Method and device for monitoring the operational status of computer room equipment

Abstract

This invention discloses a method and apparatus for monitoring the operational status of data center equipment, relating to the field of operational status monitoring technology. The method includes: real-time acquisition of multi-dimensional physical field data of a target cabinet area; construction of a reduced-order thermo-hydrodynamic model of the target cabinet; using real-time power consumption data at the equipment level as the main heat source input to drive the reduced-order thermo-hydrodynamic model for real-time simulation, obtaining first predicted physical field data; performing data assimilation processing, dynamically correcting implicit parameters, and forming a calibrated real-time digital twin; performing multi-index thermal state analysis, simulating and optimizing control strategies, and generating equipment operation intervention commands. This invention solves the technical problems of existing data center cabinet thermal state management, such as lagging thermal state perception, mismatch between control strategies and equipment status, and insufficient coordinated optimization of energy consumption and temperature control. It achieves real-time thermal state perception and intelligent control of the data center, improving the real-time performance and accuracy of data center equipment operational status monitoring.

Description

Technical Field

[0001] This invention relates to the field of operational status monitoring technology, and more specifically to a method and apparatus for monitoring the operational status of computer room equipment. Background Technology

[0002] In the field of data center cabinet equipment operation and management, traditional technologies rely heavily on manual inspections and single parameter monitoring methods. These methods suffer from problems such as incomplete three-dimensional physical field data acquisition, lagging airflow organization and heat dissipation condition monitoring, mismatch between cooling strategies and equipment power consumption and thermal characteristics, and low temperature control adjustment accuracy. They cannot achieve dynamic real-time perception of the cabinet thermal field, nor can they coordinate and optimize energy consumption control and thermal regulation. This results in insufficient stability of data center equipment operation and low energy utilization. At the same time, they lack accurate quantitative assessment and intelligent control methods for the thermal state of equipment, making it difficult to meet the market demand for refined temperature control management, energy-saving operation, and long-term stable operation of equipment in high-density cabinet scenarios.

[0003] Existing data center rack thermal management suffers from technical problems such as delayed thermal sensing, mismatch between control strategies and equipment status, and insufficient coordinated optimization of energy consumption and temperature control. Summary of the Invention

[0004] This application provides a method and apparatus for monitoring the operating status of data center equipment, which addresses the technical problems of existing data center cabinet thermal management, such as lagging thermal sensing, mismatch between control strategies and equipment status, and insufficient coordinated optimization of energy consumption and temperature control.

[0005] In view of the above problems, this application provides a method and apparatus for monitoring the operating status of computer room equipment.

[0006] A first aspect of this application provides a method for monitoring the operational status of equipment in a computer room, the method comprising:

[0007] Multidimensional physical field data of the target cabinet area is collected in real time. This multidimensional physical field data includes at least three-dimensional airflow vector field data acquired by an airflow imaging sensor grid, temperature distribution data acquired by a temperature sensor array, and real-time power consumption data at the equipment level acquired by a power monitoring device. Based on the physical structural parameters of the target cabinet, a reduced-order thermo-hydrodynamic model of the target cabinet is constructed. Using the real-time power consumption data at the equipment level as the main heat source input, the reduced-order thermo-hydrodynamic model is driven to perform real-time simulation to obtain first predicted physical field data. The multidimensional physical field data and the first predicted physical field data are assimilated to dynamically correct the implicit parameters in the reduced-order thermo-hydrodynamic model, making the model output infinitely close to the real state, forming a calibrated real-time digital twin. Based on the real-time digital twin, multi-index thermal situation analysis is performed, and based on the generated multi-dimensional thermal situation analysis results, control strategy simulation optimization is performed to generate equipment operation intervention commands.

[0008] A second aspect of this application provides a device for monitoring the operational status of computer room equipment, the device comprising:

[0009] The system includes a physical field data acquisition module for real-time acquisition of multi-dimensional physical field data of the target cabinet area. This multi-dimensional physical field data includes at least three-dimensional airflow vector field data acquired by an airflow imaging sensor grid, temperature distribution data acquired by a temperature sensor array, and real-time power consumption data at the equipment level acquired by a power monitoring device. A reduced-order model construction module is used to construct a reduced-order thermodynamic model of the target cabinet based on its physical structural parameters. A real-time simulation module uses the equipment-level real-time power consumption data as the primary heat source input to drive the reduced-order thermodynamic model in real-time simulation, obtaining first predicted physical field data. A digital twin generation module assimilates the multi-dimensional physical field data with the first predicted physical field data, dynamically correcting the implicit parameters in the reduced-order thermodynamic model to make the model output infinitely close to the real state, forming a calibrated real-time digital twin. A control strategy simulation optimization module performs multi-index thermal state analysis based on the real-time digital twin, and performs control strategy simulation optimization based on the generated multi-dimensional thermal state analysis results, generating equipment operation intervention commands.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] The system collects multidimensional physical field data of the target rack area in real time; constructs a reduced-order thermo-hydrodynamic model of the target rack; uses the real-time power consumption data at the device level as the main heat source input to drive the reduced-order thermo-hydrodynamic model for real-time simulation, obtaining first predicted physical field data; assimilates the multidimensional physical field data with the first predicted physical field data, dynamically corrects the implicit parameters in the reduced-order thermo-hydrodynamic model, making the model output infinitely close to the real state, forming a calibrated real-time digital twin; based on the real-time digital twin, performs multi-index thermal state analysis, and based on the generated multi-dimensional thermal state analysis results, simulates and optimizes control strategies to generate equipment operation intervention commands. This achieves the technical effect of realizing real-time perception and intelligent control of the data center's thermal state, improving the real-time performance and accuracy of data center equipment operation status monitoring. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0013] Figure 1 This is a schematic flowchart of a method for monitoring the operating status of computer room equipment provided in an embodiment of this application;

[0014] Figure 2 This is a schematic diagram of the structure of a device for monitoring the operating status of computer room equipment provided in an embodiment of this application.

[0015] Figure labeling: 10 Physical field data acquisition module, 20 Reduced-order model construction module, 30 Real-time simulation module, 40 Digital twin generation module, 50 Regulation strategy simulation optimization module. Detailed Implementation

[0016] This application provides a method and device for monitoring the operating status of data center equipment, which addresses the technical problems of existing data center cabinet thermal management, such as lagging thermal sensing, mismatch between control strategies and equipment status, and insufficient coordinated optimization of energy consumption and temperature control.

