CN113225994B - Intelligent air conditioner control method facing data center - Google Patents
Intelligent air conditioner control method facing data center Download PDFInfo
- Publication number
- CN113225994B CN113225994B CN202110502901.0A CN202110502901A CN113225994B CN 113225994 B CN113225994 B CN 113225994B CN 202110502901 A CN202110502901 A CN 202110502901A CN 113225994 B CN113225994 B CN 113225994B
- Authority
- CN
- China
- Prior art keywords
- energy consumption
- equipment
- data center
- node
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- H—ELECTRICITY
- H05—ELECTRIC TECHNIQUES NOT OTHERWISE PROVIDED FOR
- H05K—PRINTED CIRCUITS; CASINGS OR CONSTRUCTIONAL DETAILS OF ELECTRIC APPARATUS; MANUFACTURE OF ASSEMBLAGES OF ELECTRICAL COMPONENTS
- H05K7/00—Constructional details common to different types of electric apparatus
- H05K7/20—Modifications to facilitate cooling, ventilating, or heating
- H05K7/20709—Modifications to facilitate cooling, ventilating, or heating for server racks or cabinets; for data centers, e.g. 19-inch computer racks
- H05K7/20836—Thermal management, e.g. server temperature control
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Abstract
The invention provides an intelligent air conditioner control system design oriented to data center energy-saving control. By combining the analysis of the characteristics of the application system with the control of the air conditioning system, an intelligent energy-saving control strategy is realized. The power consumption of a large number of computing nodes is closely related to the application of the running application system, and the application with intensive operation usually needs to occupy more processor and accelerator resources, so that the power consumption of a computing cluster is improved, and some online service systems have relatively low requirements on computing capacity, so that the cluster power consumption is reduced. The same application system has different requirements for computation density with different function calls, resulting in different power consumption requirements. The invention adopts the machine learning technology to model the operation characteristics of the data center computing cluster and judge the power consumption development rule, thereby providing an effective regulation and control strategy for an air conditioner control system and realizing the purpose of ensuring the safe operation of the data center with the lowest possible power consumption.
Description
Technical Field
The invention belongs to the technical field of information, and particularly relates to a data center temperature control system based on an intelligent algorithm, which monitors and acquires the state of a data center computing cluster, performs predictive analysis on the demand of energy consumption by an artificial intelligence technology, thereby realizing a high-efficiency and energy-saving temperature control system control strategy,
background
Cloud computing has become an important technical path for informatization of various industries, and particularly in the process of IT system construction, computing power and storage are realized in the modes of public cloud, private cloud, proprietary cloud, mixed cloud and the like, and the cloud computing has become a basic option for enterprises to establish digital infrastructure. The cloud computing infrastructure-data center corresponding to the cloud computing infrastructure-data center needs to support large-scale deployment of computing, storage and network equipment to provide computing power and data storage resources with different configurations. Meanwhile, the heat dissipation problem brought by large-scale equipment deployment brings technical challenges to the configuration and control of the air conditioning equipment of the data center machine room. The server nodes in the computing cluster often have different working loads due to different service types, and an unbalanced load phenomenon is also generated in different time periods, so that the heat dissipation requirements are different due to different load states, and although the normal operation of the computing cluster system is guaranteed by adopting fixed air conditioning system power (highest heat dissipation requirement), the energy consumption of the air conditioning system is easily wasted in a low-load state.
At present, a data center air conditioner control system mainly adopts a fixed temperature control mode or carries out feedback control through detection of an environmental system sensor, adaptability adjustment can only be carried out on control measurement according to the current state, and the work load of a data center computing cluster cannot be predicted, so that a temperature control adjusting strategy with a certain time period is formulated, and the energy efficiency is improved. With the development of machine learning technology, more and more systems gradually introduce predictive technology to achieve optimization, but there is still no similar development in air conditioning control of large data centers and industrial production.
