementation, the target EER of the IT equipment and cooling system in the heterogeneous data center is determined using a linear temperature-sensitive model based on a preset reference temperature, the EER corresponding to the preset reference temperature, a preset temperature sensitivity coefficient, and the current ambient temperature.

[0029] After determining the target energy efficiency ratio, the energy efficiency conversion coefficient is further determined based on the target energy efficiency ratio, and a cooling power consumption model is constructed based on the nonlinear energy consumption model and the target energy efficiency ratio. Through the models constructed above, an accurate description of the overall power consumption of the heterogeneous data center can be achieved. The various models constructed above lay a foundation for load optimization and adjustment in subsequent steps.

[0030] Step S12: Construct a virtual power node model for the computing power adjustment based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model; the virtual power node model is a standardized equivalent model presented to the power distribution network after abstracting and encapsulating the internal characteristics of heterogeneous data centers.

[0031] In this embodiment, after constructing the nonlinear energy consumption model and cooling power consumption model of the IT equipment and determining the energy efficiency conversion coefficient, a virtual power node model for the computing power adjustment is constructed based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model. The core function of this virtual power node model is to abstract and encapsulate the complex coupling relationship between IT equipment and cooling systems within the heterogeneous data center, the nonlinear energy consumption characteristics, and the influence of ambient temperature, so that from the perspective of the power distribution network, the data center appears as a virtual node with standardized power adjustment characteristics. Through this virtual power node model, the power distribution network dispatch center does not need to focus on the heterogeneous hardware composition and complex operating logic within the data center; it only needs to obtain the power response characteristics of the data center under different computing power adjustment amounts based on this standardized model, thus providing a quantitative basis for subsequent decomposition and scheduling based on the low-carbon load baseline.

[0032] Step S13: Construct a cost minimization model for the heterogeneous data center based on the target cost of the distribution network within a preset computing power scheduling cycle, and determine the low-carbon load baseline of the heterogeneous data center based on the target computing power adjustment amount obtained by solving the cost minimization model and the virtual power node model.

[0033] In this embodiment, in order to guide heterogeneous data centers to participate in low-carbon load regulation, it is necessary to construct a cost minimization model that comprehensively considers multiple economic factors.

[0034] First, determine the net load cost. Determine the net load demand within the target area of ​​the distribution network management system, and based on the real-time time-of-use electricity price of the distribution network and the power obtained from the preset main grid, determine the net load cost corresponding to the net load demand.

[0035] In this embodiment, the power generation cost within the target area is determined. The fuel consumption and operation and maintenance losses of the coal-fired or gas-fired power unit are determined using a preset quadratic function based on the power generation capacity of the coal-fired or gas-fired power unit and a preset consumption characteristic coefficient of the coal-fired or gas-fired power unit. These fuel consumption and operation and maintenance losses are then defined as the power generation cost within the target area.

[0036] To incentivize heterogeneous data centers to provide load balancing services, it is necessary to determine the computing power compensation cost for these data centers. The computing power compensation cost for the heterogeneous data centers is determined based on the unit computing power compensation cost of the power distribution network for the heterogeneous data centers under a preset computing power task migration unit and the computing power balancing amount.

[0037] Secondly, the utilization of renewable energy is also a key factor affecting the low-carbon operation of the system. The utilization rate of renewable energy is determined based on a preset wind curtailment penalty coefficient, a preset maximum generating capacity of renewable energy, and the grid-connected absorption capacity of the distribution network. This utilization rate is then used as the cost of curtailment penalty for the renewable energy.

[0038] In this embodiment, the construction of a tiered carbon trading cost is a core element in achieving low-carbon dispatch. The total carbon emissions within the target area are determined based on the power generation capacity of the coal-fired or gas-fired units, the carbon emission intensity of thermal power units within the target area, the power output obtained from the preset main grid, and the average carbon emission intensity corresponding to the power output. The carbon quota of the distribution network is determined based on the power generation capacity, the power output, and a preset carbon quota benchmark coefficient per unit of electricity. The tiered carbon trading cost is determined based on the total carbon emissions, the carbon quota, a preset reward coefficient for issuing surplus carbon quotas, a preset basic carbon trading cost, a punitive carbon trading cost for meeting preset excess carbon emission conditions, and the length of the tiered carbon trading interval.

[0039] Finally, based on the determination of the above cost components, a cost minimization model for the heterogeneous data center is constructed based on the net load cost of the distribution network within the preset computing power scheduling cycle, the power generation cost, the computing power compensation cost for the heterogeneous data center, the renewable energy curtailment penalty cost, and the tiered carbon trading cost.

