An internal physical field prediction method, system and device of a data center

Abstract

The application discloses a kind of internal physical field prediction method, system and equipment of data center, comprising: the grid model of data center, physical property parameter and operating condition parameter are input to pre-constructed data center room-server level physical field prediction simulation model, obtain the preset physical field information of data center room-server level;The inlet parameter of each server and the thermal load data of IGBT are input to pre-constructed data center server-IGBT level physical field fast prediction model, obtain the preset physical field information of each server;The inlet parameter of each IGBT and the thermal load data of chip are input to pre-constructed data center IGBT-chip level physical field fast prediction model, obtain the preset physical field information in each IGBT and chip in data center;The application realizes the fine prediction of entire data center flow field and temperature field, greatly reduces the consumption of computing resources, and prediction time is short.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of data center heat dissipation, and particularly relates to a data center internal physical field prediction method, system and device. BACKGROUND

[0002] With the rapid development of cloud computing, big data, artificial intelligence and other technologies, data centers have gradually become the infrastructure of modern information society to undertake the important task of storing, processing and transmitting massive data, and their importance is increasingly prominent. At the same time, the energy consumption, security and reliability of data centers have also attracted much attention. The internal flow field and temperature field of data centers directly affect the heat dissipation effect of equipment and the stability of the working environment. Therefore, predicting the internal flow field and temperature field of data centers is of great significance for optimizing the design and layout of data centers, improving the heat dissipation efficiency of equipment, reducing energy consumption costs, and ensuring the safe and stable operation of internal electronic devices.

[0003] Due to the complex flow field and temperature field inside the data center, there are multiple length scales including micro-nano scales, and commercial-level data centers are usually confidential, so it is difficult to carry out in-situ testing of the internal flow field and temperature field by experimental methods. At present, scholars have established a series of computational fluid dynamics simulation models for predicting the internal flow field and temperature field of data centers, but the simulation speed is slow, and the simulation is usually carried out only at the room-server level scale of data centers. If this model is directly used for whole-field prediction of multi-scale flow field and temperature field in data centers, there are problems of long calculation time and poor economic efficiency, and the calculation resources consumed are unacceptable. Therefore, there is currently no related research and analysis on the flow field and temperature field of the whole internal model of data centers. SUMMARY

[0004] In view of the technical problems existing in the prior art, the present application provides a data center internal physical field prediction method, system and device to solve the technical problems that the computational fluid dynamics simulation model of the internal flow field and temperature field of data centers cannot be directly used for whole-field prediction of multi-scale flow field and temperature field in data centers, and the calculation resources consumed are huge and the economic efficiency is poor.

[0005] To achieve the above-mentioned purposes, the technical scheme adopted by the present application is as follows:

[0006] The present application provides a data center internal physical field prediction method, comprising:

[0007] obtaining the geometric parameters of the data center, the physical property parameters of the data center and the operating condition parameters of the data center;

[0008] drawing a grid model of the data center according to the geometric parameters of the data center;

[0009] inputting the grid model of the data center, the physical property parameters of the data center and the operation condition parameters of the data center into a pre-constructed data center room-server level physical field prediction simulation model, outputting preset physical field information of the data center room-server level, and obtaining an inlet parameter of each server in the data center;

[0010] obtaining thermal load data of each IGBT in the data center;

[0011] inputting the inlet parameter of each server in the data center and the thermal load data of the IGBT inside each server in the data center into a pre-constructed data center server-IGBT level physical field fast prediction model, outputting preset physical field information of each server in the data center, and obtaining an inlet parameter of each IGBT in the data center;

[0012] obtaining thermal load data of each chip in the data center;

[0013] inputting the inlet parameter of each IGBT in the data center and the thermal load data of the chip inside each IGBT in the data center into a pre-constructed data center IGBT-chip level physical field fast prediction model, and obtaining preset physical field information in each IGBT and chip in the data center.

[0014] Further, the geometric parameters of the data center include room sizes of the data center, sizes of air conditioners in the data center, sizes of cabinets in the data center, layout parameters of servers in the data center and layout parameters of air conditioners in the data center;

[0015] The physical property parameters of the data center include air density in the data center, air viscosity in the data center, air thermal conductivity in the data center, air specific heat capacity in the data center, solid material density in the data center, solid material thermal conductivity in the data center, solid material specific heat capacity in the data center, server resistance coefficient in the data center and server fan curve in the data center.

[0016] The operation condition parameters of the data center include air flow of the air conditioner in the data center, air outlet temperature of the air conditioner in the data center and thermal load data of each server in the data center.

[0017] Further, the pre-constructed data center room-server level physical field prediction simulation model is a data center room-server level flow field prediction simulation model or a data center room-server level temperature field prediction simulation model.

[0018] The data center room-server level flow field prediction simulation model or the data center room-server level temperature field prediction simulation model is a numerical model based on computational fluid dynamics.

[0019] The preset physical field information of the data center room-server level is flow field information or temperature field information of the data center room-server level.

[0020] Further, the inlet parameters of each server in the data center include the inlet temperature of each server and the inlet air flow rate of each server; and the inlet parameters of each IGBT in the data center include the inlet temperature of each IGBT and the inlet air flow rate of each IGBT.

[0021] Further, the pre-constructed data center server-IGBT level physical field fast prediction model includes a pre-constructed data center server-IGBT level flow field fast prediction model or a pre-constructed data center server-IGBT level temperature field fast prediction model.

[0022] The construction process of the pre-constructed data center server-IGBT level physical field fast prediction model is as follows:

[0023] According to the pre-constructed data center server-IGBT level physical field prediction simulation model, the model parameters of the data center server-IGBT level are determined.

[0024] From the model parameters of the data center server-IGBT level, the server level variable operation parameters of the data center are selected, and the server level variable operation parameters of the data center are valued to obtain a plurality of server level snapshot variable input parameters.

[0025] The plurality of server level snapshot variable input parameters are taken as the input of the pre-constructed data center server-IGBT level physical field prediction simulation model, and a plurality of server level snapshots are output.

[0026] According to the plurality of server level snapshots, a server level snapshot matrix of a preset physical field is generated.

[0027] The server level snapshot matrix of the preset physical field is subjected to matrix transformation to obtain a plurality of server level base functions, and the plurality of server level base functions are sorted in descending order according to the amount of information they capture.

[0028] The first several server level base functions are selected to construct a server-IGBT level physical field fast prediction model; wherein the server-IGBT level physical field fast prediction model is:

[0029]

[0030] wherein r is the number of physical quantity; f r is the three-dimensional physical field of the rth physical quantity under the variable input parameters of the server-IGBT level corresponding; b r,j is the weight coefficient of the three-dimensional physical field of the rth physical quantity of the server-IGBT level in the jth basis function direction; ψ r,j is the jth basis function of the rth physical quantity of the server-IGBT level; j is the number of the server level basis function; l r is the truncation order of the rth physical quantity of the server-IGBT level.

[0031] Further, the weight coefficient b r,j of the prediction model of the three-dimensional physical field of the rth physical quantity of the server-IGBT level in the jth basis function direction is determined according to the weight coefficient of the three-dimensional physical field of the rth physical quantity of the server-IGBT level in the jth basis function direction under each server level snapshot working condition.

[0032] The server level snapshot matrix of the preset physical field is projected to each server level basis function direction, and the weight coefficient of each server level snapshot working condition in the basis function direction is reconstructed.

