A control method of a liquid cooling system

Abstract

The application provides a control method of a liquid cooling system, and belongs to the technical field of liquid cooling heat dissipation. The temperature information in the liquid cooling system is acquired, and the current state of the liquid cooling system is determined as a low load state, a normal working state or an extreme working state according to the temperature information. When the state is determined as the low load state, a control mode of reducing cooling energy consumption is executed. When the state is determined as the normal working state, a control mode of maintaining or adjusting cooling capacity is executed. When the state is determined as the extreme working state, a control mode of enhancing cooling is executed. Through state determination and hierarchical control, the application realizes the matching of the cooling strategy and the real-time load, suppresses the temperature overshoot and the frequency reduction risk under the extreme load, reduces the energy consumption under the low load, and improves the operation stability under the steady state.

Description

Technical Field

[0001] This invention belongs to the field of liquid cooling technology, and more specifically, relates to a control method for a liquid cooling system. Background Technology

[0002] With the rapid development of technologies such as artificial intelligence and big data, the demand for computing power in data centers is exploding. The power density of CPUs, GPUs, and other high-power computing devices continues to rise, leading to increasingly stringent heat dissipation requirements. Traditional air-cooling technology, due to its limited heat dissipation capacity and susceptibility to localized hotspots, is no longer sufficient to meet the high-efficiency heat dissipation requirements of modern data centers. Liquid cooling technology, with its high heat dissipation efficiency and relatively low energy consumption, is gradually becoming an important technological direction for data center heat dissipation.

[0003] Currently, some liquid cooling systems have begun to adopt temperature monitoring and feedback control strategies, such as adjusting coolant flow or pump speed based on the real-time temperature of the CPU or GPU. However, existing control methods are mostly based on passive adjustment using instantaneous temperature values, meaning that they only execute response actions after the temperature reaches or exceeds a set threshold. This lag control mode is difficult to cope with rapid changes in thermal load in scenarios such as AI training, and can easily lead to temperature overshoot or frequency throttling events. Summary of the Invention

[0004] The purpose of this invention is to provide a control method for a liquid cooling system, which aims to solve the problem that existing liquid cooling systems rely on hysteresis adjustment based on instantaneous temperature, making it difficult to cope with rapid changes in heat load and easily leading to temperature overshoot or frequency reduction.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is: to provide a control method for a liquid cooling system, comprising: Obtain temperature information from the liquid cooling system; The current operating state of the liquid cooling system is determined based on the temperature information. The operating state includes low load state, normal operating state, or extreme operating state. When the load condition is determined to be low, a control mode that reduces cooling energy consumption is executed. When the system is determined to be in normal working condition, a control mode is executed to maintain or adjust the cooling capacity. When the operating condition is determined to be extreme, the enhanced cooling control mode is executed.

[0006] In one possible implementation, the temperature information includes one or more of the following: the real-time temperature of the heat source, the temperature change trend of the heat source, and the liquid supply temperature of the liquid cooling system.

[0007] In one possible implementation, the real-time temperature of the heat source is divided into a normal operating temperature range, a high-load operating temperature range, and an upper limit of the operating temperature. The temperature change trend of the heat source is divided into a temperature rising phase, a temperature stabilizing phase, and a temperature falling phase. The liquid supply temperature of the liquid cooling system includes the liquid temperature before heat exchange, the liquid temperature after heat exchange, and the liquid temperature after cooling by the cooling tower.

[0008] In one possible implementation, the criteria for determining the low load state include: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is a decreasing phase.

[0009] In one possible implementation, the conditions for determining the normal operating state include: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is in a stable phase; or, The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is in the rising phase, but the rate of temperature change is less than or equal to a preset threshold; or, The real-time temperature of the heat source is within the high-load operating temperature range, and the temperature change trend of the heat source is a decreasing phase; or, The real-time temperature of the heat source is in the high-load operating temperature range, and the temperature change trend of the heat source is in a stable phase.

[0010] In one possible implementation, the criteria for determining the extreme operating state include: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is in an upward phase, with the rate of temperature change exceeding a preset threshold; or, The real-time temperature of the heat source is in the high-load operating temperature range, and the temperature change trend of the heat source is in the rising stage.

[0011] In one possible implementation, the control mode for reducing cooling energy consumption includes: Reduce the pump speed; or, Shut down part or all of the liquid cooling system; or Increase the liquid supply temperature.

[0012] In one possible implementation, the control mode for maintaining or regulating cooling capacity includes: When the real-time temperature of the heat source is within the normal operating temperature range and the temperature change trend is in the stable phase, maintain the current pump speed or reduce the pump frequency to maintain the liquid supply temperature. When the real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend is in the temperature rising stage and the temperature change rate is less than or equal to a preset threshold, the liquid flow rate is increased. When the real-time temperature of the heat source is in the high-load operating temperature range and the temperature change trend is in the temperature decrease stage, maintain the current pump speed and maintain the liquid supply temperature. When the real-time temperature of the heat source is in the high-load operating temperature range and the temperature change trend is in the stable stage, increase the liquid flow rate.

[0013] In one possible implementation, the control mode for performing enhanced cooling includes: Increase liquid flow and / or activate solid-state cartridge cooling.

[0014] In one possible implementation, the extreme operating state also includes a cold plate blockage state, which is diagnosed by the temperature difference ΔT after heat exchange.

[0015] The beneficial effects of the control method for a liquid cooling system provided by this invention are as follows: Compared with the prior art, this invention first acquires the temperature information of the liquid cooling system to provide a basis for subsequent control decisions; then, based on the temperature information, it determines whether the liquid cooling system is currently in a low-load state, a normal working state, or an extreme working state, enabling the system to identify different load conditions; on this basis, when it is determined to be in a low-load state, a control mode that reduces cooling energy consumption is executed; when it is determined to be in a normal working state, a control mode that maintains or adjusts cooling capacity is executed; and when it is determined to be in an extreme working state, a control mode that enhances cooling is executed, thereby achieving the matching of control strategy and load state.

[0016] Therefore, under extreme operating conditions, enhanced cooling can cope with rapid changes in heat load, avoiding temperature overshoot or frequency throttling caused by lag regulation; under low load conditions, cooling energy consumption can be reduced to avoid energy waste; and under normal operating conditions, cooling capacity can be maintained or adjusted to keep the system running stably. Thus, this invention solves the problems of temperature overshoot and frequency throttling caused by lag regulation in existing technologies, while simultaneously ensuring system stability and energy consumption control. Detailed Implementation

[0017] To make the technical problems, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] The present invention will now describe a control method for a liquid cooling system. This method is applicable to the cooling control of CPUs, GPUs, and other high-power computing devices that employ liquid cooling in data centers.