[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0018] Example 1, as Figure 1As shown, this application provides a method for monitoring the operating status of computer room equipment, the method comprising:

[0019] Step S100: Real-time acquisition of multi-dimensional physical field data of the target cabinet area. The multi-dimensional physical field data includes at least three-dimensional airflow vector field data obtained by the airflow imaging sensor grid, temperature distribution data obtained by the temperature sensor array, and real-time power consumption data at the device level obtained by the power monitoring device.

[0020] Specifically, real-time acquisition of multi-dimensional physical field data is carried out on the target rack area. A grid of airflow imaging sensors is deployed to capture three-dimensional airflow vector field data within the rack, accurately obtaining vector characteristics such as airflow direction and velocity. Temperature sensor arrays are used to perform full-area temperature measurement on the rack area, acquiring temperature distribution data in the spatial dimension to reflect the real-time temperature status at different locations. Power monitoring devices are used to achieve refined energy consumption acquisition, obtaining real-time power consumption data at the device level, accurately grasping the heat source output of each device. At the same time, pressure gradient data between hot and cold aisles can be simultaneously acquired through micro differential pressure sensors deployed on the top and / or bottom of the rack, thereby forming a multi-dimensional, all-element multi-dimensional physical field dataset of the target rack area.

[0021] Step S200: Based on the physical structural parameters of the target cabinet, construct a reduced-order thermo-hydrodynamic model of the target cabinet.

[0022] Specifically, a reduced-order thermodynamic model is constructed based on the physical structural parameters of the target cabinet. First, the geometric shape, internal equipment layout, and specific locations and dimensions of ventilation openings of the target cabinet are determined based on these physical structural parameters, thereby clarifying the simulation solution domain of the model. Then, at the boundary of the simulation solution domain, the boundary conditions of the airflow inlet and outlet of the model are defined according to the actual locations and operating modes of the air conditioning supply and return air vents. Subsequently, an original high-dimensional governing equation model that can accurately describe the airflow and heat transfer process inside the cabinet is established within the simulation solution domain. Characteristic modal analysis is performed on this model to extract multiple dominant physical modes that can characterize the dominant flow and heat transfer characteristics inside the cabinet. Finally, a low-dimensional subspace is constructed based on the extracted dominant physical modes, and the original high-dimensional governing equation model is projected onto this low-dimensional subspace, ultimately obtaining a reduced-order thermodynamic model that can achieve real-time solution in seconds.

[0023] Step S300: Using the device-level real-time power consumption data as the main heat source input, drive the thermohydrodynamic reduced-order model to perform real-time simulation to obtain the first predicted physical field data.

[0024] Specifically, the collected real-time power consumption data at the device level is used as the main heat source input for the thermodynamics reduced-order model. First, the power consumption data is spatially mapped to the volumetric heat source intensity of the corresponding computing unit in the model based on the actual physical location of each device in the target cabinet. Then, the volumetric heat source intensity and the boundary conditions of the airflow inlet and outlet preset by the model are used as core input parameters and loaded into the thermodynamics reduced-order model. Subsequently, the transient solver of the model is started to carry out rapid numerical calculations to complete the real-time simulation calculation of the model. Finally, the first predicted physical field data containing the predicted three-dimensional airflow field and the predicted temperature distribution are output.

[0025] Step S400: Perform data assimilation processing on the multidimensional physical field data and the first predicted physical field data, dynamically correct the implicit parameters in the thermohydrodynamic reduced-order model, so that the model output infinitely approximates the real state, and form a calibrated real-time digital twin.

[0026] Specifically, data assimilation processing is performed on the collected multidimensional physical field data and the first predicted physical field data obtained from model simulation. First, the two types of data are spatiotemporally registered to ensure that they are compared and analyzed under the same spatial coordinates and timestamps. Then, the residuals between the measured values ​​of each physical quantity in the multidimensional physical field data and the predicted values ​​of the corresponding physical quantities in the first predicted physical field data are calculated. Subsequently, an objective function is constructed with the implicit parameters in the thermohydrodynamic reduced-order model as optimization variables. This objective function represents the weighted sum of squares of the above residuals. The implicit parameters include the flow resistance correction coefficient representing the change in local resistance caused by the absence of cables and blind plates, the surface convection heat transfer coefficient representing the actual heat dissipation efficiency of the equipment, and the equivalent leakage area coefficient representing the leakage characteristics of the cabinet gaps. Finally, the parameter values ​​of each implicit parameter are repeatedly adjusted through iterative optimization until the value of the objective function is lower than the preset threshold. The thermohydrodynamic reduced-order model that has completed parameter correction and outputs an output that can infinitely approximate the actual operating state of the cabinet is then used as the calibrated real-time digital twin.

[0027] Step S500: Based on the real-time digital twin, perform multi-index thermal situation analysis, and based on the generated multi-dimensional thermal situation analysis results, perform regulation strategy simulation optimization and generate equipment operation intervention instructions.

[0028] Specifically, multi-index thermal situation analysis is conducted based on the calibrated real-time digital twin. First, the output three-dimensional airflow vector field data is called to identify airflow vortex regions and airflow stagnation areas. The airflow short-circuit rate and server air intake uniformity index are calculated to complete the assessment of local airflow organization health and generate corresponding indicators. Then, temperature field data is called in combination with equipment air intake requirements to identify hidden thermal risk points. At the same time, cooling energy efficiency indicators are calculated based on airflow and temperature data. The above indicators are combined to generate multi-dimensional thermal situation analysis results. Subsequently, based on the results, the control strategy simulation optimization is carried out. First, adjustable control variables such as precision air conditioning supply air temperature setpoint, supply fan speed, and intelligent guide vane opening angle are defined. Different control variables are set in the real-time digital twin to simulate combinations. For each combination, fast simulation is performed to predict the corresponding thermal situation. At the same time, the total power consumption of the system, the hottest spot temperature, and the airflow short-circuit rate are calculated. Finally, with the goal of optimizing these performance indicators, the optimal control strategy is selected and extracted within the feasible domain of the control variables and converted into equipment operation intervention command output.

[0029] In one possible implementation, step S100 further includes:

[0030] The multidimensional physical field data also includes pressure gradient data between hot and cold aisles acquired by micro differential pressure sensors deployed at the top and / or bottom of the cabinet.

