Search

CN-121996035-A - High-efficiency mixed heat dissipation system and method for server based on liquid cooling and air cooling cooperation

CN121996035ACN 121996035 ACN121996035 ACN 121996035ACN-121996035-A

Abstract

The invention discloses a server efficient hybrid heat dissipation system and method based on liquid cooling and air cooling cooperation, and relates to the field of server management. The method comprises the steps of collecting a state feature set required by heat dissipation scheduling, wherein the state feature set at least comprises real-time dynamic data of a server and a heat dissipation system and preset static parameters, deciding by taking a preset global optimization target as a criterion based on the state feature set, generating an optimal collaborative heat dissipation strategy at the current moment, wherein the optimal collaborative heat dissipation strategy comprises a first control instruction of the liquid cooling subsystem and a second control instruction of the air cooling subsystem, and executing the optimal collaborative heat dissipation strategy to conduct collaborative heat dissipation control on the server. The change of the load, the environment and the economic signals is responded in real time, the globally optimal heat dissipation strategy is dynamically generated and executed, and the situation of fixed strategy and response lag of the existing hybrid heat dissipation scheme is changed.

Inventors

  • GUO YIBIN

Assignees

  • 深圳市欣欣祥荣科技有限公司

Dates

Publication Date
20260508
Application Date
20251230

Claims (10)

  1. 1. The server mixed heat dissipation scheduling method based on the cooperation of liquid cooling and air cooling is characterized by being applied to a data center provided with a liquid cooling subsystem and an air cooling subsystem, and comprises the following steps: collecting a state feature set required by heat dissipation scheduling, wherein the state feature set at least comprises real-time dynamic data of a server and a heat dissipation system and preset static parameters; based on the state feature set, making a decision by taking a preset global optimization target as a criterion, and generating an optimal cooperative heat dissipation strategy at the current moment, wherein the optimal cooperative heat dissipation strategy comprises a first control instruction for the liquid cooling subsystem and a second control instruction for the air cooling subsystem; and executing the optimal cooperative heat dissipation strategy to perform cooperative heat dissipation control on the server.
  2. 2. The efficient hybrid heat dissipation method for a server based on liquid cooling and air cooling coordination according to claim 1, wherein the state feature set comprises: Static features including nominal thermal parameters of the high heat components within the server and rated capacity parameters of the heat dissipation subsystem; The dynamic characteristics comprise real-time power consumption and temperature of a server computing part, working parameters of a heat dissipation subsystem, machine room environment parameters, and externally input energy price signals and waste heat recovery demand signals; policy features that characterize the currently validated operational objective preferences.
  3. 3. The method for efficient hybrid heat dissipation of a server based on liquid cooling and air cooling coordination according to claim 2, wherein the generating an optimal coordinated heat dissipation policy based on a state feature set comprises: judging whether the dynamic characteristics trigger a preset abnormal event or not; if the abnormal event is triggered, a preset quick response rule matched with the abnormal event is called, and the optimal cooperative heat dissipation strategy is generated; if not, generating the optimal collaborative heat dissipation strategy through optimization calculation based on the state feature set, the global optimization target and related constraint conditions.
  4. 4. The method for efficient hybrid heat dissipation of a server based on liquid cooling and air cooling synergy of claim 3, wherein generating an optimal collaborative heat dissipation policy based on a state feature set further comprises: continuously training a reinforcement learning model by utilizing historical operation data; and updating and optimizing the preset quick response rule or the model based on the optimization calculation by using the trained reinforcement learning model.
  5. 5. The efficient hybrid heat dissipation method based on the liquid cooling and air cooling cooperation server according to claim 4, wherein when the optimal cooperation heat dissipation strategy is generated, a target value of the liquid cooling subsystem backwater temperature is dynamically adjusted in response to the externally input energy price signal and the waste heat recovery demand signal, so as to optimize economical efficiency and comprehensive energy utilization.
  6. 6. The efficient hybrid heat dissipation method for the server based on the liquid cooling and air cooling coordination of claim 5, wherein the air cooling subsystem comprises a server internal fan and a machine room level air conditioner, and the process of generating the optimal coordination heat dissipation strategy is to perform unified coordination optimization on control parameters of the liquid cooling subsystem, the server internal fan and the machine room level air conditioner.
  7. 7. A server hybrid heat dissipation scheduling system based on liquid cooling and air cooling cooperation for implementing the method of any one of claims 1 to 6, the system comprising: the characteristic collection module is used for collecting the state characteristic set; The intelligent scheduling engine is connected to the characteristic acquisition module and is used for making a decision based on the state characteristic set to generate the optimal cooperative heat dissipation strategy; and the strategy execution module is connected to the intelligent scheduling engine and used for executing the optimal cooperative heat dissipation strategy and controlling the liquid cooling subsystem and the air cooling subsystem.
  8. 8. The server efficient hybrid heat dissipation system based on liquid cooling and air cooling coordination of claim 7, wherein the intelligent scheduling engine comprises: the event driving unit is used for calling a preset rule generating strategy when the dynamic characteristics trigger a preset abnormal event; and the model prediction control unit is used for generating a strategy through rolling optimization calculation based on the thermodynamic model when no abnormal event is triggered.
  9. 9. The server efficient hybrid heat dissipation system based on liquid cooling and air cooling coordination of claim 8, wherein the intelligent scheduling engine further comprises: And the reinforcement learning unit is used for performing offline training by utilizing the historical operation data and outputting an optimization strategy to update a rule base of the event driving unit or correct an internal model of the model prediction control unit.
  10. 10. The server efficient hybrid heat dissipation system based on liquid cooling and air cooling coordination according to claim 9, further comprising a digital twin module connected to the intelligent scheduling engine for providing a thermodynamic simulation environment of the server and a machine room, for the model predictive control unit to perform predictive computation, and/or for the reinforcement learning unit to perform safety exploration training.

