Immersed liquid cooling cabinet system and control method thereof

By predicting and adjusting the data to be processed in the immersed liquid-cooling cabinet system, precise heat dissipation control and energy consumption optimization of each server are achieved, and the problems of inaccurate heat dissipation control and high energy consumption in the prior art are solved.

CN118732802BActive Publication Date: 2024-11-26四川华鲲振宇智能科技有限责任公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411225234.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-11-26
Estimated Expiration
2044-09-03

AI Technical Summary

Technical Problem

The existing immersion liquid-cooled cabinet system cannot independently adapt to the different heat dissipation needs of each server, resulting in inaccurate heat dissipation control, slow response speed, and low energy consumption and efficiency.

Method used

By obtaining the historical processing data of the server, creating a pending data prediction model, predicting the data type and quantity, and pre-allocating the water pump speed of the refrigerator based on the prediction data, adjusting the water pump speed in real time to adapt to the server's cooling needs, achieving precise control and response speed promote.

Benefits of technology

It realizes the adaptation to the independent heat dissipation needs of each server, improves the response speed of temperature regulation and overall energy consumption benefits, and reduces the risk of liquid leakage, blockage and loss of the cooling medium circulation system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118732802B_ABST
    Figure CN118732802B_ABST
Patent Text Reader

Abstract

The present invention relates to an immersion liquid cooling cabinet system and a control method thereof, which obtains historical processing data of a server within a specified period; creates a prediction model for data to be processed, inputs historical processing data into the prediction model for data to be processed for training, and predicts the data to be processed based on the prediction model for data to be processed; presets the water pump speed of the corresponding refrigerator in the liquid cooling cabinet; collects the data to be processed in real time, and compares the data to be processed with the predicted processing data, distributes the real-time collected data to be processed to the corresponding server according to the pre-allocation, and makes the water pump speed of the corresponding refrigerator run at a preset value, adjusts the processing data of the corresponding server according to the comparison result between the data to be processed and the predicted processing data, and adjusts the water pump speed of the refrigerator according to the adjusted processing data. This allows the system to independently adapt to the different heat dissipation requirements of each server and realizes precise control of server heat dissipation.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of server immersion liquid cooling, and in particular relates to an immersion liquid cooling cabinet system and a control method thereof. Background Art

[0002] Immersion liquid cooling systems are increasingly used in data centers. Immersion liquid cooling is to immerse the electronic products (servers) that need to be cooled and dissipated in non-conductive liquid as a whole, and remove the heat generated by the components during operation through the flow of liquid, so as to achieve the purpose of heat dissipation and energy saving. Generally, it includes three parts: chiller, CDU (mainly including heat exchanger and liquid pump), tank (mainly liquid cooling server cabinet). The cooling medium relies on the liquid pump in the CDU to circulate between the CDU heat exchanger and the tank to remove heat. Due to the large heat capacity of the liquid, the flow rate of this liquid in the tank is actually very slow. Theoretically calculated, the average flow rate of the flow provided by the liquid pump in the tank is about 0.002m / s, which is almost negligible. In fact, the flow rate of the liquid in the tank is mainly generated by the buoyancy generated by the liquid after the heat exchange and temperature rise with the server components. Therefore, this liquid pump has almost no effect on the heat dissipation of the server components in the immersion liquid cooling system, and is only used to circulate and transport liquid. The heat exchange between the server and the liquid in the tank mainly relies on natural convection.

[0003] Since the temperature and flow rate of the cooling medium in the cabinet are uniformly controlled by the CDU, it is impossible to independently adapt to the different cooling requirements of each server, and it is difficult to achieve accurate control of server cooling. In addition, since a variety of server devices with different power consumption will be configured in the tank, and different servers will have different power consumption scenarios, relying solely on the external CDU to control the temperature of the entire cabinet will not only have a slow response speed, but also have low energy efficiency for the entire cabinet system.

