Comprehensive energy-saving control system and method for data center computer room using liquid cooling

Abstract

The present invention relates to the field of energy-saving control technology, and in particular to a comprehensive energy-saving control system and method for a data center computer room using liquid cooling, including an equipment monitoring space and a hierarchical energy-saving control module. The equipment monitoring space module monitors the working status and temperature of the computer room equipment and its liquid cooling device in real time through a status monitoring unit, and constructs a cluster monitoring network using a cluster monitoring network construction unit to achieve refined monitoring of equipment cooling and energy-saving control. The hierarchical energy-saving control module generates a first-level node coordination control strategy through a federal distributed control unit, and generates a second-level differential cooling and energy-saving control strategy through an edge difference control unit to achieve hierarchical management of energy-saving control. In addition, the system also includes a flow distribution adjustment and evaluation feedback module. The former distributes and adjusts the coolant flow according to the control strategy, and the latter evaluates the cooling energy consumption in real time and feeds back to adjust the control strategy. The present invention can achieve refined monitoring and hierarchical energy-saving control of computer room equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of energy-saving control, and in particular relates to a comprehensive energy-saving control system and method for a data center computer room using a liquid cooling method. Background Art

[0002] Data center rooms are the core places for information storage, processing and transmission. The stable operation of the equipment in the data center room has extremely high requirements for environmental temperature, humidity and other conditions. The traditional cooling method of the data center room mainly relies on air cooling technology, but the air cooling method is gradually facing challenges in dealing with the growing demand for heat density. On the one hand, with the improvement of the performance of electronic equipment, the heat generation per unit volume has increased significantly. The heat dissipation efficiency of the air cooling system is limited, and it is difficult to meet the heat dissipation requirements of high-density equipment. On the other hand, the energy consumption of the air cooling system is high, which does not conform to the current development trend of energy conservation and emission reduction. In this context, liquid cooling technology has gradually attracted attention due to its good heat dissipation performance; liquid cooling can more effectively remove the heat generated by the equipment, improve heat dissipation efficiency, and is expected to reduce overall energy consumption. However, the application of liquid cooling systems still faces many challenges in actual operation, such as the management of coolant, coordination with other equipment, and further optimization of energy-saving effects.

[0003] For example, a Chinese patent application with authorization announcement number CN118741983B discloses a method for controlling heat dissipation and energy saving in a liquid cooling room, including: reading a group of operation monitoring data of Q units in the liquid cooling room; verifying the energy efficiency of Q heat dissipation control units according to the heat dissipation and energy saving control rules of the liquid cooling room, and obtaining an optimized heat dissipation control unit and an optimized heat dissipation control target; loading heat dissipation-related scene information of the liquid cooling room; mining heat dissipation control features of the optimized heat dissipation control unit, and establishing an optimized heat dissipation control space; activating a heat dissipation and energy saving control tuning algorithm for the liquid cooling room, optimizing the heat dissipation and energy saving control of the optimized heat dissipation control unit, and obtaining a heat dissipation and energy saving control scheme; and performing heat dissipation and energy saving control on the liquid cooling room according to the optimized heat dissipation control unit and the heat dissipation and energy saving control scheme.

[0004] For example, a patent application with publication number CN114867306A discloses an energy-saving liquid cooling system for a 5G base station room using nanofluid as a medium, including a heat dissipation mechanism and a refrigeration mechanism, wherein the heat dissipation mechanism includes a liquid storage tank and multiple groups of heat exchangers, and the multiple groups of heat exchangers are respectively installed on corresponding servers, and the output end and input end of the heat exchanger are respectively connected to the input end and output end on the liquid storage tank, and a pump is installed on the liquid storage tank, and the pump is used to transport the low-temperature fluid in the liquid storage tank to the heat exchanger, and a fan assembly is installed at the heat exchanger, and the fan assembly is used to transport the airflow cooled by the heat exchanger to the server, and the refrigeration mechanism is used to cool the fluid in the liquid storage tank.

[0005] The above prior art has the following problems: the coolant temperature control in the existing liquid cooling system is not precise enough, and temperature fluctuations are prone to occur. In addition, the coolant flow distribution between different servers or nodes is uneven, which may cause insufficient heat dissipation of some equipment and over-cooling of other equipment, increasing unnecessary energy consumption, and failing to achieve hierarchical and point-to-point energy-saving and cooling control processes, resulting in low accuracy of the liquid cooling and cooling control process. For this reason, the present invention provides a comprehensive energy-saving control system and method for a data center computer room using liquid cooling. Summary of the invention

[0006] In view of the deficiencies in the prior art, the present invention proposes a comprehensive energy-saving control system and method for a data center computer room using liquid cooling, including an equipment monitoring space and a hierarchical energy-saving control module. The equipment monitoring space module monitors the working status and temperature of the computer room equipment and its liquid cooling device in real time through a status monitoring unit, and constructs a cluster monitoring network using a cluster monitoring network construction unit to achieve refined monitoring of equipment cooling and energy-saving control. The hierarchical energy-saving control module generates a first-level node coordination control strategy through a federal distributed control unit, and generates a second-level differential cooling and energy-saving control strategy through an edge difference control unit to achieve hierarchical management of energy-saving control. In addition, the system also includes a flow distribution adjustment and evaluation feedback module. The former distributes and adjusts the coolant flow according to the control strategy, and the latter evaluates the cooling energy consumption in real time and feeds back to adjust the control strategy. The present invention can achieve refined monitoring and hierarchical energy-saving control of computer room equipment, improve energy utilization efficiency, and reduce energy consumption costs.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] The comprehensive energy-saving control system of the data center room using liquid cooling includes: equipment monitoring space module and hierarchical energy-saving control module;

[0009] The equipment monitoring space module includes a state monitoring unit and a cluster monitoring network construction unit; the hierarchical energy-saving control module includes a federal distributed control unit and an edge difference control unit; the state monitoring unit is used to monitor the working state, temperature state and working state data of the liquid cooling device corresponding to each device in the computer room in real time; according to the working state data and distribution state monitored by each device and liquid cooling device, a cluster monitoring network for equipment cooling and energy-saving control is constructed through the cluster monitoring network construction unit; the cluster monitoring network consists of N It is composed of a sub-monitoring network of devices clustered by a clustering algorithm;

[0010] according to NThe device sub-monitoring network status and the overall equipment operation status data of the computer room are collected, and M edge nodes and central control nodes are constructed through the federal distributed control unit, and a first-level node coordination control strategy is generated; according to the first-level node coordination control strategy, the working status and cooling requirements of the corresponding equipment under each edge node, and the difference in the working status of the liquid cooling device corresponding to the equipment, a second-level differential cooling and energy-saving control strategy for the liquid cooling device corresponding to each device under the edge node is generated through the edge difference control unit.

