CN-121996317-A - Intelligent computation center infrastructure modularized configuration method based on energy efficiency grading

CN121996317ACN 121996317 ACN121996317 ACN 121996317ACN-121996317-A

Abstract

The invention belongs to the technical field of resource allocation, and discloses an intelligent computation center infrastructure modularized allocation method based on energy efficiency grading; the method comprises the steps of collecting key information of an intelligent computing center, determining a given energy efficiency level of the intelligent computing center, distributing the determined additional budget power consumption of facilities to each infrastructure module to obtain budget power consumption targets of each infrastructure module, generating a plurality of configuration strategies of each infrastructure module, selecting an optimal configuration strategy, judging whether the selected optimal configuration strategy is proper or not, and executing corresponding response according to analysis results. The invention carries out comprehensive analysis by combining the correction dimension coefficient and the calculated weight coefficient to correct the set basic weight, so that the method can be more suitable for individual requirements of project equipment, environment and load when the budget power consumption target is allocated to the infrastructure module, and the budget power consumption target allocation is more accurate.

Inventors

GE XIN
ZHENG MEIYING
YANG WEI
SUN TAO
TIAN XU
Xie Anchao
WANG ZHAOMENG

Assignees

北京电信规划设计院有限公司
中讯邮电咨询设计院有限公司

Dates

Publication Date: 20260508
Application Date: 20251217

Claims (10)

