An edge-computing-based intelligent operation and maintenance method for a liquid cooling system

Abstract

This invention relates to data center liquid cooling optimization, specifically to an intelligent operation and maintenance method for liquid cooling systems based on edge computing. Edge nodes acquire task request information and allocate basic task priorities according to the information. Based on a task scheduling strategy, the basic task priorities are adjusted to obtain optimized task priorities. Based on the optimized task priorities, computing resource utilization is calculated to complete task scheduling and control server heat generation. Edge nodes acquire operational data of the liquid cooling system and preprocess it. The operational data is used to train a load prediction model for the liquid cooling system, and simultaneously, it is used to train energy consumption models for each device in the liquid cooling system. Based on the pre-trained energy consumption models for each device, the total energy consumption model of the liquid cooling system is obtained. The technical solution provided by this invention effectively overcomes the shortcomings of existing technologies that make it difficult to accurately and efficiently control liquid cooling systems locally.

Description

Technical Field

[0001] This invention relates to data center liquid cooling optimization, and more specifically to an intelligent operation and maintenance method for liquid cooling systems based on edge computing. Background Technology

[0002] With the increasing deployment density of servers in large data centers, liquid cooling is widely used due to its high heat exchange efficiency. However, traditional liquid cooling system management often adopts a centralized cloud management architecture, where all collected data is uploaded to a central platform for processing. The massive data transmission can easily lead to high communication latency and high cloud computing power overhead, making it difficult to achieve millisecond-level dynamic control of the liquid cooling system.

[0003] Traditional operation and maintenance architectures separate upper-layer task scheduling from lower-layer liquid cooling control. Task priority division relies heavily on manual settings and cannot dynamically optimize task scheduling based on task type, computing power requirements, and urgency. This can easily lead to a surge in server cluster heat generation caused by the concentrated launch of batch tasks, resulting in disordered fluctuations in the data center's heat dissipation load.

[0004] Conventional liquid cooling control solutions lack load forecasting and energy consumption quantification modeling capabilities. They rely on fixed thresholds to control the parameters of primary and secondary equipment in the liquid cooling system. The supply of cold source cannot accurately match the changes in server heat source, often resulting in problems such as excessive cooling leading to unit no-load loss, or insufficient cooling causing server overheating. This limits the room's PUE optimization space.

[0005] Existing liquid cooling system operation and maintenance solutions lack localized edge computing processing capabilities, making it impossible to complete task scheduling, computing resource utilization calculations, data preprocessing, and iterative training of load and energy consumption models locally. Therefore, introducing edge computing technology to achieve collaborative optimization of task scheduling and equipment parameters at local nodes, controlling heat generation scale from the heat source end, and solving for optimal cold source operating parameters based on localized models has become an urgent need to solve the problems of high energy consumption and lagging regulation in existing liquid cooling systems. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] In view of the above-mentioned shortcomings of the existing technology, the present invention provides an intelligent operation and maintenance method for liquid cooling system based on edge computing, which can effectively overcome the defects of the existing technology that make it difficult to accurately and efficiently control the liquid cooling system locally.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] An intelligent operation and maintenance method for liquid cooling systems based on edge computing includes the following steps:

[0011] S1. Edge nodes acquire task request information and allocate basic task priorities based on the task request information;

[0012] S2. Adjust the priority of basic tasks based on the task scheduling strategy to obtain optimized task priorities;

[0013] S3. Calculate the utilization rate of computing resources based on optimized task priorities, complete task scheduling, and control server heat generation;

[0014] S4. Edge nodes acquire the operating data of the liquid cooling system and preprocess the operating data;

[0015] S5. Use the operating data to train the load prediction model of the liquid cooling system, and at the same time use the operating data to train the energy consumption model of each device in the liquid cooling system, and obtain the total energy consumption model of the liquid cooling system based on the pre-trained energy consumption models of each device.