[0004] Therefore, how to improve the existing immersion liquid cooling cabinet system to independently adapt to the different heat dissipation requirements of each server, achieve precise control of server heat dissipation, improve the response speed of temperature control of the entire cabinet, and improve the energy efficiency of the entire cabinet system is a technical problem that needs to be solved urgently. Summary of the invention

[0005] The purpose of the present invention is to provide an immersion liquid cooling cabinet system and a control method thereof, so as to improve the existing immersion liquid cooling cabinet system so as to independently adapt to the different heat dissipation requirements of each server, realize precise control of server heat dissipation, improve the response speed of temperature regulation of the whole cabinet, and improve the energy efficiency of the whole cabinet system.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, a control method for an immersion liquid cooling cabinet system is provided, comprising the following steps:

[0008] S1: Obtaining historical processing data of each server in the immersion liquid cooling cabinet during a specified period, wherein the historical processing data includes the type and quantity of processing data in each time period;

[0009] S2: creating a prediction model for data to be processed, inputting the acquired historical processing data of the server within a specified period into the prediction model for data to be processed, training the model, and predicting the data to be processed based on the prediction model for data to be processed, wherein the prediction content includes the type and quantity of the processed data;

[0010] S3: pre-allocating the predicted data to be processed according to the servers in the immersion liquid cooling cabinet, pre-allocating the predicted data to be processed to one or more servers, and presetting the water pump speed of the corresponding refrigerator in the liquid cooling cabinet based on the type and amount of the processed data;

[0011] S4: collect the data to be processed in real time, and compare the data to be processed with the predicted processing data to determine whether there is a difference between the two. If not, distribute the real-time collected data to be processed to the corresponding server according to the pre-allocation, and make the water pump speed of the corresponding refrigerator run at a preset value. If so, execute step S5;

[0012] S5: adjusting the processing data of the corresponding server according to the comparison result between the data to be processed and the predicted processing data, and adjusting the speed of the water pump of the refrigerator according to the adjusted processing data, and the water pump of the refrigerator operates based on the adjusted water pump speed;

[0013] S6: Continuously monitor the data to be processed, and execute steps S4 and S5 in a loop.

[0014] Preferably, step S1 includes the following specific process:

[0015] S11: Obtain historical processing data of each server in the immersion liquid cooling cabinet within a specified period, and summarize the historical processing data of all servers within the specified period, where the specified period is at least one month;

[0016] S12: Count the types and corresponding quantities of historical processing data within a specified period, where the statistical time unit is 24 hours, and extract the type characteristics and quantity characteristics of the historical processing data every 24 hours.

[0017] Preferably, step S2 includes the following specific process:

[0018] S21: dividing the historical processing data with 24 hours as the statistical time unit into a training data set, a test data set and a verification data set, creating a prediction model for the data to be processed, and inputting the training data set into the prediction model for the data to be processed;

[0019] S22: input the test data set into the trained prediction model for the data to be processed, and determine whether the trained prediction model for the data to be processed meets the preset requirements. If so, execute step S23; if not, re-input a new training set for training;

[0020] S23: input the verification data set into the tested data prediction model to further verify the test results, and determine whether the test results meet the preset requirements. If so, execute step S24; if not, re-execute step S21;

[0021] S24: Predict the type and quantity of the data to be processed based on the verified prediction model for the data to be processed, and output the prediction result.

[0022] Preferably, step S3 includes the following specific process:

[0023] S31: classifying the predicted data to be processed according to the importance level into important data and non-important data, and allocating the important data to the designated server in the immersion liquid cooling cabinet, and presetting the water pump speed of the refrigerator corresponding to the designated server processing the important data;

[0024] S32: Distribute the non-important data to other servers according to quantity, and preset the water pump speed of the refrigerator corresponding to the server processing the non-important data.

[0025] Preferably, step S5 includes the following specific process:

[0026] Adjust the processing data of the corresponding server according to the comparison result between the data to be processed and the predicted processing data. When the real-time collected data to be processed is equal to the predicted processing data, directly distribute it to the corresponding server according to the pre-allocated content, and make the refrigerator run according to the preset water pump speed;

[0027] When the real-time collected data to be processed is more than the predicted processing data, the data that is the same as the predicted processing data is directly allocated to the corresponding server according to the pre-allocated content, and the remaining real-time collected data to be processed is redistributed. According to the result of the redistribution, the processing data of the corresponding server is adjusted, so that the speed of the water pump of the corresponding refrigerator is increased by a fixed value, and the water pump of the refrigerator operates based on the adjusted water pump speed;

[0028] When the data to be processed collected in real time is less than the predicted processing data, the processing data less than the predicted data will be directly allocated to the corresponding server according to the pre-allocated content, and the processing data of the corresponding server will be adjusted to reduce the water pump speed of the corresponding refrigerator by a fixed value, and the water pump of the refrigerator will operate based on the adjusted water pump speed.

