WO2021042339A1 - 散热控制与模型训练方法、设备、系统及存储介质 - Google Patents

散热控制与模型训练方法、设备、系统及存储介质 Download PDF

Info

Publication number
WO2021042339A1
WO2021042339A1 PCT/CN2019/104590 CN2019104590W WO2021042339A1 WO 2021042339 A1 WO2021042339 A1 WO 2021042339A1 CN 2019104590 W CN2019104590 W CN 2019104590W WO 2021042339 A1 WO2021042339 A1 WO 2021042339A1
Authority
WO
WIPO (PCT)
Prior art keywords
refrigeration
dissipated
sample data
heat dissipation
overheating
Prior art date
Application number
PCT/CN2019/104590
Other languages
English (en)
French (fr)
Inventor
赵旭
李栈
卢毅军
宋军
奉有泉
陶原
陈钢
Original Assignee
阿里巴巴集团控股有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 阿里巴巴集团控股有限公司 filed Critical 阿里巴巴集团控股有限公司
Priority to PCT/CN2019/104590 priority Critical patent/WO2021042339A1/zh
Priority to CN201980095634.0A priority patent/CN113748386B/zh
Publication of WO2021042339A1 publication Critical patent/WO2021042339A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
    • G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
    • G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators

Definitions

  • This application relates to the field of Internet technology, and in particular to a method, equipment, system and storage medium for heat dissipation control and model training.
  • the Internet Data Center In addition to IT equipment such as computers and servers, the Internet Data Center (IDC) also includes refrigeration systems such as air conditioners and water pumps. The refrigeration system provides cooling air to IDC's computer room to ensure that the IT equipment in IDC can work normally.
  • IDC Internet Data Center
  • the refrigeration system provides cooling air to IDC's computer room to ensure that the IT equipment in IDC can work normally.
  • the refrigeration system needs to provide sufficient cooling air to maintain the IDC computer room at a constant indoor temperature to prevent the risk of overheating, but this consumes a huge amount of electricity and wastes resources.
  • Various aspects of the present application provide a heat dissipation control and model training method, equipment, system, and storage medium to reduce the energy consumption of the refrigeration system and save electrical energy resources.
  • An embodiment of the present application provides a heat dissipation control method, including: each time a heat dissipation control condition is triggered, obtaining actual power information of at least one device to be dissipated in a designated space; and inputting the actual power information of the at least one device to be dissipated into overheating Risk prediction model to obtain the probability of overheating risk in the designated space under at least one candidate refrigeration parameter; determine the target refrigeration parameter according to the probability of overheating risk in the designated space under at least one candidate refrigeration parameter; and according to the target The refrigeration parameter controls the refrigeration system to dissipate heat from at least one device to be dissipated in the designated space.
  • An embodiment of the present application also provides a model training method, including: combining a sample generation method based on real data and a sample generation method based on CFD simulation to generate multiple sets of labeled sample data; using the multiple sets of labeled sample data to perform a deep neural network Model training to obtain an overheating risk prediction model; wherein each group of labeled sample data includes at least one sample power information corresponding to the at least one device to be dissipated, sample cooling parameters corresponding to the refrigeration system, and labeled sample data in the group The result of marking whether there is a risk of overheating in the specified space described below.
  • An embodiment of the present application also provides a heat dissipation control device, including: a memory and a processor; the memory is used to store a computer program; the processor is coupled with the memory and is used to execute the computer program for use Yu: Whenever the heat dissipation control condition is triggered, obtain the actual power information of at least one device to be dissipated in the designated space; input the actual power information of the at least one device to be dissipated into the overheating risk prediction model to obtain the designated space The probability of an overheating risk occurring under at least one candidate refrigeration parameter; determining a target refrigeration parameter according to the probability of an overheating risk occurring under at least one candidate refrigeration parameter in the specified space; and controlling the refrigeration system to control the specified space according to the target refrigeration parameter At least one device to be dissipated in the heat dissipation device performs heat dissipation.
  • An embodiment of the application also provides a model training device, including: a memory and a processor; the memory is used to store a computer program; the processor is coupled with the memory and is used to execute the computer program for use Yu: Combine the sample generation method based on real data and the sample generation method based on CFD simulation to generate multiple sets of labeled sample data; use the multiple sets of labeled sample data for deep neural network model training to obtain an overheating risk prediction model; where each The set of labeled sample data includes at least one sample power information corresponding to the at least one device to be dissipated, a sample refrigeration parameter corresponding to the refrigeration system, and a labeling result of whether there is a risk of overheating in the designated space under the set of labeled sample data .
  • An embodiment of the present application also provides a computer room system, including: a computer room, and at least one device to be radiated, a refrigeration system, and a heat dissipation control device located in the computer room; the heat dissipation control device is used for whenever a heat dissipation control condition is triggered At the time, the actual power information of the at least one device to be dissipated is acquired, and the actual power information of the at least one device to be dissipated is input into the overheating risk prediction model to obtain the overheating risk of the computer room system under at least one candidate cooling parameter.
  • Probability determining the target refrigeration parameter according to the probability that the computer room system has an overheating risk under at least one candidate refrigeration parameter; according to the target refrigeration parameter, controlling the refrigeration system to dissipate heat from the at least one device to be dissipated; the The refrigeration system is used to dissipate heat of at least one device to be dissipated in the computer room under the control of the heat dissipation control device.
  • the embodiment of the application also provides a data center system, including: at least one computer room; each computer room includes: at least one device to be dissipated, a refrigeration system, and a heat dissipation control device; the heat dissipation control device is used whenever the heat dissipation control condition is When triggered, obtain the actual power information of at least one device to be dissipated in the computer room to which it belongs, and input the actual power information of the at least one device to be dissipated into the overheating risk prediction model to obtain that the computer room is overheated under at least one candidate cooling parameter
  • the probability of the risk; the target refrigeration parameter is determined according to the probability of the overheating risk of the computer room under at least one candidate refrigeration parameter; according to the target refrigeration parameter, the refrigeration system in the computer room to which it belongs is controlled to perform the operation on the at least one device to be radiated Heat dissipation;
  • the refrigeration system is used to dissipate at least one device to be dissipated in
  • the embodiment of the present application also provides another data center system, including: at least one computer room, a cooling system, and heat dissipation control equipment; wherein, each computer room includes at least one device to be radiated, and the cooling system includes Refrigeration equipment; the heat dissipation control device is used to obtain the actual power information of at least one device to be radiated in the computer room every time a heat dissipation control condition is triggered for each computer room, and to convert at least one device to be radiated in the computer room Input the actual power information into the overheating risk prediction model to obtain the probability of the overheating risk of the computer room under at least one candidate refrigeration parameter; determine the target refrigeration parameter according to the probability of the overheating risk of the computer room under the at least one candidate cooling parameter; According to the target refrigeration parameter, the refrigeration device in the computer room is controlled to dissipate heat from at least one device to be dissipated in the computer room.
  • the embodiment of the present application also provides a computer-readable storage medium storing a computer program.
  • the processor When the computer program is executed by a processor, the processor is caused to implement the steps in the heat dissipation control method provided by the embodiment of the present application.
  • the embodiment of the present application also provides a computer-readable storage medium storing a computer program.
  • the processor When the computer program is executed by a processor, the processor is caused to implement the steps in the model training method provided in the embodiment of the present application.
  • an overheating risk prediction model is obtained by pre-training, and the overheating risk relationship between the equipment power information and the cooling parameters is reflected by the model. Furthermore, on the basis of the overheating risk prediction model, the overheating risk prediction model can be based on the specified space According to the power change of the equipment to be dissipated, the refrigeration parameters of the refrigeration system are dynamically adjusted to achieve the purpose of dynamic heat dissipation control, which is beneficial to reduce the energy consumption of the refrigeration system and save electric energy resources.
  • FIG. 1 is a schematic structural diagram of a computer room system provided by an exemplary embodiment of this application;
  • Figure 2a is a schematic diagram of a model training process provided by an exemplary embodiment of the application.
  • FIG. 2b is a schematic structural diagram of an overheating risk prediction model provided by an exemplary embodiment of this application.
  • 2c is a schematic diagram of a state of a model prediction result provided by an exemplary embodiment of this application.
  • FIG. 3 is a schematic structural diagram of a data center system provided by an exemplary embodiment of this application.
  • FIG. 4 is a schematic structural diagram of another data center system provided by an exemplary embodiment of this application.
  • FIG. 5 is a schematic flowchart of a heat dissipation control method provided by an exemplary embodiment of this application.
  • FIG. 6 is a schematic flowchart of a model training method provided by an exemplary embodiment of this application.
  • FIG. 7 is a schematic structural diagram of a heat dissipation control device provided by an exemplary embodiment of this application.
  • FIG. 8 is a schematic structural diagram of a model training device provided by an exemplary embodiment of this application.
  • the overheating risk prediction model is pre-trained, and the overheating risk between the equipment power and the cooling parameters is reflected by the model. Based on the overheating risk prediction model, dynamically adjust the refrigeration parameters of the refrigeration system according to the power changes of the equipment to be dissipated in the designated space to achieve the purpose of dynamic heat dissipation control, thereby reducing the energy consumption of the refrigeration system and saving power resources .
  • FIG. 1 is a schematic structural diagram of a computer room system provided by an exemplary embodiment of this application.
  • the computer room system 100 of this embodiment includes: a computer room.
  • the computer room refers to a physical place where machinery and equipment are stored, for example, a room or a factory building.
  • the computer room system 100 further includes: at least one device to be dissipated 101, a refrigeration system 102 and a heat dissipation control device 103 located in the computer room.
  • This embodiment does not limit the number of devices 101 to be dissipated in the computer room, and it may be one or more.
  • a computer room will contain multiple devices 101 to be cooled.
  • the device to be dissipated 101 refers to an electronic device that can generate heat and has certain requirements for the ambient temperature in the working environment.
  • the device form of the device 101 to be dissipated is not limited.
  • the device 101 to be dissipated may be an IT device, but it is not limited thereto.
  • the at least one device to be dissipated 101 may include, but is not limited to, at least one of the following equipment forms: cabinet equipment, server equipment, computer equipment, printers, hubs, power supply equipment, storage equipment, network switching equipment, and so on.
  • the server device may include, but is not limited to, a conventional server, a server array, or a cloud server.
  • the power supply equipment can be storage battery equipment, dry battery equipment, or uninterruptible power supply (UPS).
  • Storage devices may include, but are not limited to: disks, disk arrays, hard disks, network storage devices (NAS), and so on.
  • At least one application or service such as cloud computing service, game service, instant messaging service, mail service, or online transaction service, runs on at least one device 101 to be dissipated in the computer room system 100.
  • the equipment to be dissipated 101 has certain requirements for the temperature in the computer room. If the temperature in the computer room is too high, the equipment to be dissipated 101 may malfunction, malfunction, or even be burned.
  • a refrigeration system 102 is also provided in the machine room.
  • the refrigeration system 102 is mainly responsible for taking away heat in the machine room and dissipating heat for the equipment 101 to be dissipated in the machine room.
  • the type and working principle of the refrigeration system 102 are not limited. For example, it may be an air conditioning system, or a water cooling system, or a combination of an air conditioning system and a water cooling system.
  • the power consumption of the equipment 101 to be dissipated in the computer room during operation is a major factor affecting the temperature of the computer room.
  • the cooling parameters are calculated according to the maximum power consumption of the device 101 to be dissipated, and according to The calculated cooling parameters control the cooling system 102 to dissipate heat in the computer room, so that regardless of the power consumption of the equipment 101 in the computer room to be cooled, the temperature of the computer room can be kept low, and the risk of overheating of the equipment 101 to be cooled is ensured.
  • the work load of the device 101 to be dissipated sometimes changes greatly, and the change in the work load will cause the power consumption of the device 101 to be dissipated to change, which means that the device 101 to be dissipated will not always be in the maximum power consumption state, so In order to control the heat dissipation of the refrigeration system 102 according to the maximum power consumption of the device 101 to be dissipated, a large amount of power resources will be wasted.
  • an overheating risk prediction model is obtained in advance, and the model can reflect the overheating risk relationship existing between the equipment power information and the cooling parameters.
  • the heat dissipation control device 103 can control the cooling system 102 to dynamically dissipate heat in the computer room based on the overheating risk prediction model and the power change of at least one device to be dissipated in the computer room.
  • the cooling system 102 can perform dynamic cooling according to actual needs.
  • the refrigeration work can reduce the energy consumption of the refrigeration system 102 and save electric energy resources.
  • the power information of the device to be dissipated 101 reflects the power consumption of the device to be dissipated 101, and may also reflect the workload of the device to be dissipated 101.
  • the model training process please refer to the subsequent embodiments, which will not be repeated here.
  • heat dissipation control conditions can be set. Whenever the heat dissipation control conditions are triggered, a heat dissipation control is performed based on the overheating risk prediction model and the actual power information of at least one device 101 to be dissipated in the computer room. It can be seen that on the basis of the overheating risk prediction model, the heat dissipation control device 103 controls the process of dynamic heat dissipation of the computer room by the refrigeration system 102 in combination with the power change of the device 101 to be dissipated in the computer room, which may include multiple heat dissipation controls. The following describes the heat dissipation control process based on the overheating risk prediction model in this embodiment:
  • the heat dissipation control device 103 obtains the actual power information of at least one device to be dissipated 101 in the computer room; inputs the actual power information of the at least one device to be dissipated 101 into the overheating risk prediction model to obtain the computer room system 100
  • the probability of an overheating risk occurring under at least one candidate refrigeration parameter; then according to the probability of an overheating risk of the computer room system 100 under at least one candidate refrigeration parameter, the target refrigeration parameter is determined; according to the target refrigeration parameter, the refrigeration system 102 is controlled to control at least one of the overheating risks in the computer room.
  • the device 101 to be dissipated dissipates heat.
  • heat dissipation can be performed on at least one device to be dissipated in the computer room.
  • the probability of the computer room system 100 having an overheating risk under a certain candidate refrigeration parameter mainly refers to the probability that if the refrigeration system 102 adopts the candidate refrigeration parameter, the equipment room system 100 will have an overheating risk for the equipment to be dissipated.
  • the number of devices to be dissipated with the risk of overheating may be one or more, which is not limited.
  • this embodiment does not limit the heat dissipation control conditions, and can be flexibly set according to the heat dissipation control requirements.
  • the following is an example of heat dissipation control conditions:
  • Example 1 In this example, considering that after the cooling system 102 adjusts the cooling parameters each time, it generally takes a certain time to achieve the expected heat dissipation effect, the heat dissipation control period can be preset, and the heat dissipation control period is used as the heat dissipation control condition . Based on this, the heat dissipation control device 103 can periodically perform heat dissipation control on the computer room according to the heat dissipation control cycle, which can not only achieve the heat dissipation effect, but also reduce the workload of the heat dissipation control device 103.
  • the actual power information of at least one device to be dissipated can be obtained; combined with the overheating risk prediction model, the refrigeration parameters of the refrigeration system 102 can be adjusted periodically to control the refrigeration system 102 to at least A device 101 to be dissipated dynamically dissipates heat.
  • This embodiment does not limit the time length of the heat dissipation control period, and can be set adaptively according to application requirements. For example, the time length of the heat dissipation control period can be 1 minute, 10 minutes, 15 minutes, or the like.
  • Example 2 the heat dissipation control device 103 can monitor the total power change range of at least one device to be dissipated in the computer room in real time, and use the total power change range of at least one device to be dissipated as a heat dissipation control condition.
  • the heat dissipation control device 103 obtains the actual power information of the at least one device to be dissipated 101; combined with the overheating risk prediction model, the refrigeration system is continuously adjusted
  • the refrigeration parameters of 102 are used to dynamically dissipate at least one device 101 to be dissipated in the computer room by controlling the refrigeration system 102.
  • the heat dissipation control device 103 can monitor the power change range of each device 101 to be dissipated in the computer room in real time, and use the power change range of each device to be dissipated as a heat dissipation control condition. Based on this, whenever it is detected that the power change range of the device to be dissipated is greater than the second range threshold, the heat dissipation control device 103 obtains the actual power information of at least one device to be dissipated 101; combined with the overheating risk prediction model, continuously adjusts the refrigeration system The refrigeration parameters of 102 are used to dynamically dissipate at least one device 101 to be dissipated in the computer room by controlling the refrigeration system 102.
  • this embodiment does not limit the values of the first amplitude threshold and the second amplitude threshold, and can be flexibly set according to application requirements.
  • the second amplitude threshold may be the same or different.
  • the corresponding second amplitude threshold can be set for each device 101 to be dissipated.
  • the actual power information of the device to be dissipated 101 is mainly used to reflect the power consumption of the device to be dissipated, and is the data basis for this heat dissipation control.
  • the implementation form of the actual power information of the device 101 to be dissipated is not limited, and it may be any data form that can reflect the power consumption of the device 101 to be dissipated.
  • the power value of the at least one device to be dissipated 101 at the moment when the heat dissipation control condition is triggered may be separately collected as the actual power information of the at least one device to be dissipated 101.
  • the average power of at least one device 101 to be dissipated during the current heat dissipation control and the last heat dissipation control is obtained as the actual power information of the at least one device 101 to be dissipated.
  • the power value of the device 101 to be dissipated may be the overall power value of the device 101 to be dissipated.
  • the overall power value of the device 101 to be dissipated may be defined as the sum of the power of the main internal components of the device 101 to be dissipated, or it may be defined as the sum of the power of all the internal components of the device 101 to be dissipated.
  • the power value of the device to be dissipated 101 may be the power of a certain internal component of the device to be dissipated 101, for example, it may be the power of a CPU, or the power of a memory.
  • the refrigeration system 102 and the heat dissipation control device 103 are in communication connection.
  • the refrigeration system 102 and the heat dissipation control device 103 may be wirelessly or wiredly connected.
  • the heat dissipation control device 103 may communicate with the refrigeration system 102 via a mobile network.
  • the network standard of the mobile network can be 2G (GSM), 2.5G (GPRS), 3G (WCDMA, TD-SCDMA, CDMA2000, UTMS), 4G (LTE), 4G+ (LTE+), 5G, WiMax or coming soon in the future Any of the new network standards, etc.
  • the heat dissipation control device 103 may also communicate with the refrigeration system 102 through Bluetooth, WiFi, infrared, zigbee, or NFC.
  • the refrigeration system 102 can be controlled to dissipate at least one device 101 to be dissipated in the computer room.
  • the heat dissipation control device 103 may directly send the target refrigeration parameter to the refrigeration system 102 based on the communication connection between it and the refrigeration system 102, so that the refrigeration system 102 can directly send the target refrigeration parameter to the refrigeration system 102 according to the target refrigeration parameter.
  • the refrigeration parameter is that at least one device 101 to be dissipated in the computer room performs heat dissipation.
  • the target refrigeration parameter sent by the heat dissipation control device 103 can be received, and the target refrigeration parameter is compared with the currently used refrigeration parameter; if the two are different, the currently used refrigeration parameter is replaced with the target refrigeration parameter, and The refrigeration work is continued according to the target refrigeration parameter; if the two are the same, the refrigeration work is continued according to the currently used refrigeration parameter, so as to achieve the purpose of dissipating at least one device 101 to be dissipated in the computer room according to the target refrigeration parameter.
