Energy consumption management system and method of data center

Abstract

The invention discloses an energy consumption management system and method for a data center, and relates to the technical field of data center thermal management, and the method comprises the steps: carrying out the noise reduction of a structured data set through quantum harmonic oscillator coupling wavelet transform, and generating a real-time three-dimensional thermal distribution diagram through a Kriging spatial interpolation algorithm, solving a Navier-Stokes equation by combining the magneto-rheological grid wind resistance value, and extracting a heat flow vector; injecting pulse voltage into the microcapsule phase-change material layer according to the dual-channel execution instruction set to trigger phase-change heat absorption, reducing the viscosity of a target path by using a space gradient magnetic field, and generating a three-dimensional wind speed vector field; the real-time three-dimensional thermal distribution diagram is generated through the Kriging space interpolation algorithm, high-precision monitoring and analysis of the complex thermal environment of the data center are achieved, the accuracy of local hot spot recognition is improved, and therefore the energy utilization efficiency and the cooling effect are remarkably improved, and the purposes of reducing energy consumption and improving the operation stability and energy efficiency of the data center are achieved.

Description

Technical Field

[0001] This invention relates to the field of data center thermal management technology, and in particular to a data center energy consumption management system and method. Background Technology

[0002] As data centers continue to expand in scale and computing density increases, the heat generated during operation also grows significantly. This is especially true for multi-core and many-core CPUs widely deployed in high-performance servers, where high-parallel computing loads exacerbate localized heat buildup. To ensure the stable operation of these critical devices, data centers generally employ a cooling method combining centralized air conditioning with hot and cold aisle isolation. In recent years, sensor network-based data center thermal management technology has been widely applied. By deploying sensors for multiple parameters such as temperature, humidity, and airflow velocity, real-time monitoring of the internal thermal environment of the computer room is achieved, and feedback control strategies are used to optimize the operation of the air conditioning system.

[0003] Existing technologies still suffer from response lag when dealing with complex dynamic heat load changes, especially when heat source distribution exhibits localized high-density characteristics. Static or semi-dynamic control strategies struggle to identify hot spot locations and implement effective local intervention measures in a timely manner. Although some studies have attempted to enhance thermal field reconstruction capabilities by introducing spatial interpolation algorithms and CFD modeling, high-resolution three-dimensional dynamic visualization of the thermal field remains difficult to achieve due to limitations in data acquisition accuracy and timeliness. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a data center energy management method to solve the problem of delayed hot spot location identification.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an energy management method for a data center, comprising: collecting the inlet and outlet temperatures of the server rack and the surface temperature of the server; obtaining the wind resistance value of the magnetorheological grid through a differential pressure sensor built into the magnetorheological grid, and generating a structured dataset; performing noise reduction processing on the structured dataset through quantum harmonic oscillator resonance, and generating a real-time three-dimensional thermal distribution map using the Kriging spatial interpolation algorithm; obtaining the magnetic viscosity coefficient by calling the magnetic field gradient-magnetic viscosity coefficient mapping table and solving the Navier-Stokes equations by combining the wind resistance value of the magnetorheological grid, and extracting the heat flux vector; calculating the heat flux density modulus of the spatial grid points based on the heat flux vector, identifying the location of spatial hot spots and the directionality of heat conduction paths, generating a phase change material activation command, and determining the phase change layer electromagnetic response based on the phase change layer electromagnetic response. The pulse voltage waveform is dynamically calculated based on the characteristics, and the magnetic field gradient distribution is measured simultaneously. The magnetorheological control parameters are calibrated through Lorentz force field simulation, and a dual-channel execution command set is output. According to the dual-channel execution command set, a pulse voltage is injected into the microcapsule phase change material layer to trigger phase change endothermy. The spatial gradient magnetic field is used to reduce the viscosity of the target path and generate a three-dimensional wind speed vector field. Based on the three-dimensional wind speed vector field, the vorticity intensity of spatial grid points is calculated and sub-regions are formed by DBSCAN clustering. The priority sequence is obtained, the air conditioning air volume is allocated according to the priority gradient, and the temperature distribution is predicted. The valve opening command is corrected according to the prediction deviation, and the air volume adjustment is completed by using a brushless motor. A physical state change report is generated, and the air conditioning energy consumption is calculated to generate an energy management report.

[0008] As a preferred embodiment of the data center energy management method of the present invention, the specific steps for generating the structured dataset are as follows:

[0009] Based on the differential pressure sensor group of the magnetorheological grid, the dynamic differential pressure measurement values ​​before and after the grid are obtained. The dynamic differential pressure measurement values ​​are input into the magnetic field strength controller to generate the wind resistance value of the magnetorheological grid. The value is then aligned with the inlet temperature, outlet temperature and server surface temperature according to the timestamp to obtain a structured dataset.

[0010] As a preferred embodiment of the data center energy management method of the present invention, the specific steps for extracting the heat flux vector are as follows:

[0011] Electromagnetic interference is eliminated by filtering the structured dataset using quantum harmonic oscillator resonance, an optimized dataset is generated, and the coordinates of temperature measurement points are extracted to solve the Kriging weight equations to obtain the weight matrix. The temperature of spatial grid points is calculated and combined with the weight matrix to generate a three-dimensional thermal distribution map using the Kriging spatial interpolation algorithm.

[0012] The magnetic field gradient-magnetic viscosity coefficient mapping table is consulted by calling the magnetorheological grid wind resistance value to obtain the magnetic viscosity coefficient. The magnetic viscosity coefficient is then substituted into the Navier-Stokes equation to obtain the three-dimensional velocity field distribution. The temperature gradient field is calculated by combining the three-dimensional thermal distribution map and then substituted into the heat flux density calculation formula to generate the heat flux vector.

