Information interaction method and system for cooperative scheduling of computing power and electric power of data center

Abstract

The invention relates to the technical field of data center interaction, and particularly discloses an information interaction method and system for computing power and electric power collaborative scheduling of a data center, which integrates multi-source heterogeneous data and constructs a space-time correlation model to realize accurate prediction and dynamic modeling of resource demands of the data center. The real-time learning ability of the DRL and the discrete decision advantage of the MIP are utilized to realize multi-objective synchronous optimization, the energy efficiency and economy of the data center are significantly improved, the multi-objective optimization can realize "computing power-electric power-heating power" coupling, the block chain technology ensures that data cannot be tampered and is transparent and credible, the security and auditing performance of the system are improved, and the system performance is improved. Carbon footprint tracking, green power authentication and PUE index dynamic energy efficiency optimization are supported, green power use is directly stimulated, the carbon footprint and operation cost is reduced, thermal data is used for cooling control, power data is used for task scheduling, resource conflicts are avoided, the limitation of traditional single resource scheduling is broken through, and the response speed is increased.

Description

Technical Field

[0001] The present invention relates to the field of data center interaction technology, and in particular to an information interaction method and system for coordinated scheduling of data center computing power and electricity. Background Art

[0002] At present, with the continuous expansion of the scale of data centers, the demand for coordinated computing power scheduling and power management is becoming increasingly urgent. The traditional independent scheduling mode can no longer meet the needs of efficient operation. The existing system usually adopts an architecture that separates the computing power monitoring module and the power management module.

[0003] However, in existing technologies, data centers have high energy consumption and a low proportion of green electricity, which does not meet the needs of low-carbon emissions. They are unable to dynamically match computing power and electricity resource supply, resulting in low resource utilization. Summary of the Invention

[0004] The purpose of the present invention is to provide an information interaction method and system for coordinated scheduling of data center computing power and electricity, aiming to solve the technical problems in the existing technology that data centers have high energy consumption, low proportion of green electricity, do not meet low-carbon emission requirements, cannot dynamically match computing power and electricity resource supply, and lead to low resource utilization.

[0005] To achieve the above objectives, the present invention adopts an information interaction method for coordinated scheduling of data center computing power and electricity, comprising the following steps: Through multimodal collection, calculation and integration of data center power load curves, computing task types, thermal system status, and carbon emission intensity data, a spatiotemporal correlation model is constructed. A multi-objective optimization algorithm (MIP) combining improved deep reinforcement learning (DRL) and mixed integer programming (MIP) was used to simultaneously optimize the computing power cost, carbon emission data, and thermal data of the data center. Use blockchain distributed ledgers to record data centers’ power sources, computing task intensity, carbon emission intensity, and thermal system status data; Data centers dynamically adjust energy efficiency based on PUE and encourage green electricity consumption; Through data interaction, the coordinated scheduling and coupling optimization of the data center's computing power, electricity, and heat networks can be achieved.

[0006] The multimodal calculation and integration of the data center's power load curve, computing task type, thermal system status, and carbon emission intensity data, and the construction of a spatiotemporal correlation model, includes the following steps: Collect data center real-time power consumption, time-of-use electricity prices, renewable energy output ratio, and grid congestion signals; Collect task priorities, computational accuracy, data dependencies, and portability of data centers; Collect data center inlet and outlet air temperature, cooling water flow, PUE value, meteorological data, and waste heat recovery efficiency; Collect data on the data center's regional power grid carbon emission factor, green electricity consumption ratio, IT equipment carbon footprint, and indirect emissions from cooling systems; Unify the above asynchronous data streams to the same time granularity and use a sliding window mechanism to process delayed data; Design a dynamic spatiotemporal correlation matrix to quantify the coupling strength between nodes; The power load curve, computing task sequence, thermal and carbon emission data are integrated through a multimodal feature fusion architecture.

[0007] Among them, the power load curve processes the spatial dependency and temporal dynamics of power data through the spatiotemporal graph convolutional network ST-GCN, and the computing power task sequence processes the high-dimensional computing power task sequence through the Mamba-SSM state space model, modeling long-distance dependencies with linear complexity.