[0031] Specifically, the collected multidimensional physical field data also includes pressure gradient data between hot and cold aisles, accurately acquired by micro differential pressure sensors specifically deployed at the top and / or bottom of the target cabinet. This type of data can reflect the pressure difference and gradient change characteristics between the hot and cold aisles in real time, accurately characterizing the pressure field distribution state of the hot and cold aisles. It supplements the key physical field measured data of the pressure dimension for the subsequent construction of thermodynamics reduction model, simulation calculation, and parameter correction of data assimilation, making the collected multidimensional physical field data more comprehensive and the data information more complete.

[0032] In one possible implementation, step S200 further includes:

[0033] Step S210: Based on the physical structure parameters, determine the geometric shape, internal equipment layout, ventilation opening location and size of the target cabinet to define the simulation solution domain of the reduced-order thermohydrodynamic model.

[0034] Step S220: On the boundary of the simulation solution domain, define the boundary conditions of the airflow inlet and outlet of the reduced-order thermodynamic model according to the position and pattern of the air supply and return air vents.

[0035] Step S230: Within the simulation solution domain, establish the original high-dimensional control equation model to describe the airflow and heat transfer process.

[0036] Step S240: Perform eigenmode analysis on the original high-dimensional control equation model to extract multiple dominant physical modes that can characterize the dominant flow and heat transfer characteristics inside the target cabinet.

[0037] Step S250: Based on the extracted multiple dominant physical modes, a low-dimensional subspace is constructed, and the original high-dimensional control equation model is projected onto the low-dimensional subspace to obtain a reduced-order thermohydrodynamic model that can be solved in real time at the second level.

[0038] Specifically, based on the actual physical structural parameters of the target rack, the overall geometric shape of the target rack, the actual installation layout and spatial position of various servers and supporting equipment inside the rack are determined one by one. At the same time, the specific layout, shape and size of various ventilation openings around the rack are clarified. These physical characteristics are transformed into spatial parameters for model construction, thereby clearly defining the simulation solution domain of the reduced-order thermodynamics model.

[0039] On various boundaries of the defined thermo-hydrodynamic reduced-order model simulation solution domain, the physical locations of the air supply and return vents of the target rack are combined with the actual air supply and return modes of the air conditioner. The air inlet and outlet locations of the model are precisely defined. At the same time, based on the actual air supply velocity, air volume, air supply temperature and return air parameters of the air conditioner, input boundary conditions such as velocity, temperature and flow rate at the air inlet are set, as well as output boundary conditions such as pressure and return flow constraints at the air outlet. This ensures that the airflow inlet and outlet boundary rules of the model are highly matched with the actual air conditioning airflow operation status of the rack.

[0040] Within the defined simulation solution domain, based on the fundamental theories of computational fluid dynamics and heat transfer, and employing the finite volume method, and based on the spatial grid discretization results of the solution domain, in each discretized computational unit, the continuity equation, momentum equation, and energy equation characterizing heat transfer are constructed and coupled to describe airflow motion. At the same time, the standard k-ε turbulence equation is introduced in combination with the airflow turbulence characteristics within the cabinet, incorporating actual physical constraints such as volumetric heat source terms for equipment heat dissipation, convective heat transfer and thermal radiation boundary source terms of solid walls, etc. Through the form of a set of equations, a high-dimensional control equation model is built that can accurately characterize the entire process of airflow, turbulent diffusion and multi-form heat transfer within the solution domain.

[0041] Eigenmode analysis was performed on the constructed original high-dimensional control equation model. Using the intrinsic orthogonal decomposition algorithm, multiple sets of flow field and temperature field snapshot data under the thermodynamic scenario of the target cabinet were first obtained through numerical simulation to construct a high-dimensional snapshot matrix containing physical quantities such as airflow velocity, pressure, and temperature. Then, singular value decomposition was performed on the snapshot matrix to obtain the corresponding eigenvalues ​​and orthogonal eigenmode vectors. A threshold was set based on the cumulative contribution rate of the eigenvalues, such as 95% and above. The top few eigenmode vectors whose cumulative contribution rate met the threshold were selected, and redundant modes with low contribution were eliminated. Finally, multiple dominant physical modes that can accurately characterize the dominant airflow law and core heat transfer characteristics inside the target cabinet were extracted, retaining the core physical information of the thermodynamic process.

[0042] Using multiple extracted dominant physical modes as core basis vectors, a reduced-order thermodynamic model is constructed using the Galerkin projection method. First, the dominant physical modes are orthogonalized and used as basis vectors to construct a low-dimensional subspace characterizing the core physical laws of the cabinet's thermodynamics. This subspace retains only the physical dimensions that play a key role in airflow and heat transfer. Then, the original high-dimensional control equation model describing the airflow and heat transfer process within the cabinet is orthogonally projected onto this low-dimensional subspace using the Galerkin projection algorithm. The high-dimensional flow field and temperature field physical variables are linearly represented by the basis vectors of the low-dimensional subspace, simplifying and reconstructing the original high-dimensional control equations. This eliminates redundant high-dimensional components with low contribution in the original model, significantly reducing the computational dimension and workload. Finally, a reduced-order thermodynamic model based on low-dimensional subspace reconstruction is obtained, which can achieve real-time solution in seconds and accurately preserves the core thermodynamic characteristics of the cabinet.

[0043] In one possible implementation, step S300 further includes:

[0044] Step S310: Based on the physical location of the device in the target cabinet, the real-time power consumption data at the device level is spatially mapped to the volumetric heat source intensity of the corresponding calculation unit in the reduced-order thermodynamic model.

[0045] Step S320: The volumetric heat source intensity and the boundary conditions of the airflow inlet and outlet are used as input parameters and loaded into the thermohydrodynamic reduced-order model.

[0046] Step S330: Start the transient solver of the thermohydrodynamic reduced-order model, perform rapid numerical calculations, and output the first predicted physical field data containing the predicted three-dimensional airflow field and the predicted temperature distribution.

[0047] Specifically, the real-time power consumption data of each server and supporting device in the target rack is first classified and organized to accurately match the real-time power consumption value of each device. Then, based on the actual physical installation position, space ratio and three-dimensional coordinate information of each device in the target rack, combined with the preset spatial grid division rules in the thermo-hydrodynamic reduced-order model, the real-time power consumption data of each device is accurately mapped to the calculation unit of the corresponding device in the model. At the same time, according to the principle of heat-work conversion, the power consumption value is converted into the volume heat source intensity of the calculation unit, so as to achieve accurate spatial correspondence between the physical quantity of heat dissipation of the actual equipment in the rack and the heat source parameters of the model calculation unit, so that the heat source input of the model is completely consistent with the actual heat dissipation distribution of the equipment in the rack.