[0013] As a preferred embodiment of the data center energy management method of the present invention, the specific steps of outputting the dual-channel execution instruction set are as follows:

[0014] The heat flux density modulus of spatial grid points is calculated using heat flux vectors to obtain high heat flux points and perform DBSCAN to form a hot spot region bounding box. The geometric center coordinates of the hot spot region bounding box are calculated and used as the hot spot position coordinates. At the same time, the main heat flux direction of the hot spot region bounding box is extracted.

[0015] The pulse voltage parameters are dynamically calculated based on the heat flux density modulus, and position quantum encryption is performed to generate phase transition activation commands, and the magnetorheological viscosity parameters are calibrated.

[0016] By aligning the principal axis of the magnetic field gradient with the direction of the main heat flow in the region, a magnetic field gradient direction is generated and aligned with the direction of the temperature gradient field. A time-varying control sequence is then generated, and hot spot location coordinates, pulse voltage parameters, and phase change activation commands are integrated to generate a phase change channel parameter package.

[0017] The magnetic field gradient direction, time-varying control sequence, and calibrated magnetorheological viscosity parameters are integrated to generate a magnetorheological change channel parameter package, which is then combined with the phase transition channel parameter package to form a dual-channel execution instruction set.

[0018] As a preferred embodiment of the data center energy management method of the present invention, the specific steps for generating the three-dimensional wind speed vector field are as follows:

[0019] Based on the pulse voltage parameters, a resonant pulse voltage is generated by an FPGA waveform generator. The resonant pulse voltage is injected into the microcapsule phase change material layer using a dual-channel execution instruction set. The electromagnetic coil array is controlled by the magnetofluidic change channel parameter package to output a spatial gradient magnetic field and generate the target path fluid viscosity reduction value.

[0020] The system measures the velocity of fluid particles, calculates the viscosity compensation coefficient based on the fluid viscosity reduction value along the target path, calibrates the fluid particle velocity, and generates a three-dimensional wind speed vector field using an optical flow algorithm.

[0021] As a preferred embodiment of the data center energy management method of the present invention, the specific steps for generating the physical state change report are as follows:

[0022] Based on the three-dimensional wind speed vector field, the vorticity intensity of spatial grid points is calculated and compared to obtain high vorticity points. Adjacent high vorticity points are merged into sub-regions using DBSCAN. The average vorticity intensity of each sub-region is calculated and sorted according to the magnitude of the average vorticity intensity of each sub-region to obtain the priority sequence of air conditioning air supply volume.

[0023] The air supply parameters are obtained by weighting the air supply volume priority sequence, and the heat flow balance equation is input to predict the future temperature distribution. The coordinates are then aligned with the actual temperature distribution to obtain the temperature deviation value.

[0024] The temperature deviation value is converted into the air volume adjustment range, and the air supply parameters are updated. The valve opening command is corrected, and the corrected valve opening command is executed by a brushless motor. The change value of the fan power is calculated to obtain the energy consumption impact, and it is integrated with the valve opening change record into a physical state change report.

[0025] As a preferred embodiment of the data center energy consumption management method of the present invention, the specific steps for calculating air conditioning energy consumption and generating an energy consumption management report based on physical state change reports are as follows:

[0026] CPU utilization data is collected and the equipment load fluctuation compensation coefficient is obtained through the principle of thermal inertia. Energy efficiency factor and temperature uniformity index are extracted from the physical state change report. A benchmark energy consumption model is established in combination with environmental parameters to obtain real-time energy saving. The energy saving ratio is corrected by using the equipment load fluctuation compensation coefficient.

[0027] Based on the revised energy-saving ratio and the operational health index of computer equipment, the failure probability is mapped, predictive maintenance instructions are generated based on the failure probability, and the revised energy-saving ratio, operational health index, and predictive maintenance instructions are packaged into an energy consumption management report.

[0028] Secondly, this invention provides an energy management system for a data center, comprising: a data acquisition module for acquiring the inlet and outlet temperatures of the server racks and the surface temperature of the servers, obtaining wind resistance values ​​through a differential pressure sensor built into a magnetorheological grid, and generating a structured dataset; a noise reduction module for performing noise reduction processing on the structured dataset using quantum harmonic oscillator coupled wavelet transform, generating a real-time three-dimensional thermal distribution map using a Kriging spatial interpolation algorithm, solving the Navier-Stokes equations based on the wind resistance values ​​of the magnetorheological grid, and extracting the heat flux vector; and an instruction module for calculating the heat flux density modulus of spatial grid points based on the heat flux vector, identifying the location of spatial hot spots and the directionality of heat conduction paths, generating phase change material activation instructions, and determining the phase change layer electromagnetic response characteristics. The system dynamically calculates pulse voltage waveforms, synchronously measures magnetic field gradient distribution, calibrates magnetorheological control parameters through Lorentz force field simulation, and outputs a dual-channel execution command set. The vector field module injects pulse voltage into the microcapsule phase change material layer to trigger phase change endothermic reaction, utilizes the spatial gradient magnetic field to reduce the target path viscosity, and generates a three-dimensional wind speed vector field. The reporting module extracts the eddy current intensity of each sub-region based on the three-dimensional wind speed vector field, generates a priority sequence, allocates air conditioning air volume according to the priority gradient, predicts temperature distribution, corrects the damper opening command in real time based on prediction deviation, uses a brushless motor to complete the air volume adjustment, generates a physical state change report, and calculates air conditioning energy consumption to generate an energy management report.

[0029] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the data center energy management method as described in the first aspect of the present invention.

[0030] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the data center energy management method as described in the first aspect of the present invention.