Description

High-efficiency mixed heat dissipation system and method for server based on liquid cooling and air cooling cooperation Technical Field The invention relates to the field of server management, in particular to a server efficient hybrid heat dissipation system and method based on liquid cooling and air cooling cooperation. Background With the explosive growth of cloud computing and artificial intelligence computing demands, the power density of single cabinets of a data center is continuously improved, and the traditional air cooling heat dissipation faces an energy efficiency bottleneck. The liquid cooling technology, particularly cold plate type liquid cooling, can efficiently solve the heat dissipation problem of high heat flux density chips such as CPU, GPU and the like, but the full liquid cooling transformation faces the challenges of high cost, poor compatibility, complex maintenance and the like, so the mixed heat dissipation mode of liquid cooling and air cooling becomes an important direction in the industry. Existing hybrid heat dissipation schemes focus on static combinations at the hardware level, such as configuring a part of the high-heat components with liquid cooling, and the rest of the components are air cooled, however, such schemes have significant drawbacks: Firstly, a heat dissipation strategy is fixed, and dynamic adjustment cannot be performed according to real-time load of a server, an external environment and an operation and maintenance target, so that energy efficiency cannot be optimized; secondly, the system-level cooperative control is lacking, and the liquid cooling subsystem and the air cooling subsystem often operate independently and even interfere with each other, so that energy consumption waste or insufficient heat dissipation capacity is caused; Finally, the potential of hybrid heat dissipation in terms of reducing overall cost of ownership has not been explored without combining the heat dissipation system with the overall economical operation of the data center. Therefore, an intelligent scheduling method is needed, which can dynamically and cooperatively configure the operation parameters of liquid cooling and air cooling based on multidimensional state characteristics, so as to realize comprehensive optimization of energy efficiency, cost and reliability. Disclosure of Invention Therefore, the invention aims to provide a server efficient hybrid heat dissipation system and a method based on the cooperation of liquid cooling and air cooling so as to realize dynamic adjustment of liquid cooling and water cooling and reduce energy consumption. In order to achieve the technical purpose, the invention provides a server mixed heat dissipation scheduling method based on liquid cooling and air cooling cooperation, which is applied to a data center provided with a liquid cooling subsystem and an air cooling subsystem, and comprises the following steps: collecting a state feature set required by heat dissipation scheduling, wherein the state feature set at least comprises real-time dynamic data of a server and a heat dissipation system and preset static parameters; based on the state feature set, making a decision by taking a preset global optimization target as a criterion, and generating an optimal cooperative heat dissipation strategy at the current moment, wherein the optimal cooperative heat dissipation strategy comprises a first control instruction for the liquid cooling subsystem and a second control instruction for the air cooling subsystem; and executing the optimal cooperative heat dissipation strategy to perform cooperative heat dissipation control on the server. Preferably, the state feature set includes: Static features including nominal thermal parameters of the high heat components within the server and rated capacity parameters of the heat dissipation subsystem; The dynamic characteristics comprise real-time power consumption and temperature of a server computing part, working parameters of a heat dissipation subsystem, machine room environment parameters, and externally input energy price signals and waste heat recovery demand signals; policy features that characterize the currently validated operational objective preferences. Preferably, the generating the optimal cooperative heat dissipation policy based on the state feature set includes: judging whether the dynamic characteristics trigger a preset abnormal event or not; if the abnormal event is triggered, a preset quick response rule matched with the abnormal event is called, and the optimal cooperative heat dissipation strategy is generated; if not, generating the optimal collaborative heat dissipation strategy through optimization calculation based on the state feature set, the global optimization target and related constraint conditions. Preferably, the generating the optimal cooperative heat dissipation policy based on the state feature set further includes: continuously training a reinforcement learni