An immersion liquid cooling heat dissipation regulation method and system, electronic equipment and storage medium

Abstract

The application provides an immersion liquid cooling heat dissipation regulation method and system, electronic equipment and a storage medium. The method comprises the following steps: acquiring a plurality of parameter values of a plurality of heat dissipation units in an immersion liquid cooling system in real time, constructing a to-be-trained data set, and the plurality of parameter values at least include temperature, temperature difference and inlet and outlet pressure; constructing an error function based on the flow distribution of each branch pipe and the heat transfer energy efficiency of the immersion liquid cooling system; training the to-be-trained DQN network based on the to-be-trained data set and the error function, obtaining a cold heat dissipation regulation model; and regulating the flow of each branch pipe in the immersion liquid cooling system based on the cold heat dissipation regulation model. The to-be-trained DQN network is trained based on the to-be-trained data and the error function, the cold heat dissipation regulation model is obtained, and the flow of each branch pipe is regulated in real time based on the cold heat dissipation regulation model, so that the flow pump group in the immersion liquid cooling system can be automatically adjusted according to the flow and temperature of each branch pipe.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of server heat dissipation, and more particularly to an immersion liquid cooling heat dissipation regulation method and system, an electronic device and a storage medium. BACKGROUND

[0002] With the popularization of big data technology, the complexity of high-throughput data calculation and image calculation is increasing. And due to the improvement of hardware integration, the power consumption of CPU, GPU and other components is rising, and the heat dissipation of electronic devices has become a key problem. Forced air cooling has become increasingly unable to meet the requirements of various computing environments, especially data scheduling and data calculation at the data center level. Immersion liquid cooling technology has emerged as the times require. Immersion liquid cooling technology is to immerse the server as a whole in a special liquid with high boiling point, insulation and no corrosion, and to use the liquid as a medium to transfer the heat generated by the CPU, memory, chip set, expansion card and other electronic devices in the server to the outside for heat dissipation, and then inject cold liquid into the interior for heat exchange to achieve temperature control effect.

[0003] At present, the immersion liquid cooling technology is still in the development stage, and the cooling detection method can only simply rely on the temperature data transmitted by the temperature sensor and the threshold value to make a judgment, and rely on the external pump to linearly adjust the overall flow. This will lead to energy waste, increase the burden of the pump and shorten the service life of the flow pump. Therefore, how to further improve the cooling efficiency of the immersion liquid cooling system is a problem to be solved. SUMMARY

[0004] The present application provides an immersion liquid cooling heat dissipation regulation method and system, an electronic device and a storage medium to solve the problem of how to further improve the cooling efficiency of the immersion liquid cooling system.

[0005] According to a first aspect of the present application, an immersion liquid cooling heat dissipation regulation method is provided, comprising:

[0006] Real-time acquisition of a plurality of parameter values of a plurality of heat dissipation units in an immersion liquid cooling system, construction of a to-be-trained data set, wherein the plurality of parameter values at least include temperature, temperature difference and inlet and outlet pressure;

[0007] Based on the flow distribution of each sub-pump in the immersion liquid cooling system and the heat transfer energy efficiency, an error function is constructed;

[0008] Based on the to-be-trained data set and the error function, a to-be-trained DQN network is trained to obtain a cold heat dissipation regulation model;

[0009] Based on the cold heat dissipation regulation model, the flow of each sub-pump in the immersion liquid cooling system is regulated.

[0010] On the basis of the above technical solutions, the application can also be improved as follows.

[0011] Preferably, the step of constructing the error function based on the flow distribution of each branch pipe in the immersion liquid cooling system and the heat transfer energy efficiency comprises:

[0012] Based on the single-pipe heat loss and flow at both ends of the flow pump corresponding to each branch pipe, a reward and punishment value is calculated.

[0013] Based on the reward and punishment value and the flow of the immersion liquid cooling system, an error function is constructed.

