Liquid cooling server cooling liquid intelligent distribution method and system

Abstract

The application discloses a liquid cooling server cooling liquid intelligent distribution method and system, and belongs to the technical field of liquid cooling heat dissipation.The application first completes the construction of a circulating system with cooling liquid compatibility verification, collects and pre-processes the operation parameters of the liquid cooling server whole cabinet in real time, constructs a heat load prediction model by using a time sequence neural network algorithm, realizes intelligent control of a cooling liquid distribution unit, operation abnormality detection and response based on model analysis results, simultaneously completes model iterative optimization by means of operation data feedback, and carries out server node adaptive load adjustment in combination with system operation monitoring results.The application can realize intelligent and accurate distribution of cooling liquid, effectively improve the operation stability and overall energy efficiency of the liquid cooling system, and adapt to the dynamic cooling demand of high-density liquid cooling servers.

Description

Technical Field

[0001] This invention relates to the field of liquid cooling technology, and in particular to a method and system for intelligent distribution of coolant in a liquid-cooled server. Background Technology

[0002] With the rapid development of cloud computing, artificial intelligence, and high-performance computing, data centers are gradually evolving towards high density and ultra-large scale. The thermal power consumption of server chips continues to increase, and the heat exchange capacity of traditional air-cooling technology is no longer sufficient to meet the heat dissipation requirements of high-density computing scenarios. Liquid cooling technology, with its higher heat exchange efficiency and lower energy consumption, has become the core direction of data center heat dissipation technology development. Currently, liquid cooling technology has achieved large-scale commercial application in high-density data centers, forming three main technical routes: cold plate, spray, and immersion. Each technical route has developed relatively mature supporting systems and application solutions. Among them, the coolant distribution unit, as the core equipment connecting the external cooling cycle and the internal cooling loop of the cabinet in the liquid cooling system, undertakes the core functions of coolant flow distribution, temperature regulation, and pressure control, and is a key link in the stable operation of the liquid cooling system. Industry research and development on liquid cooling systems is gradually moving towards intelligent management and full-process optimization. Combining data acquisition and algorithm analysis to achieve refined management of liquid cooling systems has become an important research direction in the field of liquid cooling technology. The iteration and optimization of related technologies continue to drive the improvement of the energy efficiency level of data center cooling systems.

[0003] Currently, in practical applications, liquid cooling systems mostly employ a passive adjustment mode based on real-time operating parameters for coolant distribution and management. This approach fails to anticipate and proactively adapt to changes in server heat load, resulting in a significant lag between coolant supply and actual server heat load demands. This hinders precise and intelligent coolant allocation. Furthermore, existing liquid cooling systems lack a comprehensive, interconnected operational management mechanism, making it difficult to identify and respond quickly to abnormal states during system operation. This leaves room for improvement in system stability and reliability. In addition, existing liquid cooling systems lack pre-construction verification and control of coolant and electronic component compatibility. They also struggle to adaptively adjust server node loads based on real-time system status, failing to optimize the matching of cooling resources and heat load distribution. It also hinders iterative optimization of control strategies based on continuously collected operational data. Consequently, the cooling efficiency of the liquid cooling system is not fully realized, and the optimization and management of operational energy consumption are significantly limited, failing to fully adapt to the dynamically changing cooling needs of high-density liquid-cooled servers. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for intelligent distribution of coolant in liquid-cooled servers.

[0005] The objective of this invention is achieved through the following technical solution: A method for intelligent distribution of coolant in a liquid-cooled server is provided, the method comprising the following steps: S1. Build a coolant circulation system. First, conduct coolant compatibility verification to confirm the compatibility between the coolant and electronic components. Then, connect various circulation components to complete the coolant circulation system and debug it to an operational state. S2. Collect and preprocess operating parameters. Collect operating parameters in real time for the liquid-cooled server cabinet corresponding to the completed coolant circulation system, perform preprocessing operations on the collected operating parameters, and classify and store the preprocessed operating parameters in the data collection system. S3. Construct a heat load prediction model and intelligently control the coolant distribution unit. Use a time-series neural network algorithm to process, analyze and predict the trends of the preprocessed operating parameters stored in the classification. Combine the analysis and prediction results to intelligently control the coolant distribution unit. At the same time, carry out anomaly detection and response on the operating status of the coolant circulation system. Collect the operating parameters in the anomaly detection and response process, optimize the heat load prediction model with the help of the feedback mechanism, and update it regularly. S4. Implement monitoring, early warning, and adaptive load adjustment. Based on the analysis and prediction results of the heat load prediction model and the anomaly detection and response data, continuously monitor the operating performance of the liquid-cooled server rack, analyze the operating parameters obtained by monitoring and detect potential problems, provide timely feedback on the monitoring and analysis results, and dynamically adjust the load of each server node in the liquid-cooled server rack based on the monitoring and analysis results.

[0006] Furthermore, step S1 includes the following sub-steps: S1.1. Conduct coolant compatibility verification, test and analyze various properties of the coolant, including thermophysical properties, chemical stability, thermal stability, corrosivity, boiling point characteristics and latent heat of vaporization, and confirm the compatibility between the coolant and electronic components based on the performance test and analysis results of the coolant. S1.2. Connect the circulation components and seal them sequentially. The circulation components include the cooling tower, heat exchanger, coolant pump, coolant distribution unit, coolant delivery pipeline and liquid-cooled server cabinet. S1.3. Complete the commissioning of the coolant circulation system, introduce coolant into the connected coolant circulation system, check the flow status of coolant between each circulation component, and adjust the operating components of the coolant circulation system to a stable operating state.

[0007] Furthermore, step S2 includes the following sub-steps: S2.1. Collect operating parameters. Set up parameter collection points inside the liquid-cooled server cabinet to collect various operating parameters of the liquid-cooled server cabinet in real time, including temperature, humidity, current and voltage. S2.2. Preprocessing operating parameters: Perform various preprocessing operations on the collected operating parameters, including noise reduction, normalization, and invalid data removal, to obtain valid operating parameters from the collected operating parameters; S2.3. Categorize and store operating parameters. Classify the valid operating parameters according to their parameter types and store the classified valid operating parameters in the storage area of ​​the data collection system.

