Data center carbon-electricity cooperative scheduling method and system based on carbon emission flow

Abstract

The invention provides a data center carbon electricity cooperative scheduling method and system based on a carbon emission flow, and the method comprises the steps: calculating the initial electricity purchasing power of a data center in a current scheduling period according to the load data of the data center; and inputting the initial power purchase power to a preset double-layer optimization scheduling model, so that the double-layer optimization scheduling model performs multiple times of iterative optimization under the constraint of a preset carbon emission flow analysis model according to the initial power purchase power, and a final collaborative scheduling scheme of the data center in the current scheduling period is generated, the carbon emission flow analysis model is constructed by performing mathematical modeling on operation modes of a power system and a data center and carbon emission flows among nodes; and adjusting the workload of each computing node and the operation strategy of each local power device of the data center in the current scheduling period according to the final cooperative scheduling scheme. The problem that the energy utilization efficiency and the carbon emission are difficult to optimize at the same time in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of power system dispatch optimization technology, and in particular to a data center carbon-electricity coordinated dispatch method and system based on carbon emission flow. Background Technology

[0002] With the increasing global demand for data processing and the strengthening of carbon emission control, data centers, as high-energy-consuming infrastructure, have received widespread attention for their energy management and carbon reduction. Data centers typically consist of computing load units, power distribution systems, cooling systems, and energy management systems, relying on the public power grid or renewable energy sources for energy supply and demand. However, current carbon emission assessments mostly use the fixed carbon factor method, failing to reflect the dynamic changes in renewable energy output in the power grid, resulting in insufficient accuracy in carbon emission assessments. Furthermore, existing data center load management and power system dispatch are independent of each other, lacking coordination of carbon emission signals, making it difficult to simultaneously ensure energy efficiency and optimize carbon emissions. As data centers expand in scale and power systems become more complex, traditional load management and dispatch methods are insufficient to meet the comprehensive optimization requirements of energy costs, carbon emissions, and system stability. Summary of the Invention

[0003] This application provides a carbon-electricity coordinated scheduling method and system for data centers based on carbon emission flows, which can solve the problem in the prior art that it is difficult to simultaneously optimize energy utilization efficiency and carbon emissions during the scheduling optimization process of data centers.

[0004] In a first aspect, embodiments of this application provide a data center carbon-electricity coordinated scheduling method based on carbon emission flows, comprising:

[0005] Calculate the initial power purchase of the data center in the current scheduling cycle based on the data center's load data.

[0006] The initial power purchase is input into a preset two-layer optimization scheduling model, which then performs several iterations of optimization based on the initial power purchase under the constraints of a preset carbon emission flow analysis model. This generates a final collaborative scheduling scheme for the data center in the current scheduling cycle. The final collaborative scheduling scheme includes the power system's power generation output plan, the final power purchase, the local power equipment operation scheme, and the computing task allocation scheme. The carbon emission flow analysis model is constructed by mathematically modeling the operation mode of the power system and the data center, as well as the carbon emission flow between various nodes. It includes a data center power demand model, a data center power demand adjustment model, a data center local power equipment operation model, and a carbon emission flow model.

[0007] The workload of each computing node and the operation strategy of each local power device in the data center during the current scheduling cycle are adjusted according to the final collaborative scheduling scheme.

[0008] In each iteration of the optimization process, the upper-level model of the two-layer optimization scheduling model generates the current node carbon potential signal and the current node electricity price of each node based on the previous collaborative scheduling scheme and the real-time operation data of the power system. The node carbon potential signal is the carbon emission corresponding to the unit power consumption of the node. The lower-level model of the two-layer optimization scheduling model generates the current collaborative scheduling scheme of the data center based on the current node carbon potential signal and the current node electricity price.

[0009] This application provides a carbon-electricity coordinated scheduling method for data centers based on carbon emission flows. It calculates the initial power purchase of the data center and inputs it into a two-layer optimization scheduling model. Through alternating iterative optimization using the two-layer model, a final coordinated scheduling scheme for the data center in the current scheduling cycle is generated. Based on this final coordinated scheduling scheme, the workload of the data center and the operating status of local power equipment are adjusted. In this application embodiment, the upper-layer model transmits node carbon potential signals to the lower-layer model. The lower-layer model adjusts the data center's operating strategy based on the feedback node carbon potential signals and transmits this adjustment back to the upper-layer model. Through an iterative process, the carbon-electricity signals are dynamically updated, and the scheduling strategy is gradually optimized. Ultimately, a carbon-electricity coordinated scheduling scheme with the lowest carbon-electricity cost and optimal system energy efficiency is obtained. This achieves bidirectional coordinated scheduling between the power system and the data center, as well as dynamic matching of carbon emission flows, improving energy utilization efficiency while reducing carbon emissions. Furthermore, this application also introduces a carbon emission flow analysis model to further constrain the scheduling optimization process. This model integrates multiple dimensions such as electricity demand, regulation capacity, local power generation, energy storage, and carbon flow, and tracks carbon emissions from power generation, energy storage to electricity consumption, supporting refined carbon accounting and emission reduction strategy formulation, and improving the scientific nature and comprehensiveness of coordinated scheduling.

[0010] Furthermore, the step of calculating the initial power purchase of the data center in the current scheduling cycle based on the data center's load data includes:

[0011] Obtain the load data of the data center in the current scheduling cycle. The load data includes online load, online load energy efficiency ratio, offline load, and offline load energy efficiency ratio.

[0012] The total power demand of the data center is calculated based on the load data using the carbon emission flow analysis model.

[0013] The initial power purchase capacity of the data center in the current scheduling cycle is calculated based on the output of the data center's local power generation equipment, the charging and discharging power of the energy storage equipment, and the total power demand.

[0014] This application provides a method for calculating initial power purchase capacity. The method calculates the initial power purchase capacity based on load data, local power generation equipment output, and energy storage equipment discharge power. During the calculation process, the online / offline load and its energy efficiency ratio are accurately calculated. At the same time, it takes into account that the local power generation equipment and energy storage equipment of the data center also provide a part of the power supply to the data center, ensuring the rationality and accuracy of the initial power purchase capacity calculation and providing data support for subsequent coordinated scheduling.

[0015] In one possible implementation, the upper-level model generates the current node carbon potential signal and current node electricity price for each node based on the previous coordinated scheduling scheme and real-time power system operation data, including:

[0016] Based on the previous coordinated dispatch scheme and real-time power system operation data, and under the carbon emission flow analysis model and the first constraint, the power generation output of each node is optimized with the goal of minimizing the total operating cost of the power system, and the current power generation output plan is generated.

[0017] Based on the current power generation output plan, the current node carbon potential signal of each node is calculated;

[0018] The current node electricity price is generated for each node based on the carbon potential signal of each current node.

[0019] Furthermore, the real-time operating data of the power system includes the rated capacity, ramp rate, and power generation cost coefficient of each generator unit in the power system;

[0020] The power generation cost coefficient includes a nonlinear power cost coefficient, a linear power cost coefficient, and an no-load cost coefficient. The power generation cost coefficient is obtained by actually testing the fuel consumption of the generator set at different power levels and fitting the data curve.

[0021] The first constraint includes power system power balance constraint, carbon emission flow balance constraint, and generator set power constraint.

[0022] This application provides a method for generating the current node's carbon potential signal and current node's electricity price using an upper-level model. The upper-level model establishes a scheduling optimization framework from the perspective of a power system operator, aiming to minimize the total operating cost of the power system, optimize output allocation, power flow allocation, and carbon emission flow allocation for each unit, and generate a corresponding power generation plan. Furthermore, based on the power generation plan, a node carbon potential signal and electricity price are generated. These two signals constitute the boundary conditions and economic incentives for optimizing the lower-level model, and serve as the main way for the upper layer to influence the lower layer in the entire two-layer iterative solution mechanism, promoting the efficient use of renewable energy in data centers, improving the energy utilization efficiency of data centers, and reducing carbon emissions. In addition, this embodiment improves the feasibility and economics of the optimization results by fitting the power generation cost coefficient through actual testing. Upper-level optimization is performed based on power balance constraints, carbon emission flow balance constraints, and generator power constraints. Power balance constraints and generator power constraints ensure stable operation of the power system, while carbon emission flow balance constraints ensure a dynamic balance between the carbon flow rate of each branch of the power system and the carbon flow rate of each node in the data center, ensuring the rationality of the optimization results.

[0023] Furthermore, the step of calculating the current node carbon potential signal of each node based on the current power generation output plan includes:

[0024] Based on the current power generation output plan and the topological relationship between each node, determine the active power of each node and the active power flow between each node;

[0025] Based on the active power of each node, the active power flow between nodes, the carbon emission intensity of each node, and the carbon flow density between nodes, the current node carbon potential signal of each node is calculated through the carbon emission flow analysis model.