[0033] Based on the weight coefficient of each server level snapshot working condition in the basis function direction, an interpolation method, a regression method or a machine learning method is used to obtain the prediction model of the weight coefficient b r,j of the three-dimensional physical field of the rth physical quantity of the server-IGBT level in the jth basis function direction.

[0034] Further, the pre-constructed data center IGBT-chip level physical field fast prediction model includes a pre-constructed data center IGBT-chip level flow field fast prediction model or a pre-constructed data center IGBT-chip level temperature field fast prediction model.

[0035] The construction process of the pre-constructed data center IGBT-chip level physical field fast prediction model is as follows:

[0036] According to the pre-constructed data center IGBT-chip level physical field prediction simulation model, the model parameters of the data center IGBT-chip level are determined.

[0037] From the model parameters of the data center IGBT-chip level, the IGBT level variable operating parameters of the data center are selected, and the IGBT level variable operating parameters of the data center are valued to obtain a plurality of IGBT level snapshot variable input parameters.

[0038] The plurality of IGBT level snapshot variable input parameters are taken as the input of the pre-constructed data center IGBT-chip level physical field prediction simulation model, and a plurality of IGBT level snapshots are output.

[0039] The IGBT level snapshot matrix of the preset physical field is subjected to matrix transformation to obtain a plurality of IGBT level base functions, and the plurality of IGBT level base functions are sorted according to the amount of information captured from large to small.

[0040] The IGBT level snapshot matrix of the preset physical field is subjected to matrix transformation to obtain a plurality of IGBT level base functions, and the plurality of IGBT level base functions are sorted according to the amount of information captured from large to small.

[0041] The IGBT- chip level physical field fast prediction model is constructed by selecting the base functions of the first number of IGBTs, wherein the IGBT- chip level physical field fast prediction model is:

[0042]

[0043] Wherein, r is the number of physical quantities; f r ′ is the three-dimensional physical field of the rth physical quantity under the IGBT- chip level corresponding variable input parameter; b r ′ ,j′ is the weight coefficient of the three-dimensional physical field of the rth physical quantity of the IGBT- chip level in the base function direction of the j'th IGBT level; ψ′ r,j′ is the base function of the j'th IGBT level of the rth physical quantity of the IGBT- chip level; j' is the number of IGBT level base functions; l r ′ is the truncation order of the rth physical quantity of the IGBT- chip level.

[0044] Further, the weight coefficient b r ′ ,j′ of the three-dimensional physical field of the rth physical quantity of the IGBT- chip level in the base function direction of the j'th IGBT level is obtained.

[0045] The IGBT level snapshot matrix of the preset physical field is projected to the base function direction of each IGBT level to reconstruct the weight coefficient of the base function direction under each IGBT level snapshot working condition.

[0046] Based on the weight coefficient of the base function direction under each IGBT level snapshot working condition, an interpolation method, a regression method or a machine learning method is used to obtain the prediction model of the weight coefficient b r ′of the three-dimensional physical field of the rth physical quantity of the IGBT- chip level in the base function direction of the j'th IGBT level. ,j′

[0047] The application also provides an internal physical field prediction system of a data center, characterized by comprising:

[0048] The first data acquisition module is used for acquiring the geometric parameters of the data center, the physical property parameters of the data center and the operation condition parameters of the data center.

[0049] a grid model module, configured to draw a grid model of the data center according to geometric parameters of the data center;

[0050] a room-server level physical field prediction module, configured to input the grid model of the data center, the physical property parameters of the data center and the operation condition parameters of the data center into a pre-constructed data center room-server level physical field prediction simulation model, output preset physical field information of the data center room-server level, and obtain inlet parameters of each server in the data center;

[0051] a second data acquisition module, configured to acquire thermal load data of each IGBT in the data center;

[0052] a server-IGBT level physical field prediction module, configured to input the inlet parameters of each server in the data center and the thermal load data of the IGBT inside each server in the data center into a pre-constructed data center server-IGBT level physical field fast prediction model, output preset physical field information of each server in the data center, and obtain inlet parameters of each IGBT in the data center;

[0053] a third data acquisition module, configured to acquire thermal load data of each chip in the data center;

[0054] an IGBT-chip level physical field prediction module, configured to input the inlet parameters of each IGBT in the data center and the thermal load data of the chip inside each IGBT in the data center into a pre-constructed data center IGBT-chip level physical field fast prediction model, and obtain preset physical field information in each IGBT and chip in the data center.

[0055] The application further provides a data center internal physical field prediction device, which comprises:

[0056] a memory, configured to store a computer program;

[0057] a processor, configured to execute the computer program to realize the steps of the data center internal physical field prediction method.

[0058] Compared with the prior art, the application has the following beneficial effects:

[0059] The application provides an internal physical field prediction method and system of a data center, which realizes hierarchical multi-scale whole-field prediction of preset physical fields in the data center by predicting the preset physical fields in the room-server level, the server-IGBT level and the IGBT-chip level in the data center; the preset physical fields in the server-IGBT level and the IGBT-chip level are predicted by using a data center server-IGBT level physical field fast prediction model and a data center IGBT-chip level physical field fast prediction model respectively, instead of a traditional computational fluid dynamics simulation model, so that the required calculation resources for simulation are reduced by orders of magnitude, thereby realizing fine prediction of the flow field and the temperature field in the whole data center within an acceptable calculation cost, greatly reducing the consumption of calculation resources, being good in economy and short in prediction time. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 A flow chart of the internal physical field prediction method of the data center described in the embodiment;

[0061] Figure 2 A computational fluid dynamics simulation result diagram of the temperature field of the data center server-IGBT level in the embodiment on a test section under a certain working condition;

[0062] Figure 3 A fast model prediction result diagram of the temperature field of the data center server-IGBT level in the embodiment on a test section under a certain working condition;

[0063] Figure 4 A computational fluid dynamics simulation result diagram of the flow field of the data center server-IGBT level in the embodiment on a test section under a certain working condition;

[0064] Figure 5 A fast model prediction result diagram of the flow field of the data center server-IGBT level in the embodiment on a test section under a certain working condition;

[0065] Figure 6 A computational fluid dynamics simulation result diagram of the temperature field of the data center IGBT-chip level in the embodiment under three working conditions;

[0066] Figure 7 A fast model prediction result diagram of the temperature field of the data center IGBT-chip level in the embodiment under three working conditions;

[0067] Figure 8 A highest temperature distribution diagram of each chip in each IGBT in the data center in the embodiment. DETAILED DESCRIPTION

[0068] In order to make the technical problems solved by the present application, technical solutions and beneficial effects clearer, the following specific embodiments are used to further explain the present application. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0069] The present application provides a method for predicting the internal physical field of a data center, which is used for the prediction process of the internal flow field and temperature field of the data center; wherein the method for predicting the internal physical field of the data center comprises the following steps:

[0070] Step 1, constructing a server-IGBT level physical field fast prediction model; wherein the server-IGBT level physical field fast prediction model comprises a server-IGBT level flow field fast prediction model and a server-IGBT level temperature field fast prediction model, and specifically comprises the following steps:

[0071] Step 101, constructing a prediction simulation model of the data center server-IGBT level flow field or temperature field; wherein the prediction simulation model of the data center server-IGBT level flow field or temperature field is a numerical model based on computational fluid dynamics.

[0072] Step 102, determining the model parameters of the data center server-IGBT level according to the prediction simulation model of the data center server-IGBT level flow field or temperature field.

