Energy consumption and carbon emission control method and system for supercomputing center based on artificial intelligence

Abstract

The present invention provides a method and system for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence, which relates to the field of information technology. First, the first carbon emissions of a computing node are calculated based on the first energy consumption and the second energy consumption; then, the second carbon emissions of the computing node are obtained according to the total carbon emissions of the supercomputing center and the number of computing nodes; then, the first target computing node and the second target computing node are screened out based on the first carbon emissions and the second carbon emissions, and the target computing task that needs to be scheduled in the first target computing node is determined through the computing task scheduling model; finally, the target computing task is scheduled to the second target computing node for processing. In the above scheme, by monitoring the carbon emissions of each computing node, the computing tasks in the computing nodes with larger carbon emissions are scheduled to the computing nodes with smaller emissions for processing, and the overall energy consumption and carbon emissions of the supercomputing center can be reduced while ensuring the computing power requirements.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to an artificial intelligence-based supercomputing center energy consumption and carbon emission control method and system. Background Art

[0002] Supercomputing centers are the cornerstone of modern computing science and are widely used in fields such as climate simulation, genome research, and disaster prediction. As an important infrastructure for high-performance computing, the energy consumption of supercomputing centers cannot be ignored. How to reduce the energy consumption and carbon emissions of supercomputing centers while ensuring computing power has become a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0003] In order to at least overcome the above-mentioned deficiencies in the prior art, in a first aspect, the present invention provides a method and system for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence, which is applied to a server for monitoring and controlling the supercomputing center, wherein the supercomputing center includes a plurality of computing nodes that communicate with each other, and the method includes:

[0004] Obtaining a current first energy consumption of each of the computing nodes;

[0005] Based on the surrounding supporting energy consumption, a second energy consumption required for each computing node to operate normally is obtained, and according to the first energy consumption and the second energy consumption of each computing node, a first carbon emission of each computing node is calculated;

[0006] Calculate the total energy consumption of the computing equipment of the supercomputing center based on the first energy consumption of each computing node, and calculate the total carbon emissions of the supercomputing center based on the total energy consumption of the computing equipment and the energy consumption of the surrounding supporting equipment;

[0007] Based on the total carbon emissions and the number of computing nodes in the supercomputing center, calculate a second carbon emission amount of the computing node;

[0008] Compare the first carbon emission with the second carbon emission, and use a computing node whose first carbon emission is higher than the second carbon emission as a first target computing node;

[0009] Obtaining a computing resource occupancy rate of a computing node whose first carbon emission rate is lower than that of the second carbon emission rate, and taking a computing node whose computing resource occupancy rate is lower than a preset computing resource occupancy rate as a second target computing node;

[0010] The amount of computing tasks being processed in the first target computing node, the computing resource occupancy rate, and the first carbon emissions of the first target computing node, as well as the amount of computing tasks being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emissions of the second target computing node are input into a computing task scheduling model to determine the target computing tasks that need to be scheduled from the first target computing node to the second target computing node for processing, wherein the sum of the first carbon emissions of the first target computing node and the second target computing node before scheduling is less than the sum of the first carbon emissions of the first target computing node and the second target computing node after scheduling;

[0011] The target computing task in the first target computing node is dispatched to the second target computing node for processing.

[0012] In a possible implementation, the peripheral supporting energy consumption includes cooling supporting energy consumption and power supply supporting energy consumption, and the step of obtaining the second energy consumption required to enable each computing node to work normally based on the peripheral supporting energy consumption, and calculating the first carbon emissions of each computing node according to the first energy consumption and the second energy consumption of each computing node includes:

[0013] Calculate the cooling energy consumption required for each computing node based on the type of cooling equipment;

[0014] Based on the power supply supporting energy consumption and the current first energy consumption of each computing node, the power supply energy consumption required by each computing node is calculated;

[0015] Obtaining the second energy consumption required to enable each computing node to work normally from the cooling energy consumption and the power supply energy consumption required by each computing node;

[0016] Based on the energy consumption source type of the first energy consumption and the second energy consumption of each computing node, a first carbon emission of each computing node is calculated.

[0017] In a possible implementation, the step of calculating the cooling energy consumption required for each computing node based on the type of cooling support includes:

[0018] When the type of cooling equipment is air cooling, based on the total power consumption of the cooling equipment and the number of the computing nodes, the air cooling energy consumption required for each computing node is calculated, and the air cooling energy consumption is used as the cooling energy consumption required for the computing node; or

[0019] When the type of cooling support is liquid cooling, the liquid cooling power consumption of the liquid cooling device configured for each computing node is used as the cooling energy consumption required for each computing node; or,

[0020] When the types of cooling support include air cooling and liquid cooling, the sum of the air cooling energy consumption required for each computing node and the liquid cooling power consumption of the liquid cooling device configured for the computing node is used as the cooling energy consumption required for each computing node.

[0021] In a possible implementation, the step of calculating the total energy consumption of computing equipment of the supercomputing center based on the first energy consumption of each computing node, and calculating the total carbon emissions of the supercomputing center based on the total energy consumption of computing equipment and the surrounding supporting energy consumption includes:

[0022] Adding the first energy consumption of all computing nodes to obtain the total energy consumption of computing equipment in the supercomputing center;

[0023] The total carbon emissions of the supercomputing center are calculated based on the power source type of the total energy consumption of the computing equipment of the supercomputing center and the power source type of the surrounding supporting energy consumption;

[0024] The step of calculating the second carbon emissions of the computing nodes based on the total carbon emissions and the number of computing nodes in the supercomputing center includes:

[0025] The total carbon emissions are divided by the number of computing nodes in the supercomputing center to obtain an average carbon emissions of each computing node, and the average carbon emissions are used as the second carbon emissions of the computing node.