[0026] This application provides a method for calculating node carbon potential signals. The method calculates node carbon potential signals based on active power flow and carbon emission intensity. In the calculation process, the method combines the power grid topology relationship and uses parameters such as active power flow and carbon flow density to reflect the mutual influence between nodes in the carbon emission process, dynamically reflecting the carbon emission contribution of the power transmission path and improving the accuracy of carbon potential signal calculation.

[0027] In one possible implementation, the lower-level model generates a current collaborative scheduling scheme for the data center based on the carbon potential signal of each current node and the electricity price of the current node, including:

[0028] Based on the carbon potential signals of each current node, the electricity price of the current node, and the data center's own configuration parameters, the previous collaborative scheduling scheme is optimized under the carbon emission flow analysis model and the second constraint, with the goal of minimizing electricity purchase cost and carbon emission cost, to generate the current collaborative scheduling scheme for the data center.

[0029] The second constraint includes load regulation constraint, data center memory constraint, power system-data center power balance constraint, and data center carbon emission flow balance constraint.

[0030] This application provides a method for generating a current collaborative scheduling scheme using a lower-level model. The lower-level model aims to minimize electricity purchase costs and carbon emission costs by adjusting the purchased power, local power equipment operation schemes, and computing task allocation schemes for the current scheduling cycle. This achieves dynamic response to node carbon potential signals, improving the energy utilization efficiency of the data center and reducing carbon emissions. Regarding constraint settings, load regulation constraints and data center memory constraints limit the task allocation of each computing node in the data center, preventing any computing node's computing tasks from exceeding its capacity. Power system-data center power balance constraints ensure the normal operation of the data center's power generation and energy storage equipment and maintain energy exchange between the data center and the power system within normal ranges. Data center carbon emission flow balance constraints ensure normal carbon emission behavior of each device in the data center during operation. The combined constraints of these conditions ensure that the scheduling scheme meets the actual operational needs of the data center.

[0031] In one possible implementation, the carbon emission flow analysis model is constructed by mathematically modeling the operation of the power system and data center, as well as the carbon emission flows between various nodes, including:

[0032] Based on the calculated relationship between the online load, online load energy efficiency ratio, offline load, offline load energy efficiency ratio, and power demand of each computing node in the data center, a power demand model for the data center is constructed.

[0033] Based on the calculation relationship between the transferable online load, the deferred offline load, and the adjustable power demand of each computing node in the data center, a power demand adjustment model for the data center is constructed.

[0034] Based on the calculated relationship between the real-time active power, maximum active power, conversion coefficient, power generation efficiency, and the light radiation intensity and temperature of the surrounding environment of each photovoltaic power generation device in the data center, an operation model of the photovoltaic power generation device in the data center is constructed.

[0035] Based on the calculation relationship between the real-time storage capacity, maximum storage capacity, charging and discharging efficiency, real-time charging and discharging power, and maximum charging and discharging power of each energy storage device in the data center, an operation model of the data center energy storage device is constructed.

[0036] Based on the active power of each node in the power system, the active power flow between nodes, the carbon emission intensity of each node, the carbon flow density between nodes, the node carbon potential signal and carbon emission of each node, a model for obtaining the carbon emission flow of power grid generation is constructed.

[0037] Based on the calculated relationship between the carbon emission factors, active power, and carbon emissions of each carbon emission power generation device in the data center, a carbon emission flow model for data center power generation is constructed.

[0038] Based on the charging and discharging carbon flow rate of each energy storage device in the data center, the real-time charging power of the local power generation equipment to the energy storage device and the corresponding carbon emission factor, the real-time charging power of the power grid to the energy storage device and the corresponding node carbon potential signal, as well as the calculated relationship between the discharge carbon flow density, storage carbon emissions and discharge carbon emissions, a carbon emission flow model for data center energy storage devices is constructed.

[0039] This application provides a method for constructing a carbon emission flow analysis model, which integrates multiple dimensions such as power demand, regulation capacity, local power generation, energy storage, and carbon flow to improve the scientificity and comprehensiveness of coordinated scheduling. By coupling the data center power demand model, the data center power demand regulation model, the data center local power equipment operation model, and the carbon emission flow model, it can comprehensively characterize the dynamic coupling relationship between the power output behavior of the power system generation unit, the load handling behavior of the data center, and the carbon emission behavior, realize the dynamic assessment of green power consumption, thereby generating carbon emission signals and operating characteristic parameters that reflect the carbon emission characteristics of nodes, and providing the necessary data and model foundation for subsequent carbon-electricity coordinated optimization scheduling.

[0040] Secondly, embodiments of this application provide a data center carbon-electricity coordinated scheduling system based on carbon emission flows, including a power calculation module, a two-layer optimization module, and a scheduling module;

[0041] The power calculation module is used to calculate the initial power purchase of the data center in the current scheduling cycle based on the data center's load data.

[0042] The dual-layer optimization module is used to input the initial power purchase into a preset dual-layer optimization scheduling model, so that the dual-layer optimization scheduling model performs several iterations of optimization based on the initial power purchase under the constraints of a preset carbon emission flow analysis model, generating a final collaborative scheduling scheme for the data center in the current scheduling cycle. The final collaborative scheduling scheme includes the power system power generation output plan, the final power purchase, the local power equipment operation scheme, and the computing task allocation scheme. The carbon emission flow analysis model is obtained by mathematically modeling the operation mode of the power system and the data center, as well as the carbon emission flow between various nodes, including the data center power demand model, the data center power demand adjustment model, the data center local power equipment operation model, and the carbon emission flow model.

[0043] The scheduling module is used to adjust the workload of each computing node and the operation strategy of each local power device in the data center during the current scheduling cycle according to the final collaborative scheduling scheme.

[0044] The dual-layer optimization module includes an upper-layer optimization unit and a lower-layer optimization unit. In each iteration of optimization, the upper-layer optimization unit is used to generate the current node carbon potential signal and the current node electricity price of each node based on the previous collaborative scheduling scheme and real-time power system operation data through the upper-layer model in the dual-layer optimization scheduling model. The node carbon potential signal is the carbon emission corresponding to the unit power consumption of the node. The lower-layer optimization unit is used to generate the current collaborative scheduling scheme of the data center based on the current node carbon potential signal and the current node electricity price through the lower-layer model in the dual-layer optimization scheduling model.

[0045] Furthermore, the upper-level model generates the current node carbon potential signal and current node electricity price for each node based on the previous coordinated scheduling scheme and real-time power system operation data, including:

[0046] Based on the previous coordinated dispatch scheme and real-time power system operation data, and under the carbon emission flow analysis model and the first constraint, the power generation output of each node is optimized with the goal of minimizing the total operating cost of the power system, and the current power generation output plan is generated.

[0047] Based on the current power generation output plan, the current node carbon potential signal of each node is calculated;

[0048] The current node electricity price is generated for each node based on the carbon potential signal of each current node.

[0049] Furthermore, the lower-level model generates a current collaborative scheduling scheme for the data center based on the carbon potential signal of each current node and the electricity price of each current node, including:

[0050] Based on the carbon potential signals of each current node, the electricity price of the current node, and the data center's own configuration parameters, the previous collaborative scheduling scheme is optimized under the carbon emission flow analysis model and the second constraint, with the goal of minimizing electricity purchase cost and carbon emission cost, to generate the current collaborative scheduling scheme for the data center.

[0051] The second constraint includes load regulation constraint, data center memory constraint, power system-data center power balance constraint, and data center carbon emission flow balance constraint. Attached Figure Description

[0052] Figure 1 A flowchart illustrating a data center carbon-electricity coordinated scheduling method based on carbon emission flows, provided as an embodiment of this application;

[0053] Figure 2 This is a schematic diagram of a data center carbon-electricity coordinated scheduling system based on carbon emission flow, provided as an embodiment of this application. Detailed Implementation

[0054] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0055] It should be noted that the step numbers in this document are only for the convenience of explaining the specific embodiments and are not intended to limit the order in which the steps are performed. In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0056] Example 1:

[0057] like Figure 1 As shown, to address the problem in existing technologies that it is difficult to simultaneously optimize energy efficiency and carbon emissions during data center scheduling optimization, Embodiment 1 provides a data center carbon-electricity coordinated scheduling method based on carbon emission flows, including steps S1-S3:

[0058] Step S1: Calculate the initial power purchase of the data center in the current scheduling cycle based on the data center's load data;

[0059] Step S2: Input the initial power purchase into a preset two-layer optimization scheduling model, so that the two-layer optimization scheduling model performs several iterations of optimization based on the initial power purchase under the constraints of a preset carbon emission flow analysis model, and generates the final collaborative scheduling scheme for the data center in the current scheduling cycle. The final collaborative scheduling scheme includes the power system power generation output plan, the final power purchase, the local power equipment operation scheme, and the computing task allocation scheme. The carbon emission flow analysis model is obtained by mathematically modeling the operation mode of the power system and the data center and the carbon emission flow between various nodes, including the data center power demand model, the data center power demand adjustment model, the data center local power equipment operation model, and the carbon emission flow model.