Disclosure of Invention
According to the limitation of the existing data center air conditioner control system, namely, only a fixed temperature control mode or feedback control can be carried out through detection of an environmental system sensor, and the working load of a data center computing cluster cannot be predicted, so that a targeted temperature control regulation strategy is realized. The method is based on the machine learning technology, and establishes a prediction model for the energy consumption behavior of the equipment nodes in a statistical modeling mode by collecting the state information of each equipment node, so as to assist the air conditioning system to carry out targeted heat dissipation capacity adjustment. And the optimization of the power consumption of the air conditioning system is realized on the basis of ensuring the normal operation of the whole computing cluster. The technology is used as a part of an intelligent control system and provides service for realizing energy consumption optimization of a data center.
The invention takes each computing device of the data center as a node of a device graph, namely, a server, a network switching device, a storage controller device and a data security device form the device graph G, and each device is taken as a node v of the device graph. The association relationship E between each device graph node is based on whether a network connection exists (as shown in fig. 1, a device graph is defined as G ═ V, E }, V denotes a graph node set, E denotes a set of graph edges, each node V carries an attribute { type, cost }, type denotes a node type, and cost denotes energy consumption information). The method has the advantages that the workload of each node on the equipment diagram is subjected to data acquisition, the workload state modeling is carried out in a machine learning mode, the load and energy consumption states from a single equipment node to any subgraph in the whole equipment diagram are predicted, and accordingly the temperature control prediction capability of the auxiliary air-conditioning system is realized. Specifically, the continuous optimization of the temperature control predictive regulation and the prediction model in the data center is realized through the following 5 basic steps:
1. acquiring the state of equipment nodes;
2. constructing an equipment node energy consumption model;
3. predicting the load of the equipment nodes;
4. predictive regulation of the temperature control system;
5. model iteration and updating;
step 1 requires the collection of the workload status of each device node, in particular the continuous collection of the status relating to power consumption, including for example for a server: processor utilization, storage system utilization, accelerator (GPU) utilization, and the like. After continuously collecting data for a given time period, modeling the workload state and trend of the equipment diagram (all equipment in the data center) is realized in step 2 through a machine learning technology, so that the prediction capability of the working state of each piece of equipment is realized (namely step 3). And calculating a power consumption value by combining the predicted working state of the equipment with the energy consumption range of the equipment to form energy consumption prediction on an equipment graph or a sub-graph, further deducing heat dissipation requirements, and providing predictive data for adjusting an air conditioning system (temperature control equipment). The air conditioning system can preset a temperature control strategy and correct according to the actual heat dissipation condition. Because the workload state of the data center equipment cluster changes along with the expansion of the time period, the machine learning model established in the step 2 still needs to be continuously updated to adapt to the new workload distribution characteristic, so in the step 5, data is acquired again in a fixed period (corresponding to the content in the step 1), and the machine learning model is reconstructed (corresponding to the content in the step 2), so that the high applicability of the machine learning model is realized, and the prediction accuracy is improved.
Compared with the prior art, the invention has the beneficial effects that: the existing air conditioner temperature control technology or a manual preset strategy has the defect that accurate prediction cannot be realized, so that the preset can be only carried out by experience; or self-adaptive adjustment is carried out according to real-time monitoring of the sensor, and the defect that the adjustment can only be carried out according to the current state and the problem of response delay exists. Compared with the defects of the existing system, the artificial intelligence technology based on the graph neural network is introduced, the power consumption states of the data center equipment can be predicted in different granularities (from single equipment to a cluster and then to the whole data center), accurate energy consumption trend prediction is realized based on prediction, and the strategy arrangement of the air conditioner temperature control system is guided; the error possibly caused by the machine learning model is corrected by combining with real-time monitoring; and updating the model in a self-learning mode to realize continuous adaptation of the model and the state of the application system, thereby ensuring the prediction precision.
