Convolutional neural network-based method for fast prediction of temperature field in data center room

Abstract

This invention discloses a method for rapid prediction of data center temperature fields based on convolutional neural networks, comprising: acquiring the geometric information of the data center structure; performing geometric modeling and mesh generation on the data center structure to obtain the mesh structure of the data center; acquiring the operating parameters of the equipment within the data center; acquiring the temperature field data of the data center and dividing the temperature field data according to a height plane to generate the output dataset for training the convolutional neural network model; converting the mesh structure of the data center at different heights into a spatial feature map structure and recording the corresponding height information and operating parameter information to generate the input dataset for training the convolutional neural network model; and using the trained convolutional neural network model for temperature field prediction. This method can achieve accurate and reliable rapid prediction of the temperature field of data center rooms.

Description

Technical Field

[0001] The invention belongs to the field of building simulation and involves data preprocessing technology based on spatial feature maps and black-box modeling technology based on convolutional neural networks. Specifically, it involves a method for rapid prediction of temperature field in data center computer rooms based on convolutional neural networks. Background Technology

[0002] With the rapid development of the information society and the resulting surge in demand for computing and storage resources, the energy consumption of data centers has increased significantly in recent years, becoming a major component of global energy consumption. Improving the operational energy efficiency of data centers and achieving energy conservation and emission reduction is crucial. Currently, the energy consumption of data centers mainly consists of IT equipment, cooling systems, power distribution, lighting, and other auxiliary systems. Nearly 50% of the energy consumption is used for cooling systems, which provide cold airflow to cool IT equipment. Therefore, data center cooling systems possess enormous energy-saving potential. Developing reasonable cooling system control strategies to achieve high efficiency and energy saving while ensuring the safe and reliable operation of data centers is a key step in achieving energy conservation and carbon reduction in data centers.

[0003] Accurate temperature field calculations in data center computer rooms are crucial for energy-efficient operation of air conditioning and control of localized hotspots in server racks. Currently, numerical calculation methods based on physical equations are widely used for temperature field calculations in data center computer rooms, offering high accuracy and broad applicability. However, these physical methods are computationally expensive and heavily reliant on the combined expertise of data center operations personnel and building thermal environment simulation specialists, thus failing to meet the demands for real-time, precise energy-saving control based on temperature fields.

[0004] To address the above issues, it is necessary to improve the efficiency of data center temperature field calculations while maintaining accuracy, thereby ensuring energy-saving operation and the implementation of control strategies. Improved numerical calculation methods and data-driven modeling methods are two common methods for accelerating computation. Improved numerical calculation methods aim to simplify traditional numerical calculation models by making certain assumptions. This method is faster than traditional numerical calculation methods, but it is limited by the fundamental principles of numerical calculation, and its computational speed still cannot meet the requirements of real-time control. Data-driven modeling methods aim to train a black-box model based on limited numerical calculation results to learn the statistical mapping relationship between relevant physical variables and the temperature field within the data center. This method has a fast prediction speed, greatly reducing time costs, but it is currently only applicable to temperature point prediction at a specified location. When predicting the temperature field in a two-dimensional plane, the accuracy is insufficient. Therefore, achieving rapid and high-precision prediction of the temperature field in data center rooms is of great significance. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a method for rapid prediction of temperature field in data center computer rooms based on convolutional neural networks. This method can effectively construct spatial feature information of data centers that affect temperature field distribution and reduce the training difficulty of data-driven models.

[0006] The objective of this invention is achieved through the following technical solution: a method for rapid prediction of temperature field in data center computer rooms based on convolutional neural networks, comprising:

[0007] Obtain the geometric information of the data center computer room structure, perform geometric modeling and mesh generation on the data center computer room structure, and obtain the mesh structure of the data center computer room;

[0008] Obtain operating parameters of equipment within the data center server room;

[0009] Acquire temperature field data of the data center server room, divide the temperature field data according to the height plane, and generate the output of the dataset for training the convolutional neural network model;

[0010] The grid structure of the data center server room at different heights is converted into a spatial feature map structure, and the corresponding height information and operating condition parameter information are recorded to generate the input of the dataset used to train the convolutional neural network model.