[0011] Specifically, the system also includes a flow distribution adjustment module and an evaluation feedback module; the flow distribution adjustment module includes a flow distribution unit and a temperature reduction control adjustment unit;

[0012] According to the generated first-level node coordination control strategy and second-level differential cooling and energy-saving control strategy, the coolant flow of each device is distributed and coordinated through the flow distribution unit, and the coolant rate and flow in the liquid cooling device corresponding to the device under each edge node are differentially adjusted through the cooling control adjustment unit; at the same time, the cooling energy consumption evaluation node network constructed by the evaluation feedback module is used to evaluate the cooling energy consumption of each device under each edge node in real time, and the evaluation results are fed back to the federal distributed control unit and the edge differential control unit to make real-time adjustments to the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy.

[0013] Specifically, each edge node is composed of all devices under a device sub-monitoring network and the configured liquid cooling device; the steps for constructing a cluster monitoring network for equipment cooling and energy-saving control include:

[0014] A1. Input the working status, temperature status and working status data of each device in the computer room obtained through monitoring into the distributed high-pass filter model configured on each edge node for preprocessing, and save the preprocessed status data on the corresponding edge node;

[0015] A2. Input the load, operation mode, temperature change trend under working state and working parameters of the liquid cooling device corresponding to the equipment in the state data saved by all edge nodes into the conditional clustering algorithm, and cluster the equipment that meets the same clustering conditions into one category to obtain a cluster containing N The cluster monitoring network of the equipment sub-monitoring network is composed of equipment sub-monitoring networks with the same clustering conditions. n A device and a corresponding liquid cooling device are constructed;

[0016] A3. Obtain the load status of each device in the device sub-monitoring network, and use the device with the minimum load status at the current moment as the master sub-node of the edge node corresponding to the current device sub-monitoring network, and save the pre-processed status data to the master sub-node, monitor the load status of all devices in the current device sub-monitoring network in real time, and adjust the master sub-node in real time when the load status of the device corresponding to the master sub-node is not minimum.

[0017] Specifically, the clustering conditions include that the load states of the devices are in the same load change interval, the temperature change rates of the devices per unit time are the same, and the heat dissipation rates of the corresponding liquid cooling devices per unit time are the same;

[0018] The federated distributed control unit is also configured with a federated control connection chain; the federated control connection chain includes an external connection sub-chain and an internal connection sub-chain; the external connection sub-chain is constructed through a federation algorithm and a graph algorithm based on the interaction data between M edge nodes and the central control node; the internal connection sub-chain is constructed through a graph algorithm based on the interaction data between each edge node and the interaction data between all devices in each edge node.

[0019] Specifically, the central control node is configured with a federated reinforcement model; the federated reinforcement model includes reinforcement sub-models configured by M edge nodes and a reinforcement sub-model configured by the central control node; the steps for generating the coordinated control strategy of the first-level nodes include:

[0020] B1. According to the working status of the equipment corresponding to each edge node, the equipment temperature change trend and the cooling target requirement, the required cooling liquid allocation amount for each edge node in the external connection sub-chain is obtained;

[0021] B2. Feedback the total amount of coolant distribution required by all edge nodes at the current moment to the reinforcement sub-model corresponding to the central control node, perform initial coolant distribution, and obtain the first-level node coordination control strategy. At the same time, use the cooling energy consumption evaluation node network to evaluate the initial coolant distribution energy consumption of each edge node in the externally connected sub-chain to obtain the initial distribution energy consumption evaluation value.

[0022] Specifically, the steps for building and training the federated reinforcement model include:

[0023] C1. Construct the current input state according to the working state and temperature state of each device under the corresponding edge node, the working state data of the liquid cooling device corresponding to each device, the cooling target requirements of each device, the cooling liquid volume and cooling liquid circulation rate of each device cooling target requirements, and the energy-saving and cooling state of each edge node in the internally connected sub-chain;

[0024] ;

[0025] in, Indicates the current t Momentm The input state corresponding to the edge node, Indicates m The edge node i The load status of each device at the current moment, Indicates m The edge node i The core temperature of the device at the current moment, Indicates m The edge node i The current shell temperature of a device, Indicates m The edge node i The cooling liquid flow rate of the liquid cooling device corresponding to each device at the current moment, Indicates m The edge node i The rate of temperature change of a device, Indicates m The edge node i The current inlet and outlet temperature difference of the liquid cooling device corresponding to each device is: Indicates m The edge node i The current coolant circulation rate of each device, Indicates m The edge node i The temperature threshold for normal operation of a device. Indicates k The input state parameters of the edge nodes at the current moment, Indicates m The edge node i Devices currently t The comprehensive energy consumption value of cooling according to the primary coolant distribution, secondary coolant distribution and the secondary differential cooling and energy-saving control strategy corresponding to the equipment at all times;

[0026] C2. Set the current moment trigger information according to the current moment input state, including:

[0027] When m Among the edge nodes i The corresponding device If it is not the minimum, the first action execution information is triggered, that is, the current m The master sub-node of the edge node, and the state data obtained after the current moment is stored in the adjusted master sub-node; Indicates m The edge node i The load status of each device at the current moment.

[0028] Specifically, the steps of building and training the federated reinforcement model also include:

[0029] When m Among the edge nodes i The current comprehensive average temperature of the device is greater than When the corresponding temperature change rate is greater than the temperature change rate corresponding to the normal operation of the device, the second action execution information is triggered, that is, the current i The cooling liquid circulation rate of the liquid cooling device corresponding to each device makes the device temperature less than ;in Indicates m The edge node i The temperature threshold for normal operation of a device;

[0030] When the second action execution information is triggered, the device temperature does not meet the requirement of less than , then the third action execution information is triggered, that is, while keeping the initial allocation amount of the current edge node coolant unchanged, according to the current edge node cooling target demand and the internal connection sub-chain interaction information corresponding to the current edge node, the secondary coolant allocation is redistributed so that the corresponding temperature of all devices at the current edge node is less than ;

[0031] When the third action execution information is triggered, the device temperature at the current edge node is still not less than When , the fourth action execution information is triggered, that is, the initial distribution amount of the coolant of each edge node is adjusted through the internal connection sub-chain interaction information between the edge nodes and the central control node, so that the device temperature of the devices under all edge nodes is less than ;

[0032] When the first to fourth action execution information are triggered, the fifth action execution information is triggered at the same time, that is, when the corresponding action execution information is triggered to cool down the equipment, the comprehensive energy consumption corresponding to the cooling process of the corresponding action execution information is evaluated through the cooling energy consumption evaluation node network.