[0073] Wherein the model parameters of the data center server-IGBT level include the geometric parameters of the server, the physical property parameters of the server and the operating condition parameters of the server; specifically, the geometric parameters of the server include the internal layout information of the server and the size of each component; the physical property parameters of the server include the air density of the server, the air viscosity of the server, the air thermal conductivity coefficient of the server, the air specific heat capacity of the server, the solid material density of the server, the solid material thermal conductivity coefficient of the server and the solid material specific heat capacity of the server; the operating condition parameters of the server include the inlet air flow rate of the server, the inlet temperature of the server and the heat load of the IGBT.

[0074] Step 103, selecting the server level variable operation parameters of the data center from the model parameters of the data center server-IGBT level, and taking values of the server level variable operation parameters of the data center to obtain a plurality of server level snapshot variable input parameters; wherein the server level variable operation parameters of the data center include the inlet air flow rate of the server, the inlet temperature of the server and the heat load of the IGBT.

[0075] Step 104, taking the variable input parameters of the plurality of server-level snapshots as inputs of the prediction simulation model of the data center server-IGBT-level flow field or temperature field established in step 101, solving the simulation model, and outputting a plurality of server-level snapshots.

[0076] Step 105, generating a server flow field snapshot matrix or a server-level temperature field snapshot matrix according to the plurality of server-level snapshots for the server-IGBT-level flow field or temperature field; wherein each column in the server flow field snapshot matrix or the server-level temperature field snapshot matrix represents a corresponding physical field of a server-level snapshot.

[0077] Step 106, performing matrix transformation on the server flow field snapshot matrix or the server-level temperature field snapshot matrix for the server-IGBT-level flow field or temperature field to obtain a plurality of server-level flow field basis functions or a plurality of server-level temperature field basis functions, and sorting the plurality of server-level flow field basis functions or the plurality of server-level temperature field basis functions according to the amount of information they capture from large to small; wherein the process of matrix transformation is as follows:

[0078] Performing singular value decomposition on the server flow field snapshot matrix or the server-level temperature field snapshot matrix to obtain singular values and a right singular matrix;

[0079] Right multiplying the server flow field snapshot matrix or the server-level temperature field snapshot matrix by the right singular matrix to obtain unnormalized server-level flow field basis functions or unnormalized server-level temperature field basis functions; wherein the singular values represent the amount of information captured by each basis function;

[0080] Performing normalization processing on the unnormalized server-level flow field basis functions or the unnormalized server-level temperature field basis functions to obtain a plurality of server-level flow field basis functions or a plurality of server-level temperature field basis functions.

[0081] Step 107, selecting the first several server-level flow field basis functions or the first several server-level temperature field basis functions to construct a server-IGBT-level physical field fast prediction model; wherein the server-IGBT-level physical field fast prediction model is:

[0082]

[0083] wherein r is the number of physical quantities of the server-IGBT-level; f r is the three-dimensional physical field of the rth physical quantity of the server-IGBT-level under the corresponding variable input parameters; b r,j is the weight coefficient of the three-dimensional physical field of the rth physical quantity of the server-IGBT-level in the jth basis function direction; ψ r,jis the j-th basis function of the r-th physical quantity at the server-IGBT level; j is the number of the basis function at the server level; l r is the truncation order of the r-th physical quantity at the server-IGBT level.

[0084] It should be noted that the number of selected basis functions of the flow field server level or the basis functions of the temperature field server level is determined according to the principle that singular values account for a large proportion; the weight coefficient b r,j The establishment process of the prediction model of the weight coefficient b r,j is as follows:

[0085] The server level snapshot matrix of the preset physical field is projected in each server level basis function direction to reconstruct the weight coefficient of the basis function direction under each server level snapshot working condition; based on the weight coefficient of the basis function direction under each server level snapshot working condition, an interpolation method, a regression method or a machine learning method is used to obtain the prediction model of the weight coefficient b r,j of the three-dimensional physical field of the r-th physical quantity at the server-IGBT level in the j-th basis function direction.

[0086] Step 2, constructing an IGBP-chip level physical field fast prediction model; wherein the IGBP-chip level physical field fast prediction model includes an IGBP-chip level flow field fast prediction model and an IGBP-chip level temperature field fast prediction model, and specifically includes the following steps:

[0087] Step 201, constructing a prediction simulation model of the data center IGBT-chip level flow field or temperature field; wherein the prediction simulation model of the data center IGBT-chip level flow field or temperature field is a numerical model based on computational fluid dynamics.

[0088] Step 202, determining the model parameters of the data center IGBT-chip level according to the prediction simulation model of the data center IGBT-chip level flow field or temperature field.

[0089] The model parameters of the data center IGBT-chip level include the geometric parameters of the IGBT, the physical property parameters of the IGBT and the operating condition parameters of the IGBT; specifically, the geometric parameters of the IGBT include the internal layout information of the IGBT and the size of each component; the physical property parameters of the IGBT include the air density of the IGBT, the air viscosity of the IGBT, the air thermal conductivity coefficient of the IGBT, the air specific heat capacity of the IGBT, the solid material density of the IGBT, the solid material thermal conductivity coefficient of the IGBT and the solid material specific heat capacity of the IGBT; the operating condition parameters of the IGBT include the inlet air flow rate of the IGBT, the inlet temperature of the IGBT and the heat load of the chip.

[0090] Step 203, selecting data center IGBT level variable operation parameters from the model parameters of the data center IGBT-chip level, and taking values of the data center IGBT level variable operation parameters, to obtain a plurality of IGBT level snapshot variable input parameters; wherein the data center IGBT level variable operation parameters include IGBT inlet air flow rate, IGBT inlet temperature and chip heat load.

[0091] Step 204, taking the plurality of IGBT level snapshot variable input parameters as the input of the prediction simulation model of the data center IGBT-chip level flow field or temperature field established in step 201, and outputting a plurality of IGBT level snapshots by solving the simulation model.

[0092] Step 205, for the IGBT-chip level flow field or temperature field, generating an IGBT level flow field snapshot matrix or an IGBT level temperature field snapshot matrix according to the plurality of IGBT level snapshots; wherein each column in the IGBT level flow field snapshot matrix or the IGBT level temperature field snapshot matrix represents a corresponding physical field of an IGBT level snapshot.

[0093] Step 206, for the IGBT-chip level flow field or temperature field, performing matrix transformation on the IGBT level flow field snapshot matrix or the IGBT level temperature field snapshot matrix to obtain a plurality of IGBT level flow field basis functions or a plurality of IGBT level temperature field basis functions, and sorting the plurality of IGBT level flow field basis functions or the plurality of IGBT level temperature field basis functions according to the amount of information they capture from large to small; wherein the operation of matrix transformation on the IGBT level flow field snapshot matrix or the IGBT level temperature field snapshot matrix is similar to the matrix transformation process in step 106, which will not be repeated here.

[0094] Step 207, selecting the first number of IGBT level flow field basis functions or the first number of IGBT level temperature field basis functions to construct an IGBT-chip level physical field fast prediction model; wherein the IGBT-chip level physical field fast prediction model is:

[0095]

[0096] Wherein r is the number of physical quantities; f r ′ is the three-dimensional physical field of the rth physical quantity under the IGBT-chip level corresponding variable input parameter; b r ,j′ is the weight coefficient of the three-dimensional physical field of the rth physical quantity of the IGBT-chip level in the direction of the j'th IGBT level basis function; ψ′ r,j′ ​is the j'th IGBT level basis function of the IGBT-chip level rth physical quantity; j' is the number of the IGBT level basis function; l r is the truncation order of the IGBT-chip level rth physical quantity.