[0026] In a possible implementation, before inputting the amount of computing tasks being processed in the first target computing node, the computing resource occupancy rate, the first carbon emission of the first target computing node, and the amount of computing tasks being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emission of the second target computing node into the computing task scheduling model, and determining the target computing tasks that need to be scheduled from the first target computing node to the second target computing node for processing, the method further includes the step of training the computing task scheduling model, which includes:

[0027] Create a training sample set, the training sample set including multiple training samples and sample tags corresponding to the training samples, wherein the training samples include a first computing task amount of a first computing node sample, a first computing resource occupancy rate of the first computing node sample, a first carbon emission of the first computing node sample, a second computing task amount of a second computing node sample, a first computing resource occupancy rate of the second computing node sample, and a first carbon emission of the second computing node sample, and the sample tags include a computing task amount of the first computing node sample that needs to be scheduled to the second computing node sample, a first carbon emission of the first computing node sample after scheduling, and a first carbon emission of the second computing node sample after scheduling;

[0028] Input the training sample into a deep learning network for training, and output a training result, wherein the training result includes a first training carbon emission of the first computing node sample, a second training carbon emission of the second computing node sample, and an amount of training computing tasks scheduled from the first computing node sample to the second computing node sample;

[0029] The loss function value of the deep learning network is calculated according to the training results and the sample labels. When the loss function value is greater than a preset loss function value, the network parameters in the deep learning network are adjusted, and the training process is repeated until the loss function value is no greater than the preset loss function value. The training is terminated, and the deep learning network obtained through training is used as the computing task scheduling model.

[0030] In a possible implementation, the method further includes:

[0031] Monitor the space temperature of the space where the supercomputing center is located, the ambient temperature outside the space where the supercomputing center is located, and the node operating temperature of each computing node in the supercomputing center;

[0032] When the temperature difference between the space temperature and the ambient temperature is greater than a first preset temperature difference, controlling the fan to transfer air outside the space where the supercomputing center is located to the space where the supercomputing center is located;

[0033] When the temperature difference between the space temperature and the ambient temperature is less than or equal to a first preset temperature difference, controlling the air cooling equipment in the space where the supercomputing center is located to operate;

[0034] When the temperature difference between the node operating temperature and the space temperature is greater than a second preset temperature difference, the liquid cooling device configured for the corresponding computing node is controlled to operate.

[0035] In a possible implementation, the method further includes:

[0036] Arrange a solar power generation device outside the space where the supercomputing center is located;

[0037] When the fan or the air cooling device needs to work, the solar power generation device provides electric energy to the fan or the air cooling device.

[0038] In a second aspect, the present invention further provides an artificial intelligence-based supercomputing center energy consumption and carbon emission control system, which is applied to a server for monitoring and controlling the supercomputing center, wherein the supercomputing center includes a plurality of computing nodes that communicate with each other, and the system includes:

[0039] An acquisition module, used for acquiring the current first energy consumption of each of the computing nodes;

[0040] A first calculation module is used to obtain the second energy consumption required for each computing node to work normally based on the surrounding supporting energy consumption, and calculate the first carbon emission of each computing node according to the first energy consumption and the second energy consumption of each computing node;

[0041] A second calculation module is used to calculate the total energy consumption of the computing equipment of the supercomputing center based on the first energy consumption of each computing node, and calculate the total carbon emissions of the supercomputing center based on the total energy consumption of the computing equipment and the energy consumption of the surrounding supporting equipment;

[0042] A third calculation module, configured to calculate a second carbon emission amount of the computing node based on the total carbon emission amount and the number of computing nodes in the supercomputing center;

[0043] A first target computing node confirmation module, configured to compare the first carbon emission with the second carbon emission, and to use a computing node whose first carbon emission is higher than the second carbon emission as a first target computing node;

[0044] A second target computing node confirmation module is used to obtain the computing resource occupancy rate of the computing node whose first carbon emission is lower than the second carbon emission, and use the computing node whose computing resource occupancy rate is lower than the preset computing resource occupancy rate as the second target computing node;

[0045] A computing task quantity confirmation module is used to input the computing task quantity being processed in the first target computing node, the computing resource occupancy rate, the first carbon emission of the first target computing node, and the computing task quantity being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emission of the second target computing node into a computing task scheduling model, and determine the target computing task that needs to be scheduled from the first target computing node to the second target computing node for processing, wherein the sum of the first carbon emission of the first target computing node and the second target computing node before scheduling is less than the sum of the first carbon emission of the first target computing node and the second target computing node after scheduling;

[0046] The scheduling module is used to schedule the target computing task in the first target computing node to the second target computing node for processing.

[0047] In a third aspect, the present invention also provides a server, comprising a processor, a computer-readable storage medium and a communication interface, wherein the computer-readable storage medium, the communication interface and the processor are connected via a bus system, the computer-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the computer-readable storage medium to execute the artificial intelligence-based supercomputing center energy consumption and carbon emission control method described in any possible implementation method of the first aspect.

[0048] In a fourth aspect, the present invention further provides a computer-readable storage medium, which stores instructions that, when executed, enable a server to execute an artificial intelligence-based supercomputing center energy consumption and carbon emission control method in any possible implementation of the first aspect.

[0049] In the above scheme provided by the present invention, first, the first energy consumption and the second energy consumption of each computing node are obtained, and the first carbon emission of the computing node is calculated based on the first energy consumption and the second energy consumption; then, the total carbon emission of the supercomputing center is calculated according to the total energy consumption of the computing equipment in the supercomputing center and the energy consumption of the surrounding supporting equipment, and the second carbon emission of the computing node is obtained based on the total carbon emission of the supercomputing center and the number of computing nodes; then, the first target computing node and the second target computing node that need to adjust the computing task are screened out based on the first carbon emission and the second carbon emission, and the target computing task that needs to be scheduled in the first target computing node is determined through the computing task scheduling model; finally, the target computing task is scheduled to the second target computing node for processing. In the above scheme, by monitoring the carbon emissions of each computing node, the computing tasks in the computing nodes with larger carbon emissions are scheduled to the computing nodes with smaller emissions for processing. Under the condition of ensuring the computing power demand, the overall energy consumption and carbon emissions of the supercomputing center can be reduced, and the impact of the supercomputing center on the environment can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0051] Figure 1 A flow chart of a method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence provided in an embodiment of the present application;