[0008] Furthermore, step S3 includes the following sub-steps: S3.1. Extract operating parameter features, use a time-series neural network algorithm to parse the preprocessed operating parameters of the classified storage, and extract various feature information that can reflect the overall operating status of the liquid-cooled server cabinet from the parsed preprocessed operating parameters; S3.2. Analyze and predict the operating status. Based on the extracted feature information, conduct trend analysis of the operating parameters of the liquid-cooled server cabinet and predict the future operating status to form the analysis and prediction results of the operating parameters. S3.3. Intelligent control and anomaly detection and response: Adjust the operating parameters of the coolant distribution unit based on the analysis and prediction results of the operating parameters, conduct anomaly detection and response on the overall operating status of the coolant circulation system, and implement corresponding response actions for various detected anomalies. S3.4. Optimize and update the heat load prediction model, collect various operating parameters in the prediction process and anomaly detection and response process, import the collected operating parameters into the heat load prediction model with the help of the feedback mechanism, optimize the internal parameters of the heat load prediction model, and complete the update of the heat load prediction model at fixed intervals.

[0009] Furthermore, step S4 includes the following sub-steps: S4.1. Continuously monitor the operating performance. Based on the analysis and prediction results of the heat load prediction model and the anomaly detection and response data, continuously monitor the overall operating performance of the liquid-cooled server cabinet and obtain the real-time operating parameters of the liquid-cooled server cabinet. S4.2. Analyze, detect, and report problems. Perform data parsing and operational status analysis on the real-time acquired operating parameters. Detect potential problems in the operation of the liquid-cooled server cabinet from the status analysis results. Generate feedback information for the detected potential problems and push it to the monitoring terminal. S4.3. Dynamically adjust the load of server nodes. Extract relevant data reflecting the thermal load status of the entire liquid-cooled server cabinet from the analysis results of real-time operating parameters, and dynamically adjust the load of each server node in the entire liquid-cooled server cabinet in combination with the relevant thermal load status data.

[0010] Furthermore, in step S3, the various control components of the coolant distribution unit are adjusted based on the analysis and prediction results of the operating parameters. The control components include flow control valves and temperature regulating components. At the same time, the flow control parameters and temperature control parameters of the coolant distribution unit are automatically adjusted according to the real-time operating status of the coolant circulation system.

[0011] Furthermore, in step S3, various historical operating parameters during the operation of the coolant circulation system are continuously collected. These historical operating parameters include operating parameter acquisition data, analysis and prediction result data, and anomaly detection and response data. The various historical operating parameters are classified and organized according to the operating stage. By combining the classified and organized historical operating parameters with the real-time collected operating parameters, the analysis of the heat load change trend of the liquid-cooled server cabinet and the prediction of future heat load status are carried out.

[0012] Furthermore, in step S4, the operating parameters acquired by real-time monitoring are compared and analyzed with the historical operating parameters stored in the data collection system. Based on the comparison and analysis results, the changing trend of the operating status of the liquid-cooled server cabinet is identified. Combined with the changing trend of the operating status, various judgment criteria for system monitoring and early warning are dynamically adjusted. The judgment criteria include parameter threshold criteria and anomaly identification criteria.

[0013] Furthermore, in step S1, the coolant distribution unit in the coolant circulation system is integrated and configured, and various functional components are assembled as a whole. The functional components include heat exchangers, coolant pumps, main filters, bypass filters and model operation microcontrollers. The control module of the coolant distribution unit is configured using frequency conversion control technology, and the frequency conversion adjustment of various operating parameters of the coolant distribution unit is realized through the control module.

[0014] A liquid-cooled server coolant intelligent distribution system is provided, which includes a coolant circulation component, a coolant compatibility verification module, a sensor acquisition module, a data collection system, a heat load prediction model analysis and control module, a monitoring and early warning module, a load adjustment module, and a coolant delivery pipeline. The coolant circulation system comprises various circulation components, including a cooling tower, heat exchanger, coolant pump, and coolant distribution unit. The coolant distribution unit integrates various functional components, including a heat exchanger, coolant pump, main filter, bypass filter, and model operation microcontroller unit. The coolant distribution unit is also equipped with a variable frequency control module. A sensor acquisition module is deployed within the liquid-cooled server cabinet to collect various operating parameters of the entire cabinet in real time. A coolant compatibility verification module is used to test and analyze various performance characteristics of the coolant and confirm its compatibility with electronic components. A data collection system stores pre-processed valid operating parameters and various data related to the coolant circulation system's operation. Historical operating parameters; the heat load prediction model analysis and control module incorporates a time-series neural network algorithm to extract operating parameter features, analyze and predict the operating status of the liquid-cooled server cabinet, and intelligently control the coolant distribution unit. It also optimizes and updates the heat load prediction model using a feedback mechanism. The monitoring and early warning module continuously monitors the operating performance of the liquid-cooled server cabinet, detects potential problems, and pushes feedback information. The load adjustment module dynamically adjusts the load of each server node within the liquid-cooled server cabinet based on relevant heat load data. Coolant delivery pipelines connect the coolant circulation components to the liquid-cooled server cabinet, enabling coolant flow and transmission within the system. Signal interconnection is achieved between all modules and components.

[0015] The beneficial effects of this invention are: (1) A heat load prediction model is constructed by using a time-series neural network algorithm to realize the intelligent control of the coolant distribution unit. Combined with the monitoring of the entire process operation status and the adaptive load adjustment of the server node, the coolant supply and the changes in server heat load are accurately matched. (2) Conduct integrated configuration and frequency conversion control design for the coolant distribution unit, and verify the compatibility between coolant and electronic components to simplify the overall structure of the liquid cooling system and reduce the risk of leakage and component damage during long-term operation of the equipment; (3) The heat load prediction model is iteratively optimized through the continuous feedback mechanism of operating data. Combined with the full-process anomaly detection and dynamic early warning standard adjustment, the stability of the liquid cooling system operation and the overall energy efficiency level are effectively improved. Attached Figure Description