[0014] Preferably, the reward and punishment value is:

[0015] r t =∑R i =∑ΔH i / Q i ;

[0016] Wherein, i is the serial number of the branch pipe in the immersion liquid cooling system, R i is the heat transfer efficiency corresponding to the i-th branch pipe, △H i is the single-pipe heat loss corresponding to the i-th branch pipe, and Q i is the flow of the i-th branch pipe.

[0017] The error function is:

[0018] E=r t +r*Q’-Q

[0019] Wherein, r t is the reward and punishment value, r is the discount factor, Q is the network action state value, and Q' is the network action state prediction value at the previous moment.

[0020] Preferably, before the step of training the to-be-trained DQN network based on the to-be-trained data set and the error function, the method comprises:

[0021] Based on the augmented Lagrange operator and the penalty factor, a DQN network forward propagation function is constructed with the maximum heat transfer rate as the target.

[0022] Preferably, the DQN network forward propagation function is:

[0023]

[0024] Wherein, Q t is the system flow, μ is the Lagrange factor, η is the penalty factor, H is the heat loss value, min Q -H is the maximum heat transfer rate of the system flow Q t , and P maxP is the maximum power of the flow pump group of the immersion liquid cooling system max ∑L i p i L is a constraint condition i P is the flow of the i-th sub-pipe of the immersion liquid cooling system i P is the power of the i-th flow pump of the immersion liquid cooling system.

[0025] Preferably, the step of regulating the flow of each sub-pipe in the immersion liquid cooling system based on the cooling heat dissipation regulation model comprises:

[0026] When the current temperature and / or temperature rise rate of each heat dissipation unit in the immersion liquid cooling system is greater than the preset temperature threshold and the preset temperature rise rate threshold, the temperature, temperature difference and inlet and outlet pressure of the corresponding heat dissipation unit are input into the heat dissipation regulation model, so that the corresponding flow pump is cooled according to the output flow regulation information.

[0027] Preferably, the step of cooling the corresponding flow pump according to the output flow regulation information comprises:

[0028] When the temperature drop is greater than 20% of the temperature before cooling, the flow regulation information is emptied, so that the immersion liquid cooling system stops cooling.

[0029] According to a second aspect of the present application, an immersion liquid cooling heat dissipation regulation system is provided, comprising:

[0030] A data acquisition module is configured to acquire a plurality of parameter values of a plurality of heat dissipation units in an immersion liquid cooling system in real time, and construct a training data set, wherein the plurality of parameter values at least include temperature, temperature difference and inlet and outlet pressure.

[0031] An error function module is configured to construct an error function based on the flow distribution of each sub-pipe in the immersion liquid cooling system and the heat transfer energy efficiency.

[0032] A model training module is configured to train a training DQN network based on the training data set and the error function, and obtain a cooling heat dissipation regulation model.

[0033] A flow regulation module is configured to regulate the flow of each sub-pipe in the immersion liquid cooling system based on the cooling heat dissipation regulation model.

[0034] According to a third aspect of the present application, an electronic device is provided, comprising a memory and a processor, wherein the processor is configured to execute a computer management program stored in the memory to realize the steps of any of the immersion liquid cooling heat dissipation regulation methods of the first aspect.

[0035] According to a fourth aspect of the present application, a computer readable storage medium is provided, having stored thereon a computer management program which, when executed by a processor, implements the steps of any of the above-mentioned liquid immersion cooling regulation methods.

