Predictive immersion heat dissipation system and method based on real-time power consumption of data center

Abstract

The invention relates to the technical field of immersed heat dissipation, in particular to a predictive immersed heat dissipation system and method based on real-time power consumption of a data center, and the system can obtain the power consumption data of a data center server and the temperature distribution of multiple regions through additionally arranging a data collection unit on the structure basis of an existing immersed heat dissipation system. Compared with fixed cooling logic or manual temperature adjustment, the method has the advantages that the automation degree is higher, the temperature adjustment is more accurate, and the heat dissipation efficiency can be greatly improved by arranging the heat dissipation system comprising the backflow type heat exchange unit and the evaporation type heat exchange unit through cooperation of the backflow type heat exchange unit and the evaporation type heat exchange unit. On one hand, the backflow type heat exchange unit takes away heat in cooling liquid to enable the cooling liquid to conduct heat exchange with the heat dissipation unit, the cooling liquid returns to the box body again after cooling is completed, and on the other hand, the evaporation type heat exchange unit cools the data center server, shortens the heat transfer path, accelerates heat dissipation and guarantees stable operation of the server.

Description

Technical Field

[0001] The present invention relates to the technical field of immersion heat dissipation, and in particular to a predictive immersion heat dissipation system and method based on real-time power consumption of a data center. Background Art

[0002] In today's digital age, especially with the development of AI technology, the scale of data centers has exploded, and the computing power of servers has continued to increase. However, the heat dissipation problem that comes with it has become increasingly prominent, becoming a key bottleneck restricting the development of data centers.

[0003] Traditional air cooling can no longer meet the cooling needs of high-density servers due to its low cooling efficiency. Liquid cooling technology has become the mainstream choice in the market due to its higher efficiency and lower energy consumption.

[0004] There are two main types of liquid cooling, one is indirect contact liquid cooling such as cold plate liquid cooling, and the other is direct contact liquid cooling such as immersion liquid cooling. Cold plate liquid cooling appeared earlier and has obvious advantages in heat transfer efficiency and cost compared to air cooling, but its disadvantages are also obvious, mainly in the following aspects:

[0005] 1. The installation is complex, requiring hardware modification or even disassembly of the entire system for installation;

[0006] 2. There is a risk of coolant leakage, which will damage electronic components once leaked. The manufacturing process requirements are high and the yield rate fluctuates greatly.

[0007] Immersion liquid cooling mainly relies on a unique fluid medium coolant, which can directly contact the hardware of the data center server, such as the CPU, GPU and other areas where heat is easily accumulated, and quickly take away the heat to achieve heat dissipation and cooling.

[0008] Common coolants can be pure water, deionized water, fluorinated liquid, synthetic oil or two-phase coolant. These coolants can quickly take away heat for cooling and also avoid circuit short circuits based on their own characteristics.

[0009] A Chinese patent with announcement number CN217957563U discloses a server immersion cooling oil tank, cooling system and heating system, which cools the server through a single oil immersion method. This cooling method mainly relies on the oil pump to control the rapid flow of oil to remove heat. The overall heat dissipation efficiency is limited, which is not ideal for the heat dissipation of data center servers for AI technology research and development. Summary of the invention

[0010] The purpose of the present invention is to provide a predictive immersion cooling system and method based on real-time power consumption of a data center in order to solve the above-mentioned problems.

[0011] The technical solution adopted by the present invention is as follows. A predictive immersion cooling system based on the real-time power consumption of a data center includes a box body and a data center server. A cavity is provided inside the box body, and a coolant is injected into the cavity. The data center server is placed in the cavity and immersed in the coolant. It further includes a reflux heat exchange unit, an evaporative heat exchange unit, a controller, a heat dissipation unit, and a data acquisition unit;

[0012] Both the reflux heat exchange unit and the evaporative heat exchange unit can perform heat exchange on the data center server, and the reflux heat exchange unit and the evaporative heat exchange unit are connected to the heat dissipation unit;

[0013] The heat dissipation unit is provided outside the box body;

[0014] The data acquisition unit can collect the power consumption data of the data center server and the multi-region temperature data of the data center server, and can transmit the power consumption data and the temperature data to the controller;

[0015] The controller can control the heat dissipation unit, the reflux heat exchange unit, and the evaporative heat exchange unit.

[0016] Optionally, the data acquisition unit includes a power consumption acquisition component and multiple temperature sensors;

[0017] The power consumption acquisition component is signal-connected to the data center server and can collect the real-time power consumption of the GPU and / or CPU in the data center server;

[0018] Multiple temperature sensors are distributed in the cavity and can collect the multi-region temperature data of the data center server.

[0019] Optionally, the reflux heat exchange unit includes multiple groups of submersible pumps and plate heat exchangers;

[0020] The number of the submersible pumps matches the number of the temperature sensors. The submersible pumps are arranged in the cavity and submerged in the coolant, and can pump the coolant in the cavity to the plate heat exchanger;

[0021] The plate heat exchanger is connected to the heat dissipation unit and can return the coolant input by the submersible pump to the cavity.

