Data center temperature control method and system based on multilayer perceptron

Abstract

The invention belongs to the field of data center temperature control, and discloses a data center temperature control method and system based on a multi-layer perceptron, and the method comprises the steps: collecting the temperature, load and environment parameters of a data center cabinet, inputting a trained temperature control system prediction model, taking the multi-layer perceptron as a basic model for the temperature control system prediction model, and outputting the temperature set value of the refrigeration system and the deflection angle of the guide plate, and dynamically adjusting the temperature set value according to the preset cabinet target temperature to realize temperature control. The multi-layer perceptron is used for learning and modeling historical data, and the complex relation between each parameter and the control parameter of the refrigerating system is captured. The method is remarkable in effect, and compared with a traditional CFD method, calculation is more efficient, temperature changes can be rapidly responded, and the method is suitable for real-time control; compared with the existing deep learning application, through reasonable model architecture and training, the training time can be shortened, the parameter optimization difficulty can be reduced, and the prediction precision stability can be improved, so that the temperature control efficiency is improved, the energy consumption is reduced, and the method is more suitable for variable practical application environments.

Description

Technical Field

[0001] The present invention belongs to the technical field of data center temperature control, and specifically relates to a data center temperature control method and system based on a multi-layer perceptron. Background Art

[0002] With the rapid development of information technology, especially the surge in demand for big data, artificial intelligence, and large language models, data centers have become the core infrastructure supporting modern society's digital transformation and cloud computing services. However, while providing computing services for a large number of servers, data centers are facing increasingly prominent power consumption and heat issues. Due to the high density of equipment and the nonlinear variation in power consumption, data center temperature control systems face significant challenges. How to efficiently and stably control cabinet temperature to ensure stable equipment operation and reduce energy consumption has become a key issue that needs to be addressed.

[0003] Currently, traditional data center temperature control methods primarily rely on computational fluid dynamics (CFD) models, which accurately simulate fluid flow and heat transfer to predict and regulate temperature. However, CFD methods are computationally complex, slow to solve, and only produce numerical solutions at discrete points. This lack of real-time performance and versatility makes them unsuitable for real-time control scenarios. Especially in large-scale data centers, CFD models often fail to meet the dynamic temperature control requirements and cannot quickly respond to temperature changes, resulting in inefficient temperature control and excessive energy consumption.

[0004] To improve temperature control accuracy and computational efficiency, data center temperature control research has recently begun incorporating machine learning methods, particularly deep learning techniques. By learning from historical temperature control data, these models can achieve efficient temperature control through a data-driven approach without relying on complex physical calculations. However, existing deep learning applications still face challenges such as long model training time, difficulty in parameter optimization, and unstable prediction accuracy, hindering widespread adoption in the ever-changing real-world environments. Summary of the Invention

[0005] The present invention provides a data center temperature control method and system based on a multilayer perceptron. This method combines the characteristics of the multilayer perceptron (MLP) that can learn complex nonlinear relationships, overcomes the limitations of traditional temperature control methods, achieves high-precision and high-efficiency real-time temperature control, and reduces the energy consumption of the data center.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A data center temperature control method based on a multi-layer perceptron includes: Collect cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; Input the cabinet temperature, load, and environmental parameters of the data center to be temperature-controlled into the trained temperature control system prediction model, and output the temperature set value of the cooling system and the deflection angle of the cooling system's guide plate; Based on the pre-set cabinet target temperature, the temperature set point of the cooling system is dynamically adjusted to achieve temperature regulation in the data center; Among them, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the temperature setting value of the historical refrigeration system and the guide plate deflection angle of the historical refrigeration system.

[0007] Furthermore, the load includes computing load and power consumption data of the cabinet; and the environmental parameters include indoor humidity of the data center and air-conditioning outlet temperature in the data center or indoor temperature of the data center.

[0008] Furthermore, before collecting the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated, the method includes: The multi-layer perceptron will be used as the basic model to construct a temperature control system prediction model; The data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical cooling system temperature settings, and historical cooling system guide vane deflection angles are used as data sets to train and test the temperature control system prediction model to obtain a trained temperature control system prediction model. The multilayer perceptron includes at least one fully connected layer and one output layer; the output value of the output layer is processed by the following activation function:

[0009] Where, Indicates the output value, is the activation function, h is the hidden layer output, is the output layer weight, is the output layer bias.