[0016] Figure 1 A flowchart illustrating the steps of an intelligent distribution method for coolant in a liquid-cooled server. Figure 2 The flowchart illustrates the specific steps of an intelligent distribution method for coolant in a liquid-cooled server, as provided in this embodiment. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1 See Figure 1 This embodiment provides a method for intelligent distribution of coolant in a liquid-cooled server, which includes the following steps: S1. Build a coolant circulation system. First, conduct coolant compatibility verification to confirm the compatibility between the coolant and electronic components. Then, connect various circulation components to complete the coolant circulation system and debug it to an operational state. S2. Collect and preprocess operating parameters. Collect operating parameters in real time for the liquid-cooled server cabinet corresponding to the completed coolant circulation system, perform preprocessing operations on the collected operating parameters, and classify and store the preprocessed operating parameters in the data collection system. S3. Construct a heat load prediction model and intelligently control the coolant distribution unit. Use a time-series neural network algorithm to process, analyze and predict the trends of the preprocessed operating parameters stored in the classification. Combine the analysis and prediction results to intelligently control the coolant distribution unit. At the same time, carry out anomaly detection and response on the operating status of the coolant circulation system. Collect the operating parameters in the anomaly detection and response process, optimize the heat load prediction model with the help of the feedback mechanism, and update it regularly. S4. Implement monitoring, early warning, and adaptive load adjustment. Based on the analysis and prediction results of the heat load prediction model and the anomaly detection and response data, continuously monitor the operating performance of the liquid-cooled server rack, analyze the operating parameters obtained by monitoring and detect potential problems, provide timely feedback on the monitoring and analysis results, and dynamically adjust the load of each server node in the liquid-cooled server rack based on the monitoring and analysis results.

[0019] In some embodiments, step S1 includes the following sub-steps: S1.1. Conduct coolant compatibility verification, test and analyze various properties of the coolant, including thermophysical properties, chemical stability, thermal stability, corrosivity, boiling point characteristics and latent heat of vaporization, and confirm the compatibility between the coolant and electronic components based on the performance test and analysis results of the coolant. S1.2. Connect the circulation components and seal them sequentially. The circulation components include the cooling tower, heat exchanger, coolant pump, coolant distribution unit, coolant delivery pipeline and liquid-cooled server cabinet. S1.3. Complete the commissioning of the coolant circulation system, introduce coolant into the connected coolant circulation system, check the flow status of coolant between each circulation component, and adjust the operating components of the coolant circulation system to a stable operating state.

[0020] In some embodiments, step S2 includes the following sub-steps: S2.1. Collect operating parameters. Set up parameter collection points inside the liquid-cooled server cabinet to collect various operating parameters of the liquid-cooled server cabinet in real time, including temperature, humidity, current and voltage. S2.2. Preprocessing operating parameters: Perform various preprocessing operations on the collected operating parameters, including noise reduction, normalization, and invalid data removal, to obtain valid operating parameters from the collected operating parameters; S2.3. Categorize and store operating parameters. Classify the valid operating parameters according to their parameter types and store the classified valid operating parameters in the storage area of ​​the data collection system.

[0021] In some embodiments, step S3 includes the following sub-steps: S3.1. Extract operating parameter features, use a time-series neural network algorithm to parse the preprocessed operating parameters of the classified storage, and extract various feature information that can reflect the overall operating status of the liquid-cooled server cabinet from the parsed preprocessed operating parameters; S3.2. Analyze and predict the operating status. Based on the extracted feature information, conduct trend analysis of the operating parameters of the liquid-cooled server cabinet and predict the future operating status to form the analysis and prediction results of the operating parameters. S3.3. Intelligent control and anomaly detection and response: Adjust the operating parameters of the coolant distribution unit based on the analysis and prediction results of the operating parameters, conduct anomaly detection and response on the overall operating status of the coolant circulation system, and implement corresponding response actions for various detected anomalies. S3.4. Optimize and update the heat load prediction model, collect various operating parameters in the prediction process and anomaly detection and response process, import the collected operating parameters into the heat load prediction model with the help of the feedback mechanism, optimize the internal parameters of the heat load prediction model, and complete the update of the heat load prediction model at fixed intervals.

[0022] In some embodiments, step S4 includes the following sub-steps: S4.1. Continuously monitor the operating performance. Based on the analysis and prediction results of the heat load prediction model and the anomaly detection and response data, continuously monitor the overall operating performance of the liquid-cooled server cabinet and obtain the real-time operating parameters of the liquid-cooled server cabinet. S4.2. Analyze, detect, and report problems. Perform data parsing and operational status analysis on the real-time acquired operating parameters. Detect potential problems in the operation of the liquid-cooled server cabinet from the status analysis results. Generate feedback information for the detected potential problems and push it to the monitoring terminal. S4.3. Dynamically adjust the load of server nodes. Extract relevant data reflecting the thermal load status of the entire liquid-cooled server cabinet from the analysis results of real-time operating parameters, and dynamically adjust the load of each server node in the entire liquid-cooled server cabinet in combination with the relevant thermal load status data.

[0023] In some embodiments, in step S3, the various control components of the coolant distribution unit are adjusted based on the analysis and prediction results of the operating parameters. The control components include flow control valves and temperature regulating components. At the same time, the flow control parameters and temperature control parameters of the coolant distribution unit are automatically adjusted according to the real-time operating status of the coolant circulation system.

[0024] In some embodiments, in step S3, various historical operating parameters during the operation of the coolant circulation system are continuously collected. The historical operating parameters include operating parameter acquisition data, analysis and prediction result data, and anomaly detection and response data. The various historical operating parameters are classified and organized according to the operating stage. The classified and organized historical operating parameters are combined with the real-time collected operating parameters to jointly analyze the trend of heat load change of the liquid-cooled server cabinet and predict the future heat load status.

[0025] In some embodiments, in step S4, the operating parameters acquired by real-time monitoring are compared and analyzed with the historical operating parameters stored in the data collection system. Based on the comparison and analysis results, the operating status change trend of the entire liquid-cooled server cabinet is identified. The various judgment criteria of the system monitoring and early warning are dynamically adjusted in combination with the operating status change trend. The judgment criteria include parameter threshold criteria and anomaly identification criteria.

[0026] In some embodiments, in step S1, the coolant distribution unit in the coolant circulation system is integrated and configured, and various functional components are assembled as one unit. The functional components include a heat exchanger, a coolant pump, a main filter, a bypass filter, and a model running microcontroller unit. The control module of the coolant distribution unit is configured using frequency conversion control technology, and the frequency conversion adjustment of various operating parameters of the coolant distribution unit is realized through the control module.