[0022] Optionally, the multi-region temperature data includes the high-temperature area temperature data of the GPU of the data center server, the medium-temperature area temperature data of the CPU, and the low-temperature area temperature data of the storage;

[0023] The temperature sensors include a first temperature sensor, a second temperature sensor, and a third temperature sensor;

[0024] The submersible pumps include a first submersible pump, a second submersible pump, and a third submersible pump;

[0025] The first temperature sensor is paired with the first submersible pump. The first temperature sensor is used to collect temperature data in the low-temperature storage area and transmit the temperature data to the controller. The first submersible pump is used to pump the coolant in the low-temperature storage area into the plate heat exchanger under the control of the controller;

[0026] The second temperature sensor is paired with the second submersible pump. The second temperature sensor is used to collect temperature data in the medium-temperature area of the CPU and transmit the temperature data to the controller. The second submersible pump is used to pump the coolant in the medium-temperature area of the CPU into the plate heat exchanger under the control of the controller;

[0027] The third temperature sensor is paired with the third submersible pump. The third temperature sensor is used to collect temperature data in the high-temperature area of the GPU and transmit the temperature data to the controller. The third submersible pump is used to pump the coolant in the high-temperature area of the GPU into the plate heat exchanger under the control of the controller.

[0028] Optionally, the liquid output end of the first submersible pump is connected to the first coolant input branch pipe;

[0029] The liquid output end of the second submersible pump is connected to the second coolant input branch pipe;

[0030] The liquid output end of the third submersible pump is connected to the third coolant input branch pipe;

[0031] The first coolant input branch pipe, the second coolant input branch pipe, and the third coolant input branch pipe are all communicated with the coolant input pipe.

[0032] Optionally, the plate heat exchanger is provided with a reflux inlet, a reflux outlet, a cooling inlet, and a cooling outlet;

[0033] The coolant input pipe is connected to the reflux inlet on the plate heat exchanger;

[0034] The heat dissipation unit is connected to the cooling inlet on the plate heat exchanger through a refrigeration input pipe, and the heat dissipation unit is connected to the cooling outlet on the plate heat exchanger through a refrigeration output pipe;

[0035] The reflux outlet on the plate heat exchanger is connected with a coolant reflux pipe, and the coolant is refluxed to the bottom of the cavity;

[0036] The plate heat exchanger is internally provided with heat exchange tubes, a plate heat exchange fin group, and heat exchange connecting tubes;

[0037] Both ends of the heat exchange tube are respectively connected to the coolant input pipe and the coolant reflux pipe, and the heat exchange tube can pass through the plate heat exchange fin group;

[0038] The plate heat exchange fin group is internally provided with plate heat exchange fins arranged in layers;

[0039] The heat exchange connecting pipe is arranged between adjacent plate heat exchange fins and connects adjacent plate heat exchange fins;

[0040] The refrigeration output pipe is connected to one end of the plate heat exchange fin group, and a second one-way valve is arranged on the refrigeration output pipe;

[0041] The refrigeration input pipe is connected to the other end of the plate heat exchange fin group, and a first one-way valve is arranged on the refrigeration input pipe;

[0042] The heat exchange pipe is of a U-shaped structure.

[0043] Optionally, the evaporative heat exchange unit includes an evaporator;

[0044] The evaporator is arranged in the cavity and is arranged in parallel close to the data center server. The evaporator is detachably connected to the inner wall of the cavity through a connecting block, and the evaporator is communicated with the heat dissipation unit.

[0045] Optionally, a bent coil is arranged in the evaporator, and a heat exchange agent is filled in the bent coil;

[0046] The evaporator is communicated with the refrigeration output pipe through a heat exchange agent output pipe, and the refrigeration output pipe is communicated with the heat dissipation unit;

[0047] The evaporator is communicated with the refrigeration input pipe through a heat exchange agent input pipe, and the refrigeration input pipe is communicated with the heat dissipation unit;

[0048] A third one-way valve is arranged on the heat exchange agent input pipe.

[0049] The present application also provides a predictive immersion cooling system based on the real-time power consumption of a data center, and also provides a predictive immersion cooling method based on the real-time power consumption of a data center, including the following steps:

[0050] The controller acquires the power consumption data of the data center server transmitted by the power consumption acquisition component and the temperature data of the data center server transmitted by the temperature sensor;

[0051] The controller adjusts the flow rates of the first submersible pump, the second submersible pump and the third submersible pump, the power of the evaporator and the power of the heat dissipation unit by comparing the temperature data of the high-temperature area of the GPU, the temperature data of the medium-temperature area of the CPU and the temperature data of the low-temperature area of the storage;

[0052] The controller predicts the real-time power consumption of the GPU and / or the CPU during the next power consumption acquisition component acquisition period based on the trained model, and pre-adjusts the flow rates of the first submersible pump, the second submersible pump and the third submersible pump, the power of the evaporator and the power of the heat dissipation unit.