[0060] Step S3: Adjust the workload of each computing node and the operation strategy of each local power device in the data center during the current scheduling cycle according to the final collaborative scheduling scheme;

[0061] In each iteration of the optimization process, the upper-level model of the two-layer optimization scheduling model generates the current node carbon potential signal and the current node electricity price of each node based on the previous collaborative scheduling scheme and the real-time operation data of the power system. The node carbon potential signal is the carbon emission corresponding to the unit power consumption of the node. The lower-level model of the two-layer optimization scheduling model generates the current collaborative scheduling scheme of the data center based on the current node carbon potential signal and the current node electricity price.

[0062] This application provides a carbon-electricity coordinated scheduling method for data centers based on carbon emission flows. It calculates the initial power purchase of the data center and inputs it into a two-layer optimization scheduling model. Through alternating iterative optimization using the two-layer model, a final coordinated scheduling scheme for the data center in the current scheduling cycle is generated. Based on this final coordinated scheduling scheme, the workload of the data center and the operating status of local power equipment are adjusted. In this application embodiment, the upper-layer model transmits node carbon potential signals to the lower-layer model. The lower-layer model adjusts the data center's operating strategy based on the feedback node carbon potential signals and transmits this adjustment back to the upper-layer model. Through an iterative process, the carbon-electricity signals are dynamically updated, and the scheduling strategy is gradually optimized. Ultimately, a carbon-electricity coordinated scheduling scheme with the lowest carbon-electricity cost and optimal system energy efficiency is obtained. This achieves bidirectional coordinated scheduling between the power system and the data center, as well as dynamic matching of carbon emission flows, improving energy utilization efficiency while reducing carbon emissions. Furthermore, this application also introduces a carbon emission flow analysis model to further constrain the scheduling optimization process. This model integrates multiple dimensions such as electricity demand, regulation capacity, local power generation, energy storage, and carbon flow, and tracks carbon emissions from power generation, energy storage to electricity consumption, supporting refined carbon accounting and emission reduction strategy formulation, and improving the scientific nature and comprehensiveness of coordinated scheduling.

[0063] In a preferred embodiment, an operational data system is constructed by integrating data center computing load, power consumption parameters, renewable energy output, and market signals. This system encompasses power demand calculation, renewable energy deduction, carbon emission intensity calculation, and dynamic carbon potential generation. First, the total power demand of the data center is calculated based on unit load energy consumption and energy efficiency ratio. Then, renewable energy output is deducted to obtain the grid-purchased power. Next, carbon emission intensity is calculated based on generator output and carbon emission factors in the power system, and node carbon potential signals are generated through an upper-level model. The upper-level model transmits the node carbon potential signals and dynamic electricity prices to the lower-level model, which in turn feeds back the adjusted load distribution and actual purchased power, forming a two-way closed loop. This method aims to achieve real-time assessment of green power consumption in data centers and two-way closed-loop optimization of carbon-electricity coordinated scheduling. By guiding load spatiotemporal migration through carbon potential signals, it promotes the efficient use of renewable energy in data centers and reduces carbon emissions.

[0064] In one possible implementation, the carbon emission flow analysis model is constructed by mathematically modeling the operation of the power system and data center, as well as the carbon emission flows between various nodes, including:

[0065] Based on the calculated relationship between the online load, online load energy efficiency ratio, offline load, offline load energy efficiency ratio, and power demand of each computing node in the data center, a power demand model for the data center is constructed.

[0066] Based on the calculation relationship between the transferable online load, the deferred offline load, and the adjustable power demand of each computing node in the data center, a power demand adjustment model for the data center is constructed.

[0067] Based on the calculated relationship between the real-time active power, maximum active power, conversion coefficient, power generation efficiency, and the light radiation intensity and temperature of the surrounding environment of each photovoltaic power generation device in the data center, an operation model of the photovoltaic power generation device in the data center is constructed.

[0068] Based on the calculation relationship between the real-time storage capacity, maximum storage capacity, charging and discharging efficiency, real-time charging and discharging power, and maximum charging and discharging power of each energy storage device in the data center, an operation model of the data center energy storage device is constructed.

[0069] Based on the active power of each node in the power system, the active power flow between nodes, the carbon emission intensity of each node, the carbon flow density between nodes, the node carbon potential signal and carbon emission of each node, a model for obtaining the carbon emission flow of power grid generation is constructed.

[0070] Based on the calculated relationship between the carbon emission factors, active power, and carbon emissions of each carbon emission power generation device in the data center, a carbon emission flow model for data center power generation is constructed.

[0071] Based on the charging and discharging carbon flow rate of each energy storage device in the data center, the real-time charging power of the local power generation equipment to the energy storage device and the corresponding carbon emission factor, the real-time charging power of the power grid to the energy storage device and the corresponding node carbon potential signal, as well as the calculated relationship between the discharge carbon flow density, storage carbon emissions and discharge carbon emissions, a carbon emission flow model for data center energy storage devices is constructed.

[0072] This application provides a method for constructing a carbon emission flow analysis model, which integrates multiple dimensions such as power demand, regulation capacity, local power generation, energy storage, and carbon flow to improve the scientificity and comprehensiveness of coordinated scheduling. By coupling the data center power demand model, the data center power demand regulation model, the data center local power equipment operation model, and the carbon emission flow model, it can comprehensively characterize the dynamic coupling relationship between the power output behavior of the power system generation unit, the load handling behavior of the data center, and the carbon emission behavior, realize the dynamic assessment of green power consumption, thereby generating carbon emission signals and operating characteristic parameters that reflect the carbon emission characteristics of nodes, and providing the necessary data and model foundation for subsequent carbon-electricity coordinated optimization scheduling.

[0073] In a preferred embodiment, the carbon emission flow analysis model is constructed as follows:

[0074] 1. Data Center Power Demand Model

[0075] The main energy-consuming equipment in a data center is information technology (IT) equipment and cooling systems. Energy consumption is closely related to the quantity and type of workload. Workloads in interconnected data centers are categorized into online and offline workloads based on their latency. Online workloads require real-time server responses and are therefore not time-adjustable, but each online workload typically requires a minimum amount of computing power. Cloud service providers can schedule online workloads across different data centers, offering significant spatial flexibility. Offline workloads require substantial computation and are typically processed on local servers. However, due to their lower priority, these workloads can be delayed within a certain timeframe, providing time flexibility.

[0076] For a single data center, the power demand can be expressed as:

[0077] P D,i,t =(x i,t e x +y i,t e y )U i

[0078] Among them, P D,i,t This represents the power demand of data center i at time t, where the subscript D indicates the demand to distinguish it from other power types; x i,t ,yi,t These represent the online and offline loads of data center i during the computing cycle, respectively; e x ,e y These represent the energy consumption of the online and offline loads of the data center's i-processing unit, respectively; U i This represents the ratio of data center energy consumption to IT equipment energy consumption.

[0079] 2. Data Center Power Demand Adjustment Model

[0080] Data centers offer spatial flexibility for online loads and temporal flexibility for offline loads to optimize computing power requirements. This allows for efficient adjustment of power demands across interconnected data centers. The adjustable capacity of a single data center can be calculated using the following expression:

[0081] ΔP D,i,t ={(Δx) i,out,t -Δx i,in,t )e x +Δy i,delay,t e y}U i

[0082] Where, ΔP D,i,t Δx represents the adjustment capability that data center i can provide in period t, with negative values ​​indicating a downward adjustment and positive values ​​indicating an upward adjustment; i,out,t ,Δx i,in,t Δy represents the online load of the data center during the output and input periods. i,delay,t This indicates the amount of offline load that the data center chooses to delay processing.

[0083] 3. Operation Model of Photovoltaic Power Generation Equipment in Data Centers

[0084] Due to the need for secure and low-carbon operations, data centers typically invest in distributed power and energy storage equipment. Cloud service operators can coordinate the power behavior between grid connection points, distributed power sources, and energy storage systems to effectively manage energy and meet the demand for a continuous, stable, green, and low-carbon power supply.

