Energy-saving control method and system for low-load air conditioner of data center

Abstract

The invention belongs to the technical field of energy conservation and emission reduction, and particularly relates to a data center low-load air conditioner energy conservation control method and system.The method comprises the steps that an upper computer monitors an interface in real time to conduct visual digital modeling on a data center through a data center digital modeling method, and the data center is displayed in a 3D simulation and digital mode in real time; the network data communication system realizes data interaction and real-time communication by using a network communication link construction method; the data sensing system obtains equipment operation data in real time and supports energy-saving control of the air conditioner; the temperature prediction system is applied to the temperature change prediction method, and preprocessing of collected data, data set manufacturing and training of a neural network model are achieved. The equipment parameter control system is internally provided with an equipment operation parameter optimization control method, optimal combination optimization of air conditioner internal equipment operation parameters is achieved, and an instruction is issued and transmitted to the console to achieve control. And the console adjusts and controls the related parameters of the air conditioning equipment according to the control instruction received from the network data communication system.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy conservation and emission reduction, and in particular relates to a data center low-load air conditioning energy conservation control method and system. Background Art

[0002] With the rapid progress of electronic information technology, the scale of data centers continues to expand, and the power density of single cabinets is also increasing. From the early 3kW and 4kW to the current 6kW, and even to more than 10kW, the energy consumption problem of data centers has become increasingly prominent. In particular, the energy consumption of IT equipment and air conditioning accounts for about 50% and 40% of the total energy consumption, respectively, which has become a bottleneck restricting the further development of data centers. The low-load air conditioning energy-saving control method is a solution to this problem, which can significantly reduce the energy consumption of data centers and improve energy utilization efficiency.

[0003] The air conditioning system in the data center not only ensures the normal operation of the equipment, but also consumes a lot of energy. Especially when the data center load is low, the traditional air conditioning system may not be able to effectively adapt to this change, resulting in excessive energy consumption. Therefore, how to achieve air conditioning energy saving when the data center is under low load has become the focus of the industry.

[0004] In this context, energy-saving control methods for low-load air conditioning in data centers have emerged. These methods achieve precise control of data center air conditioning systems by optimizing the operation strategies of air conditioning systems and combining advanced control technology and sensor technology. This can not only reduce the energy consumption of data centers and improve energy efficiency, but also help extend the service life of air conditioning equipment and reduce maintenance costs. At the same time, with the continuous advancement and innovation of technology, low-load air conditioning energy-saving control methods will become more mature and efficient. For example, technologies such as adaptive regulation and intelligent control based on AI algorithms will enable air conditioning systems to be more accurately adjusted dynamically according to the load conditions and environmental changes of the data center, achieving more efficient energy-saving effects.

[0005] In addition, as the concepts of green, low-carbon and sustainable development gain popularity, the demand for energy-saving technologies in the data center industry will continue to grow. As an effective means of energy saving, low-load air conditioning energy-saving control methods will be more widely used and promoted.

[0006] Finally, with the rapid development of new generation information technologies such as the Internet of Things and cloud computing, data centers will carry more services and data, and their energy consumption problems will become more prominent. Therefore, low-load air conditioning energy-saving control methods will play a more important role in the construction and operation of future data centers.

[0007] In general, the technical background of the data center low-load air conditioning energy-saving control method is the prominent energy consumption problem of the data center and the demand for energy-saving and efficient operation, which has huge market potential and social value. By applying this method, the operating cost of the data center can be effectively reduced, the energy utilization efficiency can be improved, and strong support can be provided for the sustainable development of the data center, and strong support can be provided for the energy-saving and efficient operation of the data center.

[0008] The prior art CN116817424A discloses an air-conditioning energy-saving control system and method based on artificial intelligence, including information sorting and categorizing of all valid operating parameter setting switching behaviors that appear in all historical equipment control events; capturing the characteristic use operating settings that exist when the user controls the target air-conditioning equipment; capturing all second characteristic historical switching behaviors that satisfy the switch from high loss to low loss from all first characteristic historical switching behaviors; capturing the condition information that needs to be met when each second characteristic historical switching behavior occurs respectively; when the operating status of the current operating parameter settings of the target air-conditioning equipment meets the condition information when a corresponding second characteristic historical switching behavior occurs, the current operating parameter settings of the target air-conditioning equipment are set and switched according to a second characteristic historical switching behavior, but this method can only fix a few corresponding states to make limited adjustments to parameters such as air-conditioning power, and the degree of intelligence and controllability are low. In the prior art CN117556217A, a water-cooled central air-conditioning energy-saving control method based on heterogeneous graph neural network and simulated annealing algorithm is disclosed. The method establishes a heterogeneous graph neural network model according to the actual water-cooled central air-conditioning refrigeration room environment, trains the model using a training data set, and simulates the operating environment of the refrigeration room with the trained model; then establishes a simulated annealing algorithm model, and places it in the simulated refrigeration room operating environment for training and learning, and evaluates it in combination with the real central air-conditioning operating data until the energy-saving effect meets the requirements, and then stops training, imports the trained simulated annealing algorithm model into the control terminal, and performs reasoning and issuing control instructions for energy-saving regulation. The present invention uses a heterogeneous graph neural network as a control model, and only updates whether there is a relationship between nodes, and does not need to update the weight and bias of the data, thereby reducing training time and saving training costs. Through the present invention, the control instructions are safely evaluated, and the control instructions are issued, which is predicted and reviewed, to ensure the safe operation of the system. However, this method is mainly aimed at water-cooled central air conditioning energy-saving systems, and this method is mainly used to optimize the algorithm flow of the central air conditioning energy-saving control method based on a neural network, which is not universal, and has no obvious effect on the air conditioning system equipped in the data center with high energy consumption, high degree of optimization of energy distribution, high flexibility and high dynamics. To this end, the present invention proposes a data center low-load air conditioning energy-saving control method and system. Summary of the invention

[0009] In order to solve the above problems, the present invention proposes a data center low-load air conditioning energy-saving control method and system to solve the problems existing in the above-mentioned prior art and improve the effectiveness and efficiency of data center low-load air conditioning energy-saving control.

[0010] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: a data center low-load air conditioning energy-saving control method, applied to a low-load air conditioning energy-saving control system, including a data center digital modeling method, a system information group receiving and sending method, a temperature change prediction method and an equipment operation parameter optimization control method.