[0048] The volumetric heat source intensity of each computational unit in the model obtained through spatial mapping is integrated with the airflow inlet and outlet boundary conditions defined in the previous thermodynamics reduction model. The airflow boundary conditions include core parameters such as inlet air velocity, temperature, flow rate, and outlet pressure constraints. The two types of parameters are converted and matched according to the model's preset input interface specifications and uniformly loaded into the computational kernel of the thermodynamics reduction model. This completes the configuration of all core input parameters before the model's real-time simulation, ensuring that the heat source input and airflow boundary conditions of the model simulation are highly consistent with the actual operating state of the target rack.

[0049] After loading and configuring the parameters for the volumetric heat source intensity and airflow boundary conditions, the transient solver built into the thermohydrodynamic model is activated. This solver is adapted to the low-dimensional computational characteristics of the model after dimensionality reduction and can perform efficient and rapid numerical calculations based on simplified core control equations to accurately solve the thermohydrodynamic processes that change dynamically over time within the cabinet. During the solution process, the dynamic distribution characteristics of the flow field and temperature field are calculated iteratively and simultaneously. Finally, the predicted three-dimensional airflow field data that fully reflects the airflow motion state within the cabinet and the predicted temperature distribution data that characterizes the temperature changes at various spatial locations are output. Together, these two data constitute the first predicted physical field data, providing complete data support for the model's predicted dimensions for subsequent assimilation and comparison with measured multi-dimensional physical field data.

[0050] In one possible implementation, step S400 further includes:

[0051] Step S410: Perform spatiotemporal registration between the multidimensional physical field data and the first predicted physical field data to ensure that the two are compared under the same spatial coordinates and timestamps.

[0052] Step S420: Calculate the residual between the measured value of each physical quantity in the multidimensional physical field data and the predicted value of the corresponding physical quantity in the first predicted physical field data.

[0053] Step S430: Construct an objective function with the implicit parameters of the reduced-order thermohydrodynamic model as optimization variables, wherein the objective function characterizes the weighted sum of squares of the residuals.

[0054] Step S440: Through iterative optimization, repeatedly adjust the parameter values ​​of the implicit parameters until the value of the objective function is lower than a preset threshold, and use the reduced-order thermohydrodynamic model at this time as the calibrated real-time digital twin.

[0055] Specifically, firstly, the multidimensional physical field data and the first predicted physical field data are aligned to the same global coordinate system using a coordinate transformation algorithm, while unifying the resolution and node spacing of the spatial grid. Then, time dimension registration is performed, and the sampling time and time step of the two are calibrated using a timestamp alignment method. Data at time misalignment points are filled in by interpolation, ultimately ensuring that the two are compared under completely consistent spatial coordinates and timestamps.

[0056] For physical quantities such as three-dimensional airflow velocity, temperature, and pressure, the difference between the measured values ​​and the corresponding predicted values ​​is calculated under unified spatial coordinates and timestamps after spatiotemporal registration, and residual data is obtained to quantify the degree of deviation between the measured and predicted values.

[0057] A weighted least squares algorithm is used to construct the objective function, with three key implicit parameters—flow resistance correction coefficient, surface convection heat transfer coefficient, and equivalent leakage area coefficient—as optimization variables. Under unified coordinates and timestamps after spatiotemporal registration, the residuals between the measured and predicted values ​​of the temperature and airflow fields are calculated. Differentiated weights are assigned according to the importance of physical quantities, and an objective function with the weighted sum of squares of residuals as its core structure is constructed. The function expression is the sum of the squares of the temperature and velocity residuals of all grid nodes and time steps multiplied by their corresponding weights. This function directly reflects the overall deviation between the model prediction and the measured data. By minimizing this objective function, the optimal identification of the three types of implicit parameters is achieved, providing a clear mathematical evaluation standard for subsequent iterative calibration.

[0058] A sequential quadratic programming iterative optimization algorithm is adopted, using the constructed residual weighted sum of squares objective function as the convergence criterion. Implicit parameters such as convective heat transfer coefficient, turbulence model correction coefficient, and thermal conductivity coefficient are used as iterative adjustment variables. Starting from the initial parameter values, each iteration calculates the gradient and Hessian matrix of the objective function to determine the optimal update direction and step size of the parameters, thereby adjusting the values ​​of the implicit parameters. During the iteration process, the objective function value is continuously recalculated to determine whether it is lower than the preset accuracy threshold. When the objective function meets the convergence condition, the iteration stops and the current implicit parameter combination is locked, completing the parameter calibration of the thermodynamic reduced-order model. Finally, a calibrated real-time digital twin that can accurately replicate the real thermodynamic operating state of the target cabinet and achieve second-level real-time simulation is obtained.

[0059] In one possible implementation, step S430 further includes:

[0060] Step S431: The implicit parameters include the flow resistance correction coefficient for characterizing the change in local resistance caused by the absence of cables and blind plates, the surface convection heat transfer coefficient for characterizing the actual heat dissipation efficiency of the equipment, and the equivalent leakage area coefficient for characterizing the leakage characteristics of the cabinet gaps.

[0061] Specifically, the implicit parameters include three core characterization quantities: First, the flow resistance correction coefficient, which characterizes the change in local fluid resistance caused by cable shortages. This parameter is used to quantify the change in flow resistance characteristics in areas with missing cable laying and is a key correction basis for the resistance calculation of fluid transport systems. Second, the surface convective heat transfer coefficient, which reflects the heat exchange efficiency between the equipment surface and the surrounding medium. It quantifies the convective heat transfer capacity of the equipment surface and directly determines the heat exchange performance evaluation results of the heat exchange equipment. Third, the equivalent leakage area coefficient, which characterizes the leakage characteristics of the cabinet gaps. This parameter is used to equivalently describe the flow capacity of the leakage channels in the cabinet gaps and is a core quantitative indicator for evaluating the sealing performance and leakage loss of the cabinet.

[0062] In one possible implementation, step S500 further includes:

[0063] Step S510: Call the three-dimensional airflow vector field data output by the real-time digital twin to perform a local airflow organization health assessment and generate a local airflow health index.