[0031] The beneficial effects of this invention are as follows: By generating real-time three-dimensional thermal distribution maps through the Kriging space interpolation algorithm, high-precision monitoring and analysis of the complex thermal environment of multi-core or many-core CPUs are achieved. This improves the accuracy of local hotspot identification and, through the extraction of fluid dynamic parameters, supports the formulation of refined cooling strategies, thereby significantly improving energy utilization efficiency and cooling effect, achieving the goals of reducing energy consumption and improving the operational stability and energy efficiency of data centers. Attached Figure Description

[0032] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a flowchart of a data center energy management method.

[0034] Figure 2 This is a schematic diagram of a data center's energy management system.

[0035] Figure 3 A flowchart for generating an energy management report.

[0036] Figure 4 A flowchart for generating a three-dimensional wind speed vector field. Detailed Implementation

[0037] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0038] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0039] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0040] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a data center energy management method, including the following steps:

[0041] S1. Collect the air inlet temperature, air outlet temperature, and server surface temperature of the rack, obtain the wind resistance value of the magnetorheological grid through the differential pressure sensor built into the magnetorheological grid, and generate a structured dataset.

[0042] Based on the differential pressure sensor group of the magnetorheological grid, the dynamic differential pressure measurement values ​​before and after the grid are obtained. The dynamic differential pressure measurement values ​​are input into the magnetic field strength controller to generate the magnetorheological grid wind resistance value. The inlet air temperature data, outlet air temperature data, server surface temperature and magnetorheological grid wind resistance value are aligned with the timestamp to obtain a structured dataset.

[0043] Specifically, the differential pressure sensor group of the magnetorheological grid continuously collects dynamic differential pressure measurements between the grid inlet and outlet. These measurements are transmitted in real time to the input port of the magnetic field strength controller at a millisecond-level sampling frequency. The magnetic field strength controller's internal preset fluid resistance conversion algorithm converts the dynamic differential pressure measurements into corresponding magnetorheological grid wind resistance values. The conversion process strictly follows the Reynolds number-magnetic viscosity correlation equation. For example, when the input dynamic differential pressure is 15.3 Pa, the output wind resistance value is 120 Pa·s / m³. Simultaneously, inlet air temperature data, outlet air temperature data, and multi-core or many-core CPU surface temperature data are collected. These three types of temperature data and the magnetorheological grid wind resistance values ​​are time-aligned using a nanosecond-level precision timestamp synchronization protocol, with an alignment error of no more than ±10 microseconds. The aligned four types of data are integrated into a structured dataset along the time axis. Each data record contains a timestamp field, inlet air temperature value, outlet air temperature value, multi-core or many-core CPU temperature value, and magnetorheological grid wind resistance value, resulting in a structured dataset.

[0044] It should be noted that the Reynolds number-magnetoviscous correlation equation is as follows:

[0045] ;

[0046] Where R represents the wind resistance value of the magnetorheological grid. Indicates fluid density, The characteristic length scale is represented by A, which represents the effective kinematic viscosity. Indicates the magnetic viscosity coefficient. Indicates electrical conductivity. B represents the grid channel length, C represents the magnetic field strength, and D represents the Reynolds number of the magnetic field.

[0047] S2. The structured dataset is denoised by quantum harmonic oscillator resonance, and a real-time three-dimensional thermal distribution map is generated by Kriging spatial interpolation algorithm. The magnetic viscosity coefficient is obtained by calling the magnetic field gradient-magnetic viscosity coefficient mapping table and combined with the wind resistance value of magnetorheological grid to solve the Navier-Stokes equation and extract the heat flow vector.

[0048] Electromagnetic interference is eliminated by filtering the structured dataset using quantum harmonic oscillator resonance, generating an optimized dataset. The coordinates of temperature measurement points are extracted to solve the Kriging weight equations, obtaining the weight matrix. The temperature of spatial grid points is calculated and combined with the weight matrix to generate a three-dimensional thermal distribution map using the Kriging spatial interpolation algorithm.

[0049] Specifically, a Gaussian envelope electromagnetic pulse with a frequency precisely matched to the energy level difference of the qubits is generated through the resonance of a quantum harmonic oscillator. The amplitude of the Gaussian envelope electromagnetic pulse is modulated in real time by the wind resistance value in the structured dataset, and the phase is differentially locked by the temperature data, with a timestamp synchronization accuracy of ±2ns. The electromagnetic field penetrates the insulating layer of the structured data storage unit, inducing the qubits to undergo stimulated Rabi oscillation, increasing the probability amplitude of the transition from the ground state |0> to the excited state |1> by 42.6%±5%. Then, the electromagnetic interference signal is eliminated through four techniques: hardware shielding, signal processing (power frequency notch filtering + adaptive filtering + wavelet denoising), timing avoidance (precise synchronization of switching cycle), and spatial gradient compensation, while retaining the effective temperature and magnetorheological grid wind resistance values ​​and outputting an optimized dataset. Temperature measurement point coordinates are extracted from the optimized dataset, including three sets of spatial locations: coordinates of inlet air temperature sensors, coordinates of outlet air temperature sensors, and coordinates of surface temperature sensors on multi-core or many-core CPUs. A Kriging weight equation solving algorithm is used to process all temperature sensor coordinates and fit a variation function curve. The weight values ​​of spatial grid points are calculated by matrix inversion, forming a weight matrix containing the temperature contribution of each measurement point. The spatial coordinates of the grid points are then substituted into the Kriging interpolation formula and combined with the weight matrix to calculate the temperature value of each grid point. Finally, the RGB-temperature mapping algorithm is used to convert the temperature values ​​of the grid points into color pixels, generating a three-dimensional thermal distribution map.

[0050] It should be noted that the formula for calculating the temperature value at each spatial grid point is:

[0051] ;

[0052] in, This represents the temperature value at grid point a. This represents the weight of the a-th spatial grid point. This represents the actual temperature value at the a-th spatial grid point.