[0097] It should be noted that the number of selected basis functions of the IGBT level flow field or the basis functions of the IGBT level temperature field is determined according to the principle that singular values account for a large proportion; the weight coefficient b r ′ ,j′ of the three-dimensional physical field of the IGBT-chip level rth physical quantity in the j'th IGBT level basis function direction is obtained by projecting the IGBT level snapshot matrix of the preset physical field to each IGBT level basis function direction; based on the weight coefficient of the basis function direction under each IGBT level snapshot working condition, an interpolation method, a regression method or a machine learning method is used to obtain the prediction model of the weight coefficient b r ′ ,j′ of the three-dimensional physical field of the IGBT-chip level rth physical quantity in the j'th basis function direction.

[0098] Step 3, construction of the flow field and the temperature field in the entire data center; wherein the flow field and the temperature field in the entire data center include the flow field and the temperature field information of the room-server level in the data center, the flow field and the temperature field information in each server, and the flow field and the temperature field information of each IGBT and chip; specifically, the following steps are included:

[0099] Step 301, constructing a prediction simulation model of the data center room-server level flow field or temperature field; wherein the data center room-server level flow field or temperature field prediction simulation model is a numerical model based on computational fluid dynamics.

[0100] Step 302, obtaining the geometric parameters of the data center, the physical property parameters of the data center and the operating condition parameters of the data center; specifically, the geometric parameters of the data center include the room size of the data center, the size of the air conditioner in the data center, the size of the cabinet in the data center, the layout parameters of the servers in the data center and the layout parameters of the air conditioners in the data center; the physical property parameters of the data center include the air density in the data center, the air viscosity in the data center, the air thermal conductivity in the data center, the air specific heat capacity in the data center, the solid material density in the data center, the solid material thermal conductivity in the data center, the solid material specific heat capacity in the data center, the server resistance coefficient in the data center and the server fan curve in the data center; the operating condition parameters of the data center include the air conditioner flow rate in the data center, the air conditioner outlet temperature in the data center and the thermal load data of each server in the data center.

[0101] Step 303, according to the geometric parameters of the data center, a grid model of the data center is drawn; wherein, unstructured grid is adopted in the grid model of the data center.

[0102] Step 304, the grid model of the data center, the physical property parameters of the data center and the operation condition parameters of the data center are input into the prediction simulation model of the data center room-server level flow field or temperature field constructed in step 301, and the data center room-server level flow field or temperature field information is output, and the inlet parameters of each server in the data center are obtained; wherein, the area average method is adopted to obtain the inlet parameters of each server in the data center; the inlet parameters of each server in the data center include the inlet temperature of each server and the inlet air flow rate of each server.

[0103] Step 305, the inlet air flow rate of each server in the data center, the inlet temperature of each server in the data center and the heat load data of each IGBT in the data center are substituted into the server-IGBT level physical field fast prediction model established in step 107, and the flow field or temperature field information of each server in the data center is output, and the inlet parameters of each IGBT in the data center are obtained; wherein, the area average method is adopted to obtain the inlet parameters of each IGBT in the data center; the inlet parameters of each IGBT in the data center include the inlet temperature of each IGBT and the inlet air flow rate of each IGBT.

[0104] Step 306, the inlet air flow rate of each IGBT in the data center, the inlet temperature of each IGBT in the data center and the heat load of each chip in the data center are substituted into the IGBT-chip level physical field fast prediction model established in step 207, and the flow field or temperature field information in each IGBT and chip in the data center is output.

[0105] The internal physical field prediction method of the data center provided by the application is used for the prediction of the flow field and temperature field of the server and IGBT unit with a large number of repeated characteristics, and the fast and accurate prediction model is used to replace the traditional computational fluid dynamics simulation model, so that the required calculation resources for simulation are reduced by orders of magnitude, so that the fine prediction of the flow field and temperature field in the entire data center can be realized within the acceptable calculation cost.

[0106] The application further provides an internal physical field prediction system of a data center, comprising a first data acquisition module, a grid model module, a room-server-level physical field prediction module, a second data acquisition module, a server-IGBT-level physical field prediction module, a third data acquisition module and an IGBT-chip-level physical field prediction module; the first data acquisition module is used for acquiring geometric parameters of the data center, physical property parameters of the data center and operation condition parameters of the data center; the grid model module is used for drawing a grid model of the data center according to the geometric parameters of the data center; the room-server-level physical field prediction module is used for inputting the grid model of the data center, the physical property parameters of the data center and the operation condition parameters of the data center into a pre-constructed room-server-level physical field prediction simulation model of the data center, outputting preset physical field information of the data center at the room-server level and obtaining inlet parameters of each server in the data center; the second data acquisition module is used for acquiring thermal load data of each IGBT in the data center; the server-IGBT-level physical field prediction module is used for inputting the inlet parameters of each server in the data center and the thermal load data of the IGBT inside each server in the data center into a pre-constructed server-IGBT-level physical field fast prediction model of the data center, outputting preset physical field information of each server in the data center and obtaining inlet parameters of each IGBT in the data center; the third data acquisition module is used for acquiring thermal load data of each chip in the data center; and the IGBT-chip-level physical field prediction module is used for inputting the inlet parameters of each IGBT in the data center and the thermal load data of the chip inside each IGBT in the data center into a pre-constructed IGBT-chip-level physical field fast prediction model of the data center, and obtaining preset physical field information in each IGBT and chip in the data center.

[0107] The application further provides an internal physical field prediction device of a data center, comprising a memory for storing a computer program and a processor for executing the computer program to realize the steps of the internal physical field prediction method of the data center.

[0108] The processor implements the steps of the above-mentioned internal physical field prediction method of the data center when executing the computer program, for example: obtaining the geometric parameters of the data center, the physical property parameters of the data center and the operation condition parameters of the data center; obtaining the grid model of the data center according to the geometric parameters of the data center; inputting the grid model of the data center, the physical property parameters of the data center and the operation condition parameters of the data center into the pre-constructed room-server level physical field prediction simulation model of the data center, outputting the preset physical field information of the room-server level of the data center, and obtaining the inlet parameters of each server in the data center; obtaining the thermal load data of each IGBT in the data center; inputting the inlet parameters of each server in the data center and the thermal load data of the IGBT inside each server in the data center into the pre-constructed server-IGBT level physical field fast prediction model of the data center, outputting the preset physical field information of each server in the data center, and obtaining the inlet parameters of each IGBT in the data center; obtaining the thermal load data of each chip in the data center; inputting the inlet parameters of each IGBT in the data center and the thermal load data of the chip inside each IGBT in the data center into the pre-constructed IGBT-chip level physical field fast prediction model of the data center, and obtaining the preset physical field information in each IGBT and chip in the data center.