[0040] To solve the above cost minimization model, appropriate constraints need to be set.

[0041] First, the total load of the heterogeneous data center is determined based on the virtual power node model, and the target load is determined based on the total load and the preset conventional rigid load other than the heterogeneous data center; the target power is determined based on the power obtained by the distribution network from the preset main network, the power generation of coal-fired or gas-fired units, and the grid-connected absorption power of the distribution network; the condition that the target load and the target power are equal is determined as the power balance constraint condition.

[0042] Secondly, the following conditions are defined as the constraints on the generator set: the power generation of the coal-fired unit or the gas-fired unit is within a preset power range; the grid-connected power consumption is not less than zero and not greater than a preset maximum grid-connected power consumption; and the power change value of the coal-fired unit or the gas-fired unit within a preset time period is within a preset ramp rate range.

[0043] Third, the conditions under which the computing power adjustment amount satisfies the computing power adjustment capability constraint, the total task conservation constraint, and the physical capacity security constraint are determined as the data center operation constraints.

[0044] Fourth, the condition that the active power flowing through the distribution lines of the power distribution network is within the preset active power transmission capacity range is determined as the line transmission capacity constraint condition.

[0045] Finally, based on the power balance constraints, generator unit constraints, data center operation constraints, and line transmission capacity constraints, the cost minimization model is solved to obtain the target computing power adjustment amount. This target computing power adjustment amount and the virtual power node model are then used to determine the low-carbon load baseline for the heterogeneous data center. This low-carbon load baseline reflects the standardized power response characteristics that the heterogeneous data center should exhibit to the distribution network under the objective of minimizing total system cost. The distribution network can decompose this low-carbon load baseline and distribute it to the heterogeneous data center, guiding it to operate according to the optimal adjustment scheme, thereby achieving the dual objectives of reducing system carbon emissions and operating costs.

[0046] Step S14: Define individual execution guidelines for the heterogeneous data centers to achieve preset low-carbon load targets, construct a guideline decomposition and optimization model based on the individual execution guidelines, and solve the guideline decomposition and optimization model based on the low-carbon load guidelines to obtain the target individual execution guidelines for each heterogeneous data center. Determine computing power task migration instructions based on the target individual execution guidelines so that each heterogeneous data center can perform corresponding low-carbon load adjustment operations based on the computing power task migration instructions.

[0047] In this embodiment, after obtaining the low-carbon load baseline determined in the aforementioned steps, it is necessary to decompose the global or system-level low-carbon load baseline to each heterogeneous data center to form individual execution baselines that can be specifically executed. First, the individual execution baselines for achieving the preset low-carbon load target of the heterogeneous data centers are defined. These individual execution baselines describe the execution baselines that a single heterogeneous data center needs to follow in response to low-carbon scheduling.

[0048] To optimize the allocation of individual execution guidelines, a guideline decomposition optimization model needs to be constructed based on a preset cost function, the original baseline load of heterogeneous data centers, and the individual execution guidelines. In one specific implementation, the individual execution guidelines of each heterogeneous data center and the condition equal to the low-carbon load guideline are determined as the overall guideline conservation constraint condition; simultaneously, the computing power adjustment range is determined based on the preset maximum inbound computing power task volume, the preset maximum outbound computing power task volume, and the nonlinear energy consumption model, and the condition of the individual execution guidelines within the computing power adjustment range is determined as the individual physical boundary constraint condition.

[0049] Finally, based on the overall baseline conservation constraints and the individual physical boundary constraints, the baseline decomposition optimization model is solved to obtain the target individual execution baselines for each heterogeneous data center. After obtaining the target individual execution baselines, specific computing power task migration instructions can be determined based on these instructions and sent to each heterogeneous data center. This allows each heterogeneous data center to perform corresponding low-carbon load adjustment operations based on the computing power task migration instructions, ultimately achieving the goal of low-carbon operation.