[0019] The method includes the following steps: Step S1: Obtain temperature information from the liquid cooling system.

[0020] During the operation of the liquid cooling system, temperature data is collected in real time by temperature sensors placed at various key locations in the system. Temperature sensors can be thermocouples, thermistors, or integrated temperature sensing chips. The collected temperature information includes one or more of the following: the real-time temperature of the heat source (such as a CPU or GPU chip), the temperature change trend of the heat source, and the liquid supply temperature of the liquid cooling system.

[0021] Specifically, temperature sensors can be located at the following locations: the chip surface or chip package shell, to acquire the real-time temperature of the heat source; the inlet and outlet of the liquid cooling plate, to acquire the liquid temperature before and after heat exchange; and the inlet and outlet of the cooling tower, to acquire the liquid temperature before and after cooling by the cooling tower. The sampling frequency of each temperature sensor can be set according to the control accuracy requirements, generally set to 1 to 10 times per second. The acquired temperature data is transmitted to the controller via a data bus. The controller performs preprocessing such as filtering and noise reduction on the received temperature data before using it for subsequent status determination.

[0022] Step S2: Determine the current status of the liquid cooling system based on the temperature information.

[0023] The controller comprehensively determines the current state of the liquid cooling system based on the acquired temperature information. In this embodiment, the state of the liquid cooling system can be divided into three types: low load state, normal operating state, and extreme operating state.

[0024] Specifically, the controller first determines which temperature range the real-time temperature of the heat source falls within—the normal operating temperature range, the high-load operating temperature range, or the upper limit of the operating temperature; simultaneously, it determines whether the temperature change trend of the heat source is in a rising, stable, or falling phase. Combining these two types of information with the liquid supply temperature data, the controller determines the current state of the liquid cooling system according to preset judgment logic.

[0025] Step S3: Execute the corresponding control mode according to the current state of the liquid cooling system.

[0026] When a low-load condition is detected, a control mode that reduces cooling energy consumption is implemented. For example, the pump speed can be reduced to the minimum level, or part or all of the liquid cooling system can be shut down to save energy, or the supply liquid temperature can be actively increased to reduce cooling tower energy consumption, or a combination of the above energy-saving measures can be used.

[0027] When the system is determined to be in normal operating condition, a control mode is executed to maintain or adjust the cooling capacity. For example, the current pump speed can be maintained or the pump frequency can be slightly reduced to maintain the liquid supply temperature without making large adjustments; or the liquid flow rate can be appropriately increased when the heat load increases to match the cooling capacity with the heat load.

[0028] When an extreme operating condition is detected, an enhanced cooling control mode is executed. For example, the liquid flow rate is increased to a higher level, and / or solid-state cartridge cooling is activated to actively reduce the liquid supply temperature. In addition, extreme operating conditions also include cold plate blockage. In this case, in addition to executing the enhanced cooling control mode, a maintenance alarm can be triggered to prompt maintenance personnel to check or replace the blocked cold plate.

[0029] This control scheme integrates multiple temperature information features to achieve dynamic and differentiated rapid response control of the liquid cooling system, upgrading from passive response to proactive prevention, thereby improving the reliability and cooling efficiency of the liquid cooling system.

[0030] In some embodiments, when an extreme operating condition is identified, solid-state cartridge cooling can be activated to actively reduce the liquid supply temperature, thereby reducing the energy consumption of traditional refrigeration equipment such as compressors. Solid-state cartridge cooling technology has advantages such as high energy efficiency ratio, fast response speed, and no refrigerant emissions. As an active cooling supplement in liquid cooling systems, it can effectively reduce the overall energy consumption of the system.

[0031] In some embodiments, temperature information includes one or more of the following: real-time temperature of the heat source, temperature change trend of the heat source, and liquid supply temperature of the liquid cooling system. The specific classification methods for each type of temperature information are described in detail below.

[0032] The real-time temperature of a heat source can be divided into: normal operating temperature range, high load operating temperature range, and upper limit of operating temperature.

[0033] The normal operating temperature range refers to the temperature range within which a heat source can operate stably and normally. Within this range, electronic components can operate reliably for extended periods without performance degradation. The high-load operating temperature range refers to the temperature range within which a heat source can operate for short periods, but prolonged operation within this range can accelerate component aging, shorten lifespan, and potentially lead to performance degradation and system instability. The upper operating temperature limit refers to the temperature threshold that triggers frequency throttling protection. When the heat source temperature reaches or exceeds this threshold, the system will trigger the frequency throttling protection mechanism, reducing the chip's operating frequency and voltage to decrease heat generation and prevent chip damage.

[0034] For GPUs, the normal operating temperature range is generally considered to be 30°C to 50°C (or 30°C to 65°C), the high-load operating temperature range is 65°C to 85°C, and 95°C is the upper limit of the operating temperature. Exceeding 95°C will trigger frequency reduction protection. For CPUs, the normal operating temperature range is generally considered to be 30°C to 45°C (or 30°C to 60°C), the high-load operating temperature range is 60°C to 80°C, and 90°C is the upper limit of the operating temperature. Exceeding 90°C will trigger frequency reduction protection.

[0035] In practical applications, the real-time temperature of a heat source can be obtained using temperature sensors located on the chip surface or chip package. For GPU chips, the temperature can be read using the temperature sensor integrated within the GPU via a driver; for CPU chips, it can be read using a temperature monitoring chip on the motherboard or a thermistor inside the CPU. The controller periodically reads this temperature data and compares it with preset temperature range thresholds to determine the current temperature range of the heat source.

[0036] The temperature change trend of a heat source can be divided into a temperature rising phase, a temperature stabilizing phase, and a temperature falling phase.

[0037] The temperature rise phase refers to the stage where the heat source temperature continuously increases over time. For example, when an AI training task starts, the chip load increases rapidly, and the temperature rises quickly accordingly. The temperature stabilization phase refers to the stage where the heat source temperature fluctuates within a small range and tends to remain stable overall. For example, under normal workloads, the chip temperature remains relatively stable. The temperature fall phase refers to the stage where the heat source temperature continuously decreases over time. For example, when a computing task ends and the cooling period begins, the chip load decreases, and the temperature gradually drops.