[0008] The upper-level algorithm architecture adopts DRL, and the lower-level algorithm architecture adopts MIP. Soft Actor-Critic is used in DRL to optimize the dynamic strategy. The real-time PUE value, available green electricity ratio, task queue length, cooling system COP, task migration, cooling pump frequency adjustment, and green electricity purchase volume interact with the environment. The data after each step of interaction with the environment is pushed into the experience pool to train the network. Finally, only the recommended action is output, but it is not directly executed by DRL. The upper-level algorithm architecture DRL serves as the initial solution of the lower-level algorithm architecture MIP. The lower-level algorithm architecture MIP uses the Gurobi solver to process mixed integer programming. The objective function is: Where, , , All are multi-objective weight coefficients; is the total electricity cost, in yuan; is the carbon emissions during the period, unit: kg; is the thermal constraint violation, unit ; Constraints include power balance and thermal constraints; Thermal constraints: Where, The cabinet inlet air temperature, unit: °C; Power balance: Where, The power purchased from the grid, in kW; The real-time photovoltaic power generation on site, in kW; is the battery charging and discharging power, in kW; is the total power of IT equipment, in kW; is the total power of the cooling system, in kW.

[0009] Among them, the power source, computing power task intensity, carbon emission intensity, and thermal system status are checked through the Hyperledger Fabric private chain. The power source is uploaded to the chain at a frequency of 15 minutes to verify that the green electricity certificate matches the real-time power generation. The computing power task intensity is uploaded to the chain at a frequency of 5 minutes to verify that the task resource request is consistent with the scheduling record. The carbon emission intensity is uploaded to the chain at a frequency of 60 minutes to call the IPCC coefficient calculation and cross-validation. The thermal system status is uploaded to the chain at a frequency of 1 minute, and the temperature exceeds the limit to trigger the early warning contract.

[0010] Among them, real-time calculation of data center power utilization efficiency: Where, Real-time power utilization efficiency for data centers; is the total input active power of the data center, in kW; is the total active power of IT equipment, in kW, where Collect UPS, PDU, and chiller data via Modbus TCP.

[0011] When PUE is greater than 1.3, the cooling pump frequency conversion is triggered, the current frequency is reduced by 10%, and the chilled water temperature set value is increased by 1°C each time; When PUE is less than 1.2, green electricity is consumed first and the battery is discharged to 20% SOC.

[0012] Among them, green electricity premium = benchmark electricity price × (1-green electricity ratio) × carbon price coefficient; when green electricity is surplus, the delay tolerance task is activated and compensation is 0.02~0.1 yuan / kWh.

[0013] Among them, electricity and computing power interact: the power grid dispatch system sends price signals and carbon signals, and the data center feedbacks the list of tasks that can be reduced and delay-tolerant; Computing power and thermal interaction: When the cabinet temperature exceeds the preset value, the scheduler prioritizes migrating tasks to the cooler cabinet. Interaction between heat and electricity: When the COP of the cooling system decreases, the load rate of IT equipment is reduced.

[0014] The present invention also provides an information interaction system for coordinated scheduling of computing power and electricity in a data center, including an FPGA, a multimodal acquisition module, a collaborative optimization module, a blockchain scheduling module, and an energy efficiency management module; The multimodal acquisition module is used to collect computing power, electricity, and thermal parameters and build a spatiotemporal correlation model; The collaborative optimization module is used to simultaneously optimize the computing power cost, carbon emission data and thermal data of the data center through DRL and MIP; The blockchain scheduling module is used to realize the trustworthiness of cross-domain scheduling of power sources, computing task intensity, carbon emission intensity, and thermal system status; The energy efficiency management module is used to dynamically adjust energy efficiency and encourage green electricity consumption; The FPGA is used to process real-time data plane tasks, realize the instant update of the scheduling algorithm, and is used to connect and integrate the multimodal acquisition module, the collaborative optimization module, the blockchain scheduling module and the energy efficiency management module.

