Intelligent energy consumption management system and method for data center

Abstract

The invention discloses a data center intelligent energy consumption management system and method. The system is composed of a data acquisition module, an intelligent management center module and a dynamic adjustment module. According to the method, a weighted feedback mechanism of time dimension and anomaly detection is introduced into the dynamic regulation module, so that a regulation strategy is not a simple threshold response any more, but is a dynamic optimization process with learning and adaptive capabilities. And in combination with a high-security-level data storage scheme of the intelligent management center module, the reliability and compliance of the full life cycle of the energy consumption data are ensured, and finally, the continuous optimization of the energy efficiency of the data center and the remarkable reduction of the operation cost are realized.

Description

Technical Field

[0001] This invention relates to the field of energy management technology, and in particular to a smart energy management system and method for data centers. Background Technology

[0002] With the rapid development of cloud computing, big data, and artificial intelligence technologies, the scale and energy consumption of data centers continue to surge. Statistics show that the electricity cost of a data center over three years can even be equivalent to its construction cost, with cooling systems accounting for a significant portion of this energy consumption. To address the challenge of high energy consumption, the industry generally adopts Power Usage Effectiveness (PUE) as a core energy efficiency indicator and is attempting to optimize cooling systems through artificial intelligence technology.

[0003] Existing energy management solutions often focus on single-level monitoring or regulation. For example, some systems dynamically adjust energy configuration strategies by monitoring trends in equipment energy consumption and carbon dioxide concentration; others intervene based on PUE values, identifying abnormal equipment groups and scheduling tasks or adjusting cooling capacity; still others focus on managing edge energy consumption and correcting energy consumption data based on the manager's historical adjustment effects. Furthermore, some advanced technologies are beginning to combine thermodynamic simulation and phase change materials to optimize cooling paths.

[0004] However, existing technologies still have the following shortcomings: 1) Data collection is mostly limited to energy consumption itself, lacking in-depth correlation modeling with time periods, equipment operating status, and environmental parameters; 2) The security and integrity of data storage are not adequately considered, making it difficult to meet the requirements of regulations such as the Data Security Law for data anti-tampering and anti-leakage; 3) Control strategies are often based on fixed rules or single indicators, lacking a comprehensive weighted feedback model that integrates real-time data, historical baselines, time characteristics, and anomaly levels, resulting in limited adjustment accuracy and adaptive capabilities; 4) The coupling between various modules of the system is loose, failing to form a closed-loop control system from accurate collection and secure storage to intelligent adjustment.

[0005] Therefore, there is an urgent need to design an intelligent energy management system for data centers to achieve closed-loop management of the entire process from accurate data collection and secure storage to intelligent dynamic adjustment. Summary of the Invention

[0006] The purpose of this invention is to provide a smart energy management system and method for data centers to solve the problems existing in the prior art.

[0007] To achieve the above objectives, the present invention provides the following solution: The present invention provides a data center intelligent energy consumption management system. A data center intelligent energy consumption management system includes:

[0008] The data acquisition module is used to collect the operating parameters and energy consumption data of each component of the data center at a preset sampling frequency, and to preliminarily calculate the energy consumption values ​​of each component and the total energy consumption.

[0009] The intelligent management center module is communicatively connected to the data acquisition module. It is used to clean and classify the received operating parameters and energy consumption data, and to store the data in the central database using a multi-layer encryption storage strategy.

[0010] The dynamic adjustment module interacts with the intelligent management center module to calculate its own energy consumption index and abnormal energy consumption index based on historical and real-time data, and generates control instructions based on a weighted feedback model to dynamically adjust the energy consumption of the data center.

[0011] The operating parameters of each component of the data center are divided into IT equipment energy consumption models, cooling system energy consumption models, and power distribution system energy consumption models according to the type of equipment.

[0012] The basic parameters of the IT equipment energy consumption model include the number of IT equipment servers, workload, and peak and standby power; the basic parameters of the cooling system energy consumption model include the air supply temperature of the cooling equipment; and the basic parameters of the power distribution system energy consumption model include the computer room entrance temperature and ambient temperature and humidity.