[0053] The magnetic field gradient-magnetic viscosity coefficient mapping table is consulted by calling the magnetorheological grid wind resistance value to obtain the magnetic viscosity coefficient. The magnetic viscosity coefficient is then substituted into the Navier-Stokes equation to obtain the three-dimensional velocity field distribution. The temperature gradient field is calculated by combining the three-dimensional thermal distribution map and then substituted into the heat flux density calculation formula to generate the heat flux vector.

[0054] Specifically, the system uses the magnetorheological grid drag value as input to retrieve a pre-defined magnetic field gradient-magnetic viscosity coefficient mapping table. Based on the magnetorheological grid drag value, the magnetic field gradient is calculated. Using the magnetorheological grid drag value (e.g., 18.7 kPa·s / m³), a pre-calibrated drag-magnetic viscosity coefficient mapping table (experimental calibration error ±1.2%) is consulted to obtain the magnetic viscosity coefficient. This magnetic viscosity coefficient is then substituted into the viscous force term of the Navier-Stokes equations, replacing the standard viscosity term, to complete the distributed iterative solution of the three-dimensional velocity field, outputting the velocity vector value for each spatial grid point. The temperature values ​​of the spatial grid points in the three-dimensional thermal distribution map are read, and the temperature gradient field is calculated using the central difference method. The temperature gradient field and the three-dimensional velocity field are then substituted into the heat flux density calculation formula to generate the heat flux vector for each spatial grid point.

[0055] It should be noted that the formula for calculating heat flux density is:

[0056] ;

[0057] in, Represents the heat flux vector. Indicates the thermal conductivity of a material. This indicates a temperature gradient.

[0058] The formula for the magnetic field gradient based on the wind resistance value of the magnetorheological grating is:

[0059] ;

[0060] in, The magnetic field gradient is represented by D, and the wind resistance value of the magnetorheological grating is represented by D. This is the basic wind resistance value at zero magnetic field, and 'a' represents the coefficient of the experimental fit. The coefficients represent the experimental fit, M represents the channel length of the magnetorheological grid, H represents the cross-sectional area of ​​the magnetorheological grid channel, and B represents the magnetic induction intensity.

[0061] S3. Based on the heat flux vector, calculate the heat flux density modulus of the spatial grid points, identify the location of spatial hot spots and the directionality of heat conduction paths, generate phase change material activation commands, dynamically calculate the pulse voltage waveform according to the electromagnetic response characteristics of the phase change layer, simultaneously measure the magnetic field gradient distribution, calibrate the magnetorheological control parameters through Lorentz force field simulation, and output a dual-channel execution command set.

[0062] The heat flux density modulus of spatial grid points is calculated using heat flux vectors to obtain high heat flux points, and DBSCAN is performed to form a hot spot region bounding box. The geometric center coordinates of the hot spot region bounding box are calculated and used as the hot spot position coordinates. At the same time, the main heat flux direction of the hot spot region bounding box is extracted.

[0063] Specifically, the heat flux density modulus of each spatial grid point is calculated using the heat flux vector formula. Spatial grid points with heat flux density modulus greater than the example value of 150 W / m² are selected as high heat flux points. DBSCAN cluster analysis is performed on all high heat flux points, with a neighborhood radius of 0.05 meters and a minimum number of 3 points. High heat flux points with a spatial distance of less than 0.05 meters are connected to form independent clusters. The maximum and minimum coordinate values ​​of each independent cluster in the XYZ axis directions are calculated, and an axially aligned minimum cube is constructed as the bounding box of the hot spot region. The arithmetic mean of the coordinates of the center points of the six faces of the hot spot region bounding box is taken as the hot spot position coordinates. The arithmetic mean of the heat flux vector components of all spatial grid points within the hot spot region bounding box is taken to obtain the main heat flux direction of the hot spot region bounding box.

[0064] It should be noted that the formula for calculating the heat flux density modulus of spatial grid points using heat flux vectors is:

[0065]

[0066] in, This represents the heat flux density modulus of a spatial grid point. This represents the heat flux density component of a spatial grid point in the z-direction. Indicates that the spatial grid points are in Heat flux density component in the direction, Indicates that the spatial grid points are in Heat flux density component in the direction.

[0067] The pulse voltage parameters are dynamically calculated based on the heat flux density modulus, and position quantum encryption is performed to generate phase transition activation commands, and the magnetorheological viscosity parameters are calibrated.

[0068] Specifically, pulse voltage parameters, voltage amplitude, and fixed frequency are dynamically calculated based on the heat flux density modulus.

[0069] The example value is initially fixed at 50 volts (base voltage). The current heat flux density modulus is multiplied by 15 and superimposed on the base voltage value to obtain the voltage amplitude. The example value is permanently set to a fixed frequency of 48.2 kHz (determined by the resonant characteristics of the microcapsule material and does not change with any variables). The hot spot location coordinates are formatted as a string, a 256-bit hash digest is generated using the SHA-3 algorithm, and the quantum key is bound to the IDQ Clavis3 quantum key distribution device. The Lorentz volume force formula is called. The current field reconstruction uses the magnetorheological viscosity parameters corresponding to the current field data. The velocity field uses three-dimensional velocity field distribution data. After integration, the calibrated magnetorheological viscosity parameters are output.

[0070] By aligning the principal axis of the magnetic field gradient with the direction of the main heat flow in the region, a magnetic field gradient direction is generated. This direction is then aligned with the direction of the temperature gradient field to generate a time-varying control sequence. The hot spot location coordinates, pulse voltage parameters, and phase change activation commands are integrated to generate a phase change channel parameter package.