[0109] Or, the processor implements the functions of the modules in the above system when executing the computer program, for example: a first data acquisition module for acquiring geometric parameters of a data center, physical property parameters of the data center, and operating condition parameters of the data center; a grid model module for drawing a grid model of the data center according to the geometric parameters of the data center; a room-server level physical field prediction module for inputting the grid model of the data center, the physical property parameters of the data center, and the operating condition parameters of the data center into a pre-constructed room-server level physical field prediction simulation model of the data center, outputting preset physical field information of a room-server level of the data center, and obtaining inlet parameters of each server in the data center; a second data acquisition module for acquiring thermal load data of each IGBT in the data center; a server-IGBT level physical field prediction module for inputting the inlet parameters of each server in the data center and the thermal load data of the IGBT inside each server in the data center into a pre-constructed server-IGBT level physical field fast prediction model of the data center, outputting preset physical field information of each server in the data center, and obtaining inlet parameters of each IGBT in the data center; a third data acquisition module for acquiring thermal load data of each chip in the data center; and an IGBT-chip level physical field prediction module for inputting the inlet parameters of each IGBT in the data center and the thermal load data of the chip inside each IGBT in the data center into a pre-constructed IGBT-chip level physical field fast prediction model of the data center, and obtaining preset physical field information in each IGBT and chip in the data center.

[0110] For example, 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 application. The one or more modules / units can be a series of computer program instruction segments capable of completing a preset function, which are used to describe the execution process of the computer program in the internal physical field prediction device of the data center. For example, the computer program can be divided into a first data acquisition module, a grid model module, a room-server level physical field prediction module, a second data acquisition module, a server-IGBT level physical field prediction module, a third data acquisition module, and an IGBT-chip level physical field prediction module, and the specific functions of each module are as follows: the first data acquisition module is used to acquire the geometric parameters of the data center, the physical property parameters of the data center, and the operating condition parameters of the data center; the grid model module is used to draw a grid model of the data center according to the geometric parameters of the data center; the room-server level physical field prediction module is used to input the grid model of the data center, the physical property parameters of the data center, and the operating condition parameters of the data center into a pre-constructed room-server level physical field prediction simulation model of the data center, output preset physical field information of the room-server level of the data center, and obtain the inlet parameters of each server in the data center; the second data acquisition module is used to acquire the thermal load data of each IGBT in the data center; the server-IGBT level physical field prediction module is used to input the inlet parameters of each server in the data center and the thermal load data of the IGBT inside each server in the data center into a pre-constructed server-IGBT level physical field fast prediction model of the data center, output preset physical field information of each server in the data center, and obtain the inlet parameters of each IGBT in the data center; the third data acquisition module is used to acquire the thermal load data of each chip in the data center; and the IGBT-chip level physical field prediction module is used to input the inlet parameters of each chip in the data center and the thermal load data of the chip inside each IGBT in the data center into a pre-constructed IGBT-chip level physical field fast prediction model of the data center, and obtain preset physical field information in each IGBT and chip in the data center.

[0111] The internal physical field prediction device of the data center can be a computing device such as a desktop computer, a notebook computer, a palm computer, and a cloud server. The internal physical field prediction device of the data center can include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above is an example of the internal physical field prediction device of the data center, and does not constitute a limitation on the internal physical field prediction device of the data center, and can include more components than the above, or combine certain components, or different components, for example, the internal physical field prediction device of the data center can also include an input / output device, a network access device, a bus, and the like.

[0112] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, and the like. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor and the like. The processor is the control center of the internal physical field prediction device of the data center, and connects various parts of the internal physical field prediction device of the data center through various interfaces and lines.

[0113] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the internal physical field prediction device of the data center by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory.

[0114] The memory can mainly include a program storage area and a data storage area. The program storage area can store an operating system, at least one application program required for a function (such as a sound playing function, an image playing function, and the like), and the like; and the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, and the like), and the like. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory such as a hard disk, a 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 memory devices.

[0115] Embodiments

[0116] The embodiment takes the prediction process of a three-dimensional flow field and a temperature field of an air-cooled data center as an example; wherein the air-cooled data center adopts a down air supply type.

[0117] As shown in the accompanying Figure 1 The embodiment provides a method for predicting internal physical fields of a data center, including the following steps:

[0118] Step 1, constructing a prediction simulation model of a data center server-IGBT level flow field or temperature field; wherein the prediction simulation model of the data center server-IGBT level flow field or temperature field is a numerical model based on computational fluid dynamics; wherein the equations solved in the prediction simulation model of the data center server-IGBT level flow field or temperature field include a mass conservation equation, a momentum conservation model, an energy conservation equation, and a standard k-ε turbulent flow equation; ANSYS ICEPAK software is used to physically model the server-IGBT level and solve the above equations.

[0119] Step 2, determining model parameters of the data center server-IGBT level according to the prediction simulation model of the data center server-IGBT level flow field or temperature field; wherein the model parameters of the data center server-IGBT level include geometric parameters of the server, physical property parameters of the server, and operating condition parameters of the server; specifically, the geometric parameters of the server include internal layout information of the server and sizes of various components; the physical property parameters of the server include air density of the server, air viscosity of the server, air thermal conductivity coefficient of the server, air specific heat capacity of the server, solid material density of the server, solid material thermal conductivity coefficient of the server, and solid material specific heat capacity of the server; the operating condition parameters of the server include inlet air flow rate of the server, inlet temperature of the server, and heat load of the IGBT.

[0120] Step 3, selecting data center server variable operating parameters from the model parameters of the data center server-IGBT level, and taking values of the data center server variable operating parameters to obtain a plurality of server level snapshot variable input parameters; wherein the data center server level variable operating parameters include inlet air flow rate of the server, inlet temperature of the server, and heat load of the IGBT.

[0121] In the embodiment, full factorial design is used to design the values of the data center server variable operating parameters; specifically, the server inlet air flow rate and the heat load of the IGBT are selected at 8 levels respectively, and the server inlet temperature is taken as 0, a total of 64 groups of server level snapshot variable input parameters are designed.

[0122] Step 4, the variable input parameters of the plurality of server-level snapshots are taken as inputs of a prediction simulation model of the data center server-IGBT-level flow field or temperature field constructed in step 1, and a plurality of server-level snapshots are output by solving the simulation model; wherein the plurality of server-level snapshots are output in the format of a Tecplot file.

[0123] Step 5, for the server-IGBT-level flow field or temperature field, a snapshot matrix of the server-level flow field or a snapshot matrix of the server-level temperature field is generated according to the plurality of server-level snapshots; wherein each column in the snapshot matrix of the server-level flow field or the snapshot matrix of the server-level temperature field represents a corresponding physical field of a server-level snapshot; specifically, after the plurality of server-level snapshots in the Tecplot format are parsed by self-programming in C++, the numerical values of the physical fields on each node and element are extracted, and thus the snapshot matrix of the server-level flow field or the snapshot matrix of the server-level temperature field is obtained.

[0124] Step 6, for the server-IGBT-level flow field or temperature field, the snapshot matrix of the server-level flow field or the snapshot matrix of the server-level temperature field is subjected to matrix transformation based on self-programming in C++, and a plurality of server-level flow field basis functions or a plurality of server-level temperature field basis functions are obtained, and the plurality of server-level flow field basis functions or the plurality of server-level temperature field basis functions are sorted in descending order of the amount of information they capture.

[0125] Wherein, the process of matrix transformation is specifically: singular value decomposition of the snapshot matrix of the server-level flow field or the snapshot matrix of the server-level temperature field is performed by using Jacobi algorithm to obtain singular values and a right singular matrix; the snapshot matrix of the server-level flow field or the snapshot matrix of the server-level temperature field is right multiplied by the right singular matrix to obtain unnormalized server-level flow field basis functions or unnormalized server-level temperature field basis functions; the unnormalized server-level flow field basis functions or the unnormalized server-level temperature field basis functions are subjected to normalization processing to obtain a plurality of server-level flow field basis functions or a plurality of server-level temperature field basis functions.