[0038] The trend of temperature change can be determined by calculating the derivative of temperature with respect to time (i.e., dT / dt). Specifically, the controller continuously collects real-time temperature data from the heat source and calculates the rate of temperature change between adjacent sampling points. When dT / dt is greater than a certain positive threshold (e.g., 0.5℃ / s), it is determined to be a temperature rising phase; when the absolute value of dT / dt is less than or equal to a certain small threshold (e.g., 0.5℃ / s), it is determined to be a temperature stabilization phase; when dT / dt is less than a certain negative threshold (e.g., -0.5℃ / s), it is determined to be a temperature falling phase.

[0039] Taking GPUs as an example, the real-time temperature trend of the GPU can serve as a core feedback signal for the dynamic control of the liquid cooling system. This allows for precise adjustment of coolant flow, pump speed, and fan speed, achieving an optimal balance between energy efficiency and heat dissipation. By monitoring the rate of temperature change and absolute temperature level in real time, the system can predict heat load trends and respond in advance, avoiding overheating and frequency throttling. Compared to a single instantaneous temperature value, the temperature change trend provides a forward-looking basis for the control of the liquid cooling system, thereby enabling an upgrade from a "passive response" to an "active prevention" control mode.

[0040] Feedforward control can be implemented by leveraging temperature trends. For example, when a GPU temperature is detected rising at an extremely rapid rate (e.g., dT / dt > 2℃ / s), even if the current temperature is still within a safe range, the system can immediately anticipate the impending high load and increase the coolant flow rate a few seconds in advance. Proactively cooling the heat source before it reaches its operating temperature limit can effectively suppress chip temperature spikes and prevent thermal throttling. This trend-based predictive control can reduce temperature fluctuations.

[0041] When dT / dt is very small (e.g., dT / dt ≤ 0.5℃ / s), it indicates a stable load. In this case, the controller's response sensitivity can be reduced, allowing the supply liquid temperature to fluctuate naturally within a small range, thereby reducing frequent pump and valve operations and achieving energy savings. When dT / dt is very large (e.g., dT / dt > 2℃ / s), it indicates a drastic load change. The system immediately switches to high-gain mode, rapidly and forcefully adjusting the pump speed or valve opening to suppress temperature rise.

[0042] Meanwhile, the temperature change trend of the heat source can also be used to determine the operating status of the liquid cooling system. For example, if the GPU temperature rises slowly but steadily under constant load, while the pressure drop across the cold plate increases, it may indicate microchannel blockage, triggering a maintenance alarm to remind you to clean or replace the cold plate. Another example is if the temperature difference (i.e., thermal resistance) between the GPU and the coolant gradually widens and this trend is irreversible, it may indicate pump performance degradation, prompting you to check or replace the pump. Furthermore, if the temperature dispersion of more than one CPU / GPU in the rack suddenly increases, and the temperature trends of some chips deviate significantly from those of other chips, it may indicate uneven flow distribution, prompting you to check the manifold, valves, or branches for foreign objects.

[0043] The liquid supply temperature of a liquid cooling system is one of the core parameters for controlling heat dissipation efficiency and system energy efficiency. By monitoring and adjusting the liquid temperature at different stages in real time, precise temperature control and energy consumption optimization can be achieved for high-heat-load devices such as GPUs. The liquid supply temperature includes the liquid temperature before heat exchange, the liquid temperature after heat exchange, and the liquid temperature after being cooled by the cooling tower.

[0044] The liquid temperature before heat exchange (i.e., the supply liquid temperature) refers to the temperature of the heat exchange fluid before it enters the liquid cooling plate and exchanges heat with the heat source. This temperature needs to be stable within a set range, generally set above the dew point temperature to prevent condensation. If the temperature is too high (e.g., above 45°C), the system can automatically increase the cooling power of the CDU (Cooling Distribution Unit) or increase the cooling tower fan speed to lower the supply liquid temperature; if the temperature is too low (e.g., below 40°C), the cooling intensity can be reduced to save energy. Specifically, if the monitored supply liquid temperature trend is upward, the pump flow rate can be increased in advance or the supply liquid temperature can be lowered to prevent the GPU temperature from approaching the throttling threshold.

[0045] The temperature of the liquid after heat exchange refers to the temperature of the heat exchange fluid after it has passed through the liquid cooler and completed heat exchange with the heat source, reflecting the actual heat generated by the GPU. A normal temperature rise is approximately 5°C to 10°C. If the temperature difference is too large (e.g., greater than 12°C), it indicates that the GPU is operating under high load, and the system may increase the pump speed, lower the liquid temperature, or trigger a warning. If the temperature difference is too small (e.g., less than 3°C), it may indicate insufficient flow or poor contact of the cooler plate.

[0046] The temperature of the liquid after cooling by the cooling tower represents the cooling effect of the primary side return water and should be slightly higher than the ambient wet-bulb temperature (usually 3°C to 5°C higher). If the temperature is too high after cooling, it indicates that the cooling tower efficiency has decreased, and the system can start the solid-state cassette cooling or cleaning spray system.

[0047] It should be noted that the difference between the temperature after heat exchange and the temperature before cooling is as follows: the temperature after heat exchange is the temperature of a single heat source after heat exchange, and the temperature after heat exchange may be different for different heat sources; the temperature before cooling is the temperature of the liquid when it flows into the cooling tower after different heat exchanges, and generally, one cooling tower corresponds to one temperature before cooling. After the liquid is cooled in the cooling tower, the temperature of the liquid that is ready to exchange heat with the heat source again is the temperature before heat exchange.

[0048] In some embodiments, the criteria for determining a low-load state include: the real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is a decreasing phase.

[0049] Specifically, the controller acquires the real-time temperature value of the heat source and determines whether the temperature falls within the normal operating temperature range (e.g., for a GPU, the temperature is between 30°C and 65°C); it also calculates the rate of temperature change dT / dt and determines whether it is negative (i.e., the temperature is decreasing). When both of these conditions are met, the liquid cooling system is determined to be in a low-load state.

[0050] In real-world scenarios, low-load conditions typically occur after a computing task has finished or during device standby. At this time, the chip generates relatively little heat, and full-power cooling is sufficient to maintain its temperature within a safe range.

[0051] When a low-load condition is detected, a control mode that reduces cooling energy consumption is executed. This may include one or more combinations of the following methods: Reduce the pump speed to the minimum level. The controller reduces the speed of the circulating pump via the frequency converter, thereby reducing the circulation flow rate of the heat exchange fluid. Since the power consumption of the pump is proportional to the cube of the speed, reducing the pump speed can significantly reduce the pump's energy consumption. For example, reducing the pump speed to 30% to 50% of the rated speed can reduce the pump power consumption to 2.7% to 12.5% ​​of the rated power consumption.