[0071] Specifically, the principal axis direction of the magnetic field gradient is determined through eigenvalue decomposition of the three-dimensional gradient tensor matrix. The eigenvectors of the magnetic field gradient tensor matrix are calculated, and the eigenvector corresponding to the largest eigenvalue is taken as the principal axis direction of the magnetic field gradient. This principal axis direction is vector-aligned with the main heat flow direction of the region and kept in the same direction as the temperature gradient field of the three-dimensional thermal distribution map. The temperature gradient value of the target region is obtained through the temperature gradient field extraction module. A mapping relationship between the temperature gradient value and the baseline magnetic field strength is established. A three-segment intensity configuration of the time-varying control sequence is constructed: for example, the time interval 0.0-0.5 seconds corresponds to a magnetic field strength value of 0.5×B_base, the time interval 0.5-2.0 seconds corresponds to B_base, and the time interval 2.0-3.0 seconds corresponds to 0. 7×B_base; Spatial compensation is performed by superimposing a hotspot position attenuation factor, and a time-varying control sequence is output; The target magnetic field strength value is calculated based on the temperature gradient value, and the time-varying control sequence is configured. Example values: 0.0-0.5 seconds correspond to a magnetic field strength of 4.92 Tesla / m (0.5×9.84), 0.5-2.0 seconds correspond to 9.84 Tesla / m, and 2.0-3.0 seconds correspond to 6.89 Tesla / m (0.7×9.84); The hotspot position coordinates, pulse voltage parameters (voltage amplitude 320.0 volts, frequency 48200 Hz, duty cycle 0.28), and phase transition activation command (e.g., 0x5a3f9e...c2) are integrated into a JSON structure to generate the phase transition channel parameter package.

[0072] The magnetic field gradient direction, time-varying control sequence, and calibrated magnetorheological viscosity parameters are integrated to generate a magnetorheological change channel parameter package, which is then combined with the phase transition channel parameter package to form a dual-channel execution instruction set.

[0073] Specifically, the magnetic field gradient direction vector, time-varying control sequence parameters, and magnetorheological viscosity parameters calibrated by Lorentz volume force are integrated to construct a JSON structure of the magnetorheological channel parameter package, which includes gradient direction, time-varying sequence array, and viscosity value fields; the hot spot position coordinates, pulse voltage parameters, and phase transition activation instructions in the phase transition channel parameter package are loaded synchronously; the two parameter packages are assembled according to field classification to form a dual-channel execution instruction set.

[0074] S4. According to the dual-channel execution instruction set, pulse voltage is injected into the microcapsule phase change material layer to trigger phase change endothermy. The spatial gradient magnetic field is used to reduce the viscosity of the target path and generate a three-dimensional wind speed vector field.

[0075] Based on the pulse voltage parameters, a resonant pulse voltage is generated by an FPGA waveform generator. The resonant pulse voltage is injected into the microcapsule phase change material layer using a dual-channel execution instruction set. The electromagnetic coil array is controlled by the magnetofluidic change channel parameter package to output a spatial gradient magnetic field and generate a target path fluid viscosity reduction value.

[0076] Specifically, pulse voltage parameters (example values: voltage amplitude 320.0V, frequency 48.2kHz, duty cycle 0.28) are extracted from the phase_change_params field of the dual-channel execution instruction set. A direct digital frequency synthesis algorithm is then executed using an FPGA waveform generator to generate a resonant pulse voltage waveform. This resonant pulse voltage waveform is injected into the microcapsule phase change material layer corresponding to the hot spot location coordinates through an IGBT power amplifier. Simultaneously, the content of the magneto_rheological_params field is analyzed: the magnetic field gradient direction is decomposed into X... The Y / Z three-axis components are used as an example to control the electromagnetic coil array, with the X-axis coil group loading a proportional current of 0.82, the Y-axis coil group loading a proportional current of -0.42, and the Z-axis coil group loading a proportional current of 0.37. The time-varying control sequence outputs an intensity of 4.92 T / m in the 0-0.5 second period and switches to an intensity of 9.84 T / m in the 0.5-2.0 second period. The fluid viscosity is continuously monitored by a magnetorheological grid integrated differential pressure sensor. In the example, the initial viscosity is 12.7 cP, which decreases to 3.2 cP in 0.5 seconds and stabilizes at 0.9 cP after 2.0 seconds, thus obtaining the fluid viscosity reduction value along the target path.

[0077] The system measures the velocity of fluid particles, calculates the viscosity compensation coefficient based on the fluid viscosity reduction value along the target path, calibrates the fluid particle velocity, and generates a three-dimensional wind speed vector field using an optical flow algorithm.

[0078] Specifically, based on a 128-channel ultrasonic sensor array, an example 200kHz frequency sound wave is emitted and the scattered signal is received. The velocity of fluid particles is calculated according to the Doppler frequency shift formula. Combining the target path fluid viscosity reduction value, for example, the initial viscosity η_pre=12.7cP and the current viscosity η_post=0.9cP, the initial viscosity and the current viscosity are compared to obtain the viscosity compensation coefficient. The fluid particle velocity is amplified according to the viscosity compensation coefficient to obtain the calibrated fluid particle velocity. The calibrated fluid particle velocity is input into the optical flow algorithm calculation process: between two sets of fluid particle velocities in adjacent time frames, the velocity gradient equation is solved using the Farneback polynomial expansion method. A continuous velocity field is generated at 0.01m three-dimensional spatial grid points through cubic spline interpolation, and a three-dimensional wind speed vector field is output.

[0079] It should be noted that the formula for measuring the velocity of fluid particles using the Doppler effect is:

[0080] ;

[0081] in, Indicates the velocity of fluid particles. This represents the difference between the scattered signal and the transmission frequency. Indicates the wavelength of the wave in the medium. This indicates the original frequency of the transmitted wave.

[0082] S5. Based on the three-dimensional wind speed vector field, extract the vorticity intensity of each sub-region and generate a priority sequence. Allocate the air supply volume of the air conditioner according to the priority gradient and predict the temperature distribution. Correct the valve opening command in real time according to the prediction deviation. Use a brushless motor to complete the air volume adjustment and generate a physical state change report.