[0011] Furthermore, the data center low-load air conditioning energy-saving control method and the data center digital modeling method use digital twin technology to digitally model the IT equipment and air conditioning equipment contained in the data center, and display the relevant data of each device in real-time digital visualization, specifically including the following steps:

[0012] Step 1: Data collection: Collect various data of IT equipment and air-conditioning equipment in the data center, including the name, operating time, real-time power, energy consumption, heat dissipation, equipment temperature, temperature range required for normal operation of the equipment and indoor temperature data collected by sensors and monitoring equipment.

[0013] Step 2: Data preprocessing: Clean, filter, denoise, convert and process the collected data to eliminate abnormal, missing and erroneous data, and ensure the consistency and accuracy of the data.

[0014] Step 3: Feature extraction and selection: Extract the corresponding key features of the equipment from the preprocessed data, including the power consumption, heat dissipation and indoor temperature data of the equipment at different time periods, to represent the main attributes and operating characteristics of the equipment, reduce the complexity of the model, and improve the accuracy and modeling efficiency of the model.

[0015] Step 4: Digital modeling: Based on the extracted features and artificial prior knowledge of the system, a digital twin model is constructed to simulate the behavior of the data center system.

[0016] Step 5: Model calibration: Use the data observed by the sensor in real time to calibrate the digital model to ensure the accuracy and reliability of the model, and optimize and adjust the model parameters.

[0017] Step 6: Model analysis: Use the established digital twin model to simulate and analyze the operation of IT equipment, heat dissipation, air conditioning operation, and indoor temperature changes in the data center at different time periods to verify the performance of the model and identify potential problems or optimization space.

[0018] Step 7: Visualization and interaction: Visualize the digital twin model and use interactive technology to control and operate the model.

[0019] Furthermore, in the data center low-load air conditioning energy-saving control method, the data preprocessing method includes using a joint linear regression and clustering algorithm to remove abnormal data and reduce noise interference.

[0020] Furthermore, in the data center low-load air conditioning energy-saving control method, the system information group receiving and sending method includes a network communication link construction method and a network protocol and device address design method.

[0021] Furthermore, in the method for energy-saving control of low-load air conditioning in a data center, the method for constructing a network communication link includes using two communication links, namely power carrier communication and optical fiber communication, and adopting two transmission media, namely unshielded twisted pair and trunk optical fiber. The optical fiber communication network is mainly used for data interaction between the on-site control console of the data center and each device, and between each device, and the power carrier communication network is used for data communication between the on-site control console of the data center and the upper-level management and monitoring center.

[0022] Furthermore, in the method for energy-saving control of low-load air conditioning in a data center, the network protocol and device address design method include using the TCP / IP protocol as the communication protocol between the physical layer and the data link layer; using the network number, IP address, MAC address and device number to identify each IT device and air-conditioning equipment, and supporting data interaction between the console and each device, as well as between each device.

[0023] Furthermore, in the data center low-load air conditioning energy-saving control method, the temperature change prediction method is an intelligent temperature change prediction method based on a neural network model, and the specific steps are as follows:

[0024] Step 1: Use sensors and communication networks to continuously collect data on power changes, temperature changes of each IT device in the data center, temperature changes of each air conditioner setting, and indoor temperature changes over a period of time, and build a corresponding database.

[0025] Step 2: Use the current air conditioning set temperature value as input, IT equipment power change as the influencing factor, indoor temperature and the predicted temperature of each IT device as the output of the neural network model, and the current indoor temperature and the temperature of each IT device as the model prediction bias to construct a two-dimensional data set. The predicted values ​​of the data center indoor temperature and the temperature of each IT device can be expressed by the formula:

[0026] data=F(α,β)+γ

[0027] The optimization method for equipment operation parameters is a joint neural network, where data is the temperature prediction value, α is the current temperature value set by the air conditioner, β is the power of the IT equipment, γ is the initial temperature of the IT equipment or the room, and F(·) represents the mapping relationship between α and β. The problem is then converted to finding the mapping relationship between α and β.

[0028] Step 3: Divide the two-dimensional data set constructed in step 2 into training set and test set in a ratio of 8:2.

[0029] Step 4: Construct a convolutional recurrent generative adversarial neural network model and use the neural network model to fit the mapping relationship F(·) between α and β.

[0030] Step 5: Use the constructed convolutional loop to generate an adversarial neural network model to complete model training on the training set, load the saved model weights, and complete the model performance test on the test set.

[0031] Step 6: Deploy the tested neural network model to the control system, input the current air conditioning set temperature, and continuously input the IT equipment power changes to obtain the temperature change curves for the data center room and each IT equipment in the future. According to the temperature change value, adjust the air conditioning temperature setting appropriately to keep the indoor temperature and equipment temperature near the ideal threshold at all times, ensuring the normal operation of IT equipment while avoiding energy waste caused by setting the air conditioning temperature too low.

[0032] Furthermore, the low-load air conditioning energy-saving control method for a data center, the equipment operating parameter optimization control method is an equipment operating parameter optimization control method based on a simulated annealing algorithm, and mainly performs joint optimization on the operating parameters of the fan, chilled water pump, refrigerator, cooling water pump, cooling tower, and surface cooler equipment inside the air conditioning equipment, and the specific steps are as follows:

[0033] Step 1: Conduct mathematical modeling of the fan, chilled water pump, refrigerator, cooling water pump, cooling tower, and surface cooler equipment inside the air-conditioning equipment.

[0034] Step 2: Under the premise of meeting the cooling load and maintaining the current temperature demand of the air conditioner, the total energy consumption of the air conditioning equipment is minimized, and the optimal operating parameter combination of each device is sought based on the simulated annealing algorithm.

[0035] Step 3: According to the best parameter combination obtained in step 2, adjust the parameter values ​​of the cooling and heating equipment inside the air-conditioning equipment to achieve energy-saving control of the air-conditioning equipment.

[0036] On the other hand, a data center low-load air conditioning energy-saving control system is provided, which is applied to any one of the data center low-load air conditioning energy-saving control methods described above, and the data center low-load air conditioning energy-saving control system comprises:

[0037] Real-time monitoring interface of the host computer: The real-time monitoring interface of the host computer applies the digital modeling method of the data center and uses digital twin technology to visualize the operation status, temperature, operation status, power, energy consumption, and indoor temperature of each IT equipment in the data center. It is used to monitor the operation status, temperature changes, air-conditioning operation status and indoor temperature changes of each equipment in the data center in real time, and display them in real time in 3D simulation and digital form.

[0038] Network data communication system: The network data communication system is applied to the network communication link construction method to realize data interaction and real-time communication between the control console, IT equipment, host computer real-time monitoring interface, air conditioning equipment and upper control center.

[0039] Data sensing system: The data sensing system obtains equipment operation data in real time by installing various sensors on the equipment to support air conditioning energy-saving control.