[0013] The multi-layered encryption storage strategy of the intelligent management center module includes application layer encryption, transparent database encryption, and file system encryption; the application layer encryption is based on calling external encryption devices or interfaces; the transparent database encryption is based on encrypting the stored data in the database; and the file system encryption is based on encrypting the storage medium where the database files are located.

[0014] The application layer encryption uses a UKey-based encryption device and combines SM2, SM4 and chaotic encryption algorithms.

[0015] The weighted feedback model in the dynamic adjustment module calculates the comprehensive control intensity based on the time period weight factor, its own energy consumption index, and the abnormal energy consumption index.

[0016] The dynamic adjustment module includes IT load scheduling, cooling system optimization, and equipment start-up and shutdown management.

[0017] A data center intelligent energy consumption management method includes a data center intelligent energy consumption management system, comprising the following steps: collecting energy consumption data of various sub-equipment in the data center and calculating the sub-item and total energy consumption through a data acquisition module; cleaning, classifying and encrypting the data through an intelligent management center module; and calculating the energy consumption index through a dynamic adjustment module, and judging and executing energy consumption regulation based on a weighted feedback model.

[0018] The data encryption storage specifically includes performing application-layer encryption on core indicator data; performing transparent data encryption on database storage; and performing disk encryption on database file storage volumes.

[0019] The dynamic adjustment module calculates the comprehensive control intensity based on the time period weighting factor, its own energy consumption index, and the abnormal energy consumption index; when the control intensity exceeds the threshold, it coordinates and executes one or more operations among IT load scheduling, cooling parameter optimization, and equipment start-up and shutdown control.

[0020] This invention discloses the following technical effects: By introducing a weighted feedback mechanism that combines a time dimension and anomaly detection, the control strategy is no longer a simple threshold response, but a dynamic optimization process with learning and adaptive capabilities. Combined with a high-security data storage solution, it ensures the reliability and compliance of energy consumption data throughout its entire lifecycle, ultimately achieving continuous optimization of data center energy efficiency and a significant reduction in operating costs. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the modules of the intelligent energy consumption management system for data centers of the present invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] In one specific embodiment of the present invention, a data center intelligent energy consumption management system is provided, comprising a data acquisition module, an intelligent management center module, and a dynamic adjustment module; the data acquisition module is used to acquire real-time data of various energy consumption items in the data center based on a preset sampling frequency, and to calculate the energy consumption values ​​of each item and the total energy consumption based on the energy consumption mathematical model of IT equipment, cooling system, and power distribution system.

[0026] Furthermore, the collected data includes at least the number of IT equipment servers, workload, peak and standby power, cooling equipment supply air temperature, computer room entrance temperature, and ambient temperature and humidity.

[0027] In one embodiment of the present invention, the data acquisition module consists of a sensor network, smart meters and device agents deployed in various locations of the data center; it collects raw energy consumption and environmental data at a frequency of 15 minutes to 1 hour through communication methods such as 5G or industrial Ethernet, and performs preliminary calculations of individual energy consumption items.

[0028] The intelligent management center module receives and continuously records data uploaded by the data acquisition module. This module cleans and categorizes the data (e.g., by IT equipment, cooling, power distribution, lighting, etc.), and employs a multi-layered encryption storage strategy to uniformly store the processed data in the central database to prevent data loss and tampering. The encryption strategy combines application-layer encryption, transparent database encryption, and disk encryption.

[0029] In one embodiment of the present invention, the intelligent management center module, serving as the control brain of the entire system, is deployed on a dedicated server. It includes a data cleaning engine, a classification and storage engine, an encryption engine, and a strategy analysis engine; it is used to ultimately aggregate all collected data, process it, and then store it in the central database.

