Heat flow control method and heat flow control system

Abstract

A heat flow control method is used for a data center heat dissipation system, and comprises the following steps: determining a plurality of features corresponding to a current scene of the data center heat dissipation system at a first time point; and deciding a plurality of heat dissipation parameters at a second time point according to the features, wherein at the second time point, the data center heat dissipation system performs heat flow control by using the heat dissipation parameters; wherein the second time point lags behind the first time point.

Description

Technical Field

[0001] The present invention relates to a heat flow control method and a heat flow control system, and in particular to a heat flow control method and a heat flow control system that utilize a deep learning method to determine heat dissipation parameters. Background Art

[0002] Airflow control is a key technology to improve the cooling efficiency of data centers or computer rooms. In the process of heat dissipation in computer rooms, cold air and hot air often mix improperly. The mixing of cold and hot air leads to poor cooling efficiency of the computer room and consumes too much electricity. Therefore, the flow management of data centers today usually adopts cold and hot aisle isolation and elevated ventilation floors. The design of cold and hot aisle isolation and elevated floors can reduce the improper mixing of cold and hot air. However, in order to ensure that the data center can dissipate heat in time, a more conservative flow control is usually adopted, which may cause the data center to overcool and consume energy. In this case, how to improve the flow of data centers, balance flow supply and demand, and reduce energy consumption has become one of the goals that the industry is working hard for. Summary of the invention

[0003] One of the main purposes of the present invention is to provide a heat flow control method and a heat flow control system for a data center cooling system to solve the above-mentioned problems.

[0004] The present invention provides a heat flow control method for a data center cooling system, comprising: (a) determining a plurality of features corresponding to a current scene of the data center cooling system at a first time point; and (b) determining a plurality of cooling parameters at a second time point based on the features; wherein, at the second time point, the data center cooling system uses the cooling parameters to perform heat flow control; wherein the second time point lags behind the first time point.

[0005] The present invention provides a heat flow control system for a data center cooling system, comprising a processor; and a memory coupled to the processor, storing a program code for instructing the processor to execute a transmission parameter decision method, the transmission parameter decision method comprising (a) determining a plurality of features corresponding to a current scene of the data center cooling system at a first time point; and (b) determining a plurality of cooling parameters at a second time point based on the features; wherein, at the second time point, the data center cooling system uses the cooling parameters to perform heat flow control; wherein the second time point lags behind the first time point. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 It is a schematic diagram of a heat flow control system according to an embodiment of the present invention;

[0007] Figure 2A schematic diagram of a data center cooling system according to an embodiment of the present invention;

[0008] Figure 3 It is a schematic diagram of the process of the heat flow control method according to an embodiment of the present invention;

[0009] Figure 4 A schematic diagram of a deep neural network according to an embodiment of the present invention;

[0010] Figures 5A to 5B Schematic diagram of a data center cooling system according to an embodiment of the present invention.

[0011] Explanation of symbols

[0012] 1: Heat flow control system

[0013] 2: Process

[0014] 3,5,6: Data center cooling system

[0015] 4: Deep Neural Networks

[0016] 10: Processor

[0017] 20: Memory

[0018] S200, S202, S204: Steps

[0019] 30: Cold air

[0020] 31: Hot Air

[0021] RACK: Rack

[0022] T ref :Cold air temperature

[0023] V: Cold air velocity

[0024] T in :Server inlet temperature

[0025] T out :Server outlet temperature

[0026] P: Server load power

[0027] F: Server fan speed DETAILED DESCRIPTION

[0028] Certain words are used in the specification and subsequent patent applications to refer to specific components. It should be understood by those with ordinary knowledge in the field that hardware manufacturers may use different terms to refer to the same component. This specification and subsequent patent applications do not use differences in names as a way to distinguish components, but use differences in the functions of components as the criteria for distinction. The term "including" mentioned throughout the specification and subsequent patent applications is an open term and should be interpreted as "including but not limited to". In addition, the term "coupled" here includes any direct and indirect electrical connection means. Therefore, if the text describes a first device coupled to a second device, it means that the first device can be directly electrically connected to the second device, or indirectly electrically connected to the second device through other devices or connection means.

