Low-power-consumption data center refrigeration regulation and control method and system of gas-liquid heat exchange equipment

Abstract

The invention relates to a low-power-consumption data center refrigeration regulation and control method and system for gas-liquid heat exchange equipment, and belongs to the technical field of data center temperature control. The method comprises the following steps that a transient heat flux density distribution diagram of the surface of a data center is generated; reconstructing a three-dimensional temperature field and a three-dimensional airflow velocity field in the cabinet by using a computational fluid dynamics reduced-order model according to the obtained transient heat flux density distribution diagram; an abnormal thermal environment in the cabinet is self-adaptively identified to generate a dynamic jet flow control strategy; driving a micro-spraying pipe mounted at the back of the cabinet, spraying micro-jet of dielectric cooling liquid to the abnormal thermal environment by using the micro-spraying pipe, and impacting and damaging a boundary layer of the abnormal thermal environment by using the micro-jet; the rotating speed and the inclination angle of an exhaust fan at the top of the cabinet are adaptively adjusted to form a negative pressure suction flow field in which the internal area of the abnormal thermal environment is coupled with the inclination angle of the exhaust fan, so that the fluid conveying power consumption is reduced; the data center has the advantages that the power consumption of the data center is reduced by adjusting the rotating speed and the inclination angle of the exhaust fan.

Description

Technical Field

[0001] This invention belongs to the field of data center temperature control technology, and specifically relates to a low-power data center cooling control method and system for a gas-liquid heat exchanger. Background Technology

[0002] With the rapid development of the digital economy, data centers, as the core of computing infrastructure, are experiencing continuous growth in scale and computing density, leading to increasingly prominent energy consumption issues. Cooling systems, as one of the core energy-consuming units in data centers, account for half of the total electricity consumption. Traditional data center cooling methods mostly employ compressor cooling, which is not only energy-intensive and generates large carbon emissions, but also faces bottlenecks in heat exchange efficiency when dealing with high-power heat dissipation demands.

[0003] Currently, some data centers have begun to use gas-liquid heat exchange equipment for cooling. However, existing control methods have shortcomings. Due to the rigid control strategies, constant operating parameters are often used, which cannot be adapted in real time according to the dynamic heat load of the data center. This results in serious energy waste under light load conditions. Furthermore, the selection of working fluid and flow matching in the gas-liquid heat exchange process lacks accurate calculation basis, which can easily lead to insufficient heat exchange or excessive energy consumption in the gas-liquid working fluid circulation. There is also a lack of efficient utilization mechanism for environmental cold sources, and the cooling mode has not been fully upgraded in combination with changes in outdoor ambient temperature, thus limiting the cooling efficiency.

[0004] Existing technologies employ circulating cooling systems that include energy-consuming components such as compressors, which increase the energy supply burden on data centers and are also bulky and heavy. While energy-free capillary two-phase flow (CPL) technology does not require active energy consumption, its numerous components and large size result in poor adaptability. Furthermore, existing control methods do not fully utilize the highly efficient heat exchange characteristics of gas-liquid phase change and have not established a dynamic matching relationship between heat load, environmental parameters, and heat exchange equipment operating parameters, leading to a low overall energy efficiency ratio for the cooling system and failing to meet the development requirements of low-carbon and low-power data centers. Summary of the Invention

[0005] This invention provides a low-power data center cooling control method and system for a gas-liquid heat exchanger, which solves the technical problem of high power consumption in existing data centers. By adjusting the speed and tilt angle of the exhaust fan, a negative pressure suction flow field is formed. The negative pressure suction flow field has different fluid flow rates and fluid pressures, and the exhaust fan exhibits different power, thereby reducing the power consumption of the data center.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0007] A low-power data center cooling control method for a gas-liquid heat exchanger includes the following steps:

[0008] Step S1: Use the image processing module to generate a transient heat flux density distribution map of the data center surface inside the rack;

[0009] Step S2: The data center obtains the transient heat flux density distribution map in the image processing module, and then uses the computational fluid dynamics reduced-order model to reconstruct the three-dimensional temperature field and three-dimensional airflow velocity field inside the rack.

[0010] Step S3: The data center adaptively identifies the abnormal thermal environment inside the rack and adaptively generates a dynamic jet control strategy.

[0011] The dynamic jet control strategy involves controlling the injection position and angle of each micro-nozzle, controlling the injection flow rate, injection speed and injection timing of the micro-jet, and controlling the opening or closing cycle of the micro-jet.

[0012] Step S4: Drive the micro-nozzle installed on the back of the rack based on the dynamic jet control strategy in the data center, and use the micro-nozzle to spray the dielectric coolant micro-jet into the abnormal thermal environment, and use the micro-jet to impact and destroy the boundary layer of the abnormal thermal environment.

[0013] Step S5: The data center adaptively adjusts the speed and tilt angle of the exhaust fan at the top of the rack to create a negative pressure suction flow field that couples the internal area of ​​the abnormal thermal environment with the tilt angle of the exhaust fan, thereby reducing the power consumption of fluid transport.

[0014] Optionally, in step S1, the Cattaneo-Vernotte model of thermal inertia is introduced. By using the thermal relaxation time introduced by the Cattaneo-Vernotte model, heat conduction is described as a wave rather than a simple diffusion, which can predict the thermal wave phenomenon.

[0015] Optionally, in step S2, the three-dimensional temperature field inside the reconstructed cabinet is described by a three-dimensional steady-state convection heat transfer control equation to describe the change of the three-dimensional temperature field of the air inside the cabinet with space and time.

[0016] Optionally, in step S2, for reconstructing the three-dimensional airflow velocity field inside the cabinet, the velocity components in the x, y, and z directions are treated as an overall vector and their order is reduced to ensure physical field coupling.