[0030] The dynamic adjustment module is the intelligent core of the system. Based on historical and real-time data stored in the intelligent management center module, it calculates two key indices: a time-based energy consumption index and a statistical model-based abnormal energy consumption index. This module uses a weighted feedback model to combine the weights of the time period, the energy consumption index, and the abnormal energy consumption index to generate comprehensive control commands. Specific control actions include: at the IT load level, performing task migration and scheduling across racks or servers; at the cooling level, dynamically adjusting the chiller set temperature, pump frequency, and air conditioning airflow; and at the equipment level, setting timed power-on / off strategies for idle servers.

[0031] In one specific embodiment of the present invention, the data acquisition method of the IT equipment energy consumption model is to collect the power consumption P_measured, CPU utilization U_cpu, memory utilization U_mem, and network traffic F_net of each server in real time through the server's out-of-band management port (such as IPMI) or rack PDU. Simultaneously, the number of servers in use, S_i,t, is recorded.

[0032] The energy consumption model for IT equipment is based on the load-based server energy consumption model. The energy consumption P_server of a single server in time period t can be modeled as:

[0033] P_server = [P_idle + (P_peak - P_idle) * (λ / μ)] * Δt;

[0034] Where P_idle is the standby power, P_peak is the peak power, λ is the workload (which can be obtained by weighting CPU and memory utilization), μ is the unit rate of server processing load, and Δt is the sampling time interval.

[0035] Furthermore, the total energy consumption P_i,t,servers of users and cabinets during a certain period is the sum of the energy consumption of all the associated servers.

[0036] In one specific embodiment of the present invention, the data acquisition method for the energy consumption model of the refrigeration system is through temperature sensors installed at the air conditioning supply / return vents in each computer room of the data center; the specific data collected by the temperature sensors are the supply air temperature E_sup, return air temperature E_return, and hot spot temperature of the computer room. The operating power and frequency of the chiller, cooling tower, and water pump are collected through the controllers of the chiller, cooling tower, and water pump.

[0037] The energy consumption P_cool of the IT equipment energy consumption model is directly related to the heat generated by the IT equipment and is affected by the cooling efficiency (CoP).

[0038] CoP is a non-linear function of supply air temperature:

[0039] CoP(E_sup)=0.0068*E_sup^2+0.0008*E_sup+0.458;

[0040] Therefore, the cooling energy consumption is expressed as:

[0041] P_i,t,cool=(1 / CoP(E_sup,i,t))*P_i,t,servers;

[0042] At the same time, thermal balance constraints must be met to ensure that the server inlet temperature E_in does not exceed the upper limit E_red (usually set to 25°C):

[0043] E_in,i,t=E_sup,i,t+M1*(P_i,t,servers / S_i,t)≤E_red;

[0044] Where M1 is the heat distribution coefficient.

[0045] In one specific embodiment of the present invention, the data acquisition method of the power distribution system energy consumption model is to collect the branch power of UPS, PDU, lighting, fire protection and other systems through smart meters.

[0046] The power distribution system energy consumption P_network can be modeled as a linear function related to the number of servers and the load: P_i,t,network = S_i,t * l1 + λ_i,t*l2;

[0047] Where l1 and l2 are energy consumption coefficients; P_i,t,network is the energy consumption of user i's power distribution system during time period t; S_i,t represents the number of servers activated by user i during time period t; and λ_i,t is the workload of user i during time period t.

[0048] Total energy consumption calculation: The total energy consumption of the data center within time period t is P_total,t is the sum of all individual energy consumption components:

[0049] P_total,t = ​​Σ(P_i,t,servers) + Σ(P_i,t,cool) + Σ(P_i,t,network) +P_lighting + ....

[0050] In one embodiment of the present invention, the cleaning process of the intelligent management center module is to denoise the collected raw data, remove outliers (such as negative power consumption and out-of-range temperature), and fill in missing values ​​(using time series interpolation).

[0051] Furthermore, the cleaned data is categorized, that is, stored according to categories (such as IT, cooling, power distribution, lighting) and time granularity (such as 15 minutes, hours, days, months); at the same time, the "time period label" (such as weekday / holiday, peak / off-peak / valley) of each data point is recorded.