[0040] Temperature prediction system: The temperature prediction system is applied to the temperature change prediction method, including a data processing platform and a computing power support platform, equipped with a high-computing power GPU, to realize the preprocessing of data collected by the sensor system, data set preparation, and training and testing of neural network models.

[0041] Equipment parameter control system: The equipment parameter control system is applied to the equipment operating parameter optimization control method to optimize the best combination of equipment operating parameters inside the air conditioner, and issue instructions to the control console to achieve control.

[0042] Control console: The control console adjusts and controls the relevant parameters of the air-conditioning equipment according to the control instructions received from the network data communication system.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The present invention realizes accurate and vivid modeling of the data center through digital twin technology, and realizes real-time, intuitive and quantitative monitoring of the status of each device in the data center through digital modeling and interactive technology, supporting the control system to realize accurate energy-saving control of the data center air conditioning.

[0045] 2. The present invention realizes customized high-speed data information transmission for different business flows by constructing a network data communication system of hybrid power carrier communication and optical fiber communication links supported by the TCP / IP protocol, reduces data transmission delays, and assists the data center low-load air conditioning energy-saving control system to achieve fast and accurate equipment control.

[0046] 3. The present invention uses an intelligent temperature prediction method based on a neural network model to achieve accurate prediction of the temperature of each IT equipment and indoor temperature in the data center with the dual support of data and computing power, which helps to achieve more refined adjustment and setting of the air-conditioning temperature, reduce temperature redundancy, and minimize air-conditioning energy consumption while ensuring the normal operation of IT equipment.

[0047] 4. The present invention obtains the optimal equipment parameter configuration under the premise of normal operation of the equipment and the minimum air conditioning energy consumption requirement through the equipment operation parameter optimization control method, thereby realizing precise control of the air conditioner under the low-load energy-saving goal. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the energy-saving control method for low-load air conditioning in a data center according to the present invention;

[0049] Figure 2 This is a structural diagram of the low-load air conditioning energy-saving control system for a data center according to the present invention;

[0050] Figure 3 This is the working principle diagram of data center air conditioning;

[0051] Figure 4 This is a flow chart of the combined linear regression and clustering algorithm of the present invention.

[0052] Figure 5 This is a flow chart of the simulated annealing algorithm of the present invention. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0054] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, "plurality" means two or more, unless otherwise clearly and specifically defined.

[0055] In the description of the present application, the term "for example" is used to mean "used as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any technician in the field to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes will not be elaborated in detail to avoid unnecessary details to obscure the description of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0056] In an embodiment of the present invention, a data center low-load air conditioning energy-saving control method is provided. Figure 1 , Attachment Figure 4 ~Attached Figure 5 , applied to low-load air-conditioning energy-saving control systems, including data center digital modeling methods, system information group receiving and sending methods, temperature change prediction methods and equipment operation parameter optimization control methods.

[0057] The digital modeling method of the data center uses digital twin technology to digitally model the IT equipment and air-conditioning equipment contained in the data center, and displays the heat dissipation, power, energy consumption, temperature, temperature threshold required for normal operation and indoor temperature data of each device in real time. It also uses joint linear regression and clustering algorithms to remove abnormal data and reduce noise interference.

[0058] The method for sending and receiving system information groups includes a network communication link construction method and a network protocol and device address design method, specifically including the use of two communication links, namely power carrier communication and optical fiber communication, and the use of two transmission media, namely unshielded twisted pair and trunk optical fiber. The optical fiber communication network is mainly used for data interaction between the on-site control console of the data center and each device, as well as between each device, and the power carrier communication network is used for data communication between the on-site control console of the data center and the upper-level management and monitoring center; the network protocol and device address design method includes using the TCP / IP protocol as the communication protocol between the physical layer and the data link layer; using the network number, IP address, MAC address and device number to identify each IT device and air-conditioning equipment, and supporting data interaction between the control console and each device, as well as between each device.

[0059] The temperature change prediction method is an intelligent temperature change prediction method based on a neural network model. The specific steps are as follows:

[0060] Step 1: Use sensors and communication networks to continuously collect data on power changes, temperature changes of each IT device in the data center, temperature changes of each air conditioner setting, and indoor temperature changes over a period of time, and build a corresponding database.

[0061] Step 2: Use the current air conditioning set temperature value as input, IT equipment power change as the influencing factor, indoor temperature and the predicted temperature of each IT device as the output of the neural network model, and the current indoor temperature and the temperature of each IT device as the model prediction bias to construct a two-dimensional data set. The predicted values ​​of the data center indoor temperature and the temperature of each IT device can be expressed by the formula:

[0062] data=F(α,β)+γ

[0063] The optimization method for equipment operation parameters is a joint neural network, where data is the temperature prediction value, α is the current temperature value set by the air conditioner, β is the power of the IT equipment, γ is the initial temperature of the IT equipment or the room, and F(·) represents the mapping relationship between α and β. The problem is then converted to finding the mapping relationship between α and β.

[0064] Step 3: Divide the two-dimensional data set constructed in step 2 into training set and test set in a ratio of 8:2.

[0065] Step 4: Construct a convolutional recurrent generative adversarial neural network model and use the neural network model to fit the mapping relationship F(·) between α and β.

[0066] Step 5: Use the constructed convolutional loop to generate an adversarial neural network model to complete model training on the training set, load the saved model weights, and complete the model performance test on the test set.

[0067] Step 6: Deploy the tested neural network model to the control system, input the current air conditioning set temperature, and continuously input the IT equipment power changes to obtain the temperature change curves for the data center room and each IT equipment in the future. According to the temperature change value, adjust the air conditioning temperature setting appropriately to keep the indoor temperature and equipment temperature near the ideal threshold at all times, ensuring the normal operation of IT equipment while avoiding energy waste caused by setting the air conditioning temperature too low.

[0068] The equipment operating parameter optimization control method is an equipment operating parameter optimization control method based on the simulated annealing algorithm. It mainly conducts joint optimization for the operating parameters of the fans, chilled water pumps, refrigerators, cooling water pumps, cooling towers, and surface coolers inside the air-conditioning equipment. The specific steps are as follows:

[0069] Step 1: Conduct mathematical modeling of the fan, chilled water pump, refrigerator, cooling water pump, cooling tower, and surface cooler equipment inside the air-conditioning equipment.

[0070] Step 2: Under the premise of meeting the cooling load and maintaining the current temperature demand of the air conditioner, the total energy consumption of the air conditioning equipment is minimized, and the optimal operating parameter combination of each device is sought based on the simulated annealing algorithm.