[0029] Preferably, the temperature of the server is continuously monitored. When the temperature at a certain location on the server is detected to be higher than the maximum value of a preset temperature range, the water pump speed of the corresponding refrigerator is increased by a fixed value. When the temperature at a certain location on the server is detected to be lower than the minimum value of the preset temperature range, the water pump speed of the corresponding refrigerator is reduced by a fixed value.

[0030] Preferably, the conductivity of the cooling medium is continuously monitored, and when it is detected that the conductivity of the cooling medium exceeds a preset value, an alarm is processed through an alarm module.

[0031] In a second aspect, an immersion liquid cooling cabinet system is provided, which is used to implement any one of the immersion liquid cooling cabinet control methods described above, including a liquid cooling cabinet, wherein a plurality of accommodating parts are provided in the liquid cooling cabinet, wherein the accommodating parts are used to accommodate servers and refrigerators, wherein the refrigerators include a heat exchange row, a data acquisition module, a water pump and a control module, wherein the top of the heat exchange row is respectively connected to a water inlet pipe and a water outlet pipe, and the bottom of the heat exchange row is connected to the water pump, wherein the servers and refrigerators are both immersed in a cooling medium;

[0032] The data acquisition module is connected to the control module and includes a temperature sensor and a conductivity sensor. The temperature sensor is used to measure the temperature of the cooling medium and the temperature of the water respectively; the conductivity sensor is used to detect the conductivity of the cooling medium;

[0033] The refrigerator is used to drive the water circulation inside the heat exchange row based on the operation of the water pump to take away the heat of the cooling medium near the refrigerator in the cabinet, so as to achieve rapid cooling of the server;

[0034] The control module is used to control the data acquisition module and the refrigerator.

[0035] Preferably, the alarm module is connected to the control module, and the alarm module is used to perform alarm processing when it is detected that the conductivity of the cooling medium exceeds a preset value.

[0036] Preferably, it further comprises a copper busbar power connector, and the refrigerator draws power from the copper busbar inside the liquid cooling cabinet through the copper busbar power connector.

[0037] The beneficial effects of the present invention include:

[0038] The immersion liquid cooling cabinet system and control method provided by the present invention obtain historical processing data of a server within a specified period; create a prediction model for data to be processed, input historical processing data into the prediction model for data to be processed for training, and predict the data to be processed based on the prediction model for data to be processed; preset the water pump speed of the corresponding refrigerator in the liquid cooling cabinet; collect the data to be processed in real time, and compare the data to be processed with the predicted processing data, distribute the real-time collected data to be processed to the corresponding server according to the pre-allocation, and make the water pump speed of the corresponding refrigerator run at a preset value, adjust the processing data of the corresponding server according to the comparison result of the data to be processed and the predicted processing data, and adjust the water pump speed of the refrigerator according to the adjusted processing data.

[0039] First, by creating a prediction model for data to be processed and predicting the data to be processed based on historical processing data, and pre-allocating the processed data based on the predicted data, on the one hand, the efficiency of subsequent allocation of data to be processed can be improved; on the other hand, the water pump speed of the refrigerator pre-allocated to the corresponding server can be preset to improve the response speed of temperature control of the entire cabinet.

[0040] Secondly, by collecting the data to be processed in real time and comparing the data to be processed with the predicted processing data, the real-time collected data to be processed is allocated to the corresponding server according to the pre-allocation, and the water pump speed of the corresponding refrigerator is operated at the preset value. According to the comparison result of the data to be processed and the predicted processing data, the processing data of the corresponding server is adjusted in real time, and the water pump speed of the refrigerator is adjusted according to the adjusted processing data, so as to independently adapt to the different heat dissipation requirements of each server and realize precise control of server heat dissipation.

[0041] Finally, by eliminating the external CDU (including the heat exchanger and liquid pump), the refrigeration system is made into a standard server form, which can be flexibly configured to the immersion liquid cooling cabinet system at different positions inside the liquid-cooled server cabinet. The closed liquid cooling cabinet system has no liquid inlet and outlet, and the cooling medium circulates in the cabinet without external force. There is no need to use an external pump to drive the coolant circulation of the entire cabinet for heat dissipation. The heat dissipation is completely dependent on the coolant for heat exchange circulation between the server and the refrigerator inside the cabinet, which reduces the system energy consumption and avoids the risks of leakage, blockage, and loss caused by the flow of cooling medium in the pipeline. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is a structural schematic diagram of the immersion liquid cooling cabinet system of the present invention.

