Data center hybrid air conditioning system and control method thereof

Abstract

The application provides a data center composite air conditioning system and a control method thereof, which has the characteristics of high efficiency, energy saving and intelligent self-adaptation, realizes dynamic optimization switching of a refrigeration mode by integrating various refrigeration technologies such as mechanical refrigeration, evaporative cooling and free cooling, and combining data acquisition, processing, prediction model construction and intelligent control modules. The system uses historical data to build a support vector regression prediction model to accurately predict air conditioner energy consumption and cooling demand, so as to intelligently select the optimal refrigeration mode according to real-time environmental parameters, effectively overcoming the problems of single cold source and rigid control in the prior art. The method not only fully utilizes the natural cold energy in the environment and reduces the dependence on mechanical refrigeration, but also ensures the continuous and efficient operation of the system through closed-loop monitoring and model updating, significantly improves the energy efficiency ratio, reduces energy consumption, and is particularly suitable for data center environments with variable loads.

Description

Technical Field

[0001] This invention relates to the field of air conditioning technology, and in particular to a data center composite air conditioning system and its control method. Background Technology

[0002] With the rapid development of information technology and the continuous expansion of data center scale, energy consumption has become increasingly prominent. Air conditioning systems, as a critical infrastructure of data centers, account for 30%-40% of total energy consumption, with refrigeration systems being the primary energy source. Currently, data center air conditioning systems generally employ a single mechanical refrigeration method, such as vapor compression refrigeration systems, which achieve refrigeration cycles through components such as compressors, condensers, expansion valves, and evaporators. While this system is technologically mature, it relies on a single cold source, and its refrigeration efficiency drops significantly at high ambient temperatures, resulting in persistently high energy consumption. Even when outdoor ambient temperatures are low, mechanical refrigeration systems still need to operate, failing to fully utilize natural cold sources and causing energy waste. Furthermore, existing data center air conditioning systems often employ fixed threshold control modes, such as switching operating modes based on seasons or simple temperature thresholds. However, actual environmental parameters (such as temperature, humidity, and server load) change dynamically, making it difficult for fixed control strategies to match real-time load demands, leading to low system energy efficiency.

[0003] In existing technologies, some solutions introduce hybrid refrigeration systems, such as a combination of heat pipes and compression refrigeration, to reduce energy consumption. These systems utilize shared heat exchange modules for natural cooling via heat pipes during transitional seasons or winter, reducing the operating time of mechanical refrigeration. However, such systems suffer from the following problems: First, the independent combination of heat pipe and compression refrigeration systems leads to complex equipment structures, large installation space requirements, and high initial investment costs. Second, sharing heat exchange modules increases wind resistance, placing high demands on fan performance and potentially increasing energy consumption. Finally, control strategies are mostly based on seasonality or fixed temperature thresholds, failing to adapt to real-time environmental changes, such as extreme conditions like low temperatures in summer or high temperatures in winter, resulting in a mismatch between operating modes and actual loads, limiting system energy efficiency improvements. Furthermore, existing control methods lack data-driven and intelligent prediction capabilities, failing to adjust operating modes in advance, leading to response lag and further impacting energy efficiency optimization.

[0004] Therefore, there is an urgent need in this field to solve how to design an intelligent composite air conditioning system and its control method that can dynamically integrate multiple refrigeration technologies, make full use of the ambient cooling capacity, and achieve precise control through data-driven predictive models, thereby significantly reducing energy consumption and improving system efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a data center composite air conditioning system and its control method to solve the problems existing in the prior art.

[0006] To achieve the above objectives, the present invention provides the following solution:

[0007] This invention provides a data center composite air conditioning system, comprising:

[0008] The data acquisition module is used to collect historical data and operational status data from inside and outside the data center in real time, and to obtain real-time monitoring data.

[0009] The data processing module is used to clean, normalize, and extract features from the raw data transmitted from the data acquisition module.

[0010] A prediction model building module is used to build a prediction model using processed historical data and to update the model parameters periodically.