[0126] Step 7, a plurality of server-level flow field basis functions or a plurality of server-level temperature field basis functions are selected to construct a server-IGBT-level physical field fast prediction model; wherein the server-IGBT-level physical field fast prediction model is:

[0127]

[0128] Wherein, r is the number of the physical quantity of the server-IGBT-level; f r is a three-dimensional physical field of the rth physical quantity under the corresponding variable input parameter of the server-IGBT-level; b r,jb is the weight coefficient of the three-dimensional physical field of the rth physical quantity of the server-IGBT level in the jth basis function direction; ψ r,j ψ is the jth basis function of the rth physical quantity of the server-IGBT level; j is the number of the server level basis function; l r l is the truncation order of the rth physical quantity of the server-IGBT level.

[0129] In this embodiment, the number of selected basis functions of the flow field server level or the basis functions of the temperature field server level is determined according to the principle that singular values account for a large proportion.

[0130] In this embodiment, the weight coefficient b of the three-dimensional physical field of the rth physical quantity of the server-IGBT level in the jth basis function direction is obtained by projecting the server level snapshot matrix of the preset physical field in each server level basis function direction. r,j The establishment process of the prediction model of b is as follows:

[0131] The server level snapshot matrix of the preset physical field is projected in each server level basis function direction to reconstruct the weight coefficient of the basis function direction under each server level snapshot working condition; based on the weight coefficient of the basis function direction under each server level snapshot working condition, an interpolation method, a regression method or a machine learning method is used to obtain the prediction model of the weight coefficient b of the three-dimensional physical field of the rth physical quantity of the server-IGBT level in the jth basis function direction. r,j

[0132] Specifically, based on the snapshot matrix of the server level preset physical field and the basis function of the server level, a prediction model of the weight coefficient of each basis function corresponding to any variable parameter value is obtained by training; the training can be carried out by first projecting the snapshot matrix in each basis function direction to reconstruct the weight coefficient of each basis function direction under each snapshot working condition, and then using an interpolation method, a regression method or a machine learning method based on the weight coefficient of each basis function direction under each snapshot working condition; preferably, a preselected cubic spline interpolation algorithm is used for training.

[0133] Step 8, constructing a prediction simulation model of the data center IGBT-chip level flow field or temperature field; wherein the prediction simulation model of the data center IGBT-chip level flow field or temperature field is a numerical model based on computational fluid dynamics; in this embodiment, the equations solved in the prediction simulation model of the data center IGBT-chip level flow field or temperature field include mass conservation equation, momentum conservation model, energy conservation equation and standard k-ε turbulence equation; ANSYS ICEPAK software is used for physical modeling of the IGBT-chip level and solving the above equations.

[0134] ​Step 9, determining model parameters of the data center IGBT-chip level according to the predicted simulation model of the data center IGBT-chip level flow field or temperature field; wherein the model parameters of the data center IGBT-chip level include geometric parameters of the IGBT, physical property parameters of the IGBT and operating condition parameters of the IGBT.

[0135] Specifically, the geometric parameters of the IGBT include internal layout information and component size of the IGBT; the physical property parameters of the IGBT include air density of the IGBT, air viscosity of the IGBT, air thermal conductivity coefficient of the IGBT, air specific heat capacity of the IGBT, solid material density of the IGBT, solid material thermal conductivity coefficient of the IGBT and solid material specific heat capacity of the IGBT; and the operating condition parameters of the IGBT include inlet air flow rate of the IGBT, inlet temperature of the IGBT and heat load of the chip.

[0136] Step 10, selecting data center IGBT level variable operation parameters from the model parameters of the data center IGBT-chip, and performing value selection on the data center IGBT level variable operation parameters to obtain a plurality of IGBT level snapshot variable input parameters; wherein the data center IGBT level variable operation parameters include inlet air flow rate of the IGBT, inlet temperature of the IGBT and heat load of the chip.

[0137] In this embodiment, the value selection of the data center IGBT level variable operation parameters is designed in a full factorial manner; specifically, the inlet air flow rate of the IGBT, the inlet temperature of the IGBT and the heat load of the chip are selected as 4 levels, 3 levels and 4 levels respectively, and 48 groups of IGBT level snapshot variable input parameters are designed.

[0138] Step 11, taking the plurality of IGBT level snapshot variable input parameters as inputs of the predicted simulation model of the data center IGBT-chip level flow field or temperature field constructed in step 8, and obtaining a plurality of IGBT level snapshots by solving the simulation model; wherein the plurality of IGBT level snapshots are output in the format of Tecplot files.

[0139] Step 12, generating an IGBT level flow field snapshot matrix or an IGBT level temperature field snapshot matrix according to the plurality of IGBT level snapshots for the IGBT-chip level flow field or temperature field; wherein each column in the IGBT level flow field snapshot matrix or the IGBT level temperature field snapshot matrix represents a corresponding physical field of an IGBT level snapshot.

[0140] Step 13, for the IGBT-chip-level flow field or temperature field, matrix transformation is performed on the snapshot matrix of the IGBT-level flow field or the snapshot matrix of the IGBT-level temperature field to obtain a plurality of IGBT-level flow field basis functions or a plurality of IGBT-level temperature field basis functions, and the plurality of IGBT-level flow field basis functions or the plurality of IGBT-level temperature field basis functions are sorted in descending order according to the amount of information they capture; wherein the operation of matrix transformation on the snapshot matrix of the IGBT-level flow field or the snapshot matrix of the IGBT-level temperature field is similar to the matrix transformation process in step 6, which will not be repeated here.

[0141] Step 14, select the first several IGBT-level flow field basis functions or the first several IGBT-level temperature field basis functions to construct an IGBT-chip-level physical field fast prediction model; wherein the IGBT-chip-level physical field fast prediction model is:

[0142]

[0143] Wherein, r is the number of physical quantities; f r ′ is the three-dimensional physical field of the rth physical quantity under the IGBT-chip-level corresponding variable input parameter; b r ′ ,j′ is the weight coefficient of the three-dimensional physical field of the rth physical quantity of the IGBT-chip-level in the direction of the j'th IGBT-level basis function; ψ′ r,j′ is the j'th IGBT-level basis function of the rth physical quantity of the IGBT-chip-level; j' is the number of IGBT-level basis functions; l r ′ is the truncation order of the rth physical quantity of the IGBT-chip-level.

[0144] In this embodiment, the number of selected IGBT-level flow field basis functions or IGBT-level temperature field basis functions is determined according to the principle that singular values account for a large proportion; the weight coefficient b r ′ ,j′ of the three-dimensional physical field of the rth physical quantity of the IGBT-chip-level in the direction of the j'th IGBT-level basis function is determined according to the principle that singular values account for a large proportion. r ′ ,j′ The prediction model of the weight coefficient b

[0145] Step 15, constructing a predictive simulation model of a data center room-server level flow field or temperature field; wherein the predictive simulation model of the data center room-server level flow field or temperature field is a numerical model based on computational fluid dynamics; wherein equations solved in the predictive simulation model of the data center room-server level flow field or temperature field include a mass conservation equation, a momentum conservation model, an energy conservation equation, and a standard k-ε turbulence equation; and wherein ANSYS ICEPAK software is used to physically model the room-server level and solve the above equations.