[0050] As can be seen from the above, this application constructs a nonlinear energy consumption model for IT equipment based on the computing power adjustment amount of heterogeneous data centers, determines the target energy efficiency ratio and energy efficiency conversion coefficient of IT equipment and cooling systems, and constructs a cooling power consumption model; based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model, a virtual power node model for heterogeneous data centers regarding computing power adjustment amount is established; using the target cost within the preset computing power scheduling cycle of the distribution network, a cost minimization model for heterogeneous data centers is constructed, and the target computing power adjustment amount is obtained by solving the model, and the low-carbon load baseline of heterogeneous data centers is determined by combining the virtual power node model; individual execution baselines for heterogeneous data centers to achieve preset low-carbon load targets are defined, and a baseline decomposition optimization model is constructed based on the individual execution baselines, and the target individual execution baselines of each data center are obtained by solving the low-carbon load baselines, thereby generating computing power task migration instructions to enable each heterogeneous data center to perform corresponding low-carbon load adjustment operations. In this way, this application can achieve accurate mapping and optimized allocation of system-level low-carbon target baselines to execution baselines of each data center, thereby effectively guiding data center loads to actively participate in the low-carbon scheduling of the distribution network.

[0051] The technical solutions of the embodiments of this application will be described in detail below.

[0052] First, we model the virtual power node of the data center based on the nonlinear energy consumption function.

[0053] In this step, the data center, as a flexible load with both computing power regulation and nonlinear energy consumption characteristics, has a direct impact on the grid-side input power due to its computing power task adjustments. To support the accurate generation and decomposition of the system-level load baseline in quasilinear demand response, this section constructs a virtual power node model that incorporates the coupling characteristics of nonlinear IT energy consumption and ambient temperature, used to characterize the impact of data center computing power regulation on grid load.

[0054] To achieve continuous and controllable adjustment of data center load within a quasi-linear demand response framework, a data center is defined. exist Spacetime migration of computing power tasks at any given moment This can be achieved through Geographic Load Balancing (GLB) technology. This indicates that the task will be moved to another data center or its execution will be postponed. This indicates receiving tasks from other data centers for migration or executing local tasks in advance. Data Center exist Real-time allocated computing load Represented as reference load The formula for calculating computing load, combined with the adjustment amount, can be as follows: ; Total input power of data center It mainly consists of the power consumption of IT equipment and the power consumption of auxiliary cooling. The power consumption of IT equipment depends on the operating status of the server. The physical expression can be: ; To address the power consumption of IT equipment, a nonlinear energy consumption model needs to be constructed. Considering the idle power consumption and dynamic utilization characteristics of servers, a quadratic term is introduced to characterize the energy efficiency degradation characteristics of servers in high-load regions. This establishes an analytical nonlinear mapping relationship between computing load changes and IT power consumption, providing a continuously differentiable power expression for subsequent optimization models. The corresponding formula for calculating the nonlinear power consumption of IT equipment is shown below: ; In the formula, The base keep-alive / idle power consumption of the server cluster is a constant. It is a linear growth coefficient. The coefficient is a quadratic term, representing the energy efficiency degradation under high load.

[0055] For auxiliary cooling power consumption, an environment-coupled cooling power consumption model needs to be constructed. Cooling power consumption depends on the heat generated by the IT system and the energy efficiency ratio (COP) of the cooling system. Based on the Carnot cycle principle, the COP varies with ambient temperature. The temperature rises and falls accordingly. To ensure the solvability of the optimization problem, a linear temperature-sensitive model is adopted, and the COP calculation formula can be shown below: ; In the formula, Reference temperature The rated COP below; This is the temperature sensitivity coefficient.

[0056] Furthermore, the formula for calculating the auxiliary cooling power consumption can be as follows: ; Constructing a comprehensive energy efficiency conversion coefficient PUE is the time-varying physical expression of power consumption, which reflects the total power input required for each unit of IT power consumption. The formula for calculating can be shown below: ; Therefore, a data center can be equivalently represented as a virtual power node driven by environmental parameters and computing power decision variables. The active power it injects into the power grid can be expressed as an explicit function of the decision variables. Substituting and expanding, we obtain the optimization objective in the following form: ; Furthermore, to ensure the physical security and task conservation of the data center, The following three constraints must be met: First, the constraint of computing power adjustment capability: adjustment amount It must be within the range shown in the following formula: ; in, and They are respectively Time Data Center The maximum number of tasks that can be moved in and out. This parameter takes into account both the server's remaining computing resources and the data center's communication outbound bandwidth limitations, using the smaller of the two values ​​as the adjustment boundary.

[0057] Secondly, the total task volume conservation constraint: within a scheduling cycle such as 24 hours, the total task volume of the entire data center cluster remains balanced, allowing tasks to migrate between different data centers. The corresponding constraint is shown in the following formula: ; Third, physical capacity safety constraints: the adjusted total load cannot exceed the maximum physical processing capacity of the server cluster. It cannot be lower than 0. The corresponding constraints are shown in the following formula: ; Secondly, a system-level low-carbon load baseline based on tiered carbon cost is generated.