[0011] The intelligent control module is used to calculate the optimal cooling mode and generate control signals based on the prediction results provided by the prediction model construction module and the current environmental parameters.

[0012] A composite refrigeration module, which integrates mechanical refrigeration, evaporative cooling and free cooling, and automatically switches the refrigeration mode according to the control signal of the intelligent control module;

[0013] The monitoring and feedback module is used to monitor the operating status, energy consumption performance and environmental parameters of the composite refrigeration module in real time, and feed the data back to the data acquisition module to form a closed-loop control.

[0014] Preferably, the data collected by the data acquisition module includes ambient temperature, ambient humidity, server load, air conditioning energy consumption, and cooling mode, and the data acquisition frequency is set to once per minute.

[0015] Preferably, the data cleaning module uses the Z-score method to detect and remove outliers, uses linear interpolation to fill missing values, and uses the Min-Max normalization method to scale the data to the range of [0,1].

[0016] Preferably, the prediction model building module uses the support vector regression algorithm to build the prediction model, and the model formula is:

[0017] ;

[0018] in, It is the input feature vector, including ambient temperature, ambient humidity, and server load. This is a predicted output, representing air conditioning energy consumption. It is the number of support vectors. and It is a Lagrange multiplier. It is a bias term. It is a radial basis function kernel. , These are kernel parameters. It is the i-th support vector. It is the input vector to be predicted;

[0019] Model training determines hyperparameters, including penalty and kernel parameters, through grid search and cross-validation. The loss function is an ε-insensitive loss function, and the formula for calculating the loss function value is as follows:

[0020] ;

[0021] in, This is the actual energy consumption value of the air conditioner. It is the tolerance for error.

[0022] Preferably, the intelligent control module calculates the cooling demand based on the prediction results, using the following formula:

[0023] ;

[0024] in, It is a prediction of energy consumption. It is the conversion factor;

[0025] The cooling mode is dynamically switched based on ambient temperature and predicted cooling demand. The mode selection algorithm uses a scoring function, the formula of which is:

[0026] ;

[0027] in, It refers to the cooling modes, including free cooling, evaporative cooling, and mechanical refrigeration. This is the rated cooling capacity of the model. It is the ambient temperature. It is the midpoint of the applicable temperature range of the mode. It is the energy efficiency ratio of the mode. , , These are weighting coefficients. and It is a normalization factor, and the mode with the highest score is selected as the optimal mode.

[0028] The present invention also provides a control method for a data center composite air conditioning system, comprising the following steps:

[0029] S1: Data Acquisition, real-time acquisition of historical and operational status data from inside and outside the data center, including ambient temperature, ambient humidity, server load, air conditioning energy consumption, and cooling mode;

[0030] S2: Data processing involves cleaning, normalizing, and extracting features from the collected raw data. The Z-score method is used to remove outliers, and linear interpolation is used to fill missing values. The Min-Max normalization method is used to scale the data to the range of [0,1].

[0031] S3: Predictive model construction, using processed historical data to build a support vector regression predictive model; determine hyperparameters through grid search and cross-validation, and update the model regularly;

[0032] S4: Prediction, input the current feature vector into the trained prediction model to obtain the predicted energy consumption and calculate the cooling requirement;

[0033] S5: Control. Based on the predicted cooling demand and ambient temperature, a scoring function is used to calculate the score of each cooling mode. The mode with the highest score is selected as the optimal mode, and a control signal is generated to switch the cooling mode.

[0034] Preferably, in step S2, feature extraction further includes deriving temperature-humidity index features, with the following formula:

[0035] ;

[0036] in, It is the ambient temperature. It refers to ambient humidity.

[0037] Preferably, in step S3, the error tolerance of the loss function during model training is set to 0.1.

[0038] Preferably, in step S5, the weighting coefficients in the scoring function =0.5、 =0.3、 =0.2, normalization factor and The system is dynamically adjusted based on historical data ranges; the switching threshold for cooling modes is adaptively adjusted based on prediction results.