[0146] Step 16, obtaining geometric parameters of the data center, physical property parameters of the data center, and operating condition parameters of the data center.

[0147] Specifically, the geometric parameters of the data center include room sizes of the data center, sizes of air conditioners in the data center, sizes of cabinets in the data center, layout parameters of servers in the data center, and layout parameters of air conditioners in the data center; wherein the layout form of the servers in the data center is four rows of racks, each row of racks has four columns of cabinets, and four layers of servers are arranged on each column of cabinets; and wherein the layout form of the air conditioners in the data center is a lower air supply type.

[0148] The physical property parameters of the data center include air density in the data center, air viscosity in the data center, air thermal conductivity in the data center, air specific heat capacity in the data center, solid material density in the data center, solid material thermal conductivity in the data center, solid material specific heat capacity in the data center, server resistance coefficient in the data center, and server internal fan curve in the data center.

[0149] The operating condition parameters of the data center include air conditioner flow rate in the data center, air conditioner outlet temperature in the data center, and heat load data of each server in the data center.

[0150] Step 17, physically modeling and meshing using ANSYS ICEPAK software according to the geometric parameters of the data center, to obtain a mesh model of the data center; wherein unstructured meshes are used in the mesh model of the data center.

[0151] Step 18, inputting the mesh model of the data center, the physical property parameters of the data center, and the operating condition parameters of the data center into the predictive simulation model of the data center room-server level flow field or temperature field constructed in step 15, and outputting to obtain flow field or temperature field information of the data center room-server level; and using an area average method to statistically obtain inlet temperature of each server and inlet air flow rate of each server in the data center.

[0152] Step 19, the inlet air flow rate of each server in the data center, the inlet temperature of each server in the data center, and the heat load of the IGBT inside each server in the data center are substituted into the server-IGBT level physical field fast prediction model established in step 7, and the flow field or temperature field information of each server in the data center is output; the area average method is used to obtain the inlet temperature of each IGBT in the data center and the inlet air flow rate of each IGBT.

[0153] Step 20, the inlet air flow rate of each IGBT in the data center, the inlet temperature of each IGBT in the data center, and the heat load of the chip inside each IGBT in the data center are substituted into the IGBT-chip level physical field fast prediction model established in step 14, and the flow field or temperature field information in each IGBT and chip in the data center is output.

[0154] Test verification:

[0155] The following takes sixteen groups of variable input parameters as an example to evaluate the fidelity of the server-IGBT level physical field fast prediction model and the IGBT-chip level physical field fast prediction model; wherein, the prediction error is as follows:

[0156]

[0157]

[0158] wherein, is the absolute error, is the relative error, x, y, z are coordinates, N cells is the number of grids, is the prediction result of the fast and accurate model constructed, T CFD is the simulation result of the computational fluid dynamics model under the same conditions.

[0159] As shown in the accompanying Figures 2-5 , the computational fluid dynamics simulation result figure of the temperature field of the data center server-IGBT level on the test section under a certain working condition is given in the accompanying Figure 2 , the fast model prediction result figure of the temperature field of the data center server-IGBT level on the test section under a certain working condition is given in the accompanying Figure 3 , the computational fluid dynamics simulation result figure of the flow field of the data center server-IGBT level on the test section under a certain working condition is given in the accompanying Figure 4 , the fast model prediction result figure of the flow field of the data center server-IGBT level on the test section under a certain working condition is given in the accompanying Figure 5 , the fast model prediction result figure of the flow field of the data center server-IGBT level on the test section under a certain working condition is given in the accompanying Figures 4-5It can be seen from the figures that the fast prediction model can accurately capture the characteristics of the flow field and the temperature field, and the absolute error is 1.52℃ and the relative error is 6.79% under the three working conditions.

[0160] As shown in the figures, the figures in the drawings show the simulation results of the temperature field of the IGBT-chip level in the data center under three working conditions. Figures 6-7 Figure 6 The figures in the drawings show the simulation results of the temperature field of the IGBT-chip level in the data center under three working conditions. Figure 7 The figures in the drawings show the simulation results of the temperature field of the IGBT-chip level in the data center under three working conditions. Figures 6-7 It can be seen from the figures that the fast prediction model can accurately capture the characteristics of the temperature field, and the maximum absolute error is 0.3℃ and the maximum relative error is 0.98% under the three working conditions.

[0161] As shown in the figures, the figures in the drawings show the simulation results of the temperature field of the IGBT-chip level in the data center under three working conditions. Figure 8 Figure 8 The figures in the drawings show the simulation results of the temperature field of the IGBT-chip level in the data center under three working conditions. Figure 8 It can be seen from the figures that the fast prediction model can accurately capture the characteristics of the temperature field, and the maximum absolute error is 0.3℃ and the maximum relative error is 0.98% under the three working conditions. Figure 8 According to the simulated flow field and temperature field in the entire data center, the maximum temperature of each chip in each IGBT in the data center is obtained, and the maximum chip temperature reaches 108℃.

[0162] The internal physical field prediction method of the data center described in the embodiment meets the demand for low-cost prediction of the flow field and the temperature field in the entire data center on the basis of given data center parameters, and meets the demand for temperature prediction of all chips in the design and operation of the data center; it is applicable to other geometric structures, other types of data centers and other data center models, etc.; for example, a water-cooled data center, a cold-hot channel isolation layout data center, and an independent computer room layout data center; compared with the traditional computational fluid dynamics simulation, the calculation time is greatly reduced, and for example, the CPU calculation time is reduced by 134 times compared with the computational fluid dynamics simulation for predicting the flow field and the temperature field in the IGBT.

[0163] The description of the related part of the internal physical field prediction system and the equipment of the data center provided in the embodiment can be referred to the detailed description of the corresponding part of the internal physical field prediction method of the data center described in the embodiment, and will not be repeated here.

[0164] The above embodiment is only one of the implementation manners of the technical scheme of the present application, and the scope of protection claimed by the present application is not limited to the embodiment, but also includes any changes, substitutions and other implementation manners easily thought of by those skilled in the art within the technical scope disclosed by the present application.​​

Claims

1. A method for predicting the internal physical field of a data center, characterized in that, include: Obtain the geometric parameters, physical property parameters, and operating condition parameters of the data center; Based on the geometric parameters of the data center, a grid model of the data center is drawn. The grid model of the data center, the physical property parameters of the data center, and the operating condition parameters of the data center are input into the pre-constructed data center room-server level physical field prediction simulation model. The output is the preset physical field information of the data center room-server level, and the entry parameters of each server in the data center are obtained. Acquire the heat load data for each IGBT within the data center; The input parameters of each server in the data center and the heat load data of the IGBT inside each server in the data center are input into the pre-built data center server-IGBT level physical field fast prediction model. The output is to obtain the preset physical field information of each server in the data center and the input parameters of each IGBT in the data center. Acquire thermal load data for each chip within the data center; The input parameters of each IGBT in the data center and the thermal load data of the chip inside each IGBT in the data center are input into the pre-built data center IGBT-chip-level physical field fast prediction model to obtain the preset physical field information of each IGBT and chip in the data center.

2. The method for predicting the internal physical field of a data center according to claim 1, characterized in that, The geometric parameters of the data center include the room size, the air conditioner size, the server rack size, the server layout parameters, and the air conditioner layout parameters. The physical properties of the data center include air density, air viscosity, air thermal conductivity, air specific heat capacity, solid material density, solid material thermal conductivity, solid material specific heat capacity, server drag coefficient, and fan curves within the servers. The operating parameters of the data center include the air conditioning flow rate, the air conditioning outlet temperature, and the heat load of each server in the data center.