[0053] Optionally, the power consumption acquisition component has an acquisition period of 10 ms, the temperature sensor has an acquisition period of 1 s, and the recording periods of the first, second, and third submersible pumps are 5 s.

[0054] The beneficial effects of the present invention at least include one of the following;

[0055] 1. Based on the structure of the existing immersion cooling system, by adding a data acquisition unit, the power consumption data of the data center server and the temperature distribution in multiple regions can be obtained, enabling judgment from two dimensions and selecting the optimal cooling strategy. Compared with the fixed cooling logic or manual temperature adjustment, the automation degree is higher and the temperature adjustment is more accurate.

[0056] 2. The cooling system that includes a reflux heat exchange unit and an evaporative heat exchange unit can greatly improve the cooling efficiency when they cooperate with each other.

[0057] 3. On the one hand, the reflux heat exchange unit takes away the heat in the coolant and conducts heat exchange with the heat dissipation unit, and then returns to the box again after cooling, achieving dynamic balance of the coolant volume while cooling the coolant. On the other hand, the evaporative heat exchange unit cools the data center server, shortening the heat transfer path, accelerating heat dissipation, ensuring the stable operation of the server, and reducing the temperature of the key components of the server by 10°C - 15°C.

[0058] 4. By matching the evaporator with the data center server, they can be closely arranged, saving space in the data center and improving space utilization.

[0059] 5. In the mode of matching multiple submersible pumps with multiple temperature sensors, the temperature sensors can subdivide and collect the temperature in the area where they are located, and the controller can adjust the power of the submersible pumps based on the collected results, making the heat absorption in the coolant more uniform. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a schematic connection diagram of a predictive immersion cooling system based on the real-time power consumption of a data center;

[0061] Figure 2 It is a schematic diagram of a predictive immersion cooling system based on the real-time power consumption of a data center;

[0062] Figure 3 It is a schematic connection diagram of a reflux heat exchange unit and an evaporative heat exchange unit;

[0063] Figure 4 It is a schematic connection diagram of an evaporative heat exchange unit;

[0064] Figure 5 It is a schematic diagram of a plate heat exchanger structure.

[0065] In the figure:

[0066] 1 is the box body, 2 is the cavity, 3 is the data center server, 4 is the coolant, 5 is the controller, 6 is the heat dissipation unit, 7 is the plate heat exchanger, 8 is the evaporator, 9 is the first submersible pump, 10 is the second submersible pump, 11 is the third submersible pump, 12 is the first temperature sensor, 13 is the second temperature sensor, 14 is the third temperature sensor, 15 is the third one-way valve, 16 is the coolant input pipe, 17 is the coolant return pipe, 18 is the heat exchange agent output pipe, 19 is the heat exchange agent input pipe, 20 is the refrigeration output pipe, 21 is the refrigeration input pipe, 22 is the first one-way valve, 23 is the second one-way valve, 24 is the first coolant input branch pipe, 25 is the second coolant input branch pipe, 26 is the third coolant input branch pipe, 28 is the heat exchange pipe, 29 is the plate heat exchange fin, 30 is the heat exchange connection pipe. Specific embodiments

[0067] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some but not all of the embodiments of the present invention. The components of the embodiments of the present invention usually described and illustrated in the accompanying drawings here can be arranged and designed in various different configurations.

[0068] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed present invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the scope of protection of the present invention.

[0069] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0070] It should be noted that like reference numerals and letters denote like items in the following drawings, and thus, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0071] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the invention is customarily placed during use, or the orientation or positional relationship commonly understood by those skilled in the art. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the present invention. In addition, the terms "first", "second", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0072] In the description of the present invention, it should also be noted that unless otherwise clearly specified and defined, the terms "set", "installed", "connected", "coupled" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0073] As Figure 1 and Figure 2 shown, a predictive immersion cooling system based on the real-time power consumption of a data center includes a box body 1 and a data center server 3. A cavity 2 is provided in the box body 1, and a coolant 4 is injected into the cavity 2. The data center server 3 is placed in the cavity 2 and immersed in the coolant 4. It also includes a reflux heat exchange unit, an evaporative heat exchange unit, a controller 5, a heat dissipation unit 6, and a data acquisition unit;

[0074] Both the reflux heat exchange unit and the evaporative heat exchange unit can perform heat exchange on the data center server 3, and the reflux heat exchange unit and the evaporative heat exchange unit are connected to the heat dissipation unit 6;

[0075] The heat dissipation unit 6 is provided outside the box body 2;

[0076] The data acquisition unit can collect the power consumption data of the data center server 3 and the multi-region temperature data of the data center server 3, and can transmit the power consumption data and the temperature data to the controller 5;

[0077] The controller 5 can control the heat dissipation unit 6, the reflux heat exchange unit, and the evaporative heat exchange unit.