[0064] Step S520: Call the temperature field data output by the real-time digital twin, and in combination with the equipment air intake requirements, identify potential hot spots that are not currently exceeding the standard but have poor airflow organization and are at risk of overheating, and obtain hidden thermal risk points.

[0065] Step S530: Calculate the cooling energy efficiency index based on the airflow and temperature data output by the real-time digital twin.

[0066] Step S540: Combine the local airflow health indicators, hidden thermal risk points, and cooling energy efficiency indicators to generate multivariate thermal situation analysis results.

[0067] Specifically, the system retrieves the full data of the high-fidelity three-dimensional airflow vector field output by the real-time digital twin after calibration, and conducts a systematic assessment of the local airflow organization health status within the target rack. This is achieved by identifying airflow vortex regions where streamlines form closed loops and airflow stagnation areas where wind speeds are consistently below a set threshold. The system also tracks the short-circuit trajectory of cold aisle airflow that directly enters the hot aisle without passing through the server's heat source and calculates the airflow short-circuit rate. Simultaneously, the system calculates the ratio of the standard deviation to the average value of the wind speed distribution on the grid of each server's air inlet cross-section to obtain the airflow uniformity index. The system integrates and analyzes the above-mentioned abnormal airflow area identification results with the quantitative calculation indicators to ultimately generate a local airflow health index that can comprehensively and accurately characterize the local airflow organization operation status within the rack.

[0068] By retrieving high-precision temperature field data from the real-time output of the calibrated digital twin, and combining this data with the rated inlet air temperature, inlet air temperature difference, and heat dissipation adaptation requirements of core components such as servers and supporting equipment in the data center, the system performs correlation analysis on the temperature distribution values, temperature change trends, and surrounding airflow organization status of various spatial locations within the server rack. It focuses on screening spatial locations where the current temperature value does not exceed the equipment's safe operating threshold, but where poor airflow organization issues such as airflow vortices, stagnation, and short circuits prevent effective heat removal and lead to a continuous heat accumulation trend. This accurately identifies and marks these areas with potential overheating risks, ultimately obtaining hidden thermal risk points within the target server rack and providing precise location data for early warning of thermal risks.

[0069] Based on high-precision airflow and temperature field data output from real-time digital twins, a rack-level energy efficiency quantification algorithm is used to calculate cooling energy efficiency indicators. First, key parameters such as cold aisle supply air temperature, hot aisle return air temperature, equipment inlet air temperature, and rack airflow are extracted. Then, the core indicators are solved by using the heat removal index calculation formula (the ratio of effective cooling capacity utilization to total cooling capacity) and the return air temperature index calculation formula (the normalized result of the deviation between the return air temperature and the set target value). At the same time, auxiliary parameters such as airflow short-circuit rate and cooling capacity loss rate are combined to comprehensively evaluate the cooling capacity utilization efficiency and cooling system operating performance. Finally, standardized cooling energy efficiency indicators including rack-level heat removal index, return air temperature index, and cooling capacity utilization efficiency are output, providing a quantitative basis for the optimization and energy-saving control of data center cooling systems.

[0070] First, dimensionless normalization was performed on local airflow health indicators, namely airflow short-circuit rate, inlet air uniformity index, vortex / stagnation zone ratio, and cooling efficiency indicators, namely rack-level heat removal index, return air temperature index, and cooling capacity utilization efficiency. This standardized the evaluation scale of the indicators and assigned differentiated weights. Hidden thermal risk points were then gridded and thermal distribution maps were drawn according to their spatial location and risk level. Next, a multi-dimensional correlation analysis algorithm was used to uncover the inherent coupling relationship between airflow organization health status and the formation of hidden thermal risks and low cooling efficiency indicators, combined with data center data. The system employs a dual evaluation standard for equipment operation safety and energy efficiency, constructing a hierarchical thermal status evaluation model to sequentially classify and determine local thermal status, regional thermal status, and the overall thermal status of the server rack. Finally, through visualization fusion technology, it integrates quantified indicator scores, spatial distribution of risk points, energy efficiency level ratings, and correlation analysis results between indicators to generate a multi-dimensional thermal status analysis result that includes quantitative indicator results, risk distribution maps, energy efficiency ratings, problem cause tracing, and comprehensive thermal status determination levels. This achieves a comprehensive, multi-dimensional quantitative and visual presentation of the server rack's thermal operation status.

[0071] In one possible implementation, step S510 further includes:

[0072] Step S511: In the three-dimensional airflow vector field data output by the real-time digital twin, identify the airflow vortex region where streamlines form a closed loop, and the airflow stagnation region where the wind speed is continuously lower than the set wind speed threshold.

[0073] Step S512: Track the airflow trajectory that starts from the cold aisle and enters the hot aisle directly without passing through any server heat source, and calculate the ratio of its mass flow rate to the total air supply volume as the airflow short-circuit rate.

[0074] Step S513: For each server, calculate the ratio of the standard deviation to the average value of the wind speed distribution on the grid of its air inlet section, and use it as the air inlet uniformity index.

[0075] Specifically, based on the rack-wide structured grid three-dimensional airflow vector field data output by the real-time digital twin, the airflow trajectory is first tracked node by node using the fourth-order Runge-Kutta streamline integration method. Combined with the closed-loop judgment logic of spatial coordinate threshold matching, streamline regions where the spatial deviation between the endpoint and the starting point is less than the preset accuracy threshold and forms a continuous cyclic trend are marked as airflow vortex regions. At the same time, according to the process requirements for effective heat dissipation of the equipment in the computer room, the minimum effective air intake velocity threshold is set, and the wind speed data of each grid node is continuously verified across the entire domain. Spatial regions where the wind speed is lower than the threshold for multiple consecutive time steps are screened and marked as airflow stagnation areas. The spatial coordinate positioning, range delineation and visualization marking of the two types of abnormal airflow organization areas are completed simultaneously.