[0085] Distributed power for data center investments is typically distributed photovoltaic (PV) power, and the simplified output model is as follows:

[0086]

[0087] 0≤P PV,t ≤θ PV,t P PV,max,t

[0088] Among them, P PV,max,t It is the maximum active power of distributed photovoltaic power generation at time t; I tI0 and I0 represent the light radiation intensity in the actual and standard test environments, respectively. It is the maximum active power of the photovoltaic generator set under standard testing conditions; A t A0 and μ are the temperatures in the actual and standard test environments, respectively; max It is the conversion factor; A NOC This refers to the cell temperature of the photovoltaic generator set during normal operation; P PV,t It is the active power of the photovoltaic generator at time t; θ PV,t It is the power generation efficiency of distributed photovoltaic (PV) at time t, which is related to the light-receiving area of ​​the distributed PV system.

[0089] 4. Data Center Energy Storage Equipment Operation Model

[0090] For energy storage devices, the corresponding power model is:

[0091] E S,t+1 =E S,t +η in P S,in,t -P S,out,t / η out

[0092] 0≤E S,t ≤E S,max

[0093] 0≤P S,in,t ≤P S,in,max

[0094] 0≤P S,out,t ≤P S,out,max

[0095] P S,in,t P S,out,t =0

[0096] Among them, E S,t P is the storage capacity of the energy storage device at time t; S,in,t and P S,out,t These are the charging and discharging power of the energy storage device at time t, respectively; η in and η out These are charging and discharging efficiencies, respectively; E S,max It is the maximum capacity of the energy storage device; P S,in,max and P S,out,max These are the maximum charging power and maximum discharging power of the energy storage device, respectively.

[0097] 5. Carbon emission flow model for power generation in power grids

[0098] Most carbon emissions from data centers are indirect emissions caused by electricity consumption. The corresponding power sources and carbon emission sources can be divided into grid power consumption and data center equipment power consumption. Therefore, a grid power generation carbon emission flow model is used to characterize the carbon emission process of grid power consumption, while a data center power generation carbon emission flow model and a data center energy storage equipment carbon emission flow model are used to characterize the carbon emission process of data center equipment power consumption.

[0099] In the process of power consumption in the power grid, the active power flow calculation of the branches is performed first:

[0100]

[0101] Among them, X k,j δ represents the reactance of branch (k,j); k (t), δ j (t) represents the phase angle of k and j at time t; P k,j,t This represents the active power flow from node k to node j.

[0102] Based on the calculation method for carbon emission flows, the carbon emissions corresponding to a unit of electricity consumption by a user in a power system node can be expressed as the nodal carbon potential ρ. N，j，t The representation is as follows:

[0103]

[0104] in, This represents the set of upstream nodes that connect node j, where each upstream node represents the branch direction from node j. P represents the set of generators connected by node j; B,kj,t This represents the active power flow from the connecting node to the branch (k,j) at time t; ρ B,kj,t P represents the branch carbon flux density at time t connecting the node to the (k,j) branch; G,g,t P is the active power of generator g at time t; G,g,t It is the carbon emission intensity of generator set g at time t.

[0105] Therefore, the carbon emissions C corresponding to the direct consumption of grid energy by user j at time t within a unit period are as follows. N,t for:

[0106] C N,t =ρ N,j,t P N→L,j,t

[0107] Among them, P N→L,j,t This represents the active power consumed by data center users directly from the power grid at time t.

[0108] 6. Data Center Power Generation Carbon Emission Flow Model

[0109] In the power generation process of a data center, the carbon emissions corresponding to the power generation of distributed power sources within a unit period after time t can be expressed as:

[0110] C RDG,t =ρ RDG P RDG→L,t

[0111] Among them, P RDG→L,t ρ represents the active power of the local distributed power source to the electrical load; RDG It is the carbon emission factor of local distributed power generation. Renewable energy distributed power generation, such as distributed photovoltaics, does not involve fossil fuel consumption or greenhouse gas emissions in the power generation process, and is therefore considered a zero-carbon power source.

[0112] 7. Carbon emission flow model for data center energy storage equipment

[0113] Energy storage devices achieve the time transfer of electrical energy through charging and discharging processes, accompanied by the accumulation and release of carbon emissions. The carbon emissions of energy storage devices are defined as the carbon emissions on the generation side corresponding to the stored electricity, and can be expressed as:

[0114] C S,t+1 =C S,t +R S,in,t -R S,out,t

[0115] Among them, C S,t+1 It refers to the carbon emissions stored by the user at time t+1 in the energy storage device; R S,in,t and R S,out,t These are the carbon flow rates during external charging and discharging of the energy storage device at time t, representing the carbon emissions circulating with the carbon emission flow per unit time during the charging and discharging process.

[0116] During the charging process of an energy storage device, the externally characterized charging power is P. s,in,t The actual charging power is η in P s,in,t The power loss during charging is (1-η) in )P s,in,t The carbon emissions from both actual charging power and charging power loss are quantified together to store carbon emissions. For energy storage devices in data centers, the charging power source is mainly local distributed renewable energy and the power grid; therefore, the charging carbon flow rate can be characterized as:

[0117] R S,in,t =P RDG→S,t ρ RDG +P N→S,t ρ N,j,t

[0118] Where: P RDG→S,tThis represents the active power of the local distributed power source in charging the energy storage device.

[0119] P N→S,t This represents the active power of the power grid in charging energy storage devices.

[0120] During the discharge process of an energy storage device, the external characterization of the discharge power is P. s,out,t The actual discharge power is P s,out,t / η out The power loss during the discharge process is (1-1 / η) out )P S,out,t The discharge carbon flux density of an energy storage device is defined as the ratio of stored carbon emissions to stored power, and can be expressed as:

[0121] ρ S,out,t =C S,t / E S,t

[0122] During discharge, the carbon emission stream from the energy storage device is divided into two parts. One part is related to the discharge power and flows to other devices via the discharge carbon flow rate R. S,out,t express:

[0123] R S,out,t =P S,out,t ρ S,out,t

[0124] Another portion of the energy loss is related to discharge and is measured by the actual investor in the equipment. This portion of carbon emissions is as follows:

[0125] C S,t =(1-1 / η) out )P S,out,t ρ S,out,t

[0126] It should be noted that the data center local power equipment operation model in the carbon emission flow analysis model consists of a data center photovoltaic power generation equipment operation model and a data center energy storage equipment operation model. The carbon emission flow model consists of a grid power generation carbon emission flow model, a data center power generation carbon emission flow model, and a data center energy storage equipment carbon emission flow model. Through this carbon emission flow analysis model, not only can real-time dynamic tracking of carbon emissions at each node be achieved, but it can also provide accurate and time-sensitive carbon emission signals for subsequent carbon-electricity coordinated optimization scheduling, thereby achieving precise allocation of carbon emission responsibility.

[0127] Furthermore, in step S1, calculating the initial power purchase of the data center in the current scheduling cycle based on the data center's load data includes:

[0128] Obtain the load data of the data center in the current scheduling cycle. The load data includes online load, online load energy efficiency ratio, offline load, and offline load energy efficiency ratio.

[0129] The total power demand of the data center is calculated based on the load data using the carbon emission flow analysis model.

[0130] The initial power purchase capacity of the data center in the current scheduling cycle is calculated based on the output of the data center's local power generation equipment, the charging and discharging power of the energy storage equipment, and the total power demand.

[0131] Specifically, the total power demand of the data center is first calculated based on the unit load energy consumption and energy efficiency ratio. Then, based on the idea of ​​deducting the renewable energy output from the total power demand of the data center, the sum of the total power demand of the data center and the charging power of the energy storage equipment is subtracted from the sum of the output of the local power generation equipment and the discharge power of the energy storage equipment to calculate the initial power purchase of the data center in the current scheduling cycle.

[0132] This application provides a method for calculating initial power purchase capacity. The method calculates the initial power purchase capacity based on load data, local power generation equipment output, and energy storage equipment discharge power. During the calculation process, the online / offline load and its energy efficiency ratio are accurately calculated. At the same time, it takes into account that the local power generation equipment and energy storage equipment of the data center also provide a part of the power supply to the data center, ensuring the rationality and accuracy of the initial power purchase capacity calculation and providing data support for subsequent coordinated scheduling.

[0133] In one possible implementation, in step S2, the upper-level model generates the current node carbon potential signal and current node electricity price for each node based on the previous coordinated scheduling scheme and real-time power system operation data, including:

[0134] Based on the previous coordinated dispatch scheme and real-time power system operation data, and under the carbon emission flow analysis model and the first constraint, the power generation output of each node is optimized with the goal of minimizing the total operating cost of the power system, and the current power generation output plan is generated.

[0135] Based on the current power generation output plan, the current node carbon potential signal of each node is calculated;

[0136] The current node electricity price is generated for each node based on the carbon potential signal of each current node.

[0137] Furthermore, the real-time operating data of the power system includes the rated capacity, ramp rate, and power generation cost coefficient of each generator unit in the power system;

[0138] The power generation cost coefficient includes a nonlinear power cost coefficient, a linear power cost coefficient, and an no-load cost coefficient. The power generation cost coefficient is obtained by actually testing the fuel consumption of the generator set at different power levels and fitting the data curve.