[0033] Specifically, the steps of building and training the federated reinforcement model also include:

[0034] C3, setting the current execution action according to the trigger information at the current moment, and constructing the energy-saving and cooling reward function through the comprehensive energy consumption obtained through the fifth action execution information;

[0035] C4. Repeat the C1-C3 process to obtain the current input state, current trigger information, current execution action and energy-saving and cooling reward function corresponding to each edge node, and input them into the corresponding reinforcement sub-model. Through the federal algorithm, the reinforcement sub-models configured for all edge nodes and the reinforcement sub-models corresponding to the central control node are jointly trained to obtain the trained federal reinforcement model, and the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy corresponding to the current minimum energy consumption are adjusted in real time.

[0036] The comprehensive energy-saving control method of a data center computer room using liquid cooling includes:

[0037] S1. Based on the working status, temperature status of each device in the monitored computer room and the working status data of the liquid cooling device corresponding to each device, a cluster monitoring network for equipment cooling and energy-saving control is constructed through a graph network node algorithm and a clustering algorithm;

[0038] Cluster monitoring network N It is composed of a sub-monitoring network of devices clustered by a clustering algorithm;

[0039] S2. According to N The equipment sub-monitoring network and the overall equipment operation status of the computer room after clustering by the clustering algorithm are constructed, M edge nodes and central control nodes are constructed, and the first-level node coordination control strategy is generated to perform coordinated energy-saving and cooling control on each edge node;

[0040] S3. According to the difference between the working state and cooling demand of the corresponding equipment under each edge node and the working state of the liquid cooling device corresponding to the equipment, a secondary differential cooling and energy-saving control strategy is generated for the cooling process of the liquid cooling device corresponding to each equipment under the edge node;

[0041] S4. According to the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy, the cooling liquid is coordinated and distributed to the equipment under each edge node, and the cooling liquid rate and flow rate in the corresponding liquid cooling device of the equipment are differentially adjusted;

[0042] S5. Build a cooling energy consumption assessment node network to conduct real-time assessment and monitoring of the cooling energy consumption of equipment at each edge node, and make real-time adjustments to the first-level node coordination control strategy and the second-level differential cooling energy-saving control strategy based on the monitoring results.

[0043] A computer-readable storage medium stores computer instructions, which, when executed, execute a comprehensive energy-saving control method for a data center computer room using liquid cooling.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] In view of the shortcomings of the prior art, the present invention constructs N equipment sub-monitoring networks optimized by clustering algorithms through clustering monitoring network construction units according to the equipment distribution status and the real-time monitoring of the working status, temperature status and working status data of each equipment in the computer room, which not only improves the monitoring efficiency, but also reduces the computing load of the central control system, and further expands the number of equipment in the computer room that can be monitored; secondly, hierarchical and point-to-point energy-saving and cooling control are realized through the federal distributed control unit and the edge difference control unit. N The status of each device sub-monitoring network and the overall operation status data of the computer room are used to generate a first-level node coordination control strategy, and further generate a second-level differential cooling and energy-saving control strategy based on the specific working status and cooling requirements of the equipment under each edge node. This hierarchical control mechanism ensures that each device can get the most suitable cooling solution for itself, avoiding the problem of insufficient heat dissipation or overcooling of some equipment, and greatly reducing unnecessary energy consumption;

[0046] In addition, the introduction of the flow distribution and regulation module enables the system to accurately allocate coolant flow to the equipment contained in each edge node according to the primary and secondary control strategies, and differentially adjust the coolant rate under each edge node through the cooling control and regulation unit, further optimizing the cooling effect; the evaluation and feedback module constructs a cooling energy consumption evaluation node network, conducts real-time evaluation of the cooling energy consumption of each device under each edge node, and feeds the results back to the control system to achieve dynamic adjustment, ensuring the best energy-saving and cooling effect for all equipment in the computer room. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a module diagram of a comprehensive energy-saving control system for a data center computer room using liquid cooling in accordance with Embodiment 1 of the present invention;

[0048] Figure 2 This is a diagram of the structure of a federal control connection chain according to Embodiment 1 of the present invention;

[0049] Figure 3 This is a flow chart of a comprehensive energy-saving control method for a data center computer room using liquid cooling according to Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0050] Example 1

[0051] See also Figure 1 , an embodiment provided by the present invention: a comprehensive energy-saving control system for a data center computer room using a liquid cooling method, comprising: an equipment monitoring space module, a hierarchical energy-saving control module, a flow distribution adjustment module and an evaluation feedback module;

[0052] The equipment monitoring space module is used to monitor the working status, temperature status and working status data of the liquid cooling device corresponding to each equipment in the computer room in real time, and to build a monitoring network;

[0053] The equipment monitoring space module includes a state monitoring unit and a cluster monitoring network construction unit;

[0054] The state monitoring unit is used to monitor the working state, temperature state and working state data of the liquid cooling device corresponding to each device in the computer room in real time, and pre-process the collected data;

[0055] The cluster monitoring network construction unit is used to construct a cluster monitoring network for equipment cooling and energy-saving control based on the distribution status of each device in the computer room, the distribution status of the liquid cooling device, and the working status, temperature status, and working status data of the liquid cooling device corresponding to the device through the graph network node algorithm and the clustering algorithm; further, the cluster monitoring network in this embodiment is composed of N equipment sub-monitoring networks clustered by the clustering algorithm;

[0056] Furthermore, in this embodiment, each edge node is composed of all devices under a device sub-monitoring network and configured liquid cooling devices; the steps of constructing the cluster monitoring network for device temperature reduction and energy saving control include:

[0057] A1. Input the working status, temperature status and working status data of each device in the computer room obtained through monitoring into the distributed high-pass filter model configured on each edge node for preprocessing, and save the preprocessed status data on the corresponding edge node;

[0058] A2. Input the load, operation mode, temperature change trend under working state and working parameters of the liquid cooling device corresponding to the equipment in the state data saved by all edge nodes into the conditional clustering algorithm, and cluster the equipment that meets the same clustering conditions into one category to obtain a cluster containing N Cluster monitoring network of equipment sub-monitoring networks;

[0059] Furthermore, in this embodiment, the device sub-monitoring network is composed of devices with the same clustering conditions. n A device and a corresponding liquid cooling device are constructed;

[0060] Furthermore, the clustering conditions in this embodiment include that the load states of the devices are in the same load change interval, the temperature change rates of the devices per unit time are the same, and the heat dissipation rates of the corresponding liquid cooling devices per unit time are the same.

[0061] A3. Obtain the load status of each device in the device sub-monitoring network, and use the device with the minimum load status at the current moment as the master sub-node of the edge node corresponding to the current device sub-monitoring network, and save the pre-processed status data to the master sub-node, monitor the load status of all devices in the current device sub-monitoring network in real time, and adjust the master sub-node in real time when the load status of the device corresponding to the master sub-node is not minimum.