[0036] The present application provides a liquid immersion cooling regulation method, system, electronic device and storage medium, the method comprising: acquiring a plurality of parameter values of a plurality of heat dissipation units in a liquid immersion cooling system in real time, constructing a to-be-trained data set, wherein the plurality of parameter values at least include temperature, temperature difference and inlet and outlet pressure; constructing an error function based on the flow distribution of each branch pipe and the heat transfer energy efficiency in the liquid immersion cooling system; training a to-be-trained DQN network based on the to-be-trained data set and the error function to obtain a cooling heat dissipation regulation model; and regulating the flow of each branch pipe in the liquid immersion cooling system based on the cooling heat dissipation regulation model. The present application constructs a to-be-trained data set based on a plurality of parameter values of a plurality of heat dissipation units in a liquid immersion cooling system, constructs an error function based on the flow distribution of each branch pipe and the heat transfer energy efficiency in the liquid immersion cooling system, trains a to-be-trained DQN network based on the to-be-trained data and the error function, obtains a cooling heat dissipation regulation model, and regulates the flow of each branch pipe in real time based on the cooling heat dissipation regulation model, so that the flow pump group in the liquid immersion cooling system can be automatically adjusted according to the flow and temperature of each branch pipe, and since the cooling heat dissipation regulation model can obtain a suboptimal solution according to the temperature data, the adjustment mode by threshold value in the past is changed, energy waste and the burden of the flow pump group are reduced, and the service life of the flow pump is improved. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 A flow chart of the liquid immersion cooling regulation method provided by the present application is shown in the figure;

[0038] Figure 2 A structure diagram of the liquid immersion cooling system provided by the present application is shown in the figure;

[0039] Figure 3 A schematic diagram of the DQN network forward propagation iteration provided by the present application is shown in the figure;

[0040] Figure 4 A structure diagram of the liquid immersion cooling regulation system provided by the present application is shown in the figure;

[0041] Figure 5 A hardware structure diagram of a possible electronic device provided by the present application is shown in the figure;

[0042] Figure 6 A hardware structure diagram of a possible computer readable storage medium provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0043] The specific embodiments of the present application are described in further detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application but are not intended to limit the scope of the present application.

[0044] Figure 1 A flow chart of an immersion liquid cooling heat dissipation regulation method provided by the present application is shown in Figure 1 , which comprises the following steps:

[0045] Step S100: Real-time acquisition of a plurality of parameter values of a plurality of heat dissipation units in an immersion liquid cooling system, construction of a to-be-trained data set, wherein the plurality of parameter values at least include temperature, temperature difference and inlet and outlet pressure;

[0046] It should be noted that the execution subject of the method of the present embodiment can be a computer terminal device with data processing, network communication and program running functions, such as a computer, a tablet computer, etc. It can also be a server device with the same or similar functions, and it can also be a cloud server with similar functions. The present embodiment does not limit this. In order to facilitate understanding, the present embodiment and the following embodiments will be described taking a server device as an example.

[0047] It can be understood that the above-mentioned immersion liquid cooling system can be used as a set of physical devices for dissipating heat from computer hardware, and its structure is shown in Figure 2 , Figure 2 The structure diagram of the immersion liquid cooling heat dissipation system device provided by the present application; in Figure 2 , mp is the main pipe of the immersion liquid cooling heat dissipation system, dR1 is a 1 / 4 distributor, dR2 is a 1 / 9 distributor, aR1 is a 4-in-1 collector, and aR2 is a 9-in-1 collector. Each component is provided with a corresponding temperature sensor. The above-mentioned temperature sensor can have an independent calculation module, or it can be managed by a unified central processor. After receiving the temperature threshold set by the administrator, the central processor will interact with the temperature sensor corresponding to each heat dissipation unit to obtain the real-time temperature of each heat dissipation unit of the above-mentioned immersion liquid cooling system, calculate the temperature difference with the previous time point, and collect the real-time temperature and the temperature difference.

[0048] It should be understood that the plurality of parameters in the to-be-trained data set can be constructed by real-time acquisition of a plurality of parameter values in the above-mentioned immersion liquid cooling system, or it can be constructed by acquisition of historical data. The present embodiment does not limit this.

[0049] Step S200: Construction of an error function based on the flow distribution of each branch pipe and the heat transfer energy efficiency of the immersion liquid cooling system;

[0050] Understandably, constructing the error function is the core of training the neural network. After establishing the basic network and performing iterations, energy efficiency is used as a reward / penalty value based on the idea of ​​reinforcement learning. Considering the impact of the flow resistance of each pipe on the system's energy efficiency, the error function is constructed to introduce a reinforcement learning reward / penalty value r. t This causes the network to converge toward a direction with higher energy efficiency.