[0139] The first constraint includes power system power balance constraint, carbon emission flow balance constraint, and generator set power constraint.

[0140] This application provides a method for generating the current node's carbon potential signal and current node's electricity price using an upper-level model. The upper-level model establishes a scheduling optimization framework from the perspective of a power system operator, aiming to minimize the total operating cost of the power system, optimize output allocation, power flow allocation, and carbon emission flow allocation for each unit, and generate a corresponding power generation plan. Furthermore, based on the power generation plan, a node carbon potential signal and electricity price are generated. These two signals constitute the boundary conditions and economic incentives for optimizing the lower-level model, and serve as the main way for the upper layer to influence the lower layer in the entire two-layer iterative solution mechanism, promoting the efficient use of renewable energy in data centers, improving the energy utilization efficiency of data centers, and reducing carbon emissions. In addition, this embodiment improves the feasibility and economics of the optimization results by fitting the power generation cost coefficient through actual testing. Upper-level optimization is performed based on power balance constraints, carbon emission flow balance constraints, and generator power constraints. Power balance constraints and generator power constraints ensure stable operation of the power system, while carbon emission flow balance constraints ensure a dynamic balance between the carbon flow rate of each branch of the power system and the carbon flow rate of each node in the data center, ensuring the rationality of the optimization results.

[0141] Furthermore, the step of calculating the current node carbon potential signal of each node based on the current power generation output plan includes:

[0142] Based on the current power generation output plan and the topological relationship between each node, determine the active power of each node and the active power flow between each node;

[0143] Based on the active power of each node, the active power flow between nodes, the carbon emission intensity of each node, and the carbon flow density between nodes, the current node carbon potential signal of each node is calculated through the carbon emission flow analysis model.

[0144] This application provides a method for calculating node carbon potential signals. The method calculates node carbon potential signals based on active power flow and carbon emission intensity. In the calculation process, the method combines the power grid topology relationship and uses parameters such as active power flow and carbon flow density to reflect the mutual influence between nodes in the carbon emission process, dynamically reflecting the carbon emission contribution of the power transmission path and improving the accuracy of carbon potential signal calculation.

[0145] In a preferred embodiment, the calculation of the upper-level model begins by acquiring key operational data such as generator capacity, ramp rate, and cost coefficients, followed by economic dispatch calculations. The goal is to minimize the total operating cost of the entire system, while also satisfying constraints to ensure a balance between grid power and carbon flow. After the calculations are completed, the upper-level model transmits the following results to the lower-level data center model: electricity price signals obtained from a 24-hour optimization calculation based on the latest system information at each decision point (e.g., hourly), and nodal carbon potential signals calculated based on the carbon emission flow analysis model. These two signals constitute the boundary conditions and economic incentives for optimizing the lower-level model and serve as the primary means for the upper level to influence the lower level in the entire two-layer iterative solution mechanism. After the lower-level model optimizes and feeds back its energy interaction plan based on these signals, the upper-level model can repeat this step. This process is repeated until the change in the power purchase plan between the upper and lower layers in two consecutive iterations is less than a preset convergence threshold (less than 1%). At this point, the model confirms that it has found a stable equilibrium solution, and the scheduling scheme at this point is considered the optimal final coordinated scheduling scheme for both sides. The upper-layer model establishes a scheduling optimization framework from the perspective of the power system operator, aiming to minimize the total operating cost of the power system and optimize output allocation, power flow allocation, and carbon emission flow allocation for each unit. The specific optimization process of the upper-layer model is as follows:

[0146] To accurately describe the power grid's operating status, the rated capacity of generator units is first collected in real time during the dispatching cycle T. Climbing rate R g Cost coefficient α g ,β g γ g ;

[0147] Unify the economic cost and carbon emission cost of the unit into a quadratic model:

[0148]

[0149] Among them, Ψ g P is the set of all generating units in the power system, and T is the set during the dispatch period; G,g,t a represents the effective power of the g-th power generation unit at time t; g ,b g ,c g These are the operating cost coefficients of the g-th power generation unit. These cost coefficients represent the nonlinear, linear, and no-load components of the generator cost, respectively. Their values ​​are obtained by actual testing of the generator's fuel consumption at different power levels and by fitting the data curves. This function explicitly characterizes the dual attributes of "economy-carbon emissions" in a single cost curve.

[0150] Regarding constraints, in addition to the constraints inherent in the carbon emission flow analysis model, there are also power system power balance constraints, carbon emission flow balance constraints, and generator power constraints. The power system power balance constraints are as follows:

[0151]

[0152] P B,kj,t =B kj (θ k -θ j )

[0153] -π≤θ i ≤π

[0154] Among them, Ψ i Let P represent the set of every node in the power system. D,i,t Let Θ represent the active power of user i at time t, and let B represent the set of branches in the power system. kj θ represents the admittance of the branch connecting node k to node j. i θ k θ j Let i, k, and j represent the phase angles of nodes i, k, and j, respectively.

[0155] The carbon emission flow balance constraint is:

[0156]

[0157] R B,kj,t =P B,kj,t ρ N,k,t

[0158] R G,g,t =P G,g,t ρ G,g,t

[0159] R D,d,t =P D,d,t ρ N,j,t

[0160] in, and Let R represent the set of downstream nodes of node j and the set of users connected by node j, respectively; B,kj,t R represents the branch carbon flux from the node to the branch (k,j); G,g,t R represents the carbon flow rate of generator set g; D,d,t The above three carbon flow rates are calculated by multiplying their respective active power and corresponding carbon potential.

[0161] The generator set power constraint is:

[0162] 0≤PG,g,f ≤P G,g,max

[0163] Among them, P G,g,max P represents the maximum effective power of the g-th power generation unit. G,g,f This represents the real-time power of the g-th power generation unit.

[0164] In one possible implementation, in step S2, the lower-level model generates a current collaborative scheduling scheme for the data center based on the carbon potential signal of each current node and the electricity price of the current node, including:

[0165] Based on the carbon potential signals of each current node, the electricity price of the current node, and the data center's own configuration parameters, the previous collaborative scheduling scheme is optimized under the carbon emission flow analysis model and the second constraint, with the goal of minimizing electricity purchase cost and carbon emission cost, to generate the current collaborative scheduling scheme for the data center.

[0166] The second constraint includes load regulation constraint, data center memory constraint, power system-data center power balance constraint, and data center carbon emission flow balance constraint.

[0167] This application provides a method for generating a current collaborative scheduling scheme using a lower-level model. The lower-level model aims to minimize electricity purchase costs and carbon emission costs by adjusting the purchased power, local power equipment operation schemes, and computing task allocation schemes for the current scheduling cycle. This achieves dynamic response to node carbon potential signals, improving the energy utilization efficiency of the data center and reducing carbon emissions. Regarding constraint settings, load regulation constraints and data center memory constraints limit the task allocation of each computing node in the data center, preventing any computing node's computing tasks from exceeding its capacity. Power system-data center power balance constraints ensure the normal operation of the data center's power generation and energy storage equipment and maintain energy exchange between the data center and the power system within normal ranges. Data center carbon emission flow balance constraints ensure normal carbon emission behavior of each device in the data center during operation. The combined constraints of these conditions ensure that the scheduling scheme meets the actual operational needs of the data center.

[0168] In a preferred embodiment, a lower-level model for power and carbon co-optimization of interconnected data centers is established from the perspective of cloud service providers. Its objective is to minimize the total operating cost of the data center, which comprises two key components: the cost of electricity purchased based on dynamic electricity pricing, and the carbon emission cost calculated based on carbon potential signals transmitted from the upper layer. During each optimization process, the lower-level model receives the following parameters: the data center's own configuration parameters (including computing power, energy storage system capacity, local photovoltaic power generation capacity, etc.); the carbon potential signal transmitted from the upper-level optimization; and constraints such as dynamic electricity pricing information provided by the grid side and the charging and discharging efficiency of the energy storage system.

[0169] In simple terms, the collaborative scheduling scheme increases the proportion of electricity purchased during periods when both electricity prices and carbon emissions are low, while adjusting the charging and discharging of energy storage and load transfer: when the carbon price period is low, the system will increase electricity purchases and charge energy storage; when the carbon price period is high, energy storage will be used to discharge first and load transfer will be carried out, and the computing tasks will be migrated to low carbon emission nodes.

[0170] The entire optimization process is constrained by the following: the grid's power purchase capacity is limited by the data center's access capacity limit; the energy storage system maintains its state of charge within a safe range (avoiding overcharging and over-discharging); load transfer ensures that it does not exceed the limits of computing resources and storage capacity; and the power balance equation guarantees that the total power supply always meets real-time demand. This model determines the processing of computing loads and the energy production / utilization management strategies for local power equipment.