[0071] Step 3: According to the best parameter combination obtained in step 2, adjust the parameter values ​​of the cooling and heating equipment inside the air-conditioning equipment to achieve energy-saving control of the air-conditioning equipment.

[0072] See attached Figure 2 The present invention also provides a data center low-load air conditioning energy-saving control system, including a host computer real-time monitoring interface, a network data communication system, a data sensing system, a temperature prediction system, an equipment parameter control system and a control console.

[0073] The real-time monitoring interface of the upper computer applies the digital modeling method of the data center and uses digital twin technology to visualize the operation status, temperature, operation status, power, energy consumption, and indoor temperature of each IT equipment in the data center. It is used to monitor the operation status, temperature changes, air-conditioning operation status, and indoor temperature changes of each equipment in the data center in real time, and display them in real time in 3D simulation and digital form.

[0074] The network data communication system is applied to the network communication link construction method to realize data interaction and real-time communication between the control console, IT equipment, host computer real-time monitoring interface, air-conditioning equipment and the upper control center.

[0075] The data sensing system installs various sensors on the equipment to obtain equipment operation data in real time and support air conditioning energy-saving control.

[0076] The temperature prediction system is applied to the temperature change prediction method, which includes a data processing platform and a computing power support platform. It is equipped with a high-computing power GPU to realize the preprocessing of data collected by the sensor system, data set preparation, and training and testing of neural network models.

[0077] The equipment parameter control system is applied to the equipment operating parameter optimization control method to optimize the best combination of equipment operating parameters inside the air conditioner, and send instructions to the control console to realize control.

[0078] The control console adjusts and controls the relevant parameters of the air-conditioning equipment according to the control instructions received from the network data communication system.

[0079] In a specific embodiment, the low-load air conditioning energy-saving control personnel of the data center use the built-in data center digital modeling method of the host computer real-time monitoring interface to monitor the working status of each IT equipment in the data center, air conditioning temperature, energy consumption and indoor temperature changes in real time in 3D digital mode. The data sensing system obtains the operating data of each device in real time, and transmits the sensing data back to the host computer real-time monitoring interface and temperature prediction system in real time with ultra-low latency through the network data communication system constructed by the network communication link construction method. The temperature prediction system has a built-in temperature change prediction method, which predicts the indoor temperature and temperature changes of each IT equipment at future moments through the device data transmitted by the network data communication system, and uploads the change trend curve to the host computer real-time monitoring interface. The control personnel make judgments based on the temperature change trend: if the temperature is too high in the future, it will affect the normal operation of IT equipment and even cause equipment damage. The control personnel will gradually reduce the air conditioning temperature setting in real time through the console to reduce the indoor temperature within the minimum energy consumption range; if the indoor temperature and IT equipment power and temperature continue to decrease in the future, and the indoor temperature is sufficient to support the normal operation of IT equipment, the control personnel can appropriately increase the air conditioning temperature through the console to reduce air conditioning energy consumption. The air conditioning temperature adjustment instructions of the control personnel are sent to the equipment parameter control system through the network data communication system. The equipment parameter control system has a built-in equipment operation parameter optimization control method. This method is based on the premise of achieving the air conditioning temperature setting. Under this premise, the parameter settings of each refrigeration equipment inside the air conditioner are adjusted with the goal of minimizing air conditioning energy consumption, thereby achieving the air conditioning low-load energy saving goal.

[0080] Example 1

[0081] In one embodiment, the management and control personnel perform digital twin data visualization modeling of the data center through the data center digital modeling method. First, the operating data of each IT equipment and air-conditioning equipment in the data center is collected through sensor equipment, including the operating power and temperature of the IT equipment over a long period of time, the set temperature, power, energy consumption of the air-conditioning equipment, and indoor temperature data. In order to prevent the interference of erroneous data, missing data and abnormal data on the accuracy of modeling, the management and control personnel use joint linear regression and clustering algorithms to pre-process the data, remove abnormal data, and reduce noise interference. The joint linear regression and clustering algorithms are to first input the data to be processed into the linear regression model to obtain the approximate data distribution and remove obvious errors or duplicate data. After that, the data cleaned by the linear regression algorithm is input into the clustering algorithm to perform more refined cleaning on the data and eliminate abnormal data that is not easy to find. The mathematical principle of using the linear regression algorithm to clean data is:

[0082]

[0083] Among them, y is the data distribution function curve predicted by the linear regression model, w is the model parameter, x is the characteristic data including IT equipment power, temperature, air conditioning temperature setting, energy consumption, and data center indoor temperature, and b is the bias. Draw the approximate function curve of the data distribution according to the linear regression algorithm, and remove the deviation data that obviously deviates from the curve to obtain the preliminary cleaned data. Afterwards, the control personnel will send the preliminary cleaned data to the clustering algorithm. The clustering algorithm adopts the K-Means clustering algorithm. The specific operation steps are as follows:

[0084] Step 1: Assume that there are n types of data collected by the sensor, and randomly select one data from each type of data as the initial clustering center.

[0085] Step 2: Calculate the Manhattan distance from each data point in the database to each cluster center, and assign the data point to the category corresponding to the cluster center closest to it. The calculation formula for Manhattan distance is:

[0086] d=|d1-d0|

[0087] Where d is the Manhattan distance, d1 is the data point to be classified, d0 is the data center, and |·| is the absolute value operation.

[0088] Step 3: After completing a round of clustering, calculate the average value of the data points assigned to each cluster and use the average value as the new cluster center.

[0089] Step 4: Repeat steps 2 and 3 until the cluster center no longer changes or changes very little, or the preset maximum number of iterations is reached. At this time, the data that has not been classified is the abnormal data that needs to be eliminated.

[0090] After completing the data preprocessing, the control personnel extract the corresponding key features of the equipment from the preprocessed data, including the power consumption, heat dissipation and indoor temperature data of the equipment at different time periods, to represent the main attributes and operating characteristics of the equipment, and build a digital twin model based on the extracted features and the system's artificial prior knowledge to simulate the behavior of the data center system. After the model is built, the control personnel can use the data observed by the sensor in real time to calibrate the digital model to ensure the accuracy and reliability of the model, optimize and adjust the model parameters, and use the established digital twin model to simulate and analyze the operation, heat dissipation, air conditioning operation and indoor temperature changes of the data center IT equipment at different time periods, verify the performance of the model, and find potential problems or optimization space. Finally, the digital twin model is visualized and applied to the real-time monitoring interface of the host computer, and interactive technology is used to control and operate the model.