1. The intelligent computing center infrastructure modularized configuration method based on energy efficiency grading is characterized by comprising the following steps of: step one, collecting key information of an intelligent computing center, and determining a set energy efficiency level of the intelligent computing center based on analysis of the collected key information; determining additional budget power consumption of the facility according to the predetermined energy efficiency level, and distributing the additional budget power consumption of the facility to each infrastructure module of the intelligent computation center to obtain budget power consumption targets of each infrastructure module; Generating a plurality of configuration strategies according to budget power consumption targets of the infrastructure modules, and selecting an optimal configuration strategy for the infrastructure modules based on the adaptation scores of the configuration strategies; step four, performing simulation on the intelligent computation center based on the optimal configuration strategy, obtaining energy consumption values of all infrastructure modules in the simulation, analyzing based on the energy consumption values, and judging whether the selected optimal configuration strategy is proper or not; and fifthly, executing corresponding response according to the analysis result.
2. The energy efficiency ranking based intelligent computing center infrastructure modular configuration method of claim 1, wherein the method of determining a given energy efficiency ranking for an intelligent computing center is: Extracting the characteristics of the collected key information of the intelligent computation center to obtain characteristic labels of a plurality of decision indexes, obtaining quantized values of all the decision indexes based on the characteristic labels of the decision indexes, obtaining environmental impact information of the intelligent computation center, generating impact factors based on the environmental impact information, and correcting the quantized values to obtain quantized correction values of all the decision indexes; the method comprises the steps of presetting a plurality of energy efficiency levels, setting quantized value standard intervals of each decision index under each energy efficiency level, comparing quantized correction values of each decision index with each corresponding quantized value standard interval, counting the qualified number of each decision index in each energy efficiency level, and recording the qualified number as the qualified number of the decision indexes, wherein the qualified judgment method comprises the steps of judging that the quantized correction values of the decision indexes are in the corresponding quantized value standard intervals, judging the qualified number, and selecting the energy efficiency level with the largest qualified number of the decision indexes as the set energy efficiency level of an intelligent computation center.
3. The modular configuration method for intelligent computing center infrastructure based on energy efficiency grading according to claim 2, wherein the method for obtaining the quantization correction value of each decision index is as follows: The method comprises the steps of obtaining environment influence information of an intelligent computation center, carrying out feature extraction on the environment influence information to obtain parameter values of all environment influence items, comparing the parameter values of all environment influence items with standard parameter values of corresponding environment influence items in a standard environment to obtain deviation absolute values of all environment influence items, carrying out normalization processing on the deviation absolute values to obtain deviation coefficients of all environment influence items, carrying out weighting accumulation on the deviation coefficients of all environment influence items to obtain influence factors, and multiplying the influence factors by quantized values of all decision indexes after adding one to obtain quantized corrected values of all decision indexes.
4. The energy efficiency-based hierarchical intelligent computing center infrastructure modular configuration method of claim 1, wherein the method of determining the additional budget power consumption of the facility is: The power consumption of each IT device in the intelligent computing center is obtained, the power consumption of all the IT devices is added to obtain the total power consumption of the devices, the PUE corresponding to the set energy efficiency level of the intelligent computing center is obtained, the PUE is subtracted by a constant and then multiplied by the total power consumption of the devices, and the additional budget power consumption of the facility is obtained.
5. The energy efficiency ranking based intelligent computing center infrastructure modular configuration method of claim 4 wherein the budget power consumption goal acquisition method for each infrastructure module is: Acquiring typical energy consumption duty ratio of each infrastructure module in an industry report, and setting basic weight for each infrastructure module; The method comprises the steps of collecting relevant energy consumption associated data of each infrastructure module, carrying out feature extraction on the energy consumption associated data to obtain actual parameter values of each associated feature, dividing the actual parameter values by industry reference values to obtain correction dimension coefficients of each associated feature, calculating weight coefficients of each associated feature, accumulating each associated feature by combining the weight coefficients and the correction dimension coefficients to obtain comprehensive correction coefficients of each infrastructure module, multiplying the comprehensive correction coefficients of each infrastructure module by corresponding basic weights to obtain primary correction weights of each infrastructure module, carrying out normalization processing on the primary correction weights to obtain final weights of each infrastructure module, and multiplying budget power consumption near the infrastructure by the final weights of each infrastructure module to obtain budget power consumption targets of each infrastructure module.
6. The energy efficiency ranking based intelligent computing center infrastructure modular configuration method of claim 5, wherein the weight coefficient calculation method of each associated feature is as follows: Acquiring historical energy consumption associated data of each infrastructure module under different established energy efficiency levels, extracting features of the historical energy consumption associated data to obtain historical parameter values of each associated feature under each established energy efficiency level; Combining the association features pairwise to obtain a plurality of association feature combinations, calculating the correlation coefficient of each association feature combination based on the historical quantization parameter value of each association feature to obtain the correlation coefficient of each association feature combination under each preset energy efficiency level; calculating the sum of each row in the feature matrix to obtain a first weight value of each associated feature, calculating the sum of each column in the feature matrix to obtain a second weight value of the associated feature, adding the first weight value and the second weight value to obtain the weight value of each associated feature, adding the weight values of each associated feature to obtain a total weight value, and calculating the ratio of the weight value of each associated feature to the total weight value to obtain the weight coefficient of each associated feature.
7. The energy efficiency ranking based intelligent computing center infrastructure modular configuration method of claim 1 wherein the method of selecting the optimal configuration strategy for each infrastructure module is: Combining the configuration options of each key step according to the step sequence according to the configurable options of each key step in the infrastructure module to obtain a plurality of configuration strategies; Calculating the adaptation scores of all the configuration strategies, sequencing all the configuration strategies according to the adaptation scores from large to small to obtain a configuration strategy sequencing table, and selecting the configuration strategy with the highest ranking in the configuration strategy sequencing table for configuration.
8. The energy efficiency ranking based intelligent computing center infrastructure modular configuration method of claim 7 wherein the method of calculating the fit score for each configuration policy is: The method comprises the steps of obtaining positive influence factor data and negative influence factor data of each configuration strategy, and carrying out normalization processing on the positive influence factor data and the negative influence factor data to obtain positive influence values of each positive influence factor and negative influence values of each negative influence factor; And adding the weights of the positive influence values to obtain a first influence value, adding the weights of the negative influence values to obtain a second influence value, and dividing the first influence value by the second influence value to obtain the adaptation score of each configuration strategy.
9. The energy efficiency ranking based intelligent computing center infrastructure modular configuration method of claim 1, wherein the method for determining whether the selected optimal configuration strategy is appropriate is: In simulation, energy consumption values of all infrastructure modules in a time sequence are obtained, standard deviation of the energy consumption values is calculated, an index function value is built by taking a natural constant e as a base and the standard deviation as an index, meanwhile, a real-time energy consumption value change function is built based on the energy consumption values obtained in the time sequence, an ideal energy consumption value change function in the time sequence is pre-built, difference integral of the ideal energy consumption value change function and the real-time energy consumption value change function in the time sequence is calculated, the difference integral is normalized to obtain a deviation coefficient, the deviation coefficient is multiplied by the index function value to obtain a judgment coefficient, when the judgment coefficient is larger than a preset judgment coefficient threshold, the selected optimal configuration strategy is judged to be unsuitable, and when the judgment coefficient is smaller than or equal to the preset judgment coefficient threshold, the selected optimal configuration strategy is judged to be suitable.
10. The energy efficiency ranking based intelligent computing center infrastructure modular configuration method of claim 9, wherein the method of performing the corresponding response based on the analysis result is: when the selected optimal configuration strategy is judged to be proper, the configuration strategy is selected for configuration; When the selected optimal configuration strategy is not suitable, performing iterative simulation on each configuration strategy according to the arrangement sequence in the configuration strategy ranking table until the proper configuration strategy is selected.