[0011] The trained convolutional neural network model is used for temperature field prediction.

[0012] Furthermore, the data center server room is divided into grids, including:

[0013] The finite element method was used to perform unstructured mesh generation for the data center computer room.

[0014] Furthermore, the process of converting the grid structure at different heights of the data center server room into a spatial feature map structure includes:

[0015] Orthogonal resampling is performed on the mesh structure to convert the unstructured mesh into an orthogonal mesh with a fixed resolution. Spatial feature attributes are assigned to each mesh cell, and the entire mesh is converted into a spatial feature map.

[0016] Furthermore, acquiring the temperature field data of the data center computer room includes:

[0017] The temperature field of the data center is numerically calculated using the finite element method to obtain the temperature field data of the data center.

[0018] Furthermore, the temperature field data of the data center computer room is divided into height planes and then normalized, and this normalization is used as the model output for the training process.

[0019] Furthermore, the convolutional neural network model includes: a convolutional neural network, a fully connected layer, and a transposed convolutional neural network; the input of the convolutional neural network is a spatial feature map, the output of the convolutional neural network and operating parameters are the inputs of the fully connected layer, the output of the fully connected layer is the input of the transposed convolutional neural network, and the output of the transposed convolutional neural network is the temperature distribution of the data center server room. This method combines the spatial feature extraction capability of the convolutional neural network with the ability of the fully connected layer to fuse highly nonlinear relationships. It uses the convolutional neural network to process spatial feature information and the artificial neural network to process operating condition data, aiming to achieve rapid and accurate prediction of multi-planar temperature fields in data centers.

[0020] Furthermore, the geometric information of the data center includes the shape and size of the enclosure structure, and the shape, size, and placement of the server racks, terminal air conditioners, and interior building structure within the server room.

[0021] Furthermore, the operating parameters within the data center server room include the power of the server racks, the air supply temperature of the air conditioning terminals, and the air supply speed.

[0022] The present invention also provides a data center computer room temperature field rapid prediction device based on convolutional neural network, including one or more processors, for implementing the above-mentioned data center computer room temperature field rapid prediction method based on convolutional neural network.

[0023] The present invention also provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, is used to implement the above-described method for rapid prediction of temperature field in a data center based on a convolutional neural network.

[0024] Compared with the prior art, the beneficial effects of the present invention are:

[0025] Existing data-driven methods for predicting data center temperature fields only use one-dimensional vectors or two-dimensional images of the data center as input. One-dimensional vectors cannot contain spatial information, while two-dimensional images contain too much redundant spatial information, making it difficult for data-driven models to extract effective features and thus hindering accurate temperature field predictions. To address this issue, this invention proposes a spatial feature transformation technique based on spatial feature maps. By converting a complex two-dimensional plane into an orthogonal spatial feature map mesh, this solves the input format requirements of convolutional neural networks and improves the model's efficiency in utilizing input information, enabling subsequent data-driven models to achieve more accurate predictions.

[0026] To address the low accuracy of current data-driven data center temperature prediction methods in predicting two-dimensional temperature fields, this invention proposes an improved convolutional neural network framework. This framework processes spatial features through a convolutional neural network, handles operational features through fully connected layers, and then uses a transposed convolutional neural network to predict the two-dimensional temperature field of the data center. This effectively reduces redundant information and extracts useful spatial and operational information, achieving both efficient network training and improved accuracy in predicting the two-dimensional temperature field.

[0027] Furthermore, by introducing the height plane information of the data center into the fully connected layer, this invention enables the model to be trained on large datasets with temperature fields at different height planes and to predict the temperature field at any specified height plane, thus providing richer temperature field information for data center data centers.

[0028] In summary, this invention solves the problem of accurate prediction of multi-plane temperature fields in data center computer rooms by proposing a new feature engineering and data-driven model framework, utilizing the spatial feature extraction capability of convolutional neural networks and the ability of fully connected layers to fuse highly nonlinear relationships. This not only improves the efficiency and accuracy of temperature field prediction, but also provides a scientific basis for the comprehensive optimization of airflow organization in data center computer rooms and the real-time operation of refined energy-saving control systems, demonstrating significant practical value and potential for widespread application. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1 A flowchart of the prediction method provided by the present invention;

[0031] Figure 2 This is a schematic diagram of the geometric model of a data center server room.