[0052] Furthermore, a multi-layered encryption storage strategy is employed to uniformly store the processed data in a central database to prevent data loss and tampering. Specifically, application-layer encryption includes: for highly sensitive core energy consumption statistics (such as the entire data center's PUE and customer electricity bill related data), encryption is performed using a UKey-based encryption application before the business system writes the data to the database. The encryption algorithm uses a nested combination of the Chinese national cryptographic standards SM2 and SM4, and randomness is optimized by introducing a chaotic encryption algorithm. This layer is primarily used to defend against data theft by internal personnel with Database Administrator (DBA) privileges.

[0053] The database transparent data encryption feature enables the database's own transparent data encryption (TDE) function for database tables or tablespaces storing detailed energy consumption data. This technology is completely transparent to upper-layer applications; data is automatically encrypted when written to disk and automatically decrypted when read, effectively preventing information leakage after data files are directly stolen.

[0054] Disk encryption, at the operating system level, performs full-disk encryption or file system encryption on the data volume or storage files containing the database. This solution can be bound to specific application processes, ensuring that data cannot be decrypted on other systems even if the disk is physically removed. Through this combination, end-to-end encryption is achieved from the application and database to the storage media, constructing a defense-in-depth system. Keys are centrally managed by an independent "encryption management center," supporting regular key changes and emergency loss reporting.

[0055] In one specific embodiment of the present invention, the intelligence of the dynamic adjustment module is mainly reflected in its decision-making model based on multi-factor weighted feedback; where the energy consumption index E_self: refers to the ratio of the actual energy consumption P_actual of a device or system group in the current time period to its historical average energy consumption P_history_avg in the same period (such as the same period in the past 4 weeks);

[0056] E_self = P_actual / P_history_avg, this index reflects the deviation from its own historical norm.

[0057] The abnormal energy consumption index (E_anomaly) is calculated based on a machine learning model (such as an isolated forest or an autoencoder) that performs real-time analysis of current operating parameters (power consumption, temperature, load, efficiency) and normalizes to the [0,1] range. The higher the score, the more the operating state deviates from the normal mode.

[0058] Furthermore, the judgment formula for the weighted feedback model is:

[0059] Adjust_Intensity =α*W_time +β* E_self +γ*E_anomaly;

[0060] Where Adjust_Intensity is the dynamically calculated control intensity; W_time is the time period weight factor; α, β, γ are adjustable weight coefficients;

[0061] Furthermore, the time-period weighting factor is a predefined weight for the energy-saving urgency of different time periods. For example, the weight is highest at 1 for peak electricity consumption periods (such as 14:00-16:00 on weekdays), followed by 0.7 for off-peak periods, and lowest at 0.3 for nighttime off-peak periods. This allows the system to prioritize energy-saving adjustments during periods of high electricity costs or high grid pressure.

[0062] Furthermore, α + β + γ = 1; and this coefficient can be optimized through training with historical data, or set by the administrator according to the data center strategy; if more attention is paid to instantaneous anomalies, γ can be increased; if more attention is paid to long-term energy efficiency optimization, β can be increased. When Adjust_Intensity exceeds the preset threshold, the corresponding level of adjustment action can be triggered.

[0063] In one specific embodiment of the present invention, the control action specifically includes IT load scheduling, cooling parameter optimization, and equipment start-up and shutdown control.

[0064] The IT load balancing operation is as follows: if both E_self and E_anomaly are high in a certain rack (second group) and low in another rack (first group), then some computing tasks from the first group are dynamically migrated to the second group to improve the resource utilization of the latter and prevent the former from overheating or overloading. The task scheduling algorithm needs to consider server heterogeneity and task dependencies.

[0065] The operating conditions for optimizing cooling parameters are as follows: If the E_anomaly index is high and a local hotspot is located, the system first attempts to adjust the air conditioning supply temperature E_sup or the fan speed. Simultaneously, a thermal model (such as a heat map generated using Kriging space interpolation) is used to predict the adjustment effect. If local optimization is insufficient, the chilled water supply temperature setpoint is increased, and the CoP value is increased as much as possible to reduce P_cool, while ensuring that E_in ≤ E_red.