[0076] Based on the three-dimensional airflow vector field data output by the real-time digital twin, a virtual airflow particle emission source is set at the cold aisle air inlet and the particles are given the property of following the flow. The particle trajectory is dynamically tracked throughout the entire process using the Lagrange method. At the same time, a spatial mask region for all server heat-generating elements in the rack is constructed. Spatial collision detection is performed on each particle trajectory to determine whether it passes through the heat-generating element mask region during its movement. Short-circuit particle trajectories that do not pass through any server heat-generating elements and directly enter the hot aisle from the cold aisle are filtered out. Then, the airflow mass flow rate corresponding to all short-circuit trajectories is calculated. Combined with the real-time calculation value of the total air volume of the rack in the digital twin, the ratio of short-circuit airflow mass flow rate to total air volume is calculated. This ratio is standardized and used as the airflow short-circuit rate to quantitatively characterize the airflow short-circuit problem in the rack.

[0077] The physical cross-section of the air inlet of each server in the rack is divided into structured meshes to generate equidistant wind speed monitoring mesh cells. Then, the real-time wind speed scalar value of the corresponding mesh cell of each server's air inlet is extracted from the three-dimensional airflow vector field data output by the real-time digital twin, forming a wind speed dataset for a single server's air inlet. Subsequently, statistical calculations are performed on this dataset to solve for the arithmetic mean and standard deviation of the wind speed distribution. The ratio of the standard deviation to the mean is then calculated, and this dimensionless ratio is used as the air inlet uniformity index to characterize the uniformity of the air speed distribution at a single server's air inlet. The higher the index value, the more uneven the air speed distribution at the server's air inlet, and vice versa.

[0078] In one possible implementation, step S500 further includes:

[0079] Step S550: Define a set of adjustable control variables, which include at least the supply air temperature setpoint of the precision air conditioner, the fan speed, and the opening and closing angle of the intelligent guide vanes.

[0080] Step S560: In the real-time digital twin, the control variables are set to different combinations to be simulated.

[0081] Step S570: For each combination of control variables, perform a fast simulation in the real-time digital twin to predict the thermal state under the corresponding combination and calculate the corresponding performance indicators, including total system power consumption, hottest spot temperature, and airflow short-circuit rate.

[0082] Step S580: With the goal of optimizing the performance indicators, perform an optimization search within the feasible domain of the control variables, filter and extract the optimal control strategy, and convert it into an equipment operation intervention command output.

[0083] Specifically, based on the actual operating characteristics of the data center cooling control system and the adjustable range of the equipment, a set of core adjustable control variables for optimizing the thermal state of the server rack is defined. This set of control variables at least covers the supply air temperature setpoint of the precision air conditioner, the air conditioner fan speed, and the opening and closing angle of the intelligent guide vanes for the cold / hot aisle of the server rack. At the same time, the physical adjustment range and step threshold of each control variable are clearly defined.

[0084] Based on the defined control variables and the preset feasible adjustment range and adjustment step size of each variable, the orthogonal experimental design method is used to generate multiple sets of control variable combinations to be simulated, covering the adjustment range of the variables. The parameter values ​​of each combination are mapped one by one to the corresponding control module of the real-time digital twin, completing the parameter configuration and simulation scenario construction of different control variable combinations in the digital twin, laying the foundation for subsequent thermal state simulation tests under multiple combinations.

[0085] For each combination of control variables that has been configured in the real-time digital twin, the model's second-level transient fast solver is invoked to perform thermo-hydrodynamic simulation calculations. Based on the simulation results, the overall thermal state of the target cabinet, such as the global temperature field and the three-dimensional airflow vector field, is accurately predicted. At the same time, core quantitative indicators are extracted from the simulation output data to calculate the total system power consumption of the data center cooling system and equipment operation, the hottest temperature in all spatial grid nodes in the cabinet, and the airflow short-circuit rate that characterizes the rationality of airflow organization, forming a one-to-one matching dataset of each combination of control variables and its corresponding performance indicators.

[0086] First, with the comprehensive optimization of total system power consumption, hottest temperature, and airflow short-circuit rate as the goal, weight coefficients that fit the operational needs of the data center are assigned to each performance index, and a multi-objective optimization function is constructed. Within the preset feasible domain of the control variables, the NSGA-Ⅲ multi-objective genetic algorithm is used to conduct optimization search. After multiple rounds of iterative calculations, the optimal control strategy in the Pareto optimal solution set is selected. Then, the parameters in the strategy, such as the precision air conditioning supply air temperature setpoint, the supply fan speed, and the opening and closing angle of the intelligent guide vanes, are analyzed. The parameters are standardized and converted according to the communication protocols and command formats of the control equipment in the data center. Structured operation intervention commands that can be directly sent to the corresponding equipment are generated and output, realizing the accurate implementation of the simulation optimization strategy into executable commands of the field equipment.

[0087] Example 2, based on the same inventive concept as the method for monitoring the operating status of computer room equipment in the foregoing examples, such as... Figure 2 As shown, this application provides a device for monitoring the operational status of computer room equipment. The device and method embodiments in this application are based on the same inventive concept. The device includes:

[0088] The physical field data acquisition module 10 is used to acquire multi-dimensional physical field data of the target cabinet area in real time. The multi-dimensional physical field data includes at least three-dimensional airflow vector field data obtained by the airflow imaging sensor grid, temperature distribution data obtained by the temperature sensor array, and real-time power consumption data at the device level obtained by the power monitoring device.

[0089] The reduced-order model construction module 20 is used to construct a reduced-order thermo-hydrodynamic model of the target cabinet based on the physical structural parameters of the target cabinet.

[0090] The real-time simulation module 30 is used to drive the thermohydrodynamic reduced-order model to perform real-time simulation using the device-level real-time power consumption data as the main heat source input, so as to obtain the first predicted physical field data.

[0091] The digital twin generation module 40 is used to perform data assimilation processing on the multidimensional physical field data and the first predicted physical field data, dynamically correct the implicit parameters in the thermohydrodynamic reduced-order model, so that the model output infinitely approximates the real state, and form a calibrated real-time digital twin.

[0092] The control strategy simulation optimization module 50 is used to perform multi-index thermal situation analysis based on the real-time digital twin, and to perform control strategy simulation optimization based on the generated multi-dimensional thermal situation analysis results, and generate equipment operation intervention instructions.

[0093] Furthermore, the device is also used to perform the following functions:

[0094] The multidimensional physical field data also includes pressure gradient data between hot and cold aisles acquired by micro differential pressure sensors deployed at the top and / or bottom of the cabinet.