[0058] In this step, from the perspective of the Distribution System Operator (DSO), the data center cluster is equivalent to a schedulable virtual flexible resource. Taking into account the costs of electricity purchase, generation, data center regulation and compensation, renewable energy curtailment penalties, and tiered carbon trading, a system-level generalized cost minimization model is constructed. Based on this model, a system-level low-carbon load baseline is generated within the scheduling cycle, providing a unified benchmark for the subsequent decomposition of heterogeneous data center baselines.

[0059] System-level optimization aims to minimize the total generalized cost within the scheduling cycle, encompassing electricity purchase cost, generation cost, data center regulation compensation cost, curtailment penalty cost, and tiered carbon trading cost. Among these, tiered carbon trading cost, as an endogenous variable, directly participates in the system-level baseline generation process, ensuring that the generated load baseline simultaneously reflects operational economics and carbon emission constraints. The corresponding objective function for cost minimization is shown below: ; Among them, for Considering the energy interaction between the distribution network and the upstream main grid, This represents the electricity purchase expenditure required by a DSO to meet the net load demand within its jurisdiction. The calculation formula is as follows: ; In the formula, For a moment Real-time time-of-use electricity pricing; The power that a DSO purchases from the main grid is determined by the system power balance equation.

[0060] Among them, for C Gen (t) For controllable power generation resources within the jurisdiction, the classic quadratic function is used to describe the fuel consumption and operation and maintenance losses of coal-fired / gas-fired units. C Gen (t) The calculation formula is as follows: ; In the formula, For the first Consumption characteristic coefficient of the unit.

[0061] Among them, for C Comp (t) To incentivize data centers to actively participate in grid interaction, data center management offices (DSOs) need to pay corresponding economic incentives. This cost represents the "purchase price" paid to obtain data center flexibility services. C Comp (t) The calculation formula is as follows: ; In the formula, The compensation unit price is the unit price for migrating a unit of computing power task, expressed in yuan per task.

[0062] Among them, for To promote the consumption of renewable energy, penalties are imposed on the amount of wind and solar power curtailed, including penalties for curtailed renewable energy. To maximize the utilization of renewable energy. The calculation formula is as follows: ; In the formula, This is the wind curtailment penalty coefficient; This represents the theoretical maximum power output of the new energy source. This refers to the actual grid-connected power consumption.

[0063] Among them, for C Carbon (t) Specifically, a tiered carbon trading model is introduced. This mechanism applies non-linear penalties to excessive emissions through progressively increasing carbon price signals, thereby guiding the system to prioritize the use of data center regulation capacity for peak shaving and valley filling, thus reducing the overall marginal carbon emissions of the system. (The total carbon emissions of the system...) It consists of two parts: one is the direct combustion emissions from fossil fuel units within the jurisdiction, and the other is the indirect emissions implied by purchasing electricity from the upper-level power grid due to power shortages. The calculation formula is as follows: ; In the formula, For the first The carbon emission intensity of Taiwan's local thermal power units can be expressed in kg / kWh, depending on the unit's energy efficiency characteristics. The average carbon emission intensity input from the upstream power grid, expressed in kg / kWh, characterizes the cleanliness of purchased electricity.

[0064] Based on the baseline method, the free carbon allowances obtained by the system are dynamically linked to the system's energy output. The corresponding calculation formula is shown below: ; in, These are the carbon quota benchmark coefficients per unit of electricity.

[0065] The difference between actual emissions and free allowances is calculated, and the final carbon trading cost is calculated using the piecewise function shown below. C Carbon (t) .

[0066] ; In the formula, The incentive coefficient for selling surplus quotas is reflected as a negative cost; The base carbon trading price; The punitive carbon trading price for exceeding emission limits, and ; This refers to the length of the tiered carbon trading range. Furthermore, it should be noted that the preset excess carbon emission condition refers to the state where the total carbon emissions of the system exceed the sum of the free allowance and the tier length, i.e., it meets the condition. .

[0067] To ensure the feasibility and security of the scheduling scheme at the physical level, the solution of the above objective function is subject to the following four key constraints: First, the system power balance constraint, and the corresponding constraint formula is shown below: ; in, For regular rigid loads other than data centers; The total load of the data center cluster is calculated by aggregation using the virtual power node model, and the corresponding calculation formula is shown below: ; Secondly, distributed unit operation constraints: The unit output must meet the upper and lower limits of technical output and the ramp rate limit, and the corresponding constraint formulas are as follows: ; In the formula, For the first Minimum and maximum output of the unit; These represent the unit's maximum downward and upward ramp rates, respectively, and the unit can be MW / h.