[0032] Figure 3 A schematic diagram of the temperature field at different heights in a data center server room;

[0033] Figure 4 This is a schematic diagram of the principle of spatial feature map technology.

[0034] Figure 5 This is a schematic diagram showing the transformation results of planar spatial feature maps at different heights in a data center server room.

[0035] Figure 6 An architecture diagram for improving the convolutional neural network framework;

[0036] Figure 7 To verify the prediction results visualized in a centralized manner;

[0037] Figure 8 This is a schematic diagram of a hardware structure provided for an embodiment of the present invention. Detailed Implementation

[0038] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are implemented based on the technical solution of the present invention, and provide detailed implementation methods and specific operation processes. However, the scope of protection of the present invention is not limited to the following embodiments. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0039] like Figure 1 As shown, the present invention provides a method for rapid prediction of temperature field in a data center based on a convolutional neural network, comprising the following steps:

[0040] (1) Obtain the geometric information of the data center computer room structure, perform geometric modeling and mesh generation on the data center computer room structure, and obtain the mesh structure of the data center computer room;

[0041] In one embodiment, the geometric information of the data center includes the shape and size of the enclosure structure, and the shape, size, and placement of the server racks, terminal air conditioners, and interior building structures within the data center.

[0042] In one embodiment, the data center server room is meshed, including: using the finite element method to perform unstructured meshing of the data center server room.

[0043] (2) Obtain the operating parameters of the equipment in the data center computer room;

[0044] In one embodiment, the operating parameters in the data center server room include the power of the server racks, the air supply temperature of the air conditioning terminals, and the air supply speed.

[0045] (3) Obtain temperature field data of the data center computer room, divide the temperature field data according to the height plane, and generate the output of the dataset for training the convolutional neural network model;

[0046] Specifically, the temperature field data of the data center computer room is divided into height planes and then normalized, and this is used as the model output for the training process;

[0047] (4) Convert the grid structure of the data center computer room at different heights into a spatial feature map structure, and record the corresponding height information and operating condition parameter information to generate the input of the dataset for training the convolutional neural network model;

[0048] In one embodiment, converting the grid structure of different heights in the data center server room into a spatial feature map structure includes: orthogonally resampling the grid structure, converting the unstructured grid into an orthogonal grid with a fixed resolution, assigning spatial feature attributes to each grid cell, and converting the entire grid into a spatial feature map.

[0049] (5) The trained convolutional neural network model is used for temperature field prediction.

[0050] Specifically, this also includes optimizing the hyperparameters of the convolutional neural network to obtain a set of optimal hyperparameter values; the hyperparameters include the number of convolutional / transposed convolutional layers, the size, stride, and padding of each convolutional kernel, the number of channels in the convolutional layer, the number of fully connected layers, the number of neurons in the hidden layer, and the learning rate of the model.

[0051] During training, the dataset used to train the convolutional neural network is randomly divided into a training set and a validation set at a certain ratio. The stochastic gradient descent algorithm is used on the training set to train the parameters of the convolutional neural network, including weights and biases, to ensure its accuracy on the validation set and to quickly predict the temperature field of the data center server room.

[0052] In one embodiment, the convolutional neural network model includes: a convolutional neural network, a fully connected layer, and a transposed convolutional neural network. The input of the convolutional neural network is a spatial feature map, the output of the convolutional neural network and operating parameters are the inputs of the fully connected layer, the output of the fully connected layer is the input of the transposed convolutional neural network, and the output of the transposed convolutional neural network is the temperature distribution of the data center. This model structure adds a fully connected layer with feature concatenation to a traditional convolutional neural network, thereby ensuring that the model can simultaneously process spatial information and equipment operating condition information. This method optimizes the training process of the convolutional neural network by using a data center spatial feature map and a fully connected layer with feature concatenation, thereby achieving accurate and reliable rapid prediction of the temperature field of the data center.

[0053] Example 1:

[0054] This invention uses a large data center as the research object to verify the performance of the proposed data center temperature field prediction method.