[0078] The purpose of such a design is that, based on the structure of the existing immersion cooling system, adding a data acquisition unit can obtain the power consumption data of the data center server and the temperature distribution in multiple regions, enabling judgment from two dimensions and selecting the optimal cooling strategy. Compared with the fixed cooling logic or manual temperature adjustment, the degree of automation is higher and the temperature adjustment is more accurate. By using a cooling system that includes a reflux heat exchange unit and an evaporative heat exchange unit, the two can cooperate to greatly improve the cooling efficiency. On the one hand, the reflux heat exchange unit takes away the heat in the coolant and conducts heat exchange with the cooling unit, and then returns to the box after cooling, achieving both the cooling of the coolant and the dynamic balance of the coolant volume. On the other hand, the evaporative heat exchange unit cools the data center server, shortens the heat transfer path, accelerates heat dissipation, and ensures the stable operation of the server, which can reduce the temperature of the key components of the server by 10°C - 15°C.

[0079] In this embodiment, as Figure 3 and Figure 4 shown, the data acquisition unit includes a power consumption acquisition component and multiple temperature sensors;

[0080] The power consumption acquisition component is signal-connected to the data center server 3 and can acquire the real-time power consumption of the GPU and / or CPU in the data center server 3;

[0081] The multiple temperature sensors are distributed in the cavity 2 and can acquire the temperature data of multiple regions of the data center server 3.

[0082] The reflux heat exchange unit includes multiple groups of submersible pumps and plate heat exchangers 7;

[0083] The number of the submersible pumps matches the number of the temperature sensors. The submersible pumps are arranged in the cavity 2 and submerged in the coolant 4, and can pump the coolant 4 in the cavity 2 to the plate heat exchanger 7;

[0084] The plate heat exchanger 7 is connected to the cooling unit 6 and can return the coolant 4 input by the submersible pump to the cavity 2.

[0085] The multi-region temperature data includes the high-temperature zone temperature data of the GPU of the data center server 3, the medium-temperature zone temperature data of the CPU, and the low-temperature zone temperature data of the storage;

[0086] The temperature sensors include a first temperature sensor 12, a second temperature sensor 13, and a third temperature sensor 14;

[0087] The submersible pumps include a first submersible pump 9, a second submersible pump 10, and a third submersible pump 11;

[0088] The first temperature sensor 12 is paired with the first submersible pump 9. The first temperature sensor 12 is used to collect temperature data in the low-temperature storage area and transmit the temperature data to the controller 5. The first submersible pump 9 is used to pump the coolant 4 in the low-temperature storage area into the plate heat exchanger 7 under the control of the controller 5;

[0089] The second temperature sensor 13 is paired with the second submersible pump 10. The second temperature sensor 13 is used to collect temperature data in the medium-temperature area of the CPU and transmit the temperature data to the controller 5. The second submersible pump 10 is used to pump the coolant 4 in the medium-temperature area of the CPU into the plate heat exchanger 7 under the control of the controller 5;

[0090] The third temperature sensor 14 is paired with the third submersible pump 11. The third temperature sensor 14 is used to collect temperature data in the high-temperature area of the GPU and transmit the temperature data to the controller 5. The third submersible pump 11 is used to pump the coolant 4 in the high-temperature area of the GPU into the plate heat exchanger 7 under the control of the controller 5.

[0091] Meanwhile, the liquid output end of the first submersible pump 9 is connected to the first coolant input branch pipe 24;

[0092] The liquid output end of the second submersible pump 10 is connected to the second coolant input branch pipe 25;

[0093] The liquid output end of the third submersible pump 11 is connected to the third coolant input branch pipe 26;

[0094] The first coolant input branch pipe 24, the second coolant input branch pipe 25, and the third coolant input branch pipe 26 are all communicated with the coolant input pipe 16.

[0095] The purpose of such a design is that the submersible pump can be set according to the structure of the specific data center server, usually in an upper-middle-lower three-layer setting or an inner-middle-outer three-layer setting. Here, for the convenience of description, an upper-middle-lower three-layer setting is adopted. Of course, those skilled in the art can also set it according to the actual situation in specific implementation schemes.

[0096] In this embodiment, it is assumed that the GPU of the data center server 3 is located at the bottom, the CPU is located in the middle, and the storage structure is located at the top. Therefore, the GPU high-temperature zone temperature data, the CPU medium-temperature zone temperature data, and the storage low-temperature zone temperature data are set from top to bottom. The sets composed of the submersible pumps and temperature sensors in the three divided regions are respectively denoted as T1, T2, and T3. The function of T1 is to serve as a turbulent flow generator in the high heat flux density zone, break through the laminar flow limit, and suppress vertical heat accumulation. The function of T2 is to serve as a dynamic balance valve for vertical heat conduction, block heat rise, and maintain horizontal heat balance. The function of T3 is to serve as a shock absorber in the low-temperature zone, suppress temperature fluctuations, and achieve precise temperature control. In this way, under the action of the controller, the power of the three submersible pumps is adjusted, so that the flow rate of the coolant pumped in different layers changes, so that heat cannot accumulate in the coolant. When the coolant in each layer is pumped into the plate heat exchanger, and then cooled by heat exchange and then flows back from the top of the box to the bottom, the whole coolant forms a large cycle, and the heat is continuously exported from the cavity.