[0083] Based on the three-dimensional wind speed vector field, the vorticity intensity of spatial grid points is calculated and compared to obtain high vorticity points. Adjacent high vorticity points are merged into sub-regions using DBSCAN. The average vorticity intensity of each sub-region is statistically analyzed and sorted according to the magnitude of the average vorticity intensity of each sub-region to obtain the priority sequence of air conditioning air supply volume.

[0084] Specifically, based on the velocity vectors of spatial grid points in the three-dimensional wind speed vector field, vorticity intensity is obtained, and the selected spatial grid points are marked as high vorticity points. DBSCAN clustering analysis is performed on the high vorticity points, using the connection conditions of a neighborhood example value radius of 0.05 meters and a minimum number of points of 3, merging high vorticity points with a distance of less than 0.05 meters to form independent clusters. The arithmetic mean of the vorticity intensity of all high vorticity points in each cluster is obtained by the arithmetic mean method as the average vorticity intensity of the sub-region (example: the average vorticity intensity of the cluster containing the coordinate point (1.20, 0.85, 0.30) is 38.7 s⁻¹). All sub-regions are arranged in descending order of average vorticity intensity to form a priority sequence for air conditioning supply air volume.

[0085] The air supply parameters are obtained by weighting the air supply volume according to the priority sequence of the air conditioner, and then the heat flow balance equation is input to predict the future temperature distribution. The coordinates are aligned with the actual temperature distribution to obtain the temperature deviation value.

[0086] Specifically, air supply volume weights are assigned according to the priority sequence of air conditioning supply volume. Using a baseline air volume example of 1200 m³ / h as a benchmark, the air supply volume for each area is calculated. The supply air parameters (including area supply air volume, outlet temperature, and wind speed) are input into the heat flow balance equation. The finite element method is used to solve for the temperature distribution over a future 30-second interval of the example value, with a time step of 0.5 seconds. The example output coordinates (1.20, 0.85, 0.30) m predict a temperature of 65.2℃. Simultaneously, actual temperature distribution data from a millisecond-level infrared thermal imager is acquired. The actual temperature distribution data is spatially aligned with the predicted temperature field using a grid coordinate transformation matrix to obtain the temperature deviation value.

[0087] The temperature deviation value is converted into the air volume adjustment range, and the air supply parameters are updated. The valve opening command is corrected, and the corrected valve opening command is executed by a brushless motor. The change value of the fan power is calculated to obtain the energy consumption impact, and it is integrated with the valve opening change record into a physical state change report.

[0088] Specifically, the system reads the current valve opening command from the register of the central control device via the OPC UA interface, calculates the air volume adjustment range based on the temperature deviation value, updates the regional air supply parameters using the air volume adjustment range, recalculates the valve opening based on the air volume-opening square law relationship, and sends the corrected valve opening command to the ABB M3BP brushless motor actuator to drive the valve to rotate to the target valve opening within 250 milliseconds (example: opening from 89.4% to 100.0%). The system monitors the changes in motor current, voltage, and fan power in real time, reads the valve opening status before adjustment from the DCS control device, obtains the valve opening change record, and compiles the valve opening change record and fan power change value into a physical state change report.

[0089] S6. Calculate the energy consumption management report for air conditioning based on the physical state change report.

[0090] CPU utilization data is collected and the equipment load fluctuation compensation coefficient is obtained through the principle of thermal inertia. Energy efficiency factor and temperature uniformity index are extracted from the physical state change report. A benchmark energy consumption model is established in combination with environmental parameters to obtain real-time energy saving. The energy saving ratio is then corrected using the equipment load fluctuation compensation coefficient.

[0091] Specifically, CPU utilization data from the BMC interface of multi-core or many-core CPUs is collected, especially for computing nodes integrating multi-core or many-core CPUs. The mean and standard deviation of CPU utilization are calculated (e.g., mean utilization = 62%, standard deviation = 18%). The equipment load fluctuation compensation coefficient is calculated based on the thermal inertia principle (e.g., mean utilization / standard deviation = 0.29, equipment load fluctuation compensation coefficient = 0.0203). Based on the physical state change report, structured data fields are directly read through key-based access to extract energy efficiency factors and temperature uniformity indicators. A baseline energy consumption model is established by combining ambient temperature and relative humidity parameters. The baseline energy consumption model eliminates the interference of CPU utilization fluctuations on energy saving analysis. The measured power curve is integrated over time using the Simpson integral method to obtain real-time energy savings. The energy saving ratio is corrected by applying the equipment load fluctuation compensation coefficient.

[0092] It should be noted that, to train the baseline energy consumption model, historical data such as cooling area A, measured energy efficiency factor, ambient temperature, and actual power recorded by the data center SCADA system were collected. Periods with CPU utilization fluctuations exceeding 30% were filtered out, and NTP timestamp alignment was performed. A multiple linear regression equation was constructed, and the regression coefficients were solved using the least squares method. An L2 regularization term was added to prevent overfitting. The data was divided into an 80% training set and a 20% test set. The root mean square error (RMSE) was calculated to meet the requirement of RMSE / P_avg < 3% (P_avg = 1.2MW). The validated baseline energy consumption model was then output.

[0093] Based on the revised energy-saving ratio and the operational health index of computer equipment, the failure probability is mapped, predictive maintenance instructions are generated based on the failure probability, and the revised energy-saving ratio, operational health index, and predictive maintenance instructions are packaged into an energy consumption management report.