[0043] Figure 2 It is a schematic diagram of the structure of the refrigerator in the immersion liquid cooling cabinet system of the present invention.

[0044] Figure 3 It is a schematic flow chart of a control method of an immersion liquid cooling cabinet system of the present invention. DETAILED DESCRIPTION

[0045] The following is combined with Figure 1~Figure 3 The present invention is further described in detail:

[0046] Example 1

[0047] See attached Figure 3 As shown, a control method for an immersion liquid cooling cabinet system comprises the following steps:

[0048] S1: Obtain the historical processing data of each server in the immersion liquid cooling cabinet within a specified period, wherein the historical processing data includes the type and quantity of data processed in each time period. The reason for obtaining the historical processing data within the specified period is that for the server, the type and quantity of data processed are the same within a unit of time, such as every day. Therefore, the historical processing data within the past specified period can be used to find the pattern of the type and quantity of data that the server needs to process.

[0049] S2: Create a prediction model for data to be processed, and input the historical processing data of the server obtained within a specified period into the prediction model for data to be processed, train it, and predict the data to be processed based on the prediction model for data to be processed. The prediction content includes the type and quantity of the processed data. In step S1, it is mentioned that the data processed by the server has certain rules. Therefore, by creating a prediction model for data to be processed, the rules are extracted based on the historical processing data, and then the data to be processed is predicted according to the rules of the obtained historical processing data. The predicted data explains the data that the server is about to process, providing a data basis for subsequent pre-allocation.

[0050] S3: Pre-allocate the predicted data to be processed according to the servers in the immersion liquid cooling cabinet, pre-allocate the predicted data to be processed to one or more servers, and preset the water pump speed of the corresponding refrigerator in the liquid cooling cabinet based on the type and quantity of the processed data. The pre-allocation is based on the predicted processed data in step S3. The reason for pre-allocation is that if the data type and quantity are identified after the data to be processed comes out, and then the data type and quantity are allocated based on the identification results, and then the speed of the water pump of the refrigerator of the corresponding server is set based on the allocation result, the response speed of the temperature control of the whole cabinet is low, and the energy efficiency of the whole cabinet system is low. If the data to be processed can be predicted in advance, the predicted data to be processed can be pre-allocated, and the water pump speed of the corresponding refrigerator can be preset based on the pre-allocation, the response speed of the temperature control of the whole cabinet can be effectively improved, and the energy efficiency of the whole cabinet system can be improved.

[0051] S4: collect the data to be processed in real time, and compare the data to be processed with the predicted processing data to determine whether there is a difference between the two. If not, distribute the real-time collected data to be processed to the corresponding server according to the pre-allocation, and make the water pump speed of the corresponding refrigerator run at the preset value. If so, execute step S5. In the actual working process, the difference between the predicted data to be processed and the data to be processed is not big, but there may be special circumstances at certain times, such as the increase or decrease of the amount of data to be processed, or the existence of other data types that need special processing. Therefore, when there are special circumstances, it is necessary to make adjustments through the subsequent step S5.

[0052] S5: adjusting the processing data of the corresponding server according to the comparison result between the data to be processed and the predicted processing data, and adjusting the speed of the water pump of the refrigerator according to the adjusted processing data, and the water pump of the refrigerator operates based on the adjusted water pump speed;

[0053] S6: Continuously monitor the data to be processed, and execute steps S4 and S5 in a loop.

[0054] Example 2

[0055] Based on Example 1, step S1 includes the following specific process:

[0056] S11: Obtain the historical processing data of each server in the immersion liquid cooling cabinet within a specified period, and summarize the historical processing data of all servers within the specified period. The specified period is at least one month. When obtaining the historical processing data, sufficient historical data is required so that the regular characteristics of the historical data obtained therefrom are universal. If the time of the historical data obtained is too short, the regular characteristics of the historical data obtained may be accidental, and the subsequent prediction results of the data to be processed will be inaccurate. Therefore, in this application, it takes at least one month of historical processing data for each server in the immersion liquid cooling cabinet to avoid the accidental nature of the historical processing data and improve the accuracy of subsequent predictions.