[0052] To save energy, some or all of the liquid cooling system can be shut down. Under low load conditions, if the heat source generates very little heat (such as when the equipment is in hibernation or standby mode), some of the liquid cooling system (such as shutting down some cooling towers or CDUs), or even the entire liquid cooling system, can be shut down, relying on heat capacity or natural heat dissipation to maintain the system temperature. Of course, this method must be carried out under the premise of ensuring that the heat source temperature does not exceed the safe threshold.

[0053] Actively increase the supply liquid temperature to reduce cooling tower energy consumption. The controller can appropriately increase the set value of the supply liquid temperature to reduce the heat dissipation load of the cooling tower. For example, under low load conditions, increasing the supply liquid temperature from 40°C to 45°C will correspondingly reduce the cooling tower fan speed, thereby reducing the cooling tower's energy consumption.

[0054] By using the energy-saving control modes described above, the energy consumption of the liquid cooling system can be reduced to the maximum extent under low load conditions, thereby improving the energy efficiency of the data center.

[0055] In some embodiments, the criteria for determining normal working status include the following: The first scenario: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is in a stable phase. At this time, the chip is under a stable normal workload, and the heat generation and cooling are basically balanced.

[0056] The second scenario: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend is in the rising phase, but the rate of temperature change is less than or equal to a preset threshold (e.g., dT / dt ≤ 2℃ / s). In this case, although the load increases, the temperature rise is relatively gradual and falls within the normal load fluctuation range.

[0057] The third scenario: The real-time temperature of the heat source is within the high-load operating temperature range, and the temperature change trend of the heat source is a decreasing phase. In this case, the chip is under high load but is cooling down, indicating that the cooling measures taken are effective, and the system is transitioning from a high-load state to a normal state.

[0058] The fourth scenario: The real-time temperature of the heat source is within the high-load operating temperature range, and the temperature change trend of the heat source is in a stable phase. In this case, although the chip is in the high-load operating temperature range, the temperature has stabilized, and the system is in an acceptable steady-state operating state.

[0059] When the system is determined to be in normal operating condition, a control mode is executed to maintain or adjust the cooling capacity. Specifically, this includes the following adjustment methods: When the real-time temperature of the heat source is within the normal operating temperature range and the temperature trend is in a stable phase, maintain the current pump speed or slightly reduce the pump frequency to keep the liquid supply temperature constant without making major adjustments. At this time, the system is in a relatively ideal equilibrium state, and there is no need to make significant adjustments to the control parameters to avoid unnecessary oscillations and energy consumption fluctuations.

[0060] When the real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend is in the rising phase with a rate of change less than or equal to a preset threshold, the liquid flow rate should be appropriately increased. For example, the controller can increase the pump speed by 5% to 15%, or increase the valve opening by a corresponding percentage, to increase the flow rate of the heat exchange fluid passing through the liquid cooling plate, remove more heat, and suppress the rising temperature trend.

[0061] When the real-time temperature of the heat source is in the high-load operating temperature range and the temperature trend is downward, maintain the current pump speed and supply temperature without making significant adjustments. At this time, although the chip temperature is in the high-load range, it is decreasing, indicating that the current cooling strategy is effective and no additional intervention is needed.

[0062] When the real-time temperature of the heat source is within the high-load operating temperature range and the temperature trend is in a stable phase, increase the liquid flow rate. At this time, the chip is in a high-load steady state, requiring enhanced cooling capacity to prevent further temperature increases. The controller can increase the pump speed by 10% to 30%, or use other auxiliary cooling methods (such as increasing the cooling tower fan speed) to ensure that the chip temperature does not exceed the upper limit of the operating temperature.

[0063] In some embodiments, the criteria for determining extreme working conditions include the following: The first scenario: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend is an upward phase, with the rate of temperature change exceeding a preset threshold (e.g., dT / dt > 2℃ / s). Although the current temperature is still within a safe range, the temperature rises extremely rapidly, indicating an imminent high-load condition or even potential thermal throttling, which constitutes an emergency.

[0064] The second scenario: The heat source's real-time temperature is within the high-load operating temperature range, and the temperature trend is upward. In this case, the chip is already under high load and the temperature is still rising, approaching the upper limit of its operating temperature, creating an emergency situation.

[0065] In addition, extreme operating conditions also include cold plate blockage. Cold plate blockage leads to decreased heat exchange efficiency, and even when the cooling system is operating normally, the heat source temperature may rise abnormally. Diagnostic methods for cold plate blockage are detailed in subsequent embodiments.

[0066] When an extreme operating condition is detected, an enhanced cooling control mode is activated. This includes increasing the liquid flow rate and / or activating solid-state cartridge cooling.

[0067] For the first scenario (temperature within the normal operating range but dT / dt > 2℃ / s), the corresponding control mode is: increase the liquid flow rate and activate the solid-state cartridge cooling system. The controller will increase the pump speed to a higher level (e.g., 80% to 100% of the rated speed) and activate the solid-state cartridge cooling system to actively reduce the supply liquid temperature, suppressing the temperature rise trend with the fastest response speed.

[0068] For the second scenario (temperature is in the high-load operating temperature range and is trending upwards), the corresponding control mode is as follows: Maximize the liquid flow rate and run the solid-state cartridge cooling system at full speed. The controller increases the pump speed to 100% of its rated speed and simultaneously runs the solid-state cartridge cooling system at full speed to counteract the temperature rise with maximum cooling capacity, ensuring that the chip temperature does not exceed the upper limit of the operating temperature.

[0069] In some embodiments, solid-state cassette cooling can be used to reduce the supply liquid temperature. Solid-state cassette cooling refers to a solid-state refrigeration technology that generates a cooling effect through a force-driven martensitic phase transformation of a shape memory alloy. When an axial force (tension or compression) is applied externally, the internal stress of the shape memory alloy exceeds its critical phase transformation stress, causing the shape memory alloy to transform from austenite to martensite, releasing latent heat of phase transformation, resulting in a temperature rise in the material itself under adiabatic conditions. After the external force is removed, the internal stress of the shape memory alloy is less than its critical phase transformation stress, causing the shape memory alloy to transform back from martensite to austenite, absorbing latent heat, resulting in a temperature drop in the material itself. The reverse phase transformation from martensite to austenite and the absorption of latent heat constitute the cassette cooling effect.