[0094] Specifically, by collecting CPU utilization data of multi-core or many-core central processing units (example mean 62% ± standard deviation 18%), the equipment load fluctuation compensation coefficient (0.0203) is calculated. Based on the physical status report, energy efficiency factors and temperature uniformity indicators are extracted to construct an operating health index. The corrected energy saving ratio (e.g., 31.3%) and operating health index (e.g., 0.82) are collected. The operating health index is substituted into the S-shaped function to obtain the failure probability (example operating health index 0.82 outputs failure probability 0.15). In the example, when the failure probability is ≥0.15, predictive maintenance instructions are generated. The corrected energy saving ratio, operating health index, and predictive maintenance instructions are integrated and packaged into a structured energy consumption management report.

[0095] This embodiment also provides an energy management system for a data center, including: a data acquisition module for acquiring the inlet and outlet temperatures of the server racks and the surface temperature of the servers, obtaining wind resistance values ​​through a differential pressure sensor built into the magnetorheological grid, and generating a structured dataset; a noise reduction module for performing noise reduction processing on the structured dataset through quantum harmonic oscillator coupled wavelet transform, generating a real-time three-dimensional thermal distribution map using the Kriging spatial interpolation algorithm, solving the Navier-Stokes equations by combining the wind resistance values ​​of the magnetorheological grid, and extracting the heat flux vector; and an instruction module for calculating the heat flux density modulus of the spatial grid points based on the heat flux vector, identifying the location of spatial hot spots and the directionality of heat conduction paths, generating phase change material activation instructions, and dynamically adjusting the phase change layer based on its electromagnetic response characteristics. The system calculates the pulse voltage waveform, simultaneously measures the magnetic field gradient distribution, calibrates the magnetorheological control parameters through Lorentz force field simulation, and outputs a dual-channel execution command set. The vector field module injects pulse voltage into the microcapsule phase change material layer to trigger phase change endothermic reaction, utilizes the spatial gradient magnetic field to reduce the target path viscosity, and generates a three-dimensional wind speed vector field. The report module extracts the eddy current intensity of each sub-region based on the three-dimensional wind speed vector field, generates a priority sequence, allocates air conditioning air volume according to the priority gradient, predicts temperature distribution, corrects the damper opening command in real time based on the prediction deviation, uses a brushless motor to complete the air volume adjustment, generates a physical state change report, and calculates air conditioning energy consumption to generate an energy management report.

[0096] This embodiment also provides a computer device applicable to the energy management method for data centers, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the energy management method for data centers as proposed in the above embodiment.

[0097] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0098] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the data center energy management method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0099] In summary, this invention utilizes a Kriging space interpolation algorithm to generate real-time three-dimensional thermal distribution maps, enabling high-precision monitoring and analysis of the complex thermal environment of data centers. It improves the accuracy of local hotspot identification and, through the extraction of fluid dynamic parameters, supports the formulation of refined cooling strategies, thereby significantly improving energy utilization efficiency and cooling effect, ultimately achieving the goals of reducing energy consumption and enhancing the operational stability and energy efficiency of data centers.

[0100] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A data center energy consumption management method, characterized in that: include, Collect the air inlet temperature, air outlet temperature, and server surface temperature of the rack, obtain the wind resistance value of the magnetorheological grid through the differential pressure sensor built into the magnetorheological grid, and generate a structured dataset. The structured dataset is denoised by quantum harmonic oscillator resonance, and a real-time three-dimensional thermal distribution map is generated by Kriging spatial interpolation algorithm. The magnetic viscosity coefficient is obtained by calling the magnetic field gradient-magnetic viscosity coefficient mapping table and combined with the wind resistance value of magnetorheological grid to solve the Navier-Stokes equation and extract the heat flow vector. Based on the heat flux vector, the heat flux density modulus of spatial grid points is calculated, the location of spatial hot spots and the directionality of heat conduction paths are identified, phase change material activation commands are generated, pulse voltage waveforms are dynamically calculated according to the electromagnetic response characteristics of the phase change layer, magnetic field gradient distribution is measured simultaneously, magnetorheological control parameters are calibrated through Lorentz force field simulation, and a dual-channel execution command set is output. According to the dual-channel execution instruction set, pulse voltage is injected into the microcapsule phase change material layer to trigger phase change endothermy. Spatial gradient magnetic field is used to reduce the viscosity of the target path and generate a three-dimensional wind speed vector field. Based on the three-dimensional wind speed vector field, the vorticity intensity of spatial grid points is calculated and sub-regions are formed by DBSCAN clustering. Priority sequences are obtained, air conditioning air volume is allocated according to priority gradient, and temperature distribution is predicted. The valve opening command is corrected according to the prediction deviation, and the air volume adjustment is completed by using a brushless motor. A physical state change report is generated, and the air conditioning energy consumption is calculated to generate an energy consumption management report.

2. The data center energy management method as described in claim 1, characterized in that: The specific steps for generating the structured dataset are as follows. Based on the differential pressure sensor group of the magnetorheological grid, the dynamic differential pressure measurement values ​​before and after the grid are obtained. The dynamic differential pressure measurement values ​​are input into the magnetic field strength controller to generate the wind resistance value of the magnetorheological grid. The value is then aligned with the inlet temperature, outlet temperature and server surface temperature according to the timestamp to obtain a structured dataset.

3. The data center energy management method as described in claim 1, characterized in that: The specific steps for extracting the heat flux vector are as follows: Electromagnetic interference is eliminated by filtering the structured dataset using quantum harmonic oscillator resonance, an optimized dataset is generated, and the coordinates of temperature measurement points are extracted to solve the Kriging weight equations to obtain the weight matrix. The temperature of spatial grid points is calculated and combined with the weight matrix to generate a three-dimensional thermal distribution map using the Kriging spatial interpolation algorithm. The magnetic field gradient-magnetic viscosity coefficient mapping table is consulted by calling the magnetorheological grid wind resistance value to obtain the magnetic viscosity coefficient. The magnetic viscosity coefficient is then substituted into the Navier-Stokes equation to obtain the three-dimensional velocity field distribution. The temperature gradient field is calculated by combining the three-dimensional thermal distribution map and then substituted into the heat flux density calculation formula to generate the heat flux vector.