3. The method for predicting the internal physical field of a data center according to claim 1, characterized in that, The pre-built data center room-server level physical field prediction simulation model is either a data center room-server level flow field prediction simulation model or a data center room-server level temperature field prediction simulation model. The data center room-server level flow field prediction simulation model or the data center room-server level temperature field prediction simulation model are both numerical models based on computational fluid dynamics. The preset physical field information at the data center room-server level is either the flow field information or the temperature field information at the data center room-server level.

4. The method for predicting the internal physical field of a data center according to claim 1, characterized in that, The inlet parameters for each server in the data center include the inlet temperature and the inlet air velocity of each server; the inlet parameters for each IGBT in the data center include the inlet temperature and the inlet air velocity of each IGBT.

5. The method for predicting the internal physical field of a data center according to claim 1, characterized in that, The pre-built data center server-IGBT level physical field fast prediction model includes a pre-built data center server-IGBT level flow field fast prediction model or a pre-built data center server-IGBT level temperature field fast prediction model. The construction process of the pre-built data center server-IGBT level physics field fast prediction model is as follows: Based on the pre-built predictive simulation model of the physical field of the data center server-IGBT level, the model parameters of the data center server-IGBT level are determined. From the model parameters of the data center server-IGBT level, select the variable operating parameters of the data center server level, and take values ​​for the variable operating parameters of the data center server level to obtain the variable input parameters of several server-level snapshots. The variable input parameters of the aforementioned server-level snapshots are used as input to the pre-built predictive simulation model of the data center server-IGBT level physical field, and the output is a number of server-level snapshots. Based on the aforementioned server-level snapshots, a server-level snapshot matrix of a preset physical field is generated; A matrix transformation is performed on the server-level snapshot matrix of the preset physical field to obtain several server-level basis functions, and the several server-level basis functions are sorted from largest to smallest according to the amount of information they capture. A fast prediction model for server-IGBT level physics is constructed by selecting the first few server-level basis functions; wherein, the fast prediction model for server-IGBT level physics is as follows: Where r is the physical quantity number; f r For the server-IGBT level, the three-dimensional physical field of the r-th physical quantity under variable input parameters; b r,j ψ represents the weighting coefficient of the three-dimensional physical field of the r-th physical quantity in the j-th basis function direction of the server-IGBT level; r,j Let j be the basis function of the r-th physical quantity at the server-IGBT level; j is the number of the server-level basis function; l r Let r be the truncation order of the r-th physical quantity in the server-IGBT stage.

6. The method for predicting the internal physical field of a data center according to claim 5, characterized in that, The weighting coefficient b of the three-dimensional physical field of the r-th physical quantity of the server-IGBT level in the direction of the j-th basis function. r,j The process of establishing the prediction model is as follows: Project the server-level snapshot matrix of the preset physical field onto the basis function direction of each server level to reconstruct the weight coefficients of the basis function direction under each server-level snapshot condition; Based on the weight coefficients of the basis function directions under each server-level snapshot condition, the weight coefficient b of the three-dimensional physical field of the r-th physical quantity of the server-IGBT level in the j-th basis function direction is obtained using interpolation, regression, or machine learning methods. r,j The prediction model.

7. The method for predicting the internal physical field of a data center according to claim 1, characterized in that, The pre-built data center IGBT-chip-level physical field fast prediction model includes a pre-built data center IGBT-chip-level flow field fast prediction model or a pre-built data center IGBT-chip-level temperature field fast prediction model. The construction process of the pre-built data center IGBT-chip-level physical field fast prediction model is as follows: Based on the pre-constructed predictive simulation model of the physical field of data center IGBT-chip level, the model parameters of data center IGBT-chip level are determined. From the model parameters of the data center IGBT-chip level, select the variable operating parameters of the data center IGBT level, and take values ​​for the variable operating parameters of the data center IGBT level to obtain the variable input parameters of several IGBT level snapshots. The variable input parameters of the aforementioned IGBT-level snapshots are used as input to a pre-built predictive simulation model of the data center IGBT-chip-level physical field, and the output is a number of IGBT-level snapshots. The aforementioned IGBT-level snapshots are used to generate an IGBT-level snapshot matrix of a preset physical field. A matrix transformation is performed on the IGBT snapshot matrix of the preset physical field to obtain several IGBT-level basis functions, and the several IGBT-level basis functions are sorted from largest to smallest according to the amount of information they capture; By selecting the first few basis functions of the IGBT, a fast prediction model for the IGBT-chip-level physical field is constructed; wherein, the fast prediction model for the IGBT-chip-level physical field is: Where r is the physical quantity number; f r ′ represents the three-dimensional physical field of the r-th physical quantity under the variable input parameters corresponding to the IGBT chip level; b r ′ ,j′ ψ′ represents the weighting coefficients of the basis function direction of the three-dimensional physical field of the r-th physical quantity at the IGBT chip level in the j′-th IGBT stage. r,j′ Let j′ be the basis function of the j′-th IGBT stage for the r-th physical quantity at the IGBT chip level; j′ is the number of the IGBT stage basis function; l r ′ represents the truncation order of the r-th physical quantity at the IGBT chip level.

8. The method for predicting the internal physical field of a data center according to claim 7, characterized in that, The weighting coefficient b of the three-dimensional physical field of the r-th physical quantity at the IGBT-chip level in the direction of the basis function of the j′-th IGBT level. r ′ ,j′ The process of building the prediction model is as follows: Project the IGBT-level snapshot matrix of the preset physical field onto the basis function direction of each IGBT level to reconstruct the weight coefficients of the basis function direction under the snapshot conditions of each IGBT level. Based on the weight coefficients of the basis function directions under each IGBT-level snapshot condition, the weight coefficient b of the three-dimensional physical field of the r-th physical quantity at the IGBT-chip level in the basis function direction of the j′-th IGBT level is obtained using interpolation, regression, or machine learning methods. r ′ ,j′ The prediction model.

9. An internal physical field prediction system for a data center, characterized in that, include: The first data acquisition module is used to acquire the geometric parameters, physical property parameters, and operating condition parameters of the data center. The mesh model module is used to draw a mesh model of the data center based on the geometric parameters of the data center. The room-server level physics prediction module is used to input the grid model of the data center, the physical property parameters of the data center, and the operating condition parameters of the data center into the pre-built data center room-server level physics prediction simulation model, output the preset physics information of the data center room-server level, and obtain the input parameters of each server in the data center. The second data acquisition module is used to acquire the heat load data of each IGBT in the data center; The server-IGBT level physics prediction module is used to input the inlet parameters of each server in the data center and the heat load data of the IGBT inside each server in the data center into a pre-built data center server-IGBT level physics prediction model, and output the preset physics information of each server in the data center and obtain the inlet parameters of each IGBT in the data center. The third data acquisition module is used to acquire the thermal load data of each chip in the data center; The IGBT-chip-level physical field prediction module is used to input the input parameters of each IGBT in the data center and the thermal load data of the chip inside each IGBT in the data center into a pre-built data center IGBT-chip-level physical field fast prediction model to obtain the preset physical field information of each IGBT and chip in the data center.

10. An internal physics prediction device for a data center, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the data center internal physical field prediction method as described in any one of claims 1-8 when executing the computer program.

Images