[0016] S6. Based on the load prediction model and total energy consumption model of the liquid cooling system, the optimal combination of operating parameters of the equipment in the liquid cooling system is solved by intelligent optimization algorithm, and corresponding control is executed to match the optimal cold source output according to the change of heat source.

[0017] Preferably, in S1, the edge nodes acquire task request information and allocate basic task priorities based on the task request information, including:

[0018] Based on the task type, required computing power, and urgency of each task in the task request information, assign a priority value to each task.

[0019] All tasks are sorted according to their priority values, and then added to the high-priority task queue, medium-priority task queue, and low-priority task queue in that order, awaiting scheduling.

[0020] Preferably, before adjusting the basic task priority based on the task scheduling strategy to obtain the optimized task priority in S2, the following steps are included:

[0021] Get the time when each task entered the corresponding task queue, and subtract the current time from the time when each task entered the corresponding task queue to get the task waiting time for each task;

[0022] Obtain the total computing power resources and the currently used computing power resources. Subtract the currently used computing power resources from the total computing power resources to obtain the available computing power resources.

[0023] The task scheduling strategy is determined based on the task waiting time and available computing resources for each task.

[0024] Preferably, in S2, the priority of the basic tasks is adjusted based on the task scheduling strategy to obtain the optimized task priority, including:

[0025] Tasks are first retrieved from the high-priority task queue. If the available computing resources meet the computing power required for the task, the task is assigned to the corresponding computing node for execution; otherwise, the task is waited for the available computing resources to meet the computing power required for the task.

[0026] Preferably, in S2, the priority of the basic tasks is adjusted based on the task scheduling strategy to obtain the optimized task priority, including:

[0027] For tasks in the medium-priority task queue and the low-priority task queue, if the available computing resources meet the computing power required by the task, and the task waiting time is greater than the expected task waiting time, then the priority of the task will be adjusted, and task scheduling will be performed according to the adjusted priority.

[0028] Preferably, in S3, the calculation of computing resource utilization based on optimized task priority includes:

[0029] Get the start time and finish time of each task, and subtract the start time from the finish time of each task to get the processing time of each task.

[0030] By monitoring the usage of computing resources in real time, the amount of computing resources used for each task can be obtained.

[0031] The computing resource utilization rate is calculated by combining the total time period, the processing time of each task, the total computing resources, and the computing resources used by each task.

[0032] Preferably, in S4, the edge nodes acquire the operating data of the liquid cooling system and preprocess the operating data, including:

[0033] The system acquires the operating data of the liquid cooling system and preprocesses the operating data by at least one of the following operations: interpolation to fill in missing data, replacing data with approximate environmental data with data whose error exceeds a preset error range, removing data whose missing time period exceeds a preset allowable time period, and smoothing.

[0034] Preferably, in S5, the total energy consumption model of the liquid cooling system is obtained based on the pre-trained energy consumption models of each device, including:

[0035] The total energy consumption model of the liquid cooling system is obtained by adding the pre-trained energy consumption models of each device.

[0036] Preferably, in S6, based on the load prediction model and total energy consumption model of the liquid cooling system, an intelligent optimization algorithm is used to solve for the optimal combination of operating parameters for the equipment in the liquid cooling system, including:

[0037] The total energy consumption model of the liquid cooling system is used as the objective function of the intelligent optimization algorithm, and the load values ​​at future times predicted by the load prediction model of the liquid cooling system are used as the constraints of the intelligent optimization algorithm.

[0038] Determine the parameter settings of each device in the liquid cooling system that meet the constraints, and form multiple different combinations of operating parameters;

[0039] The total energy consumption value of each combination of operating parameters is calculated using the total energy consumption model of the liquid cooling system, and the combination of operating parameters with the lowest total energy consumption value is taken as the optimal combination of operating parameters for the equipment in the liquid cooling system.