[0017] Optionally, in step S3, for the adaptive identification of abnormal thermal environment inside the cabinet, a multi-dimensional anomaly fusion method is adopted to calculate the multi-modal anomaly fusion index in the cabinet thermal environment anomaly, and the multi-source data of temperature, airflow and vibration are fused using a Bayesian framework to give the probability of anomaly.

[0018] Optionally, in step S4, microjets of dielectric coolant are sprayed into the abnormal thermal environment. The enhanced heat transfer effect of the microjets on the abnormal thermal environment is quantified using the Nusselt number correlation. The impact heat transfer of the microjets is calculated to achieve single-strand circular microjets.

[0019] Optionally, in step S4, a critical heat flux enhancement method is used to prevent the boundary layer from being damaged or burned out by the microjets impacting and destroying the abnormal thermal environment.

[0020] Optionally, in step S5, the adjustment of the speed and tilt angle of the exhaust fan at the top of the cabinet is calculated using the fan pressure-flow characteristic equation to describe the pressure-flow characteristic curve of the centrifugal or axial fan under variable speed and fan tilt angle conditions. By changing the flow field structure inside the cabinet through micro-jet, the fluid resistance is effectively reduced, which is equivalent to reducing the effective value of the pressure loss proportional coefficient caused by the fan tilt angle and the square of the flow rate, allowing the fan to operate at a lower speed and reducing fan power consumption.

[0021] A low-power data center cooling control system for a gas-liquid heat exchanger includes:

[0022] Image processing module for generating high-resolution transient heat flux density distribution maps;

[0023] The data center, connected to the image processing module, is used to obtain the transient heat flux density distribution map in the image processing module. Based on the transient heat flux density distribution map, the data center adaptively identifies the abnormal thermal environment inside the rack and adaptively generates a dynamic jet control strategy.

[0024] Micro-nozzles, connected to the data center, can spray micro-jets of dielectric coolant into abnormal thermal environments under the control of the data center.

[0025] The exhaust fan is connected to the data center, and the data center adjusts the speed and tilt angle of the exhaust fan.

[0026] The beneficial effects of this invention are:

[0027] 1. Under variable speed and tilt angle conditions, the present invention fits the pressure and flow characteristics of a real fan under different operating conditions based on the change of the total pressure it can provide with the fluid flow rate. By adjusting the variable speed and tilt angle conditions, a negative pressure suction flow field is formed. The negative pressure suction flow field exhibits different fluid flow rates and fluid pressures, which reflects different power. That is, the exhaust fan exhibits different power, thereby reducing the power consumption of the data center.

[0028] 2. After being subjected to microjets, the boundary region of the abnormal thermal environment of the present invention gradually decreases, while vortex-like gas is generated inside the abnormal thermal environment, and microjets of dielectric coolant continue to be sprayed. The microjets continue to destroy the boundary layer of the abnormal thermal environment until the abnormal thermal environment is gradually eliminated. Attached Figure Description

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

[0030] Figure 1 This is a schematic diagram of the system of the present invention;

[0031] Figure 2 This is a schematic diagram of the workflow of the present invention;

[0032] Figure 3 This is a transient heat flux density distribution diagram of the data center surface according to the present invention;

[0033] Figure 4 This is a schematic diagram of the three-dimensional temperature field inside the reconstructed cabinet of the present invention;

[0034] Figure 5 This is a schematic diagram of the three-dimensional velocity field inside the reconstructed cabinet according to the present invention;

[0035] Figure 6 This is a schematic diagram of the abnormal thermal environment inside the cabinet of the present invention;

[0036] Figure 7 This is a schematic diagram showing the generation of eddies in an abnormal thermal environment by a microjet spray of the dielectric coolant of the present invention into the abnormal thermal environment.

[0037] Figure 8 This is a schematic diagram illustrating the effect of the negative pressure suction flow field generated under the conditions of variable speed and variable tilt angle on power consumption according to the present invention. Detailed Implementation

[0038] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0039] Example 1;

[0040] like Figure 1 As shown, this embodiment provides a low-power data center cooling control system for a gas-liquid heat exchanger, comprising:

[0041] The system includes a temperature sensor, an infrared thermal imaging device, and an image processing module. The temperature sensor and the infrared thermal imaging device are connected to the image processing module. The temperature sensor collects the transient thermal temperature of the server chip surface inside the rack in real time. The infrared thermal imaging device obtains the transient thermal temperature signal from the temperature sensor. After filtering and noise reduction, the infrared thermal imaging device uses the image processing module to generate a high-resolution transient heat flux density distribution map.

[0042] The data center is connected to the image processing module to obtain the transient heat flux density distribution map in the image processing module. Based on the transient heat flux density distribution map, the data center adaptively identifies the abnormal thermal environment inside the rack and adaptively generates a dynamic jet control strategy.

[0043] The micro-nozzle is connected to the data center, and under the control of the data center, it can spray micro-jets of dielectric coolant into abnormal thermal environments.

[0044] The exhaust fans are connected to the data center, and the data center adaptively adjusts the speed and tilt angle of the exhaust fans on top of the racks.

[0045] Example 2;

[0046] Based on Example 1, such as Figure 2 As shown, this embodiment provides a low-power data center cooling control method for a gas-liquid heat exchanger, including the following steps:

[0047] Step S1: Use the image processing module to generate a transient heat flux density distribution map of the data center or server chip surface inside the rack;

[0048] The cabinet contains a temperature sensor, an infrared thermal imaging device, and an image processing module. The temperature sensor collects the transient thermal temperature of the server chip surface inside the cabinet in real time. The infrared thermal imaging device obtains the transient thermal temperature signal from the temperature sensor. After filtering and noise reduction, the infrared thermal imaging device uses the image processing module to generate a high-resolution transient heat flux density distribution map.

