Liquid cooling plate flow control method and system

Abstract

The application discloses a liquid cooling plate flow control method and system, and belongs to the field of liquid cooling control. The method comprises the following steps: predicting the heat change trend of a heat generating unit according to the obtained internal running state information of the heat generating unit; determining the physical response delay parameter of a liquid cooling branch; calculating the first instruction advance time of a flow adjustment instruction; sending the flow adjustment instruction to the adjusting valve of the corresponding liquid cooling branch at the first instruction advance time node; continuously monitoring the actual cooling effect of the liquid cooling branch and correcting the flow adjustment instruction feedback; when the heat change trend of the heat generating unit tends to decrease, according to the decreasing trend and the physical response delay parameter, the cooling liquid flow of the corresponding liquid cooling branch is reduced in advance. The application can significantly improve the heat dissipation efficiency, stability and energy utilization rate of the liquid cooling system, solve the heat dissipation problem of the heat generating component in the high-density computing environment, and avoid the local overheating risk and energy waste.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of liquid cooling control, and particularly relates to a liquid cooling plate flow control method and system. BACKGROUND

[0002] In a high-density computing environment, the server power density continues to rise, and traditional air cooling has been difficult to meet the heat dissipation demand. Direct contact liquid cooling technology has become a core solution due to the efficient heat exchange between the liquid cooling plate and the chip. The liquid cooling plate architecture is usually multi-parallel, and the cooling liquid flow of each branch is independently adjusted to provide on-demand cooling for different heat generating units.

[0003] However, the multi-parallel liquid cooling plate architecture now faces three major challenges: first, the instantaneousness and difference of dynamic load. The power of a single chip can change by milliseconds, and each heat generating unit operates independently. For example, part of the GPUs in the cluster produce a sudden surge in heat due to full load computing, while the adjacent CPUs produce little heat due to low load. The liquid cooling system only adjusts the main pump speed to control the total liquid cooling flow, which requires excessive liquid supply for peak load, but it will cause energy waste and cannot effectively reduce the overheating risk of local high heat generating units.

[0004] Second, fluid interference of parallel pipelines. Each branch shares a total pipe to form a fluid network. Adjusting the valve of a branch will change the resistance and cause the redistribution of pipe network pressure. For example, increasing the flow of a high load branch will cause the total liquid supply pipe pressure to drop, causing the flow of other branches to change passively and destroy the stability of flow distribution.

[0005] Third, time scale mismatch. Chip load mutation can cause temperature to rise suddenly in microseconds, while the liquid cooling system has inherent delays. For example, the electrically controlled valve needs tens to hundreds of milliseconds to act, and the flow field needs seconds to stabilize, which causes the flow regulation to lag behind the heat dissipation demand, easily causing chip overheating and frequency reduction, or causing waste due to continuous high flow cooling when the load drops suddenly. SUMMARY

[0006] The purpose of the present application is to overcome the defects in the prior art and provide a liquid cooling plate flow control method and system to at least solve the problem that the liquid cooling system is difficult to efficiently and accurately match the changing heat dissipation demand of each heat generating unit in a high-density computing environment, thereby possibly causing local overheating risk or unnecessary energy waste.

[0007] The specific technical solutions adopted by the present application are as follows:

[0008] In a first aspect, the present application provides a liquid cooling plate flow control method, specifically as follows:

[0009] S1: Obtain the internal operating state information of the heat generating unit, and predict the heat change trend of the heat generating unit according to the internal operating state information;

[0010] S2: determining a physical response delay parameter of the liquid cooling branch;

[0011] S3: calculating a first instruction advance time of the flow adjustment instruction according to the heat change trend and the physical response delay parameter;

[0012] S4: sending a flow adjustment instruction to the adjusting valve of the corresponding liquid cooling branch at the first instruction advance time node to synchronize the cooling liquid flow with the heat change of the heat generating unit; and adjusting the rotating speed of the main circulating pump to alleviate the fluid dynamics interference during the adjustment of the liquid cooling branch flow;

[0013] S5: continuously monitoring the actual cooling effect of the liquid cooling branch, and feeding back correction to the flow adjustment instruction according to the actual cooling effect;

[0014] S6: when predicting that the heat change trend of the heat generating unit tends to be a decreasing trend, reducing the cooling liquid flow of the corresponding liquid cooling branch in advance according to the decreasing trend and the physical response delay parameter.

[0015] As a preferred, the S3 is specifically as follows:

[0016] S31: real-time identifying the non-linear execution characteristics of the adjusting valve and the non-linear hydraulic response characteristics of the fluid system;

[0017] S32: dynamically correcting the physical response delay parameter of the liquid cooling branch according to the non-linear execution characteristics and the non-linear hydraulic response characteristics;

[0018] S33: dynamically adjusting a safety margin according to the heat change trend of the heat generating unit;

[0019] S34: calculating the first instruction advance time of the flow adjustment instruction according to the physical response delay parameter dynamically corrected by S32 and the safety margin dynamically adjusted by S33.

[0020] As a preferred, in the S4, the adjusting method of the rotating speed of the main circulating pump is specifically as follows:

[0021] S41: predicting the pressure fluctuation trend of the liquid supply main pipe, and calculating a second instruction advance time of the rotating speed adjustment instruction of the main circulating pump according to the pressure fluctuation trend;

[0022] S42: sending a rotating speed adjustment instruction to the main circulating pump at the second instruction advance time node to synchronize the rotating speed adjustment of the main circulating pump with the pressure fluctuation caused by the branch flow change, and to alleviate the fluid dynamics interference.

[0023] As a preferred, in the S41, the prediction method of the pressure fluctuation trend of the liquid supply main pipe is specifically as follows:

[0024] S411: When receiving multiple liquid cooling branch flow regulation instructions simultaneously or consecutively, receiving the flow regulation instructions of multiple liquid cooling branches;

[0025] S412: Identifying the data content of the flow regulation instructions; the data content includes the issuance sequence, time interval of the flow regulation instructions, and the expected flow change amplitude and rate of each liquid cooling branch;

[0026] S413: Based on the data content, calculate the influence of the flow change of each liquid cooling branch on the transient pressure field of the liquid supply main, and superimpose the influence time sequence to obtain the composite pressure fluctuation trend of the liquid supply main;

[0027] S414: Monitor the real-time data of multiple distributed pressure sensors on the liquid supply main, and calibrate the composite pressure fluctuation trend in combination with the real-time data.

[0028] As a preferred, the S5 is specifically as follows:

[0029] S51: Obtain real-time temperature data of multiple regions inside the heat generating unit;

[0030] S52: According to the power consumption change trend of the heat generating unit and the internal heat conduction characteristics of the chip, predict the temperature change rate and peak value of multiple regions inside the heat generating unit;

[0031] S53: When the predicted temperature change rate and peak value exceed the preset temperature threshold, send an adjustment instruction to the heat generating unit to reduce the power of the heat generating unit to reduce the heat generation of the internal region of the heat generating unit.

[0032] As a preferred, the S53 is specifically as follows:

[0033] S531: Receive multiple heat generation suppression instructions for regions inside the heat generating unit, and identify the corresponding heat generating unit region, expected suppression intensity and duration of each heat generation suppression instruction;

[0034] S532: Evaluate the influence of each heat generation suppression instruction on the overall performance of the system, and analyze whether there is a mutual conflict or superposition effect between the heat generation suppression instructions, to obtain evaluation results and analysis results;

[0035] S533: According to the evaluation results and the analysis results, adjust the intensity and execution time sequence of each heat generation suppression instruction;

[0036] S534: Send the adjusted heat generation suppression instruction to the corresponding heat generating unit.