[0029] Please refer to Figure 1 , Figure 1 The figure is a schematic diagram of a heat flow control system 1 according to an embodiment of the present invention. The heat flow control system 1 includes a processor 10 and a memory 20. The memory 20 is used to store a program code. The processor 10 executes a heat flow control method with the program code to dynamically adjust a plurality of heat dissipation parameters in a data center heat dissipation system, and performs heat flow control to balance the flow supply and demand and reduce energy consumption. In other words, the supply flow of cold air is equal to the demand flow of the data center heat dissipation system. In one embodiment, the heat flow control system 1 and the executed heat flow control method are applicable to the cooling system of high performance computing equipment (High Performance Computing, HPC) such as cloud data centers, edge computing centers or single cabinets, but are not limited thereto. It should be noted that the heat flow control system 1 only represents the necessary components required to execute the heat flow control method. Its basic structure is well known in the art, so it is not repeated. A person with ordinary knowledge in the art can appropriately add other components as needed, such as a motherboard, a power supply, a PID (Proportional-Integral-Derivative) controller, a sensor, etc., but is not limited to this, or implement the heat flow control system 1 with appropriate devices or equipment. In a more preferred embodiment, if the heat flow control method is applied to a single cabinet, it will be executed by the cabinet management controller; if the heat flow control method is applied to multiple cabinets (such as a row or a compartment unit), it will be executed by the upper management controller of these cabinets.

[0030] Please refer to Figure 2 , Figure 2 FIG. 1 is a schematic diagram of a data center cooling system 3 according to an embodiment of the present invention. Figure 2As shown, the data center cooling system 3 includes a plurality of racks and computer room air conditioning units (CRACs). The racks are, for example, 7U, 14U, or 42U racks, but not limited thereto. Each server is equipped with a server fan and a computer room air conditioning unit (CRAC) fan to guide the heat flow. For example, the CRAC fan guides the cold air 30 into the cold channel; the server fan guides the cold air 30 into the server to dissipate the heat source in the server and exhaust the hot air 31 to the hot channel. It should be noted that Figure 2 This is only an embodiment of the present invention, and a person skilled in the art can appropriately adjust the number of server fans and CRAC fans and the configuration of cold channels and hot channels as needed, without limitation. For example, the data center cooling system 3 can be configured with a raised perforated floor tile to reduce the mixing of cold air 30 and hot air 31.

[0031] The heat flow control method performed by the heat flow control system 1 in the data center cooling system 3 can be summarized as a process 2, such as Figure 3 Process 2 includes the following steps:

[0032] Step S200: Start.

[0033] Step S202: At a first time point, determine a plurality of features corresponding to a current scenario of the cooling system of the data center.

[0034] Step S204: Determine a plurality of heat dissipation parameters at a second time point according to the plurality of features.

[0035] Step S206: End.

[0036] According to process 2, in step S202, the processor 102 of the heat flow control system 1 can determine a plurality of features of a current scene (i.e., the current environment) corresponding to the data center cooling system 3 at a first time point. The plurality of features can include a cold air temperature and a cold air flow rate of the cold air generated by the computer room air conditioner, a server inlet temperature, a server outlet temperature, a server load power, and a server fan speed. In another embodiment, the plurality of features of a historical scene (i.e., the old environment) corresponding to the data center cooling system 3 can be determined at a first time point, for example, the plurality of features in the past 20 seconds. It should be noted that the plurality of features of the current scene of the heat flow control system 1 only represent the necessary features required to execute the heat flow control method. The basic meaning of the plurality of features is well known in the art, so it is not repeated. A person with ordinary knowledge in the art can appropriately add other features to determine the cooling parameters according to the needs, for example, the number of servers, the temperatures of the plurality of main components of the server, or the height of a cabinet, etc.

[0037] In step S204, the processor 102 of the heat flow control system 1 can determine a plurality of heat dissipation parameters at a second time point according to a plurality of features. It should be noted that the second time point lags behind the first time point, for example, the first time point is a current time point, and the second time point is the time point at which the heat flow control system 1 performs heat flow control according to a plurality of heat dissipation parameters next time, or the second time point is the time point at which heat flow control is performed at a certain time in the future. The processor 102 can execute a time series prediction model (Time series prediction model) according to the cold air temperature and cold air flow rate, the server inlet temperature, the server outlet temperature, the server load power and the server fan speed to perform time series prediction of a plurality of heat dissipation parameters. In one embodiment, the plurality of heat dissipation parameters can be a predicted server inlet temperature and a predicted server speed, and the heat flow control system 1 can perform heat flow control according to the time series prediction results of the predicted server inlet temperature and the predicted server speed, so that the heat flow control system 1 can make the server fan and the CRAC fan operate at an appropriate speed, achieve a balance between flow supply and demand and reduce energy consumption.