[0095] Furthermore, the device is also used to perform the following functions:

[0096] Based on the physical structural parameters, the geometric shape, internal equipment layout, ventilation opening location and size of the target cabinet are determined to define the simulation solution domain of the reduced-order thermodynamics model. At the boundary of the simulation solution domain, the boundary conditions for the airflow inlet and outlet of the reduced-order thermodynamics model are defined according to the location and pattern of the air conditioning supply and return vents. Within the simulation solution domain, an original high-dimensional governing equation model is established to describe the airflow and heat transfer processes. Characteristic mode analysis is performed on the original high-dimensional governing equation model to extract multiple dominant physical modes that characterize the dominant flow and heat transfer features inside the target cabinet. Based on the extracted dominant physical modes, a low-dimensional subspace is constructed, and the original high-dimensional governing equation model is projected onto this low-dimensional subspace to obtain a reduced-order thermodynamics model that can be solved in real time at the second level.

[0097] Furthermore, the device is also used to perform the following functions:

[0098] The device-level real-time power consumption data is spatially mapped to the volumetric heat source intensity of the corresponding computational unit in the thermodynamic reduced-order model, based on the physical location of the device within the target cabinet. The volumetric heat source intensity and the boundary conditions of the airflow inlet and outlet are used as input parameters and loaded into the thermodynamic reduced-order model. The transient solver of the thermodynamic reduced-order model is activated to perform rapid numerical calculations and output the first predicted physical field data, which includes the predicted three-dimensional airflow field and the predicted temperature distribution.

[0099] Furthermore, the device is also used to perform the following functions:

[0100] The multidimensional physical field data and the first predicted physical field data are spatiotemporally registered to ensure that they are compared under the same spatial coordinates and timestamps; the residuals between the measured values ​​of each physical quantity in the multidimensional physical field data and the predicted values ​​of the corresponding physical quantities in the first predicted physical field data are calculated; an objective function is constructed with the implicit parameters of the thermohydrodynamic reduced-order model as optimization variables, and the objective function represents the weighted sum of squares of the residuals; through iterative optimization, the parameter values ​​of the implicit parameters are repeatedly adjusted until the value of the objective function is lower than a preset threshold, and the thermohydrodynamic reduced-order model at this time is used as a calibrated real-time digital twin.

[0101] Furthermore, the device is also used to perform the following functions:

[0102] The implicit parameters include the flow resistance correction coefficient, which characterizes the change in local resistance caused by the absence of cables and blind plates; the surface convection heat transfer coefficient, which characterizes the actual heat dissipation efficiency of the equipment; and the equivalent leakage area coefficient, which characterizes the leakage characteristics of the cabinet gaps.

[0103] Furthermore, the device is also used to perform the following functions:

[0104] The system utilizes the three-dimensional airflow vector field data output by the real-time digital twin to assess the health of local airflow organization and generate local airflow health indicators. It also utilizes the temperature field data output by the real-time digital twin, combined with equipment air intake requirements, to identify potential hotspots with poor airflow organization and overheating risk, thus obtaining hidden thermal risk points. Based on the airflow and temperature data output by the real-time digital twin, it calculates cooling efficiency indicators. Finally, it integrates the local airflow health indicators, hidden thermal risk points, and cooling efficiency indicators to generate a multivariate thermal situation analysis result.

[0105] Furthermore, the device is also used to perform the following functions:

[0106] In the three-dimensional airflow vector field data output by the real-time digital twin, the airflow vortex region where streamlines form a closed loop and the airflow stagnation region where the wind speed is continuously lower than the set wind speed threshold are identified; the airflow trajectory that starts from the cold aisle and enters the hot aisle directly without passing through any server heat source are tracked, and the ratio of its mass flow rate to the total air supply is calculated as the airflow short-circuit rate; for each server, the ratio of the standard deviation to the average value of the wind speed distribution on its air inlet cross-section grid is statistically analyzed as the air inlet uniformity index.

[0107] Furthermore, the device is also used to perform the following functions:

[0108] Define a set of adjustable control variables, including at least the supply air temperature setpoint of the precision air conditioner, the supply fan speed, and the opening angle of the intelligent guide vanes; in the real-time digital twin, set the control variables to different combinations to be simulated; for each combination of control variables, perform rapid simulation in the real-time digital twin to predict the thermal state under the corresponding combination and calculate the corresponding performance indicators, including total system power consumption, hottest spot temperature, and airflow short-circuit rate; with the goal of optimizing the performance indicators, perform an optimization search within the feasible domain of the control variables, filter and extract the optimal control strategy, and convert it into equipment operation intervention command output.

[0109] It should be noted that the order of the embodiments described above is for descriptive purposes only and does not represent the superiority or inferiority of the embodiments. Specific embodiments of this specification have been described above. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0110] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0111] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A method for monitoring the operating state of equipment in a machine room, characterized in that The method includes: Real-time acquisition of multi-dimensional physical field data of the target cabinet area, the multi-dimensional physical field data including at least three-dimensional airflow vector field data obtained by airflow imaging sensor grid, temperature distribution data obtained by temperature sensor array, and real-time power consumption data at the device level obtained by power monitoring device; Based on the physical structural parameters of the target cabinet, a reduced-order thermo-hydrodynamic model of the target cabinet is constructed; Using the device-level real-time power consumption data as the main heat source input, the thermohydrodynamic reduced-order model is driven to perform real-time simulation to obtain the first predicted physical field data. The multidimensional physical field data and the first predicted physical field data are assimilated to dynamically correct the implicit parameters in the thermohydrodynamic reduced-order model, so that the model output infinitely approximates the real state, forming a calibrated real-time digital twin. Based on the real-time digital twin, multi-indicator thermal situation analysis is performed, and based on the generated multi-dimensional thermal situation analysis results, control strategy simulation optimization is performed to generate equipment operation intervention commands.

2. The operating state monitoring method for the equipment room apparatus according to Claim 1, wherein The multidimensional physical field data also includes pressure gradient data between hot and cold aisles acquired by micro differential pressure sensors deployed at the top and / or bottom of the cabinet.

