Data center energy consumption simulation optimization system based on AI
By adopting an AI-based energy consumption simulation optimization system in the data center, using data twin models and expert evaluation methods, the problem of excessive resource power consumption during data center energy consumption optimization and update is solved, and the energy consumption is significantly reduced and Savings of operating costs.
Patent Information
- Application Number
- CN202411453872.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-17
- Publication Date
- 2025-02-11
AI Technical Summary
In the prior art, when the energy consumption optimization method of the data center is updated again, it will still lead to the problem of excessive resource power consumption.
An AI-based data center energy consumption simulation optimization system is adopted, which includes a data acquisition module, a data generation module, a model construction module, an energy consumption adjustment module and an evaluation optimization module. The IoT platform collects equipment data and environmental data, preprocesses energy consumption data and generates multiple data scenarios, builds a data twin model, adjusts model parameters to optimize energy consumption, and continuously optimizes through expert evaluation and improvement solutions.
It significantly reduces the energy consumption of the data center, reduces operating costs, and provides a basis for decision-making through quantitative adjustment results, improving the flexibility and adaptability of energy consumption optimization.
Smart Images
Figure CN119416467A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of energy consumption optimization and relates to Internet of Things technology, and specifically is an AI-based data center energy consumption simulation optimization system. Background Art
[0002] Data centers consume a lot of energy, and their energy costs account for a significant portion of their operating costs. By optimizing energy consumption, energy costs can be reduced and operational efficiency can be improved. Effective management and planning of data center energy needs can meet growing computing needs and promote data center scalability and business development. For many servers in a data center, the power consumption of different servers varies. In order to avoid the problem of high energy consumption caused by overloading of some servers, it is necessary to save and reduce the resource consumption of the data center.
[0003] The invention patent with application number CN2023113600972 discloses a method and system for optimizing energy consumption of a cloud computing data center. The invention obtains the operating data of the data center server and performs dimensionality reduction to obtain principal component data, combines several principal component data to obtain sample clusters, and obtains the degree of discreteness of the sample clusters according to the difference in the average values of the sample clusters and the principal component data. The sample clusters are divided according to the difference in the degree of discreteness to obtain several random trees, and the random trees are combined to obtain an isolated forest. The energy consumption of the data center is optimized through the isolated forest. When optimizing the energy consumption of the data center, this method uses the data set corresponding to the server for analysis, one server corresponds to one data set, and the data is no longer used after use, resulting in low data utilization at the current stage. If the current optimization strategy of the data center needs to be updated again, new operating data needs to be analyzed to obtain the updated optimization strategy. Therefore, the problem of excessive resource power consumption will still occur when the analysis and update are performed again.
[0004] The present invention provides an AI-based data center energy consumption simulation and optimization system to solve the above technical problems. Summary of the invention
[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an AI-based data center energy consumption simulation optimization system to solve the technical problem in the prior art of low data utilization due to the lack of data reuse.
[0006] To achieve the above-mentioned purpose, the first aspect of the present invention provides an AI-based data center energy consumption simulation optimization system, including: a data acquisition module, a data generation module, a model building module, an energy consumption adjustment module and an evaluation optimization module;
[0007] Data acquisition module: used to collect data center equipment data, corresponding energy consumption data and environmental data;
[0008] Data generation module: used to pre-process energy consumption data to obtain standard energy consumption data; and to obtain generated data based on the standard energy consumption data;
[0009] Model building module: used to build a data twin model of simulated data center operation based on equipment data and environmental data;
[0010] Energy consumption adjustment module: used to input the generated data into the data twin model to obtain the generated energy consumption, randomly adopt the adjustment plan to adjust the parameters of the data twin model to obtain the adjusted energy consumption; the adjustment effect is obtained based on the comparison result between the generated energy consumption and the adjusted energy consumption;
[0011] Evaluation and optimization module: used by experts to evaluate and improve the adjustment effects and adjustment plans, and continuously optimize the data twin model based on the improvement plans.