[0027] A liquid-cooled server coolant intelligent distribution system is provided, which includes a coolant circulation component, a coolant compatibility verification module, a sensor acquisition module, a data collection system, a heat load prediction model analysis and control module, a monitoring and early warning module, a load adjustment module, and a coolant delivery pipeline. The coolant circulation system comprises various circulation components, including a cooling tower, heat exchanger, coolant pump, and coolant distribution unit. The coolant distribution unit integrates various functional components, including a heat exchanger, coolant pump, main filter, bypass filter, and model operation microcontroller unit. The coolant distribution unit is also equipped with a variable frequency control module. A sensor acquisition module is deployed within the liquid-cooled server cabinet to collect various operating parameters of the entire cabinet in real time. A coolant compatibility verification module is used to test and analyze various performance characteristics of the coolant and confirm its compatibility with electronic components. A data collection system stores pre-processed valid operating parameters and various data related to the coolant circulation system's operation. Historical operating parameters; the heat load prediction model analysis and control module incorporates a time-series neural network algorithm to extract operating parameter features, analyze and predict the operating status of the liquid-cooled server cabinet, and intelligently control the coolant distribution unit. It also optimizes and updates the heat load prediction model using a feedback mechanism. The monitoring and early warning module continuously monitors the operating performance of the liquid-cooled server cabinet, detects potential problems, and pushes feedback information. The load adjustment module dynamically adjusts the load of each server node within the liquid-cooled server cabinet based on relevant heat load data. Coolant delivery pipelines connect the coolant circulation components to the liquid-cooled server cabinet, enabling coolant flow and transmission within the system. Signal interconnection is achieved between all modules and components.

[0028] Example 2 This embodiment provides a specific implementation process for an intelligent coolant distribution method for liquid-cooled servers. Through a complete process including circulation system setup and debugging, operational parameter acquisition and processing, heat load prediction and intelligent control, and operational monitoring and load adjustment, intelligent distribution of coolant in liquid-cooled servers is achieved, ensuring the stable and efficient operation of the liquid cooling system. Figure 2 As shown, the specific implementation is as follows: S1. Setting up and debugging the cyclic system: This step provides the basic circulation and heat exchange conditions for the operation of the liquid cooling system, and includes the following sub-steps: S1.1. Coolant compatibility verification: Coolant compatibility refers to the characteristic that when the coolant comes into contact with electronic components inside the liquid-cooled server, it will not cause adverse effects such as corrosion, swelling, or decreased insulation performance on the electronic components. In this embodiment, compatibility verification ensures the long-term stability of the liquid cooling system. This step tests and analyzes various properties of the coolant, including thermophysical properties, chemical stability, thermal stability, corrosivity, boiling point characteristics, and latent heat of vaporization. Among them, thermophysical properties refer to properties related to heat transfer, such as thermal conductivity and specific heat capacity, which directly affect the heat exchange efficiency of the coolant; chemical stability refers to the characteristic that the coolant will not undergo chemical reactions such as decomposition or polymerization during long-term circulation, ensuring the long-term stability of the coolant's performance; thermal stability refers to the characteristic that the coolant's performance will not deteriorate due to temperature increases within a range of temperature changes; corrosivity refers to the degree of corrosion that occurs when the coolant comes into contact with the metallic or non-metallic materials of electronic components; boiling point characteristics refer to the boiling temperature range of the coolant within the set operating pressure range, ensuring that the coolant maintains a stable liquid state during operation; latent heat of vaporization refers to the heat absorbed or released when the coolant undergoes a phase change, affecting the coolant's heat exchange capacity. The compatibility between the coolant and electronic components is confirmed based on the performance test and analysis results of the coolant. Only coolants that meet the compatibility requirements can be used for the subsequent construction of the circulation system.

[0029] In some specific implementations, during coolant compatibility verification, existing accelerated aging tests can be used to conduct matching verification. Accelerated aging tests refer to existing testing methods that shorten the aging cycle of coolant contact with electronic components by increasing the temperature, pressure, and other conditions of the test environment, thereby quickly obtaining performance change data after long-term contact. In the implementation of this embodiment, the electronic component sample to be verified is completely immersed in the coolant to be tested. Accelerated aging test conditions conforming to industry standards are set, and immersion tests are continuously carried out for a set period. After the test cycle, the appearance, weight, mechanical properties, and insulation properties of the electronic component sample are tested. At the same time, the composition and performance changes of the coolant are tested. The matching between the coolant and the electronic component is determined based on the test results, shortening the compatibility verification cycle and improving verification efficiency.

[0030] S1.2. Sealed connection of the circulation component: The circulation components refer to all the equipment and parts that constitute the coolant circulation path, including cooling towers, heat exchangers, coolant pumps, coolant distribution units, coolant delivery pipelines, and the liquid-cooled server cabinet. This step involves sequentially sealing and connecting these circulation components. Following the heat exchange sequence of the coolant, the connection is made first between the cooling tower and the heat exchanger, then between the heat exchanger and the coolant pump, then between the coolant pump and the coolant distribution unit, then between the coolant distribution unit and the coolant delivery pipeline, and finally between the coolant delivery pipeline and the liquid-cooled server cabinet. Each connection point is sealed during the connection process to ensure no coolant leakage during circulation and to guarantee a smooth circulation path.

[0031] In some specific implementations, during the sealing connection of the circulation component, the existing double-compression fitting sealing connection structure can be used to connect the pipeline and the component. This double-compression fitting sealing connection structure is an existing pipeline connection structure that achieves sealing and fixation of the pipeline and joint through the clamping action of two compression fittings, and features vibration resistance, leak prevention, and convenient assembly and disassembly. In the implementation of this embodiment, a suitable double-compression fitting is selected according to the pipe diameter and material of the coolant delivery pipeline. After pretreatment at the pipeline end, the double-compression fitting is assembled with the corresponding interface of the pipeline and circulation component. After assembly, a pressure test is performed on the connection. During the pressure test, pressure changes and leakage at the connection are monitored to ensure the sealing performance and structural strength of the connection, adapt to pressure fluctuations during coolant circulation, and reduce the risk of leakage during long-term operation.