[0062] This application realizes high-precision and intelligent liquid cooling system management through the equipment monitoring space module. First, the status monitoring unit collects the working status, temperature status and working status data of each equipment in the computer room and the corresponding liquid cooling device in real time, and pre-processes the data through a distributed high-pass filtering model to ensure the accuracy and consistency of the data. This pre-processing step effectively reduces noise interference and improves the reliability of subsequent analysis; secondly, the clustering monitoring network construction unit uses a graph network node algorithm and a conditional clustering algorithm to classify the equipment that meets the same clustering conditions into one category according to multi-dimensional data such as the distribution status of the equipment, load change trend, temperature change rate and heat dissipation rate of the liquid cooling device, forming N equipment sub-monitoring networks; each equipment sub-monitoring network consists of n The control strategy of each device sub-monitoring network is always optimal by selecting the device with the smallest load at the current moment as the master node and monitoring and adjusting the master node in real time. This dynamic adjustment mechanism ensures the flexibility and response speed of the system, and can optimize the cooling strategy in time when the load changes, avoiding unnecessary energy waste. At the same time, it increases the number of devices that can be processed by the present invention and improves the adaptability of the system.

[0063] The hierarchical energy-saving control module is used to perform hierarchical control on the cooling process of each device according to the status data of each device and liquid cooling device acquired by the cluster monitoring network; the hierarchical energy-saving control module includes a federal distributed control unit and an edge difference control unit;

[0064] The federal distributed control unit is used to N The equipment sub-monitoring network and the overall equipment operation status of the computer room after clustering by the clustering algorithm are constructed, M edge nodes and central control nodes are constructed, and the first-level node coordination control strategy is generated to perform collaborative energy-saving and cooling control on each clustering node; N and M are the same in value;

[0065] For further information, see Figure 2In this embodiment, the federated distributed control unit is further configured with a federated control connection chain; the federated control connection chain includes an external connection sub-chain and an internal connection sub-chain; the external connection sub-chain is constructed by a federation algorithm and a graph algorithm based on the interaction data between M edge nodes and the central control node; the internal connection sub-chain is constructed by a graph algorithm based on the interaction data between each edge node and the interaction data between all devices in each edge node;

[0066] Figure 2 A in the figure represents the central control node, 1, 2, 3, ..., M represents the number corresponding to each edge node;

[0067] In this embodiment, a1, a2, a3 in 1 represent the serial number of each device respectively, and the corresponding letters in 2, 3, ..., M also represent the same meaning, which will not be repeated here; Figure 2 The solid line connection in the figure represents the connection relationship in the inner connection sub-chain; the dotted line represents the connection relationship between the outer connection sub-chains;

[0068] Furthermore, the central control node in this embodiment configures a federation reinforcement model; the federation reinforcement model includes M The step of generating the first-level node coordination control strategy comprises:

[0069] B1. According to the working status of the equipment corresponding to each edge node, the equipment temperature change trend and the cooling target requirement, the required cooling liquid allocation amount for each edge node in the external connection sub-chain is obtained;

[0070] B2. Feedback the total amount of coolant distribution required by all edge nodes at the current moment to the reinforcement sub-model corresponding to the central control node, perform initial coolant distribution, and obtain the first-level node coordination control strategy. At the same time, use the cooling energy consumption evaluation node network to evaluate the initial coolant distribution energy consumption of each edge node in the externally connected sub-chain to obtain the initial distribution energy consumption evaluation value.

[0071] This process achieves efficient and intelligent liquid cooling system management through the design of hierarchical energy-saving control modules, especially the federated distributed control unit and edge difference control unit. First, the federated distributed control unit builds M edge nodes and a central control node according to the N equipment sub-monitoring networks and the overall equipment operation status of the computer room, and generates a first-level node coordination control strategy. This design not only improves the flexibility of the system, but also ensures the coordinated energy-saving and cooling control among the cluster nodes. Furthermore, the external connection sub-chain and the internal connection sub-chain in the federated control connection chain are constructed through the federated algorithm and the graph algorithm, respectively processing the interaction data between the central control node and the edge node, and between the edge nodes. This multi-level connection mechanism enhances the communication efficiency and data synchronization of the system, ensuring the timely transmission and execution of control instructions. In particular, the federated reinforcement model configured by the central control node combines the reinforcement sub-models of the M edge nodes, which can accurately calculate the required cooling liquid allocation for each edge node according to the working status, temperature change trend and cooling target requirements of the corresponding equipment of each edge node.

[0072] The edge difference control unit is used to generate a secondary difference cooling and energy-saving control strategy for the energy-saving and cooling process of the liquid cooling device of each device under the edge node according to the primary node coordination control strategy, the working status of the corresponding equipment under each edge node and the cooling target demand and the working status of the liquid cooling device corresponding to the equipment;

[0073] Furthermore, the steps of constructing and training the federated reinforcement model in this embodiment include:

[0074] C1. Construct the current input state according to the working state and temperature state of each device under the corresponding edge node, the working state data of the liquid cooling device corresponding to each device, the cooling target requirements of each device, the cooling liquid volume and cooling liquid circulation rate of each device cooling target requirements, and the energy-saving and cooling state of each edge node in the internally connected sub-chain;

[0075] ;

[0076] in, Indicates the current t Moment m The input state corresponding to the edge node, Indicates m The edge node i The load state corresponding to each device at the current moment. Furthermore, the load state in this implementation is constructed by CPU usage, memory usage and disk I / O load; Indicates m The edge node i The core temperature of the device at the current moment, Indicates m The edge nodei The current shell temperature of a device, Indicates m The edge node i The cooling liquid flow rate of the liquid cooling device corresponding to each device at the current moment, Indicates m The edge node i The rate of temperature change of a device, Indicates m The edge node i The current inlet and outlet temperature difference of the liquid cooling device corresponding to each device is: Indicates m The edge node i The current coolant circulation rate of each device, Indicates m The edge node i The temperature threshold for normal operation of a device. Indicates k The input state parameters of the edge nodes at the current moment, Indicates m The edge node i Devices currently t The comprehensive energy consumption value of cooling according to the primary coolant distribution, the secondary coolant distribution and the secondary differential cooling energy-saving control strategy corresponding to the equipment at all times; further, the comprehensive energy consumption value in this embodiment includes the primary distribution energy consumption evaluation value;

[0077] Furthermore, the steps of constructing and training the federated reinforcement model in this embodiment also include:

[0078] C2. Set the current moment trigger information according to the current moment input state, including:

[0079] When m Among the edge nodes i The corresponding device If it is not the minimum, the first action execution information is triggered, that is, the current m The master sub-node of the edge node and the state data obtained after the current moment are stored in the adjusted master sub-node;

[0080] When m Among the edge nodes i The current comprehensive average temperature of the device is greater than When the corresponding temperature change rate is greater than the temperature change rate corresponding to the normal operation of the device, the second action execution information is triggered, that is, the current i The cooling liquid circulation rate of the liquid cooling device corresponding to each device makes the device temperature less than ;