[0057] S12: Count the types and corresponding quantities of historical processing data within a specified period, wherein the statistical time unit is 24 hours, and extract the type characteristics and quantity characteristics of the historical processing data every 24 hours. Since the server obviously has periodicity in the process of processing data, and this period basically cycles in 24 hours, the present invention, in the process of counting the historical processing data within the specified period, uses 24 hours as the time unit for statistics, and extracts the type characteristics and quantity characteristics of the historical processing data every 24 hours.

[0058] Step S2 includes the following specific processes:

[0059] S21: Divide the historical processing data with 24 hours as the statistical time unit into a training data set, a test data set and a verification data set, create a prediction model for the data to be processed, and input the training data set into the prediction model for the data to be processed. The training data set, the test data set and the verification data set are the data basis for the created prediction model for the data to be processed to have accurate prediction function. The prediction model for the data to be processed is trained with the training data set to have a certain prediction ability. The test data set tests the prediction ability of the trained prediction model for the data to be processed to determine whether it has the prediction ability that meets the requirements. The verification data set further verifies the prediction model for the data to be processed after the test to ensure that it has the corresponding prediction ability.

[0060] S22: input the test data set into the trained prediction model for the data to be processed, and determine whether the trained prediction model for the data to be processed meets the preset requirements. If so, execute step S23; if not, re-input a new training set for training;

[0061] S23: input the verification data set into the tested data prediction model to further verify the test results, and determine whether the test results meet the preset requirements. If so, execute step S24; if not, re-execute step S21;

[0062] S24: Predict the type and quantity of the data to be processed based on the verified prediction model for the data to be processed, and output the prediction result. The output prediction result of the data to be processed provides a data basis for subsequent pre-allocation.

[0063] In this embodiment, the prediction model for the data to be processed adopts a neural network model. The specific structure of the neural network model of this embodiment includes an input layer, a hidden layer, and an output layer. The hidden layer includes three convolutional layers, two pooling layers, and a fully connected layer. The input layer is a 32×32 structure, and the structures of the three convolutional layers are 28×28, 10×10, and 120, respectively. The structures of the two pooling layers are 14×14 and 5×5. The input layer is responsible for receiving the original data, the hidden layer is the core part of the neural network, responsible for processing and converting the input data, and the output layer outputs the data processed and converted by the hidden layer. In addition, the hidden layer also includes an activation function. The activation function in this embodiment is a Sigmoid function, which maps the output of the neuron to between 0 and 1. Weights and biases are preset in the connection between the input layer, the hidden layer, and the output layer to adjust through the learning process to minimize the prediction error.

[0064] Step S3 includes the following specific processes:

[0065] S31: classifying the predicted data to be processed according to the importance level into important data and non-important data, and allocating the important data to the designated server in the immersion liquid cooling cabinet, and presetting the water pump speed of the refrigerator corresponding to the designated server processing the important data;

[0066] S32: Distribute the non-important data to other servers according to quantity, and preset the water pump speed of the refrigerator corresponding to the server processing the non-important data.

[0067] The reason why the data to be processed should be classified according to the level of importance is that for the data to be processed, there may be important data or special data, and these important data or special data are classified as important data. Such data may need to be processed first or processed using a specific server. Other data is relatively not that special or important, so other data to be processed is classified as non-important data. The important data is then assigned to designated servers in the immersion liquid cooling cabinet. Different servers may have different performance or different processors. Therefore, for important data, it is assigned to designated servers with specific processing capabilities for processing. Other non-important data to be processed is assigned to other servers for processing according to quantity.

[0068] Example 3

[0069] Based on Example 1 or Example 2, step S5 includes the following specific process:

[0070] The processing data of the corresponding server is adjusted according to the comparison result between the data to be processed and the predicted processing data. When the real-time collected data to be processed is equal to the predicted processing data, it is directly allocated to the corresponding server according to the pre-allocated content, and the refrigerator is operated according to the preset water pump speed. The comparison between the data to be processed and the predicted processing data is to know the difference between the predicted processing data and the actual data to be processed. When the data to be processed is the same as the predicted processing data, it is directly allocated to the corresponding server according to the pre-allocated content, and the corresponding refrigerator of the server is operated according to the preset water pump speed.