[0010] Furthermore, the data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical temperature settings of the refrigeration system, and historical guide plate deflection angles of the refrigeration system are used as a data set, including: The collected historical cabinet temperatures, historical loads, historical environmental parameters, historical refrigeration system temperature settings, and historical refrigeration system guide plate deflection angles are standardized, missing values ​​are interpolated, and dimension reduction is performed; The historical cabinet temperature, historical load, historical environmental parameters, historical cooling system temperature setting values, and historical cooling system deflector deflection angles after standardization, missing value interpolation, and dimensionality reduction are used as the initial data set. The Latin hypercube sampling method is used to screen data samples from the initial data set, and the screened data samples are used as the data set of the temperature control system prediction model.

[0011] Furthermore, the input layer of the temperature control system prediction model includes cabinet temperature, load, and environmental parameters; the output layer includes the temperature setting value of the refrigeration system, the deflection angle of the guide plate of the refrigeration system, and the wind speed of the refrigeration system; the temperature control system prediction model is expressed as:

[0012] Where X represents the input data of the input layer; The temperature set point for the refrigeration system; is the wind speed of the refrigeration system; is the deflection angle of the guide plate of the refrigeration system; W and b represent the weight matrix and bias term of the temperature control system prediction model, respectively. is a nonlinear activation function and is trained using the back-propagation algorithm.

[0013] Furthermore, dynamically adjusting the temperature setting value of the refrigeration system based on the preset cabinet target temperature includes: Input the current cooling system temperature setting value and the preset cabinet target temperature into the pre-built dynamic adjustment model, and output the adjusted cooling system temperature setting value; The expression of the dynamic adjustment model is:

[0014] Where, is the temperature set point of the adjusted refrigeration system; The temperature setting value of the current refrigeration system; is the preset cabinet target temperature; λ is the preset adjustment coefficient.

[0015] A data center temperature control system based on a multi-layer perceptron includes: A data acquisition module, used to collect cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; The prediction module is used to input the cabinet temperature, load and environmental parameters of the data center to be temperature-controlled into the trained temperature control system prediction model, and output the temperature set value of the cooling system and the deflection angle of the cooling system's guide vanes; A dynamic adjustment module is used to dynamically adjust the temperature set point of the cooling system based on the preset cabinet target temperature to achieve temperature regulation in the data center; Among them, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the temperature setting value of the historical refrigeration system and the guide plate deflection angle of the historical refrigeration system.

[0016] Furthermore, it also includes: The data preprocessing module is used to standardize, interpolate missing values ​​and perform dimensionality reduction on the collected historical cabinet temperature, historical load, historical environmental parameters, historical temperature setting values ​​of the refrigeration system and historical guide plate deflection angles of the refrigeration system.

[0017] Furthermore, it also includes: The model training module is used to construct a temperature control system prediction model using a multi-layer perceptron as the basic model; The data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical cooling system temperature settings, and historical cooling system guide vane deflection angles are used as data sets to train and test the temperature control system prediction model to obtain a trained temperature control system prediction model. The multilayer perceptron includes at least one fully connected layer and one output layer; the output value of the output layer is processed by the following activation function:

[0018] Where, Indicates the output value, is the activation function, h is the hidden layer output, is the output layer weight, is the output layer bias.

[0019] Furthermore, it also includes: a feedback optimization module for dynamically updating the weight of the temperature control system prediction model according to the temperature setting value of the refrigeration system after dynamic adjustment. The specific formula is as follows:

[0020] Where, is the learning rate; is the gradient of the weight; is the loss function; is the updated weight; is the current weight; W is the weight coefficient; b is the bias.

[0021] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a data center temperature control method based on a multi-layer perceptron. This method first collects the data center cabinet temperature, load, and environmental parameters, and inputs a trained temperature control system prediction model. The temperature control system prediction model uses a multi-layer perceptron as a basic model and is trained based on multiple categories of historical data from the data center. The model outputs the temperature set point and deflection angle of the refrigeration system. The temperature set point is then dynamically adjusted based on the preset cabinet target temperature to achieve temperature control. The multi-layer perceptron is used to learn and model historical data, capturing the complex relationship between various parameters and the refrigeration system control parameters. This method is significantly effective. Compared with traditional CFD methods, it is more efficient in calculation, can quickly respond to temperature changes, and is suitable for real-time control. Compared with existing deep learning applications, through reasonable model architecture and training, it can shorten training time, reduce the difficulty of parameter optimization, and improve the stability of prediction accuracy, thereby improving temperature control efficiency, reducing energy consumption, and being more adaptable to changing actual application environments.

[0022] In the present invention, preferably, this method standardizes, interpolates missing values ​​and performs dimensionality reduction on historical data, and uses Latin hypercube sampling to screen samples, thereby improving data quality, reducing data redundancy and noise interference, making model training more efficient, and making prediction results more stable and reliable.