Drawings
FIG. 1 data center and equipment diagrammatic representation
FIG. 2 temperature control strategy prediction workflow
FIG. 3 data acquisition
FIG. 4 machine learning modeling
FIG. 5 temperature control strategy prediction and on-line correction
FIG. 6 model iterative update flow
Detailed Description
The concrete implementation mode of the 5 steps of the invention is as follows:
1. device node state collection
The purpose of the node state acquisition of the equipment is to obtain energy consumption information expressed in time series for each node, namely, the type of each node is unchanged, and cost can be expressed as a data sequence { cost based on time series0,cost1,…costnAnd the energy consumption status of the whole plant diagram at a particular instant t can be denoted GtCorresponding to each node viThe energy consumption information of (1) is costt. The running state acquisition of different equipment is realized, and 3 types of equipment are mainly and intensively processed: the server, the network switch and the storage controller are used for collecting information as shown in the following table:
the energy consumption of the server node mainly comes from a processor, an accelerator, a memory and a disk I/O; since the network switch adopts an embedded system, the power consumption state can be estimated through the network I/O of the network switch, and the power consumption state can be estimated through the data I/O (the same as the network I/O) of a storage system (a special storage system).
The acquisition mode (as shown in the attached figure 3) is as follows:
I. a server: the running state of the servers is collected and the data is transmitted back to the information collection server by presetting a data collection process on each server. For different bottom operating systems (Linux, Windows Server, etc.), corresponding acquisition programs need to be customized. For a server controlled by a virtualization platform, state acquisition can be realized through an operation interface provided by the virtualization platform, for example, VMWare, Xen, OpenStack all provide state monitoring and acquisition services of each computing node of a cluster;
network switch: monitoring the throughput of the switch by setting a state acquisition program for a server node in a network to which the switch belongs;
storage device/storage controller: the network throughput of the storage system is monitored by setting a state acquisition program at a server node in a network to which the storage system belongs or directly running a monitoring program on a storage controller.
The acquired state information is converted into energy consumption information, and the energy consumption information is calculated in the following mode:
I. a server: processor occupancy rate processor energy consumption standard + memory occupancy rate memory energy consumption standard + disk I/O rate disk energy consumption standard + accelerator occupancy rate accelerator energy consumption standard, where each energy consumption standard needs to be determined according to the upper energy consumption limit of a given component provided by an equipment manufacturer, and watt is taken as a unit;
network switch: the network I/O rate is the energy consumption standard of the switch, and all the energy consumption standards are determined according to the energy consumption upper limit provided by equipment manufacturers, and the watt is taken as a unit;
storage device/storage controller: network I/O rate switch energy consumption standards, as above, each energy consumption standard herein needs to be determined according to an energy consumption upper limit provided by a device manufacturer, in watt units.
2. Equipment node energy consumption prediction model construction
The energy consumption prediction model predicts the energy consumption state of each equipment node on the basis of the equipment graph of the data center. The definition of the device graph is: the graph nodes represent operation equipment (comprising servers, network switches and storage equipment) of the data center; edges between nodes are represented as device nodes connected through a network. The energy consumption prediction of each equipment node needs to consider the states of the equipment nodes adjacent to the equipment node, for example, two groups of servers perform data pipeline calculation, so that network and calculation intensive operations need to be performed, and for example, one group of servers participate in the map-reduce calculation of hadoop, so that data and calculation intensive operations need to be performed.
In view of the above, the machine learning prediction model needs to have the capability of performing multivariate prediction based on time series for expressing data in a graph. Variables in a multivariate time series can be seen as nodes in a device graph that express dependencies by being connected in a network. In the present invention, we select Graph Convolutional Network Graph (GCN) as the basic architecture of the prediction model (fig. 4. a). The multivariate timing data and the external graph structure are used as input to predict future values of the timing data. The variables here correspond to the nodes of the device map, and the time series data correspond to the energy consumption information of each node at different times. Because energy consumption information of the current device node in the next time period needs to be predicted, in the process of graph neural network input, energy consumption information of each node needs to be added in information aggregation of each node. The convolution operator formula of the graph neural convolution network is shown in figure 4. b.