[0049] Step S2: The data center obtains the transient heat flux density distribution map in the image processing module, and then uses the computational fluid dynamics (CFD) reduced-order model (ROM) to reconstruct the three-dimensional temperature field and three-dimensional airflow velocity field inside the rack.

[0050] The data center uses transient heat flux density distribution maps as heat source data and calls a pre-trained computational fluid dynamics reduced-order model to reconstruct the three-dimensional temperature field and three-dimensional airflow velocity field of the entire rack in real time, realizing real-time visualization and quantitative characterization of the thermal environment inside the rack.

[0051] Computational fluid dynamics (CFD) reduced-order models are mathematical models that reduce the computational cost of high-fidelity numerical simulations while maintaining physical characteristics and predictive accuracy. Through data-driven or mathematical projection methods, ultra-high-dimensional (typically millions of degrees of freedom) full-order models of complex flow systems are approximated as simplified models with significantly reduced dimensions (usually tens to hundreds). This enables applications such as rapid simulation, parametric studies, real-time control, or optimization design. It is an interdisciplinary field that utilizes computer technology and numerical methods to simulate fluid flow and physical phenomena, belonging to the application of physics and engineering. By discretizing mathematical models and combining them with numerical algorithms, fluid dynamics problems are solved and applied to engineering design and industrial optimization.

[0052] The three-dimensional temperature field is a three-dimensional distribution function with the rack's spatial coordinates (x, y, z) as independent variables, and a corresponding real-time temperature value at each spatial location. It can reflect the temperature level and temperature gradient at any point and at any time within the rack, and can intuitively identify abnormal thermal environments such as hot spots, high-temperature zones, temperature stratification, and heat retention zones inside the rack.

[0053] The three-dimensional airflow velocity field is a three-dimensional distribution with the rack spatial coordinates (x, y, z) as independent variables, corresponding to a real-time airflow velocity vector (magnitude + direction) at each spatial location. It can characterize the flow speed, flow direction, eddies, dead zones, and short-circuit airflow characteristics of air or cooling media within the rack, and serves as a basis for judging whether heat exchange is sufficient and whether airflow is stagnant.

[0054] Step S3: The data center adaptively identifies the abnormal thermal environment inside the rack and adaptively generates a dynamic jet control strategy.

[0055] Among them, hot spots, high-temperature zones, temperature stratification, and heat retention zones are areas with insufficient gas-liquid heat exchange efficiency, airflow stagnation, and excessive temperature, which will cause the overall temperature inside the cabinet to be too high and make it difficult to reduce the temperature inside the cabinet.

[0056] The dynamic jet control strategy is based on the real-time reconstruction of the three-dimensional temperature field and three-dimensional airflow velocity field inside the cabinet, as well as the identification of abnormal thermal environments. It involves adaptive jet regulation based on changes in heat source distribution, flow field state, and cooling demand. The optimization goals are to reduce abnormal thermal environments, improve heat exchange efficiency, and reduce fluid transport.

[0057] Furthermore, the dynamic jet control strategy involves controlling the injection position and angle of each micro-nozzle, controlling the injection flow rate, injection speed and injection timing of the micro-jet, and controlling the opening or closing cycle of the micro-jet.

[0058] Step S4: Drive the micro-nozzle installed on the back of the rack based on the dynamic jet control strategy in the data center, and use the micro-nozzle to spray the dielectric coolant micro-jet into the abnormal thermal environment, and use the micro-jet to impact and destroy the boundary layer of the abnormal thermal environment.

[0059] In this process, after being subjected to microjets, the boundary region of the abnormal thermal environment gradually decreases. At the same time, vortex-like gas is generated inside the abnormal thermal environment, which continues to spray microjets of dielectric coolant. The microjets continue to destroy the boundary layer of the abnormal thermal environment until the abnormal thermal environment is gradually eliminated.

[0060] Step S5: The data center adaptively adjusts the speed and tilt angle of the exhaust fan at the top of the rack to create a negative pressure suction flow field that couples the internal area of ​​the abnormal thermal environment with the tilt angle of the exhaust fan, thereby reducing the power consumption of fluid transport.

[0061] Furthermore, due to the vortex-like gas generated inside the abnormal thermal environment, by adjusting the tilt angle of the exhaust fan, the vortex-like gas inside the abnormal thermal environment can be directed to the exhaust fan. By adjusting the speed of the exhaust fan, the vortex-like gas inside the abnormal thermal environment can be directed to the exhaust fan quickly or slowly, forming a negative pressure suction flow field coupled with the tilt angle of the exhaust fan in the region inside the abnormal thermal environment.

[0062] Example 3;

[0063] Based on Example 2, in step S1, for the transient heat flux density distribution map of the surface, since the classical Fourier law assumes that the heat propagation speed is infinite (arriving instantaneously), it does not hold true in transient processes at the nanosecond or picosecond level. Therefore, the Cattaneo-Vernotte (CV) model of thermal inertia is introduced, and the thermal relaxation time introduced by the CV model is used. Describing heat conduction as a wave rather than simple diffusion allows for the prediction of thermal wave phenomena, as shown in the following formula:

[0064] ;

[0065] in, The transient heat flux density vector (unit: W / m²) represents the value at location. Location and Time At a given location, the heat power passing through a unit area is a vector quantity, with the direction of heat flow.