[0015] The present invention provides an information interaction method and system for coordinated scheduling of data center computing power and electricity, which integrates multi-source heterogeneous data and constructs a spatiotemporal correlation model to achieve accurate prediction and dynamic modeling of data center resource requirements. The spatiotemporal model can capture the load changes of the data center at different times and spaces, thereby providing a basis for subsequent optimization. The real-time learning capability of DRL and the discrete decision-making advantage of MIP are utilized to achieve synchronous optimization of multiple objectives, significantly improving the energy efficiency and economy of the data center. Multi-objective optimization can achieve "computing power-electricity-heat" coupling, and blockchain technology is used to ensure that data cannot be tampered with, is transparent and trustworthy, thereby improving the security and auditability of the system, supporting carbon footprint tracking and green electricity certification, and dynamically optimizing energy efficiency through the PUE indicator and directly incentivizing the use of green electricity, reducing carbon footprint and operating costs. The coupled optimization of computing power, electricity and heat eliminates the island effect. Thermal data is used for cooling control, and electricity data is used for task scheduling to avoid resource conflicts and break through the limitations of traditional single resource scheduling. The hybrid algorithm of DRL and MIP improves the optimization efficiency of multiple objectives and increases the response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 This is a flowchart of the steps of an information interaction method for coordinated scheduling of data center computing power and electricity of the present invention.

[0018] Figure 2 The present invention Figure 1 Flow chart of the method steps of step S1 in FIG.

[0019] Figure 3 It is a system diagram of an information interaction system for coordinated scheduling of data center computing power and electricity according to the present invention.

[0020] 1-FPGA, 2-multimodal acquisition module, 3-collaborative optimization module, 4-blockchain scheduling module, 5-energy efficiency management module. DETAILED DESCRIPTION

[0021] See also Figure 1 and Figure 2 The present invention provides an information interaction method for coordinated scheduling of computing power and power in a data center, comprising the following steps: S1: Through multimodal collection, calculation, and integration of data center power load curves, computing task types, thermal system status, and carbon emission intensity data, a spatiotemporal correlation model is constructed. S2: Uses a multi-objective optimization algorithm (MIP) that combines improved deep reinforcement learning (DRL) with mixed integer programming to simultaneously optimize data center computing costs, carbon emissions, and thermal data. S3: Uses blockchain distributed ledgers to record data center power sources, computing task intensity, carbon emission intensity, and thermal system status data; S4: Data centers dynamically adjust energy efficiency based on PUE and encourage green electricity consumption; S5: Realize coordinated scheduling and coupling optimization of the data center's computing power, electricity, and heat networks through data interaction.

[0022] In this embodiment, by integrating multi-source heterogeneous data and constructing a spatiotemporal correlation model, accurate prediction and dynamic modeling of data center resource demand can be achieved. The spatiotemporal model can capture the load changes of the data center at different times and spaces, thus providing a basis for subsequent optimization. By utilizing the real-time learning capability of DRL and the discrete decision-making advantages of MIP, simultaneous optimization of multiple objectives can be achieved, significantly improving the energy efficiency and economy of the data center. Multi-objective optimization can achieve "computing power-electricity-heat" coupling. Blockchain technology is used to ensure that data cannot be tampered with, transparent and trustworthy, thereby improving the security and auditability of the system, supporting carbon footprint tracking and green electricity certification. The PUE indicator dynamically optimizes energy efficiency and directly incentivizes the use of green electricity, reducing carbon footprint and operating costs. The coupled optimization of computing power, electricity, and heat eliminates the island effect. Thermal data is used for cooling control, and electricity data is used for task scheduling to avoid resource conflicts and break through the limitations of traditional single resource scheduling. The hybrid algorithm of DRL and MIP improves the optimization efficiency of multiple objectives and increases response speed.

[0023] Furthermore, the multimodal calculation and integration of the data center's power load curve, computing task type, thermal system status, and carbon emission intensity data, and the construction of a spatiotemporal correlation model, includes the following steps: S101: Collects data center real-time power consumption, time-of-use electricity prices, renewable energy output ratio, and grid congestion signals; S102: Collect task priorities, computing accuracy, data dependencies, and portability of the data center; S103: Collect data center inlet and outlet air temperatures, cooling water flow, PUE value, meteorological data, and waste heat recovery efficiency; S104: Collect data on the data center's regional power grid carbon emission factor, green electricity consumption ratio, IT equipment carbon footprint, and indirect emissions from the cooling system; S105: Unify the above asynchronous data streams to the same time granularity and use a sliding window mechanism to process delayed data; S106: Design a dynamic spatiotemporal correlation matrix to quantify the coupling strength between nodes; S107: Fusion of power load curves, computing task sequences, thermal and carbon emission data through a multimodal feature fusion architecture.