[0051] Furthermore, the steps for constructing the error function described above include:

[0052] Step S201: Calculate the reward / penalty value based on the heat loss and flow rate of each branch pipe at both ends of the flow pump;

[0053] Among them, the reward and punishment values ​​are:

[0054] r t =∑R i =∑ΔH i / Q i ;

[0055] Where i is the serial number of the branch pipe in the immersion liquid cooling system, R i Let ΔH be the heat transfer efficiency corresponding to the i-th branch pipe. i Q represents the heat loss of a single pipe corresponding to the i-th branch pipe. i Let i be the flow rate of the i-th branch.

[0056] Step S202: Construct an error function based on the reward / penalty value and the flow rate of the immersion liquid cooling system.

[0057] The error function is:

[0058] E = r t +r*Q'-Q

[0059] Where, r t denoted as reward / penalty value, r as discount factor, Q as network action state value, and Q' as the predicted network action state value at the previous time step.

[0060] Step S300: Based on the training dataset and the error function, train the DQN network to be trained to obtain the cold heat dissipation regulation model;

[0061] Furthermore, prior to the step of training the DQN network based on the training dataset and the error function, the following steps are included:

[0062] Step S301: Based on the augmented Lagrange operator and penalty factor, construct the forward propagation function of the DQN network with the goal of maximizing heat transfer rate.

[0063] The forward propagation function of the DQN network is:

[0064]

[0065] Among them, Q t Let μ be the system flow rate, η be the Lagrange factor, η be the penalty factor, and H be the heat loss value, min Q -H represents the system traffic Q. t The maximum heat transfer rate of the system, P max P is the maximum power of the flow pump unit of the immersion liquid cooling system. max -∑L i p i As a constraint, L i p represents the flow rate of the i-th branch pipe in the immersion liquid cooling system. i Let be the power of the i-th flow pump in the immersion liquid cooling system.

[0066] In practical implementation, the network forward design involves solving a constrained optimization problem with the goal of maximizing heat transfer rate. The augmented Lagrange multiplier method is used to transform the system transmission rate constrained optimization problem into an unconstrained problem, where the optimal problem is min... Q -H, constraint condition is P max -∑L i p i ;wherein, min Q Represents system traffic Q t The optimal solution for the system's maximum heat transfer rate, since heat is lost within the pipes, is to minimize the heat transfer rate. Q -H. Using the augmented Lagrange operator method, the multi-constraint problem is transformed into an unconstrained problem, and the partial derivatives of the augmented Lagrange equation are obtained. Substituting the penalty factor, μ is the Lagrange factor, and η is the penalty factor, we obtain the forward propagation function of the DQN network.

[0067] Furthermore, the steps for differentiating the augmented Lagrange function described above can also be found in [reference needed]. Figure 3 , Figure 3 This invention provides a schematic diagram of the forward propagation iteration of a DQN network. The derivatives of the augmented Lagrangian function for joint flow allocation L and power allocation p are calculated, and these derivatives are set as the transfer functions between deep network layers. The temperature, temperature difference, and inlet / outlet pressure information of the flow pump and its pipes are used as inputs, and the process iterates sequentially. The number of iterations determines the number of network layers. The flow distribution is initialized with a Rayleigh distribution. The constructed DNN uses the observed temperature and temperature difference as inputs and outputs the flow allocation and power allocation values ​​with the maximum heat transfer rate as the design objective. The maximum value Q (network action state value, where the maximum action reward at the policy at the maximum value Q is the suboptimal solution) represents the flow allocation as the resource allocation policy at that moment.

[0068] It is understood that the above-mentioned DNN is a deep neural network structure. The DQN network in this embodiment is based on the DNN network structure and incorporates reinforcement learning methods, hence the name DQN network.

[0069] Step S400: Adjust the flow rate of each branch pipe in the immersion liquid cooling system based on the aforementioned cold heat dissipation control model.