[0068] Third, data center operational constraints: decision variables It must satisfy the constraints established above, including the constraints on computing power adjustment capability, total task conservation, and physical capacity security.

[0069] Fourth, line transmission capacity constraints: Using the DC power flow approximation, thermal stability constraints are applied to the active power flow of the line. The corresponding constraint formulas are shown below: ; In the formula, For the first The active power flowing through each power distribution line; This represents the maximum active power transmission capacity allowed for this line.

[0070] Under the condition that the relevant parameters meet the physical feasibility, the continuous part is a convex quadratic programming problem, and the global optimal solution can be obtained by linearizing the step carbon valence term and calling a commercial solver.

[0071] Based on the model solution results, the optimal adjustment amount for each data center is obtained. At this point, according to the aforementioned The calculation formula for the optimization objective form and the calculation formula for the system-level low-carbon load baseline are as follows: ; Finally, a guideline optimization decomposition is performed for heterogeneous data centers.

[0072] In this step, to achieve accurate execution of the system-level baseline across multiple data centers, it is necessary to further decompose the system-level baseline into individual execution baselines for each data center. Referring to the Sub-CDL decomposition theory, and based on the differences in adjustment potential and cost sensitivity among each data center, the overall macro-level baseline is "projected" as follows: Individual execution standards This ensures that the allocation scheme is physically feasible and minimizes the total system adjustment cost.

[0073] Assume the first The adjustment amount for each data center is Introducing a cost function It is typically modeled as a quadratic function of the adjustment amount to characterize the business migration cost, which increases non-linearly with the depth of adjustment. The calculation formula is as follows: ; In the formula, It is the first Cost sensitivity coefficient for each data center. The larger the value, the more critical the business carried by the data center (DC), and the higher the opportunity cost or performance loss risk associated with migrating a unit of computing power; conversely, The smaller the value, the higher the adjustment flexibility of the DC. Users can report this in the first step, or it can be deduced from the capacity; the larger the capacity, the more redundancy per unit. The smaller it may be.

[0074] Based on this, with the goal of minimizing the total system adjustment cost, a directrix decomposition optimization model is constructed. Under the premise of satisfying the system-level directrix conservation, the optimal individual execution directrix distribution is solved. The corresponding solution formula is shown below: ; In the formula, For the first The original baseline load of each data center, i.e. ; The first one to be solved Individual execution baselines for each data center.

[0075] The solution to the model must satisfy the following physical and operational constraints. First is the overall headway conservation constraint: the sum of all individual execution headways must equal the system-level load headway generated in the preceding steps. The corresponding constraint formula is shown below: ; Secondly, there are individual physical boundary constraints. The decomposed individual directrixes must satisfy the adjustment capability range defined by the virtual power node model in the previous steps. The corresponding constraint formulas are as follows: ; In the formula, and These are the lower limit of computing power adjustment defined in the aforementioned steps. And computing power adjustment limit The power boundary value is obtained by substituting it into the nonlinear energy consumption function.

[0076] The aforementioned model constructs a standard quadratic programming (QP) problem with a unique globally optimal solution. By solving this model, the optimal individual execution baseline for each data center is obtained. Subsequently, based on the computing power-power analytical mapping relationship established in the preceding steps, the individual power baseline can be reversed into specific computing power task migration instructions, enabling the data center to accurately execute system-level low-carbon load adjustment targets at the physical level.

[0077] Therefore, this invention constructs a computing power-electricity analytical energy consumption function that simultaneously considers the nonlinear energy consumption characteristics of server utilization and the influence of ambient temperature. A quadratic function is used to characterize the nonlinear impact of server utilization changes on IT energy consumption. Compared to a linear model, this function more accurately reflects the energy efficiency degradation characteristics in high-load regions, thereby improving the accuracy of the mapping relationship between computing power adjustment and input power. This allows data center adjustment capabilities to be embedded into the system optimization model in an explicit functional form, improving the accuracy of adjustment capability estimation when multiple data centers participate in demand response. Furthermore, by introducing tiered carbon trading costs into the system-level optimization model, carbon emission constraints directly participate in decision-making calculations during the baseline generation stage, thus forming a system-level low-carbon load baseline that simultaneously reflects operating costs and carbon constraints.