[0171] The objective function consists of two parts: the power cost F for operating the data center. elc And carbon emission costs F car The objective function minimizes the electricity cost F. elc And carbon emission costs F car The sum of these factors is used to optimize data center operations, where electricity costs depend on dynamic electricity pricing and grid-purchased electricity (P). N→D,t The cost of carbon emissions consists of the carbon price and the real-time carbon emissions C. D,t The solution requires consideration of constraints such as power balance, energy storage charging and discharging limitations, and load flexibility. It necessitates the use of linear / mixed-integer programming or heuristic algorithms to dynamically adjust power purchase, energy storage, and load allocation strategies. The final output should be the optimal power purchase plan, energy storage scheduling scheme, and total cost. Furthermore, bidirectional collaborative optimization with the power grid should be achieved through carbon potential signals.

[0172] minF=F elc +F car

[0173]

[0174] Where Ψ represents the set of all data centers; λ elc This represents the electricity price that the data center purchases from the grid; PN→D,t λ represents the effective power that the data center purchases from the power grid at time t; car Indicates carbon emission price; C D,t This represents the amount of carbon emissions generated by the data center at time t.

[0175] In terms of constraints, in addition to the constraints inherent in the carbon emission flow analysis model, there are also load regulation constraints, data center memory constraints, power system-data center power balance constraints, and data center carbon emission flow balance constraints.

[0176] The load sizing constraints include online load sizing constraints and offline load sizing constraints. Online load is sensitive to processing latency, but the volume of individual data points is relatively small, and it can be distributed to geographically dispersed interconnected data centers for processing through front-end servers. The constraints are as follows:

[0177]

[0178] x i,t =X i,t +x in,i,t -x out,i,t

[0179] x in,i,t x out,i,t =0

[0180] Where: x i,t V represents the online load that data center i ultimately processes at time t; V represents the set of all interconnected data centers; X represents the online load that data center i ultimately processes at time t. i,t This represents the online load arriving at data center i at time t; x in,i,t This represents the online load transferred from other data centers to data center i at time t; x out,i,t This represents the online load that is transferred out of data center i at time t.

[0181] Offline workloads can be processed with latency in the data center, but are constrained by data center offline workload storage limitations, specifically:

[0182] y i,t =y i,t-1 -v i,t-1 +Y i,t

[0183] 0 <y i,t ≤y i,max

[0184]

[0185] Where: y i,t Y represents the offline load of interconnected data center i at time t. i,tThis indicates that time t is ready to reach the offline load of interconnected data center i, y i,max This indicates the maximum offline storage load of interconnected data center i. v represents the offline storage load of interconnected data center i at the end of the scheduling period. i,t-1 This represents the amount of offline load that data center i has processed at time t-1.

[0186] The specific formula for data center memory constraints is as follows:

[0187] π x,i x i,t +π y,i y i,t ≤M i

[0188] Where, π x,i and π y,i M represents the average amount of data required for data center i to process a single online load and an offline load, respectively; i This represents the memory of interconnected data center i.

[0189] The power system-data center power balance constraints need to satisfy the constraints of distributed renewable power supply and energy storage devices, as well as the constraints of the power equations, specifically:

[0190] P N→D,i,t =P N→S,i,t +P N→L,i,t

[0191] P Load,i,t =P N→L,i,t +P RDG→L,i,t +P S→L,i,t

[0192] Among them, P N→D,i,t P represents the total power flowing from the power grid to data center i. N→S,i,t and P N→L,i,t P represents the power supplied by the power grid to the energy storage device and the load, respectively. Load,i,t P represents the total power load of the data center, supplied by the power grid, distributed power sources, and energy storage devices. RDG→L,i,t P represents the power supplied to the load by the distributed power source. S→L,i,t This indicates the power supplied to the load by the energy storage device.

[0193] The specific formula for the carbon emission flow balance constraint of data centers is as follows:

[0194] C D,j,t =C S,loss,t +C RDG,t +C N,t

[0195] Among them, CD,j,t C represents the total carbon emissions of data center i at time t. S,loss,t C represents the carbon emissions generated by the loss of energy storage devices. RDG,t C represents the carbon emissions generated by locally distributed power sources. N,t This refers to carbon emissions generated from the consumption of electricity.

[0196] The data center carbon-electricity coordinated scheduling method based on carbon emission flows provided in this application includes, but is not limited to, electronic devices such as power system dispatch servers and data center energy management systems. The power system dispatch server performs calculations of the upper-level power system optimization model to determine the carbon potential of each node and the optimal power generation plan. The data center energy management system performs calculations of the lower-level optimization model, optimizing load scheduling strategies and charging / discharging plans for local energy storage devices based on received carbon potential signals. The method achieves coordinated operation of the power system and data center through a two-layer iterative optimization mechanism. The upper-level model aims to minimize grid operating costs, while the lower-level model aims to minimize the overall carbon cost of electricity for the data center. These two models achieve closed-loop optimization through dynamic carbon potential signals and electricity demand feedback. The method can be deployed on a cloud server cluster and uses blockchain technology to ensure the reliable transmission of carbon flow data and scheduling instructions.

[0197] In a preferred embodiment, collaborative optimization is achieved by constructing an IEEE-14 node test system comprising five interconnected data centers. During system initialization, detailed parameters for each data center need to be configured, including a computing power of 1500 MB / s, offline task storage capacity of 1.0 × 10^10 MB to 2.5 × 10^10 MB, and memory configuration of 7 PB to 11 PB. Simultaneously, each data center is equipped with a 10-15 MW distributed photovoltaic system and 5-7 MWh of energy storage equipment, with the maximum charging and discharging power of the energy storage system controlled within the range of 2-2.5 MW. During system operation, the power system operator calculates the dynamic carbon potential signals of each node in real time using an upper-level optimization model. These signals exhibit significant spatiotemporal differences. For example, the carbon potential of data center 1 at node 2 reaches as high as 775 kg CO2 / MWh during nighttime and evening hours, while dropping below 500 kg CO2 / MWh during peak daytime photovoltaic power generation. Data center 3 at node 9, due to its access to diverse power sources, exhibits an even larger carbon potential fluctuation range, reaching 0-323 kg CO2 / MWh.

[0198] Upon receiving carbon potential signals, the energy management systems of each data center execute refined optimization scheduling strategies. For online loads with high real-time requirements, the system employs a spatially flexible scheduling strategy. During high-carbon periods, Data Center 1 migrates 60% of its computing load to low-carbon nodes, while Data Center 3 acts as the primary load receiving node during low-carbon periods. This process must strictly ensure compliance with maximum transmission delay constraints to guarantee service quality. For computationally intensive but less time-sensitive offline loads, the system fully utilizes time flexibility for optimized scheduling. Data Center 1 processes backlogged offline tasks during peak daytime photovoltaic power generation, while Data Center 3 dynamically adjusts its processing plan based on real-time carbon potential. The energy storage system employs a "low-storage, high-discharge" optimization strategy, charging at maximum power during peak midday photovoltaic output when electricity prices are low, and discharging during high-carbon, high-electricity-price periods at night. The scheduling process requires comprehensive consideration of factors such as charging and discharging efficiency and capacity limitations.

[0199] The system employs an innovative two-layer iterative optimization mechanism to achieve collaborative operation. The lower-layer data center feeds back the optimized power demand to the upper-layer model in real time. Based on this, the upper-layer model recalculates the power flow distribution and carbon potential signal. After multiple iterations, convergence is achieved when the change in power demand is less than 1%. Throughout the optimization process, the system fully records the carbon potential signal and scheduling instructions through a blockchain platform, ensuring the authenticity and immutability of all key data. Implementation effectiveness evaluation shows that this method has achieved significant results in multiple dimensions: total data center operating costs decreased by 18.34%, carbon emissions decreased by 71.6%, and total carbon emissions at the power system level decreased by 47%. The wind and solar curtailment rate dropped significantly from 2.56% to 0.18%, fully validating the superiority and practicality of this patented method.

[0200] This method can be deployed on a cloud server cluster architecture. The upper-layer model runs on a high-performance server in the power system dispatch center, while the lower-layer model is deployed on the energy management systems of various data centers. The system supports multiple communication protocols such as MQTT and OPC UA to ensure real-time and reliable data transmission. As an optional solution, the system can use blockchain technology to achieve trusted storage and full-process traceability of carbon flow data. This collaborative optimization method based on carbon emission flows is not only applicable to the data center industry, but can also be extended to other energy-intensive fields with flexible load characteristics after appropriate adjustments, providing an innovative technical path and operable implementation plan for achieving carbon peaking and carbon neutrality goals. By establishing a carbon-electricity collaborative mechanism between the power system and electricity users, this patented method successfully solves the problem of balancing economic efficiency and environmental protection under traditional dispatching models, providing important technical support for building a new type of power system.