[0070] Such liquid cooling systems typically require two or more cooling towers. One cooling tower (which can be called the high-temperature cooling tower) stores the liquid, which has been heated to a higher temperature by the solid-state cooling device, while the other cooling tower (which can be called the low-temperature cooling tower) stores the relatively low-temperature liquid. The most direct way to reduce the supply liquid temperature is to cool the liquid before heat exchange using the solid-state cooling device: after being cooled in the low-temperature cooling tower, the liquid enters the solid-state cooling device for further cooling and then exchanges heat with the heat source; while the liquid, after being heated in the solid-state cooling device, is then transported to the high-temperature cooling tower. Of course, in other embodiments, the solid-state cooling device can also cool the liquid before or after heat exchange.

[0071] In some embodiments, extreme operating conditions also include cold plate blockage. Cold plate blockage leads to a decrease in the flow rate of the heat exchange fluid passing through the cold plate, resulting in reduced heat exchange efficiency and an abnormal increase in chip temperature at the same chip power consumption. This embodiment provides a method for diagnosing cold plate blockage based on the temperature difference ΔT after heat exchange.

[0072] The diagnostic principle of cold plate blockage is as follows: Under the same chip power consumption, cold plate blockage will cause a decrease in flow rate, a longer residence time of coolant in the plate, and more heat absorption, thereby increasing the outlet temperature and causing the heat exchange temperature difference ΔT = outlet temperature - inlet temperature to increase abnormally.

[0073] The specific diagnostic steps are as follows: The first step is to establish a normal operation model.

[0074] Under conditions of no blockage in the liquid cooling system and normal flow rate, obtain the ΔT values ​​of the chip under different loads. Specifically, run the chip under different loads for a sufficient period (e.g., 10 to 30 minutes at each load) to ensure temperature stability. Record the chip power consumption P and the corresponding heat exchange temperature difference ΔT at each load level.

[0075] Theoretically, ΔT has an approximately linear relationship with the chip power consumption P: P = m × cp × ΔT, where m is the mass flow rate and cp is the specific heat capacity of the coolant. Keeping the flow rates m and cp constant, ΔT increases linearly with P. The real-time chip power consumption P can be read through the server BMC or system management interface.

[0076] The collected data points (ΔT, P) are plotted as a scatter plot and fitted with a straight line or curve. This line is the normal baseline of ΔT, and ΔT is the expected temperature difference ΔTexpected.

[0077] The second step is to compare the real-time data with the model data.

[0078] During system operation, the current chip power consumption Pcurrent and the measured temperature difference ΔTmeasured are read in real time and compared with the expected temperature difference ΔTexpected corresponding to the same power consumption.

[0079] Among them, there can be two comparison indicators: absolute temperature difference deviation value and relative deviation rate.

[0080] The absolute temperature difference deviation is calculated as ΔTmeasured - ΔTexpected. A positive value indicates a decrease in heat exchange efficiency. Different levels of warnings can be set based on the magnitude of the deviation: minor warning: deviation > 2℃; severe fault: deviation > 5℃.

[0081] Relative deviation rate = absolute temperature difference deviation / ΔTmeasured × 100%. This indicator is normalized and compared after eliminating the influence of power consumption. A deviation consistently >10% to 15% can be considered a blockage warning.

[0082] For example: Chip power consumption P = 1400W, expected temperature difference ΔT = 10℃, actual temperature difference ΔT = 12.5℃. In this case: Absolute temperature difference deviation = 12.5 - 10.0 = 2.5℃ (exceeding the mild warning threshold of 2℃); Relative deviation rate = 2.5 / 10 × 100% = 25%, exceeding the threshold of 10% to 15%, which can be judged as a blockage anomaly.

[0083] The third step is to conduct trend analysis.

[0084] A single, short-term deviation may be caused by other factors such as measurement noise or instantaneous load fluctuations, so the dynamic trend of temperature difference should also be considered. Therefore, the rate of change of temperature difference is introduced: R_ΔT=d(ΔT) / dt.

[0085] When R_ΔT remains positive (i.e., ΔT increases slowly over time), even if the absolute value has not yet triggered an alarm, it indicates that blockage is slowly developing. Set a trend alarm: when R_ΔT > 0.1℃ / day and persists for more than 3 days, it generally indicates a slow decline in the performance of the cold plate, and maintenance or replacement is recommended. This can provide an earlier warning of blockage.

[0086] The fourth step is cross-validation comparison.

[0087] In a liquid cooling distribution unit, multiple cold plates are typically connected in parallel. If only one cold plate has a significantly higher ΔT than other cold plates of the same model GPU in the same rack, while the other cold plates are normal, it is highly likely that the cold plate is blocked, rather than due to load or environmental factors. Conversely, if the ΔT of all cold plates increases proportionally, the problem is likely in the upstream main piping or CDU, rather than a single cold plate.

[0088] As an alternative implementation, a temperature difference monitoring scheme based on NTC thermistors can also be used. High-precision NTC thermistor probes are installed on the inlet and outlet pipes of the cold plate, respectively, to monitor the inlet and outlet temperatures in real time, calculate the real-time temperature difference, and compare it with a reference temperature difference. NTC thermistors have the advantages of low cost, high accuracy, and fast response, making them suitable for large-scale deployment in data centers.

[0089] This embodiment provides a liquid cooling plate, which can be applied in the control method of the above-mentioned liquid cooling system as a heat exchange component that is in direct contact with the heat source.

[0090] A liquid cooling plate consists of a plate body. As the basic load-bearing structure of the liquid cooling plate, the plate body is usually made of a metal material with good thermal conductivity, such as copper, aluminum, or their alloys, so as to facilitate the rapid transfer of heat generated by the heat source to the heat exchange fluid inside the plate body.

[0091] The plate has an inlet and an outlet. The inlet and outlet are typically located on the same side or opposite sides of the plate. The inlet connects to an external supply line, introducing the heat exchange fluid from the coolant distribution device into the liquid-cooled plate; the outlet connects to an external return line, discharging the heat exchange fluid after heat exchange. Pipe fittings or sealing structures can be installed at the inlet and outlet ports to ensure reliable connection and sealing performance with external piping.

[0092] A heat exchange channel exists between the inlet and outlet. This channel is a fluid passageway located inside the plate, allowing the heat exchange fluid to flow between the inlet and outlet. The cross-sectional shape of the heat exchange channel can be circular, rectangular, elliptical, or other irregular shapes. A heat exchange fluid capable of exchanging heat with the heat source flows within the heat exchange channel. This heat exchange fluid is typically deionized water, a water-ethylene glycol mixture, or a dielectric coolant, possessing good thermal conductivity and chemical stability.

[0093] A throttling valve is installed inside the heat exchange channel to regulate the flow rate of the heat exchange fluid, thereby controlling the cooling capacity of the liquid cooling plate. Because the throttling valve is located inside the heat exchange channel and in direct contact with the heat exchange fluid, it can detect real-time temperature changes in the fluid.