  • the heat dissipation control device 103 may record the refrigeration parameters currently used by the refrigeration system 102, and after determining the target refrigeration parameters, compare the target refrigeration parameters with the refrigeration parameters currently used by the refrigeration system 102, and then If the target refrigeration parameter is different from the refrigeration parameter currently used by the refrigeration system 102, based on the communication connection between it and the refrigeration system 102, the target refrigeration parameter is sent to the refrigeration system 102 for the refrigeration system 102 to continue according to the target refrigeration parameter Perform refrigeration work.
  • the refrigeration system 102 can receive the target refrigeration parameter sent by the heat dissipation control device 103, adjust the currently used refrigeration parameter to the target refrigeration parameter, and continue the refrigeration work according to the target refrigeration parameter; In the case of the target refrigeration parameter, the refrigeration work can be continued according to the currently used refrigeration parameter, so as to achieve the purpose of dissipating at least one device 101 to be dissipated in the computer room according to the target refrigeration parameter.
  • the target refrigeration parameters are also different. Regardless of the refrigeration system, the target refrigeration parameters are related parameters that can affect the refrigeration effect. For example, for an air conditioning system, cooling air can be input to the computer room system 100, and the cooling air flows in a certain direction (for example, top-down or bottom-up), which will take away the heat inside the computer room system 100 to achieve the purpose of heat dissipation.
  • the air conditioning system of this embodiment includes but is not limited to the following working parameters: working temperature, working wind speed, and working mode, etc. These working parameters will affect the heat dissipation performance of the air conditioning system.
  • the working parameters of the air-conditioning system can be used as the target refrigeration parameters in this embodiment.
  • the target refrigeration parameters can include but are not limited to at least one of the working temperature, working wind speed, and working mode of the air-conditioning system.
  • liquid coolant can be provided to the computer room system 100 through a liquid carrier such as a pipe.
  • the liquid coolant can be cold water or liquid metal sodium, etc.
  • the liquid coolant flows in the computer room system 100 or surrounds the computer room system 100
  • the heat-dissipating equipment flows, thereby taking away the heat inside the computer room system 100 to achieve the purpose of heat dissipation.
  • the water cooling system of this embodiment includes but is not limited to the following working parameters: outlet water temperature, return water temperature, water flow rate, and water flow rate, etc.
  • the working parameters of the water-cooling system can be used as the target refrigeration parameters of this embodiment.
  • the target refrigeration parameters can include, but are not limited to: at least one of the outlet water temperature, the return water temperature, the water flow rate, and the water flow rate of the water-cooling system. Kind.
  • the working mode of the overheating risk prediction model is not limited.
  • the working method of the overheating risk prediction model corresponds to its training method. Different training methods can be used to train the overheating risk prediction model with different working methods. The following is an example of possible working methods of the overheating risk prediction model:
  • the actual power information of at least one device to be dissipated can be input into the overheating risk prediction model, which can independently determine at least one candidate refrigeration parameter, and can output the computer room system 100 in each The probability of overheating risk under the candidate refrigeration parameters.
  • At least one candidate refrigeration parameter is determined outside the model in advance, and the actual power information of the at least one device to be dissipated and the at least one candidate refrigeration parameter are used as input parameters into the overheating risk prediction model.
  • the probability of an overheating risk of the computer room system 100 under each candidate refrigeration parameter is output once at a time.
  • At least one candidate refrigeration parameter is determined outside the model in advance; for each candidate refrigeration parameter, the actual power information of at least one device to be dissipated and the candidate refrigeration parameter are input into the overheating risk prediction model to obtain The probability that the computer room system 100 has an overheating risk under the candidate cooling parameter.
  • the method for determining at least one candidate refrigeration parameter includes: determining at least one candidate refrigeration parameter based on manual experience; or, adjusting the refrigeration parameter currently used by the refrigeration system 102 to different ranges to obtain at least one candidate refrigeration parameter; or, according to The range of refrigeration parameters used by the overheating risk prediction model in the training phase determines at least one candidate refrigeration parameter.
  • At least one candidate refrigeration parameter is determined, that is, the candidate refrigeration parameter is within the range of the refrigeration parameter used by the overheating risk prediction model in the training phase.
  • the cooling parameters used by the overheating risk prediction model in the training phase are 19°C, 20°C, 22°C, 26°C, and 28°C
  • at least one candidate cooling parameter can be determined between 19°C and 28°C.
  • the range of refrigeration parameters used in the training phase of the overheating risk prediction model can also be used as the basic parameter range, a candidate parameter range is determined according to the basic parameter range, and at least one candidate refrigeration parameter is determined within the candidate parameter range. parameter.
  • the heat dissipation control device 103 may determine the target cooling parameter according to the probability of the computer room system 100 having an overheating risk under the at least one candidate cooling parameter.
  • determining the target refrigeration parameter can adopt but not limited to the following optional implementation manners:
  • the probability of the computer room system 100 having an overheating risk under at least one candidate refrigeration parameter may be compared, and a candidate refrigeration parameter with a lower probability may be selected as the target refrigeration parameter.
  • the candidate refrigeration parameter corresponding to the smallest probability can be selected as the target refrigeration parameter.
  • the overheating risk probability threshold corresponding to the computer room system may be determined in advance in accordance with the requirements of the application or service carried by the computer room system on the occurrence of an overheating risk. Then, based on the overheating risk probability threshold, from the probability that the computer room system has an overheating risk under at least one candidate refrigeration parameter, select a probability less than the overheating risk probability threshold as the target probability, and combine the at least one candidate refrigeration parameter with the The refrigeration parameter corresponding to the target probability is used as the target refrigeration parameter.
  • a probability less than the overheating risk probability threshold can be randomly selected as the target probability; or, the computer room system can be selected from the probability of overheating risk.
  • the maximum probability less than the overheating risk probability threshold is selected as the target probability; or, among the probability of overheating risk of the computer room system under at least one candidate refrigeration parameter, the probability of overheating risk can be selected to be less than the overheating risk probability.
  • the allowable thermal failure rate of the application or service carried by the computer room system can be obtained, and the thermal failure rate can be converted into the overheating risk probability threshold corresponding to the computer room system.
  • a statistical method may be used to convert the thermal failure rate into the overheating risk probability threshold corresponding to the computer room.
  • the thermal failure rate refers to the maximum number of times that the computer room system 100 can have an overheating risk within a certain period of time.
  • the overheating risk prediction model can be pre-trained to provide a basis for the embodiments that need to use the overheating risk prediction model.
  • multiple sets of labeled sample data can be obtained, and based on the multiple sets of labeled sample data, a supervised model training method is used to train an overheating risk prediction model.
  • a deep neural network algorithm is used for model training.
  • the sample generation method based on real data is combined with the sample generation method based on Computational Fluid Dynamics (Computational Fluid Dynamics, CFD) simulation, and the CFD simulation calculation method is used to provide the required model training.
  • the sample data can make up for the shortcomings of the sample generation method based on real data, which is conducive to obtaining sufficient sample data and effectively improving the robustness of the thermal risk prediction model.
  • a model training process includes: combining the sample generation method based on real data and the sample generation method based on CFD simulation to generate multiple sets of labeled sample data; then, use multiple sets of labeled sample data to perform Deep neural network model training, to get the overheating risk prediction model.
  • each set of labeled sample data includes at least one sample power information corresponding to at least one device to be dissipated, sample refrigeration parameters corresponding to the refrigeration system, and a labeling result of whether the computer room has an overheating risk under the set of labeled sample data.
  • multiple sets of sample data may be cleaned to improve the reliability of the sample data.
  • different data cleaning methods may be used for data cleaning.
  • combining the sample generation method based on real data and the sample generation method based on CFD simulation to generate multiple sets of labeled sample data includes: generating at least A set of marked historical sample data; and use the CFD model to perform simulation calculations between power information and cooling parameters to generate at least one set of marked simulated sample data.
  • the process of generating at least one set of labeled historical sample data includes: obtaining at least one set of unlabeled historical sample data, and each set of unlabeled historical sample data includes the history of at least one device to be dissipated in the same historical moment or historical period. Power information and historical refrigeration parameters of the refrigeration system; for each group of unlabeled historical sample data, determine whether the computer room occurs according to the temperature of at least one internal component of the device to be dissipated in the corresponding historical time or historical period and the corresponding overheating temperature threshold of the internal component The risk of overheating is marked, and at least one set of marked historical sample data is obtained.
  • the process of generating at least one set of labeled simulated sample data includes: designing at least one set of unlabeled simulated sample data, and each set of unlabeled simulated sample data includes at least one set of simulated power information corresponding to at least one device to be dissipated; Simulated refrigeration parameters corresponding to the refrigeration system; for each group of unlabeled simulation sample data, use the CFD model to simulate the group of unlabeled simulation sample data to obtain the temperature of at least one device to be dissipated, and use at least one to be dissipated
  • the temperature of the internal components of the equipment and the corresponding overheating temperature threshold of the internal components are used to mark whether the computer room has an overheating risk, and at least one set of marked simulation sample data is obtained.
  • At least one set of labeled historical samples can also be used Data, and perform parameter correction on the CFD model. After that, use the corrected CFD model to generate at least one set of labeled simulation sample data, which is conducive to improving the reliability and authenticity of the labeled simulation sample data, and is conducive to improving the accuracy of the overheating risk prediction model trained accordingly .
  • each set of labeled sample data includes: at least one sample power information, sample refrigeration parameters, and sample atmospheric temperature.
  • each set of labeled sample data includes: at least one sample power information, sample refrigeration parameters, and sample atmospheric temperature.
  • it also includes the operation of obtaining the external atmospheric temperature at the corresponding time as the sample atmospheric text and adding the sample atmospheric temperature to the sample data.
  • the outside air temperature at the time corresponding to the first set of labeled sample data is also obtained as the sample air temperature, and the first set of labeled sample data corresponds to The sample atmospheric temperature is added to the first set of labeled sample data; wherein, the first set of labeled sample data is any set of labeled sample data in the multiple sets of labeled sample data.
  • deep neural network model training can be performed according to multiple sets of labeled sample data containing sample air temperature to obtain an overheating risk prediction model.
  • the overheating risk prediction model trained based on the labeled sample data containing the sample air temperature parameter can predict the occurrence of the computer room under the candidate cooling parameter for a given set of power information, external air temperature, and a candidate cooling parameter. The probability of overheating risk.
  • the outside air temperature published by some websites or APPs can be obtained through the Internet.
  • a temperature collection device such as a temperature sensor, can also be set outside the computer room, and the outside atmospheric temperature can be collected through the temperature collection device.
  • the method of obtaining the outside air temperature is not limited in the embodiment of the present application.
  • the heat dissipation control device 103 not only obtains the actual power information of at least one device to be dissipated in the computer room, but also needs to obtain the external atmospheric temperature corresponding to when the heat dissipation control condition is triggered;
  • the actual power information of the at least one device to be dissipated and the corresponding external atmospheric temperature when the heat dissipation control condition is triggered are used as input parameters
  • Input the overheating risk prediction model to obtain the probability of overheating risk of the computer room system under at least one candidate refrigeration parameter.
  • the target refrigeration parameter is determined according to the probability of the computer room system having an overheating risk under at least one candidate refrigeration parameter; and the refrigeration system is controlled to dissipate heat from at least one device to be dissipated in the computer room according to the target refrigeration parameter.
  • the structure of an overheating risk prediction model is as shown in FIG. 2b.
  • the structure of the thermal risk prediction model presented in Fig. 2b is only an exemplary description and does not limit it.
  • the input data supported by the model includes: actual power information of device 1-device n, external atmospheric temperature, and a candidate refrigeration parameter (such as operating temperature).
  • the output is the risk of overheating in the computer room under the candidate refrigeration parameter.
  • n is a positive integer.
  • the thermal risk prediction model shown in Fig. 2b can be used to obtain the probability of an overheating risk in the computer room under the candidate refrigeration parameters.
  • the refrigeration system as an air-conditioning system as an example, as shown in Figure 2c, assuming that the candidate refrigeration parameters include four refrigeration temperatures: 20°C, 21°C, 22°C, and 23°C, after passing through the thermal risk prediction model shown in Figure 2b, The four probability values obtained are 0.76, 0.83, 0.89 and 0.92 respectively. If it is assumed that the overheating risk probability threshold corresponding to the computer room is 0.85, the temperature 21°C corresponding to the probability value of 0.83 can be selected as the target cooling temperature; then the air conditioning system can be controlled to adjust the cooling temperature to 21°C to dissipate heat in the computer room.
  • the overheating risk prediction model can also be updated.
  • the update of the thermal risk prediction model includes but is not limited to the following situations: whenever the model update cycle arrives, re-train the overheating risk prediction model; whenever the number of heat-dissipating devices in the computer room changes, re-train the model. Model training is performed on the overheating risk prediction model; whenever the topology of the equipment to be dissipated in the computer room changes, the overheating risk prediction model is re-trained.
  • the updated overheating risk prediction model can be used in the subsequent heat dissipation control process, which is beneficial to improve the accuracy and precision of the heat dissipation control. It is worth noting that the model training process provided in the foregoing embodiment is not only applicable to the foregoing embodiment of the computer room system described in this application, but also applicable to the following embodiment of the data center system described in the present application, and will not be repeated in the following embodiments.
  • the heat dissipation control principle provided in the embodiments of the present application is not only applicable to an independent computer room system, but also applicable to a data center system including one or more computer rooms.
  • the following takes the data center systems of the two structures shown in FIG. 3 and FIG. 4 as examples to illustrate the heat dissipation control principle of the embodiment of the present application.
  • FIG. 3 is a schematic structural diagram of a data center system provided by an exemplary embodiment of this application.
  • the data center system 300 includes: at least one computer room 301; each computer room 301 includes: at least one device to be dissipated, a refrigeration system, and a heat dissipation control device.
  • each computer room 301 is relatively independent, has its own refrigeration system and heat dissipation control equipment, and can independently perform heat dissipation control.
  • the heat dissipation control equipment contained in it can control the cooling system to dynamically heat the computer room 301 based on the overheating risk prediction model and the power change of at least one device to be dissipated in the computer room 301.
  • the system can perform refrigeration work according to actual needs, which can reduce the energy consumption of the refrigeration system and save electrical energy resources.
  • heat dissipation control conditions can be set. Whenever the heat dissipation control conditions are triggered, the heat dissipation control device obtains the actual power information of at least one device to be dissipated in the computer room 301 to which it belongs, and inputs the actual power information of at least one device to be dissipated into the risk of overheating Predictive model to obtain the probability of overheating risk of computer room 301 under at least one candidate cooling parameter; determine the target cooling parameter according to the probability of overheating risk of computer room 301 under at least one candidate cooling parameter; control the computer room to which it belongs according to the target cooling parameter
  • the refrigeration system in 301 dissipates heat from at least one device to be dissipated in the machine room 301 to which it belongs. For the refrigeration system, under the control of the heat dissipation control device included in the computer room 301 to which it belongs, heat dissipation can be performed on at least one device to be dissipated in the computer room 301 to which
  • the computer room 301 in this embodiment is the same as or similar to the computer room system 100 in the previous embodiment.
  • the detailed description of the computer room 301 and the heat dissipation control equipment in any computer room 301 are combined with the computer room 301 based on the overheating risk prediction model.
  • FIG. 4 is a schematic structural diagram of another data center system provided by an exemplary embodiment of this application.
  • the data center system 400 includes: at least one computer room 401, a refrigeration system, and a heat dissipation control device 403.
  • each computer room 401 includes at least one device to be radiated;
  • the refrigeration system includes a refrigeration device 402 deployed in each computer room 401.
  • the data center system 400 Different from the data center system 300 shown in FIG. 3, in the data center system 400, different computer rooms 401 share the refrigeration system and the heat dissipation control equipment 403.
  • the heat dissipation control device 403 needs to perform heat dissipation control for each computer room 401 in the data center system 400; similarly, the cooling system needs to perform heat dissipation for each computer room 401 in the data center system 400.
  • the refrigeration system in order to achieve the purpose of dissipating heat for each computer room 401 in the data center system 400, the refrigeration system includes a cooling device 402 deployed in each computer room 401, so that the heat dissipation control device 403 can pass through each computer room 401
  • the refrigeration equipment 402 performs heat dissipation control for each computer room 401, and the control logic is relatively simple, convenient, and easy to implement.
  • the heat dissipation control device 403 performs the same or similar process of heat dissipation control for each computer room 401.
  • the process of heat dissipation control for each computer room 401 by the heat dissipation control device 403 is described below:
  • the heat dissipation control device 403 is used to: each time a heat dissipation control condition is triggered, obtain the actual power information of at least one device to be dissipated in the computer room 401, and to obtain the actual power of at least one device to be dissipated in the computer room 401
  • the information is input into the overheating risk prediction model to obtain the probability of the computer room 401 having an overheating risk under at least one candidate cooling parameter; according to the probability of the computer room 401 having an overheating risk under at least one candidate cooling parameter, the target cooling parameter is determined; according to the target
  • the cooling parameter controls the cooling device 402 in the computer room 401 to dissipate heat from at least one device to be radiated in the computer room 401.
  • different computer rooms 401 can use the same heat dissipation control conditions, or different heat dissipation control conditions, which can be flexibly set according to the specific conditions of each computer room 401.
  • the heat dissipation control device 403 can control the heat dissipation of each computer room 401 in the data center system 400, the process of the heat dissipation control device 403 performing heat dissipation control on one of the computer rooms 401 is the same as the heat dissipation control device 103 in the previous embodiment.
  • the process of heat dissipation control performed by the computer room system 100 is the same or similar.
  • the process of heat dissipation control performed by the heat dissipation control device 403 on each computer room 401 please refer to the foregoing embodiment, which will not be repeated here.
  • the heat dissipation control device can be any computer device that has certain computing and communication capabilities and can perform data processing.
  • it can be a server device such as a conventional server, a cloud server, a cloud host, a virtual center, or a server array, or a smart phone, Terminal devices such as tablets, personal computers, or all-in-ones.
  • a computer room and a data center are taken as examples to illustrate the heat dissipation control principle provided in the embodiments of the present application.
  • the computer room and the data center are only two exemplary application scenarios given in the embodiments of the present application, and cannot constitute a limitation on the protection scope of the present application.