[0201] Example 2:

[0202] like Figure 2As shown, Embodiment 2 provides a data center carbon-electricity coordinated scheduling system based on carbon emission flow, including a power calculation module 10, a two-layer optimization module 20, and a scheduling module 30;

[0203] The power calculation module 10 is used to calculate the initial power purchase of the data center in the current scheduling cycle based on the data center's load data.

[0204] The dual-layer optimization module 20 is used to input the initial power purchase into a preset dual-layer optimization scheduling model, so that the dual-layer optimization scheduling model performs several iterations of optimization based on the initial power purchase under the constraints of a preset carbon emission flow analysis model, and generates a final collaborative scheduling scheme for the data center in the current scheduling cycle. The final collaborative scheduling scheme includes the power system power generation output plan, the final power purchase, the local power equipment operation scheme, and the computing task allocation scheme. The carbon emission flow analysis model is obtained by mathematically modeling the operation mode of the power system and the data center and the carbon emission flow between various nodes, including the data center power demand model, the data center power demand adjustment model, the data center local power equipment operation model, and the carbon emission flow model.

[0205] The scheduling module 30 is used to adjust the workload of each computing node and the operation strategy of each local power device in the data center during the current scheduling cycle according to the final collaborative scheduling scheme.

[0206] The dual-layer optimization module 20 includes an upper-layer optimization unit and a lower-layer optimization unit. In each iteration of optimization, the upper-layer optimization unit is used to generate the current node carbon potential signal and the current node electricity price of each node based on the previous collaborative scheduling scheme and the real-time operation data of the power system through the upper-layer model in the dual-layer optimization scheduling model. The node carbon potential signal is the carbon emission corresponding to the unit power consumption of the node. The lower-layer optimization unit is used to generate the current collaborative scheduling scheme of the data center based on the current node carbon potential signal and the current node electricity price through the lower-layer model in the dual-layer optimization scheduling model.

[0207] Furthermore, the power calculation module 10 calculates the initial power purchase of the data center in the current scheduling cycle based on the data center's load data, including:

[0208] Obtain the load data of the data center in the current scheduling cycle. The load data includes online load, online load energy efficiency ratio, offline load, and offline load energy efficiency ratio.

[0209] The total power demand of the data center is calculated based on the load data using the carbon emission flow analysis model.

[0210] The initial power purchase capacity of the data center in the current scheduling cycle is calculated based on the output of the data center's local power generation equipment, the charging and discharging power of the energy storage equipment, and the total power demand.

[0211] In one possible implementation, the upper-level model generates the current node carbon potential signal and current node electricity price for each node based on the previous coordinated scheduling scheme and real-time power system operation data, including:

[0212] Based on the previous coordinated dispatch scheme and real-time power system operation data, and under the carbon emission flow analysis model and the first constraint, the power generation output of each node is optimized with the goal of minimizing the total operating cost of the power system, and the current power generation output plan is generated.

[0213] Based on the current power generation output plan, the current node carbon potential signal of each node is calculated;

[0214] The current node electricity price is generated for each node based on the carbon potential signal of each current node.

[0215] Furthermore, the real-time operating data of the power system includes the rated capacity, ramp rate, and power generation cost coefficient of each generator unit in the power system;

[0216] The power generation cost coefficient includes a nonlinear power cost coefficient, a linear power cost coefficient, and an no-load cost coefficient. The power generation cost coefficient is obtained by actually testing the fuel consumption of the generator set at different power levels and fitting the data curve.

[0217] The first constraint includes power system power balance constraint, carbon emission flow balance constraint, and generator set power constraint.

[0218] Furthermore, the step of calculating the current node carbon potential signal of each node based on the current power generation output plan includes:

[0219] Based on the current power generation output plan and the topological relationship between each node, determine the active power of each node and the active power flow between each node;

[0220] Based on the active power of each node, the active power flow between nodes, the carbon emission intensity of each node, and the carbon flow density between nodes, the current node carbon potential signal of each node is calculated through the carbon emission flow analysis model.

[0221] In one possible implementation, the lower-level model generates a current collaborative scheduling scheme for the data center based on the carbon potential signal of each current node and the electricity price of the current node, including:

[0222] Based on the carbon potential signals of each current node, the electricity price of the current node, and the data center's own configuration parameters, the previous collaborative scheduling scheme is optimized under the carbon emission flow analysis model and the second constraint, with the goal of minimizing electricity purchase cost and carbon emission cost, to generate the current collaborative scheduling scheme for the data center.

[0223] The second constraint includes load regulation constraint, data center memory constraint, power system-data center power balance constraint, and data center carbon emission flow balance constraint.

[0224] In one possible implementation, the carbon emission flow analysis model is constructed by mathematically modeling the operation of the power system and data center, as well as the carbon emission flows between various nodes, including:

[0225] Based on the calculated relationship between the online load, online load energy efficiency ratio, offline load, offline load energy efficiency ratio, and power demand of each computing node in the data center, a power demand model for the data center is constructed.

[0226] Based on the calculation relationship between the transferable online load, the deferred offline load, and the adjustable power demand of each computing node in the data center, a power demand adjustment model for the data center is constructed.

[0227] Based on the calculated relationship between the real-time active power, maximum active power, conversion coefficient, power generation efficiency, and the light radiation intensity and temperature of the surrounding environment of each photovoltaic power generation device in the data center, an operation model of the photovoltaic power generation device in the data center is constructed.

[0228] Based on the calculation relationship between the real-time storage capacity, maximum storage capacity, charging and discharging efficiency, real-time charging and discharging power, and maximum charging and discharging power of each energy storage device in the data center, an operation model of the data center energy storage device is constructed.

[0229] Based on the active power of each node in the power system, the active power flow between nodes, the carbon emission intensity of each node, the carbon flow density between nodes, the node carbon potential signal and carbon emission of each node, a model for obtaining the carbon emission flow of power grid generation is constructed.

[0230] Based on the calculated relationship between the carbon emission factors, active power, and carbon emissions of each carbon emission power generation device in the data center, a carbon emission flow model for data center power generation is constructed.

[0231] Based on the charging and discharging carbon flow rate of each energy storage device in the data center, the real-time charging power of the local power generation equipment to the energy storage device and the corresponding carbon emission factor, the real-time charging power of the power grid to the energy storage device and the corresponding node carbon potential signal, as well as the calculated relationship between the discharge carbon flow density, storage carbon emissions and discharge carbon emissions, a carbon emission flow model for data center energy storage devices is constructed.

[0232] This application provides a carbon-electricity coordinated scheduling system for data centers based on carbon emission flows. It calculates the initial power purchase of the data center and inputs it into a two-layer optimization scheduling model. Through alternating iterative optimization using the two-layer model, a final coordinated scheduling scheme for the data center in the current scheduling cycle is generated. Based on this final coordinated scheduling scheme, the workload of the data center and the operating status of local power equipment are adjusted. In this application embodiment, the upper-layer model transmits node carbon potential signals to the lower-layer model. The lower-layer model adjusts the data center's operating strategy based on the feedback node carbon potential signals and transmits this adjustment back to the upper-layer model. Through an iterative process, the carbon-electricity signals are dynamically updated, and the scheduling strategy is gradually optimized. Ultimately, a carbon-electricity coordinated scheduling scheme with the lowest carbon-electricity cost and optimal system energy efficiency is obtained. This achieves bidirectional coordinated scheduling between the power system and the data center, as well as dynamic matching of carbon emission flows, improving energy utilization efficiency while reducing carbon emissions. Furthermore, this application also introduces a carbon emission flow analysis model to further constrain the scheduling optimization process. This model integrates multiple dimensions such as electricity demand, regulation capacity, local power generation, energy storage, and carbon flow, and tracks carbon emissions from power generation, energy storage to electricity consumption, supporting refined carbon accounting and emission reduction strategy formulation, and improving the scientific nature and comprehensiveness of coordinated scheduling.

[0233] For a more detailed explanation of the working principle and procedures of this embodiment, please refer to the relevant description in Embodiment 1.

[0234] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of this application. It should be understood that the above descriptions are merely specific embodiments of this application and are not intended to limit the scope of protection of this application. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application for those skilled in the art.