[0024] In this embodiment, by synchronously collecting power power, time-of-use electricity prices, proportion of renewable energy output, and grid congestion signals, the grid status can be dynamically perceived, providing a millisecond-level response basis for computing power scheduling, collecting task priorities, data dependencies, and portability, and optimizing task allocation logic; real-time monitoring of inlet and outlet air temperatures, PUE values, and waste heat recovery efficiency, combined with regional power grid carbon emission factors and green electricity consumption ratios, achieves coordinated optimization of energy efficiency and carbon emissions. Through real-time, coupled quantification, and heterogeneous integration, it solves the shortcomings of the traditional data center's "power-computing-heat" split optimization and provides core scheduling capabilities for data centers.

[0025] Furthermore, the power load curve processes the spatial dependency and temporal dynamics of power data through the spatiotemporal graph convolutional network ST-GCN, and the computing task sequence processes high-dimensional computing task sequences through the Mamba-SSM state space model, modeling long-distance dependencies with linear complexity.

[0026] In this implementation, the fusion architecture of ST-GCN and Mamba-SSM, through the division of labor and collaboration of spatial topology modeling and long-sequence efficient learning, optimizes the power-computing-thermal coupling to drive data center energy efficiency improvement and carbon emission reduction.

[0027] Furthermore, the upper-level algorithm architecture adopts DRL, and the lower-level algorithm architecture adopts MIP. SoftActor-Critic is used in DRL to optimize the dynamic strategy. The real-time PUE value, available green electricity ratio, task queue length, cooling system COP, task migration, cooling pump frequency adjustment, and green electricity purchase volume interact with the environment. The data after each step of interaction with the environment is pushed into the experience pool to train the network. Finally, only the recommended action is output, but it is not directly executed by DRL. The upper-level algorithm architecture DRL serves as the initial solution of the lower-level algorithm architecture MIP. The lower-level algorithm architecture MIP uses the Gurobi solver to process mixed integer programming. The objective function is: Where, , , All are multi-objective weight coefficients; is the total electricity cost, in yuan; is the carbon emissions during the period, unit: kg; is the thermal constraint violation, unit ; Constraints include power balance and thermal constraints; Thermal constraints: Where, The cabinet inlet air temperature, unit: °C; Power balance: Where, The power purchased from the grid, in kW; The real-time photovoltaic power generation on site, in kW; is the battery charging and discharging power, in kW; is the total power of IT equipment, in kW; is the total power of the cooling system, in kW.

[0028] In this implementation, dynamic response is achieved to achieve real-time cost reduction and efficiency improvement. Based on real-time data, the DRL algorithm dynamically generates recommended actions for cooling frequency regulation and task migration, reducing electricity costs by 12-15%. During peak green power output, tasks are automatically migrated to green power-rich nodes, improving green power absorption rates and reducing carbon emission costs. The lower-level algorithm MIP strictly verifies the recommendations of the upper-level algorithm DRL to avoid equipment overload and downtime, ensure system safety, and simultaneously optimize PUE, carbon emissions, and task efficiency, resolving the contradiction between "carbon reduction" and "performance maintenance", realizing continuous self-evolution, and achieving long-term stable operation.

[0029] Furthermore, the Hyperledger Fabric private chain is used to check the power source, computing power task intensity, carbon emission intensity, and thermal system status. The power source is uploaded to the chain at a frequency of 15 minutes to verify that the green electricity certificate matches the real-time power generation. The computing power task intensity is uploaded to the chain at a frequency of 5 minutes to verify that the task resource request is consistent with the scheduling record. The carbon emission intensity is uploaded to the chain at a frequency of 60 minutes to call the IPCC coefficient calculation and cross-validation. The thermal system status is uploaded to the chain at a frequency of 1 minute, and the temperature exceeding the limit triggers the early warning contract.

[0030] In this implementation, green electricity certificates are verified to match actual power generation every 15 minutes to prevent false green electricity transactions and achieve real green electricity transaction volume. Task resource applications and actual scheduling records are verified every 5 minutes to prevent virtual occupation of computing resources and improve resource utilization. IPCC coefficients are automatically called every hour to calculate carbon emissions, and multi-institutional cross-verification is used to reduce accounting errors. Temperature is monitored every minute and early warnings are triggered to increase the speed of accident impact and reduce the risk of equipment downtime.