[0023] In the present invention, the method preferably also constructs a dynamic adjustment model that, by inputting a preset cabinet target temperature, outputs an adjusted temperature setpoint for the refrigeration system, and thus outputs a real-time optimal temperature control strategy. This enables the temperature control system to quickly adjust control parameters based on real-time conditions, improving the real-time performance and accuracy of temperature control and reducing energy consumption.

[0024] The present invention also provides a data center temperature control system based on a multi-layer perceptron. This system consists of data acquisition, prediction and dynamic adjustment modules. The data acquisition module obtains the temperature, load and environmental parameters of the data center cabinets; the prediction module inputs these parameters into a trained temperature control system prediction model based on a multi-layer perceptron. The model is trained with multiple categories of historical data from the data center and outputs the temperature setting value and the deflection angle of the guide plate of the refrigeration system; the dynamic adjustment module dynamically adjusts the temperature setting value according to the preset cabinet target temperature to achieve temperature control. This system uses a multi-layer perceptron to learn from historical data and explore the relationship between various parameters and refrigeration control parameters. The use of this system overcomes the problems of complex calculations and poor real-time performance of traditional CFD methods, and also solves the problems of long training time and unstable prediction accuracy of existing deep learning applications. It can quickly respond to temperature changes, improve temperature control efficiency, reduce energy consumption, and better meet the dynamic temperature control needs of data centers. It can be widely used in various practical application environments.

[0025] In the present invention, preferably, the data preprocessing module standardizes, interpolates missing values, and performs dimensionality reduction processing on the collected historical data, thereby improving data quality, reducing data redundancy and noise interference, and enabling subsequent model training to be based on higher-quality data, which helps to improve the accuracy and stability of the temperature control system prediction model.

[0026] In the present invention, preferably, the feedback optimization module dynamically updates the weights of the temperature control system prediction model according to the control parameters and control strategy of the refrigeration system, so that the model can self-adjust and optimize according to the actual operating conditions, continuously improve the prediction accuracy, better adapt to the ever-changing temperature control needs of the data center, further improve the performance and stability of the temperature control system, and reduce energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flow chart of a data center temperature control method based on a multi-layer perceptron provided in an embodiment of the present invention; Figure 2 A schematic diagram of a data center environment provided by an embodiment of the present invention; Figure 3 A schematic diagram of the structure of a data center temperature control system based on a multi-layer perceptron according to an embodiment of the present invention; Figure 4 A flow chart of a data center temperature control method based on a multi-layer perceptron provided by the present invention; Figure 5 This is a structural schematic diagram of a data center temperature control system based on a multi-layer perceptron provided by the present invention. DETAILED DESCRIPTION

[0028] like Figure 1 As shown, the embodiment provides a data center temperature control method based on a multi-layer perceptron, including a training process of a temperature control system prediction model, a process of predicting control parameters of a refrigeration system using the temperature control system prediction model, and a process of dynamically adjusting the control parameters of the final refrigeration system based on the prediction results, wherein: The training process of the temperature control system prediction model is as follows: S1: Collect historical data such as temperature, load, and environmental parameters of each cabinet in the data center to form a data set; S2: Standardize, interpolate missing values, and reduce dimensionality of the collected data set; S3: Build and train a temperature control system prediction model by learning the nonlinear relationship between cabinet temperature requirements and temperature control parameters through model training. The temperature control system prediction model's foundational model uses a multi-layer perceptron (MLP) trained based on the data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical cooling system temperature setpoints, and historical cooling system deflector angles. The temperature control system prediction model is used to predict the control parameters of the refrigeration system, as follows: S4: Collects real-time data on the temperature, load, and environmental parameters of each cabinet in the data center, inputs it into the trained temperature control system prediction model, and predicts the control parameters of the cooling system, including the temperature setpoint, wind speed, and deflector deflection angle of the cooling system. The dynamic adjustment process of the control parameters of the final refrigeration system according to the prediction results is as follows: S5: Based on the model prediction results, specifically the temperature set value of the cooling system and the real-time collected cabinet temperature value, the two are integrated and calculated to dynamically adjust the temperature set value of the cooling system to ensure temperature stability and optimize energy efficiency.

[0029] As a preferred solution of this embodiment, step S1 specifically involves collecting historical data such as the temperature, load, environmental parameters, refrigeration system temperature setpoints, and deflector angles of each cabinet in the data center to form a data set. The cabinet temperature data includes temperature values ​​collected by sensors within the cabinets, the load data includes the calculated load and / or power consumption data of each cabinet, and the environmental parameter data includes data (indoor temperature and indoor humidity) collected by air conditioning outlets or environmental parameter sensors within the data center. All data sets are acquired in real time by the data acquisition module and transmitted to the data processing module for further analysis.