3. The operating state monitoring method for the equipment room apparatus according to Claim 1, wherein Based on the physical structural parameters of the target cabinet, a reduced-order thermo-hydrodynamic model of the target cabinet is constructed, including: Based on the physical structural parameters, the geometric shape, internal equipment layout, ventilation opening location and size of the target cabinet are determined in order to define the simulation solution domain of the reduced-order thermohydrodynamic model. On the boundary of the simulation solution domain, the boundary conditions of the airflow inlet and outlet of the reduced-order thermodynamic model are defined according to the position and pattern of the air supply and return air vents. Within the simulation solution domain, a primitive high-dimensional control equation model is established to describe the airflow and heat transfer process; Eigenmode analysis is performed on the original high-dimensional control equation model to extract multiple dominant physical modes that can characterize the dominant flow and heat transfer characteristics inside the target cabinet; Based on the extracted dominant physical modes, a low-dimensional subspace is constructed, and the original high-dimensional control equation model is projected onto this low-dimensional subspace to obtain a reduced-order thermohydrodynamic model that can be solved in real time at the second level.

4. The operating state monitoring method for the equipment room apparatus according to Claim 3, wherein Using the device-level real-time power consumption data as the main heat source input, the reduced-order thermohydrodynamic model is driven to perform real-time simulation to obtain the first predicted physical field data, including: The device-level real-time power consumption data is spatially mapped to the volumetric heat source intensity of the corresponding calculation unit in the reduced-order thermodynamic model, based on the physical location of the device within the target cabinet. The volumetric heat source intensity and the boundary conditions of the airflow inlet and outlet are used together as input parameters and loaded into the thermohydrodynamic reduced-order model. The transient solver of the reduced-order thermohydrodynamic model is activated to perform rapid numerical calculations and output the first predicted physical field data, which includes the predicted three-dimensional airflow field and the predicted temperature distribution.

5. The operating state monitoring method for the equipment room apparatus according to Claim 1, wherein The multidimensional physical field data and the first predicted physical field data are assimilated to dynamically correct the implicit parameters in the thermohydrodynamic reduced-order model, making the model output infinitely close to the real state, forming a calibrated real-time digital twin, including: The multidimensional physical field data is spatiotemporally registered with the first predicted physical field data to ensure that the two are compared under the same spatial coordinates and timestamps; Calculate the residual between the measured value of each physical quantity in the multidimensional physical field data and the predicted value of the corresponding physical quantity in the first predicted physical field data; Construct an objective function with the implicit parameters of the reduced-order thermohydrodynamic model as optimization variables, wherein the objective function characterizes the weighted sum of squares of the residuals; Through iterative optimization, the values ​​of the implicit parameters are repeatedly adjusted until the value of the objective function is lower than a preset threshold. The thermohydrodynamic reduced-order model at this point is then used as a calibrated real-time digital twin.

6. The operating state monitoring method for the equipment room apparatus according to Claim 5, wherein The implicit parameters include the flow resistance correction coefficient, which characterizes the change in local resistance caused by the absence of cables and blind plates; the surface convection heat transfer coefficient, which characterizes the actual heat dissipation efficiency of the equipment; and the equivalent leakage area coefficient, which characterizes the leakage characteristics of the cabinet gaps.

7. The operating state monitoring method for the equipment room apparatus according to Claim 1, wherein Based on the aforementioned real-time digital twin, a multi-indicator thermal trend analysis is performed, including: The three-dimensional airflow vector field data output by the real-time digital twin is invoked to assess the health of local airflow organization and generate local airflow health indicators. By calling the temperature field data output by the real-time digital twin and combining it with the equipment air intake requirements, potential hot spots with poor airflow organization and overheating risk are identified, and hidden thermal risk points are obtained. Based on the airflow and temperature data output by the real-time digital twin, the cooling energy efficiency index is calculated. By combining the local airflow health indicators, hidden thermal risk points, and cooling energy efficiency indicators, a multivariate thermal situation analysis result is generated.

8. The operating state monitoring method for the equipment room apparatus according to Claim 7, wherein The three-dimensional airflow vector field data output by the real-time digital twin is invoked to assess the health of local airflow organization and generate local airflow health indicators, including: In the three-dimensional airflow vector field data output by the real-time digital twin, identify the airflow vortex region where streamlines form a closed loop, and the airflow stagnation region where the wind speed is continuously lower than the set wind speed threshold. Track the airflow trajectory from the cold aisle without passing through any server heat-generating components and directly enter the hot aisle, and calculate the ratio of its mass flow rate to the total air supply as the airflow short-circuit rate; For each server, the ratio of the standard deviation to the average value of the wind speed distribution on the grid of its air inlet section is used as the air inlet uniformity index.

9. The operating state monitoring method for the equipment room apparatus according to Claim 8, wherein Based on the generated multivariate thermal situation analysis results, the control strategy is simulated and optimized to generate equipment operation intervention commands, including: Define a set of adjustable control variables, which include at least the supply air temperature setpoint of the precision air conditioner, the fan speed, and the opening and closing angle of the intelligent guide vanes; In the real-time digital twin, the control variables are set to different combinations to be simulated; For each combination of control variables, a fast simulation is performed in the real-time digital twin to predict the thermal state under the corresponding combination and calculate the corresponding performance indicators, including total system power consumption, hottest temperature, and airflow short-circuit rate. With the goal of optimizing the performance indicators, an optimization search is performed within the feasible domain of the control variables to select and extract the optimal control strategy, and then converted into an equipment operation intervention command output.

10. An operating condition monitoring device for a machine room apparatus, characterized by, The apparatus is used to implement the method for monitoring the operating status of computer room equipment according to any one of claims 1-9, and the apparatus comprises: The physical field data acquisition module is used to acquire multi-dimensional physical field data of the target cabinet area in real time. The multi-dimensional physical field data includes at least three-dimensional airflow vector field data obtained by the airflow imaging sensor grid, temperature distribution data obtained by the temperature sensor array, and real-time power consumption data at the device level obtained by the power monitoring device. The reduced-order model construction module is used to construct a reduced-order thermo-hydrodynamic model of the target cabinet based on the physical structural parameters of the target cabinet; The real-time simulation module is used to drive the thermodynamic reduced-order model to perform real-time simulation using the device-level real-time power consumption data as the main heat source input, so as to obtain the first predicted physical field data. The digital twin generation module is used to assimilate the multidimensional physical field data with the first predicted physical field data, dynamically correct the implicit parameters in the thermohydrodynamic reduced-order model, so that the model output infinitely approximates the real state, and form a calibrated real-time digital twin. The control strategy simulation optimization module is used to perform multi-index thermal situation analysis based on the real-time digital twin, and to perform control strategy simulation optimization based on the generated multi-dimensional thermal situation analysis results, and generate equipment operation intervention commands.

Images