[0052] Figure 2 To achieve Figure 1 Schematic diagram of the process of step S120;

[0053] Figure 3 A flow chart of implementing a training computing task scheduling model is provided for the embodiment of the present application;

[0054] Figure 4 A schematic diagram of the functional modules of the artificial intelligence-based supercomputing center energy consumption and carbon emission control system provided in an embodiment of the present invention;

[0055] Figure 5 A schematic diagram of the structural framework of a server provided in an embodiment of the present invention for implementing the above-mentioned artificial intelligence-based supercomputing center energy consumption and carbon emission control method. DETAILED DESCRIPTION

[0056] The present invention is described in detail below with reference to the accompanying drawings. The specific operation method in the method embodiment can also be applied to the system embodiment or system embodiment.

[0057] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application are only for the purpose of illustration and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate the operations implemented according to some embodiments of the embodiments of the present application.

[0058] It should be understood that the operations of the flow chart may not be implemented in order, and steps without logical contextual relationships may be reversed in order or implemented simultaneously. In addition, those skilled in the art may add one or more other operations to the flow chart, or may remove one or more operations from the flow chart under the guidance of the content of this application.

[0059] In addition, the described embodiments are only some embodiments of the present application, rather than all embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0060] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0061] See also Figure 1The artificial intelligence-based supercomputing center energy consumption and carbon emission control method provided in the embodiment of the present application can be executed by a server used to monitor and control the supercomputing center. In other embodiments, the order of some steps in the artificial intelligence-based supercomputing center energy consumption and carbon emission control method in the embodiment of the present application can be exchanged with each other according to actual needs, or some of the steps can be omitted or deleted. The detailed steps of the artificial intelligence-based supercomputing center energy consumption and carbon emission control method executed by the server are introduced as follows.

[0062] Step S110, obtaining the current first energy consumption of each computing node.

[0063] In this embodiment, each computing node can independently complete computing tasks. For example, the computing node may include multiple processors (CPUs) or a combination of multiple processors and multiple graphics processing units (GPUs). The first energy consumption is the energy consumption required for the computing node to work normally, that is, the first energy consumption is the energy consumption for the processor or the processor and the graphics processing unit in the computing node to work. For example, the first energy consumption may be the energy consumption of the computing node at a time granularity, wherein the time granularity may be 10 minutes, 30 minutes, or 1 hour, etc.

[0064] Step S120, obtaining the second energy consumption required for each computing node to work normally based on the surrounding supporting energy consumption, and calculating the first carbon emission of each computing node according to the first energy consumption and the second energy consumption of each computing node.

[0065] Peripheral supporting energy consumption is the energy consumption to ensure the normal operation of computing nodes. For example, peripheral supporting energy consumption includes cooling supporting energy consumption and power supply supporting energy consumption. Among them, cooling supporting energy consumption is the energy consumption generated by providing cooling for computing nodes in the supercomputing center, and power supply supporting energy consumption is the energy consumption generated by providing stable power supply for computing nodes in the supercomputing center. The peripheral supporting energy consumption here is the peripheral supporting energy consumption at the same time granularity as the first energy consumption.

[0066] Among them, the second energy consumption is the surrounding supporting energy consumption allocated to each computing node. The total energy consumption of each computing node can be calculated based on the first energy consumption and the second energy consumption of each computing node, and the first carbon emissions of the computing node can be calculated based on the total energy consumption of each computing node. Among them, the first carbon emissions are the actual carbon emissions of the computing nodes in the supercomputing center.

[0067] Step S130, calculating the total energy consumption of the computing equipment of the supercomputing center based on the first energy consumption of each computing node, and obtaining the total carbon emissions of the supercomputing center based on the total energy consumption of the computing equipment and the energy consumption of surrounding supporting equipment.

[0068] The total energy consumption of the supercomputing center is obtained based on the total energy consumption of computing equipment and the energy consumption of surrounding supporting equipment, and the total carbon emissions of the supercomputing center are obtained based on the total energy consumption of the supercomputing center.

[0069] Step S140, based on the total carbon emissions and the number of computing nodes in the supercomputing center, calculate the second carbon emissions of the computing nodes.

[0070] Among them, the second carbon emission is the average carbon emission of computing nodes in the supercomputing center.

[0071] Step S150: compare the first carbon emission with the second carbon emission, and use the computing node whose first carbon emission is higher than the second carbon emission as the first target computing node.

[0072] By comparing the first carbon emission amount with the second carbon emission amount, a computing node with a larger actual carbon emission amount is selected as the first target computing node.

[0073] Step S160, obtaining the computing resource occupancy rate of the computing node whose first carbon emission is lower than the second carbon emission rate, and taking the computing node whose computing resource occupancy rate is lower than the preset computing resource occupancy rate as the second target computing node.

[0074] The computing nodes with smaller actual carbon emissions and lower computing resource occupancy are selected as the second target computing nodes.

[0075] Step S170, input the amount of computing tasks being processed in the first target computing node, the computing resource occupancy rate, the first carbon emissions of the first target computing node, and the amount of computing tasks being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emissions of the second target computing node into the computing task scheduling model, and determine the target computing tasks that need to be scheduled from the first target computing node to the second target computing node for processing.

[0076] In this step, the computing task scheduling model can determine the target computing task to be scheduled from the first target computing node to the second target computing node. For example, the computing resources occupied by the target computing task on the second target computing node do not exceed 50% of the idle computing resources in the second target computing node, so as to avoid the second target computing node from heating up rapidly after too many computing tasks are transferred to the second target computing node. With such a setting, the sum of the first carbon emissions of the first target computing node and the second target computing node before scheduling can be less than the sum of the first carbon emissions of the first target computing node and the second target computing node after scheduling.

[0077] Step S180: dispatching the target computing task in the first target computing node to the second target computing node for processing.