[0030] As a preferred solution in this embodiment, step S2 specifically involves standardizing the collected dataset, interpolating missing values, and performing dimensionality reduction to ensure data quality and improve model training efficiency. To overcome the uneven data distribution and insufficient sample size issues encountered in traditional methods, this embodiment employs the Latin Hypercube Sampling (LHS) method, selecting a 1,000-sample set based on CFD model validation. The LHS method ensures a uniform distribution of sample points within the input variable space, ensuring comprehensive coverage of all variables, avoiding data imbalance, and improving the model's generalization and prediction accuracy. Furthermore, LHS sampling effectively reduces redundant computations, improves data utilization efficiency, and provides a more robust data foundation for subsequent training of the multi-layer perceptron (MLP) model.

[0031] As a preferred solution of this embodiment, step S3 is specifically as follows: by training a multi-layer perceptron (MLP) model, the nonlinear relationship between the cabinet temperature demand and the temperature control parameters is learned. In this step, by learning a large amount of historical temperature control data, the MLP can effectively extract the complex relationship between the cabinet temperature, load, environmental parameters and the air conditioning set value. Specifically, the input layer includes data such as temperature, load, and environmental parameters, and the output layer includes control parameters such as the temperature set value, wind speed, and deflection angle of the refrigeration system. The model parameters are optimized through the back propagation algorithm so that it can predict the optimal temperature control strategy in real time under a changing environment. The multi-layer perceptron includes at least one fully connected layer and one output layer; the output value of the output layer is processed by the following activation function:

[0032] Where, Indicates the output value, is the activation function, h is the hidden layer output, is the output layer weight, is the output layer bias.

[0033] As a preferred solution of this embodiment, step S4 specifically includes: performing real-time temperature control prediction based on a trained temperature control system prediction model. In this step, the system inputs the trained temperature control system prediction model based on the real-time collected data to obtain control parameters such as the temperature setting value, wind speed, and deflector deflection angle of the refrigeration system. The temperature control system prediction model is:

[0034] Where X represents the input data of the input layer, which represents input parameters such as cabinet temperature, load, and environmental parameters; The temperature set point for the refrigeration system; is the wind speed of the refrigeration system; is the deflection angle of the guide plate of the refrigeration system; W and b represent the weight matrix and bias term of the temperature control system prediction model, respectively. is a nonlinear activation function and is trained using the back-propagation algorithm.

[0035] As a preferred solution of this embodiment, step S5 specifically involves dynamically adjusting the operating parameters of the refrigeration system based on the model prediction results and the real-time collected cabinet temperature values. In this step, the system compares the predicted refrigeration system temperature setpoint with the actual collected cabinet temperature. Based on the difference between the two, the system adjusts parameters such as the refrigeration system temperature setpoint, air speed, and deflector deflection angle in real time to ensure stable cabinet temperature and optimize energy efficiency.

[0036]

[0037] Where, is the temperature set point of the adjusted refrigeration system; The temperature setting value of the current refrigeration system; is the preset cabinet target temperature; λ is the preset adjustment coefficient.

[0038] Through continuous feedback adjustment, the system can flexibly adjust the temperature control strategy according to changes in data center load and temperature fluctuations, achieving the dual goals of energy saving and precise temperature control.

[0039] Among them, step S1 builds a data center environment according to the actual scenario, the scenario example is as follows Figure 2 As shown, CFD (Computational Fluid Dynamics) simulations were used to predict and verify temperature distribution. CFD simulations analyzed temperature variations under different loads, cooling configurations, and air flow conditions. Specifically, each cabinet in the data center is equipped with a temperature sensor installed at the cabinet's air outlet to collect real-time outlet temperature. The temperature data collected by these sensors reflects the real-time temperature control status inside the cabinet and its surroundings. Humidity sensors are also installed within the data center to collect ambient humidity.

[0040] To ensure accuracy and comprehensiveness of temperature data, this embodiment also collects data on power consumption, load status, and environmental parameters within the cabinet. All data is transmitted to a central data processing system via wireless or wired communication, serving as a basis for optimizing and adjusting the temperature control system.