[0066] The operating conditions for device start-up and shutdown are as follows: for servers with a consistently low E_self value and identified as "idle", the system does not immediately shut them down, but rather based on the W_time factor. During peak hours (high W_time), it may simply put the server into a low-power state; during off-peak hours at night (low W_time), it will decisively shut down the server in conjunction with its scheduled power-on time (such as before business begins the next morning) to save basic energy consumption.

[0067] In Embodiment 1 of this invention, the application of this system is illustrated using a large cloud data center accommodating over 100,000 servers as an example. Data Acquisition: In this data center, the data acquisition module collects data every 15 minutes through intelligent PDUs deployed in each rack and an environmental sensor network for each row of racks; IT equipment energy consumption is obtained in fine-grained terms through the server BMC interface; the cooling system obtains real-time power and operating parameters of chillers, water pumps, cooling towers, and data center air conditioners through the building automation system (BAS) interface; the power distribution system collects energy consumption data from UPS and PDUs in each area through smart meters.

[0068] Data Processing and Storage: The intelligent management center module is deployed on a highly available cluster. The cleaned data is classified and stored according to multiple dimensions such as "tenant / business department", "rack row", "cooling unit", etc. The encryption scheme is as follows: 1) For the "tenant-rack-energy consumption" associated data related to billing, application layer encryption based on UKey is used; 2) Oracle TDE encryption is enabled for the core energy consumption analysis database; 3) Hardware-level disk encryption is enabled for the storage area network (SAN) that stores database files.

[0069] Intelligent Regulation: At 14:30 on a certain working day afternoon (peak period, W_time = 1.0), the system detected that the E_self index of rack row A03 was 1.8 (far exceeding the same period in history), and the E_anomaly index was 0.9 (the thermal sensor showed local overheating).

[0070] After calculation, Adjust_Intensity = 0.3*1.0 + 0.4*1.8 + 0.3*0.9 = 1.41, far exceeding the threshold of 1.0. The system immediately starts multi-level linkage regulation:

[0071] Task Scheduling: Analysis found that the load of rack row A03 was uneven, some servers were fully loaded, and some were idle. The dynamic regulation module migrated some virtual machines from the overheated servers to the servers with lower load in the same row through the cloud management platform.

[0072] Refrigeration Optimization: At the same time, the system instructs the precision air conditioner in this area to slightly raise the supply air temperature from 18°C to 19°C and increase the opening of the air valve above the overheated point. According to the prediction of the thermodynamic model, this operation can reduce the air conditioner energy consumption in this area by about 5% while ensuring temperature safety.

[0073] Policy Evaluation: One hour later, the system re-evaluated area A03. E_self dropped to 1.2, E_anomaly dropped to 0.2, and Adjust_Intensity dropped back to 0.85, below the threshold, and the regulation was suspended. During the whole process, the total PUE of the data center was optimized from 1.45 to 1.42.

[0074] In the second embodiment of the present invention, take a small and medium-sized edge data center with 50 racks located in the city as an example. Its IT load fluctuates greatly and there are limited operation and maintenance personnel. Data Collection: Limited by cost, the collection frequency is set to once an hour. The IT energy consumption is mainly collected through the rack-level PDU; the refrigeration depends on the room-level air conditioner, and its operating status and power consumption are collected through the Modbus protocol; 5 wireless temperature sensors are arranged at key points in the computer room for the environmental temperature.

[0075] Data Processing and Storage: The intelligent management center module is deployed on a single physical server. Data classification is relatively simplified, mainly categorized as "IT Equipment," "Air Conditioning," and "Power Supply and Lighting." Encryption schemes include: 1) Application-layer AES-GCM encryption for administrator operation logs and audit data; 2) Enabling the enterprise version of MySQL's TDE function for the main business databases; 3) Enabling BitLocker encryption built into the operating system for the server system disk and data disk. Intelligent Adjustment: At 2 AM on a weekend (off-peak period, W_time=0.3), the system detected a group of server racks in Zone B used for testing, with an E_self index of 0.2 (extremely low load) and an E_anomaly index of 0.1 (normal).