[0097] On this basis, as Figure 5 shown, a specific structure of a plate heat exchanger is provided in this embodiment, where

[0098] a return inlet, a return outlet, a cooling inlet, and a cooling outlet are provided on the plate heat exchanger 7;

[0099] the coolant input pipe 16 is connected to the return inlet on the plate heat exchanger 7;

[0100] the heat dissipation unit 6 is connected to the cooling inlet on the plate heat exchanger 7 through the refrigeration input pipe 21, and the heat dissipation unit 6 is connected to the cooling outlet on the plate heat exchanger 7 through the refrigeration output pipe 20;

[0101] the return outlet on the plate heat exchanger 7 is connected with a coolant return pipe 17, and the coolant 4 is returned to the bottom of the cavity 2;

[0102] heat exchange tubes 28, a plate heat exchange fin group, and heat exchange connection tubes 30 are arranged in the plate heat exchanger 7;

[0103] both ends of the heat exchange tube 28 are respectively connected with the coolant input pipe 16 and the coolant return pipe 17, and the heat exchange tube 28 can pass through the plate heat exchange fin group;

[0104] hierarchically arranged plate heat exchange fins 29 are arranged in the plate heat exchange fin group;

[0105] the heat exchange connection tubes 30 are arranged between adjacent plate heat exchange fins 29 and connect adjacent plate heat exchange fins 29;

[0106] The refrigeration output pipe 20 is communicated with one end of the plate heat exchange fin group, and a second one-way valve 23 is arranged on the refrigeration output pipe 20;

[0107] The refrigeration input pipe 21 is communicated with the other end of the plate heat exchange fin group, and a first one-way valve 22 is arranged on the refrigeration input pipe 21;

[0108] The heat exchange pipe 28 is of a U-shaped structure.

[0109] The purpose of such design is that the whole heat exchange process is a heat conduction process, and the heat is conducted from the high-temperature coolant to the low-temperature heat exchange agent, and these heat exchange agents are introduced into the plate heat exchanger through the heat dissipation unit.

[0110] In the specific heat exchange process, the heat exchange agent enters the plate heat exchange fin through the refrigeration input pipe, fills the plate heat exchange fin group through the heat exchange connecting pipe, returns to the heat exchange unit through the refrigeration output pipe, and then the exported coolant enters the heat exchange pipe. Since the heat exchange pipe passes through the plate heat exchange fin and is also of a U-shaped structure, the contact area between the heat exchange pipe and the plate heat exchange fin is greatly increased, and there is sufficient heat exchange space.

[0111] In this embodiment, in order to avoid the reverse flow phenomenon of the pipeline during transportation, one-way valves are added to several relatively important pipelines. For example, a second one-way valve 23 is arranged on the refrigeration output pipe 20, and a first one-way valve 22 is arranged on the refrigeration input pipe 21.

[0112] In this embodiment, a specific structure of an evaporative heat exchange unit is provided, and the evaporative heat exchange unit includes an evaporator 8;

[0113] The evaporator 8 is arranged in the cavity 2 and is arranged in parallel near the data center server 3. The evaporator 8 is detachably connected to the inner wall of the cavity 2 through a connecting block, and the evaporator 8 is communicated with the heat dissipation unit 6.

[0114] A bent coiled pipe is arranged in the evaporator 8, and a heat exchange agent is filled in the bent coiled pipe;

[0115] The evaporator 8 is communicated with the refrigeration output pipe 20 through a heat exchange agent output pipe 18, and the refrigeration output pipe 20 is communicated with the heat dissipation unit 6;

[0116] The evaporator 8 is communicated with the refrigeration input pipe 21 through a heat exchange agent input pipe 19, and the refrigeration input pipe 21 is communicated with the heat dissipation unit 6;

[0117] A third one-way valve 15 is arranged on the heat exchange agent input pipe 19.

[0118] The purpose of such a design is that, with its large contact surface area and high thermal conductivity, the evaporator quickly transfers the heat in the coolant near the data center server to the external heat dissipation unit, thereby cooling this area. The cooled coolant circulates around the server and continues to absorb heat, thus forming a continuous and stable heat dissipation cycle.

[0119] It should also be noted that in the technical solution provided in this embodiment, the refrigeration output pipe and the refrigeration input pipe are respectively used for heat exchange between the evaporator and the plate heat exchanger. Therefore, Freon is selected as the heat exchange agent. Considering environmental protection and other factors, those skilled in the art can select more environmentally friendly materials for corresponding replacement. This embodiment is only for principle illustration and does not limit its materials.