[0094] Throttling valves include shape memory alloy components. Shape memory alloy components are made of alloy materials with shape memory effects, such as nickel-titanium alloys, copper-based alloys (e.g., copper-zinc-aluminum, copper-aluminum-nickel), or iron-based alloys. Shape memory alloys are special metallic materials that undergo reversible phase transformations between a low-temperature phase (martensitic phase) and a high-temperature phase (austenitic phase), accompanied by shape changes.

[0095] Specifically, shape memory alloy parts exhibit a martensitic phase at lower temperatures; when the temperature rises to the austenitic phase transformation initiation temperature, the alloy begins to transform from martensite to austenite, accompanied by shape recovery; when the temperature rises to the austenitic phase transformation completion temperature, the alloy completely transforms into austenite, and the shape recovers to the preset high-temperature shape. When the temperature drops again, the alloy can transform from austenite to martensite, and the shape changes again.

[0096] The shape memory alloy component switches between a first state and a second state based on the temperature change of the heat exchange fluid. In this embodiment, the shape memory alloy component is designed to have two shape states: the first state corresponds to a lower temperature, and the second state corresponds to a higher temperature. When the heat exchange fluid temperature is below the phase transition temperature range of the shape memory alloy component, the component is in the first state; when the heat exchange fluid temperature rises above the phase transition temperature range, the component switches to the second state. Conversely, when the heat exchange fluid temperature decreases again, the component can return from the second state to the first state.

[0097] Specifically, when the shape memory alloy component is in the first state and the second state, the flow rate of the heat exchange fluid passing through the heat exchange channel per unit time is different. Specifically, when the shape memory alloy component is in the first state, it presents a first configuration in the heat exchange channel, which has a large flow resistance to the heat exchange fluid, resulting in a small flow rate of the heat exchange fluid passing through the heat exchange channel per unit time. When the shape memory alloy component is in the second state, it presents a second configuration in the heat exchange channel, which has a small flow resistance to the heat exchange fluid, resulting in a large flow rate of the heat exchange fluid passing through the heat exchange channel per unit time.

[0098] In this embodiment, when the liquid cooling plate is working, the heat exchange fluid enters the heat exchange channel through the inlet. As it flows through the heat exchange area corresponding to the heat source, it absorbs heat generated by the heat source, and its temperature gradually increases. When the heat exchange fluid passes through the location of the throttling valve, its temperature is directly transferred to the shape memory alloy component. When the heat source load is high, the heat exchange fluid absorbs more heat, and its temperature rises above the phase change temperature. The shape memory alloy component switches from a first state to a second state, increasing the flow cross-section of the heat exchange channel and allowing more heat exchange fluid to pass through, thereby enhancing the cooling effect. When the heat source load is low, the heat exchange fluid temperature is low, and the shape memory alloy component remains in the first state, reducing the flow cross-section of the heat exchange channel and decreasing the heat exchange fluid flow rate to avoid overcooling. Therefore, the liquid cooling plate can automatically adjust its flow rate according to the actual heat load of the heat source, achieving on-demand cooling.

[0099] The aforementioned liquid cooling plate structure can be used in conjunction with the control method described above. In a data center employing these liquid cooling plates, the shape memory alloy components within each liquid cooling plate can autonomously adjust the flow capacity of their internal channels according to their respective heat source load and fluid temperature changes, thus achieving a more balanced flow distribution among the liquid cooling plates. Because each liquid cooling plate possesses independent adaptive flow adjustment capabilities, it effectively avoids near-end overcooling and far-end overheating problems caused by pipeline pressure differences, thereby improving the temperature uniformity of server chips throughout the entire data center.

[0100] In some embodiments, the throttling valve is located on the side of the heat exchange channel near the liquid outlet. By positioning the throttling valve near the liquid outlet, the temperature of the heat exchange fluid after flowing through the main heat exchange area of ​​the heat exchange channel reflects the actual heating state of the heat source. At this point, flow regulation near the liquid outlet allows the temperature sensed by the shape memory alloy component to represent the overall heat load level of the liquid cooling plate. Furthermore, positioning the throttling valve near the liquid outlet avoids flow interference to the heat exchange fluid flowing through the heat exchange area, ensuring a uniform and stable flow state within the heat exchange area, which is beneficial for improving heat exchange efficiency.

[0101] In some embodiments, the heat exchange channel is sequentially connected to an inlet channel, a middle channel, and an outlet channel in the direction of heat exchange fluid flow. The inlet channel connects to the liquid inlet, the middle channel is positioned corresponding to the heat source location, and the outlet channel connects to the liquid outlet. A throttling valve is located in the outlet channel. By placing the throttling valve in the outlet channel, the heat exchange fluid flows through the throttling valve only after completing heat exchange with the heat source. At this point, the temperature of the heat exchange fluid already reflects the actual heat load of the heat source. The shape memory alloy component can more accurately sense the temperature change of the heat exchange fluid, thereby enabling more precise flow regulation.

[0102] In some embodiments, the central channel is provided with multiple heat exchange fins. The heat exchange fins are sheet-like structures protruding from the inner wall of the plate, and their material is the same as the plate, both being metal materials with good thermal conductivity. The heat exchange fins increase the contact area between the inner wall of the heat exchange channel and the heat exchange fluid, allowing more heat to be transferred to the heat exchange fluid through the fins, thus improving heat exchange efficiency. The heat exchange fins can also disrupt the flow boundary layer of the heat exchange fluid within the central channel, enhancing turbulence and promoting heat exchange. A flow gap is formed between two adjacent heat exchange fins to allow the heat exchange fluid to pass through.

[0103] In some embodiments, the inlet channel has a fluid cavity containing a turbulence structure. The turbulence structure is used to create turbulence in the heat exchange fluid flowing through the fluid cavity. When the heat exchange fluid flows into the fluid cavity at high speed from the inlet, the turbulence structure disrupts the orderly flow of the fluid, generating eddies and disturbances. This allows the heat exchange fluid to be fully mixed within the fluid cavity before entering the central channel, resulting in a more uniform flow rate and temperature distribution. This avoids the problem of uneven flow distribution within the central channel caused by the offset position of the inlet. The turbulence structure includes several turbulence columns arranged longitudinally within the fluid cavity.