[0037] As a preferred, the S533 is specifically as follows:

[0038] S5331A: Identify the workload type of the current system running;

[0039] S5332A: Query the preset performance sensitivity configuration, and dynamically adjust the performance impact evaluation weight of each heat generation suppression instruction according to the performance sensitivity configuration;

[0040] S5333A: Combine the dynamically adjusted performance impact evaluation weight with the expected adjustment amount of the suppression instruction to calculate the weighted performance loss to the overall performance of the current system.

[0041] As a preferred, the S532 is specifically as follows:

[0042] S5321: Calculate the heat generation suppression effect of each heat generation suppression instruction on the heat generation unit region itself, and the heat transfer or heat accumulation effect of each heat generation suppression instruction on the adjacent heat generation unit region through heat conduction;

[0043] S5322: Superimpose the heat generation suppression effect and the heat transfer or heat accumulation effect to obtain the actual heat suppression range and strength of each suppression instruction;

[0044] S5323: According to the actual heat suppression range and strength, determine whether there is a local thermal shock risk, and analyze the indirect coupling effect between each suppression instruction.

[0045] As a preferred, the S533 is specifically as follows:

[0046] S5331B: Identify the influence of the heat generation suppression instruction on the performance of the heat generation unit, and the critical task of the current system running;

[0047] S5332B: Obtain the minimum requirement of the critical task on performance, and determine whether the heat generation suppression instruction affects the region where the critical task is located;

[0048] S5333B: According to the minimum performance requirement of the critical task, calculate the minimum allowed suppression strength, and adjust the strength of the heat generation suppression instruction affecting the region where the critical task is located according to the minimum suppression strength;

[0049] S5334B: According to the local thermal shock risk and the influence of the overall system performance, adjust the strength and execution timing of the heat generation suppression instruction which does not affect the region where the critical task is located;

[0050] S5335B: Determine the dominant suppression instruction, adjust the strength and timing of other heat generation suppression instructions, and send the adjusted heat generation suppression instructions.

[0051] In a second aspect, the present application provides a liquid cooling plate flow control system, comprising:

[0052] a trend prediction module configured to acquire internal operating state information of a heat generating unit and predict a heat change trend of the heat generating unit according to the internal operating state information;

[0053] a delay parameter determination module configured to determine a physical response delay parameter of a liquid cooling branch;

[0054] a lead time calculation module configured to calculate a first instruction lead time of a flow adjustment instruction according to the heat change trend and the physical response delay parameter;

[0055] a flow main adjustment module configured to send the flow adjustment instruction to an adjustment valve of a corresponding liquid cooling branch at the first instruction lead time node to synchronize the cooling liquid flow with the heat change of the heat generating unit, and adjust the rotating speed of a main circulating pump to alleviate fluid dynamics interference during adjustment of the liquid cooling branch flow;

[0056] a feedback correction module configured to continuously monitor the actual cooling effect of the liquid cooling branch and perform feedback correction on the flow adjustment instruction according to the actual cooling effect;

[0057] a flow reduction module configured to reduce the cooling liquid flow of the corresponding liquid cooling branch in advance according to a decreasing trend and the physical response delay parameter when the heat change trend of the heat generating unit is predicted to be a decreasing trend.

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

[0059] By acquiring the internal operating state information of the heat generating unit and predicting the heat change trend thereof, combining the physical response delay parameter of the liquid cooling branch, accurately calculating the first instruction lead time of the flow adjustment instruction, and synchronizing the adjustment of the cooling liquid flow with the heat change of the heat generating unit through the lead control mechanism, the present application effectively solves the heat dissipation lag problem caused by the fast power change speed of the heat generating unit and the system response delay in the prior art. In addition, during adjustment of the liquid cooling branch flow, the rotating speed of the main circulating pump is also adjusted to alleviate the fluid dynamics interference, thereby overcoming the problem of interference of branch flow adjustment on other branches in a multi-branch parallel liquid cooling system, ensuring the stability of the whole system and the independence of each branch cooling. Finally, by continuously monitoring the actual cooling effect and performing feedback correction, the control precision and adaptability are further improved, so that the cooling liquid flow is reduced in advance when the heat change trend tends to decrease, avoiding unnecessary energy waste.

[0060] In summary, the application can significantly improve the heat dissipation efficiency, stability and energy utilization rate of the liquid cooling system, solve the heat dissipation problem of the heat generating components in the high-density computing environment, and avoid the risk of local overheating and energy waste. BRIEF DESCRIPTION OF DRAWINGS

[0061] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0062] Figure 1 is a flow chart of a liquid cooling plate flow control method according to an exemplary embodiment.

[0063] Figure 2 is a flow chart of step S3 according to an exemplary embodiment.

[0064] Figure 3 is a partial flow chart of step S4 according to an exemplary embodiment.

[0065] Figure 4 is a partial flow chart of step S41 according to an exemplary embodiment.

[0066] Figure 5 is a flow chart of step S5 according to an exemplary embodiment.

[0067] Figure 6 is a partial flow chart of step S53 according to an exemplary embodiment.

[0068] Figure 7 is a block diagram of a liquid cooling plate flow control system according to an exemplary embodiment. DETAILED DESCRIPTION

[0069] In order to make the above objectives, features and advantages of the application more apparent, the specific embodiments of the application will be described in detail below with reference to the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the application. However, the application can be practiced in a variety of ways beyond the specific embodiments described herein without departing from the scope of the application, and skilled persons in the art can make similar improvements without departing from the concept of the application, so the application is not limited to the specific embodiments disclosed below. The technical features in each embodiment of the application can be combined accordingly without conflict.

[0070] The application provides a liquid cooling plate flow control method, which mainly comprises the following steps: S1, obtaining internal operation state information of a heat generating unit, and predicting a heat change trend of the heat generating unit according to the internal operation state information; S2, determining a physical response delay parameter of a liquid cooling branch; S3, calculating a first instruction advance time of a flow adjustment instruction according to the heat change trend in S1 and the physical response delay parameter in S2; S4, sending the flow adjustment instruction to an adjusting valve of the corresponding liquid cooling branch at the first instruction advance time in S3, so that the cooling liquid flow is synchronized with the heat change of the heat generating unit; and adjusting the rotating speed of a main circulating pump to reduce the fluid dynamics interference during the adjustment of the liquid cooling branch flow; S5, continuously monitoring the actual cooling effect of the liquid cooling branch, and feeding back and correcting the flow adjustment instruction according to the actual cooling effect; and S6, when the heat change trend of the heat generating unit tends to be a decreasing trend, reducing the cooling liquid flow of the corresponding liquid cooling branch in advance according to the decreasing trend and the physical response delay parameter in S2.

[0071] The application will be described in detail below with reference to specific embodiments and drawings.

[0072] Embodiment 1

[0073] The embodiment provides a liquid cooling plate flow control method, as shown in the figure, which is a flow chart of the liquid cooling plate flow control method. The method comprises the following steps: Figure 1

[0074] S1, obtaining internal operation state information of a heat generating unit, and predicting a heat change trend of the heat generating unit according to the internal operation state information.

[0075] In the embodiment, the internal operation state information of the heat generating unit, such as power consumption data, core temperature and load rate, is read in real time through a sensor integrated in the heat generating unit or through a system management bus (SMBus). These internal operation state information can be directly used to predict the heat change trend of the heat generating unit. As an implementation manner, a simple linear regression model can be used to predict the future heat change according to the power consumption data in the past period of time. For example, if the power consumption of the heat generating unit continuously rises in the past 100 milliseconds, it is predicted that the heat change trend is rising.

[0076] S2, determining a physical response delay parameter of a liquid cooling branch.

[0077] In the embodiment, the physical response delay parameter reflects the time interval from the instruction issued by the control system to the actual response of the liquid cooling branch, which is determined by performing a step response test in the system debugging stage. Specifically, a transient opening degree change instruction is sent to the adjusting valve, and the time required for the cooling liquid flow to reach a new steady state is measured using a flow sensor, which is the physical response delay parameter.