[0031] Furthermore, the power utilization efficiency of the data center is calculated in real time: Where, Real-time power utilization efficiency for data centers; is the total input active power of the data center, in kW; is the total active power of IT equipment, in kW, where Collect UPS, PDU, and chiller data via Modbus TCP.

[0032] In this embodiment, by calculating the power utilization efficiency of the data center, energy efficiency can be accurately monitored and optimized. Every 0.1 reduction in PUE can reduce operating costs, reduce ineffective carbon emissions, promote green electricity consumption and heat recovery. Low PUE data centers are more easily matched to green electricity fluctuations and improve the consumption rate. The waste heat recovery efficiency and PUE are linked to optimize, which can create secondary benefits.

[0033] Furthermore, when PUE>1.3, the cooling pump frequency conversion is triggered, the frequency is reduced by 10% of the current frequency, and the chilled water temperature set point is increased by 1°C each time; When PUE is less than 1.2, green electricity is consumed first and the battery is discharged to 20% SOC.

[0034] In this implementation, the energy consumption of the cooling system is directly reduced, thereby reducing the overall PUE, avoiding overcooling and equipment damage, and extending the life of the equipment. The charge state of the energy storage battery is discharged from a high value to 20%. Discharging the battery to 20% SOC can optimize the energy storage utilization rate and avoid deep discharge damage of the battery. At the same time, it provides peak-shaving services for the power grid, and uses stored green electricity to support IT loads. The consumption of green electricity directly reduces carbon emissions, contributing to the "dual carbon" goals. Based on dynamic adjustment of the PUE value, energy efficiency is improved and clean energy is maximized, forming a virtuous cycle of "energy saving-carbon reduction-compliance".

[0035] Furthermore, the green electricity premium = benchmark electricity price × (1-green electricity ratio) × carbon price coefficient; when there is surplus green electricity, the delay tolerance task is activated and compensation is 0.02~0.1 yuan / kWh.

[0036] In this implementation, the environmental value of green electricity is internalized into economic value through a premium mechanism. At the same time, the compensation mechanism can smooth out supply and demand fluctuations, significantly reduce users' system operating costs, accurately quantify green electricity premiums, improve transaction efficiency, reduce transaction friction costs, and enhance green electricity market liquidity. It can effectively promote green electricity consumption, reduce carbon emissions, reduce dependence on fossil energy, contribute to the "dual carbon" goals, delay tolerance task compensation mechanism, and improve the flexibility and coordinated efficiency of the energy system.

[0037] Furthermore, electricity and computing power interact: the grid dispatch system sends price signals and carbon signals, and the data center feeds back a list of tasks that can be reduced and tolerate delays; Computing power and thermal interaction: When the cabinet temperature exceeds the preset value, the scheduler prioritizes migrating tasks to the cooler cabinet. Interaction between heat and electricity: When the COP of the cooling system decreases, the load rate of IT equipment is reduced.

[0038] In this embodiment, the data center feeds back a list of reducible loads and a list of delay-tolerant tasks, so that the grid dispatcher can dynamically adjust the load based on real-time electricity price / carbon price signals. The carbon price signal drives the data center to prioritize running high-load tasks during low-carbon periods. When the cabinet temperature exceeds the safety threshold, the scheduler migrates the task to a low-temperature area. Task migration reduces local cooling demand. When the cooling system COP is lower than the threshold, the IT equipment load rate is automatically reduced. By coupling the computing power, electricity, and heat of the data center and controlling the variables of the three systems, it is possible to optimize and improve energy efficiency. By adjusting the delay-tolerant tasks, the data center can improve the matching degree between the load curve and the wind and solar output.

[0039] See also Figure 3 , the present invention also provides an information interaction system for coordinated scheduling of computing power and electricity in a data center, including FPGA 1, a multimodal acquisition module 2, a collaborative optimization module 3, a blockchain scheduling module 4 and an energy efficiency management module 5; The multimodal acquisition module 2 is used to collect computing power, electricity, and thermal parameters, and to construct a spatiotemporal correlation model; The collaborative optimization module 3 is used to simultaneously optimize the computing power cost, carbon emission data and thermal data of the data center through DRL and MIP; The blockchain scheduling module 4 is used to realize the trustworthiness of cross-domain scheduling of power sources, computing task intensity, carbon emission intensity, and thermal system status; The energy efficiency management module 5 is used to dynamically adjust energy efficiency and encourage green electricity consumption; The FPGA 1 is used to process real-time data plane tasks, realize the instant update of the scheduling algorithm, and is used to connect and integrate the multimodal acquisition module 2, the collaborative optimization module 3, the blockchain scheduling module 4 and the energy efficiency management module 5.