  • the heat dissipation control principle provided by the embodiments of the present application can be applied to any physical space containing the device to be dissipated. In other words, any physical space containing the device to be dissipated can be controlled by the heat dissipation control principle provided by the embodiments of the present application.
  • office buildings various electrical equipment are installed or configured, such as office computers, servers, monitoring equipment, printers, fax machines, copiers, lights, and so on.
  • office buildings will also install or configure infrastructure such as air conditioning systems, exhaust systems, or heating to adjust the ambient temperature in the office buildings.
  • the heat dissipation control method can be used to monitor the actual power information of the equipment to be radiated in the entire office building; input the actual power information of the equipment to be radiated in the entire office building into the overheating risk prediction Model to obtain the probability of the overheating risk of the entire office building under at least one candidate cooling parameter; then determine the target cooling parameter according to the probability of the overheating risk of the entire office building under at least one candidate cooling parameter; control the cooling according to the target cooling parameter
  • the system dissipates heat from the equipment to be dissipated in the entire office building.
  • the equipment to be dissipated in the office building includes, but is not limited to: office computers, servers, monitoring equipment, printers, fax machines, copiers, etc., as well as other electrical equipment such as lights, water dispensers, microwave ovens, and vacuum cleaners.
  • the heat dissipation control method provided in the embodiments of this application can be used to monitor the actual power information of the equipment to be dissipated in the office area;
  • the actual power information is input into the overheating risk prediction model to obtain the probability of overheating risk in the office area under at least one candidate cooling parameter; and then according to the probability of overheating risk in the office area under at least one candidate cooling parameter, the target cooling parameter is determined; according to the target Refrigeration parameters control the refrigeration system to dissipate heat from the equipment to be radiated in the office area.
  • the equipment to be dissipated in the office building includes but is not limited to: office computers, servers, printers, fax machines, copiers, etc.
  • smart home devices For another example, with the development of smart home technology, more and more smart home devices appear in the home environment, and there may be a large number of smart home devices in a designated area in the home environment.
  • electrical equipment such as televisions, smart speakers, large game consoles, smart air conditioners, purifiers, home escort robots, and personal computers.
  • electrical equipment such as smart microwave ovens, ovens, refrigerators, rice cookers, and juicers. The equipment in these areas also generates a lot of heat when working.
  • the heat dissipation control method provided in the embodiments of this application can also be used to monitor the home environment.
  • the actual power information of the equipment to be dissipated in the designated area in the designated area input the actual power information of the equipment to be dissipated in the designated area in the home environment into the overheating risk prediction model to obtain the overheating risk of the designated area in the home environment under at least one candidate cooling parameter Probability; and then determine the target refrigeration parameter according to the probability of the overheating risk of the designated area in the home environment under at least one candidate refrigeration parameter; according to the target refrigeration parameter, control the refrigeration system to dissipate heat from the equipment to be radiated in the designated area in the home environment.
  • the equipment to be dissipated in the designated areas will be different.
  • the devices to be dissipated include, but are not limited to: TVs, smart speakers, large game consoles, smart air conditioners, purifiers, home escort robots, personal computers, etc.
  • FIG. 5 is a schematic flowchart of a heat dissipation control method provided by an exemplary embodiment of this application. This method is described from the perspective of heat dissipation control equipment. As shown in Figure 5, the method includes:
  • the designated space refers to any physical space capable of accommodating and accommodating equipment to be dissipated, such as but not limited to: computer rooms, data centers, office buildings, office areas in a company environment, or designated areas in a home environment.
  • a designated area in the home environment stores heat-generating devices to be dissipated, such as televisions, smart speakers, large game consoles, smart air conditioners, purifiers, home escort robots, and personal computers.
  • the designated space contains at least one device to be dissipated.
  • the equipment to be dissipated can generate heat and increase the temperature in the designated space.
  • the equipment to be radiated has certain requirements for the temperature in the designated space. If the temperature in the designated space is too high, the equipment to be radiated may malfunction, malfunction, or even be burned.
  • a refrigeration system can be provided.
  • the refrigeration system can be deployed in a designated space, but it is not limited to this.
  • the refrigeration system is mainly responsible for taking away the heat in the designated space and dissipating heat for the equipment to be radiated in the designated space.
  • the type and working principle of the refrigeration system are not limited. For example, it may be an air conditioning system, or a water cooling system, or a combination of an air conditioning system and a water cooling system.
  • an overheating risk prediction model is obtained in advance, and the model can reflect the overheating risk relationship existing between the equipment power information and the cooling parameters.
  • the heat dissipation control device can control the refrigeration system to dynamically dissipate heat in the specified space based on the overheating risk prediction model and the power change of at least one device to be dissipated in the designated space.
  • the refrigeration system can perform cooling work according to actual needs. , which can reduce the energy consumption of the refrigeration system and save power resources.
  • the power information of the device to be dissipated reflects the power consumption of the device to be dissipated, and may also reflect the workload of the device to be dissipated.
  • heat dissipation control conditions can be set. Whenever the heat dissipation control conditions are triggered, the heat dissipation control device performs a heat dissipation based on the overheating risk prediction model and the actual power information of at least one device to be dissipated in the designated space. control.
  • the probability of overheating risk in a specified space under a certain candidate refrigeration parameter mainly refers to the probability that if the refrigeration system adopts the candidate refrigeration parameter, there is a risk of overheating of the equipment to be dissipated in the specified space.
  • the number of devices to be dissipated with the risk of overheating may be one or more, which is not limited.
  • this embodiment does not limit the heat dissipation control conditions, and can be flexibly set according to the heat dissipation control requirements.
  • the following is an example of heat dissipation control conditions:
  • the heat dissipation control period may be preset, and the heat dissipation control period may be used as the heat dissipation control condition.
  • an implementation manner of step 501 is: each time the heat dissipation control period arrives, the actual power information of at least one device to be dissipated is acquired.
  • This embodiment does not limit the time length of the heat dissipation control period, and can be set adaptively according to application requirements.
  • the time length of the heat dissipation control period can be 1 minute, 10 minutes, 15 minutes, or the like.
  • the heat dissipation control device can monitor the total power change range of at least one device to be dissipated in the designated space in real time, and use the total power change range of the at least one device to be dissipated as the heat dissipation control condition. Based on this, an implementation manner of step 501 is to obtain actual power information of the at least one device to be dissipated whenever the total power variation amplitude of the at least one device to be dissipated is monitored is greater than the first amplitude threshold.
  • the heat dissipation control device can monitor the power change range of each device to be dissipated in the designated space in real time, and use the power change range of each device to be dissipated as the heat dissipation control condition. Based on this, an implementation manner of step 501 is to acquire actual power information of at least one device to be dissipated 101 whenever it is monitored that the power change range of the device to be dissipated is greater than the second range threshold.
  • this embodiment does not limit the values of the first amplitude threshold and the second amplitude threshold, and can be flexibly set according to application requirements.
  • the second amplitude threshold may be the same or different.
  • the corresponding second amplitude threshold can be set for each device to be dissipated.
  • the actual power information of the device to be dissipated is mainly used to reflect the power consumption of the device to be dissipated, and is the data basis for this heat dissipation control.
  • the implementation form of the actual power information of the device to be dissipated is not limited, and may be any data form that can reflect the power consumption of the device to be dissipated. For example, whenever the heat dissipation control condition is triggered, the power value of the at least one device to be dissipated at the moment when the heat dissipation control condition is triggered may be separately collected as the actual power information of the at least one device to be dissipated.
  • the average power value of the at least one device to be dissipated during the current heat dissipation control and the last heat dissipation control is obtained as the actual power information of the at least one device to be dissipated.
  • the information of the power value of the device to be dissipated is used.
  • the power value of the device to be dissipated may be the overall power value of the device to be dissipated.
  • the overall power value of the device to be dissipated can be defined as the sum of the power of the main internal components of the device to be dissipated, or it can be defined as the sum of the power of all internal components of the device to be dissipated.
  • the power value of the device to be dissipated may be the power of a certain internal component of the device to be dissipated, for example, it may be the power of a CPU, or the power of a memory.
  • an implementation manner of step 504 includes: sending the target refrigeration parameter to the refrigeration system, so that the refrigeration system can perform heat dissipation for at least one device to be dissipated in the designated space according to the target refrigeration parameter.
  • the refrigeration system can receive the target refrigeration parameter sent by the heat dissipation control device, and compare the target refrigeration parameter with the currently used refrigeration parameter; if the two are different, replace the currently used refrigeration parameter with the target refrigeration parameter and follow the target The refrigeration parameters continue to perform the refrigeration work; if the two are the same, the refrigeration work continues according to the currently used refrigeration parameter, so as to achieve the purpose of dissipating at least one device to be dissipated in the designated space according to the target refrigeration parameter.
  • an implementation manner of step 504 includes: comparing the target refrigeration parameter with the refrigeration parameter currently used by the refrigeration system, and when the target refrigeration parameter is different from the refrigeration parameter currently used by the refrigeration system Next, the target refrigeration parameter is sent to the refrigeration system so that the refrigeration system can continue the refrigeration work according to the target refrigeration parameter.
  • the refrigeration system can receive the target refrigeration parameter sent by the heat dissipation control device, adjust the currently used refrigeration parameter to the target refrigeration parameter, and continue the refrigeration work according to the target refrigeration parameter; if the target refrigeration parameter sent by the heat dissipation control device is not received In the case of parameters, the refrigeration work can be continued according to the currently used refrigeration parameters, so as to achieve the purpose of dissipating at least one device to be dissipated in the designated space in accordance with the target refrigeration parameters.
  • an implementation manner of step 502 includes: determining at least one candidate refrigeration parameter according to the range of the refrigeration parameter used by the overheating risk prediction model in the training phase; for each candidate refrigeration parameter, at least one candidate refrigeration parameter The actual power information of the heat sink and the candidate refrigeration parameter are input to the overheating risk prediction model to obtain the probability of the overheating risk in the specified space under the candidate refrigeration parameter.
  • an implementation manner of step 503 includes: based on the overheating risk probability threshold corresponding to the designated space, selecting a probability that is less than the overheating risk probability in the designated space under at least one candidate refrigeration parameter.
  • the threshold probability is used as the target probability, and among at least one candidate refrigeration parameter, the refrigeration parameter corresponding to the target probability is used as the target refrigeration parameter.
  • a probability that is less than the threshold of the overheating risk probability can be randomly selected from the probability that the overheating risk occurs in the designated space under at least one candidate refrigeration parameter as the target probability; or, it can be selected from the designated space in the overheating risk.
  • the maximum probability that is less than the overheating risk probability threshold is selected as the target probability; alternatively, the probability of overheating risk occurring in at least one candidate refrigeration parameter can be selected as the target probability;
  • the threshold value and the probability within the set probability range is used as the target probability.
  • the allowable thermal failure rate of the application or service carried in the designated space can be obtained, and the thermal failure rate can be converted into the overheating risk probability threshold corresponding to the designated space.
  • a statistical method may be used to convert the thermal failure rate into an overheating risk probability threshold corresponding to the designated space.
  • the thermal failure rate refers to the maximum number of overheating risks that can occur in a specified space within a certain period of time. The maximum number of times here refers to the sum of the number of overheating risks of various devices that have overheating risks in a certain period of time.
  • the following methods may be used, but not limited to, the pre-training to obtain the overheating risk prediction model:
  • each set of labeled sample data includes at least one sample power information corresponding to at least one device to be dissipated, sample refrigeration parameters corresponding to the refrigeration system, and a labeling result of whether a designated space has overheating risk under the set of labeled sample data.
  • the overheating risk prediction model trained based on these labeled sample data can predict the probability of overheating risk in a specified space under the candidate cooling parameter for a given set of power information and a candidate cooling parameter.
  • combining the sample generation method based on real data and the sample generation method based on CFD simulation to generate multiple sets of labeled sample data includes: generating at least A set of marked historical sample data; and use the CFD model to perform simulation calculations between power information and cooling parameters to generate at least one set of marked simulated sample data.
  • the process of generating at least one set of labeled historical sample data includes: obtaining at least one set of unlabeled historical sample data, and each set of unlabeled historical sample data includes the history of at least one device to be dissipated in the same historical moment or historical period. Power information and historical refrigeration parameters of the refrigeration system; for each group of unlabeled historical sample data, determine whether the space is designated according to the temperature of at least one internal component of the device to be dissipated in the corresponding historical time or historical period and the corresponding overheating temperature threshold of the internal component If the risk of overheating occurs, at least one set of marked historical sample data is obtained.
  • the process of generating at least one set of labeled simulated sample data includes: designing at least one set of unlabeled simulated sample data, and each set of unlabeled simulated sample data includes at least one set of simulated power information corresponding to at least one device to be dissipated; Simulated refrigeration parameters corresponding to the refrigeration system; for each group of unlabeled simulation sample data, use the CFD model to simulate the group of unlabeled simulation sample data to obtain the temperature of at least one device to be dissipated, and use at least one to be dissipated The temperature of the internal components of the device and the corresponding overheating temperature threshold of the internal components are marked whether the designated space has overheating risk, and at least one set of marked simulation sample data is obtained.
  • At least one set of labeled historical sample data can also be used to perform the CFD model Parameter correction. After that, use the corrected CFD model to generate at least one set of labeled simulation sample data, which is conducive to improving the reliability and authenticity of the labeled simulation sample data, and is conducive to improving the accuracy of the overheating risk prediction model trained accordingly .
  • the sample generation method based on real data is combined with the sample generation method based on CFD simulation, and the CFD simulation calculation method is used to provide the required sample data for model training, which can make up for the sample generation method based on real data.
  • the insufficiency of the sample generation method is not only conducive to obtaining sufficient sample data, but also sample data running in a relatively extreme situation in a designated space, which can effectively improve the robustness of the thermal risk prediction model.
  • each set of labeled sample data includes: at least one sample power information, sample refrigeration parameters, and sample atmospheric temperature.
  • each set of labeled sample data includes: at least one sample power information, sample refrigeration parameters, and sample atmospheric temperature.
  • it also includes the operation of acquiring the external atmospheric temperature at the corresponding time as the sample atmospheric temperature and adding the sample atmospheric temperature to the labeled sample data.
  • the outside air temperature at the time corresponding to the first set of labeled sample data is also obtained as the sample air temperature, and the first set of labeled sample data corresponds to The sample atmospheric temperature is added to the first set of labeled sample data; wherein, the first set of labeled sample data is any set of labeled sample data in the multiple sets of labeled sample data.
  • model training can be performed based on multiple sets of labeled sample data containing sample air temperature to obtain an overheating risk prediction model.
  • the overheating risk prediction model trained based on the labeled sample data containing the sample air temperature parameter can predict the specified space under the candidate cooling parameter for a given set of power information, external air temperature and a candidate cooling parameter Probability of the risk of overheating.
  • step 501 each time the heat dissipation control condition is triggered, the heat dissipation control device also needs to obtain the external atmospheric temperature corresponding to when the heat dissipation control condition is triggered.
  • step 502 the actual power information of the at least one device to be dissipated and the corresponding external atmospheric temperature when the heat dissipation control condition is triggered can be input as input parameters into the overheating risk prediction model, so as to obtain the specified space system in at least one candidate refrigeration system. The probability of the risk of overheating under the parameters.
  • the overheating risk prediction model can also be updated.
  • updating the thermal risk prediction model includes but is not limited to the following situations: whenever the model update cycle arrives, re-training the overheating risk prediction model; re-training the overheating risk prediction model whenever the number of heat-dissipating devices in the designated space changes. Perform model training on the overheating risk prediction model; whenever the topological structure between the devices to be radiated in the designated space changes, retrain the overheating risk prediction model.
  • the updated overheating risk prediction model can be used in the subsequent heat dissipation control process, which is beneficial to improve the accuracy and precision of the heat dissipation control.
  • FIG. 6 is a schematic flowchart of a model training method provided by an exemplary embodiment of this application. As shown in Figure 6, the method includes:
  • each set of labeled sample data includes at least one sample power information corresponding to at least one device to be dissipated, and sample refrigeration corresponding to the refrigeration system Parameters and marking results indicating whether there is a risk of overheating in the designated space under the set of marked sample data.
  • an implementation manner of step 601 includes: generating at least one set of marked historical sample data according to historical power information of at least one device to be dissipated and historical cooling parameters of the refrigeration system; and using the CFD model to compare power information The simulation calculation between the refrigeration parameters generates at least one set of labeled simulation sample data.
  • the process of generating at least one set of labeled historical sample data includes: obtaining at least one set of unlabeled historical sample data, and each set of unlabeled historical sample data includes the history of at least one device to be dissipated in the same historical moment or historical period. Power information and historical refrigeration parameters of the refrigeration system; for each group of unlabeled historical sample data, determine whether the space is designated according to the temperature of at least one internal component of the device to be dissipated in the corresponding historical time or historical period and the corresponding overheating temperature threshold of the internal component If the risk of overheating occurs, at least one set of marked historical sample data is obtained.
  • the process of generating at least one set of labeled simulated sample data includes: designing at least one set of unlabeled simulated sample data, and each set of unlabeled simulated sample data includes at least one set of simulated power information corresponding to at least one device to be dissipated; Simulated refrigeration parameters corresponding to the refrigeration system; for each group of unlabeled simulation sample data, use the CFD model to simulate the group of unlabeled simulation sample data to obtain the temperature of at least one device to be dissipated, and use at least one to be dissipated The temperature of the internal components of the device and the corresponding overheating temperature threshold of the internal components are marked whether the designated space has overheating risk, and at least one set of marked simulation sample data is obtained.
  • At least one set of labeled historical sample data can also be used to perform the CFD model Parameter correction. After that, use the corrected CFD model to generate at least one set of labeled simulation sample data, which is conducive to improving the reliability and authenticity of the labeled simulation sample data, and is conducive to improving the accuracy of the overheating risk prediction model trained accordingly .