[0081] When the second action execution information is triggered, the device temperature does not meet the requirement of less than , then the third action execution information is triggered, that is, while keeping the initial allocation amount of the current edge node coolant unchanged, according to the current edge node cooling target demand and the internal connection sub-chain interaction information corresponding to the current edge node, the secondary coolant allocation is redistributed so that the corresponding temperature of all devices at the current edge node is less than ;

[0082] When the third action execution information is triggered, the device temperature at the current edge node is still not less than When , the fourth action execution information is triggered, that is, the initial distribution amount of the coolant of each edge node is adjusted through the internal connection sub-chain interaction information between the edge nodes and the central control node, so that the device temperature of the devices under all edge nodes is less than ;

[0083] When the first to fourth action execution information are triggered, the fifth action execution information is triggered at the same time, that is, when the corresponding action execution information is triggered to cool down the device, the comprehensive energy consumption corresponding to the cooling process of the corresponding action execution information is evaluated through the cooling energy consumption evaluation node network;

[0084] C3, setting the current execution action according to the trigger information at the current moment, and constructing the energy-saving and cooling reward function through the comprehensive energy consumption obtained through the fifth action execution information;

[0085] C4. Repeat the C1-C3 process to obtain the current input state, current trigger information, current execution action and energy-saving and cooling reward function corresponding to each edge node, and input them into the corresponding reinforcement sub-model. Through the federal algorithm, the reinforcement sub-models configured for all edge nodes and the reinforcement sub-models corresponding to the central control node are jointly trained to obtain the trained federal reinforcement model, and the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy corresponding to the current minimum energy consumption are adjusted in real time.

[0086] The first-level node coordination control strategy is a process for allocating and adjusting the amount of coolant for each edge node, and the second-level differential cooling and energy-saving control strategy is a process for real-time adjustment of the amount of coolant for each device under the edge node, control and adjust the operating parameters of the liquid cooling device, and evaluate the energy consumption of the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy corresponding to the cooling process of each device.

[0087] This process achieves highly intelligent and refined liquid cooling system management through the edge difference control unit and the federated reinforcement model. First, the edge difference control unit generates a secondary differential cooling and energy-saving control strategy based on the primary node coordination control strategy, the working status of each device and the cooling demand, ensuring that each device can get the most suitable cooling solution for itself, avoiding the problem of insufficient heat dissipation or overcooling of some devices. The federated reinforcement model accurately processes the multi-dimensional status data of the device under each edge node, including load status, temperature change rate, coolant flow rate, etc., and adjusts the cooling strategy in real time in combination with the initial energy consumption assessment value. In particular, when the device temperature is abnormal or the master and child nodes need to be replaced, the system can quickly trigger the corresponding action execution information, dynamically adjust the coolant circulation rate or re-distribute the secondary coolant to ensure that the temperature of all devices remains within a safe range. Furthermore, the comprehensive energy consumption of each cooling process is evaluated through the fifth action execution information, and an energy-saving cooling reward function is constructed, so that the system can minimize energy consumption while ensuring the cooling effect. This closed-loop feedback mechanism ensures the system's adaptability and optimization capabilities, continuously improves the cooling strategy, and improves energy efficiency.

[0088] The flow distribution and adjustment module is used to perform flow distribution and energy-saving control and adjustment on the liquid cooling device of each device;

[0089] The flow distribution and regulation module includes a flow distribution unit and a temperature reduction control regulation;

[0090] The flow distribution unit is used to distribute the coolant flow to each device according to the secondary differential cooling and energy-saving control strategy and cooling demand corresponding to each device, and coordinately adjust the allocation amount of each edge node through the primary node coordination control strategy;

[0091] The cooling control and adjustment unit differentially adjusts the cooling liquid rate and flow rate in the liquid cooling device corresponding to the device according to the secondary differential cooling and energy-saving control strategy and cooling demand corresponding to the device under each edge node, so that the temperature of the corresponding device is lower than the normal working temperature value in real time;

[0092] The evaluation feedback module is used to construct a cooling energy consumption evaluation node network through a graph node algorithm, conduct real-time evaluation and monitoring of the cooling energy consumption of the equipment under each edge node, and feed back the monitoring results to the hierarchical energy-saving control module to make real-time adjustments to the primary node coordination control strategy and the secondary differential cooling energy-saving control strategy, so that the cooling energy consumption of each device is always lower than the cooling energy consumption before hierarchical adjustment. Furthermore, the cooling energy consumption before hierarchical adjustment in this embodiment is the cooling energy consumption after not being adjusted through the energy-saving and cooling process in this embodiment.

[0093] Furthermore, in the present embodiment, the cooling energy consumption assessment node network corresponds to the cluster monitoring network, and each sub-assessment network in the cooling energy consumption assessment node network corresponds one-to-one to each device in the cluster monitoring network; further, the cooling energy consumption assessment node network in the present embodiment is constructed through a comprehensive model evaluation algorithm.

[0094] This process achieves refined flow distribution and real-time energy-saving control of the liquid cooling system through the flow distribution adjustment module and the evaluation feedback module. The flow distribution unit accurately distributes the coolant flow according to the secondary differential cooling energy-saving control strategy and cooling requirements corresponding to each device, and ensures the coordinated adjustment between the edge nodes through the primary node coordination control strategy, avoiding the problem of uneven flow distribution; the cooling control adjustment unit further dynamically adjusts the coolant rate and flow in the liquid cooling device corresponding to the device according to the specific needs of the device under each edge node, ensuring that the device temperature is always kept within a safe range to prevent overheating or overcooling. This differential adjustment mechanism not only improves cooling efficiency, but also significantly reduces unnecessary energy consumption.

[0095] In addition, the evaluation feedback module builds a cooling energy consumption evaluation node network through the graph node algorithm, monitors the cooling energy consumption of the equipment under each edge node in real time, and feeds the results back to the hierarchical energy-saving control module. This closed-loop feedback mechanism enables the system to make real-time adjustments based on actual operating conditions, ensuring that the cooling energy consumption of each device is always lower than the level before hierarchical adjustment, that is, the optimized energy consumption performance is better than the traditional method.

[0096] Exemplarily, according to the above process, an example is presented through the following process;

[0097] Data center room equipment monitoring and clustering:

[0098] Assume that a large data center has 1,000 servers. The status monitoring unit monitors the CPU usage (working status), motherboard temperature and CPU temperature (temperature status) of each server in real time, as well as the coolant flow and pressure (working status) of the liquid cooling device. These data are collected once a second, and the collected data first enters the distributed high-pass filter model for preprocessing.

[0099] For example, the CPU usage data of a server may produce some spike noise due to momentary system fluctuations. After filtering, a smoother and more accurate usage curve is obtained, which can better reflect its actual load situation.