[0012] Preferably, the preprocessing of the energy consumption data to obtain standard energy consumption data includes:
[0013] Extracting energy consumption data; the energy consumption data includes device type and corresponding current device energy consumption;
[0014] The energy consumption data is cleaned, outliers and missing values are detected, and the outliers and missing values are supplemented and corrected by data interpolation methods, and normalized. The energy consumption data after preprocessing is marked as standard energy consumption data.
[0015] Preferably, the generating data based on the standard energy consumption data includes:
[0016] Extract standard energy consumption data;
[0017] A jitter target and a probability distribution are selected, and jitter parameters are set. Time jitter is constructed according to the jitter target, the probability distribution and the jitter parameters. The time jitter is added to the standard energy consumption data to generate a number of generated data based on the standard energy consumption data.
[0018] Preferably, the data twin model for simulating data center operation based on equipment data and environment data is constructed, including:
[0019] Extracting equipment data and environmental data; the equipment data includes equipment model parameters and equipment operation parameters; the environmental data includes temperature, humidity and airflow of the data center;
[0020] The initial three-dimensional model is constructed through three-dimensional modeling software, the working environment of the initial three-dimensional model is set according to the environmental data, and the initial three-dimensional model is supplemented based on the equipment data so that it can simulate the actual operation of the data center and obtain a data twin model.
[0021] Preferably, obtaining the adjustment effect based on the comparison result between the generated energy consumption and the adjusted energy consumption includes:
[0022] A1: Extract generation energy consumption and adjust energy consumption;
[0023] A2: Determine whether the generation energy consumption is greater than the adjustment energy consumption; if yes, mark the adjustment effect as positive optimization; if no, mark the adjustment effect as negative optimization and jump to A3;
[0024] A3: Mark the adjustment plan as a reverse plan, record the reverse plan in the database and no longer use it, select other adjustment plans and return to A2 to continue comparison.
[0025] Preferably, the continuous optimization of the data twin model according to the improvement scheme includes:
[0026] Extracting an improvement plan; the improvement plan is an improvement made by experts on the adjustment plan based on experience, so that the data twin model can better save energy consumption;
[0027] Apply the improvement plan to the data twin model to obtain the improved energy consumption; determine whether the improved energy consumption is greater than the adjusted energy consumption; if yes, mark the corresponding improvement plan as a reverse improvement plan; if no, mark the corresponding improvement plan as a forward improvement plan;
[0028] Apply the forward improvement plan to the data center and record it, continue to generate adjustment plans to obtain subsequent adjusted energy consumption, and compare the subsequent adjusted energy consumption with the improved energy consumption, and use the comparison results to optimize the data twin model;
[0029] Set a supplementary threshold; determine whether the number of generated data that has not been adjusted or improved is less than or equal to the supplementary threshold; if so, continue to generate supplementary generated data for adjustment and improvement through time jitter; if not, continue to optimize the data twin model by generating data.
[0030] Compared with the prior art, the present invention has the following beneficial effects:
[0031] 1. The present invention collects equipment data, energy consumption data and environmental data of a data center through an Internet of Things platform, and pre-processes the energy consumption data to obtain standard energy consumption data, which can ensure the integrity and continuity of the data set, and the data interpolation method can be more close to the real data, making the result more accurate; time jitter is added to the standard energy consumption data to generate a number of generated data, and a data twin model that can simulate the data center is constructed based on the equipment data and environmental data, and the parameters in the data twin model are adjusted to obtain the adjusted energy consumption, and the generated energy consumption is compared with the adjusted energy consumption, and the adjustment effect is obtained according to the comparison result, and by optimizing the parameter setting of the twin model, the energy consumption is significantly reduced, and the operating cost is reduced, and the adjustment result can also provide a quantitative basis for decision-making; experts evaluate the adjustment plan according to the adjustment effect and propose an improvement plan, optimize the data twin model according to the improvement plan, and obtain the improved energy consumption, and continuously optimize the data twin model according to the comparison result between the improved energy consumption and the adjusted energy consumption, and apply it to the data center, through detailed evaluation and comparison, the specific effect of each adjustment plan is clarified, and it can better adapt to the new adjustment plan, improve its flexibility and save costs.