[0078] ​S3、According to the heat change trend and the physical response delay parameter, calculate the first instruction advance time of the flow adjustment instruction. For example, if the heat of the heat generating unit is predicted to increase significantly in the future 1 second, and the physical response delay parameter of the liquid cooling branch is 0.5 seconds, the first instruction advance time can be simply set to 0.5 seconds, that is, the flow adjustment instruction is sent 0.5 seconds in advance.

[0079] S4、At the above-mentioned first instruction advance time node, send the flow adjustment instruction to the adjusting valve of the corresponding liquid cooling branch, so that the cooling liquid flow is synchronized with the heat change of the heat generating unit. And during the adjustment of the liquid cooling branch flow, the speed of the main circulating pump is adjusted to alleviate the fluid dynamics interference.

[0080] In this embodiment, the adjusting valve is an electric adjusting valve, which changes the valve opening degree by receiving an electric signal, thereby adjusting the cooling liquid flow. After the instruction is sent, the cooling liquid flow will start to change, and the target is to synchronize the cooling liquid flow with the heat change of the heat generating unit. At the same time, during the adjustment of the liquid cooling branch flow, the speed of the main circulating pump is adjusted to alleviate the fluid dynamics interference. For example, the speed of the main circulating pump can be adjusted according to the cumulative effect of the flow adjustment instructions of all liquid cooling branches to maintain the pressure stability of the liquid supply main pipe.

[0081] S5、Continuously monitor the actual cooling effect of the liquid cooling branch, and feedback correct the flow adjustment instruction according to the actual cooling effect.

[0082] In this embodiment, the cooling effect can be evaluated by monitoring the real-time temperature of the heat generating unit. If the temperature of the heat generating unit is higher than expected, the cooling liquid flow can be increased; if the temperature is lower than expected, the cooling liquid flow can be reduced. Through this feedback correction mechanism, the accuracy of the cooling effect can be ensured.

[0083] S6、When the heat change trend of the heat generating unit tends to decrease, according to the decreasing trend and the physical response delay parameter, the cooling liquid flow of the corresponding liquid cooling branch is reduced in advance. For example, if the power consumption of the heat generating unit is predicted to decrease significantly in the future 1 second, and the physical response delay parameter of the liquid cooling branch is 0.5 seconds, the flow reduction instruction can be sent 0.5 seconds in advance to avoid overcooling and energy waste.

[0084] The technical solutions of the above embodiments realize precise synchronization of the cooling liquid flow and the heat change of the heat generating unit by introducing the instruction advance time, the main circulating pump speed adjustment, and the feedback correction mechanism, significantly improve the heat dissipation efficiency, stability, and energy utilization rate of the liquid cooling system, and effectively solve the problems of response lag, mutual interference, and energy waste in the prior art. Specifically, first, by obtaining the internal operating state information of the heat generating unit and predicting the heat change trend, the system can predict the future heat dissipation demand. Combined with the physical response delay parameter of the liquid cooling branch, the calculated first instruction advance time enables the flow adjustment instruction to be sent in advance, so that the cooling liquid flow has already been synchronized when the heat of the heat generating unit actually changes, avoiding the risk of local overheating caused by response lag. For example, when the power consumption of the heat generating unit suddenly increases, by sending the instruction in advance, the cooling liquid flow is adjusted in place before the heat increases, thereby maintaining the stable operating temperature of the heat generating unit.

[0085] Subsequently, during the adjustment of the liquid cooling branch flow, the speed of the main circulating pump is adjusted to reduce the fluid dynamics interference. In a multi-parallel liquid cooling system, the flow adjustment of a single branch will cause pressure fluctuations in the liquid supply main pipe, which will in turn affect the flow of other branches. By synchronously adjusting the speed of the main circulating pump, the present application can effectively offset such pressure fluctuations, ensuring the independence and accuracy of the flow adjustment of each branch. For example, when the flow of a certain branch increases, the speed of the main circulating pump can be increased accordingly to maintain the stability of the main pipe pressure and avoid unintentional reduction of the flow of other branches.

[0086] After that, the actual cooling effect of the liquid cooling branch is continuously monitored and the flow adjustment instruction is corrected. Therefore, even if there is a deviation in the prediction or model, the feedback mechanism can correct it in time to ensure that the cooling effect always meets the requirements. For example, if the actual temperature of the heat generating unit is higher than expected, the system will immediately increase the cooling liquid flow to quickly reduce the temperature.

[0087] Finally, when the heat change trend of the heat generating unit tends to decrease, the cooling liquid flow is reduced in advance, not only avoiding overcooling, but also saving energy, avoiding the situation that the traditional system starts to reduce the flow only after the heat decreases, which may cause unnecessary energy consumption.

[0088] In a preferred embodiment of the present application, with reference to the accompanying Figure 2 , step S3 comprises:

[0089] S31, real-time identification of the non-linear execution characteristics of the regulating valve and the non-linear hydraulic response characteristics of the fluid system.

[0090] In this embodiment, through sensor data, system models or machine learning algorithms, the flow response curves of the regulating valve at different opening degrees and pressure differentials, as well as the dynamic behaviors of the fluid system such as pressure loss and heat conduction at different flow rates and temperatures, can be continuously obtained and analyzed. These nonlinear characteristics are key factors affecting the response speed and stability of the system.

[0091] S32, dynamically correct the physical response delay parameter of the liquid cooling branch according to the nonlinear execution characteristic and the nonlinear hydraulic response characteristic.

[0092] In this embodiment, the preset physical response delay parameter is adjusted in real time according to the nonlinear behavior of the regulating valve and the fluid system identified in real time. For example, when the regulating valve responds slowly or quickly under certain working conditions, or the fluid viscosity changes due to temperature changes, the corresponding delay parameter will be dynamically updated to more accurately reflect the actual response time of the current system. Thus, the accuracy of the physical response delay parameter is ensured, and the accuracy of the flow regulation instruction is improved.

[0093] S33, dynamically adjust the safety margin according to the heat change trend of the heat generating unit.

[0094] In this embodiment, according to the rate, amplitude and direction of the heat change of the heat generating unit, the additional time reserved when calculating the first instruction lead time is adaptively adjusted. For example, when it is predicted that the heat of the heat generating unit will rapidly and greatly increase, the safety margin can be appropriately increased to provide more adequate response time and avoid temperature overshoot; when the heat change is stable, the safety margin can be appropriately reduced to improve the response sensitivity of the system. Thus, while ensuring the cooling effect, the system resource utilization rate is optimized, and excessive cooling or response delay is avoided.

[0095] S34, calculate the first instruction lead time of the flow regulation instruction according to the dynamically corrected physical response delay parameter and the dynamically adjusted safety margin.

[0096] In this embodiment, the physically corrected response delay parameter and the safety margin dynamically adjusted according to the current heat change trend are combined to determine the optimal sending time of the flow regulation instruction by considering these factors comprehensively.

[0097] The technical solutions of the above embodiments obtain the real dynamic behavior of the system under different working conditions by identifying the nonlinear execution characteristics of the regulating valve and the nonlinear hydraulic response characteristics of the fluid system in real time. The physical response delay parameter is dynamically corrected according to the nonlinear characteristics, so that the delay parameter used is always highly matched with the current state of the system, thereby avoiding the deviation of the flow regulation instruction sending time caused by inaccurate parameters. At the same time, the safety margin is dynamically adjusted according to the heat change trend of the heat generating unit, so that the system can flexibly reserve additional response time according to the urgency and change speed of the heat load, and the insufficient response or excessive conservatism caused by the fixed safety margin is avoided. Through the above technical solutions, the calculation accuracy and adaptability of the first instruction lead time of the flow regulation instruction are significantly improved, and the nonlinear factors and dynamic heat load changes commonly existing in the liquid cooling system are effectively dealt with, so that the regulation of the cooling liquid flow can more accurately and timely respond to the heat change of the heat generating unit.