[0078] In an embodiment, due to the different hardware aging conditions of different computing nodes, when running the same number of computing tasks, the carbon emissions generated by different computing nodes may also be different. By monitoring the carbon emissions of each computing node, the computing tasks in the computing nodes with larger carbon emissions are scheduled to be processed in the computing nodes with smaller emissions. While ensuring the computing power requirements, the overall energy consumption and carbon emissions of the supercomputing center can be reduced, thereby reducing the impact of the supercomputing center on the environment.

[0079] For further information, please refer to Figure 2 , step S120 can be implemented through the following sub-steps.

[0080] Sub-step S1201, calculating the cooling energy consumption required for each computing node based on the type of cooling equipment.

[0081] Among them, the types of cooling facilities may include air cooling and liquid cooling. Air cooling refers to cooling the supercomputing center through air conditioning, and liquid cooling refers to cooling the computing nodes through liquid cooling devices configured on each computing node. It can be understood that the same supercomputing center can have both air cooling and liquid cooling facilities. Air cooling is an overall cooling of the environment in which the supercomputing center is located, and liquid cooling can be individually controlled and cooled for different computing nodes.

[0082] In this embodiment, sub-step S1201 can be implemented in three ways.

[0083] In the first implementation method, when the cooling system is air-cooled, the air-cooled energy consumption required for each computing node is calculated based on the total power consumption of the cooling equipment and the number of computing nodes, and the air-cooled energy consumption is used as the cooling energy consumption required for the computing node.

[0084] The second implementation method is that when the type of cooling support is liquid cooling, the liquid cooling power consumption of the liquid cooling device configured for each computing node is used as the cooling energy consumption required for each computing node.

[0085] A third implementation method is that when the types of cooling support include air cooling and liquid cooling, the sum of the air cooling energy consumption required for each computing node and the liquid cooling power consumption of the liquid cooling device configured for the computing node is used as the cooling energy consumption required for each computing node.

[0086] Sub-step S1202, based on the power supply supporting energy consumption and the current first energy consumption of each computing node, calculate the power supply energy consumption required by each computing node.

[0087] Exemplarily, the total energy consumption of the computing equipment of the supercomputing center can be obtained based on the current first energy consumption of each computing node, the energy consumption ratio of the computing node can be obtained based on the ratio between the current first energy consumption of the computing node and the total energy consumption of the computing equipment, and then the power consumption required by the computing node can be obtained from the energy consumption ratio of the computing node and the total energy consumption of the computing equipment of the supercomputing center.

[0088] Sub-step S1203, obtaining the second energy consumption required to enable each computing node to work normally from the cooling energy consumption and the power supply energy consumption required by each computing node.

[0089] Exemplarily, the cooling energy consumption and the power supply energy consumption required by the computing node are added together to obtain the second energy consumption required for the computing node to operate normally.

[0090] Sub-step S1204, based on the energy consumption source type of the first energy consumption and the second energy consumption of each computing node, calculate and obtain the first carbon emission of each computing node.

[0091] Exemplarily, the energy consumption source types include electricity obtained in different ways, such as thermal power generation, hydropower generation, solar power generation, wind power generation, and nuclear power generation. Among them, power generation will produce carbon emissions. For example, when thermal power generation is used to supply power to the supercomputing center, a carbon emission value will be generated, wherein the carbon emission value corresponds to the carbon emission value required to produce the corresponding consumed electricity. When hydropower generation, solar power generation, wind power generation, and nuclear power are used to supply power to the supercomputing center, the carbon emission value can be considered to be small or even 0. At present, most of the power supply in the supercomputing center is still supplied by thermal power generation.

[0092] In this embodiment, step S130 can be implemented in the following manner.

[0093] First, the first energy consumption of all computing nodes is added up to get the total energy consumption of computing equipment in the supercomputing center;

[0094] Next, the total carbon emissions of the supercomputing center are calculated based on the electricity source types of the total energy consumption of the supercomputing center's computing equipment and the electricity source types of the surrounding supporting energy consumption.

[0095] The method of calculating the first carbon emissions is the same as that of the above sub-step S1204. The total carbon emissions of the supercomputing center can be calculated based on the power source type of the total energy consumption of the computing equipment and the power source type of the surrounding supporting energy consumption.

[0096] Further, step S140 can be implemented in the following manner.

[0097] Divide the total carbon emissions by the number of computing nodes in the supercomputing center to obtain the average carbon emissions of each computing node, and use the average carbon emissions as the second carbon emissions of the computing node.

[0098] Furthermore, before step S170, the method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence provided in this embodiment also includes the step of training a computing task scheduling model, please refer to Figure 3 , this step can be achieved in the following way.

[0099] Step S201, creating a training sample set.

[0100] The training sample set includes multiple training samples and sample tags corresponding to the training samples, wherein the training samples and the sample tags are historical data of computing nodes with the same hardware configuration. Exemplarily, the training samples include a first computing task amount of a first computing node sample, a first computing resource occupancy rate of the first computing node sample, a first carbon emission of the first computing node sample, a second computing task amount of a second computing node sample, a first computing resource occupancy rate of the second computing node sample, and a first carbon emission of the second computing node sample, and the sample tag includes a computing task amount of the first computing node sample that needs to be scheduled to the second computing node sample, a first carbon emission of the first computing node sample after scheduling, and a first carbon emission of the second computing node sample after scheduling.

[0101] Step S202: input the training samples into a deep learning network for training, and output the training results.

[0102] The training result includes a first training carbon emission of the first computing node sample, a second training carbon emission of the second computing node sample, and the amount of training computing tasks scheduled from the first computing node sample to the second computing node sample.

[0103] In this embodiment, the deep learning network may include a feature extraction subnetwork, a feature processing subnetwork and a prediction subnetwork. First, the training samples are input into the feature extraction subnetwork for feature extraction to obtain extracted features, wherein the feature extraction includes feature extraction and abstraction of the training samples; then, the extracted features are input into the feature processing subnetwork for feature processing, wherein the feature processing method includes feature fusion to form a feature vector of fixed length; then, the fused feature vector is predicted by the prediction subnetwork to obtain the training result.