[0032] S1.3. Integrated configuration of coolant distribution unit: The coolant distribution unit is a device in the liquid cooling system responsible for coolant flow distribution, temperature regulation, and impurity filtration. In this embodiment, it is used to achieve precise control of the coolant delivery process and ensure stable operation of the cooling cycle. Integrated configuration refers to integrating various functional components of the coolant distribution unit into a single device structure, reducing the space occupied and connection points of the equipment, and lowering the risk of leakage. This step integrates the coolant distribution unit in the coolant circulation system, assembling various functional components into a single unit. These functional components include a heat exchanger, a coolant pump, a main filter, a bypass filter, and a model operation microcontroller unit. The heat exchanger facilitates heat exchange between the coolant and the external cooling medium, reducing the coolant temperature; the coolant pump provides power for the coolant circulation; the main filter is connected in series in the main coolant circulation path to filter solid particulate impurities in the coolant; the bypass filter is connected in parallel in the coolant circulation path to perform bypass filtration of the coolant, removing liquid impurities; and the model operation microcontroller unit is used to run the heat load prediction model, receive operating parameter data, and output control commands. This step uses variable frequency control technology to configure the control module of the coolant distribution unit. Variable frequency control technology is an existing electrical control technology that adjusts the operating speed of equipment by changing the power supply frequency. In this embodiment, it is used to adjust the operating speed of the coolant pump, so as to realize the variable frequency adjustment of various operating parameters of the coolant distribution unit and adapt to different cooling needs.

[0033] S1.4. Overall debugging of the circulation system: This step involves introducing coolant into the connected coolant circulation system, checking the flow of coolant between the various circulation components, and inspecting for any issues such as poor flow or leaks. The operating components of the coolant circulation system are then tested, including the operation of the coolant pump, the opening and closing of the valves, and the flow of the filter. The coolant circulation system is then brought to a stable operating state, providing the necessary heat exchange conditions for the subsequent operation of the liquid-cooled server.

[0034] In some embodiments, during the overall commissioning of the circulation system, temporary detection points can be added in the coolant circulation path to simultaneously detect the flow rate, pressure, and temperature parameters of the coolant. Based on the detection results, the operating status of the circulation components can be adjusted to shorten the commissioning cycle.

[0035] S2. Operation Parameter Acquisition and Processing: This step provides standardized data support for subsequent heat load prediction and intelligent control, and includes the following sub-steps: S2.1. Layout of data collection points and data acquisition for operating parameters: Operating parameters refer to various data that reflect the overall operating status of the liquid-cooled server rack and the working status of the cooling system, including temperature, humidity, current, and voltage. This step involves deploying parameter acquisition points within the liquid-cooled server rack. These points cover the heat-generating element area, the coolant inlet / outlet area, and the internal environment of the rack. Real-time acquisition of various operating parameters of the liquid-cooled server rack is conducted to ensure the continuity and integrity of the collected data. The acquired operating parameters are labeled chronologically to form a time-series data sequence.

[0036] S2.2. Preprocessing of runtime parameters: This step involves various preprocessing operations on the collected operating parameters, including denoising, normalization, and invalid data removal. Denoising removes noise data caused by environmental interference and equipment fluctuations, retaining only valid data that accurately reflects the equipment's operating status. Normalization converts operating parameters with different dimensions and value ranges into values ​​within a unified standard range, eliminating dimensional differences and facilitating subsequent algorithm analysis. Invalid data removal removes missing values ​​and outliers exceeding reasonable ranges generated during the acquisition process, ensuring data validity. Through these preprocessing operations, valid operating parameters are obtained from the collected parameters.

[0037] In some embodiments, during the preprocessing of runtime parameters, a sliding window filter can be used for data denoising to improve the efficiency and effectiveness of denoising.

[0038] S2.3. Categorized storage of operating parameters: This step categorizes valid operating parameters according to their type, including environmental, electrical, and thermal parameters. The categorized valid operating parameters are then stored in the corresponding storage area of ​​the data collection system. The data collection system refers to an existing data management system with data storage and retrieval functions. In this embodiment, it is used to store various types of data generated during the operation of the liquid cooling system, providing data support for subsequent algorithm analysis and model execution.

[0039] In some specific implementations, during the classification and storage of operational parameters, existing time-series databases can be used to manage the storage of collected operational parameters. Time-series databases are existing database systems specifically optimized for time-series data, featuring high write speeds, high compression ratios, and fast time-series data retrieval, thus meeting the storage requirements of continuously collected time-series operational parameters in this embodiment. In the implementation of this embodiment, the preprocessed valid operational parameters are structured according to the collection timestamp and parameter type, and then written into the corresponding data sequence in the time-series database. Simultaneously, data storage periods and archiving rules are set, and historical data exceeding the storage period is archived and compressed for storage, ensuring the orderliness of data storage and improving data retrieval efficiency, providing rapid data support for subsequent model analysis and trend prediction.

[0040] S3. Heat load prediction and intelligent control: This step uses a time-series neural network algorithm to predict the heat load in advance, and then uses the prediction results to achieve intelligent control of the coolant distribution unit. It is the core execution process of intelligent coolant distribution and includes the following sub-steps: S3.1. Feature extraction of runtime parameters: This step utilizes a temporal neural network algorithm to analyze the preprocessed operating parameters stored in the categorized database. Temporal neural networks are existing deep learning algorithms for processing time-series data, possessing the ability to capture long-term and short-term dependencies in time-series data and effectively extracting the changing patterns in continuously collected operating parameters. From the analyzed preprocessed operating parameters, various feature information reflecting the overall operating status of the liquid-cooled server rack is extracted. This feature information includes parameter change trends, inter-parameter correlations, and parameter fluctuations.

[0041] In some embodiments, during the feature extraction process, principal component analysis (PCA) can be used to reduce the dimensionality of the extracted features, thereby reducing redundant features and improving the efficiency of subsequent analysis. PCA is an existing statistical algorithm for data dimensionality reduction and feature extraction, capable of converting multiple correlated feature variables into a few uncorrelated principal component variables.

[0042] S3.2. Operational Status Analysis and Trend Prediction: This step analyzes the changing trends of the liquid-cooled server rack's operating parameters and predicts its future operating status based on the extracted feature information, generating analysis and prediction results for the operating parameters. The extracted feature information is integrated and processed through the fully connected layer of the heat load prediction model, outputting the analysis results of the changing trends of the liquid-cooled server rack's heat load, as well as the prediction results of its operating status within a set future time period. The analysis and prediction results include the upward or downward trend of heat load, changes in coolant demand, and the direction of change in equipment operating status, providing data support for subsequent coolant distribution unit control.