Description

Intelligent computation center infrastructure modularized configuration method based on energy efficiency grading Technical Field The invention relates to the technical field of resource allocation, in particular to an intelligent computing center infrastructure modularized allocation method based on energy efficiency grading. Background The intelligent computing center is used for efficiently supporting open sharing of data, intelligent ecological construction and industrial innovation aggregation through production, aggregation, scheduling and release of computing power, and is capable of promoting AI industrialization, industrial AI and government control intellectualization. The intelligent computing system is not only a physical carrier for intelligent computing, but also a novel infrastructure for integrating computing power, data and algorithms, so that the AI technology is promoted to be practically applied from a laboratory, and the commercialization process of a large model is accelerated. With the rapid development of the fields of artificial intelligence, big data, scientific computing and the like, an intelligent computing center is taken as a key digital infrastructure, and the scale and the energy consumption of the intelligent computing center are in explosive growth, so that the intelligent computing center is particularly important to select a proper configuration method for each infrastructure module of the intelligent computing center. In addition, when the existing intelligent computation center infrastructure module is subjected to budget division, the existing intelligent computation center infrastructure module is not strongly bound with a preset energy efficiency level of the intelligent computation center, and the intelligent computation center infrastructure module is divided only according to the rated power consumption of IT equipment as a single basis, so that the problems of substandard energy efficiency or resource redundancy waste are easy to occur. In view of this, the present invention proposes an intelligent computing center infrastructure modular configuration method based on energy efficiency grading to solve the above-mentioned problems. Disclosure of Invention In order to overcome the defects in the prior art and achieve the purposes, the invention provides the following technical scheme: The intelligent computing center infrastructure modularized configuration method based on energy efficiency grading is characterized by comprising the following steps of: step one, collecting key information of an intelligent computing center, and determining a set energy efficiency level of the intelligent computing center based on analysis of the collected key information; determining additional budget power consumption of the facility according to the predetermined energy efficiency level, and distributing the additional budget power consumption of the facility to each infrastructure module of the intelligent computation center to obtain budget power consumption targets of each infrastructure module; Generating a plurality of configuration strategies according to budget power consumption targets of the infrastructure modules, and selecting an optimal configuration strategy for the infrastructure modules based on the adaptation scores of the configuration strategies; step four, performing simulation on the intelligent computation center based on the optimal configuration strategy, obtaining energy consumption values of all infrastructure modules in the simulation, analyzing based on the energy consumption values, and judging whether the selected optimal configuration strategy is proper or not; and fifthly, executing corresponding response according to the analysis result. Further, the method for determining the given energy efficiency level of the intelligent computing center comprises the following steps: Extracting the characteristics of the collected key information of the intelligent computation center to obtain characteristic labels of a plurality of decision indexes, obtaining quantized values of all the decision indexes based on the characteristic labels of the decision indexes, obtaining environmental impact information of the intelligent computation center, generating impact factors based on the environmental impact information, and correcting the quantized values to obtain quantized correction values of all the decision indexes; the method comprises the steps of presetting a plurality of energy efficiency levels, setting quantized value standard intervals of each decision index under each energy efficiency level, comparing quantized correction values of each decision index with each corresponding quantized value standard interval, counting the qualified number of each decision index in each energy efficiency level, and recording the qualified number as the qualified number of the decision indexes, wherein the qualified judgment method comprises the steps of judging that t