[0066] The inertial rate of change of heat flux reflects that heat flux depends not only on the current temperature gradient (unit: K / m), but also on the historical rate of change of the heat flux itself. Thermal relaxation time, measured in seconds (s), represents the time required for microscopic heat carriers (such as phonons and electrons) inside a material to establish a stable heat flow after a disturbance. The transient rate of change of heat flux density represents the current time. The rate of change of heat flux density at that location;

[0067] The classical Fourier heat flux term describes the linear relationship between the heat flux density vector and the temperature gradient in a continuous medium. Specifically, for example... Figure 3 As shown, Figure 3 The transient heat flux density distribution on the surface further demonstrates the linear relationship between heat flux density and temperature gradient in a continuous medium. Thermal conductivity (unit: W / (m·K)); Indicates the location Location and Time The temperature gradient vector at a certain location represents the temperature gradient vector at that location. In time The rate of temperature change, and the temperature is rising; The negative sign indicates that heat flows from a high temperature to a low temperature, which is a manifestation of the second law of thermodynamics. This is because heat always flows spontaneously from a high temperature region to a low temperature region. This indicates that the temperature is rising, so adding a negative sign means that the temperature is decreasing due to the heat flow.

[0068] The Catteneo-Vernotte (CV) model in this formula is the constitutive equation of the heat conduction model, also known as the non-Fourier heat flow law. It is a major modification of the classical Fourier law, revealing that in ultrafast, microscopic or high-gradient transient processes, heat transfer no longer follows instantaneous diffusion, but exhibits the physical nature of wave-like properties, inertia and finite propagation speed.

[0069] This formula is a modification of the classical Fourier law, used to describe the thermal inertia and finite heat propagation speed phenomena that occur at extremely short time scales (nanoseconds, picoseconds) or extremely small spatial scales (nanometers). The heat flow at the current moment depends not only on the current temperature gradient, but also on the rate of change of the heat flow itself (historical memory). It is applied to the instantaneous hot spot formation on the outside of server chips under microsecond-level burst loads.

[0070] Example 4;

[0071] Based on Example 2, in step S2, as follows Figure 4As shown, the reconstructed three-dimensional temperature field inside the server rack is characterized by an abnormal thermal environment due to unsteady flow, turbulence, or low-speed thermal convection within the rack, resulting in hot spots, high-temperature zones, temperature stratification, and thermal stagnation zones. A three-dimensional steady-state convection heat transfer control equation is used to describe the spatial and temporal variations of the three-dimensional temperature field of the air inside the rack. Specifically:

[0072] ;

[0073] in, The convection term (i.e., energy transport) is the movement of air to carry heat from high-temperature areas to low-temperature areas, reflecting the dominant role of convective heat transfer; The density of the gas is (kg / m³). Specific heat capacity of air at constant pressure (J / (kg·K)); , and These are the velocity components or vectors of the airflow in the x, y, and z directions, respectively (m / s). , and Representing temperature The spatial gradients in the x, y, and z directions represent the rate at which temperature changes with location and are the driving force for heat transfer.

[0074] For the thermal conduction and diffusion term, heat diffuses from the high-temperature region to the low-temperature region through the thermal conduction of air molecules, which reflects the thermal conduction effect. It represents the net heat gained or lost per unit volume of air in three-dimensional space through thermal conduction (including turbulent diffusion). The effective thermal conductivity (W / (m·K)) combines laminar and turbulent thermal conductivity of air and is used to describe the overall thermal conductivity of air in actual flow. , and These are the partial derivative operators in the x, y, and z directions, respectively; Represents the net flux change rate of heat in the x-direction. Represents the rate of change of net heat flux in the y-direction and This represents the rate of change of net heat flux in the z-direction;

[0075] For internal heat sources, this refers to the intensity of internal heat sources per unit volume (W / m³). In a server rack scenario, this mainly refers to the heat generation power of electronic devices such as servers and power supplies. The heat generated by these devices acts as an energy source, directly transferring heat to the air.

[0076] The three-dimensional steady-state convective heat transfer control equation states that the rate of change of energy per unit volume of air = energy carried in / out by convection + energy transferred by conduction + energy generated by internal heat sources. In cabinet thermal management, this is the core control mechanism for CFD simulation and reconstructing the three-dimensional temperature field using reduced-order models.

[0077] To reconstruct the three-dimensional airflow velocity field inside the cabinet, the velocity components in the x, y, and z directions are treated as a whole vector and their order is reduced to ensure physical field coupling. Specifically:

[0078] ;

[0079] in, Represents any spatial point within the server rack The three-dimensional velocity vector or components in the x, y, and z directions. Any spatial point within the server rack The velocity vector in the x-direction. Any spatial point within the server rack The velocity vector in the y-direction. Any spatial point within the server rack The velocity vector in the z-direction;

[0080] The average ground state is used to describe the statistical average of the three-dimensional velocity vector in space. It represents the steady-state part of the flow field and reflects the average velocity vector field. It is obtained from a large number of CFD snapshots and is used to eliminate random fluctuations.

[0081] This is a modal superposition term, which is formed by the linear superposition of low-dimensional dominant modes and represents the dynamic changes or disturbances in the flow field. For the past The dominant energy modes are linearly superimposed, and high-dimensional invalid noise modes are discarded. The order of the reduced CFD is less than the number of grid points in the original CFD, which reflects the dimensionality reduction method. For the first The first-order modal coefficients are low-dimensional variables that can be obtained in real time through fitting with a small amount of sensor data; For the first The first-order POD basis functions (POD basis functions usually refer to a set of optimal orthogonal basis functions extracted by the Proper Orthogonal Decomposition (POD) method. This set of basis functions can capture the main energy or feature information in the dynamic behavior of complex systems to the greatest extent with the fewest number of modes) are fixed spatial modes obtained through offline training, representing typical flow structures of the velocity field (such as eddies and jets). It is a three-dimensional velocity vector in space. , and Composed of three-dimensional velocity vectors in space .