Claims

1. A data center carbon electricity co-scheduling method based on carbon emission flow, characterized in that, include: Calculate the initial power purchase of the data center in the current scheduling cycle based on the data center's load data. The initial power purchase is input into a preset two-layer optimization scheduling model, which then performs several iterations of optimization based on the initial power purchase under the constraints of a preset carbon emission flow analysis model. This generates a final collaborative scheduling scheme for the data center in the current scheduling cycle. The final collaborative scheduling scheme includes the power system's power generation output plan, the final power purchase, the local power equipment operation scheme, and the computing task allocation scheme. The carbon emission flow analysis model is constructed by mathematically modeling the operation mode of the power system and the data center, as well as the carbon emission flow between various nodes. It includes a data center power demand model, a data center power demand adjustment model, a data center local power equipment operation model, and a carbon emission flow model. The workload of each computing node and the operation strategy of each local power device in the data center during the current scheduling cycle are adjusted according to the final collaborative scheduling scheme. In each iteration of the optimization process, the upper-level model of the two-layer optimization scheduling model generates the current node carbon potential signal and the current node electricity price of each node based on the previous collaborative scheduling scheme and the real-time operation data of the power system. The node carbon potential signal is the carbon emission corresponding to the unit power consumption of the node. The lower-level model of the two-layer optimization scheduling model generates the current collaborative scheduling scheme of the data center based on the current node carbon potential signal and the current node electricity price.

2. The data center carbon-electricity coordinated scheduling method based on carbon emission flows as described in claim 1, characterized in that, The step of calculating the initial power purchase capacity of the data center in the current scheduling cycle based on the data center's load data includes: Obtain the load data of the data center in the current scheduling cycle. The load data includes online load, online load energy efficiency ratio, offline load, and offline load energy efficiency ratio. The total power demand of the data center is calculated based on the load data using the carbon emission flow analysis model. The initial power purchase capacity of the data center in the current scheduling cycle is calculated based on the output of the data center's local power generation equipment, the charging and discharging power of the energy storage equipment, and the total power demand.

3. The data center carbon-electricity coordinated scheduling method based on carbon emission flows as described in claim 1, characterized in that, The upper-level model generates the current node carbon potential signal and current node electricity price for each node based on the previous coordinated scheduling scheme and real-time power system operation data, including: Based on the previous coordinated dispatch scheme and real-time power system operation data, and under the carbon emission flow analysis model and the first constraint, the power generation output of each node is optimized with the goal of minimizing the total operating cost of the power system, and the current power generation output plan is generated. Based on the current power generation output plan, the current node carbon potential signal of each node is calculated; The current node electricity price is generated for each node based on the carbon potential signal of each current node.

4. The data center carbon-electricity coordinated scheduling method based on carbon emission flows as described in claim 3, characterized in that, The real-time operating data of the power system includes the rated capacity, ramp rate, and power generation cost coefficient of each generator unit in the power system. The power generation cost coefficient includes a nonlinear power cost coefficient, a linear power cost coefficient, and an no-load cost coefficient. The power generation cost coefficient is obtained by actually testing the fuel consumption of the generator set at different power levels and fitting the data curve. The first constraint includes power system power balance constraint, carbon emission flow balance constraint, and generator set power constraint.

5. The data center carbon-electricity coordinated scheduling method based on carbon emission flows as described in claim 3, characterized in that, The step of calculating the current node carbon potential signal of each node according to the current power generation output plan includes: Based on the current power generation output plan and the topological relationship between each node, determine the active power of each node and the active power flow between each node; Based on the active power of each node, the active power flow between nodes, the carbon emission intensity of each node, and the carbon flow density between nodes, the current node carbon potential signal of each node is calculated through the carbon emission flow analysis model.

6. The data center carbon-electricity coordinated scheduling method based on carbon emission flows as described in claim 1, characterized in that, The lower-level model generates a current collaborative scheduling scheme for the data center based on the carbon potential signal and electricity price of each current node, including: Based on the carbon potential signals of each current node, the electricity price of the current node, and the data center's own configuration parameters, the previous collaborative scheduling scheme is optimized under the carbon emission flow analysis model and the second constraint, with the goal of minimizing electricity purchase cost and carbon emission cost, to generate the current collaborative scheduling scheme for the data center. The second constraint includes load regulation constraint, data center memory constraint, power system-data center power balance constraint, and data center carbon emission flow balance constraint.

7. A data center carbon-electricity coordinated scheduling method based on carbon emission flows as described in any one of claims 1-6, characterized in that, The carbon emission flow analysis model is obtained by mathematically modeling the operation of power systems and data centers, as well as the carbon emission flows between various nodes. This includes: Based on the calculated relationship between the online load, online load energy efficiency ratio, offline load, offline load energy efficiency ratio, and power demand of each computing node in the data center, a power demand model for the data center is constructed. Based on the calculation relationship between the transferable online load, the deferred offline load, and the adjustable power demand of each computing node in the data center, a power demand adjustment model for the data center is constructed. Based on the calculated relationship between the real-time active power, maximum active power, conversion coefficient, power generation efficiency, and the light radiation intensity and temperature of the surrounding environment of each photovoltaic power generation device in the data center, an operation model of the photovoltaic power generation device in the data center is constructed. Based on the calculation relationship between the real-time storage capacity, maximum storage capacity, charging and discharging efficiency, real-time charging and discharging power, and maximum charging and discharging power of each energy storage device in the data center, an operation model of the data center energy storage device is constructed. Based on the active power of each node in the power system, the active power flow between nodes, the carbon emission intensity of each node, the carbon flow density between nodes, the node carbon potential signal and carbon emission of each node, a model for obtaining the carbon emission flow of power grid generation is constructed. Based on the calculated relationship between the carbon emission factors, active power, and carbon emissions of each carbon emission power generation device in the data center, a carbon emission flow model for data center power generation is constructed. Based on the charging and discharging carbon flow rate of each energy storage device in the data center, the real-time charging power of the local power generation equipment to the energy storage device and the corresponding carbon emission factor, the real-time charging power of the power grid to the energy storage device and the corresponding node carbon potential signal, as well as the calculated relationship between the discharge carbon flow density, storage carbon emissions and discharge carbon emissions, a carbon emission flow model for data center energy storage devices is constructed.

8. A data center carbon-electricity coordinated scheduling system based on carbon emission flows, characterized in that, It includes a power calculation module, a two-layer optimization module, and a scheduling module; The power calculation module is used to calculate the initial power purchase of the data center in the current scheduling cycle based on the data center's load data. The dual-layer optimization module is used to input the initial power purchase into a preset dual-layer optimization scheduling model, so that the dual-layer optimization scheduling model performs several iterations of optimization based on the initial power purchase under the constraints of a preset carbon emission flow analysis model, generating a final collaborative scheduling scheme for the data center in the current scheduling cycle. The final collaborative scheduling scheme includes the power system power generation output plan, the final power purchase, the local power equipment operation scheme, and the computing task allocation scheme. The carbon emission flow analysis model is obtained by mathematically modeling the operation mode of the power system and the data center, as well as the carbon emission flow between various nodes, including the data center power demand model, the data center power demand adjustment model, the data center local power equipment operation model, and the carbon emission flow model. The scheduling module is used to adjust the workload of each computing node and the operation strategy of each local power device in the data center during the current scheduling cycle according to the final collaborative scheduling scheme. The dual-layer optimization module includes an upper-layer optimization unit and a lower-layer optimization unit. In each iteration of optimization, the upper-layer optimization unit is used to generate the current node carbon potential signal and the current node electricity price of each node based on the previous collaborative scheduling scheme and real-time power system operation data through the upper-layer model in the dual-layer optimization scheduling model. The node carbon potential signal is the carbon emission corresponding to the unit power consumption of the node. The lower-layer optimization unit is used to generate the current collaborative scheduling scheme of the data center based on the current node carbon potential signal and the current node electricity price through the lower-layer model in the dual-layer optimization scheduling model.

9. A data center carbon-electricity coordinated scheduling system based on carbon emission flows as described in claim 8, characterized in that, The upper-level model generates the current node carbon potential signal and current node electricity price for each node based on the previous coordinated scheduling scheme and real-time power system operation data, including: Based on the previous coordinated dispatch scheme and real-time power system operation data, and under the carbon emission flow analysis model and the first constraint, the power generation output of each node is optimized with the goal of minimizing the total operating cost of the power system, and the current power generation output plan is generated. Based on the current power generation output plan, the current node carbon potential signal of each node is calculated; The current node electricity price is generated for each node based on the carbon potential signal of each current node.

10. A data center carbon-electricity coordinated scheduling system based on carbon emission flows as described in claim 8, characterized in that, The lower-level model generates a current collaborative scheduling scheme for the data center based on the carbon potential signal and electricity price of each current node, including: Based on the carbon potential signals of each current node, the electricity price of the current node, and the data center's own configuration parameters, the previous collaborative scheduling scheme is optimized under the carbon emission flow analysis model and the second constraint, with the goal of minimizing electricity purchase cost and carbon emission cost, to generate the current collaborative scheduling scheme for the data center. The second constraint includes load regulation constraint, data center memory constraint, power system-data center power balance constraint, and data center carbon emission flow balance constraint.

Images