  • each set of labeled sample data includes: at least one sample power information, sample refrigeration parameters, and sample atmospheric temperature.
  • each set of labeled sample data includes: at least one sample power information, sample refrigeration parameters, and sample atmospheric temperature.
  • it also includes the operation of acquiring the external atmospheric temperature at the corresponding time as the sample atmospheric temperature and adding it to the corresponding labeled sample data.
  • the outside air temperature at the time corresponding to the first set of labeled sample data is also obtained as the sample air temperature, and the first set of labeled sample data corresponds to The sample atmospheric temperature is added to the first set of labeled sample data; wherein, the first set of labeled sample data is any set of labeled sample data in the multiple sets of labeled sample data.
  • model training can be performed based on multiple sets of labeled sample data containing sample air temperature to obtain an overheating risk prediction model.
  • the overheating risk prediction model trained based on the labeled sample data containing the sample air temperature parameter can predict the specified space under the candidate cooling parameter for a given set of power information, external air temperature and a candidate cooling parameter Probability of the risk of overheating.
  • the overheating risk prediction model can also be updated.
  • updating the thermal risk prediction model includes but is not limited to the following situations: whenever the model update cycle arrives, re-training the overheating risk prediction model; re-training the overheating risk prediction model whenever the number of heat-dissipating devices in the designated space changes. Perform model training on the overheating risk prediction model; whenever the topological structure between the devices to be radiated in the designated space changes, retrain the overheating risk prediction model.
  • the sample generation method based on real data is combined with the sample generation method based on CFD simulation.
  • the CFD simulation calculation method provides the required sample data for model training, which can compensate for the sample generation based on real data.
  • the insufficiency of the method is not only conducive to obtaining sufficient sample data, but also sample data running in a relatively extreme situation in the designated space, which can effectively improve the robustness of the thermal risk prediction model.
  • the execution subject of each step of the method provided in the foregoing embodiment may be the same device, or different devices may also be the execution subject of the method.
  • the execution subject of steps 501 to 504 may be device A; for another example, the execution subject of steps 501 and 502 may be device A, and the execution subject of steps 503 and 504 may be device B; and so on.
  • FIG. 7 is a schematic structural diagram of a heat dissipation control device provided by an exemplary embodiment of this application. As shown in FIG. 7, the device includes: a memory 71 and a processor 72.
  • the memory 71 is used to store computer programs, and can be configured to store other various data to support operations on the heat dissipation control device. Examples of such data include instructions for any application or method operating on the heat dissipation control device, contact data, phone book data, messages, pictures, videos, etc.
  • the processor 72 is coupled with the memory 71, and is configured to execute a computer program in the memory 71, so as to obtain the actual power information of at least one device to be dissipated in the designated space every time a heat dissipation control condition is triggered;
  • the actual power information of the heat sink is input into the overheating risk prediction model to obtain the probability of overheating risk in the designated space under at least one candidate refrigeration parameter; determine the target refrigeration parameter according to the probability of overheating risk in the designated space under at least one candidate refrigeration parameter;
  • the refrigeration system is controlled to dissipate heat from at least one device to be dissipated in the designated space.
  • the designated space refers to any physical space capable of accommodating and accommodating equipment to be dissipated, such as but not limited to: computer rooms, data centers, office buildings, office areas in a company environment, or designated areas in a home environment.
  • a designated area in the home environment stores heat-generating devices to be dissipated, such as televisions, smart speakers, large game consoles, smart air conditioners, purifiers, home escort robots, and personal computers.
  • a refrigeration system can be provided.
  • the refrigeration system can be deployed in a designated space, but it is not limited to this.
  • the refrigeration system is mainly responsible for taking away the heat in the designated space and dissipating heat for the equipment to be radiated in the designated space.
  • the processor 72 is specifically configured to perform at least one of the following operations when acquiring the actual power information of at least one device to be dissipated in the designated space:
  • the actual power information of the device to be dissipated is acquired;
  • the actual power information of the at least one device to be dissipated is acquired.
  • the processor 72 when acquiring the actual power information of at least one device to be dissipated in the designated space, the processor 72 is specifically configured to:
  • the average power value of the at least one device to be dissipated during the current heat dissipation control and the last heat dissipation control is respectively obtained as the actual power information of the at least one device to be dissipated.
  • the processor 72 when the processor 72 obtains the probability of the overheating risk occurring in the designated space under at least one candidate cooling parameter, it is specifically configured to: determine at least the range of the cooling parameter used by the overheating risk prediction model in the training phase according to the A candidate refrigeration parameter; for each candidate refrigeration parameter, the actual power information of at least one device to be dissipated and the candidate refrigeration parameter are input into the overheating risk prediction model to obtain the probability of overheating risk in the specified space under the candidate refrigeration parameter.
  • the processor 72 determines the target refrigeration parameter, it is specifically configured to: from the probability of an overheating risk occurring in the specified space under at least one candidate refrigeration parameter, select the one that is less than the overheating risk probability threshold corresponding to the specified space. Target probability; among at least one candidate refrigeration parameter, the refrigeration parameter corresponding to the target probability is taken as the target refrigeration parameter.
  • the processor 72 is specifically configured to select the maximum probability that is less than the overheating risk probability threshold from the probability of the overheating risk occurring in the specified space under at least one candidate refrigeration parameter as the target probability.
  • the processor 72 is further configured to: before selecting a target probability that is less than the overheating risk probability threshold corresponding to the designated space, convert the allowable thermal failure rate of the application or service carried by the designated space into the overheating risk probability corresponding to the designated space Threshold; the thermal failure rate indicates the maximum number of times that the risk of overheating can occur in a specified space within a certain period of time.
  • the processor 72 is further configured to: combine a sample generation method based on real data and a sample generation method based on CFD simulation to generate multiple sets of labeled sample data; and use multiple sets of labeled sample data to perform a deep neural network model Trained to get an overheating risk prediction model.
  • each set of labeled sample data includes at least one sample power information corresponding to at least one device to be dissipated, sample refrigeration parameters corresponding to the refrigeration system, and a labeling result of whether a designated space has overheating risk under the set of labeled sample data.
  • the processor 72 when the processor 72 generates multiple sets of labeled sample data, it is specifically configured to: generate at least one set of labeled historical sample data according to historical power information of the at least one device to be dissipated and historical cooling parameters of the refrigeration system; and Use the CFD model to perform simulation calculations between power information and cooling parameters to generate at least one set of labeled simulation sample data.
  • the processor 72 when the processor 72 generates at least one set of labeled historical sample data, it is specifically configured to: obtain at least one set of unlabeled historical sample data, and each set of unlabeled historical sample data includes the same historical moment or history. Historical power information of at least one device to be cooled and historical cooling parameters of the refrigeration system in a time period; for each group of unmarked historical sample data, according to the temperature and internal components of at least one internal component of the device to be cooled in the corresponding historical time or historical period The corresponding overheating temperature threshold is used to mark whether an overheating risk occurs in the designated space, and at least one set of marked historical sample data is obtained.
  • the processor 72 when it generates at least one set of labeled simulated sample data, it is specifically configured to: design at least one set of unlabeled simulated sample data, and each set of unlabeled simulated sample data includes at least one set of simulated sample data to be dissipated At least one simulated power information corresponding to the equipment and simulated refrigeration parameters corresponding to the refrigeration system; for each group of unmarked simulation sample data, the CFD model is used to simulate the group of unmarked simulation sample data to obtain at least one inside of the equipment to be dissipated For the temperature of the device, the temperature of the internal device of at least one device to be dissipated and the corresponding overheating temperature threshold of the internal component are used to mark whether the designated space has overheating risk, and at least one set of marked simulation sample data is obtained.
  • the processor 72 is further configured to: before generating at least one set of labeled simulated sample data, use at least one set of labeled historical sample data to perform parameter correction on the CFD model.
  • the processor 72 is further configured to: in the process of generating multiple sets of labeled sample data, for the first set of labeled sample data, obtain the outside air temperature at the time corresponding to the first set of labeled sample data as the sample atmosphere The temperature of the sample is added to the first set of labeled sample data; wherein, the first set of labeled sample data is any one of multiple sets of labeled sample data.
  • the processor 72 is also used to obtain the corresponding outside air temperature when the heat dissipation control condition is triggered.
  • the processor 72 obtains the probability of an overheating risk in the specified space under at least one candidate cooling parameter, it is specifically configured to: combine the actual power information of the at least one device to be dissipated and the corresponding external atmospheric temperature when the heat dissipation control condition is triggered
  • the overheating risk prediction model is input as an input parameter to obtain the probability of overheating risk occurring in the designated space under at least one candidate refrigeration parameter.
  • processor 72 is further configured to perform at least one of the following operations:
  • the overheating risk prediction model is retrained.
  • processor 72 controls the refrigeration system to dissipate heat from at least one device to be dissipated in the designated space according to the target refrigeration parameter, it is specifically configured to:
  • the target refrigeration parameter is sent to the refrigeration system for the refrigeration system to perform refrigeration work according to the target refrigeration parameter.
  • the heat dissipation control device further includes: a communication component 73, a display 74, a power supply component 75, an audio component 76 and other components. Only some components are schematically shown in FIG. 7, which does not mean that the heat dissipation control device only includes the components shown in FIG. 7. In addition, according to the different implementation forms of the heat dissipation control device, the components in the dashed box in FIG. 7 are optional components, not mandatory components.
  • the heat dissipation control device when the heat dissipation control device is implemented as a terminal device such as a smart phone, a tablet computer, or a desktop computer, it may include the components in the dashed box in Figure 7; when the heat dissipation control device is implemented as a conventional server, cloud server, data center or server array, etc. When the server device is used, the components in the dashed box in Figure 7 may not be included.
  • the heat dissipation control device adopts a pre-trained overheating risk prediction model, which reflects the overheating risk relationship between the device power information and the cooling parameters, and further, on the basis of the overheating risk prediction model, According to the power changes of the equipment to be dissipated in the designated space, the refrigeration parameters of the refrigeration system are dynamically adjusted to achieve the purpose of dynamic heat dissipation control, which is beneficial to reduce the energy consumption of the refrigeration system and save power resources.
  • an embodiment of the present application also provides a computer-readable storage medium storing a computer program.
  • the processor can cause the processor to implement the steps in the foregoing heat dissipation control method embodiment, or implement the foregoing system implementation.
  • the corresponding operation in the example For detailed steps or operations, please refer to the foregoing embodiment, which will not be repeated here.
  • FIG. 8 is a schematic structural diagram of a model training device provided by an exemplary embodiment of this application. As shown in FIG. 8, the device includes: a memory 81 and a processor 82.
  • the memory 81 is used to store computer programs, and can be configured to store various other data to support operations on the model training device. Examples of such data include instructions for any application or method used to operate on the model training device, contact data, phone book data, messages, pictures, videos, etc.
  • the processor 82 coupled with the memory 81, is configured to execute the computer program in the memory 81, and is used to: combine the sample generation method based on real data and the sample generation method based on CFD simulation to generate multiple sets of labeled sample data;
  • the labeled sample data is trained with a deep neural network model to obtain an overheating risk prediction model; where each set of labeled sample data includes at least one sample power information corresponding to at least one device to be dissipated, sample refrigeration parameters corresponding to the refrigeration system, and data in the group Mark the result of marking whether there is a risk of overheating in the designated space under the sample data.
  • the designated space refers to any physical space capable of accommodating and accommodating equipment to be dissipated, such as but not limited to: computer rooms, data centers, office buildings, office areas in a company environment, or designated areas in a home environment.
  • a designated area in the home environment stores heat-generating devices to be dissipated, such as televisions, smart speakers, large game consoles, smart air conditioners, purifiers, home escort robots, and personal computers.
  • a refrigeration system can be provided.
  • the refrigeration system can be deployed in a designated space, but it is not limited to this.
  • the refrigeration system is mainly responsible for taking away the heat in the designated space and dissipating heat for the equipment to be radiated in the designated space.
  • the processor 82 when the processor 82 generates multiple sets of labeled sample data, it is specifically configured to: generate at least one set of labeled history based on historical power information of at least one device to be dissipated and historical cooling parameters of the refrigeration system Sample data; and use the CFD model to perform simulation calculations between power information and cooling parameters to generate at least one set of labeled simulation sample data.
  • the processor 82 when the processor 82 generates at least one set of marked historical sample data, it is specifically configured to: obtain at least one set of unmarked historical sample data, and each set of unmarked historical sample data includes the same historical moment or history. Historical power information of at least one device to be cooled and historical cooling parameters of the refrigeration system in a time period; for each group of unmarked historical sample data, according to the temperature and internal components of at least one internal component of the device to be cooled in the corresponding historical time or historical period The corresponding overheating temperature threshold is used to mark whether an overheating risk occurs in the designated space, and at least one set of marked historical sample data is obtained.
  • the processor 82 when the processor 82 generates at least one set of labeled simulated sample data, it is specifically configured to: design at least one set of unlabeled simulated sample data, and each set of unlabeled simulated sample data includes at least one set of simulated sample data to be dissipated At least one simulated power information corresponding to the equipment and simulated refrigeration parameters corresponding to the refrigeration system; for each group of unmarked simulation sample data, the CFD model is used to simulate the group of unmarked simulation sample data to obtain at least one inside of the equipment to be dissipated For the temperature of the device, the temperature of the internal device of at least one device to be dissipated and the corresponding overheating temperature threshold of the internal component are used to mark whether the designated space has overheating risk, and at least one set of marked simulation sample data is obtained.
  • the processor 82 is further configured to: before using the CFD model to perform simulation calculations between the power information and the cooling parameters to generate at least one set of labeled simulation sample data, use at least one set of labeled historical sample data , Carry on parameter correction to the CFD model.
  • the model training device further includes: a communication component 83, a display 84, a power supply component 85, an audio component 86 and other components. Only part of the components are schematically shown in FIG. 8, which does not mean that the model training device only includes the components shown in FIG. 8.
  • the components in the dashed box in FIG. 8 are optional components, not mandatory components.
  • the model training device when the model training device is implemented as a terminal device such as a smart phone, a tablet computer, or a desktop computer, it can include the components in the dashed box in Figure 8; when the model training device is implemented as a conventional server, cloud server, data center, or server array, etc.
  • the components in the dashed box in Figure 8 may not be included.
  • the model training device provided in this embodiment combines the sample generation method based on real data and the sample generation method based on CFD simulation. Through the method of CFD simulation calculation, it provides the required sample data for model training, which can compensate for the sample data based on real data.
  • the insufficiency of the sample generation method is not only conducive to obtaining sufficient sample data, but also sample data running in a relatively extreme situation in a designated space, which can effectively improve the robustness of the thermal risk prediction model.
  • an embodiment of the present application also provides a computer-readable storage medium storing a computer program.
  • the processor can cause the processor to implement the steps in the above-mentioned model training method embodiment, or implement the above-mentioned system implementation.
  • the corresponding operation in the example For detailed steps or operations, please refer to the foregoing embodiment, which will not be repeated here.
  • the communication components in FIGS. 7 and 8 described above are configured to facilitate wired or wireless communication between the device where the communication component is located and other devices.
  • the device where the communication component is located can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G/LTE, 5G and other mobile communication networks, or a combination of them.
  • the communication component receives a broadcast signal or broadcast related information from an external broadcast management system via a broadcast channel.
  • the communication component may further include a near field communication (NFC) module, radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra wideband (UWB) technology, Bluetooth (BT) technology, and the like.
  • NFC near field communication
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra wideband
  • Bluetooth Bluetooth
  • the above-mentioned display in FIGS. 7 and 8 includes a screen, and the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user.
  • the touch panel includes one or more touch sensors to sense touch, sliding, and gestures on the touch panel. The touch sensor may not only sense the boundary of a touch or slide action, but also detect the duration and pressure related to the touch or slide operation.
  • the power supply components in Figures 7 and 8 above provide power for various components of the equipment where the power supply component is located.
  • the power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device where the power supply component is located.
  • the audio components in FIGS. 7 and 8 described above may be configured to output and/or input audio signals.
  • the audio component includes a microphone (MIC).
  • the microphone When the device where the audio component is located is in an operating mode, such as call mode, recording mode, and voice recognition mode, the microphone is configured to receive external audio signals.
  • the received audio signal can be further stored in a memory or sent via a communication component.
  • the audio component further includes a speaker for outputting audio signals.
  • the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
  • computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
  • These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing equipment to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including the instruction device.
  • the device implements the functions specified in one process or multiple processes in the flowchart and/or one block or multiple blocks in the block diagram.
  • These computer program instructions can also be loaded on a computer or other programmable data processing equipment, so that a series of operation steps are executed on the computer or other programmable equipment to produce computer-implemented processing, so as to execute on the computer or other programmable equipment.
  • the instructions provide steps for implementing the functions specified in one process or multiple processes in the flowchart and/or one block or multiple blocks in the block diagram.
  • the computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
  • processors CPUs
  • input/output interfaces network interfaces
  • memory volatile and non-volatile memory
  • the memory may include non-permanent memory in a computer readable medium, random access memory (RAM) and/or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer readable media.
  • RAM random access memory
  • ROM read-only memory
  • flash RAM flash memory
  • Computer-readable media include permanent and non-permanent, removable and non-removable media, and information storage can be realized by any method or technology.
  • the information can be computer-readable instructions, data structures, program modules, or other data.
  • Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, Magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media can be used to store information that can be accessed by computing devices. According to the definition in this article, computer-readable media does not include transitory media, such as modulated data signals and carrier waves.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Automation & Control Theory (AREA)
  • Cooling Or The Like Of Electrical Apparatus (AREA)
  • Air Conditioning Control Device (AREA)