[0100] Then the cluster monitoring network construction unit clusters the processed data according to clustering conditions such as equipment load status (such as CPU usage in different intervals such as 0-20%, 20%-40%), equipment temperature change rate per unit time (such as temperature rise of no more than 0.5°C per minute and other same change rate ranges), and heat dissipation rate of the corresponding liquid cooling device per unit time (such as heat dissipation power per second within a certain range); assuming there are 5 edge nodes, after clustering, each edge node manages about 200 servers with similar characteristics, forming 5 equipment sub-monitoring networks. For example, the servers managed by edge node 1 may all be devices with low load, slow temperature changes, and similar liquid cooling requirements.

[0101] In the device sub-monitoring network, suppose one of the servers managed by edge node 1 has the lowest load at a certain moment, and it is selected as the primary sub-node at the current moment. If the load of this server increases and is no longer the lowest, the system will monitor it in real time and adjust the primary sub-node to the server with the lowest load at the moment to ensure the optimization of the management strategy.

[0102] Federated Distributed Control and Policy Generation:

[0103] The federal distributed control unit builds five edge nodes and one central control node based on these five equipment sub-monitoring networks and the overall equipment operation status of the computer room (such as overall energy consumption, average temperature, etc.), and generates a first-level node coordination control strategy.

[0104] For example, according to the working status of the equipment corresponding to each edge node, the temperature change trend of the equipment and the cooling target requirements (assuming that the target temperature of the server is 25°C), the required amount of coolant distribution for each edge node in the external connection sub-chain is calculated. If the overall temperature of the server managed by edge node 1 has an upward trend and the load increases, the reinforcement sub-model corresponding to the central control node will calculate that the edge node needs to increase the amount of coolant distribution, perform the initial distribution of coolant, and obtain the first-level node coordination control strategy. At the same time, the cooling energy consumption evaluation node network will evaluate the energy consumption of the initial coolant distribution of edge node 1, assuming that the evaluation value is 100 units of energy consumption (this is a hypothetical value, which is actually obtained based on the evaluation algorithm).

[0105] When building and training the federated reinforcement model configured by the central control node, the current input status will be constructed based on the detailed status data of each device under the corresponding edge node.

[0106] For example, the current CPU usage of a server under edge node 1 is 30% (load status), the CPU core temperature is 50°C (core temperature), the motherboard temperature is 45°C (shell temperature), the current coolant flow of the liquid cooling device is 5L / min (secondary distributed coolant flow), the temperature change rate is 0.3°C per minute, the inlet and outlet temperature difference of the liquid cooling device is 5°C, the coolant circulation rate is 0.5m / s, the normal working temperature threshold is 25°C, and the average temperature of adjacent edge nodes in the inner connected sub-chain is 40°C. Parameters such as these construct the input state.

[0107] When a server in edge node 1 meets the triggering conditions, such as the temperature change rate of the server exceeds the temperature change rate corresponding to normal operation (assuming that the normal change rate is 0.2°C per minute, and the server reaches 0.4°C per minute) and the comprehensive average temperature is greater than 30°C, the second action execution information is triggered, and the system will adjust the coolant circulation rate of the equipment liquid cooling device to reduce its temperature. If the temperature still does not meet the requirements after adjustment, the third action execution information will be triggered. While keeping the initial allocation of coolant at the current edge node unchanged, the secondary coolant allocation will be redistributed according to the current edge node cooling demand at the current moment and the internal connection sub-chain interaction information corresponding to the current edge node (such as whether the adjacent edge node has remaining cooling capacity to borrow, etc.).

[0108] Edge difference control and strategy adjustment:

[0109] The edge difference control unit generates a secondary difference cooling and energy-saving control strategy for the liquid cooling device of each device under the edge node based on the primary node coordination control strategy, the working status and cooling requirements of the corresponding equipment under each edge node, and the working status of the liquid cooling device corresponding to the equipment.

[0110] For example, for a high-load server under edge node 1, according to its specific load conditions (such as CPU usage rate continuously above 80%) and cooling requirements (quick cooling to the target temperature), based on the first-level node coordination control strategy, its liquid cooling device is secondary coolant distribution, the coolant flow is increased, and the liquid cooling pump speed is adjusted (liquid cooling device operation parameter control adjustment) to improve heat dissipation efficiency. At the same time, the energy consumption of the server corresponding to the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy during the cooling process is evaluated. It is assumed that after adjustment, the energy consumption is reduced from the original 50 units to 40 units, achieving the energy saving target.

[0111] Flow distribution adjustment and energy consumption evaluation feedback:

[0112] The flow distribution unit distributes the coolant flow to each device according to the corresponding secondary differential cooling and energy-saving control strategy and cooling requirements of each device. For example, for the above-mentioned high-load server, the coolant flow is increased from the original 5L / min to 8L / min according to its secondary strategy, and the overall allocation of edge node 1 is coordinated and adjusted through the primary node coordination control strategy to ensure that the flow of other devices is not affected and the coolant distribution of the entire edge node is reasonable.

[0113] The cooling control and regulation unit makes differential adjustments to the cooling liquid rate and flow rate in the corresponding liquid cooling device of each edge node according to the corresponding secondary differential cooling and energy-saving control strategy and cooling requirements of the equipment under each edge node. For example, in edge node 1, for servers whose temperature is close to the upper limit of normal working temperature (such as 24°C), the cooling liquid rate is appropriately reduced to reduce energy consumption; while for servers with higher temperature (such as 30°C), the cooling liquid rate is increased to speed up heat dissipation.

[0114] The evaluation feedback module uses a cooling energy consumption evaluation node network constructed by a graph node algorithm to conduct real-time evaluation and monitoring of the cooling energy consumption of equipment at each edge node.

[0115] For example, the energy consumption of each server under edge node 1 is continuously monitored. If it is found that the energy consumption of a server is still higher than expected after adjustment, the monitoring result is fed back to the hierarchical energy-saving control module. Based on the feedback, the hierarchical energy-saving control module adjusts the primary node coordination control strategy and the secondary differential cooling energy-saving control strategy of the server again, such as further optimizing the coolant flow distribution or adjusting the operating parameters of the liquid cooling device, so that the cooling energy consumption of the server is always lower than the cooling energy consumption before hierarchical adjustment, achieving continuous energy-saving optimization.