[0040] Preferably, the method further includes:

[0041] The accuracy of the load prediction model of the liquid cooling system is monitored in real time. If the model accuracy does not meet the preset accuracy requirements for a period of time exceeding the preset duration, the load prediction model is retrained and redeployed.

[0042] The accuracy of the total energy consumption model of the liquid cooling system is monitored in real time. If the model accuracy does not meet the preset accuracy requirements for a period of time exceeding the preset duration, the total energy consumption model is retrained and redeployed.

[0043] (III) Beneficial Effects

[0044] Compared with existing technologies, the intelligent operation and maintenance method for liquid cooling systems based on edge computing provided by this invention has the following beneficial effects:

[0045] 1) Optimize computing power and cooling source allocation to improve overall resource utilization.

[0046] This invention breaks away from the traditional data center's independent management model of computing power scheduling and liquid cooling control. It leverages edge nodes to achieve hierarchical task scheduling, dynamically adjusting task priorities based on task type, computing power requirements, urgency, task waiting time, and available computing resources. This avoids problems such as high-priority task blocking, low-priority task backlog, and idle or overloaded computing resources. Simultaneously, by accurately calculating computing resource utilization, it proactively balances the heat load of the server cluster from the heat source end, preventing sudden bursts of high heat load. Furthermore, by constructing energy consumption models for liquid cooling equipment and total energy consumption models, and combining predicted loads, it accurately matches the optimal cold source output, avoiding energy waste caused by liquid cooling equipment blindly operating at full capacity. This achieves optimal bidirectional allocation of computing and cooling resources, significantly improving the overall resource utilization of the data center.

[0047] 2) Edge-localized computing improves overall system operating efficiency.

[0048] By leveraging edge computing technology, the entire process of localized data processing and intelligent decision-making is completed, eliminating the drawbacks of traditional centralized cloud management, such as high transmission latency, slow computation, and high bandwidth consumption. Edge nodes complete the entire process of task priority adjustment, computing resource utilization calculation, data preprocessing, model training, and parameter optimization locally, resulting in faster response speed and stronger real-time decision-making. At the same time, through intelligent optimization algorithms, the optimal combination of operating parameters for equipment in the liquid cooling system is found under load constraints, accurately locking the lowest energy consumption operating strategy, realizing dynamic changes in heat sources and adaptive matching of cold sources, effectively improving the dynamic control efficiency of the liquid cooling system and the overall operating efficiency of the data center.

[0049] 3) Coordinated and precise control enhances data center operational security.

[0050] This invention constructs a linked protection system of "front-end computing power scheduling and control of heat sources, and back-end intelligent optimization and matching of cold sources," which completely solves the problem of poor adaptability of traditional fixed parameter control mode. By dynamically optimizing task scheduling to balance server heat generation, it avoids the safety hazards of local equipment high temperature and sudden load changes from the source. At the same time, relying on a high-precision load prediction model to predict heat dissipation demand in advance, it can effectively avoid server overheating caused by lag or insufficient cooling supply. The whole mechanism realizes refined, adaptive, and highly reliable intelligent operation and maintenance of liquid cooling system, and comprehensively ensures the long-term stable and safe operation of data center servers and liquid cooling equipment. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0052] Figure 1 This is a schematic diagram of the process of the present invention;

[0053] Figure 2 This is a schematic diagram illustrating the process of using intelligent optimization algorithms to solve for the optimal combination of operating parameters for equipment in a liquid cooling system in this invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0055] The following describes the specific process of the intelligent operation and maintenance method for liquid cooling systems based on edge computing provided by this invention, using specific examples (e.g.) Figure 1 (as shown) and technical effects.

[0056] S1. Edge nodes acquire task request information and allocate basic task priorities based on the task request information, including:

[0057] Based on the task type, required computing power, and urgency of each task in the task request information, assign a priority value to each task.

[0058] All tasks are sorted according to their priority values, and then added to the high-priority task queue, medium-priority task queue, and low-priority task queue in that order, awaiting scheduling.

