Built-in centralized power supply device of immersed liquid cooling computing power module and control method of built-in centralized power supply device

Abstract

The embodiment of the invention provides an immersed liquid cooling computing power module built-in centralized power supply device and a control method thereof, and relates to the technical field of computing power module built-in power supply technologies. The device comprises a liquid cooling cavity; the plurality of power supply modules are arranged, and the power supply modules are installed in the liquid cooling cavity in a pluggable mode; and the busbar is electrically connected with the power supply module. According to the invention, the problem of large power supply occupied space of an immersed liquid cooling system is solved, and the effect of reducing the power supply occupied space is achieved.

Description

Technical Field

[0001] This invention relates to the field of built-in power supply technology for computing modules, and more specifically, to a centralized power supply device and control method for an immersion liquid-cooled computing module. Background Technology

[0002] With the explosive growth in demand for artificial intelligence and high-performance computing, the power density of single-rack systems has increased dramatically, and immersion liquid cooling technology has gradually become mainstream due to its efficient heat dissipation capabilities. However, existing technical solutions face numerous challenges in the power supply architecture design of immersion liquid cooling systems.

[0003] The existing technical solution is to attach the Powershelf (power module) to the outside of the Tank (liquid cooling cavity). This solution requires an additional complex integrated fixing structure to be designed for the "Tank cavity + Powershelf" structure, which increases the overall size and footprint of the system.

[0004] There is currently no good solution to the above problems. Summary of the Invention

[0005] This invention provides a centralized power supply device and control method for an immersion liquid-cooled computing module, which at least solves the problem of large space occupation for power supply in immersion liquid-cooled systems in related technologies.

[0006] According to one embodiment of the present invention, a centralized power supply device built into an immersion liquid-cooled computing module is provided, comprising:

[0007] Liquid-cooled cavity;

[0008] Several power modules are provided, and the power modules are plugged into and installed in the liquid cooling cavity.

[0009] The busbar is electrically connected to the power module.

[0010] In one exemplary embodiment, the busbar includes a positive copper busbar, a negative copper busbar, and an insulating layer located between them, arranged in parallel.

[0011] According to another embodiment of the present invention, a control method for a centralized power supply device built into an immersion liquid-cooled computing module is provided, applied to the aforementioned centralized power supply device built into an immersion liquid-cooled computing module, comprising:

[0012] Obtain the number of power modules in place and control the power modules to perform a pre-charge soft start to establish the base voltage of the bus;

[0013] Collect aggregated current data on the busbar and temperature data of the immersion medium in the liquid cooling cavity, and correct the physical impedance value of the busbar based on the temperature data to obtain a real-time load profile.

[0014] Based on the real-time load profile, the overall system efficiency value is determined, and the target number of activated power modules is determined according to the preset redundancy safety constraints.

[0015] The transmission line voltage drop is calculated based on the aggregated current data and the corrected physical impedance value, and a voltage regulation command is generated based on the transmission line voltage drop to control the power output.

[0016] In one exemplary embodiment, the control power module performs a pre-charge soft-start to establish the bus base voltage, including:

[0017] Send a precharge command to all power modules that have detected an in-situ signal to control the power modules to inject charge into the bus with a precharge current;

[0018] Monitor the terminal voltage of the bus;

[0019] When the terminal voltage rises to the preset rated operating voltage threshold, the power module is controlled to switch to constant voltage output mode and the main output circuit is closed.

[0020] In one exemplary embodiment, correcting the physical impedance value of the bus based on the temperature data includes:

[0021] Obtain the basic impedance value of the bus at a standard reference temperature and the temperature coefficient of resistance of the bus material;

[0022] Calculate the temperature difference between the temperature data of the immersion medium and the standard reference temperature;

[0023] Based on the basic impedance value, the temperature coefficient of resistance, and the temperature difference, the actual physical impedance value at the current temperature is determined.