Abstract

一种散热控制与模型训练方法、设备、系统及存储介质。该方法包括：预先训练得到过热风险预测模型，通过该模型体现设备功率与制冷参数之间存在的过热风险关系，进而，在该过热风险预测模型的基础上，可根据指定空间内待散热设备的功率变化情况，动态调整制冷系统的制冷参数，达到动态散热控制的目的，有利于降低制冷系统的能耗，节约电能资源。

Description

散热控制与模型训练方法、设备、系统及存储介质 技术领域
本申请涉及互联网技术领域，尤其涉及一种散热控制与模型训练方法、设备、系统及存储介质。
背景技术
互联网数据中心(Internet Data Center，IDC)除了包括计算机、服务器等IT设备之外，还包括空调、水泵等制冷系统。制冷系统向IDC的机房提供冷却空气，以保证IDC中的IT设备能够正常工作。
在现有技术中，制冷系统需要提供足够的冷却空气，将IDC机房维持在恒定的室内温度，以防出现过热风险，但是这会消耗巨额电量，存在资源浪费。
发明内容
本申请的多个方面提供一种散热控制与模型训练方法、设备、系统及存储介质，用以降低制冷系统的能耗，节约电能资源。
本申请实施例提供一种散热控制方法，包括：每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息；将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述指定空间在至少一个候选制冷参数下发生过热风险的概率；根据所述指定空间在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据所述目标制冷参数，控制制冷系统对所述指定空间内的至少一个待散热设备进行散热。
本申请实施例还提供一种模型训练方法，包括：结合基于真实数据的样 本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；利用所述多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型；其中，每组标记样本数据包括与所述至少一个待散热设备对应的至少一个样本功率信息、与所述制冷系统对应的样本制冷参数以及在该组标记样本数据下所述指定空间是否发生过热风险的标记结果。
本申请实施例还提供一种散热控制设备，包括：存储器和处理器；所述存储器，用于存储计算机程序；所述处理器，与所述存储器耦合，用于执行所述计算机程序，以用于：每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息；将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述指定空间在至少一个候选制冷参数下发生过热风险的概率；根据所述指定空间在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据所述目标制冷参数，控制制冷系统对所述指定空间内的至少一个待散热设备进行散热。
本申请实施例还提供一种模型训练设备，包括：存储器和处理器；所述存储器，用于存储计算机程序；所述处理器，与所述存储器耦合，用于执行所述计算机程序，以用于：结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；利用所述多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型；其中，每组标记样本数据包括与所述至少一个待散热设备对应的至少一个样本功率信息、与所述制冷系统对应的样本制冷参数以及在该组标记样本数据下所述指定空间是否发生过热风险的标记结果。
本申请实施例还提供一种机房系统，包括：机房，以及位于所述机房内的至少一个待散热装置、制冷系统以及散热控制设备；所述散热控制设备，用于每当散热控制条件被触发时，获取所述至少一个待散热设备的实际功率信息，将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述机房系统在至少一个候选制冷参数下发生过热风险的概率；根据所述机房系统在至少一个候选制冷参数下发生过热风险的概率，确定目标制 冷参数；根据所述目标制冷参数，控制所述制冷系统对所述至少一个待散热设备进行散热；所述制冷系统，用于在所述散热控制设备的控制下，对所述机房内的至少一个待散热设备进行散热。
本申请实施例还提供一种数据中心系统，包括：至少一个机房；每个机房包括：至少一个待散热设备、制冷系统以及散热控制设备；所述散热控制设备，用于每当散热控制条件被触发时，获取其所属机房内至少一个待散热设备的实际功率信息，将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述机房在至少一个候选制冷参数下发生过热风险的概率；根据所述机房在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据所述目标制冷参数，控制其所属机房内的制冷系统对所述至少一个待散热设备进行散热；所述制冷系统，用于在所述散热控制设备的控制下，对其所属机房内的至少一个待散热设备进行散热。
本申请实施例还提供另一种数据中心系统，包括：至少一个机房、制冷系统以及散热控制设备；其中，每个机房包括至少一个待散热设备，所述制冷系统包括部署在每个机房内的制冷设备；所述散热控制设备，用于针对每个机房，每当散热控制条件被触发时，获取所述机房内至少一个待散热设备的实际功率信息，将所述机房内至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述机房在至少一个候选制冷参数下发生过热风险的概率；根据所述机房在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据所述目标制冷参数，控制所述机房内的制冷设备对所述机房内的至少一个待散热设备进行散热。
本申请实施例还提供一种存储有计算机程序的计算机可读存储介质，当所述计算机程序被处理器执行时，致使所述处理器实现本申请实施例提供的散热控制方法中的步骤。
本申请实施例还提供一种存储有计算机程序的计算机可读存储介质，当所述计算机程序被处理器执行时，致使所述处理器实现本申请实施例提供的模型训练方法中的步骤。
在本申请实施例中，预先训练得到过热风险预测模型，通过该模型体现设备功率信息与制冷参数之间存在的过热风险关系，进而，在该过热风险预测模型的基础上，可根据指定空间内待散热设备的功率变化情况，动态调整制冷系统的制冷参数，达到动态散热控制的目的，有利于降低制冷系统的能耗，节约电能资源。
附图说明
此处所说明的附图用来提供对本申请的进一步理解，构成本申请的一部分，本申请的示意性实施例及其说明用于解释本申请，并不构成对本申请的不当限定。在附图中：
图1为本申请示例性实施例提供的一种机房系统的结构示意图；
图2a为申请示例性实施例提供的模型训练过程的示意图；
图2b为本申请示例性实施例提供的一种过热风险预测模型的结构示意图；
图2c为本申请示例性实施例提供的一种模型预测结果的状态示意图；
图3为本申请示例性实施例提供的一种数据中心系统的结构示意图；
图4为本申请示例性实施例提供的另一种数据中心系统的结构示意图；
图5为本申请示例性实施例提供的一种散热控制方法的流程示意图；
图6为本申请示例性实施例提供的一种模型训练方法的流程示意图；
图7为本申请示例性实施例提供的一种散热控制设备的结构示意图；
图8为本申请示例性实施例提供的模型训练设备的结构示意图。
具体实施方式
为使本申请的目的、技术方案和优点更加清楚，下面将结合本申请具体实施例及相应的附图对本申请技术方案进行清楚、完整地描述。显然，所描 述的实施例仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
针对现有制冷系统存在的功耗较大，存在资源浪费等技术问题，在本申请一些实施例中，预先训练得到过热风险预测模型，通过该模型体现设备功率与制冷参数之间存在的过热风险关系，在该过热风险预测模型的基础上，根据指定空间内待散热设备的功率变化情况，动态调整制冷系统的制冷参数，达到动态散热控制的目的，从而降低制冷系统的能耗，节约电能资源。
以下结合附图，详细说明本申请各实施例提供的技术方案。
图1为本申请示例性实施例提供的一种机房系统的结构示意图。如图1所示，本实施例的机房系统100包括：机房，机房是指存放机器设备的物理场所，例如可以是一个房间或厂房等。进一步，如图1所示，机房系统100还包括：位于机房内的至少一个待散热设备101、制冷系统102以及散热控制设备103。本实施例并不限定机房内待散热设备101的数量，可以是一个，也可以是多个。一般来说，一个机房内会包含多个待散热设备101。
其中，待散热设备101是指可产生热量，且对工作环境中的环境温度具有一定要求的电子设备。在本实施例中，并不限定待散热设备101的设备形态。可选地，待散热设备101可以是IT设备，但不限于此。例如，至少一个待散热设备101可以包括但不限于以下至少一种设备形态：机柜设备、服务器设备、计算机设备、打印机、集线器、电源设备、存储设备以及网络交换设备等。服务器设备可以是包括但不限于：常规服务器、服务器阵列或云服务器等。电源设备可以是蓄电池设备、干电池设备、或不间断电源(UPS)等。存储设备可以包括但不限于：磁盘、磁盘阵列、硬盘、网络存储设备(NAS)等。在机房系统100的至少一个待散热设备101上运行有至少一种应用或服务，例如云计算服务、游戏服务，即时通信服务、邮件服务或在线交易服务等等。
待散热设备101对机房内的温度有一定要求，如果机房内的温度过高， 待散热设备101可能会发生故障、失灵，甚至被烧毁。为了给待散热设备101提供良好的工作环境，机房内还设置有制冷系统102，制冷系统102主要负责带走机房内的热量，为机房内的待散热设备101进行散热。在本实施例中，并不限定制冷系统102的类型和工作原理，例如可以是空调系统，或者是水冷系统，或者是空调系统与水冷系统的组合。
其中，机房内待散热设备101运行中的功耗是影响机房温度的一种主要因素。但是，鉴于机房系统100的复杂性，在现有技术中无法证明待散热设备101的功耗与制冷之间的关系，为此，会按照待散热设备101的最大功耗计算制冷参数，并按照计算出的制冷参数控制制冷系统102对机房进行散热，这样不论机房内待散热设备101的功耗是多少都能让机房保持较低的温度，确保待散热设备101不发生过热风险。但是，待散热设备101的工作负载有时会发生较大变化，工作负载的变化，会导致待散热设备101的功耗发生变化，这意味着待散热设备101不会一直处于最大功耗状态，因此，按照待散热设备101的最大功耗对制冷系统102进行散热控制，会浪费大量电能资源。
在本实施例中，通过模型训练，预先得到过热风险预测模型，该模型可体现设备功率信息与制冷参数之间存在的过热风险关系。基于此，散热控制设备103可在该过热风险预测模型的基础上，结合机房内至少一个待散热设备101的功率变化情况，控制制冷系统102对机房进行动态散热，制冷系统102可根据实际需要进行制冷工作，可降低制冷系统102的能耗，节约电能资源。其中，待散热设备101的功率信息反应待散热设备101的功耗，也可反映待散热设备101的工作负载。其中，关于模型训练过程可参见后续实施例，在此不再赘述。
在本实施例中，可以设置散热控制条件，每当散热控制条件被触发时，就在过热风险预测模型的基础上，结合机房内至少一个待散热设备101的实际功率信息进行一次散热控制。由此可见，散热控制设备103在过热风险预测模型的基础上，结合机房内待散热设备101的功率变化情况，控制制冷系统102对机房进行动态散热的过程，可以包括多次散热控制。下面对本实施 例基于过热风险预测模型的散热控制过程进行说明：
每当散热控制条件被触发时，散热控制设备103获取机房内至少一个待散热设备101的实际功率信息；将至少一个待散热设备101的实际功率信息输入过热风险预测模型，以得到机房系统100在至少一个候选制冷参数下发生过热风险的概率；然后根据机房系统100在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据目标制冷参数，控制制冷系统102对机房内的至少一个待散热设备101进行散热。对制冷系统102来说，可在散热控制设备103的控制下，对机房内的至少一个待散热设备进行散热。
其中，机房系统100在某个候选制冷参数下发生过热风险的概率，主要是指假定制冷系统102采用该候选制冷参数，则机房系统100中出现待散热设备发生过热风险情况的概率。其中，发生过热风险的待散热设备的数量可以是一个，也可以是多个，对此不做限定。
值得说明的是，本实施例不对散热控制条件进行限定，可根据散热控制需求灵活设定。下面对散热控制条件进行举例说明：
示例1：在该示例中，考虑到制冷系统102在每次调整制冷参数之后，一般需要经过一定时间才能达到预期的散热效果，则可以预先设定散热控制周期，将散热控制周期作为散热控制条件。基于此，散热控制设备103可以按照该散热控制周期，周期性地对机房进行散热控制，这既可以达到散热效果，又可以减轻散热控制设备103的工作负担。基于此，每当散热控制周期到达时，可以获取至少一个待散热设备的实际功率信息；结合过热风险预测模型，周期性调整制冷系统102的制冷参数，以通过控制制冷系统102对机房内的至少一个待散热设备101进行动态散热。本实施例并不限定散热控制周期的时间长度，可以根据应用需求适应性设置。例如，散热控制周期的时间长度可以是1分钟、10分钟或15分钟等。
示例2：在该示例中，散热控制设备103可以实时监测机房内至少一个待散热设备101的总功率变化幅度，将至少一个待散热设备的总功率变化幅度作为散热控制条件。基于此，每当监测至少一个待散热设备101的总功率变 化幅度大于第一幅度阈值时，散热控制设备103获取至少一个待散热设备101的实际功率信息；结合过热风险预测模型，不断调整制冷系统102的制冷参数，以通过控制制冷系统102对机房内的至少一个待散热设备101进行动态散热。
示例3：在该示例中，散热控制设备103可以实时监测机房内每个待散热设备101的功率变化幅度，将每个待散热设备的功率变化幅度作为散热控制条件。基于此，每当监测到出现待散热设备的功率变化幅度大于第二幅度阈值的情况时，散热控制设备103获取至少一个待散热设备101的实际功率信息；结合过热风险预测模型，不断调整制冷系统102的制冷参数，以通过控制制冷系统102对机房内的至少一个待散热设备101进行动态散热。
在此说明，本实施例并不限定第一幅度阈值和第二幅度阈值的数值，可根据应用需求灵活设定。另外，对不同待散热设备101来说，第二幅度阈值可以相同，也可以不相同。例如，可以为每个待散热设备101分别设置其对应的第二幅度阈值。
在此说明，上述示例1-3中的散热控制条件，可以单独使用，也可以以任意组合方式组合使用，对此不做限定。
在本申请各实施例中，待散热设备101的实际功率信息主要用于反应待散热设备的功率消耗情况，是本次散热控制的数据基础。在本申请各实施例中，并不限定待散热设备101的实际功率信息的实现形式，可以是任何能够反映待散热设备101的功率消耗情况的数据形式。
例如，每当散热控制条件被触发时，可以分别采集至少一个待散热设备101在散热控制条件被触发时刻的功率值，作为至少一个待散热设备101的实际功率信息。
又例如，每当散热控制条件被触发时，分别获取至少一个待散热设备101在本次散热控制与上次散热控制期间的功率均值，作为至少一个待散热设备101的实际功率信息。
可选地，在上述两个可选实施例中，都用到了待散热设备101的功率值 这一信息。可选地，待散热设备101的功率值可以是待散热设备101的整体功率值。例如，待散热设备101的整体功率值可以定义为待散热设备101的主要内部组件的功率之和，也可以定义为待散热设备101的所有内部组件的功率之和。或者，待散热设备101的功率值可以是待散热设备101的某个内部组件的功率，例如可以是CPU的功率，或者内存的功率等。
在本实施例中，制冷系统102与散热控制设备103之间通信连接。制冷系统102与散热控制设备103之间可以是无线或有线连接。可选地，散热控制设备103可以通过移动网络与制冷系统102进行通信连接。其中，移动网络的网络制式可以为2G(GSM)、2.5G(GPRS)、3G(WCDMA、TD-SCDMA、CDMA2000、UTMS)、4G(LTE)、4G+(LTE+)、5G、WiMax或者未来即将出现的新网络制式等中的任意一种。可选地，散热控制设备103也可以通过蓝牙、WiFi、红外、zigbee或NFC等方式与制冷系统102进行通信连接。
对散热控制设备103来说，可基于其与制冷系统102之间的通信连接，根据目标制冷参数，控制制冷系统102对机房内的至少一个待散热设备101进行散热。
在一可选实施例中，散热控制设备103在确定目标制冷参数之后，可以基于其与制冷系统102之间的通信连接，直接将目标制冷参数发送给制冷系统102，以供制冷系统102根据目标制冷参数为机房内的至少一个待散热设备101进行散热。对制冷系统102来说，可接收散热控制设备103发送的目标制冷参数，将目标制冷参数与当前使用的制冷参数进行比较；若两者不同，将当前使用的制冷参数替换为目标制冷参数，并按照目标制冷参数继续进行制冷工作；若两者相同，则继续根据当前使用的制冷参数进行制冷工作，从而达到按照目标制冷参数对机房内至少一个待散热设备101进行散热的目的。
在另一可选实施例中，散热控制设备103可记录制冷系统102当前使用的制冷参数，在确定出目标制冷参数之后，将目标制冷参数与制冷系统102当前使用的制冷参数进行比较，并在目标制冷参数与制冷系统102当前使用的制冷参数不相同的情况下，基于其与制冷系统102之间的通信连接，将目 标制冷参数发送给制冷系统102，以供制冷系统102根据目标制冷参数继续进行制冷工作。对制冷系统102来说，可接收散热控制设备103发送的目标制冷参数，将当前使用的制冷参数调整为目标制冷参数，并按照目标制冷参数继续进行制冷工作；在未接收到散热控制设备103发送的目标制冷参数的情况下，可以继续根据当前使用的制冷参数进行制冷工作，从而达到按照目标制冷参数对机房内至少一个待散热设备101进行散热的目的。
值得说明的是，根据制冷系统102的类型和工作原理的不同，目标制冷参数也会有所不同。不论是哪种制冷系统，目标制冷参数都是可以影响制冷效果的相关参数。例如，对于空调系统，可向机房系统100输入冷却空气，冷却空气按照一定方向(例如自上而下或自下而上)流动，会带走机房系统100内部的热量，达到散热目的。可选地，本实施例的空调系统包括但不限于以下工作参数：工作温度、工作风速以及工作方式等，这些工作参数会影响空调系统的散热性能。空调系统的工作参数可作为本实施例的目标制冷参数，则对空调系统而言，目标制冷参数可以包括但不限于：空调系统的工作温度、工作风速以及工作方式中的至少一种。又例如，对于水冷系统，可通过管道等液体承载体向机房系统100提供液体冷却剂，液体冷却剂可以是冷水或液态金属钠等，液体冷却剂在机房系统100内流动或环绕机房系统100内的待散热设备流动，从而带走机房系统100内部的热量，达到散热目的。可选地，本实施例的水冷系统包括但不限于以下工作参数：出水温度、回水温度、水流速以及水流量等，这些工作参数会影响水冷系统的散热性能。水冷系统的工作参数可作为本实施例的目标制冷参数，则对水冷系统而言，目标制冷参数可以包括但不限于：水冷系统的出水温度、回水温度、水流速以及水流量中的至少一种。
在本申请各实施例中，不对过热风险预测模型的工作方式进行限定。过热风险预测模型的工作方式与其训练方式相对应，采用不同训练方式可以训练出具有不同工作方式的过热风险预测模型。下面对过热风险预测模型可能 的工作方式进行举例说明：
在一种可选方式中，可以将至少一个待散热设备的实际功率信息输入过热风险预测模型，该过热风险预测模型可自主确定出至少一个候选制冷参数，并可一次性输出机房系统100在各个候选制冷参数下发生过热风险的概率。
在另一种可选方式中，预先在模型外部确定至少一个候选制冷参数，将至少一个待散热设备的实际功率信息和至少一个候选制冷参数一并作为输入参数输入过热风险预测模型，该模型一次性输出机房系统100在各个候选制冷参数下发生过热风险的概率。
在又一种可选方式中，预先在模型外部确定至少一个候选制冷参数；针对每个候选制冷参数，将至少一个待散热设备的实际功率信息和该候选制冷参数输入过热风险预测模型，以得到机房系统100在该候选制冷参数下发生过热风险的概率。
其中，确定至少一个候选制冷参数的方式包括：根据人工经验，确定至少一个候选制冷参数；或者，对制冷系统102当前使用的制冷参数进行不同幅度的调整，得到至少一个候选制冷参数；或者，根据过热风险预测模型在训练阶段使用的制冷参数的范围，确定至少一个候选制冷参数。
进一步可选地，在根据过热风险预测模型在训练阶段使用的制冷参数的范围，确定至少一个候选制冷参数的可选方案中，可以在过热风险预测模型在训练阶段使用的制冷参数的范围之内确定至少一个候选制冷参数，即候选制冷参数位于过热风险预测模型在训练阶段使用的制冷参数的范围之内。例如，假设过热风险预测模型在训练阶段使用的制冷参数为19℃、20℃、22℃、26℃和28℃，则可以在19℃与28℃之间确定至少一个候选制冷参数。
当然，除上述方式之外，还可以将过热风险预测模型在训练阶段使用的制冷参数的范围作为基础参数范围，根据基础参数范围确定一个候选参数范围，在该候选参数范围内确定至少一个候选制冷参数。
在得到机房系统100在至少一个候选制冷参数下发生过热风险的概率之后，散热控制设备103可根据机房系统100在至少一个候选制冷参数下发生 过热风险的概率，确定目标制冷参数。其中，确定目标制冷参数可以采用但不限于下述可选实施方式：
在一可选实施方式中，可以将机房系统100在至少一个候选制冷参数下发生过热风险的概率进行比较，从中选择概率较小的候选制冷参数作为目标制冷参数。例如，可以选择最小概率对应的候选制冷参数作为目标制冷参数。
在另一可选实施方式中，可结合机房系统承载的应用或服务对发生过热风险的要求，预先确定机房系统对应的过热风险概率阈值。则，可基于该过热风险概率阈值，从机房系统在至少一个候选制冷参数下发生过热风险的概率中，选择小于该过热风险概率阈值的概率作为目标概率，将至少一个候选制冷参数中，与该目标概率对应的制冷参数作为目标制冷参数。
进一步可选地，在选择目标概率时，可以从机房系统在至少一个候选制冷参数下发生过热风险的概率中，随机选择一个小于过热风险概率阈值的概率作为目标概率；或者，可以从机房系统在至少一个候选制冷参数下发生过热风险的概率中，选择小于过热风险概率阈值的最大概率作为目标概率；或者，可以机房系统在至少一个候选制冷参数下发生过热风险的概率中，选择小于过热风险概率阈值且位于设定概率范围内的概率作为目标概率。
在一可选实施例中，在使用过热风险概率阈值之前，可以获取机房系统承载的应用或服务允许的热故障率，将该热故障率转换为机房系统对应的过热风险概率阈值。可选地，可以采用统计学的方法，将该热故障率转换为机房对应的过热风险概率阈值。
其中，热故障率是指机房系统100在一定时间内能够发生过热风险的最大次数。这里的最大次数是指在一定时间内各个发生过热风险的设备发生过热风险的次数之和。例如，假设热故障率表示机房系统100在一个月内最多能够发生15次过热风险，散热控制设备103每25分钟进行一次散热控制，则过热风险概率阈值＝15/(30*24*4)。
在本申请一可选实施例中，可以预先训练过热风险预测模型，为需要使 用过热风险预测模型的实施例提供基础。在该可选实施例中，可获取多组标记样本数据，基于多组标记样本数据，采用有监督的模型训练方法训练出过热风险预测模型。在本实施例中，鉴于深度神经网络(DNN)具有处理具有大量输入参数的复杂情况的能力，因此采用深度神经网络算法进行模型训练。
为了提高过热风险预测模型的判别能力，需要积累机房系统在极端运行情况下的样本数据，但是考虑到机房系统的安全性，无法通过实际改变机房系统的热负载和电负载使之真正运行在比较极端的环境下。为此，在本实施例中，将基于真实数据的样本生成方式与基于计算流体力学(Computational Fluid Dynamics，CFD)模拟的样本生成方式相结合，通过CFD模拟计算的方法，为模型训练提供所需的样本数据，可以弥补基于真实数据的样本生成方式的不足，有利于获取足够的样本数据，有效提高热风险预测模型的鲁棒性。