[0104] In some embodiments, the shape memory alloy (MME) component has multiple hollowed-out protrusions along its thickness direction, and the hollowed-out protrusions have flow-through holes along the length and / or width direction of the MME component. When the MME component is in a first state, the flow-through holes are arranged at an angle to the flow direction of the heat exchange fluid; when the MME component is in a second state, the flow-through holes are in the same direction as the flow direction of the heat exchange fluid. By providing a MME component with hollowed-out protrusions and flow-through holes, the MME component can switch states without significantly changing its overall volume occupied in the heat exchange channel; the flow rate can be adjusted simply by changing the direction of the flow-through holes, resulting in a compact structure and rapid response.

[0105] In some embodiments, a flow guide is provided between the liquid inlet and the heat exchange channel to guide the heat exchange fluid to enter the heat exchange channel uniformly. The flow guide can be a flow guide plate or a flow guide cone, which disperses and guides the heat exchange fluid flowing in from the liquid inlet to different areas of the heat exchange channel in multiple directions, avoiding the heat exchange fluid from concentrating and rushing into the heat exchange channel at high speed from the direction directly opposite the liquid inlet.

[0106] This embodiment provides another liquid cooling plate, which can also be applied to the control method of the above-mentioned liquid cooling system.

[0107] The liquid cooling plate consists of a plate body and an integrated interface. The plate body is generally flat and plate-shaped, made of a metal material with good thermal conductivity, such as copper or aluminum alloy. The plate body is hollow inside and has heat exchange channels for the flow of heat exchange fluid. Heat exchange sections are located within the heat exchange channels to increase the contact area with the heat exchange fluid and improve heat exchange efficiency. The heat exchange fluid flows within the heat exchange channels for heat exchange with the heat exchange sections.

[0108] The integrated interface is located on the same edge of the plate and has a columnar or block-shaped structure that protrudes outward from the edge of the plate. The integrated interface has an interface channel inside, and the interface channel has a partition that divides the internal space of the interface channel into two independent and sealed chambers, namely the liquid inlet channel and the liquid outlet channel.

[0109] The inlet and outlet channels are connected to the two ends of the heat exchange channel, respectively. That is, the inlet channel is connected to the inlet end of the heat exchange channel, and the outlet channel is connected to the outlet end of the heat exchange channel. The partition can be integrally formed with the integrated interface, for example, by casting or machining, or it can be an independent partition fixed in the interface channel by plugging or welding.

[0110] The heat exchange fluid enters the inlet channel of the integrated interface from the external pipeline, and then flows into the heat exchange channel inside the plate. During the flow through the heat exchange channel, it exchanges heat with the heat exchange part and the inner wall of the plate. After absorbing heat, the temperature rises, and then it flows out of the integrated interface through the outlet channel and flows back to the external cooling system for circulation cooling.

[0111] Through the integrated interface design described above, the inlet and outlet channels are combined into a single interface, replacing the two separate inlet and outlet interfaces found in traditional technology. This design significantly reduces the space occupied by the external interfaces of the cold plate, making the overall structure of the cold plate more compact. Simultaneously, external liquid cooling pipelines only need to be centrally connected at the integrated interface location, avoiding pipeline crossings and detours caused by separate inlet and outlet interfaces, simplifying the pipeline layout, reducing the complexity of the pipeline system, and facilitating installation and maintenance.

[0112] In some embodiments, the inlet of the inlet channel and the outlet of the outlet channel open in the same direction. That is, both the inlet and outlet face the same side of the integrated interface and are arranged side by side on the same end face of the integrated interface. External pipelines can be connected to both the inlet and outlet simultaneously through an integrated dual-pipe connector, enabling quick insertion and removal.

[0113] As an alternative implementation, the inlet direction of the liquid inlet channel and the outlet direction of the liquid outlet channel can also be opposite. That is, the inlet and outlet open towards opposite sides of the integrated interface along the axial direction. This is suitable for scenarios where pipes need to be routed on both sides of the cold plate.

[0114] In some embodiments, the heat exchange channel is sequentially connected to a first channel, a second channel, and a third channel in the direction of heat exchange fluid flow. The heat exchange section is located within the second channel. Through the three-section channel structure, after the heat exchange fluid enters from the inlet channel, it first passes through the first channel for flow direction guidance and flow distribution, allowing the fluid to enter the second channel more evenly and exchange heat with the heat exchange section. After heat exchange, the fluid is concentrated and guided to the outlet channel in the third channel, improving the overall heat exchange efficiency.

[0115] In some embodiments, guide channels are provided in the first channel and / or the third channel. The guide channels in the first channel are used to guide the heat exchange fluid to be evenly distributed from the inlet channel to the heat exchange section. The guide channels in the third channel are used to guide the heat exchange fluid to be concentrated from the heat exchange section to the outlet channel.

[0116] In some embodiments, a guide channel is provided between the liquid inlet channel and the heat exchange channel to guide the heat exchange fluid after passing through the liquid inlet channel toward the high-temperature region of the heat source corresponding to the heat exchange channel. When the heat exchange fluid flows out from the liquid inlet channel through the guide channel, under the guidance of the guide channel, the fluid impacts the inner wall surface of the plate corresponding to the high-temperature region of the heat source vertically or at a large angle, forming an impact jet heat exchange, which quickly removes the heat from the high-temperature region.

[0117] In some embodiments, the heat exchange section includes a plurality of heat exchange elements arranged at intervals, with a heat exchange gap formed between adjacent heat exchange elements for the passage of heat exchange fluid. The heat exchange elements can be finned structures with rectangular cross-sections, or columnar structures with circular, elliptical, or rhomboid cross-sections. In one specific embodiment, the heat exchange element is a finned structure with a thickness of 0.1 mm to 0.2 mm, preferably 0.13 mm. The width of the heat exchange gap between two adjacent finned structures is 0.1 mm to 0.3 mm, preferably 0.13 mm. The height of the finned structure is 2 mm to 5 mm, preferably 2.6 mm.

[0118] In some embodiments, the heat exchanger includes a plurality of first heat exchange fins spaced apart, with a flow splitting gap formed between adjacent first heat exchange fins. Second heat exchange fins are disposed within the flow splitting gap, and the first and second heat exchange fins are arranged at an angle. The first heat exchange fins are arranged parallel to the main flow direction of the heat exchange fluid, and the length direction of the second heat exchange fins forms a certain angle with the length direction of the first heat exchange fins, which can be from 15 degrees to 90 degrees, preferably 90 degrees. Through the staggered arrangement of the first and second heat exchange fins, multiple changes in flow direction and fluid splitting and merging are created within the heat exchange gap, increasing the turbulence and disturbance effect of the fluid, and enhancing the convective heat transfer efficiency.