[0039] Preferably, it further includes:

[0040] S6: Closed-loop monitoring, which monitors the operating status and energy consumption performance of the cooling mode in real time, and feeds the data back to the data acquisition step to update the predictive model and optimize the control strategy.

[0041] The present invention achieves the following beneficial technical effects compared to the prior art:

[0042] This invention provides a data center integrated air conditioning system and its control method, featuring high efficiency, energy saving, and intelligent self-adaptation. By integrating multiple cooling technologies such as mechanical refrigeration, evaporative cooling, and free cooling, and combining data acquisition, processing, predictive model construction, and intelligent control modules, it achieves dynamic optimization switching of cooling modes. The system utilizes historical data to construct a support vector regression predictive model, accurately predicting air conditioning energy consumption and cooling demand. Based on real-time environmental parameters, it intelligently selects the optimal cooling mode, effectively overcoming the problems of single cold source and rigid control in existing technologies. This method not only fully utilizes the natural cooling capacity of the environment and reduces reliance on mechanical refrigeration, but also ensures continuous and efficient system operation through closed-loop monitoring and model updates, significantly improving the energy efficiency ratio and reducing energy consumption. It is particularly suitable for data center environments with variable loads. Attached Figure Description

[0043] 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.

[0044] Figure 1 This is a schematic diagram of the control relationship of a data center composite air conditioning system provided by the present invention;

[0045] Figure 2 A flowchart illustrating the control method for a data center composite air conditioning system provided by this invention. Detailed Implementation

[0046] 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.

[0047] The purpose of this invention is to provide a composite air conditioning system for data centers and its control method. By integrating multiple cooling technologies and data-driven intelligent control, it solves the problems of low energy efficiency and single cooling source in existing data center air conditioning systems. The core of the system lies in using historical data to build a predictive model, dynamically optimizing the switching of cooling modes, making full use of ambient cooling capacity, and reducing energy consumption.

[0048] 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.

[0049] Example 1:

[0050] refer to Figure 1 The system consists of six main modules, including:

[0051] The data acquisition module collects data such as ambient temperature, humidity, server load, air conditioning energy consumption, and current cooling mode every minute through a sensor network deployed inside and outside the data center. This data is transmitted to the data processing module via wired or wireless communication protocols.

[0052] The data processing module first performs data cleaning: it uses the Z-score method to detect outliers and calculates the Z-value for each data point using the following formula: in, For data point values, The mean, The standard deviation is Z. If the absolute value of the Z-value is greater than 3, it is considered an outlier and removed. For missing values, linear interpolation is used to fill them, that is, the average of the two adjacent data points is used instead. The formula is: Subsequently, the data was scaled to the [0,1] range using the Min-Max normalization method, as shown in the formula. ,in and These are the minimum and maximum values ​​from historical data. In addition, the module also derives a temperature-humidity index feature, with the formula: ,in, It is the ambient temperature. This refers to ambient humidity, which comprehensively reflects environmental comfort. The processed data is then sent to the predictive model building module.

[0053] The prediction model building module uses the support vector regression algorithm to train the prediction model. The model formula is as follows: in, It is the input feature vector, including ambient temperature, ambient humidity, and server load. This is a predicted output, representing air conditioning energy consumption. It is the number of support vectors. and It is a Lagrange multiplier. It is a bias term. It is a radial basis function kernel. , These are kernel parameters. It is the i-th support vector. This is the input vector to be predicted. During model training, the hyperparameters (penalty parameter C and kernel parameter) are determined through grid search and cross-validation. The loss function used is the ε-insensitive loss function, and the formula is as follows: ,in, This is the actual energy consumption value of the air conditioner. It is the tolerance for error. =0.1. The model is updated monthly to adapt to changes in data distribution. After training is complete, the model parameters are transferred to the intelligent control module.