[0038] In one embodiment, the time series prediction model can use a deep learning method or a reinforcement learning method to predict a plurality of heat dissipation parameters in a complex heat flow environment. For example, the time series prediction model can use a deep neural network (DNN) architecture. Figure 4Schematic diagram of a deep neural network 4 according to an embodiment of the present invention. The deep neural network 4 includes an input layer, a hidden layer, and an output layer. The processor 102 can input the cold air temperature T ref And cold air velocity V, server inlet temperature T in , Server outlet temperature T out , server load power P and server fan speed F are input to the input layer, and a plurality of heat dissipation parameters are obtained from the output layer. In this way, the heat flow control system 1 can make the server fan and the CRAC fan operate at an appropriate speed at the second time point according to a plurality of heat dissipation parameters, so as to achieve a balance between flow supply and demand and reduce energy consumption. It should be noted that the operating principle of the deep neural network should be well known in the art, so it will not be repeated. In addition, the deep learning method or the reinforcement learning method may also adopt the architecture of a deep belief network (DBN), a convolutional neural network (CNN) and a convolutional deep belief network (CDBN), but is not limited to this.

[0039] In one embodiment, the heat flow control system 1 can obtain nine sets of heat dissipation parameters according to a plurality of characteristics and a plurality of rotation speeds of the CRAC fans (for example, 20%, 30% to 90%, a total of nine types). The heat flow control system 1 can calculate nine sets of CRAC fan powers P according to the nine rotation speeds of the CRAC fans and the corresponding nine sets of heat dissipation parameters. AF and server fan power P SF In this way, the heat flow control system 1 can select the speed at which the CRC fan operates, and the total power (CRAC fan power P AF and server fan power P SF The sum of ) is the smallest.

[0040] In one embodiment, please refer to Tables 1 and 2 for the power consumption of the heat flow control system 1 of the embodiment of the present invention. Table 1 shows the power consumption of the heat flow control system 1 when executing the traditional method, and Table 2 shows the power consumption of the heat flow control system 1 when executing the heat flow control method of the embodiment of the present invention. Comparing Tables 1 and 2, it can be seen that the inlet temperatures of servers 1 to 15 of the embodiment of the present invention are all 23 degrees, which means that the heat flow of cold air and hot air is balanced. On the other hand, the inlet temperatures of servers 9 to 15 of the traditional method are between 24 and 28 degrees, which means that the heat flow of cold air and hot air is unbalanced, that is, the hot air overflows to the inlet of servers 9 to 15. Further, since the embodiment of the present invention makes the heat flow of cold air and hot air balanced, the fan can operate at a lower fan speed compared to the traditional method, so that the power consumption is significantly reduced. It can be seen from Tables 1 and 2 that the power consumption of the fan of the embodiment of the present invention is reduced by 38% (1-769.47 / 474.9) compared to the traditional method.

[0041] Table 1

[0042]

[0043]

[0044] Table 2

[0045]

[0046]

[0047] Furthermore, the data center cooling system applicable to the present invention may also include a water-cooled cooling pump (pump), using liquid to remove heat flow, but is not limited to this. For example, please refer to Figures 5A to 5B , Figures 5A to 5B Schematic diagram of data center cooling systems 5 and 6 according to an embodiment of the present invention. Data center cooling systems 5 and 6 respectively use an internal water cooling loop and an external water cooling loop to remove the heat flow of the server inside the rack, and data center cooling systems 5 and 6 execute the heat flow control method of the present invention, so that the server fan and pump operate at an appropriate speed at the second time point, achieving a balance between flow supply and demand and reducing energy consumption.