[0076] Adjust_Intensity = 0.3*0.3 + 0.4*0.2 + 0.3*0.1 = 0.2, which does not reach the control threshold. However, based on the preset "device start-up and shutdown strategy" and the energy-saving strategy during off-peak hours, the system determines that this group of servers is an "idle server" that can be shut down. The dynamic adjustment module sends orderly shutdown commands to these servers through out-of-band management and synchronizes this status to the resource pool management interface. At 8:00 AM the next morning, the system automatically wakes up these servers before business demand arrives, according to the preset "scheduled power-on period". Through this fine-grained start-up and shutdown management based on time periods, the edge data center saves approximately 30% of IT infrastructure energy consumption during off-peak hours at night, without affecting daytime business operations.

[0077] In the description of this invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0078] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A data center intelligent energy management system, characterized in that, include: The data acquisition module is used to collect the operating parameters and energy consumption data of each component of the data center at a preset sampling frequency, and to preliminarily calculate the energy consumption values ​​of each component and the total energy consumption. The intelligent management center module is communicatively connected to the data acquisition module. It is used to clean and classify the received operating parameters and energy consumption data, and to store the data in the central database using a multi-layer encryption storage strategy. The dynamic adjustment module interacts with the intelligent management center module to calculate its own energy consumption index and abnormal energy consumption index based on historical and real-time data, and generates control instructions based on a weighted feedback model to dynamically adjust the energy consumption of the data center.

2. The intelligent energy management system for data centers according to claim 1, characterized in that: The operating parameters of each component of the data center are divided into IT equipment energy consumption models, cooling system energy consumption models, and power distribution system energy consumption models according to the type of equipment. The basic parameters of the IT equipment energy consumption model include the number of IT equipment servers, workload, and peak and standby power; the basic parameters of the cooling system energy consumption model include the air supply temperature of the cooling equipment; and the basic parameters of the power distribution system energy consumption model include the computer room entrance temperature and ambient temperature and humidity.

3. The intelligent energy management system for data centers according to claim 1, characterized in that: The multi-layered encryption storage strategy of the intelligent management center module includes application layer encryption, transparent database encryption, and file system encryption; the application layer encryption is based on calling external encryption devices or interfaces; the transparent database encryption is based on encrypting the stored data in the database; and the file system encryption is based on encrypting the storage medium where the database files are located.

4. The intelligent energy management system for data centers according to claim 1, characterized in that: The application layer encryption uses a UKey-based encryption device and combines SM2, SM4 and chaotic encryption algorithms.

5. The intelligent energy management system for data centers according to claim 1, characterized in that: The weighted feedback model in the dynamic adjustment module calculates the comprehensive control intensity based on the time period weight factor, its own energy consumption index, and the abnormal energy consumption index.

6. The intelligent energy management system for data centers according to claim 1, characterized in that: The dynamic adjustment module includes IT load scheduling, cooling system optimization, and equipment start-up and shutdown management.

7. A data center intelligent energy consumption management method, comprising the data center intelligent energy consumption management system as described in any one of claims 1-6, characterized in that, Includes the following steps: The data acquisition module collects energy consumption data of each component of the data center and calculates the individual and total energy consumption. The intelligent management center module cleans, classifies, and encrypts the data; the dynamic adjustment module calculates the energy consumption index and judges and executes energy consumption control based on the weighted feedback model.

8. The intelligent energy consumption management method for data centers according to claim 7, characterized in that: The data encryption storage specifically includes performing application-layer encryption on core indicator data; performing transparent data encryption on database storage; and performing disk encryption on database file storage volumes.

9. A data center intelligent energy consumption management method according to claim 7, characterized in that: The dynamic adjustment module calculates the comprehensive control intensity based on the time period weighting factor, its own energy consumption index, and the abnormal energy consumption index. When the control intensity exceeds the threshold, one or more of the following operations are coordinated and executed: IT load scheduling, cooling parameter optimization, and equipment start-up and shutdown control.

Images