基于上述，如图2a所示，一种模型训练的过程包括：结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；之后，利用多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型。其中，每组标记样本数据包括与至少一个待散热设备对应的至少一个样本功率信息、与制冷系统对应的样本制冷参数以及在该组标记样本数据下机房是否发生过热风险的标记结果。
可选地，如图2a所示，在利用多组标记样本数据进行深度神经网络模型训练之前，可以对多组样本数据进行数据清洗，以提高样本数据的可靠性。可选地，针对基于真实数据的样本生成方式生成的标记样本数据，与基于CFD模拟的样本生成方式生成的标记样本数据，可以采用不同的数据清洗方式进行数据清洗。
可选地，结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据，包括：根据至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据；以及利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据。
进一步，生成至少一组带标记的历史样本数据的过程包括：获取至少一组未标记的历史样本数据，每组未标记的历史样本数据包括同一历史时刻或历史时段内至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数；针对每组未标记的历史样本数据，根据至少一个待散热设备内部组件在相应历史时刻或历史时段内的温度与内部组件对应的过热温度阈值进行机房是否发生过热风险的标记，得到至少一组带标记的历史样本数据。
进一步，生成至少一组带标记的模拟样本数据的过程包括：设计至少一组未标记的模拟样本数据，每组未标记的模拟样本数据包括与至少一个待散热设备对应的至少一个模拟功率信息以及与制冷系统对应的模拟制冷参数；针对每组未标记的模拟样本数据，利用CFD模型对该组未标记的模拟样本数据进行模拟以得到至少一个待散热设备内部器件的温度，利用至少一个待散热设备内部器件的温度与内部组件对应的过热温度阈值进行机房是否发生过热风险的标记，得到至少一组带标记的模拟样本数据。
进一步可选地，如图2a所示，在利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据之前，还可以利用至少一组带标记的历史样本数据，对CFD模型进行参数校正。之后，利用校正后的CFD模型生成至少一组带标记的模拟样本数据，有利于提高带标记的模拟样本数据的可靠性和真实性，有利于提高据此训练出的过热风险预测模型的准确度。
除上述生成过热风险预测模型的实施例之外，在本申请另一可选实施例中，除了依据待散热设备的功率信息和制冷系统的制冷参数之外，还可以结合外部大气温度。外部大气温度是指机房系统100外部的大气温度。在该可选实施例中，每组标记样本数据包括：至少一个样本功率信息、样本制冷参数和样本大气温度。相应地，在生成多组标记样本数据的过程中，还包括获取对应时刻的外部大气温度作为样本大气文本并在样本数据中添加样本大气温度的操作。以第一组标记样本数据为例，在生成第一组标记样本数据的过程中，还获取第一组标记样本数据对应时刻的外部大气温度作为样本大气温 度，将第一组标记样本数据对应的样本大气温度加入第一组标记样本数据中；其中，第一组标记样本数据是多组标记样本数据中任一组标记样本数据。
在获取多组包含样本大气温度的标记样本数据之后，可根据多组包含样本大气温度的标记样本数据进行深度神经网络模型训练，得到过热风险预测模型。基于这些包含样本大气温度这一参数的标记样本数据训练出的过热风险预测模型，可以针对给定的一组功率信息、外部大气温度和一个候选制冷参数，预测出机房在该候选制冷参数下发生过热风险的概率。
可选地，可以通过互联网获取一些网站或APP发布的外部大气温度。或者，也可以在机房外部设置温度采集设备，例如温度传感器等，通过温度采集设备采集外部大气温度。对于获取外部大气温度的方式，本申请实施例不做限定。
基于上述，每当散热控制条件被触发时，散热控制设备103除了获取机房内至少一个待散热设备的实际功率信息之外，还需要获取散热控制条件被触发时对应的外部大气温度；进而，在根据过热风险预测模型得到机房系统100在至少一个候选制冷参数下发生过热风险的概率的过程中，将至少一个待散热设备的实际功率信息以及散热控制条件被触发时对应的外部大气温度作为输入参数输入过热风险预测模型，以得到机房系统在至少一个候选制冷参数下发生过热风险的概率。进而，根据机房系统在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据目标制冷参数，控制制冷系统对机房内的至少一个待散热设备进行散热。
在一实施例中，假设一种过热风险预测模型的结构如图2b所示。图2b中所呈现的热风险预测模型的结构只是示例性说明，并不对此进行限定。如图2b所示，该模型支持的输入数据包括：设备1-设备n的实际功率信息、外部大气温度以及一个候选制冷参数(如工作温度)，输出为机房在该候选制冷参数下发生过热风险的概率。其中，n是正整数。在有多个候选制冷参数的情况下，通过图2b所示热风险预测模型可以得到机房在该候选制冷参数下发生过热风险的概率。以制冷系统为空调系统为例，则如图2c所示，假设候选 制冷参数包括：20℃、21℃、22℃和23℃四个制冷温度，在经过图2b所示热风险预测模型后，得到4个概率值分别为0.76、0.83、0.89和0.92。若假设机房对应的过热风险概率阈值为0.85，则可以选择概率值0.83对应的温度21℃作为目标制冷温度；进而可控制空调系统将制冷温度调整为21℃，以对机房进行散热。
值得说明的是，为了保证过热风险预测模型的准确度，还可以对过热风险预测模型进行更新。例如，在更新触发条件被触发时，可以对过热风险预测模型进行更新。其中，对热风险预测模型进行更新包括但不限于以下几种情况：每当模型更新周期到达时，重新对过热风险预测模型进行模型训练；每当机房内待散热设备的数量变化时，重新对过热风险预测模型进行模型训练；每当机房内待散热设备之间的拓扑结构发生变化时，重新对过热风险预测模型进行模型训练。
在对过热风险预测模型进行更新之后，在后续散热控制过程中可以使用更新后的过热风险预测模型，有利于提高散热控制的准确度和精度。值得说明的是，上述实施例提供的模型训练过程不仅适用于本申请上述描述机房系统的实施例，同样适用于本申请下述描述数据中心系统的实施例，在下述实施例中不再赘述。
本申请实施例提供的散热控制原理不仅适用于独立的机房系统，也适用于包括一个或多个机房的数据中心系统。下面以图3和图4所示两种结构的数据中心系统为例对本申请实施例的散热控制原理进行示例性说明。
图3为本申请示例性实施例提供的一种数据中心系统的结构示意图。如图3所示，该数据中心系统300包括：至少一个机房301；每个机房301包括：至少一个待散热设备、制冷系统以及散热控制设备。
在本实施例中，从制冷系统以及散热控制的角度来看，每个机房301相对独立，分别具有自己的制冷系统和散热控制设备，可独立进行散热控制。对任意一个机房301来说，其包含的散热控制设备可在过热风险预测模型的 基础上，结合该机房301内至少一个待散热设备的功率变化情况，控制制冷系统对机房301进行动态散热，制冷系统可根据实际需要进行制冷工作，可降低制冷系统的能耗，节约电能资源。
具体地，可以设置散热控制条件，每当散热控制条件被触发时，散热控制设备获取其所属机房301内至少一个待散热设备的实际功率信息，将至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到机房301在至少一个候选制冷参数下发生过热风险的概率；根据机房301在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据目标制冷参数，控制其所属机房301内的制冷系统对其所属机房301内至少一个待散热设备进行散热。对制冷系统来说，可在其所属机房301包含的散热控制设备的控制下，对其所属机房301内的至少一个待散热设备进行散热。
本实施例中的机房301与前述实施例中的机房系统100相同或相似，关于该机房301的详细描述以及任一机房301中的散热控制设备在过热风险预测模型的基础上，结合该机房301内至少一个待散热设备的功率变化情况，控制制冷系统对机房301进行动态散热的详细实现或过程，均可参见前述实施例，在此不再赘述。
图4为本申请示例性实施例提供的另一种数据中心系统的结构示意图。如图4所示，该数据中心系统400包括：至少一个机房401、制冷系统以及散热控制设备403。其中，每个机房401包括至少一个待散热设备；制冷系统包括部署在每个机房401内的制冷设备402。
与图3所示数据中心系统300不同，在数据中心系统400中，不同机房401共享制冷系统和散热控制设备403。散热控制设备403需对数据中心系统400中每个机房401进行散热控制；同理，制冷系统需为数据中心系统400中每个机房401进行散热。在本实施例中，为了实现给数据中心系统400中每个机房401进行散热的目的，制冷系统包括部署在每个机房401内的制冷设备402，这样散热控制设备403可以通过每个机房401内的制冷设备402对每个机房401进行散热控制，控制逻辑相对简单、方便、易于实施。
其中，散热控制设备403对每个机房401进行散热控制的过程相同或相似，下面对散热控制设备403对每个机房401进行散热控制的过程进行说明：
针对每个机房401，散热控制设备403用于：每当散热控制条件被触发时，获取该机房401内至少一个待散热设备的实际功率信息，将该机房内401至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到该机房401在至少一个候选制冷参数下发生过热风险的概率；根据该机房401在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据该目标制冷参数，控制该机房401内的制冷设备402对该机房401内的至少一个待散热设备进行散热。
在此说明，不同机房401可以使用相同的散热控制条件，也可以使用不同的散热控制条件，可根据各机房401的具体情况灵活设定。
值得说明的是，虽然散热控制设备403可以对数据中心系统400中各个机房401进行散热控制，但是散热控制设备403对其中一个机房401进行散热控制的过程，与前述实施例中散热控制设备103对机房系统100进行散热控制的过程相同或相似，关于散热控制设备403对每个机房401进行散热控制的过程的详细描述，可参见前述实施例，在此不再赘述。
在此说明，在本申请各实施例中，并不限定散热控制设备的设备形态。散热控制设备可以是任何具有一定计算和通信能力，并能进行数据处理的计算机设备，例如可以是常规服务器、云服务器、云主机、虚拟中心或服务器阵列等服务端设备，也可以是智能手机、平板电脑、个人电脑或一体机等终端设备。
另外，在本申请上述实施例中，以机房和数据中心为例，对本申请实施例提供的散热控制原理进行了示例性说明。显然，机房和数据中心仅是本申请实施例给出的两种示例性的应用场景，并不能构成对本申请保护范围的限定。本申请实施例提供的散热控制原理可以应用到任何容纳有待散热设备的物理空间，换句话说，凡是容纳有待散热设备的物理空间，都可以采用本申请实施例提供的散热控制原理进行散热控制。
例如，在一些办公楼宇中，安装或配置有各种电器设备，例如办公用的电脑、服务器、监控设备、打印机、传真机、复印机、照明灯等等。当然，办公楼宇中也会安装或配置空调系统、排风系统或暖气等基础设施，用以调整办公楼宇中的环境温度。为了避免整个办公楼宇发生过热风险，可以采用本申请实施例提供的散热控制方法，监控整个办公楼宇内待散热设备的实际功率信息；将整个办公楼宇内待散热设备的实际功率信息输入过热风险预测模型，以得到整个办公楼宇在至少一个候选制冷参数下发生过热风险的概率；进而根据整个办公楼宇在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据目标制冷参数，控制制冷系统对整个办公楼宇内的待散热设备进行散热。办公楼宇内的待散热设备包括但不限于：办公用的电脑、服务器、监控设备、打印机、传真机、复印机等，以及照明灯、饮水机、微波炉、吸尘器等其它电器设备。
又例如，在互联网时代的公司环境中，办公区域内会安装或配置有大量办公用的电脑，有些公司还会在办公区域内配置公司服务器、打印机、复印机、传真机等，这些电脑、服务器、打印机、复印机、传真机等在工作时都会产生热量。为了避免办公区域因为环境过热致使这些设备故障、失灵，甚至被烧毁，可以采用本申请实施例提供的散热控制方法，监控办公区域内待散热设备的实际功率信息；将办公区域内待散热设备的实际功率信息输入过热风险预测模型，以得到办公区域在至少一个候选制冷参数下发生过热风险的概率；进而根据办公区域在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据目标制冷参数，控制制冷系统对办公区域内的待散热设备进行散热。办公楼宇内的待散热设备包括但不限于：办公用的电脑、服务器、打印机、传真机、复印机等。
又例如，随着智能家居技术的发展，越来越多的智能家居设备出现在家庭环境中，在家庭环境中的指定区域内可能存在大量智能家居设备。例如，在家庭环境中的客厅区域中，可能存在电视机、智能音箱、大型游戏机、智能空调、净化器、家庭陪护机器人、个人电脑等电器设备。又例如，在家庭 环境中的厨房区域中，可能存在智能微波炉、烤箱、冰箱、电饭煲、榨汁机等电气设备。这些区域内的设备在工作时也会产生大量热量，为了避免这些区域内的设备因环境过热导致发生故障、失灵，甚至被烧毁，也可以采用本申请实施例提供的散热控制方法，监控家庭环境中指定区域内待散热设备的实际功率信息；将家庭环境中指定区域内待散热设备的实际功率信息输入过热风险预测模型，以得到家庭环境中指定区域在至少一个候选制冷参数下发生过热风险的概率；进而根据家庭环境中指定区域在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据目标制冷参数，控制制冷系统对家庭环境中指定区域内的待散热设备进行散热。根据指定区域的不同，指定区域内的待散热设备也会有所不同。以家庭环境中的客厅区域为例，其包含的待散热设备包括但不限于：电视机、智能音箱、大型游戏机、智能空调、净化器、家庭陪护机器人、个人电脑等。
图5所示为本申请示例性实施例提供的一种散热控制方法的流程示意图。该方法是从散热控制设备的角度进行的描述，如图5所示，该方法包括：
501、每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息。
502、将至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到指定空间在至少一个候选制冷参数下发生过热风险的概率。
503、根据指定空间在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数。
504、根据目标制冷参数，控制制冷系统对指定空间内的至少一个待散热设备进行散热。
其中，指定空间是指任何能够容纳且容纳有待散热设备的物理空间，例如可以是但不限于：机房、数据中心、办公楼宇、公司环境中的办公区域、或者家庭环境中的指定区域等。其中，家庭环境中的指定区域中存储有会产生热量且需要散热的待散热设备，例如电视机、智能音箱、大型游戏机、智能空调、净化器、家庭陪护机器人、个人电脑等。
在本实施例中，指定空间内包含至少一个待散热设备。待散热设备可产生热量，使指定空间内的温度升高。待散热设备对指定空间内的温度有一定要求，如果指定空间内的温度过高，待散热设备可能会发生故障、失灵，甚至被烧毁。为了给待散热设备提供良好的工作环境，可提供制冷系统。可选地，制冷系统可部署于指定空间内，但不限于此。制冷系统主要负责带走指定空间内的热量，为指定空间内的待散热设备进行散热。在本实施例中，并不限定制冷系统的类型和工作原理，例如可以是空调系统，或者是水冷系统，或者是空调系统与水冷系统的组合。
在本实施例中，通过模型训练，预先得到过热风险预测模型，该模型可体现设备功率信息与制冷参数之间存在的过热风险关系。基于此，散热控制设备可在该过热风险预测模型的基础上，结合指定空间内至少一个待散热设备的功率变化情况，控制制冷系统对指定空间进行动态散热，制冷系统可根据实际需要进行制冷工作，可降低制冷系统的能耗，节约电能资源。其中，待散热设备的功率信息反应待散热设备的功耗，也可反映待散热设备的工作负载。其中，关于模型训练过程可参见后续实施例，在此不再赘述。
在本实施例中，可以设置散热控制条件，每当散热控制条件被触发时，散热控制设备就在过热风险预测模型的基础上，结合指定空间内至少一个待散热设备的实际功率信息进行一次散热控制。
其中，指定空间在某个候选制冷参数下发生过热风险的概率，主要是指假定制冷系统采用该候选制冷参数，则指定空间中出现待散热设备发生过热风险情况的概率。其中，发生过热风险的待散热设备的数量可以是一个，也可以是多个，对此不做限定。
值得说明的是，本实施例不对散热控制条件进行限定，可根据散热控制需求灵活设定。下面对散热控制条件进行举例说明：
例如，在一种示例中，可以预先设定散热控制周期，将散热控制周期作为散热控制条件。基于此，步骤501的一种实施方式为：每当散热控制周期到达时，获取至少一个待散热设备的实际功率信息。本实施例并不限定散热 控制周期的时间长度，可以根据应用需求适应性设置。例如，散热控制周期的时间长度可以是1分钟、10分钟或15分钟等。
又例如，在一种示例中，散热控制设备可以实时监测指定空间内至少一个待散热设备的总功率变化幅度，将至少一个待散热设备的总功率变化幅度作为散热控制条件。基于此，步骤501的一种实施方式为：每当监测至少一个待散热设备的总功率变化幅度大于第一幅度阈值时，获取至少一个待散热设备的实际功率信息。
再例如，在一种示例中，散热控制设备可以实时监测指定空间内每个待散热设备的功率变化幅度，将每个待散热设备的功率变化幅度作为散热控制条件。基于此，步骤501的一种实施方式为：每当监测到出现待散热设备的功率变化幅度大于第二幅度阈值的情况时，获取至少一个待散热设备101的实际功率信息。
在此说明，本实施例并不限定第一幅度阈值和第二幅度阈值的数值，可根据应用需求灵活设定。另外，对不同待散热设备来说，第二幅度阈值可以相同，也可以不相同。例如，可以为每个待散热设备分别设置其对应的第二幅度阈值。
在此说明，上述几种示例中的散热控制条件，可以单独使用，也可以以任意组合方式组合使用，对此不做限定。
在本实施例中，待散热设备的实际功率信息主要用于反应待散热设备的功率消耗情况，是本次散热控制的数据基础。在本实施例中，并不限定待散热设备的实际功率信息的实现形式，可以是任何能够反映待散热设备的功率消耗情况的数据形式。例如，每当散热控制条件被触发时，可以分别采集至少一个待散热设备在散热控制条件被触发时刻的功率值，作为至少一个待散热设备的实际功率信息。又例如，每当散热控制条件被触发时，分别获取至少一个待散热设备在本次散热控制与上次散热控制期间的功率均值，作为至少一个待散热设备的实际功率信息。
可选地，在上述两个可选实施例中，都用到了待散热设备的功率值这一 信息。可选地，待散热设备的功率值可以是待散热设备的整体功率值。例如，待散热设备的整体功率值可以定义为待散热设备的主要内部组件的功率之和，也可以定义为待散热设备的所有内部组件的功率之和。或者，待散热设备的功率值可以是待散热设备的某个内部组件的功率，例如可以是CPU的功率，或者内存的功率等。
在一可选实施例中，步骤504的一种实施方式包括：将目标制冷参数发送给制冷系统，以供制冷系统根据目标制冷参数为指定空间内的至少一个待散热设备进行散热。对制冷系统来说，可接收散热控制设备发送的目标制冷参数，将目标制冷参数与当前使用的制冷参数进行比较；若两者不同，将当前使用的制冷参数替换为目标制冷参数，并按照目标制冷参数继续进行制冷工作；若两者相同，则继续根据当前使用的制冷参数进行制冷工作，从而达到按照目标制冷参数对指定空间内至少一个待散热设备进行散热的目的。
在另一可选实施例中，步骤504的一种实施方式包括：将目标制冷参数与制冷系统当前使用的制冷参数进行比较，并在目标制冷参数与制冷系统当前使用的制冷参数不相同的情况下，将目标制冷参数发送给制冷系统，以供制冷系统根据目标制冷参数继续进行制冷工作。对制冷系统来说，可接收散热控制设备发送的目标制冷参数，将当前使用的制冷参数调整为目标制冷参数，并按照目标制冷参数继续进行制冷工作；在未接收到散热控制设备发送的目标制冷参数的情况下，可以继续根据当前使用的制冷参数进行制冷工作，从而达到按照目标制冷参数对指定空间内至少一个待散热设备进行散热的目的。
在一可选实施例中，步骤502的一种实施方式包括：根据过热风险预测模型在训练阶段使用的制冷参数的范围，确定至少一个候选制冷参数；针对每个候选制冷参数，将至少一个待散热设备的实际功率信息和该候选制冷参数输入过热风险预测模型，以得到指定空间在该候选制冷参数下发生过热风险的概率。
在一可选实施例中，步骤503的一种实施方式包括：基于指定空间对应 的过热风险概率阈值，从指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于该过热风险概率阈值的概率作为目标概率，将至少一个候选制冷参数中，与该目标概率对应的制冷参数作为目标制冷参数。
进一步可选地，在选择目标概率时，可以从指定空间在至少一个候选制冷参数下发生过热风险的概率中，随机选择一个小于过热风险概率阈值的概率作为目标概率；或者，可以从指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于过热风险概率阈值的最大概率作为目标概率；或者，可以指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于过热风险概率阈值且位于设定概率范围内的概率作为目标概率。
进一步可选地，在使用过热风险概率阈值之前，可以获取指定空间承载的应用或服务允许的热故障率，将该热故障率转换为指定空间对应的过热风险概率阈值。可选地，可以采用统计学的方法，将该热故障率转换为指定空间对应的过热风险概率阈值。其中，热故障率是指指定空间在一定时间内能够发生过热风险的最大次数。这里的最大次数是指在一定时间内各个发生过热风险的设备发生过热风险的次数之和。
在一可选实施例中，可以采用但不限于下述方式，预先训练得到过热风险预测模型：
结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；之后，利用多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型。其中，每组标记样本数据包括与至少一个待散热设备对应的至少一个样本功率信息、与制冷系统对应的样本制冷参数以及在该组标记样本数据下指定空间是否发生过热风险的标记结果。基于这些标记样本数据训练出的过热风险预测模型，可以针对给定的一组功率信息与一个候选制冷参数，预测出指定空间在该候选制冷参数下发生过热风险的概率。
可选地，结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据，包括：根据至少一个待散热设备的历史功率 信息和制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据；以及利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据。
进一步，生成至少一组带标记的历史样本数据的过程包括：获取至少一组未标记的历史样本数据，每组未标记的历史样本数据包括同一历史时刻或历史时段内至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数；针对每组未标记的历史样本数据，根据至少一个待散热设备内部组件在相应历史时刻或历史时段内的温度与内部组件对应的过热温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的历史样本数据。