[0091] Example 2

[0092] In one embodiment, in order to ensure efficient, accurate and real-time transmission of data and control information and realize data communication between devices, the control personnel build a network data communication system through a system information group sending and receiving method. The system information group sending and receiving method includes a network communication link construction method and a network protocol and device address design method, which specifically includes the use of two communication links, power carrier communication and optical fiber communication, and the use of two transmission media, unshielded twisted pair and trunk optical fiber. The optical fiber communication network is mainly used for data exchange between the data center field console and each device, and between each device. The power carrier communication network is used for data communication between the data center field console and the upper management and monitoring center; the network protocol and device address design method includes using the TCP / IP protocol as the communication protocol between the physical layer and the data link layer; using the network number, IP address, MAC address and device number to identify each IT device and air conditioning equipment, and supporting the data exchange between the console and each device, and between each device.

[0093] Example 3

[0094] In one embodiment, the data center controller hopes to predict the indoor temperature of the data center and the temperature of each IT device in the future based on the daily working power change of the IT equipment in the data center and the historical data of the indoor temperature change, and adjust the air conditioning temperature step by step in advance, so as to avoid the high energy consumption caused by the large adjustment of the air conditioning set temperature in a short period of time while ensuring that the indoor temperature and the temperature of each IT device are within the normal working range. The controller achieves the above goal through the temperature change prediction method, and the specific steps include:

[0095] Step 1: Use sensors and communication networks to continuously collect data on power changes, temperature changes of each IT device in the data center, temperature changes of each air conditioner setting, and indoor temperature changes over a period of time, and build a corresponding database.

[0096] Step 2: Use the current air conditioning set temperature value as input, IT equipment power change as the influencing factor, indoor temperature and the predicted temperature of each IT device as the output of the neural network model, and the current indoor temperature and the temperature of each IT device as the model prediction bias to construct a two-dimensional data set. The predicted values ​​of the data center indoor temperature and the temperature of each IT device can be expressed by the formula:

[0097] data=F(α,β)+γ

[0098] The optimization method for equipment operation parameters is a joint neural network, where data is the temperature prediction value, α is the current temperature value set by the air conditioner, β is the power of the IT equipment, γ is the initial temperature of the IT equipment or the room, and F(·) represents the mapping relationship between α and β. The problem is then converted to finding the mapping relationship between α and β.

[0099] Step 3: Divide the two-dimensional data set constructed in step 2 into training set and test set in a ratio of 8:2.

[0100] Step 4: Construct a convolutional recurrent generative adversarial neural network model. The convolutional recurrent generative adversarial network model is a combination of a convolutional recurrent neural network and a generative adversarial network. The initial temperature change results predicted by the convolutional recurrent neural network are sent to the generative adversarial network model to generate refined prediction results. The convolutional recurrent neural network contains three convolutional recurrent layers, which use 128 1*1, 3*3, and 5*5 convolution kernels of three types with step-by-step gradient ascent. Batch normalization operations and LeakyReLU nonlinear activation functions are used after each convolutional recurrent layer. The mean square error (MSE) loss function is used for the training of the recurrent neural network model. The calculation formula is as follows:

[0101]

[0102] Among them, y i For actual results, Predict results for the neural network.

[0103] The generative adversarial network consists of two parts: the generator and the discriminator. The generative adversarial network uses adversarial loss as the training objective function. The mathematical expression of adversarial loss is:

[0104]

[0105] The generator G transforms a noise sample z~P n (z) as input, and then generate a target sample x according to the target unknown data distribution, where x~G(z); the discriminator D randomly samples the output of the generator to obtain features S1, and at the same time obtains samples from the original real sample to obtain S2. The discriminator is trained using the difference between the features of the two samples collected to maximize the ability to know that S1 is the feature of the fake sample, and S2 is the feature of the real sample; the first item It represents the mathematical expectation of the probability that the discriminator determines that the sample is a real sample from the real data distribution; the second term Indicates that for the noise P n The samples sampled from the (z) distribution are generated by the generator to obtain the generated data, and then sent to the discriminator. After being judged as true by the discriminator, the output is the refined prediction result.

[0106] Step 5: Use the constructed convolutional loop generative adversarial neural network model to complete model training on the training set, so that the neural network model can well fit the mapping relationship F(·) between α and β. Load the saved model weights and complete the model performance test on the test set.

[0107] Step 6: Deploy the tested neural network model to the control system, input the current air conditioning set temperature, and continuously input the IT equipment power changes to obtain the temperature change curves for the data center room and each IT equipment in the future. According to the temperature change value, adjust the air conditioning temperature setting appropriately to keep the indoor temperature and equipment temperature near the ideal threshold at all times, ensuring the normal operation of IT equipment while avoiding energy waste caused by setting the air conditioning temperature too low.

[0108] Example 4

[0109] In one embodiment, the data center air conditioning equipment is essentially composed of internal fans, chilled water pumps, refrigerators, cooling water pumps, cooling towers, and surface coolers. When the air conditioner is maintained at a fixed set temperature, obtaining the best parameter combination of the above internal equipment through an optimization algorithm is the best way to reduce air conditioning energy consumption and achieve energy conservation and emission reduction. The management and control personnel use the equipment operation parameter optimization control method to achieve joint optimization of the operating parameters of the fans, chilled water pumps, refrigerators, cooling water pumps, cooling towers, and surface coolers inside the air conditioning equipment. The specific steps are as follows:

[0110] Step 1: Conduct mathematical modeling of the fan, chilled water pump, refrigerator, cooling water pump, cooling tower, and surface cooler equipment inside the air-conditioning equipment.

[0111] Fan modeling: Fan energy consumption is related to air supply volume, and its energy consumption mathematical model is modeled as follows:

[0112]

[0113] η a =a1+b1β a +c1β a 2 +d1β a 3 +e1β a 4

[0114]

[0115]

[0116]

[0117] Among them, P fan is the fan power; η a is the partial load factor; α fan is the total efficiency of the fan; ρ air is the air density; m a_N is the rated flow rate of the fan; a1~e1 are the fan characteristic fitting coefficients; β a is the fan partial load rate; ma is the actual air volume of the fan; Q S is the indoor cooling load; T N Set the indoor temperature; T S is the air supply temperature, 1.01 represents the specific heat of dry air at constant pressure; ΔT fan The temperature rise of the fan supply air.

[0118] Chilled water pump modeling: The operating power of the chilled water pump is related to the chilled water flow rate, and its energy consumption model is as follows:

[0119] P e =η e P e_N

[0120] η e =a2+b2β e +c2β e 2 +d2β e 3

[0121]

[0122]

[0123] Among them, P e is the power of the chilled water pump; η e is the partial load factor of the chilled water pump; P e_N is the rated power of the chilled water pump; a2~d2 are the chilled water pump characteristic fitting coefficients; β e is the partial load rate of the chilled water pump; m e is the chilled water flow rate; m e_N is the rated power of the chilled water pump; T1 is the inlet air temperature of the surface cooler; T2 is the outlet air temperature of the surface cooler; C p is the specific heat of water; ΔT e It is the temperature difference between supply and return water of chilled water.