[0070] Furthermore, when the current temperature and / or temperature rise rate of each heat dissipation unit in the immersion liquid cooling system is greater than the preset temperature threshold and the preset temperature rise rate threshold, the temperature, temperature difference and inlet / outlet pressure of the corresponding heat dissipation unit are input into the heat dissipation control model so that the corresponding flow pump can cool down according to the output flow rate adjustment information.

[0071] In the specific implementation, the data acquisition module, mainly composed of temperature and pressure sensors, collects the temperature of each branch and the pressure at both ends, and transmits them to the embedded DQN module for processing via the data bus. The network prediction result is the branch flow allocation strategy of the joint suboptimal solution. The flow adjustment information of the flow allocation strategy is transmitted to the liquid pump via the data bus to achieve the corresponding flow control.

[0072] Furthermore, when the temperature drop is greater than 20% of the temperature before cooling, the flow regulation information is cleared so that the immersion liquid cooling system stops cooling.

[0073] In practice, temperature and temperature rise rate thresholds are set by the user. When the temperature or temperature rise rate exceeds the specified threshold, the corresponding branch data is collected and transmitted via a D / A converter to adjust the flow rate in the corresponding pipeline. Cooling stops when the temperature drops by more than 20% of the original temperature.

[0074] Understandably, based on the deficiencies in the background technology, this embodiment of the invention proposes an immersion liquid cooling heat dissipation control method. The method includes: acquiring multiple parameter values ​​of multiple heat dissipation units in an immersion liquid cooling system in real time, constructing a training dataset, wherein the multiple parameter values ​​include at least temperature, temperature difference, and inlet / outlet pressure; constructing an error function based on the flow distribution and heat transfer energy efficiency of each branch pipe in the immersion liquid cooling system; training a DQN network based on the training dataset and the error function to obtain a cold heat dissipation control model; and controlling the flow of each branch pipe in the immersion liquid cooling system based on the cold heat dissipation control model. This invention constructs a training dataset based on multiple parameter values ​​of multiple heat dissipation units in an immersion liquid cooling system, and constructs an error function based on the flow distribution and heat transfer energy efficiency of each branch pipe in the immersion liquid cooling system. The DQN network to be trained is then trained based on the training dataset and the error function to obtain a cooling and heat dissipation control model. This model is then used to control the flow rate of each branch pipe in real time, enabling the flow pump group in the immersion liquid cooling system to automatically adjust according to the flow rate and temperature of each branch pipe. Furthermore, since the cooling and heat dissipation control model can obtain suboptimal solutions based on the temperature data, it changes the previous adjustment method based on threshold judgment, reducing energy waste and the burden on the flow pump group, and extending the service life of the flow pump.

[0075] Please see Figure 4 , Figure 4 This is a schematic diagram of an immersion liquid cooling heat dissipation control system provided in an embodiment of the present invention, as shown below. Figure 4 As shown, an immersion liquid cooling heat dissipation control system includes a data acquisition module 100, an error function module 200, a model training module 300, and a flow control module 400, wherein:

[0076] The data acquisition module 100 is used to acquire multiple parameter values ​​of multiple heat dissipation units in the immersion liquid cooling system in real time and construct a training dataset. The multiple parameter values ​​include at least temperature, temperature difference, and inlet / outlet pressure. The error function module 200 is used to construct an error function based on the flow distribution and heat transfer energy efficiency of each branch pipe in the immersion liquid cooling system. The model training module 300 is used to train the DQN network to be trained based on the training dataset and the error function to obtain a cold heat dissipation control model. The flow control module 400 is used to control the flow of each branch pipe in the immersion liquid cooling system based on the cold heat dissipation control model.

[0077] It is understood that the immersion liquid cooling heat dissipation control system provided by the present invention corresponds to the immersion liquid cooling heat dissipation control method provided in the foregoing embodiments. The relevant technical features of the immersion liquid cooling heat dissipation control system can be referred to the relevant technical features of the immersion liquid cooling heat dissipation control method, and will not be repeated here.