[0116] Example 2

[0117] Another embodiment provided by the present invention is a workflow of a comprehensive energy-saving control system for a data center computer room using liquid cooling, including:

[0118] First, the state monitoring unit is used to monitor and obtain the working state, temperature state and working state data of the liquid cooling device corresponding to each device in the computer room in real time;

[0119] Second, according to the working status data and distribution status of each device and the liquid cooling device, a cluster monitoring network for equipment cooling and energy saving control is constructed through the cluster monitoring network construction unit; the cluster monitoring network consists of N It is composed of a sub-monitoring network of devices clustered by a clustering algorithm;

[0120] Third, according to N The equipment sub-monitoring network status and the overall equipment operation status data of the computer room are constructed through the federal distributed control unitM edge nodes and central control nodes, and generate a first-level node coordination control strategy;

[0121] Fourth, according to the primary node coordination control strategy, the working status and cooling requirements of the corresponding equipment under each edge node, and the difference in the working status of the liquid cooling device corresponding to the equipment, the edge difference control unit generates a secondary difference cooling and energy-saving control strategy corresponding to the liquid cooling device configured for each device under the edge node;

[0122] Fifth, according to the generated first-level node coordination control strategy and second-level differential cooling and energy-saving control strategy, the flow distribution unit is used to distribute and coordinate the cooling liquid flow of each device, and the cooling control adjustment unit is used to differentially adjust the cooling liquid rate and flow in the liquid cooling device corresponding to the device under each edge node; at the same time, the cooling energy consumption evaluation node network constructed by the evaluation feedback module is used to evaluate the cooling energy consumption of each device under each edge node in real time, and the evaluation results are fed back to the federal distributed control unit and the edge differential control unit, and the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy are adjusted in real time, so that the cooling energy consumption of each device is always lower than the cooling energy consumption before the hierarchical adjustment.

[0123] Example 3

[0124] See also Figure 3 Another embodiment provided by the present invention is a comprehensive energy-saving control method for a data center computer room using liquid cooling, the steps comprising:

[0125] S1. Based on the working status, temperature status of each device in the monitored computer room and the working status data of the liquid cooling device corresponding to each device, a cluster monitoring network for equipment cooling and energy-saving control is constructed through a graph network node algorithm and a clustering algorithm;

[0126] The cluster monitoring network consists of N It is composed of a sub-monitoring network of devices clustered by a clustering algorithm;

[0127] S2. According to N The equipment sub-monitoring network and the overall equipment operation status of the computer room after clustering by the clustering algorithm are constructed, M edge nodes and central control nodes are constructed, and the first-level node coordination control strategy is generated to perform coordinated energy-saving and cooling control on each edge node;

[0128] S3. According to the difference between the working state and cooling demand of the corresponding equipment under each edge node and the working state of the liquid cooling device corresponding to the equipment, a secondary differential cooling and energy-saving control strategy is generated for the cooling process of the liquid cooling device corresponding to each equipment under the edge node;

[0129] S4. According to the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy, the cooling liquid is coordinated and distributed to the equipment under each edge node, and the cooling liquid rate and flow rate in the corresponding liquid cooling device of the equipment are differentially adjusted;

[0130] S5. Build a cooling energy consumption assessment node network to conduct real-time assessment and monitoring of the cooling energy consumption of equipment at each edge node, and make real-time adjustments to the first-level node coordination control strategy and the second-level differential cooling energy-saving control strategy based on the monitoring results.

[0131] Example 4

[0132] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, a comprehensive energy-saving control method for a data center computer room using a liquid cooling method is implemented.

[0133] A computer-readable storage medium stores computer instructions, which, when executed, execute a comprehensive energy-saving control method for a data center computer room using liquid cooling.

[0134] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are within the protection of the present invention.

[0135] If the disclosed technical solution involves personal information, the product using the disclosed technical solution has clearly informed the personal information processing rules and obtained the individual's voluntary consent before processing the personal information. If the disclosed technical solution involves sensitive personal information, the product using the disclosed technical solution has obtained the individual's separate consent before processing the sensitive personal information, and at the same time meets the "explicit consent" requirement. For example, on personal information collection devices such as cameras, clear and prominent signs are set to inform that the personal information collection scope has been entered and personal information will be collected. If the individual voluntarily enters the collection scope, it is deemed that he or she agrees to the collection of his or her personal information; or on the device that processes personal information, when the personal information processing rules are notified by obvious signs / information, the individual's authorization is obtained through pop-up information or by asking the individual to upload his or her personal information; among them, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the type of personal information processed.

Claims

1. A comprehensive energy-saving control system for a data center computer room using liquid cooling, characterized in that: include: Equipment monitoring space module and hierarchical energy-saving control module; The equipment monitoring space module includes a state monitoring unit and a cluster monitoring network construction unit; the hierarchical energy-saving control module includes a federal distributed control unit and an edge difference control unit; the state monitoring unit is used to monitor the working state, temperature state and working state data of the liquid cooling device corresponding to each device in the computer room in real time; according to the working state data and distribution state monitored by each device and the liquid cooling device, a cluster monitoring network for equipment cooling and energy-saving control is constructed through the cluster monitoring network construction unit; the cluster monitoring network consists of N It is composed of a sub-monitoring network of devices clustered by a clustering algorithm; according to N The device sub-monitoring network status and the overall equipment operation status data of the computer room are collected, and M edge nodes and central control nodes are constructed through a federal distributed control unit, and a first-level node coordination control strategy is generated; according to the first-level node coordination control strategy, the working status and cooling requirements of the corresponding equipment under each edge node, and the difference in the working status of the liquid cooling device corresponding to the equipment, a second-level differential cooling and energy-saving control strategy for the liquid cooling device corresponding to each device under the edge node is generated through the edge difference control unit.

2. The comprehensive energy-saving control system for a data center computer room using liquid cooling as claimed in claim 1, characterized in that: The system also includes a flow distribution adjustment module and an evaluation feedback module; the flow distribution adjustment module includes a flow distribution unit and a temperature reduction control adjustment unit; According to the generated first-level node coordination control strategy and second-level differential cooling and energy-saving control strategy, the flow distribution unit is used to distribute and coordinate the coolant flow of each device, and the cooling control adjustment unit is used to differentially adjust the coolant rate and flow in the liquid cooling device corresponding to the device under each edge node; at the same time, the cooling energy consumption evaluation node network constructed by the evaluation feedback module is used to evaluate the cooling energy consumption of each device under each edge node in real time, and the evaluation results are fed back to the federal distributed control unit and the edge differential control unit, so as to make real-time adjustments to the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy.

3. The comprehensive energy-saving control system for a data center computer room using liquid cooling as claimed in claim 2, characterized in that: Each edge node is composed of all devices under a device sub-monitoring network and configured liquid cooling devices; the steps of constructing the cluster monitoring network for equipment cooling and energy-saving control include: A1. Input the working status, temperature status and working status data of each device in the computer room obtained through monitoring into the distributed high-pass filter model configured on each edge node for preprocessing, and save the preprocessed status data on the corresponding edge node; A2. Input the load, operation mode, temperature change trend under working state and working parameters of the liquid cooling device corresponding to the equipment in the state data saved by all edge nodes into the conditional clustering algorithm, and cluster the equipment that meets the same clustering conditions into one category to obtain a cluster containing N The device sub-monitoring network is composed of devices with the same clustering condition. n A device and a corresponding liquid cooling device are constructed; A3. Obtain the load status of each device in the device sub-monitoring network, and use the device with the minimum load status at the current moment as the master sub-node of the edge node corresponding to the current device sub-monitoring network, and save the pre-processed status data to the master sub-node, monitor the load status of all devices in the current device sub-monitoring network in real time, and adjust the master sub-node in real time when the load status of the device corresponding to the master sub-node is not minimum.