[0104] In this embodiment, during the model training phase, the training results may include the first training carbon emissions of the first computing node sample, the second training carbon emissions of the second computing node sample, and the amount of training computing tasks scheduled from the first computing node sample to the second computing node sample. After the computing task scheduling model is obtained after the model training is completed, the computing task scheduling model may only output the amount of computing tasks that need to be scheduled.

[0105] Step S203, calculate the loss function value of the deep learning network according to the training results and sample labels. When the loss function value is greater than the preset loss function value, adjust the network parameters in the deep learning network, and repeat the above training process until the loss function value is no greater than the preset loss function value. The training is completed and the trained deep learning network is used as a computing task scheduling model.

[0106] In this embodiment, the deep learning network may also include a feedback subnetwork, which calculates the loss function value of the deep learning network based on the training results output by the prediction subnetwork and the sample labels of the corresponding training samples. Exemplarily, the feedback subnetwork calculates the loss function value based on the first training carbon emissions of the first computing node sample, the second training carbon emissions of the second computing node sample, the amount of training computing tasks scheduled from the first computing node sample to the second computing node sample, the amount of computing tasks that the first computing node sample needs to be scheduled to the second computing node sample, the first carbon emissions of the first computing node sample after scheduling, and the first carbon emissions of the second computing node sample after scheduling.

[0107] When the loss function value is greater than the preset loss function value, adjust the network parameters of the feature extraction subnetwork, feature processing subnetwork and prediction subnetwork in the deep learning network, and repeat the above training process until the loss function value is no greater than the preset loss function value. The training ends and the trained deep learning network is used as a computing task scheduling model.

[0108] Furthermore, the artificial intelligence-based supercomputing center energy consumption and carbon emission control method provided in this embodiment also includes the following steps.

[0109] Monitor the space temperature of the space where the supercomputing center is located, the ambient temperature outside the space where the supercomputing center is located, and the node operating temperature of each computing node in the supercomputing center.

[0110] Specifically, the temperature of the space can be monitored by a temperature sensor installed in the space where the supercomputing center is located, the ambient temperature can be monitored by a temperature sensor installed outside the space where the supercomputing center is located, and the node operating temperature of each computing node can be monitored by a temperature sensor configured in each computing node. These temperature sensors send the monitored temperatures to the server.

[0111] When the temperature difference between the space temperature and the ambient temperature is greater than a first preset temperature difference, the fan is controlled to transfer air outside the space where the supercomputing center is located to the space where the supercomputing center is located.

[0112] When the temperature difference between the space temperature and the ambient temperature is less than or equal to the first preset temperature difference, the air cooling equipment in the space where the supercomputing center is located is controlled to operate.

[0113] When the temperature difference between the node operating temperature and the space temperature is greater than a second preset temperature difference, the liquid cooling device configured for the corresponding computing node is controlled to operate.

[0114] In this embodiment, among the three different cooling modes, the fan cooling mode requires the lowest energy consumption and its corresponding carbon emissions are also the lowest; the air cooling mode requires the second highest energy consumption and its corresponding carbon emissions are also the second highest; the liquid cooling mode requires the highest energy consumption and its corresponding carbon emissions are also the highest. Through the above design, the power consumption and overall carbon emissions of the supercomputing center can be minimized while ensuring the computing power of the supercomputing center.

[0115] In this embodiment, one of the fan cooling mode and the air cooling mode can be operated alone, or one of the fan cooling mode and the air cooling mode can be operated together with the liquid cooling mode. Different computing nodes can selectively turn on the liquid cooling mode according to their own heat dissipation conditions. That is, in the supercomputing center, the liquid cooling mode can be selectively turned on or off for a single computing node.

[0116] Furthermore, the artificial intelligence-based supercomputing center energy consumption and carbon emission control method provided in this embodiment may also include the following steps.

[0117] A solar power generation device is configured outside the space where the supercomputing center is located.

[0118] When the fan or air cooling device needs to work, the solar power generation device provides electric energy for the fan or air cooling device.

[0119] Since the power consumption of fans and air-cooling equipment is relatively low, the power supply problem of the two can be at least partially solved by configuring solar power generation equipment outside the space where the supercomputing center is located, thus reducing the carbon emissions of the entire supercomputing center.

[0120] Based on the same inventive concept, this embodiment also provides an artificial intelligence-based supercomputing center energy consumption and carbon emission control system, please refer to Figure 4 , Figure 4The functional module diagram of the supercomputing center energy consumption and carbon emission control system 100 based on artificial intelligence provided in this embodiment, this embodiment can divide the functional modules of the supercomputing center energy consumption and carbon emission control system 100 based on artificial intelligence according to the above method embodiment, that is, the following functional modules corresponding to the supercomputing center energy consumption and carbon emission control system 100 based on artificial intelligence can be used to execute the above method embodiments. Among them, the supercomputing center energy consumption and carbon emission control system 100 based on artificial intelligence can include an acquisition module 110, a first calculation module 120, a second calculation module 130, a third calculation module 140, a first target computing node confirmation module 150, a second target computing node confirmation module 160, a computing task volume confirmation module 170 and a scheduling module 180. The functions of each functional module of the supercomputing center energy consumption and carbon emission control system 100 based on artificial intelligence are described in detail below.

[0121] The acquisition module 110 is used to acquire the current first energy consumption of each computing node.

[0122] In this embodiment, each computing node can independently complete computing tasks. For example, the computing node may include multiple processors (CPUs) or a combination of multiple processors and multiple graphics processing units (GPUs). The first energy consumption is the energy consumption required for the computing node to work normally, that is, the first energy consumption is the energy consumption for the processor or the processor and the graphics processing unit in the computing node to work.

[0123] The acquisition module 110 may be used to execute the above step S110 . The detailed implementation of the acquisition module 110 may refer to the above detailed description of step S110 .