[0124] Refrigerator modeling: The refrigerator provides the cooling capacity required by the system. Its model can be based on three performance curves ηη ch1 , η ch2 , η ch3 To describe, its energy consumption model is as follows:

[0125]

[0126] Q avail =Q N η ch1

[0127] Among them, P ch is the refrigerator power; Q avail C is the available cooling capacity of the refrigerator;COP_N is the rated energy efficiency ratio of the refrigerator; Q N is the rated cooling capacity of the refrigerator; η ch1 The curve showing the relationship between cooling capacity and temperature; η ch2 Indicates E EIR The relationship curve between E EIR is the ratio of shaft power to cooling capacity, defined as the inverse of the energy efficiency ratio, which varies with the chilled water outlet temperature and the cooling water return temperature; η ch3 Indicates E EIR The relationship curve between the load factor and the partial load factor is as follows:

[0128]

[0129]

[0130]

[0131]

[0132] Q e =C p m e ΔT e

[0133] Among them, a3~f3, a4~f4, a5~c5 are the characteristic fitting coefficients of the three performance curves respectively; T eo T is the chilled water outlet temperature of the refrigerator; ci is the cooling water return temperature; ch Indicates the partial load rate of the refrigerator; Q e is the cooling capacity of the refrigerator.

[0134] Cooling water pump and cooling tower modeling: The cooling water pump uses a fixed frequency water pump, and the power consumed by the cooling water pump and the cooling tower is relatively small. In order to reduce the complexity of the model, the cooling water pump power and the cooling tower power are simplified and both take the rated power. The specific modeling is as follows:

[0135] P c =P c_N

[0136] P ct =P ct_N

[0137] Among them, P c P is the cooling water pump power; c_N is the rated power of the cooling water pump; P ct is the cooling tower power; P ct_N is the rated power of the cooling tower.

[0138] Modeling of the total thermal efficiency of the surface cooler: The surface cooler connects the air cycle with the chilled water cycle and does not consume any power. Its efficiency model is described. g The total heat exchange efficiency of the surface cooler is:

[0139]

[0140] Where T2 is the outlet air temperature of the surface cooler. If the temperature rise of the pipeline is not considered, then T2 = T S Taking into account the influence of fan temperature rise, T1 can be calculated as follows:

[0141] T1=T N +ΔT fan

[0142] Step 2: Under the premise of meeting the cooling load and maintaining the current temperature demand of the air conditioner, the total energy consumption of the air conditioning equipment is minimized, and the optimal operating parameter combination of each device is sought based on the simulated annealing algorithm.

[0143] The optimization objective function is defined as:

[0144] minP=min(P fan +P e +P ch +P c +P ct )

[0145] The simulated annealing algorithm process is shown in the attached Figure 5 , specifically including the following steps:

[0146] (1) Let T = T0 represent the initial temperature at the beginning of annealing, and randomly assign an initial solution P fan_0 ,P e_0 ,P ch_0 ,P c_0 ,P ct_0 As the solution to be optimized corresponding to the power of each device at the initial time T0, the corresponding objective function P(T0) is calculated.

[0147] (2) Let T t =kT (t-1) , where k takes a value between 0 and 1, representing the rate at which the temperature drops, that is, the descent gradient.

[0148] (3) For the current solution P fan_t ,P e_t ,P ch_t ,P c_t ,P ct_t Apply random perturbations to generate a new corresponding solution P in its neighborhood fan_(t+1) ,P e_(t+1) ,P ch_(t+1) ,P c_(t+1) ,Pct_(t+1) , and calculate the corresponding objective function value P(T t+1 ).

[0149] (4) Calculate ΔP = P(T t+1 )-P(T t ).

[0150] (5) If ΔP < 0, accept the new solution as the current solution, otherwise follow the probability e -ΔE / kT Decide whether to accept the new solution.

[0151] (6) At temperature T, repeat the perturbation and acceptance process multiple times, that is, execute (3) and (4).

[0152] (7) Determine whether the temperature reaches the termination temperature level. If so, terminate the algorithm; otherwise, return to (2).

[0153] Step 3: According to the best parameter combination obtained in step 2, adjust the parameter values ​​of the cooling and heating equipment inside the air-conditioning equipment to achieve energy-saving control of the air-conditioning equipment.

[0154] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0155] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A data center low-load air conditioning energy-saving control method, comprising a data center digital modeling method, a system information group receiving and sending method, a temperature change prediction method and an equipment operation parameter optimization control method, characterized in that: The data center digital modeling method adopts digital twin technology to digitally model the IT equipment and air-conditioning equipment contained in the data center, and displays the heat dissipation, power, energy consumption and other related data of each device in real time digital visualization; the system information group receiving and sending method includes a network communication link construction method and a network protocol and device address design method; the temperature change prediction method is an intelligent temperature change prediction method based on a neural network model; the equipment operation parameter optimization control method is an equipment operation parameter optimization control method based on a simulated annealing algorithm.

2. A data center low-load air conditioning energy-saving control method according to claim 1, characterized in that: The data center digital modeling method uses digital twin technology to digitally model the IT equipment and air-conditioning equipment contained in the data center, and displays the relevant data of each device in real-time digital visualization, specifically including the following steps: Step 1: Data collection: Collect various data of IT equipment and air conditioning equipment in the data center, including the name, operating time, real-time power, energy consumption, heat dissipation, equipment temperature, temperature range required for normal operation of equipment and indoor temperature data of IT equipment and air conditioning equipment collected by sensors and monitoring equipment; Step 2: Data preprocessing: Clean, filter, denoise and convert the format of the collected data, process it, eliminate abnormal data, missing data and erroneous data, and ensure the consistency and accuracy of the data; Step 3: Feature extraction and selection: Extract the corresponding key features of the device from the preprocessed data, including the power consumption, heat dissipation and indoor temperature data of the device at different time periods, to represent the main attributes and operating characteristics of the device, reduce the complexity of the model, and improve the accuracy and modeling efficiency of the model; Step 4: Digital modeling: Based on the extracted features and artificial prior knowledge of the system, a digital twin model is constructed to simulate the behavior of the data center system; Step 5: Model calibration: Use the data observed by the sensor in real time to calibrate the digital model to ensure the accuracy and reliability of the model, and optimize and adjust the model parameters; Step 6: Model analysis: Use the established digital twin model to simulate and analyze the operation of IT equipment, heat dissipation, air conditioning operation, and indoor temperature changes in the data center at different times to verify the performance of the model and identify potential problems or optimization space; Step 7: Visualization and interaction: Visualize the digital twin model and use interactive technology to control and operate the model.