4. The comprehensive energy-saving control system for a data center computer room using liquid cooling as claimed in claim 3, characterized in that: The clustering conditions include that the load states of the devices are in the same load change interval, the temperature change rates of the devices per unit time are the same, and the heat dissipation rates of the corresponding liquid cooling devices per unit time are the same; The federal distributed control unit is also configured with a federal control connection chain; the federal control connection chain includes an external connection sub-chain and an internal connection sub-chain; the external connection sub-chain is constructed through a federal algorithm and a graph algorithm based on the interaction data between M edge nodes and the central control node; the internal connection sub-chain is constructed through a graph algorithm based on the interaction data between each edge node and the interaction data between all devices in each edge node.

5. The integrated energy-saving control system for a data center computer room using liquid cooling as claimed in claim 4, characterized in that: The central control node is configured with a federated reinforcement model; the federated reinforcement model includes reinforcement sub-models configured by M edge nodes and a reinforcement sub-model configured by the central control node; The step of generating the first-level node coordination control strategy includes: B1. According to the working status of the equipment corresponding to each edge node, the equipment temperature change trend and the cooling target requirement, the required cooling liquid allocation amount for each edge node in the external connection sub-chain is obtained; B2. Feedback the total amount of coolant distribution required by all edge nodes at the current moment to the reinforcement sub-model corresponding to the central control node, perform initial coolant distribution, and obtain the first-level node coordination control strategy. At the same time, use the cooling energy consumption evaluation node network to evaluate the initial coolant distribution energy consumption of each edge node in the externally connected sub-chain to obtain the initial distribution energy consumption evaluation value.

6. The data center computer room comprehensive energy-saving control system using liquid cooling as claimed in claim 5, characterized in that: The steps of building and training the federated reinforcement model include: C1. Construct the current input state according to the working state and temperature state of each device under the corresponding edge node, the working state data of the liquid cooling device corresponding to each device, the cooling target requirements of each device, the cooling liquid volume and cooling liquid circulation rate of each device cooling target requirements, and the energy-saving and cooling state of each edge node in the internally connected sub-chain; C2. Set the current moment trigger information according to the current moment input state, including: When m Among the edge nodes i The corresponding device If it is not the minimum, the first action execution information is triggered, that is, the current m The master sub-node of the edge node, and the state data obtained after the current moment is stored in the adjusted master sub-node; Indicates m The edge node i The load status of each device at the current moment.

7. The integrated energy-saving control system for a data center computer room using liquid cooling as claimed in claim 6, characterized in that: The steps of building and training the federated reinforcement model also include: When m Among the edge nodes i The current comprehensive average temperature of the device is greater than When the corresponding temperature change rate is greater than the temperature change rate corresponding to the normal operation of the device, the second action execution information is triggered, that is, the current i The cooling liquid circulation rate of the liquid cooling device corresponding to each device makes the device temperature less than ;in Indicates m The edge node i The temperature threshold for normal operation of a device; When the second action execution information is triggered, the device temperature does not meet the requirement of less than , then the third action execution information is triggered, that is, while keeping the initial allocation amount of the current edge node coolant unchanged, according to the current edge node cooling target demand and the internal connection sub-chain interaction information corresponding to the current edge node, the secondary coolant allocation is redistributed so that the corresponding temperature of all devices at the current edge node is less than ; When the third action execution information is triggered, the device temperature at the current edge node is still not less than When , the fourth action execution information is triggered, that is, the initial distribution amount of the coolant of each edge node is adjusted through the internal connection sub-chain interaction information between the edge nodes and the central control node, so that the device temperature of the devices under all edge nodes is less than ; When the first to fourth action execution information are triggered, the fifth action execution information is triggered at the same time, that is, when the corresponding action execution information is triggered to cool down the equipment, the comprehensive energy consumption corresponding to the cooling process of the corresponding action execution information is evaluated through the cooling energy consumption evaluation node network.

8. The integrated energy-saving control system for a data center computer room using liquid cooling as claimed in claim 7, characterized in that: The steps of building and training the federated reinforcement model also include: C3, setting the current execution action according to the trigger information at the current moment, and constructing the energy-saving and cooling reward function through the comprehensive energy consumption obtained through the fifth action execution information; C4. Repeat the C1-C3 process to obtain the current input state, current trigger information, current execution action and energy-saving and cooling reward function corresponding to each edge node, and input them into the corresponding reinforcement sub-model. Through the federal algorithm, the reinforcement sub-models configured for all edge nodes and the reinforcement sub-models corresponding to the central control node are jointly trained to obtain the trained federal reinforcement model, and the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy corresponding to the current minimum energy consumption are adjusted in real time.

9. A comprehensive energy-saving control method for a data center computer room using liquid cooling, which is implemented based on the comprehensive energy-saving control system for a data center computer room using liquid cooling as claimed in any one of claims 1 to 8, characterized in that: include: S1. Based on the working status, temperature status of each device in the monitored computer room and the working status data of the liquid cooling device corresponding to each device, a cluster monitoring network for equipment cooling and energy-saving control is constructed through a graph network node algorithm and a clustering algorithm; The cluster monitoring network consists of N It is composed of a sub-monitoring network of devices clustered by a clustering algorithm; S2. According to N The equipment sub-monitoring network and the overall equipment operation status of the computer room after clustering by the clustering algorithm are constructed, M edge nodes and central control nodes are constructed, and the first-level node coordination control strategy is generated to perform coordinated energy-saving and cooling control on each edge node; S3. According to the difference between the working state and cooling demand of the corresponding equipment under each edge node and the working state of the liquid cooling device corresponding to the equipment, a secondary differential cooling and energy-saving control strategy is generated for the cooling process of the liquid cooling device corresponding to each equipment under the edge node; S4. According to the first-level node coordination control strategy and the second-level differential cooling and energy-saving control strategy, the cooling liquid is coordinated and distributed to the equipment under each edge node, and the cooling liquid rate and flow rate in the corresponding liquid cooling device of the equipment are differentially adjusted; S5. Build a cooling energy consumption assessment node network to conduct real-time assessment and monitoring of the cooling energy consumption of equipment at each edge node, and make real-time adjustments to the first-level node coordination control strategy and the second-level differential cooling energy-saving control strategy based on the monitoring results.

10. A computer-readable storage medium, characterized in that: Computer instructions are stored thereon, and when the computer instructions are executed, the comprehensive energy-saving control method for a data center computer room using liquid cooling as described in claim 9 is executed.

Images