[0098] In one example, it is assumed that the CPU heat generating unit in a server cabinet of a data center is performing a high-intensity computing task, and its power consumption rapidly jumps from 100W under low load to 300W under high load and is maintained for a short time. The traditional liquid cooling plate flow control method may only use a preset physical response delay parameter and a fixed safety margin to calculate the first instruction lead time of the flow regulation instruction. However, in actual operation, when the cooling liquid temperature rises, its viscosity will decrease, causing the hydraulic response characteristics of the fluid system to change, thereby affecting the actual physical response delay. At the same time, the execution characteristics of the regulating valve may also be nonlinear during the rapid switching from low flow to high flow.

[0099] The above technical solutions of the present application first monitor these changes in real time. For example, through sensors installed on the regulating valve and pressure and temperature sensors in the fluid system, the system can identify the actual response speed of the regulating valve under the current working condition (nonlinear execution characteristics) and the influence of cooling liquid viscosity change on fluid resistance (nonlinear hydraulic response characteristics) in real time.

[0100] Based on these real-time data, the physical response delay parameter of the liquid cooling branch is dynamically corrected to more accurately reflect the real response time of the current system. At the same time, due to the rapid jump of CPU power consumption, the system identifies that the heat change trend of the heat generating unit is rapidly rising, and therefore dynamically increases the safety margin to ensure that even in the most unfavorable case, the cooling liquid flow can be fully in place before the temperature of the heat generating unit reaches the critical value.

[0101] Finally, combined with the dynamically corrected physical response delay parameter and the dynamically adjusted safety margin, a more accurate first instruction advance time is calculated, so that the flow regulation instruction can be sent at the best opportunity, so that when the CPU power consumption increases rapidly, the coolant flow can almost synchronously increase, effectively inhibiting the rapid rise of CPU temperature, avoiding performance degradation or overheating protection.

[0102] As a preferred embodiment of the present application, refer to the accompanying drawings Figure 3 In step S4, the specific method of "adjusting the rotation speed of the main circulating pump to alleviate the fluid dynamics interference during the adjustment of the liquid cooling branch flow" includes:

[0103] S41, predict the pressure fluctuation trend of the liquid supply main pipe, and calculate the second instruction advance time of the main circulating pump rotation speed adjustment instruction according to the pressure fluctuation trend.

[0104] In this embodiment, before the liquid cooling branch flow changes, the trend of possible pressure rise or fall in the liquid supply main pipe is predicted by analyzing the expected flow regulation instruction, system model or historical data. Among them, the liquid supply main pipe is the main pipe connecting the main circulating pump and each liquid cooling branch, and its pressure stability directly affects the flow distribution and cooling effect of each branch. According to the pressure fluctuation trend, the second instruction advance time of the main circulating pump rotation speed adjustment instruction is calculated, which aims to ensure that the rotation speed adjustment of the main circulating pump can be ahead of the actual pressure fluctuation, so as to realize active rather than passive interference suppression. The second instruction advance time can be accurately calculated according to the propagation speed of fluid in the pipe, the response characteristics of the main circulating pump, and the processing delay of the system controller and other factors.

[0105] S42, at the second instruction advance time node, send the rotation speed adjustment instruction to the main circulating pump, so that the rotation speed adjustment of the main circulating pump is synchronized with the pressure fluctuation caused by the change of branch flow, and the fluid dynamics interference is alleviated. For example, when the expected increase of branch flow leads to the drop of the liquid supply main pipe pressure, the rotation speed of the main circulating pump should be raised in advance; on the contrary, when the expected decrease of branch flow leads to the rise of the liquid supply main pipe pressure, the rotation speed of the main circulating pump should be lowered in advance.

[0106] The technical solutions of the above embodiments introduce pressure fluctuation trend prediction and instruction lead time calculation, ensure that the speed adjustment of the main circulating pump is highly synchronized with the pressure fluctuation caused by the change in branch flow, and thus significantly reduces the pressure transients and oscillations in the system. First, by predicting the pressure fluctuation trend of the liquid supply main pipe, the system can anticipate the upcoming pressure changes. Then, the second instruction lead time of the main circulating pump speed adjustment instruction is calculated, and the instruction is sent at this lead time node, so that the speed adjustment of the main circulating pump can be accurately synchronized with the pressure fluctuation caused by the change in branch flow. This forward-looking control strategy avoids the pressure transients and system oscillations that may be caused by traditional lagging responses, thereby fundamentally reducing the fluid dynamics interference. It improves the operation stability of the cooling system, reduces energy consumption, and also helps to prolong the service life of key components such as liquid cooling plates and pumps, ensuring that the heat generating units can obtain stable and efficient cooling effect under various working conditions.

[0107] In one example, assume that a liquid cooling system contains multiple liquid cooling branches, each connected to a heat generating unit. When the internal operating state information of multiple heat generating units indicates that the heat change trend needs to simultaneously or continuously adjust the cooling liquid flow of the corresponding liquid cooling branch, for example, multiple heat generating units simultaneously enter a high load operating state and need to increase the cooling liquid flow.

[0108] At this time, the system will predict that these flow adjustment instructions will cause the pressure fluctuation trend of the liquid supply main pipe. Based on the analysis of these expected flow changes, the system calculates the second instruction lead time of the main circulating pump speed adjustment instruction. For example, if it is predicted that the branch flow will increase significantly at time T, causing the main pipe pressure to drop, the system will send a speed increase instruction to the main circulating pump at time T minus the second instruction lead time (e.g., T-Δt). As a result, when the actual flow change occurs, the speed of the main circulating pump has already been adjusted, and the head it provides can effectively offset the pressure fluctuation caused by the change in branch flow, thereby maintaining the stability of the liquid supply main pipe pressure, ensuring accurate control of the flow of each liquid cooling branch, and reducing fluid dynamics interference.

[0109] As a preferred embodiment of the present application, refer to the accompanying Figure 4 In step S41, the method of "predicting the pressure fluctuation trend of the liquid supply main pipe" further includes:

[0110] S411, when receiving multiple liquid cooling branch flow adjustment instructions simultaneously or continuously, receiving multiple liquid cooling branch flow adjustment instructions.

[0111] In this embodiment, when multiple heat generating units in the system need to adjust the cooling liquid flow, their corresponding liquid cooling branches will generate corresponding flow adjustment instructions. It can be understood that the flow adjustment instructions may be issued simultaneously in a very short time, or may be issued continuously within a certain time interval.

[0112] S412, identify the data content of the flow adjustment instruction; wherein the data content includes the issuing order of the flow adjustment instruction, the time interval, and the expected flow change amplitude and rate of each branch.

[0113] In this embodiment, each received flow adjustment instruction is parsed to extract key data content. The data content includes but is not limited to the issuing order of the instruction, i.e. which branch's instruction is issued first and which is issued later; the time interval between instructions, i.e. the time difference between adjacent instructions; and the expected flow change amplitude and rate of each branch, such as adjusting the flow of a certain branch from X liters per minute to Y liters per minute, and the adjustment process is expected to be completed within Z seconds.

[0114] S413, based on the data content, calculate the influence of the flow change of each branch on the transient pressure field of the liquid supply main, and superimpose the influence time sequence to obtain the composite pressure fluctuation trend of the liquid supply main.