[0043] The heat load prediction model comprises four layers: an input layer, a time-series feature extraction layer, a fully connected layer, and an output layer, all connected in series. The input layer receives preprocessed time-series operating parameter data; the time-series feature extraction layer employs a gated recurrent unit structure of a time-series neural network to extract multi-dimensional features from the input time-series data, capturing the correlation patterns of parameter changes over time; the fully connected layer performs nonlinear mapping and integration of the extracted feature information, eliminating redundant features; and the output layer outputs the predicted heat load change trend and the probability data of operational anomalies.

[0044] The training steps for the heat load prediction model include: First, constructing a training dataset, which includes historical operating parameter data and corresponding actual heat load change data; Second, setting key model training parameters, including learning rate, number of iterations, batch size, and loss function, using the mean squared error function; Third, dividing the training dataset into training and validation sets proportionally, and inputting them into the model for iterative training; Fourth, validating the accuracy of the trained model using the validation set, adjusting the model's internal parameters, and completing the model training. The model's input data consists of the operating parameters of the liquid-cooled server cabinet collected at fixed time intervals, and the output data includes the predicted heat load change trend over a future set time period, suggested values ​​for adjusting the operating parameters of the coolant distribution unit, and the probability value of abnormal operating conditions.

[0045] S3.3. Intelligent control and anomaly detection response of the coolant distribution unit: This step adjusts the operating parameters of the coolant distribution unit based on the analysis and prediction results of the operating parameters, performs anomaly detection and response on the overall operating status of the coolant circulation system, and implements corresponding response actions for various detected anomalies. Based on the analysis and prediction results, various control components of the coolant distribution unit are adjusted. These control components include flow control valves and temperature regulation components. The flow control valves regulate the flow rate of the coolant, and the temperature regulation components regulate the output temperature of the coolant. Simultaneously, the flow control parameters and temperature control parameters of the coolant distribution unit are automatically adjusted according to the real-time operating status of the coolant circulation system to match the coolant supply with the heat load changes of the liquid-cooled server. Anomalies in the operating status of the coolant circulation system are detected using the anomaly probability data output by the heat load prediction model. Anomalies include coolant leakage, flow abnormalities, pressure abnormalities, and temperature abnormalities. For each detected anomaly, corresponding response actions are implemented, including adjusting the operating parameters of the coolant distribution unit, issuing anomaly warning messages, and shutting down the circulation path of the corresponding faulty area to ensure the safety of system operation.

[0046] S3.4. Optimization and Update of Heat Load Prediction Model: This step collects various operational parameters during the analysis and prediction process and the anomaly detection and response process. Using a feedback mechanism, the collected operational parameters are imported into the heat load prediction model. The internal parameters of the heat load prediction model are optimized, and the model is updated periodically. Historical operational parameters are continuously collected during the operation of the coolant circulation system. These historical parameters include collected data, analysis and prediction results, and anomaly detection and response data. These historical parameters are categorized and organized according to operational stages. Combining these categorized historical parameters with real-time collected parameters, the heat load variation trend of the liquid-cooled server cabinet is analyzed, and future heat load conditions are predicted, providing data support for the optimization and updating of the heat load prediction model. At a fixed update cycle, the collected operational data and historical operational data are integrated to form a model optimization dataset. This optimized dataset is input into the heat load prediction model, and the model's internal weight parameters are iteratively adjusted to optimize its feature extraction capabilities and prediction accuracy. This completes the model version update, enabling the heat load prediction model to adapt to changes in the operating status of the liquid-cooled server cabinet.

[0047] In some embodiments, during the update process of the heat load prediction model, a verification threshold for the model update can be set. The model version update is completed only when the prediction accuracy of the optimized model reaches the set verification threshold, thus ensuring the stability of the model operation.

[0048] In some specific implementations, during the optimization and updating of the heat load prediction model, existing incremental learning methods can be used to achieve iterative optimization of the model. Incremental learning refers to existing machine learning methods that enable the model to continuously learn new feature patterns based on new operating data without forgetting previously learned feature patterns, thus avoiding the problem of forgetting historical data features during model updates. In the implementation of this embodiment, the new operating data collected within a fixed update cycle is organized into an incremental training dataset. Regularization constraints for incremental learning are set, and the incremental training dataset is input into the heat load prediction model. The parameters of the fully connected layers of the model are fine-tuned and optimized, while the original parameter weights of the time-series feature extraction layers are retained. This completes the incremental update of the model, enabling it to adapt to new changes in operating status while retaining the ability to identify and predict historical operating features, ensuring the stability of the model's prediction accuracy.

[0049] S4. Operation Monitoring and Load Adjustment: This step ensures stable system operation through continuous monitoring and optimizes the heat load distribution of the entire cabinet through adaptive load adjustment. It includes the following sub-steps: S4.1. Continuous monitoring of operational performance: This step, based on the analysis and prediction results of the heat load prediction model and anomaly detection and response data, continuously monitors the overall operational performance of the liquid-cooled server rack, acquiring real-time operating parameters. The monitoring includes the operating status of the coolant circulation system, the working status of the liquid-cooled server, and changes in heat load. Continuous monitoring obtains complete real-time operating parameters, providing a data foundation for subsequent problem detection and load adjustment.

[0050] S4.2. Potential Problem Detection and Information Feedback: This step involves parsing and analyzing the real-time acquired operating parameters to detect potential problems in the operation of the liquid-cooled server rack. Feedback information is generated for each detected problem and pushed to the monitoring terminal. The real-time operating parameters are compared with historical operating parameters stored in the data collection system. Based on the comparison analysis, the operating status trend of the liquid-cooled server rack is identified. Various judgment criteria for system monitoring and early warning are dynamically adjusted based on these trends. These criteria include parameter threshold standards and anomaly identification standards, ensuring that monitoring and early warning judgments adapt to changes in system operating status and improving the accuracy of potential problem detection. For each detected potential problem, standardized feedback information is generated and pushed to the corresponding monitoring terminal, enabling maintenance personnel to promptly obtain information on changes in the system's operating status.