[0071] When the real-time collected data to be processed is more than the predicted processing data, the data that is the same as the predicted processing data is directly allocated to the corresponding server according to the pre-allocated content, and the remaining real-time collected data to be processed is redistributed. According to the result of the redistribution, the processing data of the corresponding server is adjusted so that the water pump speed of the corresponding refrigerator is increased by a fixed value, and the water pump of the refrigerator runs based on the adjusted water pump speed. There may be a gap between the data to be processed and the predicted processing data, so the part with the difference needs to be reallocated, and the time for reallocation of the difference part is also less than the time for identifying the type and quantity of all the data to be processed.

[0072] When the real-time collected data to be processed is less than the predicted processing data, the processed data less than the predicted data will be directly allocated to the corresponding server according to the pre-allocated content, and the processing data of the corresponding server will be adjusted to reduce the water pump speed of the corresponding refrigerator by a fixed value, and the water pump of the refrigerator will operate based on the adjusted water pump speed. In the actual operation process, the data to be processed is generally predicted 24 hours in advance, which effectively improves the data allocation efficiency and the response efficiency of the refrigerator of the liquid cooling cabinet.

[0073] Continuously monitor the temperature of the server. When the temperature of a certain part of the server is higher than the maximum value of the preset temperature range, the corresponding refrigerator water pump speed is increased by a fixed value. When the temperature of a certain part of the server is lower than the minimum value of the preset temperature range, the corresponding refrigerator water pump speed is reduced by a fixed value. Continuously monitor the conductivity of the cooling medium. When the conductivity of the cooling medium is detected to exceed the preset value, an alarm is processed through the alarm module.

[0074] See also Figure 1 and Figure 2 The above-mentioned, an immersion liquid cooling cabinet system is used to implement any one of the above-mentioned immersion liquid cooling cabinet control methods, including a liquid cooling cabinet, a plurality of accommodating parts are provided in the liquid cooling cabinet, the accommodating parts are used to accommodate servers and refrigerators, the refrigerators include a heat exchange row, a data acquisition module, a water pump and a control module, the top of the heat exchange row is respectively connected to a water inlet pipe and a water outlet pipe, the bottom of the heat exchange row is connected to the water pump, and the server and the refrigerator are both immersed in the cooling medium. The water inlet and the water outlet are respectively connected to the heat exchange row, and the heat exchange row is connected to the water pump. The water circuit inside the heat exchange row is driven by the operation of the water pump to circulate the coolant temperature near the refrigerator in the cabinet, so as to achieve rapid cooling of the server.

[0075] The data acquisition module is connected to the control module, and includes a temperature sensor and a conductivity sensor. The temperature sensor is used to measure the temperature of the cooling medium and the temperature of the water respectively; the conductivity sensor is used to detect the conductivity of the cooling medium; the refrigerator is used to drive the water circulation inside the heat exchange row based on the operation of the water pump to take away the heat of the cooling medium near the refrigerator in the cabinet, so as to achieve rapid cooling of the server. The control module is used to control the data acquisition module and the refrigerator. The alarm module is connected to the control module, and the alarm module is used to perform alarm processing when it is detected that the conductivity of the cooling medium exceeds the preset value. It also includes a copper busbar power connector, and the refrigerator draws power from the copper busbar inside the liquid cooling cabinet through the copper busbar power connector. The connector is connected to the power conversion board connector, and the water pump, monitoring and sensor inside the refrigerator are powered through the power conversion board.

[0076] Since the liquid cooling cabinet is closed and has no inlet and outlet, the coolant circulates in the cabinet without external force. It only relies on the temperature change caused by heat exchange with the server and the refrigerator for internal circulation, which can avoid the risks of leakage, blockage, loss, etc. caused by the flow of coolant in the pipeline. The appearance, structure and installation method of the refrigerator in the cabinet are consistent with the server (the height U number of the equipment will be different). The installation position of the refrigerator is flexible and can be configured with the server in 1:1 or 1:n according to actual needs. The servers are connected through signal lines, and the cooler and the corresponding server are connected through the panel interface. The inlet and outlet are set on the refrigerator panel, which are connected to the water cooling circulation system pipeline through the inlet and outlet pipes respectively. The inlet and outlet pipes are buried inside the cabinet, one end is connected to the underground pipeline, and the other end is reserved for the joint to connect to the inlet and outlet of the refrigerator.

[0077] The control module is connected to the conductivity sensor to obtain the conductivity information of the cabinet coolant. The water temperature sensor is set on the heat exchanger to monitor the outlet water temperature in the heat exchanger. The control module is connected to the water temperature sensor to obtain the outlet water temperature data. The monitoring board is connected to the water pump, and the speed of the water pump can be adjusted through the control signal.