[0115] In this embodiment, the flow change of each liquid cooling branch will cause local pressure disturbance in the liquid supply main. This disturbance is transient and will propagate along the fluid. Through the fluid mechanics model or the preset lookup table, the specific influence of the flow change of each branch on the transient pressure field of the liquid supply main can be calculated according to the flow change amplitude, rate and position of each branch in the main. Subsequently, considering the issuing order and time interval of the instructions of different branches, these independent transient pressure influences are superimposed according to their occurrence time sequence, thereby obtaining the composite pressure fluctuation trend of the entire liquid supply main in the future period of time. The above-mentioned superposition considers the mutual enhancement or offset effect that may be produced by the flow changes of different branches.

[0116] S414, monitor the real-time data of multiple distributed pressure sensors on the liquid supply main, and calibrate the composite pressure fluctuation trend in combination with the real-time data.

[0117] In this embodiment, in order to ensure the accuracy of the prediction, multiple pressure sensors are deployed at different positions of the liquid supply main to collect real-time pressure data. The real-time data is used to compare and correct the predicted composite pressure fluctuation trend. For example, through algorithms such as Kalman filtering, the trend output by the prediction model is fused with the actual measurement data, and the parameters of the model are dynamically adjusted or the prediction result is corrected, so as to improve the precision and robustness of the prediction, and to cope with the model errors or unexpected disturbances that may occur in actual operation.

[0118] The technical solutions of the above embodiments achieve accurate prediction of the pressure fluctuation trend of the liquid supply main pipe by comprehensively and meticulously analyzing the flow regulation instructions of multiple liquid cooling branches and calibrating in combination with real-time monitoring data. Specifically, when the system receives the flow regulation instructions of multiple liquid cooling branches, the order of issuance, time interval, and expected flow change amplitude and rate of each branch of the instructions are identified and utilized. By deeply mining these key data contents, the system can calculate the specific influence of the flow change of each branch on the transient pressure field of the liquid supply main pipe based on the principle of fluid mechanics. Subsequently, by accurately superimposing these independent transient pressure influences according to their occurrence time sequence, the system can construct the composite pressure fluctuation trend of the liquid supply main pipe, thereby overcoming the prediction deviation that may be caused by considering only a single branch or simple superposition. Finally, by introducing the real-time data of multiple distributed pressure sensors on the liquid supply main pipe, the predicted composite pressure fluctuation trend is dynamically calibrated, further enhancing the accuracy and adaptability of the prediction, ensuring that the prediction result can truly reflect the current operating state of the system. The more accurate and comprehensive prediction of the pressure fluctuation trend of the liquid supply main pipe enables the speed regulation instruction of the main circulating pump to be more accurately synchronized with the pressure fluctuation caused by the flow change of the branch, reducing fluid dynamics interference and avoiding adverse effects on system stability and cooling efficiency caused by excessive pressure fluctuation.

[0119] As a more preferred embodiment of the present application, refer to the accompanying drawings Figure 5 In step S5, the specific implementation of "continuously monitoring the actual cooling effect of the liquid cooling branch" includes:

[0120] S51, acquiring real-time temperature data of multiple regions inside the heat-generating unit.

[0121] In this embodiment, real-time temperature data is acquired by integrating multiple miniature temperature sensors inside the heat-generating unit. These miniature temperature sensors are strategically placed in key hotspot regions of the heat-generating unit, such as processor cores, memory controllers, or graphics processing units, etc.

[0122] S52, predicting the temperature change rate and peak value of multiple regions inside the heat-generating unit according to the power consumption change trend of the heat-generating unit and the internal thermal conduction characteristics of the chip.

[0123] In this embodiment, a pre-established thermal model based on physical models or machine learning algorithms is used to accurately predict the temperature change trend of each region over a period of time. The purpose is to achieve early warning of potential overheating risks. It can be understood that the above-mentioned thermal model can combine the current power consumption data, historical operation data of the heat-generating unit, and the thermal conductivity, heat capacity, and other characteristic parameters of the internal materials of the chip.

[0124] S53、When the predicted temperature change rate and peak value exceed the preset temperature threshold, send an adjustment instruction to the heat generating unit to reduce the power of the heat generating unit to reduce the heat generation in the internal region of the heat generating unit.

[0125] In this embodiment, the system sends instructions to the power management unit or operating system of the heat generating unit based on the prediction results, triggering measures such as CPU / GPU frequency reduction (frequency reduction), voltage reduction (voltage reduction), core disabling, or task scheduling optimization. These measures aim to directly reduce the heat generation inside the heat generating unit, thereby intervening before the heat accumulates to a dangerous level and avoiding overheating.

[0126] The technical solutions of the above embodiments can more finely grasp the heat distribution and heat evolution law inside the heat generating unit by introducing real-time temperature data acquisition and prediction mechanisms for multiple regions inside the heat generating unit. When it is predicted that a local region inside the heat generating unit may overheat, i.e., the temperature change rate or peak value will exceed the preset threshold, the system actively sends an adjustment instruction to the heat generating unit to directly reduce the power of the heat generating unit. This direct intervention of the heat source can quickly and effectively reduce heat generation, thereby inhibiting the accumulation of heat from the source before it is conducted to the liquid cooling plate and carried away by the cooling liquid. This compensates for the limitations of simply relying on liquid cooling flow regulation, such as response lag and insufficient local region control, and achieves faster and more accurate local thermal management.

[0127] In one example, assume that a central processing unit (CPU) in a high-performance computing server is executing an intensive computing task. During task execution, the power consumption of a core region of the CPU increases sharply due to a sudden increase in workload.

[0128] At this time, the system acquires real-time temperature data of the core region through multiple temperature sensors integrated inside the CPU. At the same time, based on the power consumption trend of the CPU and a preset internal chip heat conduction model, the system predicts that the temperature of the core region will rise at a high rate in a short time and may exceed the preset temperature threshold (e.g., 90°C) in a few seconds. To avoid overheating of the core region, the system immediately sends an adjustment instruction to the power management unit of the CPU. The instruction may include reducing the operating frequency of the core, reducing its supply voltage, or even temporarily disabling part of the core's functions. For example, the frequency of the core is reduced from 4.0 GHz to 3.0 GHz, and the voltage is reduced accordingly. Through these measures, the heat generation of the core region is quickly suppressed, and its temperature rising trend is effectively curbed, thereby avoiding the actual temperature reaching or exceeding the preset threshold, ensuring the stable operation of the CPU, and giving the liquid cooling system more time to adjust the cooling liquid flow. By actively intervening the heat source and passively adjusting the liquid cooling flow, a more comprehensive and efficient thermal management strategy is formed.

[0129] As a preferred embodiment of the present invention, refer to the appendix. Figure 6 Step S53 includes:

[0130] S531. Receive heat generation suppression commands from multiple internal regions of heating units, and identify the heating unit region, expected suppression intensity, and duration corresponding to each heat generation suppression command.

[0131] In this embodiment, the system receives power reduction requests from different sensors, monitoring modules, or upper-level control logic for different heat-generating areas or different time points within the same heat-generating area. Upon receiving these instructions, the system identifies them and extracts key information, such as which specific heat-generating unit area the instruction targets (e.g., a CPU core, a GPU computing unit, a memory module chip, etc.), the expected power reduction magnitude (i.e., suppression intensity), and the duration of the suppression.

[0132] S532. Evaluate the impact of each heat generation suppression command on the overall system performance, and analyze whether there are mutual conflicts or superimposed effects between heat generation suppression commands, and obtain evaluation results and analysis results.

[0133] In this embodiment, the potential impact of single or multiple power reduction operations on key performance indicators such as system throughput, response time, and task completion rate is quantified. For example, if a heat-generating unit area is responsible for performing a critical task, power suppression on it may lead to task delays or failures. Analyzing whether there are conflicts or cumulative effects between heat generation suppression commands refers to identifying potential interactions between different commands. Conflicts may manifest as two commands making contradictory demands on the same area, or the execution of one command hindering the execution of another. Cumulative effects may refer to multiple commands acting simultaneously on adjacent areas, resulting in excessive local temperature drops or excessive concentration of cooling resources, or affecting other areas through heat conduction.