[0078] Furthermore, this invention constructs a heterogeneous baseline optimization decomposition model that takes into account differences in adjustment cost sensitivity. Under the premise of satisfying system-level baseline conservation, it determines individual execution baselines based on the differences in adjustment capabilities and costs of each data center, achieving optimal allocation of system-level load baselines to execution baselines across multiple data centers. This method can minimize the total system adjustment cost while avoiding resource allocation deviations caused by traditional proportional allocation. Through the established computing power-power analytical mapping relationship, individual power baselines are directly converted into computing power task migration instructions, enabling the executable implementation of system-level load targets on the data center side.

[0079] Accordingly, see Figure 2 As shown, this application provides a low-carbon load baseline generation and decomposition device for heterogeneous data centers, comprising: The first model construction module 11 is used to construct a nonlinear energy consumption model of IT equipment in the heterogeneous data center based on the computing power adjustment amount of the heterogeneous data center, determine the target energy efficiency ratio of the IT equipment and cooling system in the heterogeneous data center, determine the energy efficiency conversion coefficient based on the target energy efficiency ratio, and construct a cooling power consumption model based on the nonlinear energy consumption model and the target energy efficiency ratio. The second model construction module 12 is used to construct a virtual power node model for the computing power adjustment based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model; the virtual power node model is a standardized equivalent model presented to the power distribution network after abstracting and encapsulating the internal characteristics of the heterogeneous data center. The baseline determination module 13 is used to construct a cost minimization model of the heterogeneous data center based on the target cost of the distribution network within a preset computing power scheduling cycle, and to determine the low-carbon load baseline of the heterogeneous data center based on the target computing power adjustment amount obtained by solving the cost minimization model and the virtual power node model. The guideline decomposition module 14 is used to define individual execution guidelines for the heterogeneous data centers to achieve preset low-carbon load targets, construct a guideline decomposition optimization model based on the individual execution guidelines, and solve the guideline decomposition optimization model based on the low-carbon load guidelines to obtain the target individual execution guidelines for each heterogeneous data center. The computing power task migration instructions are determined based on the target individual execution guidelines so that each heterogeneous data center can perform corresponding low-carbon load adjustment operations based on the computing power task migration instructions.

[0080] In some specific embodiments, the first model construction module 11 specifically includes: The load determination unit is used to determine the computing load of the heterogeneous data center based on the computing power adjustment amount of the heterogeneous data center and the preset benchmark load of the heterogeneous data center. The first model building unit is used to build a nonlinear energy consumption model of IT equipment in the heterogeneous data center based on the idle power consumption of the server cluster in the heterogeneous data center, the preset linear growth coefficient, the preset quadratic coefficient, and the computing load; the preset quadratic coefficient is a coefficient that characterizes the degree of energy efficiency degradation of the server cluster under a preset high load.

[0081] In some specific embodiments, the first model construction module 11 specifically includes: An energy efficiency ratio determination unit is used to determine the target energy efficiency ratio of the IT equipment and cooling system in the heterogeneous data center based on a preset reference temperature, the energy efficiency ratio corresponding to the preset reference temperature, a preset temperature sensitivity coefficient, and the current ambient temperature using a linear temperature-sensitive model. The second model building unit is used to determine the energy efficiency conversion coefficient based on the target energy efficiency ratio, and to build a cooling power consumption model based on the nonlinear energy consumption model and the target energy efficiency ratio.

[0082] In some specific implementations, the computing power adjustment amount is a variable that satisfies the constraints of computing power adjustment capability, the constraints of total task conservation, and the constraints of physical capacity security.

[0083] In some specific embodiments, the alignment determination module 13 specifically includes: The third model construction unit is used to construct a cost minimization model for the heterogeneous data center based on the net load cost, power generation cost, computing power compensation cost, power curtailment penalty cost, and tiered carbon trading cost of the distribution network within a preset computing power scheduling cycle.

[0084] In some specific embodiments, the third model building unit includes: The first cost determination subunit is used to determine the computing power compensation cost of the heterogeneous data center based on the unit computing power compensation cost of the distribution network for the heterogeneous data center under a preset computing power task migration unit and the computing power adjustment amount. The second cost determination subunit is used to determine the tiered carbon trading cost based on total carbon emissions, carbon allowance amount, preset reward coefficient for issuing surplus carbon allowances, preset basic carbon trading cost, punitive carbon trading cost for meeting preset excess carbon emission conditions, and the length of the tiered carbon trading interval.