[0082] like Figure 5 As shown, this formula assumes an arbitrarily complex three-dimensional velocity field and superimposes an average flow field into a linear combination of a series of typical flow modes. Figure 5 The velocity components in the x, y, and z directions are combined into a single three-dimensional space. The values ​​on the x, y, and z coordinates represent the magnitude of the velocity, in m / s. This is achieved by retaining the previous... The flow field is reconstructed with extremely high accuracy (typically energy retention ≥99%) using the mode with the highest energy proportion. Modal coefficients are fitted using data. This allows for online reconstruction of the entire velocity field, meeting the real-time monitoring requirements of rack thermal management. The same POD framework as the temperature field ensures physical consistency between the flow and temperature fields. Discrete point velocity measurements are transformed into a continuous representation of the global three-dimensional velocity field. Through a decomposition method combining average ground state and modal superposition, the complex turbulent flow field within the rack is reconstructed.

[0083] Example 5;

[0084] Based on Example 2, in step S3, for the adaptive identification of abnormal thermal environment inside the cabinet, a multi-dimensional anomaly fusion method is used to calculate the multi-modal anomaly fusion index in the cabinet thermal environment anomaly. Multi-source data of temperature, airflow, and vibration are fused using a Bayesian framework to give the probability of the anomaly. Specifically:

[0085] ;

[0086] in, For at any time The multimodal anomaly fusion index has a value range of [0,1]. The closer the value is to 1, the higher the credibility of the anomaly. The total number of modes (i.e., monitoring dimensions) participating in the fusion, such as temperature, power, airflow, and vibration; For the first Index of each modality; For the first The modality at time... Observed values; This assumes a normal operating condition, i.e., a baseline operating condition without any abnormalities. This indicates that under the assumption of a normal state, the first... Modal observations The probability density of occurrence;

[0087] This is a probability product term, representing the joint probability of all modal observations occurring simultaneously under the assumption of normal conditions. If all modes are within the normal range, each... Both are close to 1, and the product is also close to 1, indicating that the current state is highly consistent with the normal state; if one or more modes are abnormal, the corresponding This will significantly decrease, leading to a substantial drop in the product, indicating that the current state deviates from the normal state.

[0088] The entire formula represents When the joint probability is close to 1, A value close to 0 indicates no anomalies; when the joint probability is close to 0, A value close to 1 indicates a high degree of abnormality.

[0089] like Figure 6 As shown, the entire formula integrates information from multiple physical quantities, including temperature, fluctuation or vibration time, and wind speed, into a unified index, avoiding misjudgments caused by drift of a single sensor or local anomalies; it enables the quantification of anomalies with reliable measurement. The higher the value, the more monitoring indicators deviate from normal simultaneously, and the higher the reliability and severity of the anomaly. This can be achieved by analyzing each... The magnitude of the signal can quickly pinpoint which mode (e.g., excessively high temperature, power overload, insufficient wind speed) is the main cause of the anomaly.

[0090] Example 4;

[0091] Based on Example 2, in step S4, a micro-jet of dielectric coolant is sprayed into the abnormal thermal environment. The enhanced heat transfer effect of the micro-jet on the abnormal thermal environment is quantified using the Nusselt number correlation. By calculating the micro-jet impact heat transfer method, the impact of a single circular micro-jet is achieved, as specifically shown below:

[0092] ;

[0093] in, It is the average Nusselt number, which represents the ratio of convective heat transfer intensity to pure conduction, and is dimensionless; It is an empirical constant, obtained by fitting experimental or numerical simulations, and depends on the geometry and flow state; The jet Reynolds number is used to measure the magnitude of the jet's kinetic energy and the degree of turbulence, characterizing the first... The ratio of jet inertial force to viscous force in each mode determines whether the flow is turbulent. Prandtl number, reflecting the coupling relationship between the fluid's own thermal conductivity and flow heat transfer, characterizes... The ratio of momentum diffusivity to thermal diffusivity at a given location reflects the fluid's thermal conductivity. The dimensionless jet distance describes the degree of energy attenuation and diffusion of the jet as it travels from ejection to impact with the wall. The spray distance is the vertical distance from the nozzle to the surface being cooled. Where is the nozzle diameter, and is the inner diameter of the micro-nozzle outlet; , and These are the first, second, and third empirical indices, determined by regression analysis of experimental data, and their values ​​vary under different working conditions.

[0094] Empirical constants First Experience Index Second experience index and the third experience index The results were obtained through extensive experiments or CFD simulation fitting, and were obtained by fitting the nozzle shape (circular, square, slit), data center surface type (flat, curved, ribbed), and flow state (laminar or turbulent).

[0095] This indicates the effect of jet intensity on heat transfer. A higher jet Reynolds number means more severe fluid disturbance, a thinner boundary layer, and stronger convective heat transfer. It is usually a positive value (generally between 0.5 and 0.8).

[0096] This indicates the influence of fluid thermal properties on heat transfer. A higher Prandtl number indicates stronger viscous diffusion relative to thermal diffusion, a thinner thermal boundary layer, and stronger convective heat transfer. Therefore... It is usually a positive value (1 / 3 or 0.4 in conventional techniques).

[0097] This indicates the effect of the relative distance between the nozzle and the target surface on the heat transfer intensity, within the spray range set for the nozzle and target surface. The jet core region directly impacts the target surface, resulting in stronger disturbances and enhanced heat transfer. Usually a negative value;

[0098] Within the spray range set between the nozzle and the target surface, when If the nozzle is too small, fluid diffusion is limited, which will actually weaken heat transfer; within the spray range set between the nozzle and the target surface, when When the flow rate is too high, the jet flow weakens, and heat exchange also decreases.

[0099] This formula serves as a performance dashboard for microjets impact cooling systems. By adjusting the Reynolds number, Prandtl number, and jet distance, the degree of thermal boundary layer disruption can be precisely controlled, achieving efficient heat dissipation under extreme thermal environments. The intensity of microjets impact heat transfer is determined by the fluid's impact force (Reynolds number), the degree of matching between the fluid's thermal conductivity and flow characteristics (Prandtl number), and the geometric efficiency of energy transfer (jet distance ratio). Essentially, it is a quantitative model of energy transfer efficiency. Figure 7 As shown, microjets of dielectric coolant are sprayed into an abnormal thermal environment, generating eddies within the abnormal thermal environment.