[0032] 2. When the present invention continuously optimizes the data twin model through the improvement scheme, the improved energy consumption is obtained by applying the improvement scheme, and the improved energy consumption is compared with the adjusted energy consumption to obtain the improvement effect of the improvement scheme. The improvement effect is determined based on whether to record and apply it, which can further improve the energy consumption optimization effect and achieve significant energy consumption reduction and cost savings. A supplementary threshold is set, and the number of generated data that has not been adjusted and improved is compared with the supplementary threshold. If the number of generated data that has not been adjusted and improved is less than the supplementary threshold, it means that the number of remaining data for optimization is low. The optimization is continued by generating supplementary generated data through a scheme of adding time jitter, ensuring that the data volume is sufficient and avoiding data bottlenecks in the optimization process. This can increase the diversity and richness of the data set, improve the generalization ability of the model, and improve the energy consumption optimization effect of the data center. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0034] Figure 1 The following is a flowchart of an embodiment of the present invention.
[0035] Figure 2 It is a system composition diagram of the present invention.
[0036] Figure 3A complete flow chart of obtaining the adjustment effect of an adjustment scheme in one embodiment of the present invention. DETAILED DESCRIPTION
[0037] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0038] See also Figure 1-Figure 3 , the first aspect of the present invention provides an AI-based data center energy consumption simulation optimization system, including: a data acquisition module, a data generation module, a model building module, an energy consumption adjustment module and an evaluation optimization module;
[0039] Data acquisition module: used to collect data center equipment data, corresponding energy consumption data and environmental data.
[0040] Exemplarily, the device data and corresponding energy consumption data of all devices in the data center are collected through the Internet of Things platform, and the environmental data of the data center at that time is collected through the temperature sensors, humidity sensors and wind speed sensors installed in the data center.
[0041] Data generation module: used to pre-process energy consumption data to obtain standard energy consumption data; and to obtain generated data based on the standard energy consumption data.
[0042] Exemplarily, the energy consumption data is clarified, outliers and missing values in the energy consumption data are detected, interpolation data identical to the non-missing data in the missing values are generated by the radial basis function interpolation method, the missing values are replaced by the interpolation data, abnormal interpolation data identical to the non-abnormal data in the abnormal values are generated by the radial basis function interpolation method, the abnormal interpolation data replaces the abnormal values, and the supplementary correction of the abnormal values is completed; the supplementary and corrected energy consumption data is marked as supplementary energy consumption data, the supplementary energy consumption data is normalized, and the supplementary energy consumption data after processing is marked as standard energy consumption data.
[0043] In this embodiment, the jitter target is set to simulated generated data, the probability distribution is selected as normal distribution, the jitter parameter is set to variance, the range of normal distribution is limited to 0 to 1, time jitter is constructed, a normal distribution model is selected, standard energy consumption data is input into the normal distribution model as basic data, and time jitter is added to the normal distribution model to obtain some generated data.
[0044] The initial three-dimensional model of the data center is constructed using CAD software. The environmental data obtained in this embodiment are a temperature of 25 degrees, a humidity of 45%, and a wind speed of 2 meters per second. The surrounding environment data of the initial three-dimensional model is set to be the same as the environmental data, and the equipment data is input into the initial three-dimensional model so that it can simulate the actual operation of the data center to obtain a data twin model.
[0045] Energy consumption adjustment module: used to input the generated data into the data twin model to obtain the generated energy consumption, randomly adopt adjustment plans to adjust the parameters of the data twin model to obtain the adjusted energy consumption; and obtain the adjustment effect based on the comparison results of the generated energy consumption and the adjusted energy consumption.
[0046] Exemplarily, a set of generated data is input into the data twin model to run the data twin model, and the total energy consumption of all devices in the data twin model is counted and marked as generated energy consumption; the device parameters in the data twin model are randomly adjusted, and the total energy consumption of the adjusted data twin model is obtained and marked as adjusted energy consumption.