[0055] Step one is to represent the geometric spatial characteristics of the data center server room. First, the geometric information of the server room's spatial structure is obtained. The server room is 29m long, 28m wide, and 5.5m high, containing 10 micro-modules. Each micro-module contains 27-28 server racks. The air conditioning uses a horizontal diffusion air supply system with air-wall AHUs (Air Handling Units). The hot aisles on the sides of the server racks are sealed. The AHUs diffuse cool air to the cold aisle (room cold pool), eliminating the need for a raised floor. The cool air carries away heat from the server racks and is exhausted from the rear of the racks to the sealed hot aisle, then returns to the rear return air vents of the air conditioning units via hot return air ducts (ceiling).

[0056] Then, operating parameters were acquired, including rack power, air conditioning set temperature, and air conditioning airflow velocity. The rack power and air conditioning set temperature were set values, while the air conditioning airflow velocity adaptively met the cooling load requirements of the data center during numerical calculations and was acquired after the calculations. To generate a temperature field dataset for training the neural network, the operating parameters needed to be diverse. In this study, the rack power was set to vary between 1 and 6 kW, and the air conditioning set temperature to vary between 18 and 27°C, generating a total of 60 operating condition combinations.

[0057] Then, geometric modeling and mesh generation are performed, resulting in a geometric model of the data center computer room as follows: Figure 2 As shown, the data center server room space is divided into unstructured meshes using the finite element method. In this embodiment, a total of 4,801,752 mesh cells are divided, with a minimum mesh size of 0.01m.

[0058] Then, data center temperature field data is generated. Based on the finite element method, different boundary conditions are obtained by changing the operating parameters in the established data center geometric model mesh. Numerical calculations are then performed based on these different boundary conditions to obtain the data center temperature field dataset. The temperature field data is then divided according to height planes to generate the output dataset for training convolutional neural networks. In this embodiment, a height plane is extracted every 0.5m, resulting in 12 temperature field samples from each height plane for each operating condition combination. Figure 3 As shown.

[0059] Finally, a grid-to-spatial feature map conversion is performed, transforming the grid structure at different heights in the data center server room into a spatial feature map structure, and recording the corresponding height and operating condition information to generate the input dataset for training the convolutional neural network. A schematic diagram of the spatial feature map technology is shown below. Figure 4 As shown, in this embodiment, orthogonal resampling is first performed on the grid at a specified height. In this example, the sampling resolution is 0.2m. Then, each resampled grid cell is assigned a value according to the actual physical meaning of the computer room's geometric space. In this example, solids (including cabinets, air conditioners, ceilings, columns, baffles, and cabling) are assigned a value of 1, and watersheds are assigned a value of 0. The spatial feature map division results of different planes in this example are shown below. Figure 5 As shown, the white area is assigned a value of 1, and the black area is assigned a value of 0.

[0060] Step two is neural network training. First, data preprocessing is performed, using min-max normalization to normalize the temperature values ​​of all two-dimensional temperature field samples to the range [0, 1].

[0061] Then, the model structure is built, as follows: Figure 6 As shown, the proposed model includes model input, convolutional neural network, fully connected layer, transposed convolutional neural network and model output.

[0062] 1) In convolutional neural networks, convolutional layers effectively capture local spatial features by applying multiple convolutional kernels (or filters) to the input data. Pooling layers reduce the dimensionality of the data through downsampling operations (such as max pooling or average pooling), reducing computational cost and preventing overfitting. In the neural network framework proposed in this invention, the input to the convolutional network is a spatial feature map representing the spatial information of the computer room. Processing this feature map through convolutional layers effectively captures the latent spatial features of the computer room. Finally, the output data serves as the input to the fully connected layers.

[0063] 2) The fully connected layer consists of a large number of fully connected artificial neurons, which are connected by weights to form a complex network structure. Each neuron in the neural network receives multiple input signals and calculates the output signal through a weighted summation activation function, thereby realizing the nonlinear transformation of information. In the neural network framework proposed in this invention, the input layer of the fully connected layer is a multidimensional vector obtained by concatenating the output vector of the convolutional network with operating parameters (air conditioning temperature, cabinet power, plane height), which is used to represent the integration result of system spatial information and operating condition information. By performing high-dimensional nonlinear mapping on this vector through the hidden layer, the fused global feature vector can be obtained, which serves as the input of the transposed convolutional module.