[0134] S533. Based on the evaluation and analysis results, adjust the intensity and execution sequence of each heat generation suppression command.

[0135] In this embodiment, specific adjustments include: for instructions that have a significant impact on system performance, appropriately reducing their suppression intensity or delaying their execution timing; for conflicting instructions, prioritizing the execution of critical instructions or coordinating conflicting instructions; for instructions with cumulative effects, adjusting their intensity or staggering their execution timing to avoid excessive or uneven suppression.

[0136] S534. Send adjusted heat generation suppression commands to the corresponding heating units to ensure that these commands can be executed in an optimized manner, thereby achieving efficient and lossless thermal management.

[0137] The technical solutions of the above embodiments achieve fine management of heat generation suppression in the internal region of the heat generating unit by introducing mechanisms for receiving, identifying, evaluating, analyzing, and adjusting multiple heat generation suppression instructions. Specifically, by uniformly coordinating and optimizing multiple heat generation suppression instructions, the system effectively avoids unnecessary impact on the overall performance of the system caused by simple and rough power reduction strategies due to local overheating. Subsequently, by identifying and handling conflicts and superposition effects among the instructions, the system ensures the synergy and efficiency of the thermal management strategy, prevents excessive cooling or insufficient cooling in local regions, and ultimately improves the adaptability and robustness of the liquid cooling plate flow control system under complex working conditions, achieving more intelligent system thermal management.

[0138] In one example, assume that in a high-performance computing system, two cores of a CPU (Core A and Core B) and an independent graphics processor (GPU) module simultaneously detect that the temperature is too high and trigger three heat generation suppression instructions respectively. Instruction 1 requires Core A to reduce power by 20% for 10 seconds; Instruction 2 requires Core B to reduce power by 15% for 8 seconds; and Instruction 3 requires GPU to reduce power by 10% for 12 seconds.

[0139] First, the system receives and identifies the three heat generation suppression instructions, obtaining their respective heat generating unit regions, expected suppression intensities, and durations.

[0140] Next, the system evaluates the impact of each instruction on the overall performance of the system. For example, if Core A is executing a critical task sensitive to delay, while Core B and GPU are executing non-critical background tasks, the impact weight of Instruction 1 on system performance will be higher. At the same time, the system analyzes whether there are conflicts or superposition effects among the instructions. For example, Core A and Core B are physically adjacent, and the superposition effect of Instruction 1 and Instruction 2 may cause the temperature in the local region to drop too quickly, or cause a transient peak in the demand for cooling liquid flow.

[0141] Based on the evaluation results and analysis results, the system makes adjustments. For example, due to the critical task of Core A, the system may adjust the suppression intensity of Instruction 1 from 20% to 10% and slightly delay its execution timing by 2 seconds to give Core A some buffer time to complete the current critical computation. For Core B and GPU, considering that they are non-critical tasks and there is a superposition effect, the system may keep the suppression intensity of Instruction 2 at 15% unchanged, but stagger its execution timing with Instruction 1, such as starting to execute 2 seconds after Instruction 1, to smooth the demand for cooling liquid flow. The suppression intensity of Instruction 3 for GPU may remain at 10% unchanged, but its execution timing may be adjusted to start after the suppression instructions for Core A and Core B end, to avoid causing too much impact on the overall performance of the system.

[0142] Finally, the system sends the adjusted heat generation suppression instructions to core A, core B, and the GPU. Through this coordination and optimization, the system can effectively control the temperature of the heat-generating units while minimizing the impact on critical task performance and avoiding potential conflicts or efficiency problems caused by the simultaneous execution of multiple suppression instructions.

[0143] In a feasible design, step S533 includes:

[0144] S5331A, identify the type of workload currently running on the system.

[0145] In this embodiment, by monitoring system indicators such as CPU utilization, memory usage, input / output (I / O) throughput, network traffic, etc., and combining predefined load patterns (such as high computational load, high I / O load, idle state, video rendering, database query, etc.), the current running mode of the system is determined.

[0146] S5332A, query the preset performance sensitivity configuration, and dynamically adjust the performance impact evaluation weight of each heat generation suppression instruction according to the performance sensitivity configuration.

[0147] In this embodiment, a set of performance sensitivity parameters corresponding to different workload types is read from the storage medium. These parameters are pre-calibrated according to the performance of the system under different loads and user experience requirements. Then, according to the currently identified workload type, the corresponding performance sensitivity parameters are selected from the preset configuration, and based on this, different weights are given to the performance impact that each heat generation suppression instruction may cause. For example, under a workload with extremely high performance requirements, any suppression instruction that may cause performance degradation will be given a higher weight to limit its execution intensity or delay its execution timing; while under a workload with lower performance requirements, a lower weight can be given to allow more aggressive suppression strategies, so that the performance evaluation is more in line with the actual running scenario, avoiding unnecessary performance loss.

[0148] S5333A, combine the adjusted performance impact evaluation weight with the expected adjustment amount of the suppression instruction to calculate the weighted performance loss to the overall performance of the current system.

[0149] In this embodiment, the expected power reduction amount or duration of each heat generation suppression instruction is multiplied or weighted sum operation with the performance impact evaluation weight given to the instruction under the current workload, so as to obtain the comprehensive impact of the instruction on the overall performance of the system. Finally, the weighted performance loss of all instructions is further superimposed to obtain the weighted performance loss prediction value of the overall system.

[0150] The technical solutions of the above embodiments realize intelligent and adaptive adjustment of the heat generation suppression instructions. Compared with adjustment based only on general evaluation results, the application can dynamically adjust the performance impact evaluation weight according to the current workload type of the system, so that the adjustment of the strength and execution timing of the suppression instructions is more refined and rational. Thus, while ensuring effective heat dissipation, the negative impact on the overall performance of the system is minimized, unnecessary performance loss is effectively avoided when executing critical tasks, and the running efficiency and user experience of the system are improved.

[0151] In a feasible design, the specific implementation of "analyzing whether there is a mutual conflict or superposition effect between the heat generation suppression instructions" in step S532 includes:

[0152] S5321, calculating the heat generation suppression effect of each heat generation suppression instruction on the self-heating unit region, and the heat transfer or heat accumulation effect of each heat generation suppression instruction on the adjacent heating unit region through heat conduction.

[0153] In this embodiment, for each transmitted heat generation suppression instruction, the power reduction or heat reduction directly caused in the target heating unit region is accurately quantified. The accurate quantification of the suppression effect can be obtained by consulting a preset suppression instruction parameter table, performing simulation or monitoring the power consumption change of the heating unit in real time. When the heat generation of a certain heating unit region is suppressed, the temperature drop may affect the temperature distribution of the adjacent region. Specifically, heat may be transferred from the adjacent region with a higher temperature to the suppressed region, or in some cases, suppressing the heat generation of one region may change the heat accumulation effect of other regions. By establishing a refined heat model inside the heating unit, combined with finite element analysis or computational fluid dynamics (CFD) simulation, the conduction path and rate of heat between different regions are predicted, so as to quantify the indirect heat influence.

[0154] S5322, superimposing the heat generation suppression effect and the heat transfer or heat accumulation effect to obtain the actual heat suppression range and strength of each suppression instruction.

[0155] In this embodiment, the direct suppression effect in the target region and the indirect influence generated through heat conduction are algebraically superimposed or model fused. For example, if an instruction directly reduces the heat of a certain region by 10W, while causing 2W of heat to be transferred to the region through heat conduction from the adjacent region, the actual net suppression effect of the instruction on the target region is 8W. Through such superposition, the real heat influence distribution diagram of each suppression instruction inside the entire heating unit can be obtained, including the boundary of its influence and the suppression strength at different positions.