[0059] S2. Adjust the priority of basic tasks based on the task scheduling strategy to obtain optimized task priorities.

[0060] Before adjusting the basic task priorities based on the task scheduling strategy in S2 to obtain the optimized task priorities, the following steps are included:

[0061] Get the time when each task entered the corresponding task queue, and subtract the current time from the time when each task entered the corresponding task queue to get the task waiting time for each task;

[0062] Obtain the total computing power resources and the currently used computing power resources. Subtract the currently used computing power resources from the total computing power resources to obtain the available computing power resources.

[0063] The task scheduling strategy is determined based on the task waiting time and available computing resources for each task.

[0064] In S2, the priority of basic tasks is adjusted based on the task scheduling strategy to obtain optimized task priorities, including:

[0065] Tasks are first retrieved from the high-priority task queue. If the available computing resources meet the computing power required for the task, the task is assigned to the corresponding computing node for execution; otherwise, the process waits until the available computing resources meet the computing power required for the task.

[0066] For tasks in the medium-priority task queue and the low-priority task queue, if the available computing resources meet the computing power required by the task, and the task waiting time is greater than the expected task waiting time, then the priority of the task will be adjusted, and task scheduling will be performed according to the adjusted priority.

[0067] S3. Calculate the utilization rate of computing resources based on optimized task priorities, complete task scheduling, and control server heat generation.

[0068] S3 calculates computing resource utilization based on optimized task priorities, including:

[0069] Get the start time and finish time of each task, and subtract the start time from the finish time of each task to get the processing time of each task.

[0070] By monitoring the usage of computing resources in real time, the amount of computing resources used for each task can be obtained.

[0071] The computing resource utilization rate is calculated by combining the total time period, the processing time of each task, the total computing resources, and the computing resources used by each task.

[0072] S4. Edge nodes acquire operational data from the liquid cooling system and preprocess the data, including:

[0073] The system acquires the operating data of the liquid cooling system and preprocesses the operating data by at least one of the following operations: interpolation to fill in missing data, replacing data with approximate environmental data with data whose error exceeds a preset error range, removing data whose missing time period exceeds a preset allowable time period, and smoothing.

[0074] S5. Use the operating data to train the load prediction model of the liquid cooling system, and at the same time use the operating data to train the energy consumption model of each device in the liquid cooling system. Based on the pre-trained energy consumption models of each device, obtain the total energy consumption model of the liquid cooling system.

[0075] In S5, the total energy consumption model of the liquid cooling system is obtained based on the pre-trained energy consumption models of each device, such as... Figure 2 As shown, it includes:

[0076] The total energy consumption model of the liquid cooling system is obtained by adding the pre-trained energy consumption models of each device.

[0077] S6. Based on the load prediction model and total energy consumption model of the liquid cooling system, the optimal combination of operating parameters of the equipment in the liquid cooling system is solved by intelligent optimization algorithm, and corresponding control is executed to match the optimal cold source output according to the change of heat source.

[0078] In S6, a load prediction model and a total energy consumption model based on a liquid cooling system are used to solve for the optimal combination of operating parameters for the equipment in the liquid cooling system using intelligent optimization algorithms, such as... Figure 2 As shown, it includes:

[0079] The total energy consumption model of the liquid cooling system is used as the objective function of the intelligent optimization algorithm, and the load values ​​at future times predicted by the load prediction model of the liquid cooling system are used as the constraints of the intelligent optimization algorithm.

[0080] Determine the parameter settings of each device in the liquid cooling system that meet the constraints, and form multiple different combinations of operating parameters;

[0081] The total energy consumption value of each combination of operating parameters is calculated using the total energy consumption model of the liquid cooling system, and the combination of operating parameters with the lowest total energy consumption value is taken as the optimal combination of operating parameters for the equipment in the liquid cooling system.