[0064] 3) Transposed convolutional networks are similar in principle to convolutional networks; their process can be seen as the inverse of convolution, i.e., generating a larger feature map from a smaller one. It expands the feature map size by inserting additional pixel locations and padding. In this invention, the transposed convolutional module receives the global feature vector output by the fully connected layer and predicts the temperature distribution of the target data center server room plane through the inverse operation of convolution.

[0065] Then, the model input is selected. A spatial feature map is chosen as the spatial variable input to the convolutional neural network (CNN) model, while the operating parameters most relevant to the data center temperature field are selected as the global variable input. The temperature field is used as the output of the CNN model to construct a neural network for quickly predicting the data center temperature field. In this embodiment, the spatial feature map used as the spatial variable input is obtained through a grid-to-spatial feature map conversion step. In this example, the operating parameters most relevant to the data center temperature field are selected as rack power, air conditioning temperature setpoint, and the height of the plane to be predicted.

[0066] Then, model hyperparameter optimization is performed to obtain a set of optimal hyperparameter values. In this embodiment, the model hyperparameters to be optimized are the number of convolutional / transposed convolutional layers, the size, stride, and padding of each convolutional kernel, the number of channels in the convolutional layers, the number of fully connected layers, the number of neurons in the hidden layers, and the model's learning rate. A grid search algorithm is used to optimize the hyperparameters, and the final optimization results are as follows: the number of convolutional and transposed convolutional layers is 2, each layer uses a 3*3 kernel, the stride and padding are 1, the number of channels in the first and second convolutional layers are 32 and 64 respectively, the number of channels in the first and second transposed convolutional layers are 64 and 32 respectively; the number of artificial neural network layers is 2, the number of neurons in the hidden layers is 1024; and the model's learning rate is 0.0001.

[0067] Finally, model parameter training is performed. The dataset of the data center server room temperature field is randomly divided into a training set and a validation set according to a certain proportion. On the training set, the stochastic gradient descent algorithm is used to train the parameters of the convolutional neural network, including weights and biases, to ensure its accuracy on the validation set and to quickly predict the data center server room temperature field. In this embodiment, 90% of the dataset is divided into a training set and 10% into a validation set. During training, the loss function used is the mean squared error (MSE), which is calculated using the following formula:

[0068]

[0069] The trained surrogate model was evaluated on the test dataset. This study used Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Coefficient of Determination (R²) as model evaluation metrics, calculated as follows:

[0070]

[0071] In the formula, n is the number of samples, and y i It is the actual value of the i-th sample. It is the predicted value of the i-th sample. It is the average value of the sample.

[0072] This embodiment compares the proposed framework for rapid prediction of multi-planar temperature fields in data center computer rooms based on spatial feature maps and convolutional neural networks with commonly used artificial neural network models in this field to verify the performance of the proposed method. Table 1 shows the performance comparison of the models on the validation set. Although artificial neural network models lacking spatial feature processing capabilities have demonstrated their advantages in single-point temperature prediction, they are not suitable for temperature field prediction, exhibiting large prediction errors and failing to provide effective information for airflow organization and control. In contrast, the prediction model that introduces convolutional networks for spatial feature extraction and encoding / decoding exhibits significantly better performance than the artificial neural network model. Figure 7 The model-predicted temperature fields at different heights in the validation set were randomly selected for comparison with the actual temperature fields after data preprocessing. It can be seen that under different computer room heights and operating conditions, the proposed prediction framework can accurately predict the overall temperature field distribution. This provides a scientific basis for the comprehensive optimization of airflow organization in data center computer rooms and the real-time operation of refined energy-saving control systems, demonstrating significant practical value and potential for wider application.

[0073] Table 1 Comparison of model accuracy on the validation set

[0074]

[0075] Corresponding to the aforementioned embodiment of a method for rapid prediction of temperature field in a data center based on a convolutional neural network, the present invention also provides an embodiment of a device for rapid prediction of temperature field in a data center based on a convolutional neural network.