[0040] In this embodiment, the FPGA1 uses an on-chip temperature sensor to collect cabinet / chip temperature in real time, predict heat dissipation requirements, and dynamically calibrate the cooling strategy. It uses PCIe SR-IOV technology to split FPGA1 into virtual functions, monitor the computing load and delay tolerance of each operator in a fine-grained manner, directly analyze the carbon intensity signal and real-time electricity price of the power grid, and implement signal decoding and priority sorting through hardware logic circuits. FPGA1 directly connects sensors and actuators through logic circuits, eliminating operating system scheduling overhead and achieving 5μs-level closed-loop control to meet high-priority task requirements. Through the "hardware reconfigurable × real-time control × energy efficiency closed-loop" three-in-one architecture, FPGA1 becomes the core engine for the coordinated scheduling of computing power, electricity, and heat. Through deep coupling of multiple modules and hardware acceleration innovation, a closed-loop system of "perception-decision-scheduling-execution-verification" is constructed, achieving breakthrough improvements in economy, environmental protection, reliability, and response speed.

[0041] The above disclosure is only a preferred embodiment of the present invention, and certainly cannot be used to limit the scope of the rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiment and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for information interaction between data center computing power and power coordinated scheduling, characterized in that: The steps include: Through multimodal collection, calculation and integration of data center power load curves, computing task types, thermal system status, and carbon emission intensity data, a spatiotemporal correlation model is constructed. A multi-objective optimization algorithm (MIP) combining improved deep reinforcement learning (DRL) and mixed integer programming (MIP) was used to simultaneously optimize the computing power cost, carbon emission data, and thermal data of the data center. Use blockchain distributed ledgers to record data centers’ power sources, computing task intensity, carbon emission intensity, and thermal system status data; Data centers dynamically adjust energy efficiency based on PUE and encourage green electricity consumption; Through data interaction, the coordinated scheduling and coupling optimization of the data center's computing power, electricity, and heat networks can be achieved.

2. The information interaction method for coordinated scheduling of computing power and electricity in a data center according to claim 1, wherein the multimodal calculation and integration of the data center's power load curve, computing power task type, thermal system status, and carbon emission intensity data, and the construction of a spatiotemporal correlation model, is characterized in that: The steps include: Collect data center real-time power consumption, time-of-use electricity prices, renewable energy output ratio, and grid congestion signals; Collect task priorities, computational accuracy, data dependencies, and portability of data centers; Collect data center inlet and outlet air temperature, cooling water flow, PUE value, meteorological data, and waste heat recovery efficiency; Collect data on the data center's regional power grid carbon emission factor, green electricity consumption ratio, IT equipment carbon footprint, and indirect emissions from cooling systems; Unify the above asynchronous data streams to the same time granularity and use a sliding window mechanism to process delayed data; Design a dynamic spatiotemporal correlation matrix to quantify the coupling strength between nodes; The power load curve, computing task sequence, thermal and carbon emission data are integrated through a multimodal feature fusion architecture.

3. The information interaction method for coordinated scheduling of computing power and electricity in a data center according to claim 2, characterized in that: The power load curve processes the spatial dependency and temporal dynamics of power data through the spatiotemporal graph convolutional network ST-GCN, and the computing task sequence processes high-dimensional computing task sequences through the Mamba-SSM state space model, modeling long-distance dependencies with linear complexity.