[0047] Determine whether the generated energy consumption is greater than the adjusted energy consumption. In this embodiment, the generated energy consumption is 500 kilowatts. After adjusting and reducing the chilled water flow rate in the cooling system, the adjusted energy consumption is 480 kilowatts. It is determined that the generated energy consumption is greater than the adjusted energy consumption, indicating that this adjustment has reduced energy consumption, and the adjustment effect is marked as positive optimization.
[0048] In some other embodiments, if the adjusted energy consumption after adjustment is greater than or equal to the generated energy consumption, it means that this adjustment fails to optimize the energy consumption. The current adjustment plan is marked as a reverse plan, and this adjustment plan will no longer be used in subsequent adjustments. The data twin model will continue to be adjusted, and the adjustment effects will be compared.
[0049] It should be noted that when adjusting the data twin model, you can obtain the adjustment direction based on expert advice or experience and make adjustments based on the adjustment direction.
[0050] Evaluation and optimization module: used by experts to evaluate and improve the adjustment effects and adjustment plans, and continuously optimize the data twin model based on the improvement plans.
[0051] Exemplarily, after obtaining the forward optimized adjustment plan, the expert team analyzed the adjustment plan and evaluated it as feasible. The improvement plan was to increase the initial flow rate of chilled water; the improvement plan was applied to increase the initial flow rate of chilled water in the cooling system of the data twin model, and further reduce the chilled water flow rate. The total energy consumption of the data twin model after applying the improvement plan was counted and marked as improved energy consumption; it was determined whether the improved energy consumption was greater than the adjusted energy consumption. In this embodiment, the improved energy consumption was 420 kilowatts, and it was determined that the improved energy consumption was less than the adjusted energy consumption, indicating that the improvement plan was more energy-saving than the adjustment plan, and the improvement plan was marked as a forward improvement plan; the forward improvement plan was applied to the cooling system of the data center, and the forward improvement plan was recorded. The forward improvement plan continued to be adjusted in the direction of the improvement plan and applied to the data twin model for continued optimization.
[0052] In this embodiment, the supplementary threshold is set to 5, and it is determined whether the number of generated data that has not been adjusted or improved is less than or equal to the supplementary threshold. If it is determined that the current number of generated data that has not been adjusted and improved is 9, which is greater than the supplementary threshold, the data twin model continues to be optimized by generating data.
[0053] In another embodiment, if the number of generated data that has not been adjusted and improved is less than the supplementary threshold, the data twin model is continuously optimized by adding time jitter to the generated data to generate supplementary generated data, and the range of the time jitter is between 0 and 1. The supplementary generated data used to optimize the model can be supplemented in time, and the data twin model can be continuously optimized to improve the optimization effect.
[0054] Part of the data in the above formula is calculated by removing the dimension and taking its numerical value. The formula is a formula closest to the actual situation obtained by software simulation of a large amount of collected data; the preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained through simulation of a large amount of data.
[0055] Working principle of the present invention:
[0056] The present invention collects equipment data, energy consumption data and environmental data of a data center through an Internet of Things platform, pre-processes the energy consumption data to obtain standard energy consumption data, and adds time jitter to the standard energy consumption data to obtain generated data; a data twin model capable of simulating the operation of a data center is constructed based on the equipment data and the environmental data, the generated data is input into the data twin model to obtain generated energy consumption, the parameters in the data twin model are adjusted to obtain adjusted energy consumption and an adjustment plan, the generated energy consumption is compared with the adjusted energy consumption to obtain a comparison result, and an adjustment effect of the adjustment plan is obtained according to the comparison result. If the adjustment effect of the adjustment plan is positive optimization, the adjustment plan is evaluated by an expert and an improvement plan is proposed, the improvement plan is applied to the data twin model to obtain improved energy consumption, the improved energy consumption is compared with the adjusted energy consumption, and the improvement effect of the improvement plan is obtained according to the comparison result, and the adjustment plan is continuously generated; a supplementary threshold is set, and the number of generated data that has not been adjusted and improved is compared with the supplementary threshold. If the number of generated data is less than the supplementary threshold, supplementary generated data is continuously generated through time jitter for adjustment and optimization, so as to achieve continuous optimization of the energy consumption of the data center.