[0120] Furthermore, in specific operations, the heat dissipation unit is similar to the outdoor unit of an air conditioner in function. Therefore, this embodiment selects this as the heat dissipation unit, and those skilled in the art can also select other suitable structures.

[0121] In this embodiment, a predictive immersion heat dissipation method based on the real-time power consumption of a data center is also provided for the predictive immersion heat dissipation system of the data center, including the following steps:

[0122] The controller 5 acquires the power consumption data of the data center server 3 transmitted by the power consumption acquisition component and the temperature data of the data center server 3 transmitted by the temperature sensor;

[0123] The controller 5 adjusts the flow rates of the first submersible pump 9, the second submersible pump 10, and the third submersible pump 11, the power of the evaporator 8, and the power of the heat dissipation unit 6 by comparing the temperature data of the high-temperature GPU area, the medium-temperature CPU area, and the low-temperature storage area;

[0124] The controller 5 predicts the real-time power consumption of the GPU and / or CPU during the next power consumption acquisition component acquisition cycle based on the trained model, and pre-adjusts the flow rates of the first submersible pump 9, the second submersible pump 10, and the third submersible pump 11, the power of the evaporator 8, and the power of the heat dissipation unit 6.

[0125] Meanwhile, the acquisition cycle of the power consumption acquisition component is 10 ms, the acquisition cycle of the temperature sensor is 1 s, and the recording cycle of the first submersible pump 9, the second submersible pump 10, and the third submersible pump 11 is 5 s.

[0126] In a specific implementation scenario, the method provided by this application makes a judgment from two dimensions: the real-time power consumption of the data center and the temperature distribution in key areas, and selects the optimal cooling strategy. When predicting cooling based on the real-time power consumption of the data center, the purpose is to cope with the upcoming increase in GPU and / or CPU temperature caused by the increase in power consumption. The way to do this is to increase the temperature of the cooling components in the corresponding area in advance, so that the overall temperature will not fluctuate greatly.

[0127] In a specific prediction, an existing large model can be selected, or a large model can be trained independently for temperature prediction. Taking the LSTM-TCN hybrid model as an example, first construct a sample group. Each sample group contains 9,000 power consumption data, among which there are 3,000 data for T1, T2, and T3 respectively. The sampling period is 10 ms, and the sampling time is 30 s.

[0128] At the same time, add the flow data of the first submersible pump, the second submersible pump, and the third submersible pump, and correlate the flow with the power consumption.

[0129] Furthermore, set the T1 temperature zone to be triggered when the temperature > 40°C or the predicted temperature rise based on the power consumption data model > 2°C;

[0130] - High-speed stage: 12 L / min (lasting for 3 seconds);

[0131] - Low-speed stage: 8 L / min (lasting for 1 second);

[0132] - Basic flow regulation: For every 1°C increase in temperature (25°C → 45°C), the flow linearly increases by 0.25 L / min (25°C → 5 L / min, 45°C → 10 L / min).

[0133] Set the T2 temperature zone. When the predicted temperature difference between T1 and T2 based on the power consumption data model ≤ 4°C, the flow rate is reduced by 20%;

[0134] Based on the power consumption data model, for every 1°C decrease in the temperature difference between T2 and T3, the flow rate linearly decreases by 0.3 L / min;

[0135] Basic flow regulation: For every 1°C increase in temperature (25°C → 45°C), the flow linearly increases by 0.2 L / min (25°C → 4 L / min, 45°C → 8 L / min).

[0136] For the T3 temperature zone, when the temperature collected by the temperature sensor < 25°C, turn off or pause the operation of the submersible pump;

[0137] Basic flow regulation: For every 1°C increase in temperature (25°C → 45°C), the flow linearly increases by 0.125 L / min (25°C → 1 L / min, 45°C → 3.5 L / min).

[0138] Based on this scenario, the temperature data-based controller controls the three submersible pumps and the evaporator as follows. When the temperature sensor collects that the temperature in area T1 rises by 2 °C or more, the flow rate of the submersible pump in this area is increased by 10% to 15%, and at the same time, the power of the evaporator is increased in coordination, that is, the power of the heat dissipation unit is adjusted actively to regulate the flow rate of the heat exchange agent in the evaporator.

[0139] When all three temperature zones are within the set threshold range, and at the same time the temperature difference between T1 and T2 > 4 °C, it indicates that the temperature difference between the two temperature zones is too large. To ensure the heat dissipation in area T1, the flow rate of the submersible pump in area T2 is actively reduced. In this scenario, it is reduced by 20% to prevent reverse heat absorption, and the power of the evaporator is locked to avoid interfering with the temperature control.