4. The data center energy management method as described in claim 1, characterized in that: The specific steps for executing the dual-channel instruction set are as follows. The heat flux density modulus of spatial grid points is calculated using heat flux vectors to obtain high heat flux points and perform DBSCAN to form a hot spot region bounding box. The geometric center coordinates of the hot spot region bounding box are calculated and used as the hot spot position coordinates. At the same time, the main heat flux direction of the hot spot region bounding box is extracted. The pulse voltage parameters are dynamically calculated based on the heat flux density modulus, and position quantum encryption is performed to generate phase transition activation commands, and the magnetorheological viscosity parameters are calibrated. By aligning the principal axis of the magnetic field gradient with the direction of the main heat flow in the region, a magnetic field gradient direction is generated and aligned with the direction of the temperature gradient field. A time-varying control sequence is then generated, and hot spot location coordinates, pulse voltage parameters, and phase change activation commands are integrated to generate a phase change channel parameter package. The magnetic field gradient direction, time-varying control sequence, and calibrated magnetorheological viscosity parameters are integrated to generate a magnetorheological change channel parameter package, which is then combined with the phase transition channel parameter package to form a dual-channel execution instruction set.

5. The data center energy management method as described in claim 1, characterized in that: The specific steps for generating the three-dimensional wind speed vector field are as follows: Based on the pulse voltage parameters, a resonant pulse voltage is generated by an FPGA waveform generator. The resonant pulse voltage is injected into the microcapsule phase change material layer using a dual-channel execution instruction set. The electromagnetic coil array is controlled by the magnetofluidic change channel parameter package to output a spatial gradient magnetic field and generate the target path fluid viscosity reduction value. The system measures the velocity of fluid particles, calculates the viscosity compensation coefficient based on the fluid viscosity reduction value along the target path, calibrates the fluid particle velocity, and generates a three-dimensional wind speed vector field using an optical flow algorithm.

6. The data center energy management method as described in claim 1, characterized in that: The specific steps for generating the physical state change report are as follows. Based on the three-dimensional wind speed vector field, the vorticity intensity of spatial grid points is calculated and compared to obtain high vorticity points. Adjacent high vorticity points are merged into sub-regions using DBSCAN. The average vorticity intensity of each sub-region is calculated and sorted according to the magnitude of the average vorticity intensity of each sub-region to obtain the priority sequence of air conditioning air supply volume. The air supply parameters are obtained by weighting the air supply volume priority sequence, and the heat flow balance equation is input to predict the future temperature distribution. The coordinates are then aligned with the actual temperature distribution to obtain the temperature deviation value. The temperature deviation value is converted into the air volume adjustment range, and the air supply parameters are updated. The valve opening command is corrected, and the corrected valve opening command is executed by a brushless motor. The change value of the fan power is calculated to obtain the energy consumption impact, and it is integrated with the valve opening change record into a physical state change report.

7. The data center energy management method as described in claim 6, characterized in that: The specific steps for calculating the air conditioning energy generation and energy consumption management report are as follows. CPU utilization data is collected and the equipment load fluctuation compensation coefficient is obtained through the principle of thermal inertia. Energy efficiency factor and temperature uniformity index are extracted from the physical state change report. A benchmark energy consumption model is established in combination with environmental parameters to obtain real-time energy saving. The energy saving ratio is corrected by using the equipment load fluctuation compensation coefficient. Based on the revised energy-saving ratio and the operational health index of computer equipment, the failure probability is mapped, predictive maintenance instructions are generated based on the failure probability, and the revised energy-saving ratio, operational health index, and predictive maintenance instructions are packaged into an energy consumption management report.

8. A data center energy management system, based on the data center energy management method according to any one of claims 1 to 7, characterized in that: include, The data acquisition module is used to collect the temperature of the air inlet and outlet of the cabinet and the surface temperature of the server. It obtains the wind resistance value through the differential pressure sensor built into the magnetorheological grid and generates a structured dataset. The noise reduction module is used to perform noise reduction on the structured dataset by coupling wavelet transform with quantum harmonic oscillators, and to generate a real-time three-dimensional thermal distribution map by using the Kriging spatial interpolation algorithm. It calls the magnetic field gradient-magnetic viscosity coefficient mapping table to obtain the magnetic viscosity coefficient and combines it with the wind resistance value of the magnetorheological grid to solve the Navier-Stokes equations and extract the heat flow vector. The instruction module is used to calculate the heat flux density modulus of spatial grid points based on the heat flux vector, identify the location of spatial hot spots and the directionality of heat conduction paths, generate phase change material activation instructions, dynamically calculate pulse voltage waveforms based on the electromagnetic response characteristics of the phase change layer, simultaneously measure the magnetic field gradient distribution, calibrate magnetorheological control parameters through Lorentz force field simulation, and output a dual-channel execution instruction set. The vector field module is used to inject pulse voltage into the microcapsule phase change material layer according to the dual-channel execution instruction set to trigger phase change endothermy, and use the spatial gradient magnetic field to reduce the viscosity of the target path to generate a three-dimensional wind speed vector field. The reporting module is used to calculate the vorticity intensity of spatial grid points based on the three-dimensional wind speed vector field, form sub-regions through DBSCAN clustering, obtain priority sequences, allocate air conditioning air volume according to priority gradient, predict temperature distribution, correct the damper opening command according to prediction deviation, use a brushless motor to complete the air volume adjustment, generate a physical state change report, and calculate air conditioning energy consumption to generate an energy management report.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the data center energy management method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the data center energy management method according to any one of claims 1 to 7.

Images