[0041] like Figure 3 As shown, this embodiment also provides a data center temperature control system based on a multi-layer perceptron, which is used to implement the steps of the above-mentioned temperature control method, including a data acquisition module, a data preprocessing module, a multi-layer perceptron (MLP) model, a prediction and optimization module, a control module, and a display and feedback module. Among them: the data acquisition module is used to collect temperature, power consumption, and environmental parameter data of each cabinet in the data center in real time; the data preprocessing module is used to perform standardization, missing value interpolation, and dimensionality reduction on the collected data; the multi-layer perceptron (MLP) model is used to learn the nonlinear mapping relationship between cabinet temperature requirements and temperature control parameters; the prediction and optimization module is used to predict and optimize the temperature setpoint and wind speed of the cooling system in real time based on the trained MLP model; the control module is used to adjust the operating parameters of the cooling system based on the prediction results to achieve temperature control of each cabinet in the data center; and the display and feedback module is used to display the temperature control results in real time and adjust the control strategy based on the actual temperature control requirements.

[0042] As a preferred solution of this embodiment, the data acquisition module further includes a temperature sensor, a power sensor and an environmental sensor, which are respectively used to collect cabinet interior, equipment power consumption and environmental parameter data; and also includes a humidity sensor to provide more comprehensive environmental data.

[0043] As a preferred solution in this embodiment, a multi-layer perceptron model serves as the core intelligent computing module. Using real-time data, it learns the nonlinear relationship between cabinet temperature and temperature control parameters. Based on historical data and real-time feedback, the model runs on a high-performance computing platform, inferring and predicting outputs using real-time input data.

[0044] As a preferred solution of this embodiment, the prediction and optimization module calculates and optimizes the temperature setting value, wind speed and guide plate deflection angle of the refrigeration system in real time based on the prediction results of the MLP model, ensuring that the temperature of each cabinet is controlled within the set range and optimizing energy efficiency.

[0045] As a preferred solution of this embodiment, the control module adjusts the refrigeration system's control parameters, such as the temperature setting value, wind speed, and deflector deflection angle, based on the output values ​​of the prediction and optimization module. This module implements real-time adjustment of the refrigeration system through hardware.

[0046] As a preferred solution of this embodiment, the display and feedback module is a user interface that displays temperature control results, system status, and adjustment suggestions in real time. This module also provides feedback to the user on deviations and optimization needs during the temperature control process, allowing for further adjustments to the control strategy to ensure the temperature control system remains efficient during actual operation.

[0047] As a preferred solution of this embodiment, the feedback optimization module is used to dynamically update the weights of the temperature control system prediction model based on the control parameters and control strategy of the refrigeration system. By adjusting the operating parameters and control strategy of the air-conditioning system, it can adapt to the dynamically changing temperature control requirements and environmental conditions. The specific formula is as follows:

[0048] Where, is the learning rate; is the gradient of the weight; is the loss function; is the updated weight; is the current weight; W is the weight coefficient; b is the bias.

[0049] As a preferred solution of this embodiment, the data acquisition module and each control module exchange data through wired or wireless communication to ensure the real-time performance of the temperature control system.

[0050] As a preferred solution of this embodiment, during the control process of the refrigeration system, an adaptive optimization algorithm is used to dynamically correct the output of the MLP model. The specific optimization strategies include: using a weighted loss function to improve the model's sensitivity to abnormal temperature points; introducing time series analysis and combining historical temperature control data to improve temperature control accuracy; and combining reinforcement learning methods to achieve autonomous adjustment of temperature control strategies.

[0051] According to the air conditioning performance, the output of the temperature control model is modified, and the optimization formula is as follows:

[0052] Where, Set the temperature of the refrigeration system for the final output; The temperature control value of the refrigeration system predicted by the temperature control system prediction model; is the weighted average of historical temperature control data; The outlet temperature setpoints of each cabinet are input to the MLP model; Provide real-time feedback of outlet temperature for each cabinet; 、 、 is the weight parameter, which is determined by optimization calculation.

[0053] It should be noted that the above formula, as an example, applies to the control scenario of a single cooling system. In actual applications, all relevant variables should be expressed in matrix form. The size of the matrix depends on the layout and number of equipment and cooling systems in the data center. The number of rows and columns in these matrices directly reflects the relationship between the cabinets and cooling units in the data center, ensuring that the temperature control strategy can accurately adapt to different space structures and cooling requirements.

[0054] For example, Figure 4 As shown, this embodiment provides a data center temperature control method based on a multi-layer perceptron, including the following steps: Collect cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; Input the cabinet temperature, load, and environmental parameters of the data center to be temperature-controlled into the trained temperature control system prediction model, and output the temperature set value of the cooling system and the deflection angle of the cooling system's guide plate; Based on the pre-set cabinet target temperature, the temperature set point of the cooling system is dynamically adjusted to achieve temperature regulation in the data center; Among them, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the temperature setting value of the historical refrigeration system and the guide plate deflection angle of the historical refrigeration system.