[0085] In some specific embodiments, the alignment determination module 13 specifically includes: The first condition determination unit is used to determine the conditions under which the computing power adjustment amount satisfies the computing power adjustment capability constraint, the total task conservation constraint, and the physical capacity security constraint as data center operation constraints. The baseline determination unit is used to solve the cost minimization model based on the power balance constraints between the heterogeneous data center and the power distribution network, the unit constraints of the coal-fired or gas-fired units, the operation constraints of the data center, and the line transmission capacity constraints of the power distribution network, to obtain the target computing power adjustment amount, and to determine the low-carbon load baseline of the heterogeneous data center using the target computing power adjustment amount and the virtual power node model.

[0086] In some specific embodiments, the directrix decomposition module 14 specifically includes: The second condition determination unit is used to determine the condition that the sum of the individual execution guidelines of each heterogeneous data center is equal to the low-carbon load guideline as the overall guideline conservation constraint condition. The third condition determination unit is used to determine the computing power adjustment range based on the preset maximum inbound computing power task volume, the preset maximum outbound computing power task volume, and the nonlinear energy consumption model, and to determine the conditions of the individual execution guideline within the computing power adjustment range as individual physical boundary constraint conditions. The guideline decomposition unit is used to solve the guideline decomposition optimization model based on the overall guideline conservation constraint and the individual physical boundary constraint to obtain the target individual execution guideline of each heterogeneous data center.

[0087] In some specific implementations, the constraint on computing power adjustment capability is a condition that limits the amount of computing power adjustment to not less than the preset maximum amount of computing power tasks that can be moved into the heterogeneous data center and not greater than the preset maximum amount of computing power tasks that can be moved out of the heterogeneous data center.

[0088] In some specific implementations, the total task conservation constraint is a condition that limits the total computing power task quantity related to the computing power adjustment amount to meet the preset task quantity balance condition within a preset computing power scheduling period.

[0089] In some specific implementations, the physical capacity safety constraint is a condition that limits the difference between the computing load and the computing power adjustment amount to not less than zero and not greater than the preset maximum physical processing capacity of the server cluster.

[0090] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the low-carbon load baseline generation and decomposition method for heterogeneous data centers disclosed in any of the foregoing embodiments. Furthermore, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0091] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0092] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0093] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the low-carbon load baseline generation and decomposition method for heterogeneous data centers executed by the electronic device 20 as disclosed in any of the foregoing embodiments, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0094] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned disclosed method for generating and decomposing low-carbon load baselines for heterogeneous data centers. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0095] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0096] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0097] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0098] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0099] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for generating and decomposing low-carbon load baselines for heterogeneous data centers, characterized in that, include: Based on the computing power adjustment of the heterogeneous data center, a nonlinear energy consumption model of the IT equipment in the heterogeneous data center is constructed. The target energy efficiency ratio of the IT equipment and the cooling system in the heterogeneous data center is determined. The energy efficiency conversion coefficient is determined based on the target energy efficiency ratio. A cooling power consumption model is constructed based on the nonlinear energy consumption model and the target energy efficiency ratio. A virtual power node model for the computing power adjustment is constructed based on the energy efficiency conversion coefficient, the nonlinear energy consumption model, and the cooling power consumption model; the virtual power node model is a standardized equivalent model presented to the power distribution network after abstracting and encapsulating the internal characteristics of heterogeneous data centers. Based on the target cost of the power distribution network within a preset computing power scheduling cycle, a cost minimization model for the heterogeneous data center is constructed. Based on the target computing power adjustment amount obtained from solving the cost minimization model and the virtual power node model, the low-carbon load guideline for the heterogeneous data center is determined. Define individual execution guidelines for the heterogeneous data centers to achieve preset low-carbon load targets. Construct a guideline decomposition and optimization model based on the individual execution guidelines, and solve the guideline decomposition and optimization model based on the low-carbon load guidelines to obtain the target individual execution guidelines for each heterogeneous data center. Determine computing power task migration instructions based on the target individual execution guidelines so that each heterogeneous data center can perform corresponding low-carbon load adjustment operations based on the computing power task migration instructions.

2. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 1, characterized in that, The nonlinear energy consumption model for IT equipment in the heterogeneous data center, constructed based on the computing power adjustment of the heterogeneous data center, includes: The computing load of the heterogeneous data center is determined based on the computing power adjustment amount of the heterogeneous data center and the preset baseline load of the heterogeneous data center. A nonlinear energy consumption model for IT equipment in the heterogeneous data center is constructed based on the idle power consumption of the server cluster in the heterogeneous data center, the preset linear growth coefficient, the preset quadratic coefficient, and the computing load; the preset quadratic coefficient is a coefficient characterizing the degree of energy efficiency degradation of the server cluster under a preset high load.

3. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 1, characterized in that, The process of determining the target energy efficiency ratio (EER) of the IT equipment and cooling system in the heterogeneous data center, determining the energy efficiency conversion coefficient based on the target EER, and constructing a cooling power consumption model based on the nonlinear energy consumption model and the target EER includes: The target energy efficiency ratio of the IT equipment and cooling system in the heterogeneous data center is determined by a linear temperature-sensitive model based on a preset reference temperature, the energy efficiency ratio corresponding to the preset reference temperature, a preset temperature sensitivity coefficient, and the current ambient temperature. The energy efficiency conversion coefficient is determined based on the target energy efficiency ratio, and a cooling power consumption model is constructed based on the nonlinear energy consumption model and the target energy efficiency ratio.

4. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 2, characterized in that, The computing power adjustment amount is a variable that satisfies the constraints of computing power adjustment capability, total task conservation, and physical capacity security.

5. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 4, characterized in that, The constraint on the computing power adjustment capability is that the amount of computing power adjustment is not less than the preset maximum amount of computing power tasks that can be moved into the heterogeneous data center and not greater than the preset maximum amount of computing power tasks that can be moved out of the heterogeneous data center.

6. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 4, characterized in that, The total task conservation constraint is a condition that limits the total computing power task quantity related to the computing power adjustment amount to meet the preset task quantity balance condition within a preset computing power scheduling cycle.

7. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 4, characterized in that, The physical capacity safety constraint is a condition that limits the difference between the computing load and the computing power adjustment amount to not less than zero and not greater than the preset maximum physical processing capacity of the server cluster.

8. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to any one of claims 1 to 4, characterized in that, The method for constructing a cost minimization model for the heterogeneous data center based on the target cost of the power distribution network within a preset computing power scheduling period includes: A cost minimization model for the heterogeneous data center is constructed based on the net load cost, power generation cost, computing power compensation cost, power curtailment penalty cost, and tiered carbon trading cost of the distribution network within a preset computing power scheduling cycle.

9. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 8, characterized in that, The process of determining the computing power compensation cost includes: The computing power compensation cost of the heterogeneous data center is determined based on the unit computing power compensation cost of the power distribution network for the heterogeneous data center under a preset computing power task migration unit and the computing power adjustment amount. Furthermore, the process for determining the tiered carbon trading costs includes: The tiered carbon trading cost is determined based on total carbon emissions, carbon allowances, a pre-set reward coefficient for issuing surplus carbon allowances, a pre-set basic carbon trading cost, a punitive carbon trading cost for meeting pre-set excess carbon emission conditions, and the length of the tiered carbon trading range.

10. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 4, characterized in that, The process of determining the low-carbon load baseline of the heterogeneous data center based on the target computing power adjustment amount obtained from solving the cost minimization model and the virtual power node model includes: The conditions under which the computing power adjustment amount satisfies the computing power adjustment capability constraint, the total task conservation constraint, and the physical capacity security constraint are determined as the data center operation constraints. The cost minimization model is solved based on the power balance constraints between the heterogeneous data center and the power distribution network, the unit constraints of the coal-fired or gas-fired units, the operation constraints of the data center, and the line transmission capacity constraints of the power distribution network to obtain the target computing power adjustment amount. The low-carbon load baseline of the heterogeneous data center is then determined using the target computing power adjustment amount and the virtual power node model.

11. The method for generating and decomposing low-carbon load baselines for heterogeneous data centers according to claim 5, characterized in that, The optimization model based on the low-carbon load baseline is solved to obtain the target individual execution baseline for each heterogeneous data center, including: The condition that the sum of the individual execution guidelines of each heterogeneous data center is equal to the low-carbon load guideline is determined as the overall guideline conservation constraint condition. Based on the preset maximum load capacity, the preset maximum load capacity, and the nonlinear energy consumption model, the computing power adjustment range is determined, and the conditions of the individual execution guideline within the computing power adjustment range are determined as individual physical boundary constraints. The guideline decomposition optimization model is solved based on the overall guideline conservation constraint and the individual physical boundary constraint to obtain the target individual execution guideline for each heterogeneous data center.

Images