[0024] In one exemplary embodiment, determining the target number of activated power modules based on preset redundancy safety constraints includes:

[0025] Obtain the rated power parameters of a single power module;

[0026] Calculate the minimum number of power supply modules required to meet the total load power of the current system;

[0027] The number of candidate activations is determined based on the preset redundancy mode and the minimum number of power modules.

[0028] If the number of candidate activations is less than or equal to the number in place, the number of candidate activations is determined as the target number of activated power modules.

[0029] In one exemplary embodiment, the method further includes:

[0030] Record the cumulative operating time of each of the power modules;

[0031] When the number of the target activated power modules is less than the number in place, the power module with the longest cumulative running time is controlled to enter a sleep state, and the sleep power module with the shortest cumulative running time is controlled to enter an activated state.

[0032] According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0033] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0034] This invention achieves blind-plugging and hot-swapping of the computing module and power module by embedding the power module inside an immersion liquid-cooled cavity and using a cableless connection method with busbars and current module connectors. This eliminates the obstruction of the heat dissipation field by traditional cables, significantly improving system integration and maintenance efficiency while reducing the space required. Therefore, it solves the problem of large space occupation and improves space utilization efficiency. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of a centralized power supply device built into an immersion liquid-cooled computing module according to an embodiment of the present invention.

[0036] Figure 2 This is a schematic diagram of the power module structure according to an embodiment of the present invention;

[0037] Figure 3 This is a partial structural diagram of a centralized power supply device built into an immersion liquid-cooled computing module according to an embodiment of the present invention;

[0038] Figure 4 This is a top view of a partial structure of a centralized power supply device built into an immersion liquid-cooled computing module according to an embodiment of the present invention.

[0039] Figure 5This is a flowchart of a control method for a centralized power supply device built into an immersion liquid-cooled computing module according to an embodiment of the present invention.

[0040] In the diagram, 1 is the liquid cooling cavity; 2 is the power module; 21 is the elastic clamping connector; 31 is the first part; 32 is the second part; 33 is the third part; and 4 is the computing module. Detailed Implementation

[0041] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0042] In the following description, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0043] Furthermore, in this application, directional terms such as "upper," "lower," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and may change accordingly depending on the orientation of the components in the accompanying drawings.

[0044] In this application, unless otherwise expressly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a direct connection or an indirect connection through an intermediate medium. Furthermore, the term "coupled" can refer to an electrical connection that enables signal transmission.

[0045] As used herein, “about,” “approximately,” or “approximately” includes the stated value and the average value within an acceptable range of deviation from the given value, wherein the acceptable range of deviation is determined by a person skilled in the art taking into account the measurement under discussion and the error associated with the measurement of the given quantity (i.e., the limitations of the measurement system).

[0046] Example 1

[0047] This embodiment provides a centralized power supply device built into an immersion liquid-cooled computing module. This device, through deep integration of its physical structure and a cableless design, solves the problems of messy cables and wasted space in existing technologies.

[0048] like Figure 1-4As shown, the device includes an immersion liquid-cooled chamber 1 (Tank), a power module 2 (Powershelf), a busbar system, and a power management unit (PMU). A computing module 4 is connected to the busbar. Both the power module 2 and the computing module 4 are equipped with elastic clamping connectors 21 (Clip) for clamping and connecting with the busbar. The liquid-cooled chamber 1 is filled with insulating coolant such as fluorinated liquid or mineral oil to immerse all heat-generating electronic components.

[0049] The power module 2 is embedded in the internal space of one side of the liquid cooling cavity 1. Specifically, the power module 2 is configured as a vertically inserted frame structure.

[0050] The busbar is fixedly arranged on the inner wall of the liquid-cooled cavity 1 and is completely immersed in the insulating coolant. Specifically, it includes a vertically arranged first part 31, a horizontally arranged second part 32 located at the top of the first part 31, and a third part 33 located at the bottom of the first part 31 and extending along the length of the liquid-cooled cavity 1. The busbar includes a positive copper busbar, a negative copper busbar, and an insulating isolation layer located between them, and the positive copper busbar and the negative copper busbar are made of high-purity oxygen-free copper (C10100).