[0054] The intelligent control module calculates the optimal cooling mode based on prediction results and real-time environmental parameters. First, it calculates the cooling demand based on predicted energy consumption. ,in, It is a prediction of energy consumption. It is the conversion factor. =1. Then, a scoring function is used. ;

[0055] in, It refers to the cooling modes, including free cooling, evaporative cooling, and mechanical refrigeration. This is the rated cooling capacity of the model. It is the ambient temperature. It is the midpoint of the applicable temperature range of the mode. It is the energy efficiency ratio of the mode. , , These are weighting coefficients, with values ​​of 0.5, 0.3, and 0.2 respectively. and It is a normalization factor, dynamically calculated based on a range of historical data. The intelligent control module selects the mode with the highest score and generates a control signal (such as a relay switch or PWM signal) to send to the composite cooling module.

[0056] The composite refrigeration module comprises three subsystems: mechanical refrigeration, evaporative cooling, and free cooling. The mechanical refrigeration subsystem uses a compression cycle and includes a compressor, condenser, expansion valve, and evaporator. The evaporative cooling subsystem absorbs heat through water evaporation for cooling. The free cooling subsystem utilizes outdoor air or a water-side economizer for heat exchange. Upon receiving intelligent control signals, the module switches the refrigeration mode via solenoid valves or a frequency converter. Free cooling is prioritized in low-temperature environments, evaporative cooling is activated in medium-temperature environments, and mechanical refrigeration is switched to in high-load or high-temperature environments. The module is equipped with flow and temperature sensors to ensure smooth mode switching.

[0057] The monitoring and feedback module collects real-time data on the operating status of the composite refrigeration module (such as compressor frequency and water pump speed), energy consumption data (such as power consumption and COP), and environmental parameters through an embedded system, and displays the system status through a human-machine interface (such as a touch screen). This module feeds the data back to the data acquisition module, forming a closed-loop control. If the system energy efficiency is lower than the threshold, an early warning is triggered and control parameters are adjusted.

[0058] Example 2:

[0059] refer to Figure 2The specific flow of the control method is as follows: The data acquisition step collects ambient temperature, humidity, server load, air conditioning energy consumption, and cooling mode data every minute. The data processing step performs data cleaning, normalization, and feature extraction, including calculating the temperature-humidity index. The prediction model building step trains a support vector regression model using the processed data, optimizes hyperparameters, and updates it periodically. The prediction step inputs the current feature vector into the model, outputs predicted energy consumption, and calculates cooling demand. The control step selects the optimal cooling mode based on cooling demand and ambient temperature using a scoring function and sends a control signal. The closed-loop monitoring step monitors the operating status and feeds the data back to the data acquisition step for model updates and strategy optimization.

[0060] This invention achieves dynamic optimization of cooling modes through the aforementioned system and method, significantly reducing data center air conditioning energy consumption. In actual tests, the system prioritizes free cooling during transitional seasons, reducing energy consumption by more than 40% compared to traditional mechanical cooling; in high-temperature environments, it switches to a high-efficiency mechanical cooling mode in advance through a predictive model, improving the energy efficiency ratio by 25%. Furthermore, closed-loop monitoring ensures the long-term stability and adaptability of the system.

[0061] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0062] It should be noted that the components mentioned in the above embodiments are all general standard parts or components known to those skilled in the art. Their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0063] This invention has illustrated its principles and implementation methods using specific examples. The descriptions of these embodiments are merely illustrative of the method and its core ideas; furthermore, those skilled in the art will recognize that modifications may be made to the specific implementation methods and application scope based on the principles of this invention. Therefore, the content of this specification should not be construed as limiting the invention.

Claims

1. A data center integrated air conditioning system, characterized in that, include: The data acquisition module is used to collect historical data and operational status data from inside and outside the data center in real time, and to obtain real-time monitoring data. The data processing module is used to clean, normalize, and extract features from the raw data transmitted from the data acquisition module. A prediction model building module is used to build a prediction model using processed historical data and to update the model parameters periodically. The intelligent control module is used to calculate the optimal cooling mode and generate control signals based on the prediction results provided by the prediction model construction module and the current environmental parameters. A composite refrigeration module, which integrates mechanical refrigeration, evaporative cooling and free cooling, and automatically switches the refrigeration mode according to the control signal of the intelligent control module; The monitoring and feedback module is used to monitor the operating status, energy consumption performance and environmental parameters of the composite refrigeration module in real time, and feed the data back to the data acquisition module to form a closed-loop control.