[0076] See Figure 8 The present invention provides a data center computer room temperature field rapid prediction device based on convolutional neural network, comprising one or more processors, for implementing a data center computer room temperature field rapid prediction method based on convolutional neural network in the above embodiments.

[0077] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0078] An embodiment of the data center computer room temperature field rapid prediction device based on convolutional neural networks of the present invention can be applied to any device with data processing capabilities, such as a computer. The device embodiment can be implemented in software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution. From a hardware perspective, such as... Figure 8 The diagram shown is a hardware structure diagram of any data processing-capable device, including a data center computer room temperature field rapid prediction device based on a convolutional neural network according to the present invention. Except for... Figure 8 In addition to the processor, memory, network interface, and non-volatile memory shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0079] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0080] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the present invention according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0081] This invention also provides a computer-readable storage medium storing a program thereon, which, when executed by a processor, implements a method for rapid prediction of data center temperature field based on a convolutional neural network as described in the above embodiments.

[0082] The computer-readable storage medium can be an internal storage unit of any data processing device described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be any data processing device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data processing device, and can also be used to temporarily store data that has been output or will be output.

[0083] The above embodiments are only used to illustrate the design concept and features of the present invention, and their purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. The protection scope of the present invention is not limited to the above embodiments. Therefore, all equivalent changes or modifications made based on the principles and design ideas disclosed in the present invention are within the protection scope of the present invention.

Claims

1. A method for fast prediction of temperature field in data center machine room based on convolutional neural network, characterized in that, include: Obtain the geometric information of the data center computer room structure, perform geometric modeling and mesh generation on the data center computer room structure, and obtain the mesh structure of the data center computer room; Obtain operating parameters of equipment within the data center server room; Acquire temperature field data of the data center server room, divide the temperature field data according to the height plane, and generate the output of the dataset for training the convolutional neural network model; The grid structure of the data center server room at different heights is converted into a spatial feature map structure, and the corresponding height information and operating condition parameter information are recorded to generate the input of the dataset used to train the convolutional neural network model. The trained convolutional neural network model is used for temperature field prediction; The process of converting the grid structure at different heights of the data center server room into a spatial feature map structure includes: Orthogonal resampling is performed on the mesh structure to convert the unstructured mesh into an orthogonal mesh with a fixed resolution. Spatial feature attributes are assigned to each mesh cell, and the entire mesh is converted into a spatial feature map. The convolutional neural network model includes: a convolutional neural network, a fully connected layer, and a transposed convolutional neural network; the input of the convolutional neural network is a spatial feature map, the output of the convolutional neural network, the operating parameters, and the plane height are the inputs of the fully connected layer, the output of the fully connected layer is the input of the transposed convolutional neural network, and the output of the transposed convolutional neural network is the temperature distribution of the data center server room.

2. The method of claim 1, wherein, The data center server room is divided into grids, including: The finite element method was used to perform unstructured mesh generation for the data center computer room.

3. The method of claim 1, wherein, The acquisition of temperature field data in the data center includes: The temperature field of the data center is numerically calculated using the finite element method to obtain the temperature field data of the data center.

4. The method of claim 1, wherein, The temperature field data of the data center is divided into height planes and then normalized, and this normalization is used as the model output for the training process.

5. The method of claim 1, wherein, The geometric information of the data center includes the shape and size of the enclosure structure, and the shape, size, and placement of the server racks, terminal air conditioners, and interior building structure.

6. The method of claim 1, wherein, The operating parameters within the data center server room include the power of the server racks, the air supply temperature of the air conditioning terminals, and the air supply speed.

7. A device for fast prediction of temperature field in a data center machine room based on a convolutional neural network, characterized in that, It includes one or more processors for implementing a method for rapid prediction of data center temperature field based on a convolutional neural network, as described in any one of claims 1-6.

8. A computer-readable storage medium having a program stored thereon, characterized in that, When executed by the processor, the program is used to implement the method for rapid prediction of data center temperature field based on convolutional neural network as described in any one of claims 1-6.

Images