[0055] In this embodiment, the load includes the computing load and power consumption data of the cabinet; the environmental parameters include the indoor humidity of the data center and the air-conditioning outlet temperature in the data center or the indoor temperature of the data center.

[0056] In this embodiment, before collecting the cabinet temperature, load, and environmental parameters of the data center to be temperature-regulated, the following steps are included: The multi-layer perceptron will be used as the basic model to construct a temperature control system prediction model; The data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical cooling system temperature settings, and historical cooling system guide vane deflection angles are used as data sets to train and test the temperature control system prediction model to obtain a trained temperature control system prediction model. The multilayer perceptron includes at least one fully connected layer and one output layer; the output value of the output layer is processed by the following activation function:

[0057] Where, Indicates the output value, is the activation function, h is the hidden layer output, is the output layer weight, is the output layer bias.

[0058] In this embodiment, the data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical refrigeration system temperature settings, and historical refrigeration system deflector angles are used as a data set, including: The collected historical cabinet temperatures, historical loads, historical environmental parameters, historical refrigeration system temperature settings, and historical refrigeration system guide plate deflection angles are standardized, missing values ​​are interpolated, and dimension reduction is performed; The historical cabinet temperature, historical load, historical environmental parameters, historical cooling system temperature setting values, and historical cooling system deflector deflection angles after standardization, missing value interpolation, and dimensionality reduction are used as the initial data set. The Latin hypercube sampling method is used to screen data samples from the initial data set, and the screened data samples are used as the data set of the temperature control system prediction model.

[0059] In this embodiment, the input layer of the temperature control system prediction model includes the cabinet temperature, load, and environmental parameters; the output layer includes the temperature setting value of the refrigeration system, the deflection angle of the refrigeration system's guide plate, and the wind speed of the refrigeration system; the temperature control system prediction model is expressed as:

[0060] Where X represents the input data of the input layer; The temperature set point for the refrigeration system; is the wind speed of the refrigeration system; is the deflection angle of the guide plate of the refrigeration system; W and b represent the weight matrix and bias term of the temperature control system prediction model, respectively. is a nonlinear activation function and is trained using the back-propagation algorithm.

[0061] In this embodiment, dynamically adjusting the temperature setting value of the refrigeration system based on the preset cabinet target temperature includes: Input the current cooling system temperature setting value and the preset cabinet target temperature into the pre-built dynamic adjustment model, and output the adjusted cooling system temperature setting value; The expression of the dynamic adjustment model is:

[0062] Where, is the temperature set point of the adjusted refrigeration system; The temperature setting value of the current refrigeration system; is the preset cabinet target temperature; λ is the preset adjustment coefficient.

[0063] like Figure 5 As shown, this embodiment also provides a data center temperature control system based on a multi-layer perceptron, including: a data acquisition module for collecting the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; a prediction module for inputting the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated into a trained temperature control system prediction model, and outputting the temperature setting value of the refrigeration system and the deflection angle of the guide plate of the refrigeration system; a dynamic adjustment module for dynamically adjusting the temperature setting value of the refrigeration system based on a preset cabinet target temperature to achieve temperature regulation of the data center; wherein, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the historical temperature setting value of the refrigeration system and the historical deflection angle of the refrigeration system. The present invention also provides an electronic device, comprising: a memory for storing a computer program; and a processor for implementing the steps of the data center temperature control method based on a multi-layer perceptron when executing the computer program.

[0064] When the processor executes the computer program, the steps of the above-mentioned data center temperature control based on the multi-layer perceptron are implemented, for example: collecting the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; inputting the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated into a trained temperature control system prediction model, and outputting the temperature setting value of the refrigeration system and the deflection angle of the guide plate of the refrigeration system; based on the preset cabinet target temperature, dynamically adjusting the temperature setting value of the refrigeration system to achieve temperature regulation of the data center; wherein, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the historical temperature setting value of the refrigeration system and the historical deflection angle of the refrigeration system.

[0065] Alternatively, when the processor executes the computer program, it implements the functions of each module in the above-mentioned system, for example: a data acquisition module, used to collect the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; a prediction module, used to input the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated into a trained temperature control system prediction model, and output the temperature setting value of the refrigeration system and the guide plate deflection angle of the refrigeration system; a dynamic adjustment module, used to dynamically adjust the temperature setting value of the refrigeration system based on a preset cabinet target temperature to achieve temperature regulation of the data center; wherein, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the historical temperature setting value of the refrigeration system and the historical guide plate deflection angle of the refrigeration system.