2. The data center composite air conditioning system according to claim 1, characterized in that, The data acquisition module collects data including ambient temperature, ambient humidity, server load, air conditioning energy consumption, and cooling mode, with the data acquisition frequency set to once per minute.

3. The data center composite air conditioning system according to claim 1, characterized in that, The data cleaning module uses the Z-score method to detect and remove outliers, and linear interpolation is used to fill missing values. The data normalization method uses the Min-Max normalization method to scale the data to the range of [0,1].

4. The data center composite air conditioning system according to claim 1, characterized in that, The prediction model construction module uses the support vector regression algorithm to construct the prediction model, and the model formula is as follows: ;in, It is the input feature vector, including ambient temperature, ambient humidity, and server load. This is a predicted output, representing air conditioning energy consumption. It is the number of support vectors. and It is a Lagrange multiplier. It is a bias term. It is a radial basis function kernel. , These are kernel parameters. It is the i-th support vector. The input vector to be predicted is used; the model training determines the hyperparameters, including penalty parameters and kernel parameters, through grid search and cross-validation. The loss function is an ε-insensitive loss function, and the formula for calculating the loss function value is: ;in, This is the actual energy consumption value of the air conditioner. It is the tolerance for error.

5. The data center composite air conditioning system according to claim 1, characterized in that, The intelligent control module calculates the cooling demand based on the prediction results, using the following formula: ;in, It is a prediction of energy consumption. This is the conversion factor; the cooling mode is dynamically switched based on ambient temperature and predicted cooling demand. The mode selection algorithm uses a scoring function, and the formula is: ;in, It refers to the cooling modes, including free cooling, evaporative cooling, and mechanical refrigeration. This is the rated cooling capacity of the model. It is the ambient temperature. It is the midpoint of the applicable temperature range of the mode. It is the energy efficiency ratio of the mode. , , These are weighting coefficients. and It is a normalization factor, and the mode with the highest score is selected as the optimal mode.

6. A control method for a data center integrated air conditioning system, characterized in that, Includes the following steps: S1: Data Acquisition, real-time acquisition of historical and operational status data from inside and outside the data center, including ambient temperature, ambient humidity, server load, air conditioning energy consumption, and cooling mode; S2: Data processing involves cleaning, normalizing, and extracting features from the collected raw data. The Z-score method is used to remove outliers, and linear interpolation is used to fill missing values. The Min-Max normalization method is used to scale the data to the range of [0,1]. S3: Predictive model construction, using processed historical data to build a support vector regression predictive model; determine hyperparameters through grid search and cross-validation, and update the model regularly; S4: Prediction, input the current feature vector into the trained prediction model to obtain the predicted energy consumption and calculate the cooling requirement; S5: Control. Based on the predicted cooling demand and ambient temperature, a scoring function is used to calculate the score of each cooling mode. The mode with the highest score is selected as the optimal mode, and a control signal is generated to switch the cooling mode.

7. The control method for a data center composite air conditioning system according to claim 6, characterized in that, In step S2, feature extraction also includes deriving temperature-humidity index features, the formula of which is: ;in, It is the ambient temperature. It refers to ambient humidity.

8. The control method for a data center composite air conditioning system according to claim 6, characterized in that, In step S3, the error tolerance of the loss function during model training is set to 0.

1.

9. The control method for a data center composite air conditioning system according to claim 6, characterized in that, In step S5, the weighting coefficients in the scoring function =0.5、 =0.3、 =0.2, normalization factor and Dynamically adjusted based on historical data range; The switching threshold for cooling modes is adaptively adjusted based on the prediction results.

10. The control method for a data center composite air conditioning system according to claim 6, characterized in that, Also includes: S6: Closed-loop monitoring, which monitors the operating status and energy consumption performance of the cooling mode in real time, and feeds the data back to the data acquisition step to update the predictive model and optimize the control strategy.

Images