[0051] It should be noted that the elastic clamping connectors 21 are respectively disposed at the output end of the power module 2 and the input end of the computing module 4 to achieve "blind insertion" and "electrical conduction". The elastic clamping connectors 21 include paired beryllium copper alloy springs. In order to adapt to the immersion environment, the surface of the beryllium copper alloy springs is plated with a hard gold layer with a thickness of not less than 3 micrometers to prevent oxidation and reduce contact resistance. In particular, the clamping force of the beryllium copper alloy springs is configured to be greater than the preset oil film breaking threshold. This is because in the immersion environment, the insulating coolant will form a very thin insulating oil film on the conductor surface. In order to establish a reliable electrical connection, the oil film must be pierced at the moment of insertion and removal. Therefore, the preset oil film breaking threshold is set to 50 (N). In this way, when the computing module 4 or the computing module 2 is inserted, the elastic clamping connectors 21 engage the bus, and the springs on one side apply a normal pressure of more than 25N. Combined with the micro-scraping action during the insertion and removal process, the oil film on the contact surface is effectively removed, and the contact resistance is controlled within the specified range. the following.

[0052] The power management unit (PMU) is typically integrated on the backplane of power module 2. The power management unit communicates with all power modules 2 in place via I2C or SMBus bus and controls all power modules 2.

[0053] Example 2

[0054] This embodiment provides a control method for a centralized power supply device built into an immersion liquid-cooled computing module. This method is applied to the aforementioned centralized power supply device built into the immersion liquid-cooled computing module, such as... Figure 5 As shown, the specific steps include:

[0055] S11, the power management unit detects the physical connection status of the elastic clamping connector 21 to identify the number of power modules 2 in place, and controls the power modules 2 to perform a pre-charge soft start to establish the base voltage of the bus.

[0056] In this embodiment, the power management unit first determines the total number of power modules actually inserted and locked in the current system by detecting the physical presence pin on each power slot. .

[0057] For example, assuming power module 2 has 8 slots, and the PMU detects that the Present Pin signal of slots 1 to 6 is low (active), while slots 7 and 8 are high (floating), then it identifies... .

[0058] Subsequently, a pre-charge soft start is performed. Since the bus connects a large number of computing modules 4, the input terminals of these modules are usually connected in parallel with large-capacity filter capacitors. If the full voltage is directly applied at this time, the huge surge current will instantly burn the contacts of the elastic clamp connector, or even cause the protection to trip. Therefore, it is necessary to control the power module 2 to perform a pre-charge soft start to establish the base voltage of the bus, which includes the following sub-steps:

[0059] S111, a precharge command is sent to all power modules 4 that have detected an in-situ signal, controlling the power module 2 to inject charge into the busbar with a limited precharge current.

[0060] For example, the PMU sends a "Pre-charge Enable" command to these 6 power modules. The controller inside power module 2 responds to this command and limits its output current to no more than 5% of the rated current (for example, if the rated current is 55.5A, it is limited to about 2.7A). At this time, the 6 modules charge the bus and node capacitors together with a total current of about 16.2A.

[0061] S112, monitor the terminal voltage of the bus, and when the terminal voltage rises to the preset rated operating voltage threshold, control the power supply module 2 to switch to constant voltage output mode and close the main output circuit.

[0062] For example, the PMU monitors the bus voltage at a sampling rate of 10ms; when the voltage rises linearly from 0V to 50V (set to 92% of the rated voltage of 54V), the PMU determines that the pre-charge is complete, which takes about 300ms to 500ms; at this time, the PMU sends the "Main Output Enable" command, the power module 4 releases the current limit and enters the normal constant voltage closed-loop control mode, stabilizing the voltage at 54V.

[0063] Step S12: Collect aggregated current data on the bus and temperature data of the immersion medium (i.e., the aforementioned insulating coolant) at a preset sampling period, and correct the physical impedance value of the bus based on the temperature data to construct a real-time load profile.

[0064] In this embodiment, the current load status and environmental status of the system are accurately sensed to provide a data basis for subsequent optimized control.