3. A data center low-load air conditioning energy-saving control method according to claim 2, characterized in that: The data preprocessing method is to combine linear regression and clustering algorithms. First, the linear regression model is input to obtain the approximate data distribution and remove obvious errors or duplicate data. Then, the data cleaned by the linear regression algorithm is input into the clustering algorithm to perform more refined data cleaning and remove abnormal data that is difficult to find. The mathematical principle of linear regression algorithm data cleaning is: Among them, y is the data distribution function curve predicted by the linear regression model, w is the model parameter, x is the characteristic data including IT equipment power, temperature, air conditioning temperature setting, energy consumption, and data center indoor temperature, and b is the offset; The clustering algorithm uses the K-Means clustering algorithm. The specific steps are as follows: Step 1: Assume that there are n types of data collected by the sensor, and randomly select one data from each type of data as the initial cluster center; Step 2: Calculate the Manhattan distance from each data point in the database to each cluster center, and assign the data point to the category corresponding to the cluster center closest to it. The calculation formula for Manhattan distance is: d=|d1-d0| Where d is the Manhattan distance, d1 is the data point to be classified, d0 is the data center, and |·| is the absolute value operation; Step 3: After completing a round of clustering, calculate the average value of the data points assigned to each cluster and use the average value as the new cluster center; Step 4: Repeat steps 2 and 3 until the cluster center no longer changes or changes very little, or the preset maximum number of iterations is reached. At this time, the data that has not been classified is the abnormal data that needs to be eliminated.

4. A data center low-load air conditioning energy-saving control method according to claim 1, characterized in that: The network communication link construction method includes using two communication links, namely power carrier communication and optical fiber communication, and adopting two transmission media, namely unshielded twisted pair and trunk optical fiber. The optical fiber communication network is mainly used for data interaction between the data center on-site control console and various devices, and between various devices, and the power carrier communication network is used for data communication between the data center on-site control console and the upper-level management and monitoring center.

5. A data center low-load air conditioning energy-saving control method according to claim 1, characterized in that: The network protocol and device address design method includes using the TCP / IP protocol as the communication protocol between the physical layer and the data link layer; using the network number, IP address, MAC address and device number to identify each IT device and air-conditioning equipment, and supporting data interaction between the console and each device, as well as between each device.

6. A data center low-load air conditioning energy-saving control method according to claim 1, characterized in that: The specific steps of the intelligent temperature change prediction method based on the neural network model include: Step 1: Use sensors and communication networks to continuously collect data on power changes and temperature changes of each IT device in the data center, changes in the set temperatures of each air conditioner, and changes in indoor temperature over a period of time, and build a corresponding database; Step 2: Use the current air conditioning set temperature value as input, IT equipment power change as the influencing factor, indoor temperature and the predicted temperature of each IT device as the output of the neural network model, and the current indoor temperature and the temperature of each IT device as the model prediction bias to construct a two-dimensional data set. The predicted values ​​of the data center indoor temperature and the temperature of each IT device can be expressed by the formula: data=F(α,β)+γ The optimization method for equipment operation parameters is a joint neural network, where data is the temperature prediction value, α is the current temperature value set for the air conditioner, β is the power of the IT equipment, γ is the initial temperature of the IT equipment or the room, and F(·) represents the mapping relationship between α and β. The problem is then transformed into finding the mapping relationship between α and β. Step 3: Divide the two-dimensional data set constructed in step 2 into training set and test set in a ratio of 8:2; Step 4: Construct a convolutional recurrent generative adversarial neural network model. The convolutional recurrent generative adversarial network model is a combination of a convolutional recurrent neural network and a generative adversarial network. The initial temperature change results predicted by the convolutional recurrent neural network are sent to the generative adversarial network model to generate refined prediction results. The convolutional recurrent neural network contains three convolutional recurrent layers, which use 128 1*1, 3*3, and 5*5 convolution kernels of three types with gradual gradient increase. Batch normalization operation and LeakyReLU nonlinear activation function are used after each convolutional recurrent layer. The mean square error (MSE) loss function is used for the training of some recurrent neural network models. The calculation formula is as follows: Among them, y i For actual results, Predict results for the neural network; The generative adversarial network consists of two parts: the generator and the discriminator. The generative adversarial network uses adversarial loss as the training objective function. The mathematical expression of adversarial loss is: The generator G transforms a noise sample z~P n (z) as input, and then generate a target sample x according to the target unknown data distribution, where x~G(z); the discriminator D randomly samples the output of the generator to obtain features S1, and at the same time obtains samples from the original real sample to obtain S2. The discriminator is trained using the difference between the features of the two samples collected to maximize the ability to know that S1 is the feature of the fake sample, and S2 is the feature of the real sample; the first item It represents the mathematical expectation of the probability that the discriminator determines that the sample is a real sample from the real data distribution; the second term Indicates that for the noise P n (z) The samples sampled from the distribution are generated by the generator to obtain the generated data, which is then sent to the discriminator. After being judged as true by the discriminator, the output is the refined prediction result; Step 5: Complete model training of the constructed convolutional loop generative adversarial neural network model on the training set, so that the neural network model can well fit the mapping relationship F(·) between α and β; load the saved model weights and complete the model performance test on the test set; Step 6: Deploy the tested neural network model to the control system, input the current air conditioning set temperature, and continuously input the IT equipment power changes to obtain the temperature change curves of the data center room and each IT equipment over a period of time in the future.

7. A data center low-load air conditioning energy-saving control method according to claim 1, characterized in that: The device operating parameter optimization control method is a device operating parameter optimization control method based on device mathematical modeling and simulated annealing algorithm.

8. A data center low-load air conditioning energy-saving control method according to claim 7, characterized in that: The equipment operating parameter optimization control method mainly realizes the joint optimization of the operating parameters of the fan, chilled water pump, refrigerator, cooling water pump, cooling tower and surface cooler equipment inside the air-conditioning equipment.