[0078] Please seeFigure 5 , Figure 5 A schematic diagram illustrating an embodiment of the electronic device provided in this invention. For example... Figure 5 As shown, this embodiment of the invention provides an electronic device, including a memory 1310, a processor 1320, and a computer program 1311 stored in the memory 1310 and executable on the processor 1320. When the processor 1320 executes the computer program 1311, it performs the following steps:

[0079] Multiple parameter values ​​of multiple heat dissipation units in the immersion liquid cooling system are acquired in real time to construct a training dataset. The multiple parameter values ​​include at least temperature, temperature difference, and inlet / outlet pressure. An error function is constructed based on the flow distribution and heat transfer energy efficiency of each branch pipe in the immersion liquid cooling system. Based on the training dataset and the error function, the DQN network to be trained is trained to obtain a cold heat dissipation control model. The flow of each branch pipe in the immersion liquid cooling system is controlled based on the cold heat dissipation control model.

[0080] Please see Figure 6 , Figure 6 This is a schematic diagram illustrating an embodiment of a computer-readable storage medium provided by the present invention. (See diagram below.) Figure 6 As shown, this embodiment provides a computer-readable storage medium 1400, on which a computer program 1411 is stored. When the computer program 1411 is executed by a processor, it performs the following steps:

[0081] Multiple parameter values ​​of multiple heat dissipation units in the immersion liquid cooling system are acquired in real time to construct a training dataset. The multiple parameter values ​​include at least temperature, temperature difference, and inlet / outlet pressure. An error function is constructed based on the flow distribution and heat transfer energy efficiency of each branch pipe in the immersion liquid cooling system. Based on the training dataset and the error function, the DQN network to be trained is trained to obtain a cold heat dissipation control model. The flow of each branch pipe in the immersion liquid cooling system is controlled based on the cold heat dissipation control model.

[0082] This invention provides a method, system, electronic device, and storage medium for controlling immersion liquid cooling. The method includes: acquiring multiple parameter values ​​of multiple heat dissipation units in an immersion liquid cooling system in real time to construct a training dataset, wherein the multiple parameter values ​​include at least temperature, temperature difference, and inlet / outlet pressure; constructing an error function based on the flow distribution and heat transfer energy efficiency of each branch pipe in the immersion liquid cooling system; training a DQN network based on the training dataset and the error function to obtain a cooling control model; and controlling the flow of each branch pipe in the immersion liquid cooling system based on the cooling control model. This invention constructs a training dataset based on multiple parameter values ​​of multiple heat dissipation units in an immersion liquid cooling system, and constructs an error function based on the flow distribution and heat transfer energy efficiency of each branch pipe in the immersion liquid cooling system. The DQN network to be trained is then trained based on the training dataset and the error function to obtain a cooling and heat dissipation control model. This model is then used to control the flow rate of each branch pipe in real time, enabling the flow pump group in the immersion liquid cooling system to automatically adjust according to the flow rate and temperature of each branch pipe. Furthermore, since the cooling and heat dissipation control model can obtain suboptimal solutions based on the temperature data, it changes the previous adjustment method based on threshold judgment, reducing energy waste and the burden on the flow pump group, and extending the service life of the flow pump.

[0083] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0084] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0087] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0088] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0089] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for regulating immersion liquid cooling heat dissipation, characterized in that, The method includes: In real time, multiple parameter values ​​of multiple heat dissipation units in an immersion liquid cooling system are acquired to construct a training dataset. The multiple parameter values ​​include at least temperature, temperature difference, and inlet / outlet pressure. Based on the flow distribution and heat transfer efficiency of each branch pipe in the immersion liquid cooling system, an error function is constructed, including: Based on the heat loss and flow rate of each pipe at both ends of the flow pump in the immersion liquid cooling system, calculate the reward / penalty value: ； Where i is the serial number of the branch pipe in the immersion liquid cooling system, R i Let ΔH be the heat transfer efficiency corresponding to the i-th branch pipe. i Q represents the heat loss of a single pipe corresponding to the i-th branch pipe. i Let i be the flow rate of the i-th branch. Based on the reward / penalty value and the flow rate of the immersion liquid cooling system, an error function is constructed: ； Where, r t , where r is the reward / penalty value, r is the discount factor, Q is the network action state value, and Q' is the predicted network action state value at the previous time step; Based on the training dataset and the error function, the DQN network to be trained is trained to obtain the cold heat dissipation regulation model. The flow rate of each branch pipe in the immersion liquid cooling system is regulated based on the aforementioned cold heat dissipation control model.