The purpose of GCN-based machine learning modeling is to build a prediction Model GCN _ Model that can predict the plant diagram energy consumption state at time t + x from the plant diagram energy consumption information at time t: cost (G)t+x)＝GCN_Model(Gt) And x is a time period (usually 1 to 8 hours, or 24 hours). The concrete modeling steps are as follows:
I. collecting device graph energy consumption information in given time periodInformation: through the data collection of step 1 of the present invention, a sequence of energy consumption information { G ] corresponding to a time period (one week or one month may be used) of an equipment diagram is collected0,G1,…Gn}; representing the energy consumption information of the equipment graph corresponding to the time state i as each node v in the graphiThe cost information of (1), the calculation method is as shown in step 1;
each element of the data set D is denoted Di＝{Gi,cost(Gi+x)},GiI.e. all information of the device graph at time i, including energy consumption information and constant information of each node (node type) and association information between nodes (edge of graph: network connection of device), GiAs an input to the GCN; cost (G)i+x) Energy consumption information of each node represented as G at time i + x, cost (Gi + x) as a prediction target of the GCN; the choice of the value x can be determined according to the application scenario, typically with reference to a time period of 1 to 8 hours;
for each node v in the device graphiIn the invention, the GCN needs to aggregate the information of the neighbor nodes and carry out convolution calculation, and the type and energy consumption information of the node vi also comprises convolution calculation;
and IV, constructing a GCN _ Model by using the D as a training data set for GCN modeling.
The graph convolution neural network expresses the graph information of input and output in an adjacency matrix, so the size of the graph (the number of nodes and edges) affects the processing efficiency and the memory usage. The amount of equipment in a data center is typically above 1000 levels, i.e., requiring greater than 1000 x 1000 of matrix processing capacity. For very large matrices (e.g. for>10000 x 10000), we reduce the storage pressure in a partition mode, i.e. one device graph is decomposed into a plurality of subgraphs, and the subgraphs are modeled respectively. For example: if a device graph G is divided into a plurality of sub-graphs Gp0,Gp1,…GpmThe GCN corresponding to each subgraph is GCN _ Modelp0,GCN_Model p1,…GCN_Model pm. Each subgraph avoids the existence of association (namely network connection) as much as possible, so that the subgraphs can be regarded as networks independent of each other, and prediction errors are reduced.
3. Device node load prediction
After the prediction Model GCN _ Model based on the graph neural convolution network is constructed, the energy consumption information can be predicted in the following way: acquiring energy consumption information of data center equipment, and taking the equipment diagram state information in the step 2 as an input mode GtTo predict cost (G)t+x) (i.e., the device map energy consumption information at time t + x, including the energy consumption status of each device node).
4. Predictive regulation of temperature control system
The predictive regulation of the temperature control system presets the power control of the air conditioning system according to the energy consumption state prediction of the future time period (x time period) of the data center equipment, thereby effectively avoiding energy loss possibly caused by full-load operation. The energy consumption of the data center equipment is predicted by the mode in step 3, namely cost (G)t+x)＝GCN_Model(Gt) To predict energy consumption information, cost (G), for each device nodet+x) Representing energy consumption information for all device nodes in the device graph (i.e., cost (v))0),cost(v1),…cost(vn) And carrying out synthesis: cost (v) sum0)+cost(v1)+…+cost(vn) To obtain energy consumption information for the device graph or a portion of the nodes therein.
In order to avoid the possible prediction error of the machine learning model, the air conditioner temperature control system still needs to introduce a monitoring sensor to monitor the current temperature state so as to perform temperature control strategy adjustment on the possible prediction error (as shown in fig. 5).
5. Model iteration and update
The energy consumption information of the equipment is closely related to a service system running on the equipment, and the service system is not invariable but changes according to time and customer service requirements, so that an energy consumption prediction model obtained by data sampling and model training in a certain time period cannot realize prediction covering the whole time period of a data center. Therefore, Model reconstruction needs to be performed within a certain time interval, that is, after a time period y (the value may be week or 1 month), steps 1 and 2 are repeated, that is, the device node energy consumption information is re-acquired, and the GCN _ Model is retrained. Since the temperature control operation performed in steps 3 and 4 requires continuous energy consumption information acquisition in the data center, the data required for model reconstruction can be obtained from daily energy consumption information acquisition, and continuous model iteration is realized (as shown in fig. 6).