[0156] S5323, determine whether there is a local thermal shock risk based on the actual heat suppression range and intensity, and analyze the indirect coupling effect between instructions.

[0157] In this embodiment, based on the actual heat suppression range and intensity data obtained above, it is identified whether there is a local temperature gradient too large, temperature fluctuation too large and other phenomena caused by rapid change or uneven distribution of heat in the heat generating unit, which will adversely affect the chip life or performance, i.e. local thermal shock risk. At the same time, by analyzing the actual heat suppression range and intensity of different suppression instructions, it is identified whether there is an indirect coupling effect of overlapping, canceling or enhancing between them, for example, two instructions may be implemented in different areas, but their heat conduction effect produces superposition in a common area, resulting in unexpected temperature changes.

[0158] The technical solutions of the above embodiments can accurately obtain the actual heat suppression range and intensity of each suppression instruction in the heat generating unit by finely calculating the direct suppression effect and indirect heat conduction effect of each heat generation suppression instruction and superimposing them. Subsequently, the system effectively determines whether there is a local thermal shock risk based on these actual heat suppression range and intensity data, and deeply analyzes the possible indirect coupling effect between different instructions, ensuring that when implementing the heat suppression strategy, not only the direct effect of a single instruction is considered, but also the complex influence of the entire thermal environment is comprehensively considered, thereby avoiding potential problems caused by local overcooling or uneven heat distribution.

[0159] In a feasible design, step S533 includes:

[0160] S5331B, identify the influence of the heat generation suppression instruction on the performance of the heat generating unit, and the key task of the current system operation.

[0161] In this embodiment, by using the pre-set performance model or real-time monitoring data, the influence of each suppression instruction on the calculation ability, response speed or data throughput of the heat generating unit when reducing the power of the heat generating unit is evaluated. At the same time, the key task of the current system operation refers to the task that has strict requirements on system stability, response time or data integrity, such as the core computing task of the data center, the decision module in the automatic driving system or the vital sign monitoring function in the medical device. These key tasks usually have pre-set performance priority and minimum performance requirements.

[0162] S5332B, obtain the minimum requirement of the key task on the performance, and determine whether the heat generation suppression instruction affects the region where the key task is located.

[0163] In this embodiment, the minimum processing capacity, upper limit of latency and other performance indicators required to ensure the normal operation of the critical task are obtained from the configuration database or through a dynamic negotiation mechanism. The determination of whether the heat generation suppression instruction affects the region where the critical task is located can be achieved by comparing the physical location or logical mapping relationship of the heat generating unit region affected by the suppression instruction and the heat generating unit region relied on by the critical task.

[0164] S5333B, according to the minimum performance requirement of the critical task, the minimum suppression strength allowed is calculated, and the strength of the heat generation suppression instruction affecting the region where the critical task is located is adjusted according to the minimum suppression strength.

[0165] In this embodiment, the minimum suppression strength is the maximum allowed amplitude of the suppression of the power of the heat generating unit under the premise of not being lower than the minimum performance requirement of the critical task, and the minimum suppression strength is used as the upper limit of the strength of the heat generation suppression instruction affecting the region where the critical task is located.

[0166] S5334B, according to the local heat shock risk and the overall system performance influence, the strength and execution timing of the heat generation suppression instruction not affecting the region where the critical task is located are adjusted.

[0167] In this embodiment, for those heat generating unit regions that do not directly affect the critical task, the system flexibly adjusts the suppression strength and execution time point according to the previously evaluated local heat shock risk (for example, avoiding the temperature of a certain region from dropping or rising too fast) and the comprehensive influence on the overall system performance (for example, ensuring that other non-critical tasks can also operate normally).

[0168] S5335 B, determine the dominant suppression instruction, adjust the strength and timing of other heat generation suppression instructions, and send the adjusted heat generation suppression instructions.

[0169] In this embodiment, determining the dominant suppression instruction means that among multiple interrelated or conflicting suppression instructions, according to the preset priority rules, heat management strategies or system performance targets, a suppression instruction with the highest decision-making power or the most priority execution is selected. Subsequently, the strength and timing of other heat generation suppression instructions are adjusted according to the execution plan of the dominant suppression instruction, so as to ensure the coordination and effectiveness of the overall heat management strategy. Finally, the adjusted heat generation suppression instructions will be sent to the corresponding heat generating unit to achieve fine power adjustment.

[0170] The technical solutions of the above embodiments solve the problem that the system key function may be negatively affected when heat is inhibited by introducing the identification of key tasks and the consideration of performance requirements, and the comprehensive evaluation of the local thermal shock risk and the overall system performance impact. Specifically, by identifying the influence of the heat generation inhibition instruction on the performance of the heat generating unit and the key tasks of the current system operation, the system can distinguish the importance of different areas and tasks. Then, the minimum requirement of the key task on the performance is obtained and it is judged whether the inhibition instruction affects the area where the key task is located, so that the system can draw a performance protection zone for the key task. On this basis, the minimum inhibition intensity allowed is calculated and the inhibition instruction intensity affecting the area where the key task is located is adjusted accordingly, ensuring that the key task can still meet its minimum performance requirement during the heat management process, and avoiding the performance decline caused by excessive inhibition. At the same time, the inhibition instruction not affecting the area where the key task is located is adjusted according to the local thermal shock risk and the overall system performance impact, so that the system can perform more flexible and optimized heat inhibition in the non-key area, thereby maximizing the overall system heat dissipation efficiency and stability while ensuring the performance of the key task. Finally, by determining the dominant inhibition instruction and coordinating the intensity and timing of other instructions, intelligent collaborative management of multiple heat generation inhibition instructions is realized, ensuring the global optimality of the heat management strategy.

[0171] Embodiment 2

[0172] The present embodiment provides a liquid cooling plate flow control system, referring to the accompanying Figure 7 The system mainly includes:

[0173] The trend prediction module 01 is configured to obtain internal operating state information of the heat generating unit, and predict the heat change trend of the heat generating unit according to the internal operating state information.

[0174] The delay parameter determination module 02 is configured to determine the physical response delay parameter of the liquid cooling branch.

[0175] The lead time calculation module 03 is configured to calculate a first instruction lead time of the flow adjustment instruction according to the heat change trend and the physical response delay parameter.

[0176] The flow main adjustment module 04 is configured to send the flow adjustment instruction to the adjustment valve of the corresponding liquid cooling branch at the first instruction lead time node, so that the cooling liquid flow is synchronized with the heat change of the heat generating unit, and the speed of the main circulating pump is adjusted during the adjustment of the liquid cooling branch flow to reduce the fluid dynamics interference.

[0177] The feedback correction module 05 is configured to continuously monitor the actual cooling effect of the liquid cooling branch, and feedback correct the flow adjustment instruction according to the actual cooling effect.

[0178] The flow reduction module 06 is configured to reduce the flow of the cooling liquid in advance according to the decreasing trend and the physical response delay parameter when the heat change trend of the heat generating unit tends to be a decreasing trend.

[0179] Therefore, by obtaining the internal operation state information of the heat generating unit and predicting the heat change trend thereof, and combining the physical response delay parameter of the liquid cooling branch, the first instruction advance time of the flow adjustment instruction is accurately calculated, and the adjustment of the cooling liquid flow is synchronized with the heat change of the heat generating unit through the advance control mechanism, thereby effectively solving the heat dissipation lag problem caused by the fast power change speed of the heat generating unit and the system response delay in the prior art. In addition, during the adjustment of the flow of the liquid cooling branch, the speed of the main circulating pump is also adjusted to reduce the fluid dynamics interference, thereby overcoming the problem of interference of the branch flow adjustment on other branches in the multi-parallel liquid cooling system, and ensuring the stability of the whole system and the independence of the cooling of each branch. Finally, by continuously monitoring the actual cooling effect and performing feedback correction, the control accuracy and adaptability are further improved, so that the flow of the cooling liquid is reduced in advance when the heat change trend tends to be a decreasing trend, thereby avoiding unnecessary energy waste.