[0124] The first calculation module 120 is used to obtain the second energy consumption required for each computing node to operate normally based on the surrounding supporting energy consumption, and calculate the first carbon emissions of each computing node according to the first energy consumption and the second energy consumption of each computing node.

[0125] Peripheral supporting energy consumption is the energy consumption to ensure the normal operation of computing nodes. For example, peripheral supporting energy consumption includes cooling supporting energy consumption and power supply supporting energy consumption. Among them, cooling supporting energy consumption is the energy consumption generated by providing cooling for computing nodes in the supercomputing center, and power supply supporting energy consumption is the energy consumption generated by providing stable power supply for computing nodes in the supercomputing center.

[0126] Among them, the second energy consumption is the surrounding supporting energy consumption allocated to each computing node. The first computing module 120 can calculate the total energy consumption of each computing node based on the first energy consumption and the second energy consumption of each computing node, and calculate the first carbon emissions of the computing node based on the total energy consumption of each computing node, wherein the first carbon emissions are the actual carbon emissions of the computing nodes in the supercomputing center.

[0127] The first calculation module 120 may be used to execute the above step S120. The detailed implementation of the first calculation module 120 may refer to the above detailed description of step S120.

[0128] The second calculation module 130 is used to calculate the total energy consumption of the computing equipment of the supercomputing center based on the first energy consumption of each computing node, and to calculate the total carbon emissions of the supercomputing center based on the total energy consumption of the computing equipment and the energy consumption of surrounding supporting equipment.

[0129] In this embodiment, the second calculation module 130 obtains the total energy consumption of the supercomputing center based on the total energy consumption of the computing equipment and the energy consumption of the surrounding supporting equipment, and obtains the total carbon emissions of the supercomputing center based on the total energy consumption of the supercomputing center.

[0130] The second calculation module 130 executes the above step S130 . The detailed implementation of the second calculation module 130 may refer to the above detailed description of step S130 .

[0131] The third calculation module 140 is used to calculate the second carbon emissions of the computing nodes based on the total carbon emissions and the number of computing nodes in the supercomputing center.

[0132] Among them, the second carbon emissions are the average carbon emissions of computing nodes in the supercomputing center.

[0133] The third calculation module 140 may be used to execute the above step S140. The detailed implementation of the third calculation module 140 may refer to the above detailed description of step S140.

[0134] The first target computing node confirmation module 150 is used to compare the first carbon emission with the second carbon emission, and to use a computing node whose first carbon emission is higher than the second carbon emission as a first target computing node.

[0135] The first target computing node confirmation module 150 compares the first carbon emission with the second carbon emission to select a computing node with a larger actual carbon emission as the first target computing node.

[0136] The first target computing node confirmation module 150 may be used to execute the above step S150. The detailed implementation of the first target computing node confirmation module 150 may refer to the above detailed description of step S150.

[0137] The second target computing node confirmation module 160 is used to obtain the computing resource occupancy rate of the computing node whose first carbon emission is lower than the second carbon emission, and use the computing node whose computing resource occupancy rate is lower than the preset computing resource occupancy rate as the second target computing node.

[0138] The second target computing node confirmation module 160 selects computing nodes with smaller actual carbon emissions and lower computing resource occupancy rates as second target computing nodes.

[0139] The second target computing node confirmation module 160 may be used to execute the above step S160 . The detailed implementation of the second target computing node confirmation module 160 may refer to the above detailed description of step S160 .

[0140] The computing task quantity confirmation module 170 is used to input the computing task quantity being processed in the first target computing node, the computing resource occupancy rate, the first carbon emission of the first target computing node, and the computing task quantity being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emission of the second target computing node into the computing task scheduling model, and determine the target computing task that needs to be scheduled from the first target computing node to the second target computing node for processing.

[0141] In this embodiment, the computing task amount confirmation module 170 can determine the target computing task scheduled from the first target computing node to the second target computing node through the computing task scheduling model. For example, the computing resources occupied by the target computing task on the second target computing node do not exceed 50% of the idle computing resources in the second target computing node, so as to avoid the second target computing node from heating up rapidly after too many computing tasks are transferred to the second target computing node. In this way, the sum of the first carbon emissions of the first target computing node and the second target computing node before scheduling can be less than the sum of the first carbon emissions of the first target computing node and the second target computing node after scheduling.

[0142] The computing task volume confirmation module 170 may be used to execute the above step S170 . The detailed implementation of the computing task volume confirmation module 170 may refer to the above detailed description of step S170 .

[0143] The scheduling module 180 is used to schedule the target computing task in the first target computing node to the second target computing node for processing.

[0144] The scheduling module 180 may be used to execute the above step S180 . The detailed implementation of the scheduling module 180 may refer to the detailed description of the above step S170 .

[0145] It should be noted that it should be understood that the division of the various modules of the above system is only a division of logical functions. In actual implementation, all or part of them can be integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software (for example, open source software) called by processing elements. It can also be implemented in the form of hardware. Some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the calculation task amount confirmation module 170 can be a separately established processing element, or it can be integrated in a certain chip of the above system for implementation. In addition, it can also be stored in the memory of the above system in the form of program code, and called and executed by a certain processing element of the above system. The function of the above calculation task amount confirmation module 170. The implementation of other modules is similar. In addition, all or part of these modules can be integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.

[0146] Please refer to Figure 5 , Figure 5 The hardware structure diagram of the server 10 provided in the embodiment of the present disclosure for implementing the above-mentioned supercomputing center energy consumption and carbon emission control method based on artificial intelligence is shown. The server 10 can be implemented on a cloud server. Figure 5 As shown, the server 10 may include a processor 101 , a computer-readable storage medium 102 , a bus 103 , and a communication interface 104 .

[0147] In a specific implementation process, at least one processor 101 executes computer-executable instructions (e.g., Figure 4 ), so that the processor 101 can execute the supercomputing center energy consumption and carbon emission control method based on artificial intelligence as described in the above method embodiment, wherein the processor 101, the computer-readable storage medium 102 and the communication interface 104 are connected via a bus 103, and the processor 101 can be used to control the sending and receiving actions of the communication interface 104.