[0051] S4.3. Dynamic load adjustment for server nodes: This step extracts relevant data reflecting the thermal load status of the entire liquid-cooled server cabinet from the analysis results of real-time operating parameters, and dynamically adjusts the load of each server node within the cabinet based on this data. According to the thermal load data, the thermal load distribution of each server node within the cabinet is identified. For server nodes with high thermal loads, their operating load is reduced; for server nodes with low thermal loads, their operating load is rationally allocated, resulting in a more uniform thermal load distribution within the cabinet, matching the cooling capacity of the coolant, and ensuring the stable operation of the entire liquid-cooled server cabinet.

[0052] In some embodiments, during the dynamic adjustment of server node load, load distribution can be carried out in combination with the service operation priority of the server node to achieve a uniform distribution of heat load while ensuring the normal operation of services.

[0053] In some specific implementations, during the dynamic adjustment of server node load, an existing distributed task scheduling framework can be used to achieve dynamic load allocation. This framework is an existing software framework capable of unified scheduling, allocation, and management of tasks across multiple distributed computing nodes, adjusting task allocation strategies based on the operating status of each node. In this embodiment, each server node within the liquid-cooled server rack is connected to the distributed task scheduling framework. Real-time data on the thermal load status, operating load, and task processing capacity of each server node is acquired. Task scheduling constraints are set, and the task allocation for each node is adjusted based on the thermal load status data. New tasks are allocated to nodes with high thermal loads, while an appropriate amount of tasks is allocated to nodes with low thermal loads. This ensures the normal processing of business tasks while achieving a uniform distribution of thermal load across all server nodes within the rack, reducing the risk of excessive localized thermal loads, and adapting to the cooling capacity of the coolant circulation system.

[0054] The intelligent coolant distribution system for liquid-cooled servers in this embodiment includes a coolant circulation assembly, a coolant compatibility verification module, a sensor acquisition module, a data collection system, a heat load prediction model analysis and control module, a monitoring and early warning module, a load adjustment module, and coolant delivery pipelines. The coolant circulation assembly comprises various circulation components, including a cooling tower, heat exchanger, coolant pump, and coolant distribution unit. The coolant distribution unit integrates various functional components, including a heat exchanger, coolant pump, main filter, bypass filter, and model operation microcontroller unit, and is equipped with a frequency converter control module. The sensor acquisition module is deployed within the liquid-cooled server cabinet to collect various operating parameters of the entire cabinet in real time. The coolant compatibility verification module is used to detect and analyze various performance characteristics of the coolant and confirm the compatibility between the coolant and electronic components. The data collection system stores pre-processed valid operating parameters and various historical operating parameters of the coolant circulation system. The load prediction model analysis and control module incorporates a time-series neural network algorithm to extract operating parameter features, analyze and predict the operating status of the liquid-cooled server cabinet, and intelligently control the coolant distribution unit. It also uses a feedback mechanism to optimize and update the heat load prediction model. The monitoring and early warning module continuously monitors the operating performance of the liquid-cooled server cabinet, detects potential problems, and pushes feedback information. The load adjustment module dynamically adjusts the load of each server node within the liquid-cooled server cabinet based on relevant heat load data. Coolant delivery pipelines connect the coolant circulation components to the liquid-cooled server cabinet, enabling coolant flow and transmission within the system. Signal interconnection is achieved between all modules and components to ensure the normal transmission of data and control commands within the system.

[0055] This embodiment achieves intelligent distribution of coolant for liquid-cooled servers through a complete technical process including coolant circulation system construction, operational parameter acquisition and processing, heat load prediction and intelligent control, and operation monitoring and load adjustment. The solution ensures the compatibility between the cooling medium and electronic components through coolant compatibility verification, reducing the risk of electronic component damage during long-term operation of the liquid-cooled system. It achieves advance prediction of heat load changes through a time-series neural network algorithm, enabling the coolant distribution unit control to adapt to changes in server heat load, improving the cooling system's response speed and adaptability. The integrated coolant distribution unit configuration simplifies the liquid-cooled system structure, reduces system connection points, and lowers leakage risk and maintenance difficulty. Adaptive load adjustment of server nodes optimizes the heat load distribution within the entire cabinet, improving the cooling system's operating efficiency. Overall, the solution optimizes the cooling effect of liquid-cooled servers, reduces the cooling system's energy consumption, improves the stability and reliability of the liquid-cooled system, and adapts to the cooling requirements of high-density liquid-cooled servers.

[0056] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A method for intelligent distribution of coolant in a liquid-cooled server, characterized in that, Includes the following steps: S1. Build a coolant circulation system. First, conduct coolant compatibility verification to confirm the compatibility between the coolant and electronic components. Then, connect various circulation components to complete the coolant circulation system and debug it to an operational state. S2. Collect and preprocess operating parameters. Collect operating parameters in real time for the liquid-cooled server cabinet corresponding to the completed coolant circulation system, perform preprocessing operations on the collected operating parameters, and classify and store the preprocessed operating parameters in the data collection system. S3. Construct a heat load prediction model and intelligently control the coolant distribution unit. Use a time-series neural network algorithm to process, analyze and predict the trends of the preprocessed operating parameters stored in the classification. Combine the analysis and prediction results to intelligently control the coolant distribution unit. At the same time, carry out anomaly detection and response on the operating status of the coolant circulation system. Collect the operating parameters in the anomaly detection and response process, optimize the heat load prediction model with the help of the feedback mechanism, and update it regularly. S4. Implement monitoring, early warning, and adaptive load adjustment. Based on the analysis and prediction results of the heat load prediction model and the anomaly detection and response data, continuously monitor the operating performance of the liquid-cooled server rack, analyze the operating parameters obtained by monitoring and detect potential problems, provide timely feedback on the monitoring and analysis results, and dynamically adjust the load of each server node in the liquid-cooled server rack based on the monitoring and analysis results.

2. The method according to claim 1, characterized in that, Step S1 includes the following sub-steps: S1.

1. Conduct coolant compatibility verification, test and analyze various properties of the coolant, including thermophysical properties, chemical stability, thermal stability, corrosivity, boiling point characteristics and latent heat of vaporization, and confirm the compatibility between the coolant and electronic components based on the performance test and analysis results of the coolant. S1.

2. Connect the circulation components and seal them sequentially. The circulation components include the cooling tower, heat exchanger, coolant pump, coolant distribution unit, coolant delivery pipeline and liquid-cooled server cabinet. S1.