[0078] In summary, the immersion liquid cooling cabinet system and control method thereof provided by the present invention obtain historical processing data of a server within a specified period; create a prediction model for data to be processed, input historical processing data into the prediction model for data to be processed for training, and predict the data to be processed based on the prediction model for data to be processed; preset the water pump speed of the corresponding refrigerator in the liquid cooling cabinet; collect data to be processed in real time, and compare the data to be processed with the predicted processing data, allocate the real-time collected data to be processed to the corresponding server according to the pre-allocation, and make the water pump speed of the corresponding refrigerator run at a preset value, adjust the processing data of the corresponding server according to the comparison result of the data to be processed and the predicted processing data, and adjust the water pump speed of the refrigerator according to the adjusted processing data.

[0079] By creating a prediction model for data to be processed, predicting the data to be processed based on historical processing data, and pre-allocating the processed data based on the predicted data, on the one hand, the efficiency of subsequent data allocation can be improved, and on the other hand, the water pump speed of the refrigerator pre-allocated to the corresponding server can be preset to improve the response speed of the temperature control of the entire cabinet. By collecting data to be processed in real time and comparing the data to be processed with the predicted processing data, the real-time collected data to be processed is allocated to the corresponding server according to the pre-allocation, and the water pump speed of the corresponding refrigerator is operated at the preset value. The processing data of the corresponding server is adjusted in real time according to the comparison results of the data to be processed and the predicted processing data, and the water pump speed of the refrigerator is adjusted according to the adjusted processing data, so as to achieve independent adaptation to the different heat dissipation requirements of each server and achieve precise control of server heat dissipation. By eliminating the external heat exchanger and liquid pump, the refrigeration system is made into a standard server form. It can be flexibly configured to the immersion liquid cooling cabinet system at different positions inside the liquid-cooled server cabinet. The closed liquid cooling cabinet system has no liquid inlet and outlet. The cooling medium circulates in the cabinet without external force. There is no need to use an external pump to drive the coolant circulation of the entire cabinet for heat dissipation. The heat dissipation is completely dependent on the coolant for heat exchange circulation between the server and the refrigerator inside the cabinet, which reduces the system energy consumption and avoids the risks of leakage, blockage, and loss caused by the flow of cooling medium in the pipeline.

Claims

1. A control method for an immersion liquid cooling cabinet system, characterized in that: The following steps are involved: S1: Obtaining historical processing data of each server in the immersion liquid cooling cabinet during a specified period, wherein the historical processing data includes the type and quantity of processing data in each time period; S2: creating a prediction model for data to be processed, inputting the acquired historical processing data of the server within a specified period into the prediction model for data to be processed, training the model, and predicting the data to be processed based on the prediction model for data to be processed, wherein the prediction content includes the type and quantity of the processed data; S3: pre-allocating the predicted data to be processed according to the servers in the immersion liquid cooling cabinet, pre-allocating the predicted data to be processed to one or more servers, and presetting the water pump speed of the corresponding refrigerator in the liquid cooling cabinet based on the type and amount of the processed data; S4: collect the data to be processed in real time, and compare the data to be processed with the predicted processing data to determine whether there is a difference between the two. If not, distribute the real-time collected data to be processed to the corresponding server according to the pre-allocation, and make the water pump speed of the corresponding refrigerator run at a preset value. If so, execute step S5; S5: adjusting the processing data of the corresponding server according to the comparison result between the data to be processed and the predicted processing data, and adjusting the speed of the water pump of the refrigerator according to the adjusted processing data, and the water pump of the refrigerator operates based on the adjusted water pump speed; S6: Continuously monitor the data to be processed, and execute steps S4 and S5 in a loop.

2. The control method of an immersion liquid cooling cabinet system according to claim 1, characterized in that: Step S1 includes the following specific processes: S11: Obtain historical processing data of each server in the immersion liquid cooling cabinet within a specified period, and summarize the historical processing data of all servers within the specified period, where the specified period is at least one month; S12: Count the types and corresponding quantities of historical processing data within a specified period, where the statistical time unit is 24 hours, and extract the type characteristics and quantity characteristics of the historical processing data every 24 hours.