进一步，生成至少一组带标记的模拟样本数据的过程包括：设计至少一组未标记的模拟样本数据，每组未标记的模拟样本数据包括与至少一个待散热设备对应的至少一个模拟功率信息以及与制冷系统对应的模拟制冷参数；针对每组未标记的模拟样本数据，利用CFD模型对该组未标记的模拟样本数据进行模拟以得到至少一个待散热设备内部器件的温度，利用至少一个待散热设备内部器件的温度与内部组件对应的过热温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的模拟样本数据。
进一步可选地，在利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据之前，还可以利用至少一组带标记的历史样本数据，对CFD模型进行参数校正。之后，利用校正后的CFD模型生成至少一组带标记的模拟样本数据，有利于提高带标记的模拟样本数据的可靠性和真实性，有利于提高据此训练出的过热风险预测模型的准确度。
在上述可选实施例中，将基于真实数据的样本生成方式与基于CFD模拟的样本生成方式相结合，通过CFD模拟计算的方法，为模型训练提供所需的样本数据，可以弥补基于真实数据的样本生成方式的不足，既有利于获取足够的样本数据，又可以获取指定空间运行在比较极端情况下的样本数据，可有效提高热风险预测模型的鲁棒性。
除上述生成过热风险预测模型的实施例之外，在本申请另一可选实施例 中，除了依据待散热设备的功率信息和制冷系统的制冷参数之外，还可以结合外部大气温度。外部大气温度是指指定空间外部的大气温度。在该可选实施例中，每组标记样本数据包括：至少一个样本功率信息、样本制冷参数和样本大气温度。相应地，在生成多组标记样本数据的过程中，还包括获取对应时刻的外部大气温度作为样本大气温度并将样本大气温度添加至标记样本数据中的操作。以第一组标记样本数据为例，在生成第一组标记样本数据的过程中，还获取第一组标记样本数据对应时刻的外部大气温度作为样本大气温度，将第一组标记样本数据对应的样本大气温度加入第一组标记样本数据中；其中，第一组标记样本数据是多组标记样本数据中任一组标记样本数据。在获取多组包含样本大气温度的标记样本数据之后，可根据多组包含样本大气温度的标记样本数据进行模型训练，得到过热风险预测模型。基于这些包含样本大气温度这一参数的标记样本数据训练出的过热风险预测模型，可以针对给定的一组功率信息、外部大气温度和一个候选制冷参数，预测出指定空间在该候选制冷参数下发生过热风险的概率。
基于上述，在步骤501中，每当散热控制条件被触发时，散热控制设备还需要获取散热控制条件被触发时对应的外部大气温度。相应地，在步骤502中，可将至少一个待散热设备的实际功率信息以及散热控制条件被触发时对应的外部大气温度作为输入参数输入过热风险预测模型，以得到指定空间系统在至少一个候选制冷参数下发生过热风险的概率。
进一步，为了保证过热风险预测模型的准确度，还可以对过热风险预测模型进行更新。例如，在更新触发条件被触发时，可以对过热风险预测模型进行更新。其中，对热风险预测模型进行更新包括但不限于以下几种情况：每当模型更新周期到达时，重新对过热风险预测模型进行模型训练；每当指定空间内待散热设备的数量变化时，重新对过热风险预测模型进行模型训练；每当指定空间内待散热设备之间的拓扑结构发生变化时，重新对过热风险预测模型进行模型训练。
在对过热风险预测模型进行更新之后，在后续散热控制过程中可以使用 更新后的过热风险预测模型，有利于提高散热控制的准确度和精度。
图6为本申请示例性实施例提供的一种模型训练方法的流程示意图。如图6所示，该方法包括：
601、结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据。
602、利用多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型；其中，每组标记样本数据包括与至少一个待散热设备对应的至少一个样本功率信息、与制冷系统对应的样本制冷参数以及在该组标记样本数据下指定空间是否发生过热风险的标记结果。
可选地，步骤601的一种实施方式包括：根据至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据；以及利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据。
进一步，生成至少一组带标记的历史样本数据的过程包括：获取至少一组未标记的历史样本数据，每组未标记的历史样本数据包括同一历史时刻或历史时段内至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数；针对每组未标记的历史样本数据，根据至少一个待散热设备内部组件在相应历史时刻或历史时段内的温度与内部组件对应的过热温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的历史样本数据。
进一步，生成至少一组带标记的模拟样本数据的过程包括：设计至少一组未标记的模拟样本数据，每组未标记的模拟样本数据包括与至少一个待散热设备对应的至少一个模拟功率信息以及与制冷系统对应的模拟制冷参数；针对每组未标记的模拟样本数据，利用CFD模型对该组未标记的模拟样本数据进行模拟以得到至少一个待散热设备内部器件的温度，利用至少一个待散热设备内部器件的温度与内部组件对应的过热温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的模拟样本数据。
进一步可选地，在利用CFD模型进行功率信息与制冷参数之间的模拟计 算，生成至少一组带标记的模拟样本数据之前，还可以利用至少一组带标记的历史样本数据，对CFD模型进行参数校正。之后，利用校正后的CFD模型生成至少一组带标记的模拟样本数据，有利于提高带标记的模拟样本数据的可靠性和真实性，有利于提高据此训练出的过热风险预测模型的准确度。
在一可选实施例中，在进行模型训练过程中，除了依据待散热设备的功率信息和制冷系统的制冷参数之外，还可以结合外部大气温度。外部大气温度是指指定空间外部的大气温度。在该可选实施例中，每组标记样本数据包括：至少一个样本功率信息、样本制冷参数和样本大气温度。相应地，在生成多组标记样本数据的过程中，还包括获取对应时刻的外部大气温度作为样本大气温度并添加至对应标记样本数据中的操作。以第一组标记样本数据为例，在生成第一组标记样本数据的过程中，还获取第一组标记样本数据对应时刻的外部大气温度作为样本大气温度，将第一组标记样本数据对应的样本大气温度加入第一组标记样本数据中；其中，第一组标记样本数据是多组标记样本数据中任一组标记样本数据。在获取多组包含样本大气温度的标记样本数据之后，可根据多组包含样本大气温度的标记样本数据进行模型训练，得到过热风险预测模型。基于这些包含样本大气温度这一参数的标记样本数据训练出的过热风险预测模型，可以针对给定的一组功率信息、外部大气温度和一个候选制冷参数，预测出指定空间在该候选制冷参数下发生过热风险的概率。
进一步，为了保证过热风险预测模型的准确度，还可以对过热风险预测模型进行更新。例如，在更新触发条件被触发时，可以对过热风险预测模型进行更新。其中，对热风险预测模型进行更新包括但不限于以下几种情况：每当模型更新周期到达时，重新对过热风险预测模型进行模型训练；每当指定空间内待散热设备的数量变化时，重新对过热风险预测模型进行模型训练；每当指定空间内待散热设备之间的拓扑结构发生变化时，重新对过热风险预测模型进行模型训练。
在本实施例中，将基于真实数据的样本生成方式与基于CFD模拟的样本 生成方式相结合，通过CFD模拟计算的方法，为模型训练提供所需的样本数据，可以弥补基于真实数据的样本生成方式的不足，既有利于获取足够的样本数据，又可以获取指定空间运行在比较极端情况下的样本数据，可有效提高热风险预测模型的鲁棒性。
需要说明的是，上述实施例所提供方法的各步骤的执行主体均可以是同一设备，或者，该方法也由不同设备作为执行主体。比如，步骤501至步骤504的执行主体可以为设备A；又比如，步骤501和502的执行主体可以为设备A，步骤503和504的执行主体可以为设备B；等等。
另外，在上述实施例及附图中的描述的一些流程中，包含了按照特定顺序出现的多个操作，但是应该清楚了解，这些操作可以不按照其在本文中出现的顺序来执行或并行执行，操作的序号如501、502等，仅仅是用于区分开各个不同的操作，序号本身不代表任何的执行顺序。另外，这些流程可以包括更多或更少的操作，并且这些操作可以按顺序执行或并行执行。需要说明的是，本文中的“第一”、“第二”等描述，是用于区分不同的消息、设备、模块等，不代表先后顺序，也不限定“第一”和“第二”是不同的类型。
图7为本申请示例性实施例提供的一种散热控制设备的结构示意图。如图7所示，该设备包括：存储器71和处理器72。
存储器71，用于存储计算机程序，并可被配置为存储其它各种数据以支持在散热控制设备上的操作。这些数据的示例包括用于在散热控制设备上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。
处理器72，与存储器71耦合，用于执行存储器71中的计算机程序，以用于：每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息；将至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到指定空间在至少一个候选制冷参数下发生过热风险的概率；根据指定空间在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据目标制冷参数，控制制冷系统对指定空间内的至少一个待散热设备 进行散热。
其中，指定空间是指任何能够容纳且容纳有待散热设备的物理空间，例如可以是但不限于：机房、数据中心、办公楼宇、公司环境中的办公区域、或者家庭环境中的指定区域等。其中，家庭环境中的指定区域中存储有会产生热量且需要散热的待散热设备，例如电视机、智能音箱、大型游戏机、智能空调、净化器、家庭陪护机器人、个人电脑等。为了给待散热设备提供良好的工作环境，可提供制冷系统。可选地，制冷系统可部署于指定空间内，但不限于此。制冷系统主要负责带走指定空间内的热量，为指定空间内的待散热设备进行散热。
在一可选实施例中，处理器72在获取指定空间内至少一个待散热设备的实际功率信息获取时，具体用于执行以下至少一种操作：
每当散热控制周期到达时，获取至少一个待散热设备的实际功率信息；
每当监测至少一个待散热设备的总功率变化幅度大于第一幅度阈值时，获取至少一个待散热设备的实际功率信息；
每当监测到出现待散热设备的功率变化幅度大于第二幅度阈值的情况时，获取至少一个待散热设备的实际功率信息。
在一可选实施例中，处理器72在获取指定空间内至少一个待散热设备的实际功率信息获取时，具体用于：
每当散热控制条件被触发时，分别采集至少一个待散热设备在散热控制条件被触发时刻的功率值，作为至少一个待散热设备的实际功率信息；或者
每当散热控制条件被触发时，分别获取至少一个待散热设备在本次散热控制与上次散热控制期间的功率均值，作为至少一个待散热设备的实际功率信息。
在一可选实施例中，处理器72在得到指定空间在至少一个候选制冷参数下发生过热风险的概率时，具体用于：根据过热风险预测模型在训练阶段使用的制冷参数的范围，确定至少一个候选制冷参数；针对每个候选制冷参数，将至少一个待散热设备的实际功率信息和候选制冷参数输入过热风险预测模 型，以得到指定空间在候选制冷参数下发生过热风险的概率。
在一可选实施例中，处理器72在确定目标制冷参数时，具体用于：从指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于指定空间对应的过热风险概率阈值的目标概率；将至少一个候选制冷参数中，与目标概率对应的制冷参数作为目标制冷参数。
进一步可选地，处理器72在选择目标概率时，具体用于：从指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于过热风险概率阈值的最大概率作为目标概率。
进一步可选地，处理器72还用于：在选择小于指定空间对应的过热风险概率阈值的目标概率之前，将指定空间承载的应用或服务允许的热故障率转换为指定空间对应的过热风险概率阈值；热故障率表示指定空间在一定时间内能够发生过热风险的最大次数。
在一可选实施例中，处理器72还用于：结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；利用多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型。其中，每组标记样本数据包括与至少一个待散热设备对应的至少一个样本功率信息、与制冷系统对应的样本制冷参数以及在该组标记样本数据下指定空间是否发生过热风险的标记结果。
可选地，处理器72在生成多组标记样本数据时，具体用于：根据至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据；以及利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据。
进一步可选地，处理器72在生成至少一组带标记的历史样本数据时，具体用于：获取至少一组未标记的历史样本数据，每组未标记的历史样本数据包括同一历史时刻或历史时段内至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数；针对每组未标记的历史样本数据，根据至少一个待散热设备内部组件在相应历史时刻或历史时段内的温度与内部组件对应的过热 温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的历史样本数据。
进一步可选地，处理器72在生成至少一组带标记的模拟样本数据时，具体用于：设计至少一组未标记的模拟样本数据，每组未标记的模拟样本数据包括与至少一个待散热设备对应的至少一个模拟功率信息以及与制冷系统对应的模拟制冷参数；针对每组未标记的模拟样本数据，利用CFD模型对该组未标记的模拟样本数据进行模拟以得到至少一个待散热设备内部器件的温度，利用至少一个待散热设备内部器件的温度与内部组件对应的过热温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的模拟样本数据。
进一步可选地，处理器72还用于：在生成至少一组带标记的模拟样本数据之前，利用至少一组带标记的历史样本数据，对CFD模型进行参数校正。
在一可选实施例中，处理器72还用于：在生成多组标记样本数据的过程中，针对第一组标记样本数据，获取第一组标记样本数据对应时刻的外部大气温度作为样本大气温度，并将样本大气温度加入第一组标记样本数据中；其中，第一组标记样本数据是多组标记样本数据中的任一组。
基于上述，每当散热控制条件被触发时，处理器72还用于获取散热控制条件被触发时对应的外部大气温度。相应地，处理器72在得到指定空间在至少一个候选制冷参数下发生过热风险的概率时，具体用于：将至少一个待散热设备的实际功率信息以及散热控制条件被触发时对应的外部大气温度作为输入参数输入过热风险预测模型，以得到指定空间在至少一个候选制冷参数下发生过热风险的概率。
在一可选实施例中，处理器72还用于执行以下至少一种操作：
每当模型更新周期到达时，重新对过热风险预测模型进行模型训练；
每当指定空间内待散热设备的数量变化时，重新对过热风险预测模型进行模型训练；
每当指定空间内待散热设备之间的拓扑结构发生变化时，重新对过热风 险预测模型进行模型训练。
在一可选实施例中，处理器72在根据目标制冷参数控制制冷系统对指定空间内的至少一个待散热设备进行散热时，具体用于：
将目标制冷参数发送给制冷系统，以供制冷系统根据目标制冷参数进行制冷工作；或者
在目标制冷参数与制冷系统当前使用的制冷参数不同时，将目标制冷参数发送给制冷系统，以供制冷系统根据目标制冷参数进行制冷工作。
进一步，如图7所示，该散热控制设备还包括：通信组件73、显示器74、电源组件75、音频组件76等其它组件。图7中仅示意性给出部分组件，并不意味着散热控制设备只包括图7所示组件。另外，根据散热控制设备的实现形态的不同，图7中虚线框内的组件为可选组件，而非必选组件。例如，当散热控制设备实现为智能手机、平板电脑或台式电脑等终端设备时，可以包括图7中虚线框内的组件；当散热控制设备实现为常规服务器、云服务器、数据中心或服务器阵列等服务端设备时，可以不包括图7中虚线框内的组件。
本实施例提供的散热控制设备，采用预先训练得到的过热风险预测模型，通过该模型体现设备功率信息与制冷参数之间存在的过热风险关系，进而，在该过热风险预测模型的基础上，可根据指定空间内待散热设备的功率变化情况，动态调整制冷系统的制冷参数，达到动态散热控制的目的，有利于降低制冷系统的能耗，节约电能资源。
相应地，本申请实施例还提供一种存储有计算机程序的计算机可读存储介质，计算机程序被处理器执行时，致使处理器能够实现上述散热控制方法实施例中的步骤，或者实现上述系统实施例中的相应操作。关于详细步骤或操作，可参见前述实施例，在此不再赘述。
图8为本申请示例性实施例提供的一种模型训练设备的结构示意图。如图8所示，该设备包括：存储器81和处理器82。
存储器81，用于存储计算机程序，并可被配置为存储其它各种数据以支持在模型训练设备上的操作。这些数据的示例包括用于在模型训练设备上操 作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。
处理器82，与存储器81耦合，用于执行存储器81中的计算机程序，以用于：结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；利用多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型；其中，每组标记样本数据包括与至少一个待散热设备对应的至少一个样本功率信息、与制冷系统对应的样本制冷参数以及在该组标记样本数据下指定空间是否发生过热风险的标记结果。
其中，指定空间是指任何能够容纳且容纳有待散热设备的物理空间，例如可以是但不限于：机房、数据中心、办公楼宇、公司环境中的办公区域、或者家庭环境中的指定区域等。其中，家庭环境中的指定区域中存储有会产生热量且需要散热的待散热设备，例如电视机、智能音箱、大型游戏机、智能空调、净化器、家庭陪护机器人、个人电脑等。为了给待散热设备提供良好的工作环境，可提供制冷系统。可选地，制冷系统可部署于指定空间内，但不限于此。制冷系统主要负责带走指定空间内的热量，为指定空间内的待散热设备进行散热。
在一可选实施例中，处理器82在生成多组标记样本数据时，具体用于：根据至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据；以及利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据。
进一步可选地，处理器82在生成至少一组带标记的历史样本数据时，具体用于：获取至少一组未标记的历史样本数据，每组未标记的历史样本数据包括同一历史时刻或历史时段内至少一个待散热设备的历史功率信息和制冷系统的历史制冷参数；针对每组未标记的历史样本数据，根据至少一个待散热设备内部组件在相应历史时刻或历史时段内的温度与内部组件对应的过热温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的历史样本数据。
进一步可选地，处理器82在生成至少一组带标记的模拟样本数据时，具体用于：设计至少一组未标记的模拟样本数据，每组未标记的模拟样本数据包括与至少一个待散热设备对应的至少一个模拟功率信息以及与制冷系统对应的模拟制冷参数；针对每组未标记的模拟样本数据，利用CFD模型对该组未标记的模拟样本数据进行模拟以得到至少一个待散热设备内部器件的温度，利用至少一个待散热设备内部器件的温度与内部组件对应的过热温度阈值进行指定空间是否发生过热风险的标记，得到至少一组带标记的模拟样本数据。
进一步可选地，处理器82还用于：在利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据之前，利用至少一组带标记的历史样本数据，对CFD模型进行参数校正。
进一步，如图8所示，该模型训练设备还包括：通信组件83、显示器84、电源组件85、音频组件86等其它组件。图8中仅示意性给出部分组件，并不意味着模型训练设备只包括图8所示组件。另外，根据模型训练设备的实现形态的不同，图8中虚线框内的组件为可选组件，而非必选组件。例如，当模型训练设备实现为智能手机、平板电脑或台式电脑等终端设备时，可以包括图8中虚线框内的组件；当模型训练设备实现为常规服务器、云服务器、数据中心或服务器阵列等服务端设备时，可以不包括图8中虚线框内的组件。
本实施例提供的模型训练设备，将基于真实数据的样本生成方式与基于CFD模拟的样本生成方式相结合，通过CFD模拟计算的方法，为模型训练提供所需的样本数据，可以弥补基于真实数据的样本生成方式的不足，既有利于获取足够的样本数据，又可以获取指定空间运行在比较极端情况下的样本数据，可有效提高热风险预测模型的鲁棒性。
相应地，本申请实施例还提供一种存储有计算机程序的计算机可读存储介质，计算机程序被处理器执行时，致使处理器能够实现上述模型训练方法实施例中的步骤，或者实现上述系统实施例中的相应操作。关于详细步骤或操作，可参见前述实施例，在此不再赘述。
上述图7和图8中的通信组件被配置为便于通信组件所在设备和其他设备之间有线或无线方式的通信。通信组件所在设备可以接入基于通信标准的无线网络，如WiFi，2G、3G、4G/LTE、5G等移动通信网络，或它们的组合。在一个示例性实施例中，通信组件经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实施例中，通信组件还可以包括近场通信(NFC)模块，射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术等。
上述图7和图8中的显示器包括屏幕，其屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。
上述图7和图8中的电源组件，为电源组件所在设备的各种组件提供电力。电源组件可以包括电源管理系统，一个或多个电源，及其他与为电源组件所在设备生成、管理和分配电力相关联的组件。
上述图7和图8中的音频组件，可被配置为输出和/或输入音频信号。例如，音频组件包括一个麦克风(MIC)，当音频组件所在设备处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器或经由通信组件发送。在一些实施例中，音频组件还包括一个扬声器，用于输出音频信号。
本领域内的技术人员应明白，本发明的实施例可提供为方法、系统、或计算机程序产品。因此，本发明可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本发明可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本发明是参照根据本发明实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程 图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上，使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
在一个典型的配置中，计算设备包括一个或多个处理器(CPU)、输入/输出接口、网络接口和内存。
内存可能包括计算机可读介质中的非永久性存储器，随机存取存储器(RAM)和/或非易失性内存等形式，如只读存储器(ROM)或闪存(flash RAM)。内存是计算机可读介质的示例。
计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁带磁 磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算机可读介质不包括暂存电脑可读媒体(transitory media)，如调制的数据信号和载波。
还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
以上所述仅为本申请的实施例而已，并不用于限制本申请。对于本领域技术人员来说，本申请可以有各种更改和变化。凡在本申请的精神和原理之内所作的任何修改、等同替换、改进等，均应包含在本申请的权利要求范围之内。