Claims (7)
1. The data center-oriented intelligent air conditioner control method is characterized in that distributed management is carried out on a data center air conditioner temperature control system in a gridding mode, each air conditioner controller corresponds to a computing environment, modeling based on artificial intelligence is carried out through an environment characteristic set to realize fine-grained and gridded environment control, and the method comprises the following steps:
taking each computing device of the data center as a node of the equipment graph, establishing an incidence relation of the node of each equipment graph according to the fact that whether network connection exists between the computing devices, and taking the incidence relation as an edge of the equipment graph;
acquiring equipment node state information, and converting the acquired state information into energy consumption information;
constructing an equipment node energy consumption prediction model based on GCN machine learning modeling;
acquiring energy consumption information of data center equipment, and predicting the energy consumption information of equipment nodes through the energy consumption prediction model;
performing preset predictive regulation of a temperature control system on the power control of the air conditioning system according to the energy consumption state prediction of the equipment node in the future time period;
iteratively updating the energy consumption prediction model according to a preset time period through the acquired daily energy consumption information;
wherein the machine learning modeling based on GCN establishes a prediction model according to the time
To predict device graph energy consumption information over time
The device map energy consumption state;
acquiring equipment diagram energy consumption information in a given time period to generate a data set D;
each element of the data set D is represented as
,
Representing the plant diagram in time
All information of the GCN, including energy consumption information, constant information of each node and association information among the nodes, is used as the input of the GCN;
expressed as each node being at time
As a predicted output target of the GCN;
each node in the GCN aggregation equipment graph
And carrying out convolution calculation on the information of the neighbor nodes, and constructing an energy consumption prediction model by taking the data set D as a training data set for GCN modeling.
2. The intelligent air-conditioning control method facing the data center according to claim 1, wherein the environment feature set of the computing environment corresponding to each air-conditioning controller is composed of features of an application software layer and a physical equipment layer, and the application software layer is represented by computing task features: comprises one or more than two combinations of calculation density, storage density, memory access density and communication density;
the physical device layer is represented by the running states of various devices, and comprises the following steps: CPU occupancy rate, GPU occupancy rate, CPU, GPU memory occupancy rate, network I/O rate, and disk I/O rate.
3. The intelligent air conditioner control method oriented to the data center as claimed in claim 1, wherein the energy consumption prediction model predicts the energy consumption state of each equipment node based on an equipment graph of the data center; the graph nodes in the device graph represent operation devices of the data center, and comprise servers, network switches and storage devices, and edges between the nodes represent device nodes connected through a network.
4. The intelligent air-conditioning control method oriented to the data center as claimed in claim 1, wherein the energy consumption information of each equipment node in the equipment diagram is predicted through the energy consumption prediction model, and the energy consumption information of each equipment node is integrated to obtain the whole or partial energy consumption information of the equipment diagram.
5. The intelligent air conditioner control method for the data center according to claim 1, further comprising: and acquiring a current temperature state through the monitoring sensor for monitoring, adjusting strategy arrangement of the temperature control system according to the current temperature state, and correcting errors existing in the energy consumption prediction model.
6. The intelligent air conditioner control method oriented to the data center as claimed in claim 1, wherein a time period is set according to time and changes of customer service demands, energy consumption information of equipment nodes is collected again according to the time period, the energy consumption prediction model is retrained for model reconstruction, and data required by the model reconstruction is obtained from daily energy consumption information.