[0065] The sampling period is typically set to 10ms to 50ms to balance response speed and processing load; aggregated current data The temperature data of the immersion medium is obtained by reading and summing the output current values ​​reported by all activated power modules 2; Temperature data is obtained from temperature sensors distributed throughout the cavity.

[0066] Correcting the physical impedance value of the bus based on the temperature data includes:

[0067] S121, obtain the basic impedance value of the bus at the standard reference temperature and the temperature coefficient of resistance of the bus material.

[0068] For example, for a copper busbar with a length of 2 meters, at the standard reference temperature Below, the basic impedance of its circuit is measured as The temperature coefficient of resistance of copper for .

[0069] S122, calculate the temperature difference between the temperature data of the immersion medium and the standard reference temperature.

[0070] For example, if the temperature of the immersion medium currently collected... for The temperature difference .

[0071] S123 calculates the actual physical impedance value at the current temperature based on the basic impedance value, the temperature coefficient of resistance, and the temperature difference.

[0072] For example, the actual impedance at the current temperature The calculation is as follows:

[0073] .

[0074] The system is thus able to detect an approximately 8% increase in impedance due to the rise in temperature.

[0075] Then build a real-time load profile:

[0076] S124 calculates the first-order difference of aggregated current data over multiple consecutive sampling periods.

[0077] S125, when the first-order difference value exceeds the preset load step threshold, a load change signal is generated, and subsequent steps are immediately executed, ignoring the current calculation cycle waiting time.

[0078] For example, if in time ,exist time Given that the time interval is 10ms, the first-order difference is... If the preset threshold is If the system determines that computing module 4 has suddenly entered a high-load computing state (such as starting an AI training task), it needs to adjust the power strategy immediately without waiting for the regular polling once per minute, and so on. No limit is set here.

[0079] S13: Based on the real-time load profile, query the total system efficiency value corresponding to different activation numbers in the pre-stored power efficiency-load curve table, and determine the target number of activated power modules according to the preset redundancy safety constraints.

[0080] In this embodiment, traditional N+N redundancy often simply turns on all modules, resulting in each module only sharing a very small current under low load, operating in an inefficient region (e.g., efficiency <85%). This embodiment optimizes the solution to address this issue.

[0081] Specifically, based on preset redundancy safety constraints, the number of target activated power modules is determined as follows:

[0082] S131, obtain the rated power parameters of a single power module 2.

[0083] S132, calculate the minimum number of power supply modules required to meet the total load power of the current system.

[0084] Assuming the rated power of a single power module Current real-time load power (Depend on (Calculated), at this point the minimum required quantity indivual.

[0085] S133, based on the preset N+N redundancy mode or N+1 redundancy mode, add the corresponding redundancy number to the minimum number of power modules to obtain the candidate activation number.

[0086] If the system is configured in a high-reliability N+N mode, then the number of redundancies is equal to... The corresponding number of candidate activations indivual.

[0087] If the system is configured in an economical N+1 mode, the redundancy is 1, corresponding to the number of candidate activations. One, and so on, without limitation here.

[0088] S134, under the condition that the number of candidate activations is less than or equal to the number of in-situ, the number of candidate activations is determined as the target number of activated power modules.

[0089] Assuming the current actual incumbent If the N+N model is adopted, then At this time, the individual load is The load rate is At this point, the efficiency curve shows that this power supply model is most efficient (e.g., 96%) in the 40%-60% load range, while its efficiency is lower at 100% full load or 10% light load. Therefore, keeping all 8 modules running is the optimal solution at present.

[0090] Conversely, if the current load Down to ,but In the corresponding N+N mode Individual load; at this time, the single unit load With a load rate of 33%, the efficiency is approximately 95%. If this algorithm is not executed and all 8 are kept running, the single unit load is only 250W, the load rate is 8.3%, and the efficiency may drop below 80%. It is easy to understand that by reducing the number of active units from 8 to 2, the system saves a lot of fixed losses.

[0091] Of course, the following equilibrium strategies can also be used:

[0092] S135 records the cumulative running time of each power module 2.