[0082] like Figure 2 As shown, this method also includes:

[0083] The accuracy of the load prediction model of the liquid cooling system is monitored in real time. If the model accuracy does not meet the preset accuracy requirements for a period of time exceeding the preset duration, the load prediction model is retrained and redeployed.

[0084] The accuracy of the total energy consumption model of the liquid cooling system is monitored in real time. If the model accuracy does not meet the preset accuracy requirements for a period of time exceeding the preset duration, the total energy consumption model is retrained and redeployed.

[0085] This technical solution breaks away from the traditional independent management model of computing power scheduling and liquid cooling control in data centers. It leverages edge nodes to achieve hierarchical task scheduling, dynamically adjusting task priorities based on task type, computing power requirements, urgency, task waiting time, and available computing resources. This avoids problems such as high-priority task blocking, low-priority task backlog, and idle or overloaded computing resources. Simultaneously, by accurately calculating computing resource utilization, it proactively balances the heat load of the server cluster from the heat source end, preventing sudden bursts of high heat load. Furthermore, by constructing energy consumption models for individual liquid cooling devices and a total energy consumption model, combined with predicted load, it accurately matches the optimal cold source output, avoiding energy waste caused by blindly operating liquid cooling equipment at full capacity. This achieves optimal bidirectional allocation of computing and cooling resources, significantly improving the overall resource utilization of the data center.

[0086] Secondly, a coordinated protection system was constructed, consisting of "front-end computing power scheduling and control of heat sources, and back-end intelligent optimization and matching of cold sources." This system completely solves the problem of poor adaptability of traditional fixed-parameter control modes. By dynamically optimizing task scheduling to balance server heat generation, it avoids the safety hazards of localized high temperatures and sudden load changes in equipment from the source. Simultaneously, relying on a high-precision load prediction model to anticipate heat dissipation needs in advance, it effectively avoids server overheating caused by delayed or insufficient cooling supply. The entire mechanism achieves refined, adaptive, and highly reliable intelligent operation and maintenance of the liquid cooling system, comprehensively ensuring the long-term stable and safe operation of data center servers and liquid cooling equipment.

[0087] Furthermore, edge computing technology enables end-to-end localized data processing and intelligent decision-making, eliminating the drawbacks of traditional centralized cloud management, such as high transmission latency, computational lag, and high bandwidth consumption. Edge nodes complete the entire process locally, including task priority adjustment, computing resource utilization calculation, data preprocessing, model training, and parameter optimization, resulting in faster response times and more real-time decision-making. Simultaneously, intelligent optimization algorithms identify the optimal combination of operating parameters for equipment in the liquid cooling system under load constraints, precisely locking in the lowest energy consumption operating strategy. This achieves dynamic matching between heat source changes and cold source adaptation, effectively improving the dynamic control efficiency of the liquid cooling system and the overall operating efficiency of the data center.

[0088] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent operation and maintenance of a liquid cooling system based on edge computing, characterized in that: Includes the following steps: S1. Edge nodes acquire task request information and allocate basic task priorities based on the task request information; S2. Adjust the priority of basic tasks based on the task scheduling strategy to obtain optimized task priorities; S3. Calculate the utilization rate of computing resources based on optimized task priorities, complete task scheduling, and control server heat generation; S4. Edge nodes acquire the operating data of the liquid cooling system and preprocess the operating data; S5. Use the operating data to train the load prediction model of the liquid cooling system, and at the same time use the operating data to train the energy consumption model of each device in the liquid cooling system, and obtain the total energy consumption model of the liquid cooling system based on the pre-trained energy consumption models of each device. S6. Based on the load prediction model and total energy consumption model of the liquid cooling system, the optimal combination of operating parameters of the equipment in the liquid cooling system is solved by intelligent optimization algorithm, and corresponding control is executed to match the optimal cold source output according to the change of heat source.

2. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 1, characterized in that: In S1, edge nodes acquire task request information and allocate basic task priorities based on the task request information, including: Based on the task type, required computing power, and urgency of each task in the task request information, assign a priority value to each task. All tasks are sorted according to their priority values, and then added to the high-priority task queue, medium-priority task queue, and low-priority task queue in that order, awaiting scheduling.

3. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 2, characterized in that: Before adjusting the basic task priorities based on the task scheduling strategy in S2 to obtain the optimized task priorities, the following steps are included: Get the time when each task entered the corresponding task queue, and subtract the current time from the time when each task entered the corresponding task queue to get the task waiting time for each task; Obtain the total computing power resources and the currently used computing power resources. Subtract the currently used computing power resources from the total computing power resources to obtain the available computing power resources. The task scheduling strategy is determined based on the task waiting time and available computing resources for each task.

4. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 3, characterized in that: In S2, the priority of basic tasks is adjusted based on the task scheduling strategy to obtain optimized task priorities, including: Tasks are first retrieved from the high-priority task queue. If the available computing resources meet the computing power required for the task, the task is assigned to the corresponding computing node for execution; otherwise, the task is waited for the available computing resources to meet the computing power required for the task.

5. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 4, characterized in that: In S2, the priority of basic tasks is adjusted based on the task scheduling strategy to obtain optimized task priorities, including: For tasks in the medium-priority task queue and the low-priority task queue, if the available computing resources meet the computing power required by the task, and the task waiting time is greater than the expected task waiting time, then the priority of the task will be adjusted, and task scheduling will be performed according to the adjusted priority.

6. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 5, characterized in that: S3 calculates computing resource utilization based on optimized task priorities, including: Get the start time and finish time of each task, and subtract the start time from the finish time of each task to get the processing time of each task. By monitoring the usage of computing resources in real time, the amount of computing resources used for each task can be obtained. The computing resource utilization rate is calculated by combining the total time period, the processing time of each task, the total computing resources, and the computing resources used by each task.

7. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 1, characterized in that: In S4, edge nodes acquire operational data from the liquid cooling system and preprocess the data, including: The system acquires the operating data of the liquid cooling system and preprocesses the operating data by at least one of the following operations: interpolation to fill in missing data, replacing data with approximate environmental data with data whose error exceeds a preset error range, removing data whose missing time period exceeds a preset allowable time period, and smoothing.

8. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 7, characterized in that: In S5, the total energy consumption model of the liquid cooling system is obtained based on the pre-trained energy consumption models of each device, including: The total energy consumption model of the liquid cooling system is obtained by adding the pre-trained energy consumption models of each device.

9. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 8, characterized in that: S6 uses a load prediction model and total energy consumption model based on a liquid cooling system, and employs intelligent optimization algorithms to solve for the optimal combination of operating parameters for equipment in the liquid cooling system, including: The total energy consumption model of the liquid cooling system is used as the objective function of the intelligent optimization algorithm, and the load values ​​at future times predicted by the load prediction model of the liquid cooling system are used as the constraints of the intelligent optimization algorithm. Determine the parameter settings of each device in the liquid cooling system that meet the constraints, and form multiple different combinations of operating parameters; The total energy consumption value of each combination of operating parameters is calculated using the total energy consumption model of the liquid cooling system, and the combination of operating parameters with the lowest total energy consumption value is taken as the optimal combination of operating parameters for the equipment in the liquid cooling system.

10. The intelligent operation and maintenance method for liquid cooling systems based on edge computing according to claim 9, characterized in that: The method also includes: The accuracy of the load prediction model of the liquid cooling system is monitored in real time. If the model accuracy does not meet the preset accuracy requirements for a period of time exceeding the preset duration, the load prediction model is retrained and redeployed. The accuracy of the total energy consumption model of the liquid cooling system is monitored in real time. If the model accuracy does not meet the preset accuracy requirements for a period of time exceeding the preset duration, the total energy consumption model is retrained and redeployed.

Images