[0100] For boundary layers in abnormal thermal environments where microjets impact and disrupt the flow, a critical heat flux (CHF) enhancement method is employed. (Critical heat flux (CHF) is a limiting parameter in engineering thermophysics and heat transfer; once the heat load exceeds this value, the cooling mechanism fails instantly, causing the temperature of the cooled object (e.g., chips, nuclear fuel rods) to spike dramatically, typically resulting in burnout or melting within milliseconds to seconds.) If the abnormal thermal environment temperature is extremely high, causing the dielectric coolant to undergo a phase change (boiling), the impact of the microjets can significantly increase the critical heat flux (CHF), preventing boundary layer damage or burnout. Specifically, an enhanced model of critical heat flux (CHF) in microjets impact cooling is used to predict the maximum heat load the boundary layer can withstand under jet cooling conditions without burnout or drying out. This is applicable to scenarios where dielectric coolant is directly sprayed onto high-temperature electronic components or combustion chamber walls. The enhanced model of critical heat flux in microjets impact cooling is as follows:

[0101] ;

[0102] in, The critical heat flux under jet impact is the maximum heat flux density that the system can withstand during microjet cooling (without burning out). The critical heat flux for pool boiling is the maximum heat flux density when there is no jet flow and only natural convection or pool boiling. These are parameters obtained from experimental fitting and that inversely reflect the effects of flow geometry and fluid properties. The density of the liquid is the density of the coolant; For the first The velocity of each jet outlet, the fluid velocity at the nozzle; Surface tension is the interfacial tension between the liquid and vapor. It is the acceleration due to gravity; It is the liquid phase density, which is the density of the saturated liquid. It is the gas phase density, and the saturated vapor density;

[0103] The form is the square root of the Weber number (We), which has been dimensionless to match the effect of gravity. The local pressure generated when the jet impacts the wall is the key driving force for pushing bubbles to detach and inhibiting the formation of a gas film. The coupling effect of buoyancy and surface tension determines the behavior of bubbles on the wall surface, i.e., whether bubbles easily form a stable gas film (leading to a decrease in CHF). It reflects the influence of buoyancy on bubble detachment, and therefore is closer to a dimensionless parameter that corrects for the Weber number or the ratio of jet kinetic head to buoyancy; surface tension To prevent bubbles from growing and detaching, Archimedes' buoyancy drives the bubble to rise and detach from the wall. This indicates that the jet can blow away air bubbles, prevent air film from covering the wall, delay burn-out, and increase CHF.

[0104] This formula utilizes micro-jet technology to enhance the detachment of bubbles from the heated surface by applying additional dynamic pressure head, thereby delaying the occurrence of film boiling and enabling safe operation under higher heat loads. Without a jet (pool boiling), when the heat flux reaches... At this point, the wall or boundary layer is covered by a stable vapor film, causing the boundary layer thermal conductivity to deteriorate rapidly and burn out. After adding microjets, the high-speed fluid continuously scours the boundary layer, blowing away newly formed bubbles and preventing them from merging into a continuous gas film, allowing for a higher heat input to reach the critical state.

[0105] Example 6;

[0106] Based on Example 2, in step S5, the adjustment of the speed and tilt angle of the exhaust fan at the top of the cabinet is calculated using the fan pressure-flow characteristic equation to describe the centrifugal or axial fan at variable speeds. and fan tilt angle Pressure-flow characteristic curve under the following conditions:

[0107] ;

[0108] in, The total pressure difference generated by the fan is the fan's ability to overcome fluid resistance and propel airflow; it is related to the variable speed. and fan tilt angle The function, by adjusting the speed and fan tilt angle This can change the fan's performance curve;

[0109] The maximum pressure at zero flow rate, when the fluid flow rate is... When the value is 0 (i.e., the fan outlet is completely blocked), the maximum pressure that the fan can generate is Maximum pressure With variable speed It is proportional to the square of . Indicates the fan tilt angle Below, the square of the unit variable speed The proportional coefficient for zero flow head that can be generated;

[0110] This is a linear loss term or slip effect term, representing the dynamic pressure drop introduced by the flow rate, reflecting the flow inside the fan and its effect with the fluid flow rate. The increase, due to fluid inertia, blade slippage, and internal friction, will cause the actual output pressure to decrease, compared to... Proportional This reflects the energy loss caused by the coupling of rotational speed and flow rate. Reflects the fan tilt angle Below, the fan's sensitivity coefficient to changes in airflow;

[0111] This is a secondary loss term or fluid resistance term, expressed as a fraction of the square of the fluid flow rate. This is directly proportional to the friction loss and local resistance in the duct (such as elbows, grilles, or filters), and also includes the turbulence losses within the fan itself. In most ventilation systems, the fluid resistance curve is inherently such that the fluid resistance is less than the fluid flow rate, thus this factor makes the fan curve more closely resemble real-world scenarios. Characterized by fan tilt angle Below, the square of the flow rate The proportionality coefficient of the pressure loss caused;

[0112] This formula reflects the situation at variable speeds. and fan tilt angle Under certain conditions, the total pressure that the fan can provide With fluid flow rate The changes in performance are represented by the fan's performance curves, with each curve corresponding to a fan tilt angle. It is used to accurately fit the pressure-flow characteristics of real fans under different operating conditions, and is especially suitable for real-time prediction and adjustment of fan behavior in adaptive control systems. For example, Figure 8 As shown, variable speed and fan tilt angle Under different conditions, different fluid flow rates and fluid pressures result in different power outputs. Figure 8 In this context, SK represents a negative pressure suction flow field, which indicates the flow rate of fluid flowing towards the fan in an abnormal thermal environment.