4. The information interaction method for coordinated scheduling of computing power and electricity in a data center according to claim 1, wherein the multi-objective optimization algorithm MIP combining improved deep reinforcement learning (DRL) and mixed integer programming is used to simultaneously optimize the computing power cost, carbon emission data, and thermal data of the data center, characterized in that: The upper-level algorithm architecture uses DRL, and the lower-level algorithm architecture uses MIP. Soft Actor-Critic is used in DRL to optimize dynamic strategies. The real-time PUE value, available green electricity ratio, task queue length, cooling system COP, task migration, cooling pump frequency adjustment, and green electricity purchase volume interact with the environment. The data after each step of interaction with the environment is pushed into the experience pool to train the network. Finally, only recommended actions are output, but they are not directly executed by DRL. The upper-level algorithm architecture DRL serves as the initial solution for the lower-level algorithm architecture MIP, which processes mixed integer programming through the Gurobi solver. The objective function is: Where, , , All are multi-objective weight coefficients; is the total electricity cost, in yuan; is the carbon emissions during the period, unit: kg; is the thermal constraint violation, unit ; Constraints include power balance and thermal constraints; Thermal constraints: Where, The cabinet inlet air temperature, unit: °C; Power balance: Where, The power purchased from the grid, in kW; The real-time photovoltaic power generation on site, in kW; is the battery charging and discharging power, in kW; is the total power of IT equipment, in kW; is the total power of the cooling system, in kW.

5. The method for information interaction for coordinated scheduling of computing power and electricity in a data center according to claim 1, wherein the blockchain distributed ledger is used to record the power source, computing power task intensity, carbon emission intensity, and thermal system status data of the data center, characterized in that: Through the Hyperledger Fabric private chain, the power source, computing power task intensity, carbon emission intensity, and thermal system status are recorded. The power source is uploaded to the chain every 15 minutes to verify that the green electricity certificate matches the real-time power generation. The computing power task intensity is uploaded to the chain every 5 minutes to verify that the task resource request is consistent with the scheduling record. The carbon emission intensity is uploaded to the chain every 60 minutes to call the IPCC coefficient calculation and cross-validation. The thermal system status is uploaded to the chain every 1 minute, and temperature exceeding the limit triggers the early warning contract.

6. The method for information interaction for coordinated scheduling of computing power and electricity in a data center according to claim 1, wherein the data center dynamically adjusts energy efficiency according to PUE and encourages green electricity consumption, characterized in that: Real-time calculation of data center power utilization efficiency: Where, Real-time power utilization efficiency for data centers; is the total input active power of the data center, in kW; is the total active power of IT equipment, in kW, where Collect UPS, PDU, and chiller data via Modbus TCP.

7. The method for information interaction for coordinated scheduling of computing power and electricity in a data center according to claim 6, characterized in that: When PUE>1.3, the cooling pump frequency conversion is triggered, the current frequency is reduced by 10%, and the chilled water temperature set value is increased by 1°C each time; When PUE is less than 1.2, green electricity is consumed first and the battery is discharged to 20% SOC.

8. The method for information interaction for coordinated scheduling of computing power and electricity in a data center according to claim 6, characterized in that: Green electricity premium = benchmark electricity price × (1-green electricity ratio) × carbon price coefficient; when green electricity is in surplus, the delay tolerance task is activated and compensation is 0.02~0.1 yuan / kWh.

9. The method for information interaction for coordinated scheduling of computing power and electricity in a data center according to claim 1, wherein the coordinated scheduling and coupling optimization of computing power, electricity, and heat networks in the data center are achieved through data interaction, characterized in that: Interaction between electricity and computing power: The grid dispatch system sends price and carbon signals, and the data center provides feedback on load reduction and delay-tolerant task lists. Computing power and thermal interaction: When the cabinet temperature exceeds the preset value, the scheduler prioritizes migrating tasks to the cooler cabinet. Interaction between heat and electricity: When the COP of the cooling system decreases, the load rate of IT equipment is reduced.

10. An information interaction system for coordinated scheduling of computing power and electricity in a data center, characterized in that: Including FPGA, multimodal acquisition module, collaborative optimization module, blockchain scheduling module and energy efficiency management module; The multimodal acquisition module is used to collect computing power, electricity, and thermal parameters and build a spatiotemporal correlation model; The collaborative optimization module is used to simultaneously optimize the computing power cost, carbon emission data and thermal data of the data center through DRL and MIP; The blockchain scheduling module is used to realize the trustworthiness of cross-domain scheduling of power sources, computing task intensity, carbon emission intensity, and thermal system status; The energy efficiency management module is used to dynamically adjust energy efficiency and encourage green electricity consumption; The FPGA is used to process real-time data plane tasks, realize the instant update of the scheduling algorithm, and is used to connect and integrate the multimodal acquisition module, the collaborative optimization module, the blockchain scheduling module and the energy efficiency management module.

Images