3. The control method of an immersion liquid cooling cabinet system according to claim 2, characterized in that: Step S2 includes the following specific processes: S21: dividing the historical processing data with 24 hours as the statistical time unit into a training data set, a test data set and a verification data set, creating a prediction model for the data to be processed, and inputting the training data set into the prediction model for the data to be processed; S22: input the test data set into the trained prediction model for the data to be processed, and determine whether the trained prediction model for the data to be processed meets the preset requirements. If so, execute step S23; if not, re-input a new training set for training; S23: input the verification data set into the tested data prediction model to further verify the test results, and determine whether the test results meet the preset requirements. If so, execute step S24; if not, re-execute step S21; S24: Predict the type and quantity of the data to be processed based on the verified prediction model for the data to be processed, and output the prediction result.

4. The control method of an immersion liquid cooling cabinet system according to claim 1, characterized in that: Step S3 includes the following specific processes: S31: classifying the predicted data to be processed according to the importance level into important data and non-important data, and allocating the important data to the designated server in the immersion liquid cooling cabinet, and presetting the water pump speed of the refrigerator corresponding to the designated server processing the important data; S32: Distribute the non-important data to other servers according to the quantity, and preset the water pump speed of the refrigerator corresponding to the server processing the non-important data.

5. The control method of an immersion liquid cooling cabinet system according to claim 1, characterized in that: Step S5 includes the following specific processes: Adjust the processing data of the corresponding server according to the comparison result between the data to be processed and the predicted processing data. When the real-time collected data to be processed is equal to the predicted processing data, directly distribute it to the corresponding server according to the pre-allocated content, and make the refrigerator run according to the preset water pump speed; When the real-time collected data to be processed is more than the predicted processing data, the data that is the same as the predicted processing data is directly allocated to the corresponding server according to the pre-allocated content, and the remaining real-time collected data to be processed is redistributed. According to the result of the redistribution, the processing data of the corresponding server is adjusted, so that the speed of the water pump of the corresponding refrigerator is increased by a fixed value, and the water pump of the refrigerator operates based on the adjusted water pump speed; When the data to be processed collected in real time is less than the predicted processing data, the processing data less than the predicted data will be directly allocated to the corresponding server according to the pre-allocated content, and the processing data of the corresponding server will be adjusted to reduce the water pump speed of the corresponding refrigerator by a fixed value, and the water pump of the refrigerator will operate based on the adjusted water pump speed.

6. The control method of an immersion liquid cooling cabinet system according to claim 1, characterized in that: The temperature of the server is continuously monitored. When the temperature of a certain part of the server is higher than the maximum value of the preset temperature range, the water pump speed of the corresponding refrigerator is increased by a fixed value. When the temperature of a certain part of the server is lower than the minimum value of the preset temperature range, the water pump speed of the corresponding refrigerator is reduced by a fixed value.

7. The control method of an immersion liquid cooling cabinet system according to claim 6, characterized in that: The electrical conductivity of the cooling medium is continuously monitored. When it is detected that the electrical conductivity of the cooling medium exceeds a preset value, an alarm is processed through the alarm module.

8. An immersion liquid cooling cabinet system, used to implement an immersion liquid cooling cabinet control method according to any one of claims 1 to 7, comprising a liquid cooling cabinet, characterized in that: The liquid cooling cabinet is provided with a plurality of accommodating parts, which are used to accommodate servers and refrigerators. The refrigerator includes a heat exchange row, a data acquisition module, a water pump and a control module. The top of the heat exchange row is respectively connected with a water inlet pipe and a water outlet pipe, and the bottom of the heat exchange row is connected with the water pump. The server and refrigerator are immersed in the cooling medium. The data acquisition module is connected to the control module and includes a temperature sensor and a conductivity sensor. The temperature sensor is used to measure the temperature of the cooling medium and the temperature of the water respectively; the conductivity sensor is used to detect the conductivity of the cooling medium; The refrigerator is used to drive the water circulation inside the heat exchange row based on the operation of the water pump to take away the heat of the cooling medium near the refrigerator in the cabinet, so as to achieve rapid cooling of the server; The control module is used to control the data acquisition module and the refrigerator.

9. The immersion liquid cooling cabinet system according to claim 8, characterized in that: It also includes an alarm module, which is connected to the control module and is used to perform an alarm process when it is detected that the conductivity of the cooling medium exceeds a preset value.

10. The immersion liquid cooling cabinet system according to claim 8, characterized in that: It also includes a copper busbar power connector, through which the refrigerator draws power from the copper busbar inside the liquid cooling cabinet.