[0148] The specific implementation process of the processor 101 can refer to the various method embodiments executed by the above-mentioned server 10. The implementation principles and technical effects are similar and will not be repeated in this embodiment.

[0149] The computer-readable storage medium 102 may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk storage.

[0150] The bus 103 can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the bus in the drawings of the present invention is not limited to only one bus or one type of bus.

[0151] In addition, an embodiment of the present invention also provides a computer-readable storage medium, which stores computer execution instructions. When the processor executes the computer execution instructions, the above-mentioned supercomputing center energy consumption and carbon emission control method based on artificial intelligence is implemented.

[0152] In summary, the technical solution provided by the embodiment of the present invention first obtains the first energy consumption and the second energy consumption of each computing node, and calculates the first carbon emission of the computing node based on the first energy consumption and the second energy consumption; then, the total carbon emission of the supercomputing center is calculated according to the total energy consumption of the computing equipment in the supercomputing center and the energy consumption of the surrounding supporting equipment, and the second carbon emission of the computing node is obtained based on the total carbon emission of the supercomputing center and the number of computing nodes; then, the first target computing node and the second target computing node that need to adjust the computing task are screened out based on the first carbon emission and the second carbon emission, and the target computing task that needs to be scheduled in the first target computing node is determined through the computing task scheduling model; finally, the target computing task is scheduled to the second target computing node for processing. In the above scheme, by monitoring the carbon emissions of each computing node, the computing tasks in the computing nodes with larger carbon emissions are scheduled to the computing nodes with smaller emissions for processing. Under the condition of ensuring the computing power demand, the overall energy consumption and carbon emissions of the supercomputing center can be reduced, and the impact of the supercomputing center on the environment can be reduced.

[0153] The above is a description of a specific embodiment of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence, characterized in that: A server is applied to monitor and control the supercomputing center, wherein the supercomputing center includes a plurality of computing nodes communicating with each other, and the method includes: Acquire a current first energy consumption of each of the computing nodes, wherein the first energy consumption is energy consumption for a processor or a processor and an image processor in the computing node to work at a time granularity; Based on the peripheral supporting energy consumption, the second energy consumption required to enable each of the computing nodes to work normally is obtained, and according to the first energy consumption and the second energy consumption of each of the computing nodes, the first carbon emissions of each computing node are calculated, wherein the peripheral supporting energy consumption includes cooling supporting energy consumption and power supply supporting energy consumption, the peripheral supporting energy consumption is the peripheral supporting energy consumption at the same time granularity as the first energy consumption, and the second energy consumption is the peripheral supporting energy consumption allocated to each of the computing nodes; Calculate the total energy consumption of the computing equipment of the supercomputing center based on the first energy consumption of each computing node, and calculate the total carbon emissions of the supercomputing center based on the total energy consumption of the computing equipment and the energy consumption of the surrounding supporting equipment; Based on the total carbon emissions and the number of computing nodes in the supercomputing center, calculate a second carbon emission amount of the computing node; Compare the first carbon emission with the second carbon emission, and use a computing node whose first carbon emission is higher than the second carbon emission as a first target computing node; Obtaining a computing resource occupancy rate of a computing node whose first carbon emission rate is lower than that of the second carbon emission rate, and taking a computing node whose computing resource occupancy rate is lower than a preset computing resource occupancy rate as a second target computing node; The amount of computing tasks being processed in the first target computing node, the computing resource occupancy rate, and the first carbon emissions of the first target computing node, as well as the amount of computing tasks being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emissions of the second target computing node are input into a computing task scheduling model to determine the target computing tasks that need to be scheduled from the first target computing node to the second target computing node for processing, wherein the sum of the first carbon emissions of the first target computing node and the second target computing node before scheduling is less than the sum of the first carbon emissions of the first target computing node and the second target computing node after scheduling; The target computing task in the first target computing node is dispatched to the second target computing node for processing.

2. The method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence according to claim 1, characterized in that: The peripheral supporting energy consumption includes cooling supporting energy consumption and power supply supporting energy consumption. The step of obtaining the second energy consumption required to enable each computing node to work normally based on the peripheral supporting energy consumption, and calculating the first carbon emissions of each computing node according to the first energy consumption and the second energy consumption of each computing node includes: Calculate the cooling energy consumption required for each computing node based on the type of cooling equipment; Based on the power supply supporting energy consumption and the current first energy consumption of each computing node, the power supply energy consumption required by each computing node is calculated; Obtaining the second energy consumption required to enable each computing node to work normally from the cooling energy consumption and the power supply energy consumption required by each computing node; Based on the energy consumption source type of the first energy consumption and the second energy consumption of each computing node, a first carbon emission of each computing node is calculated.

3. The method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence according to claim 2, characterized in that: The step of calculating the cooling energy consumption required for each computing node based on the type of cooling equipment includes: When the type of cooling support is air cooling, based on the total power consumption of the cooling equipment and the number of the computing nodes, the air cooling energy consumption required for each computing node is calculated, and the air cooling energy consumption is used as the cooling energy consumption required for the computing node; or When the type of cooling support is liquid cooling, the liquid cooling power consumption of the liquid cooling device configured for each computing node is used as the cooling energy consumption required for each computing node; or, When the types of cooling support include air cooling and liquid cooling, the sum of the air cooling energy consumption required for each computing node and the liquid cooling power consumption of the liquid cooling device configured for the computing node is used as the cooling energy consumption required for each computing node.