[0057] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. An AI-based data center energy consumption simulation and optimization system, characterized in that: include: Data acquisition module, data generation module, model building module, energy consumption adjustment module and evaluation and optimization module; Data acquisition module: used to collect data center equipment data, corresponding energy consumption data and environmental data; Data generation module: used to pre-process energy consumption data to obtain standard energy consumption data; and to obtain generated data based on the standard energy consumption data; Model building module: used to build a data twin model of simulated data center operation based on equipment data and environmental data; Energy consumption adjustment module: used to input the generated data into the data twin model to obtain the generated energy consumption, randomly adopt the adjustment plan to adjust the parameters of the data twin model to obtain the adjusted energy consumption; the adjustment effect is obtained based on the comparison result between the generated energy consumption and the adjusted energy consumption; Evaluation and optimization module: used by experts to evaluate and improve the adjustment effects and adjustment plans, and continuously optimize the data twin model based on the improvement plans.
2. According to claim 1, the AI-based data center energy consumption simulation and optimization system is characterized in that: The preprocessing of the energy consumption data to obtain standard energy consumption data includes: Extracting energy consumption data; the energy consumption data includes device type and corresponding current device energy consumption; The energy consumption data is cleaned, outliers and missing values are detected, and the outliers and missing values are supplemented and corrected by data interpolation methods, and normalized. The energy consumption data after preprocessing is marked as standard energy consumption data.
3. According to claim 1, the AI-based data center energy consumption simulation and optimization system is characterized in that: The generating data based on the standard energy consumption data includes: Extract standard energy consumption data; A jitter target and a probability distribution are selected, and jitter parameters are set. Time jitter is constructed according to the jitter target, the probability distribution and the jitter parameters. The time jitter is added to the standard energy consumption data to generate a number of generated data based on the standard energy consumption data.
4. According to claim 1, the AI-based data center energy consumption simulation and optimization system is characterized in that: The data twin model for simulating data center operation based on device data and environment data includes: Extracting equipment data and environmental data; the equipment data includes equipment model parameters and equipment operation parameters; the environmental data includes temperature, humidity and airflow of the data center; The initial three-dimensional model is constructed through three-dimensional modeling software, the working environment of the initial three-dimensional model is set according to the environmental data, and the initial three-dimensional model is supplemented based on the equipment data so that it can simulate the actual operation of the data center and obtain a data twin model.
5. According to claim 1, the AI-based data center energy consumption simulation and optimization system is characterized in that: The obtaining of the adjustment effect based on the comparison result between the generated energy consumption and the adjusted energy consumption includes: A1: Extract generation energy consumption and adjust energy consumption; A2: Determine whether the generation energy consumption is greater than the adjustment energy consumption; if yes, mark the adjustment effect as positive optimization; if no, mark the adjustment effect as negative optimization and jump to A3; A3: Mark the adjustment plan as a reverse plan, record the reverse plan in the database and no longer use it, select other adjustment plans and return to A2 to continue comparison.
6. The AI-based data center energy consumption simulation and optimization system according to claim 1 is characterized in that: The continuous optimization of the data twin model according to the improvement plan includes: Extracting an improvement plan; the improvement plan is an improvement made by experts on the adjustment plan based on experience, so that the data twin model can better save energy consumption; Apply the improvement plan to the data twin model to obtain the improved energy consumption; determine whether the improved energy consumption is greater than the adjusted energy consumption; if yes, mark the corresponding improvement plan as a reverse improvement plan; if no, mark the corresponding improvement plan as a forward improvement plan; Apply the forward improvement plan to the data center and record it, continue to generate adjustment plans to obtain subsequent adjusted energy consumption, and compare the subsequent adjusted energy consumption with the improved energy consumption, and use the comparison results to optimize the data twin model; Set a supplementary threshold; determine whether the number of generated data that has not been adjusted or improved is less than or equal to the supplementary threshold; if so, continue to generate supplementary generated data for adjustment and improvement through time jitter; if not, continue to optimize the data twin model by generating data.