[0113] Microjets alter the flow field structure within the server rack, effectively reducing fluid resistance, which is equivalent to reducing the fan tilt angle. Flow square The proportionality of pressure loss caused The effective value allows the fan to operate at a lower speed, reducing fan power consumption.

[0114] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope described in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A low-power data center cooling control method for a gas-liquid heat exchanger, characterized in that, Includes the following steps: Step S1: Use the image processing module to generate a transient heat flux density distribution map of the data center surface inside the rack; Step S2: The data center obtains the transient heat flux density distribution map in the image processing module, and then uses the computational fluid dynamics reduced-order model to reconstruct the three-dimensional temperature field and three-dimensional airflow velocity field inside the rack. Step S3: The data center adaptively identifies the abnormal thermal environment inside the rack and adaptively generates a dynamic jet control strategy. The dynamic jet control strategy involves controlling the injection position and angle of each micro-nozzle, controlling the injection flow rate, injection speed and injection timing of the micro-jet, and controlling the opening or closing cycle of the micro-jet. Step S4: Drive the micro-nozzle installed on the back of the rack based on the dynamic jet control strategy in the data center, and use the micro-nozzle to spray the dielectric coolant micro-jet into the abnormal thermal environment, and use the micro-jet to impact and destroy the boundary layer of the abnormal thermal environment. Step S5: The data center adaptively adjusts the speed and tilt angle of the exhaust fan at the top of the rack to create a negative pressure suction flow field that couples the internal area of ​​the abnormal thermal environment with the tilt angle of the exhaust fan, thereby reducing the power consumption of fluid transport.

2. The low-power data center cooling control method for a gas-liquid heat exchanger according to claim 1, characterized in that, In step S1, the Catteneo-Vernotte model of thermal inertia is introduced, and the thermal relaxation time introduced by the Catteneo-Vernotte model is... Describing heat conduction as a wave rather than simple diffusion allows for the prediction of heat wave phenomena, as shown in the following formula: ； in, The transient heat flux density vector represents the value at position. Location and Time At a given location, the heat power passing through a unit area is a vector quantity, with the direction of heat flow. The inertial rate of change of heat flux reflects that heat flux depends not only on the current temperature gradient, but also on the historical rate of change of heat flux itself. Thermal relaxation time represents the time required for microscopic heat carriers inside the material to establish a stable heat flow after a disturbance. The transient rate of change of heat flux density represents the current time. The rate of change of heat flux density at that location; The classical Fourier heat flux term describes the linear relationship between the heat flux density vector and the temperature gradient in a continuous medium, reflecting the linear relationship between heat flux density and temperature gradient in a continuous medium. Thermal conductivity; Indicates the location Location and Time The temperature gradient vector at a certain location represents the temperature gradient vector at that location. In time The rate of temperature change, and the temperature is rising; The negative sign indicates that heat flows from a high temperature to a low temperature, reflecting the spontaneous flow of heat from a high temperature region to a low temperature region.

3. The low-power data center cooling control method for a gas-liquid heat exchanger according to claim 1, characterized in that, In step S2, the reconstructed three-dimensional temperature field inside the cabinet is described by a three-dimensional steady-state convection heat transfer control equation, which describes the change of the three-dimensional temperature field of the air inside the cabinet with space and time. Specifically: ； in, Convection, or energy transport, is the process by which airflow carries heat from high-temperature regions to low-temperature regions, demonstrating the dominant role of convective heat transfer. The density of the gas; The specific heat capacity of air at constant pressure; , and These are the velocity components or vectors of the airflow in the x, y, and z directions, respectively; , and Representing temperature The spatial gradients in the x, y, and z directions represent the rate at which temperature changes with location and are the driving force for heat transfer. For thermal conduction and diffusion, heat diffuses from the high-temperature region to the low-temperature region through the thermal conduction of air molecules, which reflects the thermal conduction effect and represents the net heat gained or lost per unit volume of air in three-dimensional space through thermal conduction. The effective thermal conductivity coefficient is used to describe the overall thermal conductivity of air in actual flow. , and These are the partial derivative operators in the x, y, and z directions, respectively; Represents the net flux change rate of heat in the x-direction. Represents the rate of change of net heat flux in the y-direction and This represents the rate of change of net heat flux in the z-direction; For internal heat source terms, it represents the intensity of internal heat source per unit volume, which uses the heat generated by the equipment as an energy source to directly transfer heat to the air.

4. A low-power data center cooling control method for a gas-liquid heat exchanger according to claim 1, characterized in that, In step S2, the reconstructed three-dimensional airflow velocity field inside the cabinet is achieved by reducing the order of the velocity components in the x, y, and z directions as a whole vector to ensure physical field coupling. Specifically: ； in, Represents any spatial point within the server rack The three-dimensional velocity vector or components in the x, y, and z directions. Any spatial point within the server rack The velocity vector in the x-direction. Any spatial point within the server rack The velocity vector in the y-direction. Any spatial point within the server rack The velocity vector in the z-direction; The average ground state is used to describe the statistical average of the three-dimensional velocity vector in space. It represents the steady-state part of the flow field and reflects the average velocity vector field. This is a modal superposition term, which is formed by the linear superposition of low-dimensional dominant modes and represents the dynamic changes or disturbances in the flow field. For the past The dominant energy modes are linearly superimposed, and high-dimensional invalid noise modes are discarded. This reduces the order, reflecting the method of dimensionality reduction; For the first The first-order modal coefficients are low-dimensional variables, obtained through real-time fitting using a small amount of sensor data; For the first The first-order POD basis functions are fixed spatial modes obtained through offline training, representing typical flow structures of the velocity field; It is a three-dimensional velocity vector in space.