9. A data center low-load air conditioning energy-saving control method according to claim 7, characterized in that: The specific steps of the equipment operation parameter optimization control method include: Step 1: Conduct mathematical modeling of the fan, chilled water pump, refrigerator, cooling water pump, cooling tower, and surface cooler equipment inside the air-conditioning equipment; Fan modeling: Fan energy consumption is related to air supply volume, and its energy consumption mathematical model is modeled as follows: Among them, P fan is the fan power; a is the partial load factor; α fan is the total efficiency of the fan; ρ air is the air density; m a_N is the rated flow rate of the fan; a1~e1 are the fan characteristic fitting coefficients; β a is the fan partial load rate; m a is the actual air volume of the fan; Q S is the indoor cooling load; T N Set the indoor temperature; T S is the air supply temperature, 1.01 represents the specific heat of dry air at constant pressure; ΔT fan The temperature rise of the air supplied by the fan; Chilled water pump modeling: The operating power of the chilled water pump is related to the chilled water flow rate, and its energy consumption model is as follows: P.S e η e P.S e_N Among them, P e is the power of the chilled water pump; η e is the partial load factor of the chilled water pump; P e_N is the rated power of the chilled water pump; a2~d2 are the chilled water pump characteristic fitting coefficients; β e is the partial load rate of the chilled water pump; m e is the chilled water flow rate; m e_N is the rated power of the chilled water pump, kg / s; T1 is the inlet air temperature of the surface cooler; T2 is the outlet air temperature of the surface cooler; C p is the specific heat of water; ΔT e The temperature difference between the supply and return water of chilled water; Refrigerator modeling: The refrigerator provides the cooling capacity required by the system. Its model can be based on three performance curves ηη ch1 , η ch2 , η ch3 To describe, its energy consumption model is as follows: Q avail =Q N the ch1 Among them, P ch is the refrigerator power; Q avail C is the available cooling capacity of the refrigerator; COP_N is the rated energy efficiency ratio of the refrigerator; Q N is the rated cooling capacity of the refrigerator; η ch1 The curve showing the relationship between cooling capacity and temperature; η ch2 Indicates E EIR The relationship curve between E and temperature EIR is the ratio of shaft power to cooling capacity, defined as the inverse of the energy efficiency ratio, which varies with the chilled water outlet temperature and the cooling water return temperature; η ch3 Indicates E EIR The relationship curve between the load factor and the partial load factor is as follows: Q e =C p m e ΔT e Among them, a3~f3, a4~f4, a5~c5 are the characteristic fitting coefficients of the three performance curves respectively; T eo T is the chilled water outlet temperature of the refrigerator; ci is the cooling water return temperature; ch Indicates the partial load rate of the refrigerator; Q e is the cooling capacity of the refrigerator; Cooling water pump and cooling tower modeling: The cooling water pump uses a fixed frequency water pump, and the power consumed by the cooling water pump and the cooling tower is relatively small. In order to reduce the complexity of the model, the cooling water pump power and the cooling tower power are simplified and both take the rated power. The specific modeling is as follows: P c =P c_N P ct =P ct_N Among them, P c P is the cooling water pump power; c_N is the rated power of the cooling water pump; P ct is the cooling tower power; P ct_N is the rated power of the cooling tower; Modeling of the total thermal efficiency of the surface cooler: The surface cooler connects the air cycle with the chilled water cycle and does not consume any power. Its efficiency model is described. g The total heat exchange efficiency of the surface cooler is: Where T2 is the outlet air temperature of the surface cooler. If the temperature rise of the pipeline is not considered, then T2 = T S ; Considering the influence of fan temperature rise, T1 can be calculated as follows: T1=T N +ΔT fan Step 2: Under the premise of meeting the cooling load and maintaining the current temperature demand of the air conditioner, the total energy consumption of the air conditioner is minimized, and the optimal operating parameter combination of each device is sought according to the simulated annealing algorithm; The optimization objective function is defined as: minP=min(P fan +P e +P ch +P c +P ct ) The simulated annealing algorithm specifically includes the following steps: (1) Let T = T0 represent the initial temperature at the beginning of annealing, and randomly assign an initial solution P fan_0 ,P e_0 ,P ch_0 ,P c_0 ,P ct_0 As the solution to be optimized corresponding to the power of each device at the initial time T0, and calculate the corresponding objective function P(T0); (2) Let T t =kT (t-1) , where k is a value between 0 and 1, representing the rate at which the temperature drops, that is, the gradient of the drop; (3) For the current solution P fan_t ,P e_t ,P ch_t ,P c_t ,P ct_t Apply random perturbations to generate a new corresponding solution P in its neighborhood fan_(t+1) ,P e_(t+1) ,P ch_(t+1) ,P c_(t+1) ,P ct_(t+1) , and calculate the corresponding objective function value P(T t+1 ); (4) Calculate ΔP = P(T t+1 )-P(T t ); (5) If ΔP < 0, accept the new solution as the current solution, otherwise follow the probability e -ΔE / kT Decide whether to accept the new solution; (6) Repeat the perturbation and acceptance process multiple times at temperature T, i.e., execute (3) and (4); (7) Determine whether the temperature reaches the termination temperature level. If so, terminate the algorithm; otherwise, return to (2); Step 3: According to the best parameter combination obtained in step 2, adjust the parameter values ​​of the cooling and heating equipment inside the air-conditioning equipment to achieve energy-saving control of the air-conditioning equipment.

10. A data center low-load air conditioning energy-saving control system, characterized in that: A data center low-load air conditioning energy-saving control method according to any one of claims 1 to 9 is used to implement a data center low-load air conditioning energy-saving control system, and the data center low-load air conditioning energy-saving control system comprises: Host computer real-time monitoring interface: The host computer real-time monitoring interface uses the data center digital modeling method and the digital twin technology to visualize the operation status, temperature, air conditioning equipment operation status, power, energy consumption, and indoor temperature of each IT equipment in the data center. It is used to monitor the operation status, temperature changes, air conditioning operation status, and indoor temperature changes of each equipment in the data center in real time, and display them in real time in 3D simulation and digital form; Network data communication system: The network data communication system is applied to the network communication link construction method to realize data interaction and real-time communication between the control console, IT equipment, host computer real-time monitoring interface, air conditioning equipment and upper control center; Data sensing system: The data sensing system obtains equipment operation data in real time by installing various sensors on the equipment to support air conditioning energy-saving control; Temperature prediction system: The temperature prediction system is applied to the temperature change prediction method, including a data processing platform and a computing power support platform, equipped with a high-computing power GPU, to realize the preprocessing of data collected by the sensor system, data set preparation, and training and testing of neural network models; Equipment parameter control system: The equipment parameter control system is applied to the equipment operation parameter optimization control method to optimize the best combination of equipment operation parameters inside the air conditioner, and issue instructions to the control console to realize control; Control console: The control console adjusts and controls the relevant parameters of the air-conditioning equipment according to the control instructions received from the network data communication system.

Images