Claims (31)

  1. 一种散热控制方法，其特征在于，包括：
    每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息；
    将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述指定空间在至少一个候选制冷参数下发生过热风险的概率；
    根据所述指定空间在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；
    根据所述目标制冷参数，控制制冷系统对所述指定空间内的至少一个待散热设备进行散热。
  2. 根据权利要求1所述的方法，其特征在于，每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息，包括以下至少一种：
    每当散热控制周期到达时，获取所述至少一个待散热设备的实际功率信息；
    每当监测所述至少一个待散热设备的总功率变化幅度大于第一幅度阈值时，获取所述至少一个待散热设备的实际功率信息；
    每当监测到出现待散热设备的功率变化幅度大于第二幅度阈值的情况时，获取所述至少一个待散热设备的实际功率信息。
  3. 根据权利要求1或2所述的方法，其特征在于，每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息，包括：
    每当散热控制条件被触发时，分别采集所述至少一个待散热设备在所述散热控制条件被触发时刻的功率值，作为所述至少一个待散热设备的实际功率信息；或者
    每当散热控制条件被触发时，分别获取所述至少一个待散热设备在本次散热控制与上次散热控制期间的功率均值，作为所述至少一个待散热设 备的实际功率信息。
  4. 根据权利要求1所述的方法，其特征在于，将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述指定空间在至少一个候选制冷参数下发生过热风险的概率，包括：
    根据所述过热风险预测模型在训练阶段使用的制冷参数的范围，确定所述至少一个候选制冷参数；
    针对每个候选制冷参数，将所述至少一个待散热设备的实际功率信息和所述候选制冷参数输入所述过热风险预测模型，以得到所述指定空间在所述候选制冷参数下发生过热风险的概率。
  5. 根据权利要求1所述的方法，其特征在于，根据所述指定空间在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数，包括：
    从所述指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于所述指定空间对应的过热风险概率阈值的目标概率；
    将所述至少一个候选制冷参数中，与所述目标概率对应的制冷参数作为所述目标制冷参数。
  6. 根据权利要求5所述的方法，其特征在于，从所述指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于所述指定空间对应的过热风险概率阈值的目标概率，包括：
    从所述指定空间在至少一个候选制冷参数下发生过热风险的概率中，选择小于所述过热风险概率阈值的最大概率作为所述目标概率。
  7. 根据权利要求5所述的方法，其特征在于，在选择小于所述指定空间对应的过热风险概率阈值的目标概率之前，还包括：
    将所述指定空间承载的应用或服务允许的热故障率转换为所述指定空间对应的过热风险概率阈值；所述热故障率表示所述指定空间在一定时间内能够发生过热风险的最大次数。
  8. 根据权利要求1、2和4-7中任一项所述的方法，其特征在于，在将所述至少一个待散热设备的实际功率信息输入过热风险预测模型之前， 还包括：
    结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；
    利用所述多组标记样本数据进行深度神经网络模型训练，得到所述过热风险预测模型；
    其中，每组标记样本数据包括与所述至少一个待散热设备对应的至少一个样本功率信息、与所述制冷系统对应的样本制冷参数以及在该组标记样本数据下所述指定空间是否发生过热风险的标记结果。
  9. 根据权利要求8所述的方法，其特征在于，结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据，包括：
    根据所述至少一个待散热设备的历史功率信息和所述制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据；以及
    利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据。
  10. 根据权利要求9所述的方法，其特征在于，根据所述至少一个待散热设备的历史功率信息和所述制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据，包括：
    获取至少一组未标记的历史样本数据，每组未标记的历史样本数据包括同一历史时刻或历史时段内所述至少一个待散热设备的历史功率信息和所述制冷系统的历史制冷参数；
    针对每组未标记的历史样本数据，根据所述至少一个待散热设备内部组件在相应历史时刻或历史时段内的温度与内部组件对应的过热温度阈值进行所述指定空间是否发生过热风险的标记，得到至少一组带标记的历史样本数据。
  11. 根据权利要求9所述的方法，其特征在于，利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据， 包括：
    设计至少一组未标记的模拟样本数据，每组未标记的模拟样本数据包括与所述至少一个待散热设备对应的至少一个模拟功率信息以及与所述制冷系统对应的模拟制冷参数；
    针对每组未标记的模拟样本数据，利用CFD模型对该组未标记的模拟样本数据进行模拟以得到所述至少一个待散热设备内部器件的温度，利用所述至少一个待散热设备内部器件的温度与内部组件对应的过热温度阈值进行所述指定空间是否发生过热风险的标记，得到至少一组带标记的模拟样本数据。
  12. 根据权利要求9所述的方法，其特征在于，在利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据之前，还包括：
    利用所述至少一组带标记的历史样本数据，对所述CFD模型进行参数校正。
  13. 根据权利要求8所述的方法，其特征在于，在生成多组标记样本数据的过程中，还包括：
    针对第一组标记样本数据，获取所述第一组标记样本数据对应时刻的外部大气温度作为样本大气温度，并将所述样本大气温度加入所述第一组标记样本数据中；其中，所述第一组标记样本数据是所述多组标记样本数据中的任一组。
  14. 根据权利要求13所述的方法，其特征在于，每当散热控制条件被触发时，还包括：获取散热控制条件被触发时对应的外部大气温度；
    将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述指定空间在至少一个候选制冷参数下发生过热风险的概率，包括：
    将所述至少一个待散热设备的实际功率信息以及散热控制条件被触发时对应的外部大气温度作为输入参数输入所述过热风险预测模型，以得到 所述指定空间在至少一个候选制冷参数下发生过热风险的概率。
  15. 根据权利要求8所述的方法，其特征在于，还包括以下至少一种操作：
    每当模型更新周期到达时，重新对所述过热风险预测模型进行模型训练；
    每当所述指定空间内待散热设备的数量变化时，重新对所述过热风险预测模型进行模型训练；
    每当所述指定空间内待散热设备之间的拓扑结构发生变化时，重新对所述过热风险预测模型进行模型训练。
  16. 根据权利要求1、2和4-7任一项所述的方法，其特征在于，根据所述目标制冷参数，控制所述制冷系统对所述指定空间内的至少一个待散热设备进行散热，包括：
    将所述目标制冷参数发送给所述制冷系统，以供所述制冷系统根据所述目标制冷参数进行制冷工作；或者
    在所述目标制冷参数与所述制冷系统当前使用的制冷参数不同时，将所述目标制冷参数发送给所述制冷系统，以供所述制冷系统根据所述目标制冷参数进行制冷工作。
  17. 根据权利要求1、2和4-7任一项所述的方法，其特征在于，所述指定空间为机房、数据中心、办公楼宇、公司环境中的办公区域或家庭环境中的指定区域。
  18. 一种模型训练方法，其特征在于，包括：
    结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；
    利用所述多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型；
    其中，每组标记样本数据包括与所述至少一个待散热设备对应的至少一个样本功率信息、与所述制冷系统对应的样本制冷参数以及在该组标记 样本数据下所述指定空间是否发生过热风险的标记结果。
  19. 根据权利要求18所述的方法，其特征在于，结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据，包括：
    根据所述至少一个待散热设备的历史功率信息和所述制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据；以及
    利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据。
  20. 根据权利要求19所述的方法，其特征在于，根据所述至少一个待散热设备的历史功率信息和所述制冷系统的历史制冷参数，生成至少一组带标记的历史样本数据，包括：
    获取至少一组未标记的历史样本数据，每组未标记的历史样本数据包括同一历史时刻或历史时段内所述至少一个待散热设备的历史功率信息和所述制冷系统的历史制冷参数；
    针对每组未标记的历史样本数据，根据所述至少一个待散热设备内部组件在相应历史时刻或历史时段内的温度与内部组件对应的过热温度阈值进行所述指定空间是否发生过热风险的标记，得到至少一组带标记的历史样本数据。
  21. 根据权利要求19所述的方法，其特征在于，利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据，包括：
    设计至少一组未标记的模拟样本数据，每组未标记的模拟样本数据包括与所述至少一个待散热设备对应的至少一个模拟功率信息以及与所述制冷系统对应的模拟制冷参数；
    针对每组未标记的模拟样本数据，利用CFD模型对该组未标记的模拟样本数据进行模拟以得到所述至少一个待散热设备内部器件的温度，利用所述至少一个待散热设备内部器件的温度与内部组件对应的过热温度阈值 进行所述指定空间是否发生过热风险的标记，得到至少一组带标记的模拟样本数据。
  22. 根据权利要求19-21任一项所述的方法，其特征在于，在利用CFD模型进行功率信息与制冷参数之间的模拟计算，生成至少一组带标记的模拟样本数据之前，还包括：
    利用所述至少一组带标记的历史样本数据，对所述CFD模型进行参数校正。
  23. 一种散热控制设备，其特征在于，包括：存储器和处理器；
    所述存储器，用于存储计算机程序；
    所述处理器，与所述存储器耦合，用于执行所述计算机程序，以用于：
    每当散热控制条件被触发时，获取指定空间内至少一个待散热设备的实际功率信息；
    将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述指定空间在至少一个候选制冷参数下发生过热风险的概率；
    根据所述指定空间在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；
    根据所述目标制冷参数，控制制冷系统对所述指定空间内的至少一个待散热设备进行散热。
  24. 根据权利要求23所述的设备，其特征在于，所述处理器还用于：
    结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；
    利用所述多组标记样本数据进行深度神经网络模型训练，得到所述过热风险预测模型；
    其中，每组标记样本数据包括与所述至少一个待散热设备对应的至少一个样本功率信息、与所述制冷系统对应的样本制冷参数以及在该组标记样本数据下所述指定空间是否发生过热风险的标记结果。
  25. 一种模型训练设备，其特征在于，包括：存储器和处理器；
    所述存储器，用于存储计算机程序；
    所述处理器，与所述存储器耦合，用于执行所述计算机程序，以用于：
    结合基于真实数据的样本生成方式与基于CFD模拟的样本生成方式，生成多组标记样本数据；
    利用所述多组标记样本数据进行深度神经网络模型训练，得到过热风险预测模型；
    其中，每组标记样本数据包括与所述至少一个待散热设备对应的至少一个样本功率信息、与所述制冷系统对应的样本制冷参数以及在该组标记样本数据下所述指定空间是否发生过热风险的标记结果。
  26. 一种机房系统，其特征在于，包括：机房，以及位于所述机房内的至少一个待散热装置、制冷系统以及散热控制设备；
    所述散热控制设备，用于每当散热控制条件被触发时，获取所述至少一个待散热设备的实际功率信息，将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述机房系统在至少一个候选制冷参数下发生过热风险的概率；根据所述机房系统在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据所述目标制冷参数，控制所述制冷系统对所述至少一个待散热设备进行散热；
    所述制冷系统，用于在所述散热控制设备的控制下，对所述机房内的至少一个待散热设备进行散热。
  27. 根据权利要求26所述的机房系统，其特征在于，所述制冷系统为空调系统，所述目标制冷参数为所述空调系统的工作温度、工作风速以及工作方式中的至少一种；或者
    所述制冷系统为水冷系统，所述目标制冷参数为所述水冷系统的出水温度、回水温度、水流速以及水流量中的至少一种。
  28. 一种数据中心系统，其特征在于，包括：至少一个机房；每个机房包括：至少一个待散热设备、制冷系统以及散热控制设备；
    所述散热控制设备，用于每当散热控制条件被触发时，获取其所属机 房内至少一个待散热设备的实际功率信息，将所述至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述机房在至少一个候选制冷参数下发生过热风险的概率；根据所述机房在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据所述目标制冷参数，控制其所属机房内的制冷系统对所述至少一个待散热设备进行散热；
    所述制冷系统，用于在所述散热控制设备的控制下，对其所属机房内的至少一个待散热设备进行散热。
  29. 一种数据中心系统，其特征在于，包括：至少一个机房、制冷系统以及散热控制设备；其中，每个机房包括至少一个待散热设备，所述制冷系统包括部署在每个机房内的制冷设备；
    所述散热控制设备，用于针对每个机房，每当散热控制条件被触发时，获取所述机房内至少一个待散热设备的实际功率信息，将所述机房内至少一个待散热设备的实际功率信息输入过热风险预测模型，以得到所述机房在至少一个候选制冷参数下发生过热风险的概率；根据所述机房在至少一个候选制冷参数下发生过热风险的概率，确定目标制冷参数；根据所述目标制冷参数，控制所述机房内的制冷设备对所述机房内的至少一个待散热设备进行散热。
  30. 一种存储有计算机程序的计算机可读存储介质，其特征在于，当所述计算机程序被处理器执行时，致使所述处理器执行权利要求1-17任一项所述方法中的步骤。
  31. 一种存储有计算机程序的计算机可读存储介质，其特征在于，当所述计算机程序被处理器执行时，致使所述处理器执行权利要求18-22任一项所述方法中的步骤。
PCT/CN2019/104590 2019-09-05 2019-09-05 散热控制与模型训练方法、设备、系统及存储介质 WO2021042339A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
PCT/CN2019/104590 WO2021042339A1 (zh) 2019-09-05 2019-09-05 散热控制与模型训练方法、设备、系统及存储介质
CN201980095634.0A CN113748386B (zh) 2019-09-05 2019-09-05 散热控制与模型训练方法、设备、系统及存储介质

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2019/104590 WO2021042339A1 (zh) 2019-09-05 2019-09-05 散热控制与模型训练方法、设备、系统及存储介质

Publications (1)

Publication Number Publication Date
WO2021042339A1 true WO2021042339A1 (zh) 2021-03-11

Family

ID=74852163

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/104590 WO2021042339A1 (zh) 2019-09-05 2019-09-05 散热控制与模型训练方法、设备、系统及存储介质

Country Status (2)

Country Link
CN (1) CN113748386B (zh)
WO (1) WO2021042339A1 (zh)

Cited By (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113111589A (zh) * 2021-04-25 2021-07-13 北京百度网讯科技有限公司 预测模型的训练方法、预测供热温度的方法、装置和设备
CN113204883A (zh) * 2021-05-10 2021-08-03 奇瑞汽车股份有限公司 汽车散热器目标进风量自动计算装置及方法
CN113438865A (zh) * 2021-06-25 2021-09-24 上海安畅网络科技股份有限公司 一种控制服务器机柜散热的方法、系统、存储介质及设备
CN113866536A (zh) * 2021-09-07 2021-12-31 深圳Tcl数字技术有限公司 温升检测方法、装置和温升检测系统
CN113993343A (zh) * 2021-10-11 2022-01-28 科华数据股份有限公司 制冷设备的节能控制方法、装置、终端及存储介质
CN115023117A (zh) * 2022-06-24 2022-09-06 深圳市中科创激光技术有限公司 环境参数控制方法、防尘箱、计算机设备及可读存储介质
CN115484800A (zh) * 2022-11-14 2022-12-16 联通(广东)产业互联网有限公司 一种数据中心的液冷散热系统及其控制方法
CN115969152A (zh) * 2022-12-31 2023-04-18 云南钜盛电器科技有限公司 基于气流倍增模式的吹风设备检验控制系统
CN116796208A (zh) * 2023-08-23 2023-09-22 广州吉谷电器有限公司 基于历史数据的电水壶过热保护方法及装置
CN117062404A (zh) * 2023-08-15 2023-11-14 湖南恩智测控技术有限公司 电池模拟器控制方法、装置及存储介质

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116661574B (zh) * 2023-07-13 2023-12-12 之江实验室 一种计算设备散热的方法、装置、存储介质及电子设备
CN118554084B (zh) * 2024-07-30 2024-09-27 深圳联钜自控科技有限公司 控制器温度补偿方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102213475A (zh) * 2011-03-22 2011-10-12 曙光信息产业(北京)有限公司 一种数据中心功耗自适应管理方法
CN102541015A (zh) * 2012-01-10 2012-07-04 深圳市嘉力达实业有限公司 一种能效智能控制系统及方法
CN104698843A (zh) * 2015-02-06 2015-06-10 同济大学 一种基于模型预测控制的数据中心节能控制方法
US20190075687A1 (en) * 2017-09-06 2019-03-07 Vertiv Corporation Cooling unit energy optimization via smart supply air temperature setpoint control
CN109600956A (zh) * 2018-12-10 2019-04-09 国家电网有限公司信息通信分公司 一种数据中心服务器放置位置优化算法和系统

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103025120B (zh) * 2011-09-22 2016-03-30 联想(北京)有限公司 散热系统、散热方法及其电子设备
CN103322553B (zh) * 2013-07-04 2015-03-04 东南大学 一种火电机组过热汽温多模型扰动估计预测控制方法
JP6412093B2 (ja) * 2016-12-26 2018-10-24 ファナック株式会社 学習モデル構築装置およびオーバーヒート予知装置
CN108197396A (zh) * 2018-01-09 2018-06-22 国网福建省电力有限公司 一种基于pso-svm的高压隔离开关过热状态预测方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102213475A (zh) * 2011-03-22 2011-10-12 曙光信息产业(北京)有限公司 一种数据中心功耗自适应管理方法
CN102541015A (zh) * 2012-01-10 2012-07-04 深圳市嘉力达实业有限公司 一种能效智能控制系统及方法
CN104698843A (zh) * 2015-02-06 2015-06-10 同济大学 一种基于模型预测控制的数据中心节能控制方法
US20190075687A1 (en) * 2017-09-06 2019-03-07 Vertiv Corporation Cooling unit energy optimization via smart supply air temperature setpoint control
CN109600956A (zh) * 2018-12-10 2019-04-09 国家电网有限公司信息通信分公司 一种数据中心服务器放置位置优化算法和系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
MOORE ; CHASE ; RANGANATHAN: "Weatherman: Automated, Online and Predictive Thermal Mapping and Management for Data Centers", AUTONOMIC COMPUTING, 2006. ICAC '06. IEEE INTERNATIONAL CONFERENCE ON DUBLIN, IRELAND 13-16 JUNE 2006, PISCATAWAY, NJ, USA,IEEE, 1 January 2006 (2006-01-01), pages 155 - 164, XP031016274, ISBN: 978-1-4244-0175-8, DOI: 10.1109/ICAC.2006.1662394 *

Cited By (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113111589A (zh) * 2021-04-25 2021-07-13 北京百度网讯科技有限公司 预测模型的训练方法、预测供热温度的方法、装置和设备
CN113204883A (zh) * 2021-05-10 2021-08-03 奇瑞汽车股份有限公司 汽车散热器目标进风量自动计算装置及方法
CN113204883B (zh) * 2021-05-10 2023-11-24 奇瑞汽车股份有限公司 汽车散热器目标进风量自动计算装置及方法
CN113438865A (zh) * 2021-06-25 2021-09-24 上海安畅网络科技股份有限公司 一种控制服务器机柜散热的方法、系统、存储介质及设备
CN113438865B (zh) * 2021-06-25 2022-12-20 上海安畅网络科技股份有限公司 一种控制服务器机柜散热的方法、系统、存储介质及设备
CN113866536A (zh) * 2021-09-07 2021-12-31 深圳Tcl数字技术有限公司 温升检测方法、装置和温升检测系统
CN113993343A (zh) * 2021-10-11 2022-01-28 科华数据股份有限公司 制冷设备的节能控制方法、装置、终端及存储介质
CN115023117A (zh) * 2022-06-24 2022-09-06 深圳市中科创激光技术有限公司 环境参数控制方法、防尘箱、计算机设备及可读存储介质
CN115484800B (zh) * 2022-11-14 2023-03-07 联通(广东)产业互联网有限公司 一种数据中心的液冷散热系统及其控制方法
CN115484800A (zh) * 2022-11-14 2022-12-16 联通(广东)产业互联网有限公司 一种数据中心的液冷散热系统及其控制方法
CN115969152A (zh) * 2022-12-31 2023-04-18 云南钜盛电器科技有限公司 基于气流倍增模式的吹风设备检验控制系统
CN115969152B (zh) * 2022-12-31 2023-06-30 云南钜盛电器科技有限公司 基于气流倍增模式的吹风设备检验控制系统
CN117062404A (zh) * 2023-08-15 2023-11-14 湖南恩智测控技术有限公司 电池模拟器控制方法、装置及存储介质
CN117062404B (zh) * 2023-08-15 2024-02-23 湖南恩智测控技术有限公司 电池模拟器控制方法、装置及存储介质
CN116796208A (zh) * 2023-08-23 2023-09-22 广州吉谷电器有限公司 基于历史数据的电水壶过热保护方法及装置
CN116796208B (zh) * 2023-08-23 2023-11-24 广州吉谷电器有限公司 基于历史数据的电水壶过热保护方法及装置

Also Published As

Publication number Publication date
CN113748386B (zh) 2023-01-17
CN113748386A (zh) 2021-12-03

Similar Documents

Publication Publication Date Title
WO2021042339A1 (zh) 散热控制与模型训练方法、设备、系统及存储介质
Hossain et al. A belief rule based expert system for datacenter PUE prediction under uncertainty
Xia et al. Phone2Cloud: Exploiting computation offloading for energy saving on smartphones in mobile cloud computing
TWI454008B (zh) 電源分配單元及其所適用的電源管理系統
US10003502B1 (en) Integrated wireless sensor network (WSN) and massively parallel processing database management system (MPP DBMS)
CN104635663A (zh) 用于装置的分级管理的方法和系统
US20180003402A1 (en) Real-time control of highly variable thermal loads
WO2021046774A1 (zh) 资源调度与信息预测方法、设备、系统及存储介质
US20210123634A1 (en) Water heater controller and methods for controlling water temperature
CN109324679A (zh) 一种服务器能耗控制方法及装置
CN104595223A (zh) 一种基于天气信息的风机转速控制方法及系统
Gupta et al. Auxiliary informatics system: an advancement towards a smart home environment
WO2013040850A1 (zh) 基于云计算的空气处理设备管理控制系统及方法
Zhang et al. A new energy efficient VM scheduling algorithm for cloud computing based on dynamic programming
Sivakumar et al. A novel energy optimization framework to enhance the performance of sensor nodes in Industry 4.0
Khan et al. Advanced data analytics modeling for evidence-based data center energy management
WO2024113906A1 (zh) 一种服务器集群温度调节方法和装置
JP2017070190A (ja) 電気負荷分離の方法、システム及び非一時的なコンピュータ可読記憶媒体
Sathupadi Ai-driven energy optimization in sdn-based cloud computing for balancing cost, energy efficiency, and network performance
WO2019080807A1 (zh) 一种供能设备管理的方法、系统及终端设备
Mahmud et al. Power profiling of context aware systems: a contemporary analysis and framework for power conservation
Kumar et al. Effect of cooling systems on the energy efficiency of data centers: machine learning optimisation
CN111083201B (zh) 一种工业物联网中数据驱动制造服务的节能资源分配方法
Kumar et al. Power usage efficiency (PUE) optimization with counterpointing machine learning techniques for data center temperatures
Kanso et al. A Review of Energy Aware Cyber-Physical Systems

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 19944044

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 19944044

Country of ref document: EP

Kind code of ref document: A1