3. Complete the commissioning of the coolant circulation system, introduce coolant into the connected coolant circulation system, check the flow status of coolant between each circulation component, and adjust the operating components of the coolant circulation system to a stable operating state.

3. The method according to claim 1, characterized in that, Step S2 includes the following sub-steps: S2.

1. Collect operating parameters. Set up parameter collection points inside the liquid-cooled server cabinet to collect various operating parameters of the liquid-cooled server cabinet in real time, including temperature, humidity, current and voltage. S2.

2. Preprocessing operating parameters: Perform various preprocessing operations on the collected operating parameters, including noise reduction, normalization, and invalid data removal, to obtain valid operating parameters from the collected operating parameters; S2.

3. Categorize and store operating parameters. Classify the valid operating parameters according to their parameter types and store the classified valid operating parameters in the storage area of ​​the data collection system.

4. The method according to claim 1, characterized in that, Step S3 includes the following sub-steps: S3.

1. Extract operating parameter features, use a time-series neural network algorithm to parse the preprocessed operating parameters of the classified storage, and extract various feature information that can reflect the overall operating status of the liquid-cooled server cabinet from the parsed preprocessed operating parameters; S3.

2. Analyze and predict the operating status. Based on the extracted feature information, conduct trend analysis of the operating parameters of the liquid-cooled server cabinet and predict the future operating status to form the analysis and prediction results of the operating parameters. S3.

3. Intelligent control and anomaly detection and response: Adjust the operating parameters of the coolant distribution unit based on the analysis and prediction results of the operating parameters, conduct anomaly detection and response on the overall operating status of the coolant circulation system, and implement corresponding response actions for various detected anomalies. S3.

4. Optimize and update the heat load prediction model, collect various operating parameters in the prediction process and anomaly detection and response process, import the collected operating parameters into the heat load prediction model with the help of the feedback mechanism, optimize the internal parameters of the heat load prediction model, and complete the update of the heat load prediction model at fixed intervals.

5. The method according to claim 1, characterized in that, Step S4 includes the following sub-steps: S4.

1. Continuously monitor the operating performance. Based on the analysis and prediction results of the heat load prediction model and the anomaly detection and response data, continuously monitor the overall operating performance of the liquid-cooled server cabinet and obtain the real-time operating parameters of the liquid-cooled server cabinet. S4.

2. Analyze, detect, and report problems. Perform data parsing and operational status analysis on the real-time acquired operating parameters. Detect potential problems in the operation of the liquid-cooled server cabinet from the status analysis results. Generate feedback information for the detected potential problems and push it to the monitoring terminal. S4.

3. Dynamically adjust the load of server nodes. Extract relevant data reflecting the thermal load status of the entire liquid-cooled server cabinet from the analysis results of real-time operating parameters, and dynamically adjust the load of each server node in the entire liquid-cooled server cabinet in combination with the relevant thermal load status data.

6. The method according to claim 1, characterized in that, In step S3, the various control components of the coolant distribution unit are adjusted based on the analysis and prediction results of the operating parameters. The control components include flow control valves and temperature regulating components. At the same time, the flow control parameters and temperature control parameters of the coolant distribution unit are automatically adjusted according to the real-time operating status of the coolant circulation system.

7. The method according to claim 1, characterized in that, In step S3, various historical operating parameters during the operation of the coolant circulation system are continuously collected. These historical operating parameters include operating parameter acquisition data, analysis and prediction result data, and anomaly detection and response data. The various historical operating parameters are classified and organized according to the operating stage. By combining the classified and organized historical operating parameters with the real-time collected operating parameters, the analysis of the heat load change trend of the liquid-cooled server cabinet and the prediction of future heat load status are carried out.

8. The method according to claim 1, characterized in that, In step S4, the operating parameters acquired in real time are compared and analyzed with the historical operating parameters stored in the data collection system. Based on the comparison and analysis results, the changing trend of the operating status of the liquid-cooled server cabinet is identified. Combined with the changing trend of the operating status, the various judgment criteria of the system monitoring and early warning are dynamically adjusted. The judgment criteria include parameter threshold criteria and anomaly identification criteria.

9. The method according to claim 1, characterized in that, In step S1, the coolant distribution unit in the coolant circulation system is integrated and configured, and various functional components are assembled into one unit. The functional components include heat exchangers, coolant pumps, main filters, bypass filters and model operation microcontrollers. The control module of the coolant distribution unit is configured using frequency conversion control technology, and the frequency conversion adjustment of various operating parameters of the coolant distribution unit is realized through the control module.

10. A liquid-cooled server coolant intelligent distribution system, used to perform the method as described in any one of claims 1-9, characterized in that, The system includes a coolant circulation assembly, a coolant compatibility verification module, a sensor acquisition module, a data collection system, a heat load prediction model analysis and control module, a monitoring and early warning module, a load adjustment module, and a coolant delivery pipeline. The coolant circulation assembly includes various circulation components, such as a cooling tower, heat exchanger, coolant pump, and coolant distribution unit. The coolant distribution unit integrates various functional components, including a heat exchanger, coolant pump, main filter, bypass filter, and model operation microcontroller unit. The coolant distribution unit is also equipped with a frequency converter control module. The sensor acquisition module is located inside the liquid-cooled server cabinet and is used to collect various operating parameters of the liquid-cooled server cabinet in real time. The coolant compatibility verification module is used to test and analyze various properties of the coolant and confirm the compatibility between the coolant and electronic components; The data collection system stores pre-processed valid operating parameters and various historical operating parameters of the coolant circulation system; the heat load prediction model analysis and control module has a built-in time-series neural network algorithm to extract operating parameter features, analyze and predict the operating status of the liquid-cooled server cabinet, and intelligently control the coolant distribution unit, while optimizing and updating the heat load prediction model with the help of a feedback mechanism; the monitoring and early warning module continuously monitors the operating performance of the liquid-cooled server cabinet, detects potential problems, and pushes feedback information. The load adjustment module is used to dynamically adjust the load of each server node in the liquid-cooled server cabinet based on relevant data on thermal load status; the coolant delivery pipeline is used to connect the coolant circulation component and the liquid-cooled server cabinet to realize the flow and transmission of coolant in the system; and signal interconnection is realized between various modules and components.

Images