5. A low-power data center cooling control method for a gas-liquid heat exchanger according to claim 1, characterized in that, In step S3, for the adaptive identification of abnormal thermal environment inside the cabinet, a multi-dimensional anomaly fusion method is used to calculate the multi-modal anomaly fusion index in the cabinet thermal environment anomaly. Multi-source data of temperature, airflow, and vibration are fused using a Bayesian framework to give the probability of the anomaly. Specifically: ； in, For at any time The multimodal anomaly fusion index has a value range of [0,1]. The total number of modalities or monitoring dimensions participating in the fusion; For the first Index of each modality; For the first The modality at time... Observed values; This is an assumption of normal conditions, reflecting a baseline operating condition without any abnormalities; This indicates that under the assumption of a normal state, the first... Modal observations The probability density of occurrence; This is a probability product term, representing the joint probability of all modal observations occurring simultaneously under the assumption of normal conditions.

6. A low-power data center cooling control method for a gas-liquid heat exchanger according to claim 1, characterized in that, In step S4, a microjets of dielectric coolant are sprayed into the abnormal thermal environment. The enhanced heat transfer effect of the microjets on the abnormal thermal environment is quantified using the Nusselt number correlation. The impact heat transfer of the microjets is calculated to achieve the impact of a single circular microjets, as specifically shown below: ； in, It is the average Nusselt number, which represents the ratio of convective heat transfer intensity to pure conduction. It is an empirical constant, obtained by fitting experimental or numerical simulations, and depends on the geometry and flow state; Let be the jet Reynolds number, representing the th The ratio of jet inertial force to viscous force in each mode determines whether the flow is turbulent. Prandtl number, reflecting the coupling relationship between the fluid's own thermal conductivity and flow heat transfer, characterizes... The ratio of momentum diffusivity to thermal diffusivity at a given location; The dimensionless jet distance describes the degree of energy attenuation and diffusion of the jet as it travels from ejection to impact with the wall. The spray distance is the vertical distance from the nozzle to the surface being cooled. Where is the nozzle diameter, and is the inner diameter of the micro-nozzle outlet; , and These are the first, second, and third empirical indices, determined by regression analysis of experimental data, and their values ​​vary under different working conditions. This indicates the effect of jet intensity on heat transfer; This indicates the effect of fluid thermal properties on heat transfer; This indicates the effect of the relative distance between the nozzle and the target surface on the heat transfer intensity.

7. A low-power data center cooling control method for a gas-liquid heat exchanger according to claim 1, characterized in that, In step S4, for the boundary layer that is impacted and damaged by the microjets and disrupts the abnormal thermal environment, a critical heat flux enhancement method is adopted to prevent the boundary layer from being damaged or burned. Specifically: ； in, The critical heat flux under jet impact is the maximum heat flux density that the system can withstand during microjet cooling. The critical heat flux for pool boiling is the maximum heat flux density when there is no jet flow and only natural convection or pool boiling. These are parameters obtained from experimental fitting and that inversely reflect the effects of flow geometry and fluid properties. The density of the liquid is the density of the coolant; For the first The velocity of each jet outlet, the fluid velocity at the nozzle; Surface tension is the interfacial tension between the liquid and vapor. It is the acceleration due to gravity; It is the liquid phase density, which is the density of the saturated liquid. It is the gas phase density, and the saturated vapor density; The form of the square root of the Weber number was made dimensionless to match the gravitational effect; The local pressure generated when the jet impacts the wall is the key driving force for pushing bubbles to detach and inhibiting the formation of a gas film. The coupling effect of buoyancy and surface tension determines the behavior of bubbles on the wall surface; surface tension... To prevent bubbles from growing and detaching, Archimedes' buoyancy drives the bubble to rise and detach from the wall.

8. A low-power data center cooling control method for a gas-liquid heat exchanger according to claim 1, characterized in that, In step S5, the adjustment of the speed and tilt angle of the exhaust fan at the top of the cabinet is calculated using the fan pressure-flow characteristic equation to describe the centrifugal or axial fan operating at variable speeds. and fan tilt angle Pressure-flow characteristic curve under the following conditions: ； in, The total pressure difference generated by the fan is the fan's ability to overcome fluid resistance and propel airflow; it is related to the variable speed. and fan tilt angle The function; Maximum pressure at zero flow rate, maximum pressure With variable speed It is proportional to the square of . Indicates the fan tilt angle Below, the square of the unit variable speed The proportional coefficient for zero flow head that can be generated; This is a linear loss term or slip effect term, representing the dynamic pressure drop introduced by the flow rate, reflecting the flow inside the fan and its effect with the fluid flow rate. The increase, due to fluid inertia, blade slippage, and internal friction, will cause the actual output pressure to decrease, compared to... Proportional This reflects the energy loss caused by the coupling of rotational speed and flow rate. Reflects the fan tilt angle Below, the fan's sensitivity coefficient to changes in airflow; This is a secondary loss term or fluid resistance term, expressed as a fraction of the square of the fluid flow rate. It is proportional to the frictional resistance and local resistance in the pipe, and also includes the turbulent loss inside the fan itself, so that the fluid resistance is less than the fluid flow rate. Characterized by fan tilt angle Below, the square of the flow rate The proportionality coefficient of the resulting pressure loss.

9. A low-power data center cooling control system for a gas-liquid heat exchanger, used to execute a low-power data center cooling control method for a gas-liquid heat exchanger according to any one of claims 1-8, characterized in that, include: Image processing module for generating high-resolution transient heat flux density distribution maps; The data center is connected to the image processing module to obtain the transient heat flux density distribution map in the image processing module. The data center adaptively identifies the abnormal thermal environment inside the rack based on the transient heat flux density distribution map and adaptively generates a dynamic jet control strategy. Micro-nozzles, connected to the data center, can spray micro-jets of dielectric coolant into abnormal thermal environments under the control of the data center. The exhaust fan is connected to the data center, and the data center adjusts the speed and tilt angle of the exhaust fan.

Images