[0093] S136, when the number of target activated power modules is less than the number in place, the power module 2 with the longest cumulative running time is prioritized to enter the sleep state, and the power module 2 with the shortest cumulative running time is prioritized to enter the active state.

[0094] For example, if two modules need to be activated from eight modules, and the cumulative running times of modules #1 to #8 are [100h, 120h, ...], the system will select the two modules with the shortest running times (e.g., #1 and #5) to activate, while the rest will remain dormant, thus achieving a rotation system and extending the overall system's mean time between failures (MTBF).

[0095] Step S14: Calculate the transmission line voltage drop based on the aggregated current data and the corrected physical impedance value, and generate a voltage regulation command based on the transmission line voltage drop to control the power module 2 in the active state to increase the output voltage to compensate for the transmission line voltage drop.

[0096] In this embodiment, although the impedance is low due to the use of bus transmission, the voltage drop under high current is still not negligible; in order to ensure the voltage stability of the input terminal of the computing node (e.g., strictly stable at 54V), active compensation must be performed.

[0097] Specifically, generating voltage regulation commands based on transmission line voltage drop includes:

[0098] S141, the preset target load terminal voltage is superimposed with the voltage drop of the transmission line to obtain the target voltage of the power output terminal.

[0099] For example, the target load terminal voltage Current aggregation current The thermally corrected impedance calculated from the aforementioned steps Voltage drop in transmission lines Therefore, the target voltage at the power supply output terminal .

[0100] S142, a fine-tuning command containing the target voltage value of the power output terminal is sent to all the power modules in the active state via the digital communication bus.

[0101] The PMU sends a "VOUT_TRIM" command via the PMBus protocol with a parameter value of 54.216V. All activated power modules 2 respond to this command and synchronously increase their output voltage. In this way, after a bus transmission loss of 0.216V, the voltage reaching the input terminal of the computing module 4 returns to exactly 54.0V, ensuring the absolute stability of the power supply to the computing chip.

[0102] Example 3

[0103] Unlike Example 2, traditional compensation only considers the total busbar current and average temperature. However, in an immersion liquid-cooled box, the busbar section near the inlet has a lower temperature (lower impedance) and a higher temperature (higher impedance) near the outlet. Furthermore, the load on nodes at different locations varies, resulting in a stepped decrease in current distribution on the busbar. To address this, this example improves control accuracy by constructing a topological segmented impedance matrix. The system anticipates the physical location of each node on the busbar (distance from the power source), reads the independent current of each node (reported via PDB or estimated), estimates the segmented liquid temperature, and calculates a highly accurate position-weighted voltage drop, achieving targeted voltage regulation for the furthest node.

[0104] Specifically:

[0105] S151, Set topology parameters.

[0106] The preset physical topology table includes the first Busbar length between node access point and power output terminal For example, in a liquid-cooled enclosure that houses 16 nodes, node #1 is 10cm from the power supply, node #16 is 200cm from the power supply, and the busbar segment length between every two nodes is... .

[0107] S152, calculate local thermal-electric coupling.

[0108] Since the coolant typically enters from one side (e.g., node #1) and exits from the other side (node ​​#16), the coolant temperature gradually increases along the flow path. This is used to calculate the... Temperature of the busbar :

[0109]

[0110] in, These are the actual measured values ​​from the inlet and outlet sensors. For the first The power consumption of the nth node is then calculated. Real-time resistance of segment That is, the resistivity per unit length of the busbar section near the outlet will be corrected to be higher.

[0111] S153, cumulative step-like pressure drop.

[0112] Flowing through the Current in the busbar It is not the total current, but the sum of the currents at all nodes after this segment:

[0113]

[0114] Therefore, the total voltage drop to the farthest node (node ​​#N) For the accumulation of each voltage drop segment:

[0115]

[0116] Compared to simple The model in this embodiment takes into account the dual nonlinear effects of current decay along the path and temperature rise along the path.