[0066] Exemplarily, the computer program can be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of completing preset functions, and the instruction segments are used to describe the execution process of the computer program in the data center temperature control device based on the multi-layer perceptron. For example, the computer program can be divided into a data acquisition module, a prediction module, and a dynamic adjustment module; the data acquisition module is used to collect the cabinet temperature, load, and environmental parameters of the data center to be temperature-controlled; the prediction module is used to input the cabinet temperature, load, and environmental parameters of the data center to be temperature-controlled into a trained temperature control system prediction model and output the temperature set value of the cooling system and the deflector deflection angle of the cooling system; the dynamic adjustment module is used to dynamically adjust the temperature set value of the cooling system based on a preset cabinet target temperature to achieve temperature control of the data center; wherein the basic model of the temperature control system prediction model adopts a multi-layer perceptron and is trained based on the historical cabinet temperature, historical load, historical environmental parameters, historical cooling system temperature set value, and historical cooling system deflector deflection angle.

[0067] The multi-layer perceptron-based data center temperature control device can be a computing device such as a desktop computer, laptop, PDA, or cloud server. The multi-layer perceptron-based data center temperature control device can include, but is not limited to, a processor and memory. Those skilled in the art will appreciate that the above examples of multi-layer perceptron-based data center temperature control devices do not limit the full scope of multi-layer perceptron-based data center temperature control devices. These devices can include more components than those described above, or combinations of certain components, or different components. For example, the multi-layer perceptron-based data center temperature control device can also include input / output devices, network access devices, buses, and the like.

[0068] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The processor serves as the control center of the multilayer perceptron-based data center temperature control system, connecting various components of the multilayer perceptron-based data center temperature control system using various interfaces and circuits.

[0069] The memory can be used to store the computer program and / or module, and the processor implements various functions of the data center temperature control device based on the multi-layer perceptron by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory.

[0070] The memory may primarily include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as sound playback or image playback); the data storage area may store data generated based on the use of the mobile phone (such as audio data and a phone book). Furthermore, the memory may include high-speed random access memory (RAM) and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0071] The present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of the data center temperature control method based on a multi-layer perceptron.

[0072] If the integrated module / unit of the data center temperature control system based on the multi-layer perceptron is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0073] Based on this understanding, the present invention can implement all or part of the processes in the aforementioned multi-layer perceptron-based data center temperature control method by using a computer program to instruct related hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the aforementioned multi-layer perceptron-based data center temperature control method. The computer program includes computer program code, which can be in source code form, object code form, executable file, or a pre-defined intermediate form.

[0074] The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0075] It should be noted that the content contained in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practices, computer-readable storage media do not include electrical carrier signals and telecommunication signals.

[0076] In summary, the present invention provides a data center temperature control method and system based on a multi-layer perceptron, which has the following advantages: First, the present invention realizes the precise mapping between temperature control parameters and cabinet temperature through a temperature control system based on a multi-layer perceptron (MLP) model, and can provide a fast and accurate temperature control solution in a large-scale data center environment.

[0077] Second, the present invention can dynamically optimize the temperature control parameters of the refrigeration system based on real-time environmental data and prediction results, which not only improves the accuracy of temperature control, but also significantly reduces energy consumption, achieving the effect of energy conservation and emission reduction.

[0078] Third, the temperature control system of the present invention has high scalability and flexibility, can adapt to the temperature control needs of data centers of different sizes and types, and has broad engineering application prospects.

[0079] The above embodiment is only one of the implementation methods that can realize the technical solution of the present invention. The scope of protection claimed by the present invention is not limited only to this embodiment, but also includes changes, replacements and other implementation methods that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention.

[0080] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered by the scope of protection of the claims of the present invention.

Claims

1. A data center temperature control method based on a multi-layer perceptron, characterized in that: include: Collect cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; Input the cabinet temperature, load, and environmental parameters of the data center to be temperature-controlled into the trained temperature control system prediction model, and output the temperature set value of the cooling system and the deflection angle of the cooling system's guide plate; Based on the pre-set cabinet target temperature, the temperature set point of the cooling system is dynamically adjusted to achieve temperature regulation in the data center; Among them, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the temperature setting value of the historical refrigeration system and the guide plate deflection angle of the historical refrigeration system.

2. The data center temperature control method based on a multi-layer perceptron according to claim 1 is characterized in that: The load includes the computing load and power consumption data of the cabinet; the environmental parameters include the indoor humidity of the data center and the air-conditioning outlet temperature in the data center or the indoor temperature of the data center.