[0140] When all three temperature zones are within the set threshold range, and at the same time the temperature difference between T2 and T3 > 5 °C, it indicates that the temperature difference between the two temperature zones is too large. Since area T2 is between areas T1 and T3, the flow rate of the submersible pump in area T2 is increased in a slow stepwise manner. In this scenario, it is increased by 0.5 L / min per second until the temperature difference returns to the range, and at the same time, the power of the evaporator is increased.

[0141] In this scenario, the power consumption data-based controller controls the three submersible pumps and the evaporator as follows. When it is detected that the GPU and / or CPU are suddenly fully loaded, the controller increases the flow rate of the submersible pumps in areas T1 and T2 in advance, and at the same time increases the power of the evaporator.

[0142] When the large model predicts that the GPU load will climb within 60 s, active cooling is carried out in two stages:

[0143] Pre-cooling startup stage (0 - 30 s):

[0144] Basic flow rate of the submersible pump in area T1: 28 °C → 7 L / min (0.25 L / min / °C);

[0145] Anti-heat countercurrent of the submersible pump in area T2: T1 - T2 temperature difference = 3.8 °C (< 4 °C threshold), flow rate from 5 L / min → 4 L / min (decrease by 20%);

[0146] Evaporator: power 1.5 kW;

[0147] Load climbing stage (30 - 60 s):

[0148] Emergency response in area T1: predicted temperature rise > 2 °C → the submersible pump adopts a pulse mode (12 L / min × 3 s → 8 L / min × 1 s);

[0149] Flow rate control in area T2: T2 - T3 temperature difference 7 °C → the flow rate of the submersible pump from 4 L / min → 6.5 L / min (+0.5 L per second);

[0150] Evaporator: Power 2.8 kW

[0151] Steady-state recovery (after 60 seconds)

[0152] Flow normalization: Submersible pump in T1 area = 10 L / min (45 °C), submersible pump in T2 area = 5 L / min (40 °C), submersible pump in T3 area = 2 L / min (30 °C).

[0153] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A predictive immersion cooling system based on the real-time power consumption of a data center, comprising a box body (1) and a data center server (3). A cavity (2) is arranged inside the box body (1), and a coolant (4) is injected into the cavity (2). The data center server (3) is placed in the cavity (2) and immersed in the coolant (4). It is characterized in that, It also includes a reflux heat exchange unit, an evaporative heat exchange unit, a controller (5), a heat dissipation unit (6), and a data acquisition unit; Both the reflux heat exchange unit and the evaporative heat exchange unit can perform heat exchange on the data center server (3), and the reflux heat exchange unit and the evaporative heat exchange unit are connected to the heat dissipation unit (6); The heat dissipation unit (6) is arranged outside the box body (2); The data acquisition unit can collect the power consumption data of the data center server (3) and the multi-region temperature data of the data center server (3), and can transmit the power consumption data and the temperature data to the controller (5); The controller (5) can control the heat dissipation unit (6), the reflux heat exchange unit, and the evaporative heat exchange unit.

2. The predictive immersion cooling system based on the real-time power consumption of the data center according to claim 1, wherein The data acquisition unit includes a power consumption acquisition component and multiple temperature sensors; The power consumption acquisition component is signal-connected to the data center server (3) and can collect the real-time power consumption of the GPU and / or CPU in the data center server (3); Multiple temperature sensors are distributed in the cavity (2) and can collect the multi-region temperature data of the data center server (3).

3. The predictive immersion cooling system based on the real-time power consumption of the data center according to claim 2, wherein The reflux heat exchange unit includes multiple groups of submersible pumps and a plate heat exchanger (7); The number of the submersible pumps matches the number of the temperature sensors. The submersible pumps are arranged in the cavity (2) and submerged in the coolant (4), and can pump the coolant (4) in the cavity (2) into the plate heat exchanger (7); The plate heat exchanger (7) is connected to the heat dissipation unit (6) and can return the coolant (4) input by the submersible pumps to the cavity (2).

4. The predictive immersion cooling system based on the real-time power consumption of the data center according to claim 2, wherein The multi-region temperature data includes the high-temperature area temperature data of the GPU in the data center server (3), the medium-temperature area temperature data of the CPU, and the low-temperature area temperature data of the storage; The temperature sensors include a first temperature sensor (12), a second temperature sensor (13), and a third temperature sensor (14); The submersible pumps include a first submersible pump (9), a second submersible pump (10), and a third submersible pump (11); The first temperature sensor (12) is paired with the first submersible pump (9). The first temperature sensor (12) is used to collect the low-temperature area temperature data of the storage and transmit the temperature data to the controller (5). The first submersible pump (9) is used to pump the coolant (4) in the low-temperature area of the storage into the plate heat exchanger (7) under the control of the controller (5); The second temperature sensor (13) is paired with the second submersible pump (10). The second temperature sensor (13) is used to collect the medium-temperature area temperature data of the CPU and transmit the temperature data to the controller (5). The second submersible pump (10) is used to pump the coolant (4) in the medium-temperature area of the CPU into the plate heat exchanger (7) under the control of the controller (5); The third temperature sensor (14) is paired with the third submersible pump (11). The third temperature sensor (14) is used to collect the high-temperature area temperature data of the GPU and transmit the temperature data to the controller (5). The third submersible pump (11) is used to pump the coolant (4) in the high-temperature area of the GPU into the plate heat exchanger (7) under the control of the controller (5).