[0119] In some embodiments, a flow-diverting fin is provided at the end of the heat exchanger near the inflow of the heat exchange fluid. The flow-diverting fin is wedge-shaped or streamlined, with its tip facing the direction of the incoming flow of the heat exchange fluid and its tail facing the interior of the heat exchange gap, for guiding the heat exchange fluid to flow into the heat exchange gaps on both sides of the corresponding heat exchanger.

[0120] In some embodiments, the heat exchange section includes a high-density region and a low-density region. The high-density region corresponds to the high-heat location of the heat source, and the low-density region corresponds to the low-heat location of the heat source. The height of the heat exchanger in the high-density region is greater than that in the low-density region, and the width of the heat exchange gap in the high-density region is smaller than that in the low-density region. Through this differentiated heat exchange section design, the cold plate can precisely dissipate heat to address the non-uniform heating characteristics of the chip, reducing the temperature difference between the high-temperature and low-temperature regions and improving the chip's operational reliability.

[0121] This embodiment provides a data center, including the liquid-cooled plates described in the above embodiments. Specifically, the data center includes a server rack, multiple server nodes disposed within the server rack, and a cooling system for dissipating heat from the server nodes. The cooling system includes a coolant distribution device, a supply pipeline, a return pipeline, and multiple liquid-cooled plates. Each liquid-cooled plate is respectively disposed on a heat-generating chip in each server node, for direct contact liquid cooling of the heat-generating chip.

[0122] The coolant distribution device connects to the inlet of each liquid-cooled plate via a supply line, and the outlet of each liquid-cooled plate connects to the coolant distribution device via a return line, forming a circulating cooling loop. The coolant distribution device pumps the cooled heat exchange fluid to each liquid-cooled plate via the supply line. After absorbing heat within each liquid-cooled plate, the temperature of the heat exchange fluid rises, and it returns to the coolant distribution device via the return line for secondary cooling, thus completing the cycle.

[0123] In this data center, the control method of the liquid cooling system is as described in the above embodiment. By acquiring the temperature information of the liquid cooling system, the current state of the liquid cooling system is determined based on the temperature information, and the corresponding control mode is executed according to the state.

[0124] In data centers employing the aforementioned liquid cooling plates, the shape memory alloy components within each plate can autonomously adjust the flow capacity of their internal channels according to their respective heat source load and fluid temperature changes, resulting in a more balanced flow distribution among the plates. Simultaneously, the integrated interface design allows each liquid cooling plate to connect to external inlet and outlet pipes via only one integrated interface, leading to a clearer and more organized pipe layout. This saves space occupied by pipes within server nodes and on the back panel of the server rack, facilitating high-density server deployment.

[0125] Compared to traditional data centers, this data center effectively avoids the problem of near-end overcooling and far-end overheating caused by pipeline pressure differences, improves the temperature uniformity of server chips throughout the entire data center, reduces the risk of localized hotspots, and ensures reliable heat dissipation in high-power-density scenarios. Simultaneously, this adjustment method requires no external control signals or additional power, does not increase the control complexity or energy consumption of the data center, and has the advantages of simple structure, low cost, and high reliability. Integrated interfaces reduce the number of sealing points, lower the risk of coolant leakage, and ensure the safe and stable operation of electronic equipment within the data center.

[0126] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A control method for a liquid cooling system, characterized in that, include: Obtain temperature information from the liquid cooling system; The current operating state of the liquid cooling system is determined based on the temperature information. The operating state includes low load state, normal operating state, or extreme operating state. When the load condition is determined to be low, a control mode that reduces cooling energy consumption is executed. When the system is determined to be in normal working condition, a control mode is executed to maintain or adjust the cooling capacity. When the operating condition is determined to be extreme, the enhanced cooling control mode is executed.

2. The control method for a liquid cooling system as described in claim 1, characterized in that, The temperature information includes one or more of the following: the real-time temperature of the heat source, the temperature change trend of the heat source, and the liquid supply temperature of the liquid cooling system.

3. The control method for a liquid cooling system as described in claim 2, characterized in that, The real-time temperature of the heat source is divided into a normal operating temperature range, a high-load operating temperature range, and an upper limit of the operating temperature. The temperature change trend of the heat source is divided into a temperature rising phase, a temperature stabilizing phase, and a temperature falling phase. The liquid supply temperature of the liquid cooling system includes the liquid temperature before heat exchange, the liquid temperature after heat exchange, and the liquid temperature after cooling by the cooling tower.

4. The control method for a liquid cooling system as described in claim 3, characterized in that, The criteria for determining the low load state include: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is a decreasing phase.

5. The control method for a liquid cooling system as described in claim 3, characterized in that, The conditions for determining the normal working state include: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is in a stable phase; or, The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is in the rising phase, but the rate of temperature change is less than or equal to a preset threshold; or, The real-time temperature of the heat source is within the high-load operating temperature range, and the temperature change trend of the heat source is a decreasing phase; or, The real-time temperature of the heat source is in the high-load operating temperature range, and the temperature change trend of the heat source is in a stable phase.

6. The control method for a liquid cooling system as described in claim 3, characterized in that, The criteria for determining the extreme working state include: The real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend of the heat source is in an upward phase, with the rate of temperature change exceeding a preset threshold; or, The real-time temperature of the heat source is in the high-load operating temperature range, and the temperature change trend of the heat source is in the rising stage.

7. The control method for a liquid cooling system as described in claim 1, characterized in that, The control mode for reducing cooling energy consumption includes: Reduce the pump speed; or, Shut down part or all of the liquid cooling system; or Increase the liquid supply temperature.

8. The control method for a liquid cooling system as described in claim 1, characterized in that, The control modes for maintaining or regulating cooling capacity include: When the real-time temperature of the heat source is within the normal operating temperature range and the temperature change trend is in the stable phase, maintain the current pump speed or reduce the pump frequency to maintain the liquid supply temperature. When the real-time temperature of the heat source is within the normal operating temperature range, and the temperature change trend is in the temperature rising stage and the temperature change rate is less than or equal to a preset threshold, the liquid flow rate is increased. When the real-time temperature of the heat source is in the high-load operating temperature range and the temperature change trend is in the temperature decrease stage, maintain the current pump speed and maintain the liquid supply temperature. When the real-time temperature of the heat source is in the high-load operating temperature range and the temperature change trend is in the stable stage, increase the liquid flow rate.

9. The control method for a liquid cooling system as described in claim 1, characterized in that, The control mode for implementing enhanced cooling includes: Increase liquid flow and / or activate solid-state cartridge cooling.

10. The control method for a liquid cooling system as described in claim 1, characterized in that, The extreme operating conditions also include cold plate blockage, which is diagnosed by the temperature difference ΔT after heat exchange.