3. The data center temperature control method based on multi-layer perceptron according to claim 1 is characterized in that: Before collecting the cabinet temperature, load and environmental parameters of the data center to be temperature-regulated, the method includes: The multi-layer perceptron will be used as the basic model to construct a temperature control system prediction model; The data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical cooling system temperature settings, and historical cooling system guide vane deflection angles are used as data sets to train and test the temperature control system prediction model to obtain a trained temperature control system prediction model. The multilayer perceptron includes at least one fully connected layer and one output layer; the output value of the output layer is processed by the following activation function: Where, Indicates the output value, is the activation function, h is the hidden layer output, is the output layer weight, is the output layer bias.

4. The data center temperature control method based on a multi-layer perceptron according to claim 3 is characterized in that: The data set includes: historical cabinet temperatures, historical loads, historical environmental parameters, historical temperature settings of the cooling system, and historical deflector deflection angles of the cooling system. The collected historical cabinet temperatures, historical loads, historical environmental parameters, historical refrigeration system temperature settings, and historical refrigeration system guide plate deflection angles are standardized, missing values ​​are interpolated, and dimension reduction is performed; The historical cabinet temperature, historical load, historical environmental parameters, historical cooling system temperature setting values, and historical cooling system deflector deflection angles after standardization, missing value interpolation, and dimensionality reduction are used as the initial data set. The Latin hypercube sampling method is used to screen data samples from the initial data set, and the screened data samples are used as the data set of the temperature control system prediction model.

5. The data center temperature control method based on multi-layer perceptron according to claim 1 is characterized in that: The input layer of the temperature control system prediction model includes cabinet temperature, load, and environmental parameters; the output layer includes the temperature setting value of the refrigeration system, the deflection angle of the refrigeration system's guide plate, and the wind speed of the refrigeration system; the temperature control system prediction model is expressed as: Where X represents the input data of the input layer; The temperature set point for the refrigeration system; is the wind speed of the refrigeration system; is the deflection angle of the guide plate of the refrigeration system; W and b represent the weight matrix and bias term of the temperature control system prediction model, respectively. is a nonlinear activation function and is trained using the back-propagation algorithm.

6. The data center temperature control method based on a multi-layer perceptron according to claim 1 is characterized in that: The dynamically adjusting the temperature setting value of the refrigeration system based on the preset cabinet target temperature includes: Input the current cooling system temperature setting value and the preset cabinet target temperature into the pre-built dynamic adjustment model, and output the adjusted cooling system temperature setting value; The expression of the dynamic adjustment model is: Where, is the temperature set point of the adjusted refrigeration system; The temperature setting value of the current refrigeration system; is the preset cabinet target temperature; λ is the preset adjustment coefficient.

7. A data center temperature control system based on a multi-layer perceptron, characterized in that: include: A data acquisition module, used to collect cabinet temperature, load and environmental parameters of the data center to be temperature-regulated; The prediction module is used to input the cabinet temperature, load and environmental parameters of the data center to be temperature-controlled into the trained temperature control system prediction model, and output the temperature set value of the cooling system and the deflection angle of the cooling system's guide vanes; A dynamic adjustment module is used to dynamically adjust the temperature set point of the cooling system based on the preset cabinet target temperature to achieve temperature regulation in the data center; Among them, the basic model of the temperature control system prediction model adopts a multi-layer perceptron, which is trained based on the historical cabinet temperature, historical load and historical environmental parameters of the data center, the temperature setting value of the historical refrigeration system and the guide plate deflection angle of the historical refrigeration system.

8. The data center temperature control system based on the multi-layer perceptron according to claim 7 is characterized in that: Also includes: The data preprocessing module is used to standardize, interpolate missing values ​​and perform dimensionality reduction on the collected historical cabinet temperature, historical load, historical environmental parameters, historical temperature setting values ​​of the refrigeration system and historical guide plate deflection angles of the refrigeration system.

9. The data center temperature control system based on a multi-layer perceptron according to claim 7, characterized in that: Also includes: The model training module is used to construct a temperature control system prediction model using a multi-layer perceptron as the basic model; The data center's historical cabinet temperatures, historical loads, historical environmental parameters, historical cooling system temperature settings, and historical cooling system guide vane deflection angles are used as data sets to train and test the temperature control system prediction model to obtain a trained temperature control system prediction model. The multilayer perceptron includes at least one fully connected layer and one output layer; the output value of the output layer is processed by the following activation function: Where, Indicates the output value, is the activation function, h is the hidden layer output, is the output layer weight, is the output layer bias.

10. The data center temperature control system based on a multi-layer perceptron according to claim 7, characterized in that: Also includes: The feedback optimization module is used to dynamically update the weight of the temperature control system prediction model according to the temperature set value of the dynamically adjusted refrigeration system. The specific formula is as follows: Where, is the learning rate; is the gradient of the weight; is the loss function; is the updated weight; is the current weight; W is the weight coefficient; b is the bias.

Images