[0117] For example, in a full-load test, the calculated voltage drop is 0.5V, while the voltage drop calculated by this topology model is 0.42V (because the large current is mainly concentrated at the front end, and the current at the back end is small). If the compensation is 0.5V, it will lead to the front end node voltage being too high (such as 54.08V), increasing unnecessary power consumption. However, if the compensation is 0.42V, the voltage at the farthest end can be accurately controlled at 54.0V, achieving ultimate energy efficiency control. And so on, without limitation here.

[0118] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0119] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0120] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0121] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0122] Embodiments of the present invention also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0123] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0124] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0125] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0126] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0127] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0128] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0129] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An immersed liquid-cooled computing power module built-in centralized power supply device, characterized in that, The application relates to a centralized power supply device for an immersed liquid cooling computing module, comprising: a liquid cooling cavity; a plurality of power supply modules, which are plug-in installed in the liquid cooling cavity; a bus bar, which is electrically connected with the power supply modules.

2. The apparatus of claim 1, wherein, The bus bar comprises a positive copper bar, a negative copper bar and an insulating layer arranged between the two copper bars.

3. A method for controlling a centralized power supply device built-in an immersion liquid-cooled computing power module, characterized in that, The application is applied to the centralized power supply device for the immersed liquid cooling computing module, and comprises: acquiring the number of in-situ power supply modules and controlling the power supply modules to perform pre-charging soft starting to establish a basic voltage of the bus bar; collecting aggregated current data on the bus bar and temperature data of the immersed medium in the liquid cooling cavity, and correcting the physical impedance value of the bus bar based on the temperature data to obtain a real-time load image; based on the real-time load image, determining a system total efficiency value, and determining a target activated power supply module number according to a preset redundancy safety constraint; calculating a transmission line voltage drop based on the aggregated current data and the corrected physical impedance value, and generating a voltage regulation instruction based on the transmission line voltage drop to control power output.

4. The method of claim 3, wherein, The control of the power supply modules to perform pre-charging soft starting to establish the basic voltage of the bus bar comprises: sending a pre-charging instruction to all the power supply modules detected with in-situ signals to control the power supply modules to inject charges to the bus bar with a pre-charging current; monitoring the terminal voltage of the bus bar; when the terminal voltage rises to a preset rated working voltage threshold, controlling the power supply modules to switch to a constant voltage output mode and closing a main output loop.

5. The method of claim 3, wherein, The correction of the physical impedance value of the bus bar based on the temperature data comprises: acquiring a basic impedance value of the bus bar at a standard reference temperature and a resistance temperature coefficient of the bus bar material; calculating a temperature difference value between the temperature data of the immersed medium and the standard reference temperature; based on the basic impedance value, the resistance temperature coefficient and the temperature difference value, determining an actual physical impedance value at a current temperature.

6. The method of claim 3, wherein, The determination of the target activated power supply module number according to the preset redundancy safety constraint comprises: acquiring a rated power parameter of a single power supply module; calculating a minimum power supply module number required to meet a current system total load power; determining a candidate activated number according to a preset redundancy mode and the minimum power supply module number; under the condition that the candidate activated number is less than or equal to the in-situ number, determining the candidate activated number as the target activated power supply module number.

7. The method of claim 3, wherein, The method further comprises: recording the cumulative running time of each power supply module; when the determined target activated power supply module number is less than the in-situ number, controlling the power supply module with the longest cumulative running time to enter a dormant state and controlling the power supply module with the shortest cumulative running time to enter an activated state.

8. The method of claim 3, wherein, The generation of the voltage regulation instruction based on the transmission line voltage drop comprises: determining a power supply output terminal target voltage based on a preset target load terminal voltage and the transmission line voltage drop; sending a voltage regulation instruction containing the power supply output terminal target voltage value to all the power supply modules in the activated state.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, wherein the computer program is configured to execute the method in any one of claims 1 to 5 when running. 10.An electronic device comprising a memory and a processor, the electronic device characterized by, The memory stores a computer program, and the processor is configured to execute the computer program to execute the method in any one of claims 1 to 5.

Images