[0048] It should be noted that the heat flow control system 1 is an embodiment of the present invention. A person with ordinary knowledge in the art can combine, modify or change the above-mentioned embodiments according to the spirit of the present invention, but is not limited thereto. All the above descriptions, steps, and / or processes (including recommended steps) can be implemented through hardware, software, firmware (i.e., a combination of hardware devices and computer instructions, the data in the hardware devices are read-only software data), electronic systems, or a combination of the above devices. The hardware may include analog, digital and hybrid circuits (i.e., microcircuits, microchips or silicon chips). The electronic system may include a system on chip (SoC), a system in package (SiP), a computer on module (CoM) and a heat flow control system 1. The process steps and embodiments of the present invention may exist in the form of program codes or instructions and be stored in the memory 20. The memory 20 may be a computer-readable recording medium, and the memory 20 may include a read-only memory (ROM), a flash memory (Flash Memory), a random-access memory (RAM), a subscriber identity module (SIM), a hard disk, and a CD-ROM / DVD-ROM / BD-ROM, but is not limited thereto. The above-mentioned processes and embodiments may be compiled into program codes or instructions and stored in the memory 20. The processor 10 may be used to read and execute the program codes or instructions stored in the memory 20 to implement all the aforementioned steps and functions.

[0049] It should be noted that in this embodiment, the server of the present invention can be used for artificial intelligence (AI) computing, edge computing, and can also be used as a 5G server, cloud server or Internet of Vehicles server.

[0050] In summary, the heat flow control system and the heat flow control method of the present invention predict the heat dissipation parameters at a future time point based on a plurality of features of the current scene or the historical scene. Thus, compared with the traditional technology, the heat flow control system and the heat flow control method of the present invention can make the server fan and the CRAC fan operate at an appropriate speed, achieve a balance between flow supply and demand and reduce energy consumption.

[0051] The above description is only a preferred embodiment of the present invention. All equivalent changes and modifications made according to the scope of the patent application of the present invention should fall within the scope of the present invention.

Claims

1. A heat flow control method, characterized in that: For a data center cooling system, the heat flow control method includes: (a) at a first time point, determining a plurality of features corresponding to a current scenario of the cooling system of the data center; and (b) determining a plurality of heat dissipation parameters at a second time point based on the characteristics; Wherein, at the second time point, the data center cooling system uses the cooling parameters to perform heat flow control; The second time point lags behind the first time point.

2. The heat flow control method according to claim 1, characterized in that: The characteristics include a cooling air temperature, a cooling air velocity, a server inlet temperature, a server outlet temperature, a server load power, a plurality of main component temperatures, a server fan speed or a server quantity.

3. The heat flow control method according to claim 1, characterized in that: Wherein step (b) further comprises: A deep learning method is used to perform decision fusion on the features to generate the heat dissipation parameters.

4. The heat flow control method according to claim 3, characterized in that: The deep learning method adopts at least one architecture of a deep neural network, a deep belief network, a convolutional neural network and a convolutional deep belief network.

5. The heat flow control method according to claim 1, characterized in that: The heat dissipation parameters include a predicted server inlet temperature and a predicted server rotation speed.

6. A heat flow control system, characterized in that: Used in a data center cooling system, including: a processor; and A memory, coupled to the processor, stores a program code for instructing the processor to execute a transmission parameter decision method, the transmission parameter decision method comprising: (a) at a first time point, determining a plurality of features corresponding to a current scenario of the cooling system of the data center; and (b) determining a plurality of heat dissipation parameters at a second time point based on the characteristics; Wherein, at the second time point, the data center cooling system uses the cooling parameters to perform heat flow control; The second time point lags behind the first time point.

7. The heat flow control system according to claim 6, characterized in that: The characteristics include a cooling air temperature, a cooling air velocity, a server inlet temperature, a server outlet temperature, a server load power, a plurality of main component temperatures, a server fan speed or a server quantity.

8. The heat flow control system according to claim 6, characterized in that: Wherein step (b) further comprises: A deep learning method is used to perform decision fusion on the features to generate the heat dissipation parameters.

9. The heat flow control system according to claim 8, characterized in that: The deep learning method adopts at least one architecture of a deep neural network, a deep belief network, a convolutional neural network and a convolutional deep belief network.

10. The heat flow control system according to claim 6, characterized in that: The heat dissipation parameters include a predicted server inlet temperature and a predicted server rotation speed.

Images