5. The predictive immersion cooling system based on the real-time power consumption of the data center according to claim 4, characterized in that, The liquid output end of the first submersible pump (9) is connected to the first coolant input branch pipe (24); The liquid output end of the second submersible pump (10) is connected to the second coolant input branch pipe (25); The liquid output end of the third submersible pump (11) is connected to the third coolant input branch pipe (26); The first coolant input branch pipe (24), the second coolant input branch pipe (25) and the third coolant input branch pipe (26) are all communicated with the coolant input pipe (16).

6. The predictive immersion cooling system based on the real-time power consumption of the data center according to claim 5, characterized in that, The plate heat exchanger (7) is provided with a reflux inlet, a reflux outlet, a cooling inlet and a cooling outlet; The coolant input pipe (16) is connected to the reflux inlet on the plate heat exchanger (7); The heat dissipation unit (6) is connected to the cooling inlet on the plate heat exchanger (7) through the refrigeration input pipe (21), and the heat dissipation unit (6) is connected to the cooling outlet on the plate heat exchanger (7) through the refrigeration output pipe (20); The reflux outlet on the plate heat exchanger (7) is connected with a coolant reflux pipe (17), and the coolant (4) is refluxed to the bottom of the cavity (2); The plate heat exchanger (7) is internally provided with a heat exchange pipe (28), a plate heat exchange fin group and a heat exchange connecting pipe (30); Both ends of the heat exchange pipe (28) are respectively connected to the coolant input pipe (16) and the coolant reflux pipe (17), and the heat exchange pipe (28) can pass through the plate heat exchange fin group; The plate heat exchange fin group is internally provided with plate heat exchange fins (29) arranged in layers; The heat exchange connecting pipe (30) is arranged between adjacent plate heat exchange fins (29) and communicates the adjacent plate heat exchange fins (29); The refrigeration output pipe (20) is communicated with one end of the plate heat exchange fin group, and a second one-way valve (23) is arranged on the refrigeration output pipe (20); The refrigeration input pipe (21) is communicated with the other end of the plate heat exchange fin group, and a first one-way valve (22) is arranged on the refrigeration input pipe (21); The heat exchange pipe (28) is in a U-shaped structure.

7. The predictive immersion cooling system based on the real-time power consumption of the data center according to claim 1, wherein The evaporative heat exchange unit includes an evaporator (8); The evaporator (8) is arranged in the cavity (2), is arranged in parallel near the data center server (3), is detachably connected to the inner wall of the cavity (2) through a connecting block, and the evaporator (8) is communicated with the heat dissipation unit (6).

8. The predictive immersion cooling system based on the real-time power consumption of the data center according to claim 7, wherein, The evaporator (8) is internally provided with a bent coiled pipe, and the bent coiled pipe is filled with a heat exchange agent; The evaporator (8) is communicated with the refrigeration output pipe (20) through the heat exchange agent output pipe (18), and the refrigeration output pipe (20) is communicated with the heat dissipation unit (6); The evaporator (8) is communicated with the refrigeration input pipe (21) through the heat exchange agent input pipe (19), and the refrigeration input pipe (21) is communicated with the heat dissipation unit (6); A third one-way valve (15) is arranged on the heat exchange agent input pipe (19).

9. A predictive immersion cooling method based on the real-time power consumption of a data center, implemented based on the predictive immersion cooling system for a data center based on real-time power consumption according to any one of claims 4 to 6, characterized in that, Including the following steps: The controller (5) acquires the power consumption data of the data center server (3) transmitted by the power consumption acquisition component and the temperature data of the data center server (3) transmitted by the temperature sensor; The controller (5) adjusts the flow rates of the first submersible pump (9), the second submersible pump (10), and the third submersible pump (11), the power of the evaporator (8), and the power of the heat dissipation unit (6) by comparing the temperature data of the high-temperature area of the GPU, the temperature data of the medium-temperature area of the CPU, and the temperature data of the low-temperature area of the storage. The controller (5) pre-adjusts the flow rates of the first submersible pump (9), the second submersible pump (10), and the third submersible pump (11), the power of the evaporator (8), and the power of the heat dissipation unit (6) by predicting the real-time power consumption of the GPU and / or the CPU during the acquisition period of the next power consumption acquisition component based on the trained model.

10. The predictive immersion cooling method based on the real-time power consumption of a data center according to claim 9, characterized in that, The acquisition period of the power consumption acquisition component is 10 ms, the acquisition period of the temperature sensor is 1 s, and the recording period of the first submersible pump (9), the second submersible pump (10), and the third submersible pump (11) is 5 s.

Images