4. The method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence according to claim 1, characterized in that: The step of calculating the total energy consumption of the computing equipment of the supercomputing center based on the first energy consumption of each computing node, and calculating the total carbon emissions of the supercomputing center based on the total energy consumption of the computing equipment and the surrounding supporting energy consumption, comprises: Adding the first energy consumption of all computing nodes to obtain the total energy consumption of computing equipment in the supercomputing center; The total carbon emissions of the supercomputing center are calculated based on the power source type of the total energy consumption of the computing equipment of the supercomputing center and the power source type of the surrounding supporting energy consumption; The step of calculating the second carbon emissions of the computing nodes based on the total carbon emissions and the number of computing nodes in the supercomputing center includes: The total carbon emissions are divided by the number of computing nodes in the supercomputing center to obtain an average carbon emissions of each computing node, and the average carbon emissions are used as the second carbon emissions of the computing node.

5. The method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence according to claim 1, characterized in that: Before inputting the amount of computing tasks being processed in the first target computing node, the computing resource occupancy rate, the first carbon emission of the first target computing node, and the amount of computing tasks being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emission of the second target computing node into the computing task scheduling model, and determining the target computing tasks that need to be scheduled from the first target computing node to the second target computing node for processing, the method further includes the step of training the computing task scheduling model, which includes: Create a training sample set, the training sample set including multiple training samples and sample tags corresponding to the training samples, wherein the training samples include a first computing task amount of a first computing node sample, a first computing resource occupancy rate of the first computing node sample, a first carbon emission of the first computing node sample, a second computing task amount of a second computing node sample, a first computing resource occupancy rate of the second computing node sample, and a first carbon emission of the second computing node sample, and the sample tags include a computing task amount of the first computing node sample that needs to be scheduled to the second computing node sample, a first carbon emission of the first computing node sample after scheduling, and a first carbon emission of the second computing node sample after scheduling; Input the training sample into a deep learning network for training, and output a training result, wherein the training result includes a first training carbon emission of the first computing node sample, a second training carbon emission of the second computing node sample, and an amount of training computing tasks scheduled from the first computing node sample to the second computing node sample; The loss function value of the deep learning network is calculated according to the training results and the sample labels. When the loss function value is greater than a preset loss function value, the network parameters in the deep learning network are adjusted, and the training process is repeated until the loss function value is no greater than the preset loss function value. The training is terminated, and the deep learning network obtained through training is used as the computing task scheduling model.

6. The method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence according to claim 1, characterized in that: The method further comprises: Monitor the space temperature of the space where the supercomputing center is located, the ambient temperature outside the space where the supercomputing center is located, and the node operating temperature of each computing node in the supercomputing center; When the temperature difference between the space temperature and the ambient temperature is greater than a first preset temperature difference, controlling the fan to transfer air outside the space where the supercomputing center is located to the space where the supercomputing center is located; When the temperature difference between the space temperature and the ambient temperature is less than or equal to a first preset temperature difference, controlling the air cooling equipment in the space where the supercomputing center is located to operate; When the temperature difference between the node operating temperature and the space temperature is greater than a second preset temperature difference, the liquid cooling device configured for the corresponding computing node is controlled to operate.

7. The method for controlling energy consumption and carbon emissions of a supercomputing center based on artificial intelligence according to claim 6, characterized in that: The method further comprises: Arrange a solar power generation device outside the space where the supercomputing center is located; When the fan or the air cooling device needs to work, the solar power generation device provides electric energy to the fan or the air cooling device.

8. An artificial intelligence-based supercomputing center energy consumption and carbon emission control system, characterized in that: A server used for monitoring and controlling the supercomputing center, wherein the supercomputing center includes a plurality of computing nodes communicating with each other, and the system includes: An acquisition module, used for acquiring a current first energy consumption of each computing node, wherein the first energy consumption is energy consumption for a processor or a processor and an image processor in the computing node to work at a time granularity; A first calculation module is used to obtain a second energy consumption required to enable each of the computing nodes to work normally based on the surrounding supporting energy consumption, and calculate the first carbon emissions of each computing node according to the first energy consumption and the second energy consumption of each computing node, wherein the surrounding supporting energy consumption includes cooling supporting energy consumption and power supply supporting energy consumption, the surrounding supporting energy consumption is the surrounding supporting energy consumption at the same time granularity as the first energy consumption, and the second energy consumption is the surrounding supporting energy consumption allocated to each of the computing nodes; A second calculation module is used to calculate the total energy consumption of the computing equipment of the supercomputing center based on the first energy consumption of each computing node, and calculate the total carbon emissions of the supercomputing center based on the total energy consumption of the computing equipment and the energy consumption of the surrounding supporting equipment; A third calculation module, configured to calculate a second carbon emission amount of the computing node based on the total carbon emission amount and the number of computing nodes in the supercomputing center; A first target computing node confirmation module, configured to compare the first carbon emission with the second carbon emission, and to use a computing node whose first carbon emission is higher than the second carbon emission as a first target computing node; A second target computing node confirmation module is used to obtain the computing resource occupancy rate of the computing node whose first carbon emission is lower than the second carbon emission, and use the computing node whose computing resource occupancy rate is lower than the preset computing resource occupancy rate as the second target computing node; A computing task quantity confirmation module is used to input the computing task quantity being processed in the first target computing node, the computing resource occupancy rate, the first carbon emission of the first target computing node, and the computing task quantity being processed in the second target computing node, the computing resource occupancy rate, and the first carbon emission of the second target computing node into a computing task scheduling model, and determine the target computing task that needs to be scheduled from the first target computing node to the second target computing node for processing, wherein the sum of the first carbon emission of the first target computing node and the second target computing node before scheduling is less than the sum of the first carbon emission of the first target computing node and the second target computing node after scheduling; The scheduling module is used to schedule the target computing task in the first target computing node to the second target computing node for processing.

9. A server, characterized in that: The server includes a processor, a computer-readable storage medium and a communication interface. The computer-readable storage medium, the communication interface and the processor are connected via a bus system. The computer-readable storage medium is used to store programs, instructions or codes. The processor is used to execute the programs, instructions or codes in the computer-readable storage medium to execute the artificial intelligence-based supercomputing center energy consumption and carbon emission control method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions which, when executed, enable the server to execute the artificial intelligence-based supercomputing center energy consumption and carbon emission control method as described in any one of claims 1 to 7.

Images