2. The immersion liquid cooling heat dissipation control method according to claim 1, characterized in that, Before the step of training the DQN network based on the training dataset and the error function, the following steps are included: Based on the augmented Lagrange operator and the penalty factor, a forward propagation function for the DQN network is constructed with the goal of maximizing the heat transfer rate.

3. The immersion liquid cooling heat dissipation control method according to claim 2, characterized in that, The forward propagation function of the DQN network is: ； in, Q t For system traffic, μ For Lagrange factors, As a penalty factor, H This represents the heat loss value. Indicates system traffic Regarding heat loss value H The function of influence, the objective of the forward propagation function of the DQN network, is to influence the system traffic. The maximum heat transfer rate of the system is achieved by minimizing the heat loss value H, expressed as: min Q - H;P max The maximum power of the flow pump set of the immersion liquid cooling system. As constraints, L i For the immersion liquid cooling system i The flow rate of each branch pipe p i For the immersion liquid cooling system i The power of each flow pump.

4. The immersion liquid cooling heat dissipation control method according to claim 1, characterized in that, The step of regulating the flow rate of each branch pipe in the immersion liquid cooling system based on the cold heat dissipation regulation model includes: When the current temperature and / or temperature rise rate of each heat dissipation unit in the immersion liquid cooling system is greater than the preset temperature threshold and the preset temperature rise rate threshold, the temperature, temperature difference and inlet / outlet pressure of the corresponding heat dissipation unit are input into the heat dissipation control model so that the corresponding flow pump can cool down according to the output flow rate adjustment information.

5. The immersion liquid cooling heat dissipation control method according to claim 4, characterized in that, After the step of cooling the corresponding flow pump according to the output flow rate adjustment information, the following steps are included: When the temperature drop is greater than 20% of the temperature before cooling, the flow regulation information is cleared so that the immersion liquid cooling system stops cooling.

6. An immersion liquid cooling heat dissipation control system, characterized in that, include The data acquisition module is used to acquire multiple parameter values ​​of multiple heat dissipation units in the immersion liquid cooling system in real time and construct a training dataset. The multiple parameter values ​​include at least temperature, temperature difference and inlet / outlet pressure. The error function module is used to construct an error function based on the flow distribution and heat transfer efficiency of each branch pipe in the immersion liquid cooling system; it includes: Based on the heat loss and flow rate of each pipe at both ends of the flow pump in the immersion liquid cooling system, calculate the reward / penalty value: ； in, i The serial number of the branch pipe in the immersion liquid cooling system. R i For the first i The heat transfer efficiency corresponding to each branch pipe △H i For the first i The heat loss per pipe corresponds to the heat loss of each branch pipe. Q i For the first i The flow rate of each branch; Based on the reward / penalty value and the flow rate of the immersion liquid cooling system, an error function is constructed: ； in, r t As a reward and punishment value, r As a discount factor, Q This represents the network action state value. Q’ The predicted value of the network action state at the previous moment; The model training module is used to train the DQN network to be trained based on the training dataset and the error function to obtain the cooling and heat dissipation regulation model. The flow control module is used to control the flow rate of each branch pipe in the immersion liquid cooling system based on the cold heat dissipation control model.

7. An electronic device, characterized in that, It includes a memory and a processor, wherein the processor is used to implement the steps of the immersion liquid cooling heat dissipation control method as described in any one of claims 1-5 when executing a computer management program stored in the memory.

8. A computer-readable storage medium, characterized in that, It stores a computer management program, which, when executed by a processor, implements the steps of the immersion liquid cooling heat dissipation control method as described in any one of claims 1-5.

Images