[0180] In summary, the present application can significantly improve the heat dissipation efficiency, stability and energy utilization rate of the liquid cooling system, solve the heat dissipation problem of the heat generating components in the high-density computing environment, and avoid the risk of local overheating and energy waste.

[0181] The above-mentioned embodiments are only a preferred scheme of the present application, and are not intended to limit the present application. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application. Therefore, any technical scheme obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present application.

Claims

1. A method for controlling the flow rate of a liquid-cooled plate, characterized in that, Specifically as follows: S1: Obtain the internal operating status information of the heating unit, and predict the heat change trend of the heating unit based on the internal operating status information; S2: Determine the physical response delay parameters of the liquid cooling branch; S3: Calculate the first command lead time of the flow regulation command based on the heat change trend and the physical response delay parameter; S4: At the time node of the first instruction advance, send a flow regulation instruction to the regulating valve of the corresponding liquid cooling branch to synchronize the coolant flow with the heat change of the heating unit; and during the adjustment of the flow of the liquid cooling branch, adjust the speed of the main circulation pump to reduce hydrodynamic interference. S5: Continuously monitor the actual cooling effect of the liquid cooling branch, and provide feedback correction to the flow rate adjustment command based on the actual cooling effect; S6: When the predicted heat change trend of the heating unit tends to decrease, the coolant flow rate of the corresponding liquid cooling branch is reduced in advance based on the decreasing trend and the physical response delay parameter.

2. The liquid cooling plate flow control method according to claim 1, characterized in that, S3 is specifically as follows: S31: Real-time identification of the nonlinear actuation characteristics of control valves and the nonlinear hydraulic response characteristics of fluid systems; S32: Based on the nonlinear execution characteristics and the nonlinear hydraulic response characteristics, dynamically correct the physical response delay parameters of the liquid cooling branch; S33: Dynamically adjust the safety margin according to the heat change trend of the heating unit; S34: Calculate the first command lead time of the flow regulation command based on the physical response delay parameter dynamically corrected in S32 and the safety margin dynamically adjusted in S33.

3. The liquid cooling plate flow control method according to claim 1, characterized in that, In step S4, the method for adjusting the speed of the main circulation pump is as follows: S41: Predict the pressure fluctuation trend of the main supply pipe, and calculate the second command lead time of the main circulation pump speed adjustment command based on the pressure fluctuation trend; S42: At the second instruction advance time node, a speed adjustment instruction is sent to the main circulation pump to synchronize the speed adjustment of the main circulation pump with the pressure fluctuation caused by the change of branch flow, thereby reducing fluid dynamic interference.

4. The liquid cooling plate flow control method according to claim 3, characterized in that, In step S41, the method for predicting the pressure fluctuation trend of the main supply pipe is as follows: S411: When multiple liquid cooling branches receive flow adjustment commands simultaneously or continuously, receive the flow adjustment commands of the multiple liquid cooling branches. S412: Identify the data content of the flow regulation command; the data content includes the order of issuance of the flow regulation command, the time interval, and the expected flow change amplitude and rate of each liquid cooling branch; S413: Based on the data content, calculate the impact of the flow rate change of each liquid cooling branch on the transient pressure field of the liquid supply main pipe, and superimpose the impact time sequence to obtain the composite pressure fluctuation trend of the liquid supply main pipe; S414: Monitor real-time data from multiple distributed pressure sensors on the main supply line, and calibrate the composite pressure fluctuation trend based on the real-time data.

5. The liquid cooling plate flow control method according to claim 1, characterized in that, S5 is specifically as follows: S51: Acquire real-time temperature data of multiple areas inside the heating unit; S52: Based on the power consumption change trend of the heating unit and the internal thermal conduction characteristics of the chip, predict the temperature change rate and peak value of multiple regions inside the heating unit; S53: When the predicted rate of temperature change and peak value exceed the preset temperature threshold, an adjustment command is sent to the heating unit to reduce the power of the heating unit in order to reduce the heat generation in the internal area of ​​the heating unit.

6. The liquid cooling plate flow control method according to claim 5, characterized in that, S53 is specifically as follows: S531: Receive heat generation suppression commands for the internal regions of the multiple heat generation units, and identify the heat generation suppression command corresponding to the heat generation suppression region, the expected suppression intensity, and the duration; S532: Evaluate the impact of each heat generation suppression command on the overall system performance, and analyze whether there are mutual conflicts or superimposed effects among the heat generation suppression commands, and obtain evaluation results and analysis results; S533: Based on the evaluation results and the analysis results, adjust the intensity and execution sequence of each of the heat generation suppression commands; S534: Send an adjusted heat generation suppression command to the corresponding heating unit.

7. The liquid cooling plate flow control method according to claim 6, characterized in that, The specific details of S533 are as follows: S5331A: Identifies the type of workload currently running on the system; S5332A: Query the preset performance sensitivity configuration, and dynamically adjust the performance impact evaluation weight of each heat generation suppression command according to the performance sensitivity configuration; S5333A: Combine the dynamically adjusted performance impact assessment weights with the expected adjustment amount of the suppression command to calculate the weighted performance loss on the overall performance of the current system.

8. The liquid cooling plate flow control method according to claim 6, characterized in that, The specific details of S532 are as follows: S5321: Calculate the heat suppression effect of each heat generation suppression command on its own heat generation unit area, and the heat transfer or heat accumulation effect of each heat generation suppression command on adjacent heat generation unit areas through heat conduction; S5322: By superimposing the heat generation inhibition effect and the heat transfer or heat accumulation effect, the actual heat suppression range and intensity of each inhibition command are obtained; S5323: Based on the actual heat suppression range and intensity, determine whether there is a risk of local thermal shock, and analyze the indirect coupling effect between the suppression commands.

9. The liquid cooling plate flow control method according to claim 8, characterized in that, The specific details of S533 are as follows: S5331B: Identify the impact of the heat generation suppression command on the performance of the heating unit, and the key tasks of the current system operation; S5332B: Obtain the minimum performance requirements of the critical task and determine whether the heat generation suppression command affects the area where the critical task is located; S5333B: Calculate the minimum allowable suppression intensity based on the minimum performance requirements of the critical task, and adjust the intensity of the heat generation suppression command affecting the area where the critical task is located based on the minimum suppression intensity; S5334B: Based on the impact of local thermal shock risk and overall system performance, adjust the intensity and execution timing of the heat generation suppression command that does not affect the area where the critical task is located; S5335B: Determine the dominant suppression command, adjust the strength and timing of other heat generation suppression commands, and send the adjusted heat generation suppression command.

10. A liquid-cooled plate flow control system, characterized in that, include: The trend prediction module acquires the internal operating status information of the heating unit and predicts the heat change trend of the heating unit based on the internal operating status information. The delay parameter determination module is used to determine the physical response delay parameters of the liquid cooling branch; The lead time calculation module is used to calculate the first command lead time of the flow regulation command based on the heat change trend and the physical response delay parameter. The main flow control module is used to send a flow control command to the control valve of the corresponding liquid cooling branch at the time node before the first command, so as to synchronize the coolant flow with the heat change of the heating unit; and during the flow control of the liquid cooling branch, the speed of the main circulation pump is adjusted to reduce hydrodynamic interference. The feedback correction module is used to continuously monitor the actual cooling effect of the liquid cooling branch and correct the flow rate adjustment command based on the actual cooling effect. The flow reduction module is used to reduce the coolant flow rate of the corresponding liquid cooling branch in advance when the predicted heat change trend of the heating unit tends to decrease, based on the decreasing trend and the physical response delay parameter.

Images