7. The intelligent air conditioner control method oriented to the data center as claimed in claim 1, wherein machine learning modeling independent of each other is performed on device graph subgraphs without connection to each other and the capability of predicting power consumption information is realized.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110502901.0A CN113225994B (en) | 2021-05-08 | 2021-05-08 | Intelligent air conditioner control method facing data center |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110502901.0A CN113225994B (en) | 2021-05-08 | 2021-05-08 | Intelligent air conditioner control method facing data center |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113225994A CN113225994A (en) | 2021-08-06 |
| CN113225994B true CN113225994B (en) | 2022-06-24 |
Family
ID=77094559
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110502901.0A Active CN113225994B (en) | 2021-05-08 | 2021-05-08 | Intelligent air conditioner control method facing data center |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113225994B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114002952A (en) * | 2021-09-29 | 2022-02-01 | 苏州浪潮智能科技有限公司 | Cluster centralized heat dissipation regulation and control system and method |
| CN114253159B (en) * | 2021-12-10 | 2022-10-04 | 安徽中家智锐科技有限公司 | Control method and system for electric appliance test environment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106300349A (en) * | 2016-11-05 | 2017-01-04 | 北京普锐电子有限公司 | A kind of region electricity consumption intelligent management system |
| CN107065582A (en) * | 2017-03-31 | 2017-08-18 | 苏州科技大学 | The Intelligent Indoor Air regulating system and adjusting method of a kind of parameter based on environment |
| CN108534295A (en) * | 2018-03-02 | 2018-09-14 | 珠海格力电器股份有限公司 | The methods, devices and systems of display data |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105868487B (en) * | 2016-04-11 | 2019-04-16 | 中国建筑科学研究院 | A kind of energy saving expert system based on energy consumption equipment model |
| CN111178626A (en) * | 2019-12-30 | 2020-05-19 | 苏州科技大学 | Building energy consumption prediction method and monitoring prediction system based on WGAN algorithm |
| CN112616292B (en) * | 2020-11-27 | 2022-04-22 | 湖南大学 | Data center energy efficiency optimization control method based on neural network model |
-
2021
- 2021-05-08 CN CN202110502901.0A patent/CN113225994B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106300349A (en) * | 2016-11-05 | 2017-01-04 | 北京普锐电子有限公司 | A kind of region electricity consumption intelligent management system |
| CN107065582A (en) * | 2017-03-31 | 2017-08-18 | 苏州科技大学 | The Intelligent Indoor Air regulating system and adjusting method of a kind of parameter based on environment |
| CN108534295A (en) * | 2018-03-02 | 2018-09-14 | 珠海格力电器股份有限公司 | The methods, devices and systems of display data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113225994A (en) | 2021-08-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109800066B (en) | Energy-saving scheduling method and system for data center | |
| JP6584040B2 (en) | Control power output using battery DC characteristics | |
| CN102004671B (en) | Resource management method of data center based on statistic model in cloud computing environment | |
| US20170315523A1 (en) | Using forecasting to control target systems | |
| Henri et al. | A supervised machine learning approach to control energy storage devices | |
| CN113225994B (en) | Intelligent air conditioner control method facing data center | |
| Su et al. | An agent-based distributed real-time optimal control strategy for building HVAC systems for applications in the context of future IoT-based smart sensor networks | |
| CN109491760B (en) | High-performance data center cloud server resource autonomous management method | |
| KR102363303B1 (en) | Decentralized Planning, Scheduling and Control of Multi-Agent Flow Control Systems | |
| Hsu et al. | Self-aware workload forecasting in data center power prediction | |
| US9964933B2 (en) | System for distributing electrical energy over a cluster of electrical devices, method for distributing electrical energy over a cluster of electrical devices of such a system and controller for use in such a system | |
| US10054997B2 (en) | Multivariable control for power-latency management to support optimization of data centers or other systems | |
| CN114899885A (en) | Power scheduling method, system and storage medium | |
| CN114626562A (en) | Intelligent monitoring method and system for running state of large public building equipment | |
| CN108108837B (en) | Regional new energy power supply structure optimization prediction method and system | |
| Dharaniya et al. | Ai agents at different data centers to minimize the energy spending | |
| CN113632132A (en) | Computer-aided energy management method and energy management system | |
| Petri et al. | Ensemble-based network edge processing | |
| KR20120133362A (en) | Optimized production scheduling system using loading simulation engine with dynamic feedback scheduling algorithm | |
| CN111210060B (en) | Method for predicting temperature of machine room during working days | |
| CN115185683A (en) | Cloud platform stream processing resource allocation method based on dynamic optimization model | |
| Mörzinger et al. | A modular, holistic optimization approach for industrial appliances | |
| CN107608237B (en) | Hardware resource optimization control method based on photovoltaic system semi-physical simulation | |
| Sun et al. | Probability-Based Online Algorithm for Switch Operation of Energy Efficient Data Center | |
| Mehenni et al. | An optimal big data processing for smart grid based on hybrid MDM/R architecture to strengthening RE integration and EE in datacenter |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |