Machine room temperature control adjusting method and device, equipment, storage medium and program product

Abstract

The embodiment of the invention provides a machine room temperature control adjusting method and device, equipment, a storage medium and a program product. A cooling capacity demand sequence in a machine room is obtained, environment factor characteristic data at the current moment in the machine room is obtained, the cooling capacity demand sequence and the environment factor characteristic data at the current moment are input into a machine room cooling capacity demand prediction model, and cooling capacity demand prediction at the future moment is carried out through the machine room cooling capacity demand prediction model. Determining the cooling capacity demand of the machine room in the future; the current temperature and the current humidity of the machine room and the cooling capacity demand at the future moment are input into the machine room temperature control regulation model, the air conditioner regulation and control strategy is obtained through the machine room temperature control regulation model, and the temperature and air conditioner regulation and control strategy at the future moment can be actively pre-judged by pre-judging the cooling capacity demand at the future moment; the temperature and the humidity in the machine room are always maintained within the safe operation range of the equipment, so that the service life of the equipment is prolonged.

Description

Technical Field

[0001] This application belongs to the field of computer room air conditioning control technology, and in particular relates to a computer room temperature control and regulation method, device, equipment, storage medium and program product. Background Technology

[0002] With the rapid development of the data center industry, the stability and energy efficiency of the temperature control system in the computer room, as the core location for the operation of computer equipment, have become key technical requirements. A large number of servers, switches, and other equipment in the computer room generate a significant amount of heat during operation. Improper temperature control can not only affect the lifespan of the equipment but may also lead to system failures. At the same time, as one of the main sources of energy consumption in the computer room, optimizing the energy efficiency of the temperature control system is crucial for reducing overall operating costs.

[0003] In existing technologies, computer room temperature control mainly uses fixed settings to adjust the air conditioning control strategy. In other words, the air conditioning control strategy in the computer room is controlled based on the set threshold of the temperature sensor. For example, when the temperature sensor detects that the current temperature has reached the preset threshold, the air conditioner is turned on; or when the temperature sensor detects that the current temperature has reached the required temperature, the air conditioner is turned off to save energy.

[0004] However, existing technologies rely solely on a single, fixed setting to adjust the air conditioning control strategy, resulting in untimely temperature control in the computer room and affecting the lifespan of the equipment. Summary of the Invention

[0005] This application provides a method, device, equipment, storage medium, and program product for controlling the temperature of a computer room. It can determine an air conditioning control strategy that adapts to the complex and ever-changing environment of the computer room, adjust the air conditioning in the computer room in a timely manner, maintain the environment in the computer room within the range of safe equipment operation, and thus extend the service life of the equipment.

[0006] In a first aspect, embodiments of this application provide a method for regulating the temperature of a computer room, the method comprising: Obtain the cooling demand sequence in the computer room, and obtain the environmental factor characteristic data of the computer room at the current moment. The cooling demand sequence is a cooling demand sequence formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. The cooling demand sequence and the environmental factor characteristic data at the current moment are input into the data center cooling demand prediction model. The data center cooling demand prediction model is used to predict the cooling demand at future moments to determine the cooling demand of the data center at future moments. The current temperature and humidity of the computer room, as well as the future cooling demand, are input into the computer room temperature control model, and the air conditioning control strategy is obtained through the computer room temperature control model.

[0007] Secondly, embodiments of this application provide a computer room temperature control and regulation device, the device comprising: The acquisition module is used to acquire the cooling demand sequence in the computer room and the environmental factor characteristic data of the computer room at the current moment. The cooling demand sequence is a cooling demand sequence formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. The prediction module is used to input the cooling demand sequence and the environmental factor characteristic data at the current moment into the computer room cooling demand prediction model, and to predict the cooling demand at future moments through the computer room cooling demand prediction model to determine the cooling demand at future moments of the computer room. The control module is used to input the current temperature and humidity of the computer room and the future cooling demand into the computer room temperature control model, and obtain the air conditioning control strategy through the computer room temperature control model.

[0008] Thirdly, embodiments of this application provide an electronic device, the device including: a processor, and a memory storing computer program instructions; The processor reads and executes computer program instructions to implement a computer room temperature control method as described in any one of the first aspects.

[0009] Fourthly, embodiments of this application provide a computer storage medium storing computer program instructions, which, when executed by a processor, implement a computer room temperature control method as described in any one of the first aspects.

[0010] Fifthly, embodiments of this application provide a computer program product in which instructions, when executed by the processor of an electronic device, cause the electronic device to perform a computer room temperature control method as described in the first aspect.

[0011] The data center temperature control method, apparatus, equipment, storage medium, and program product provided in this application obtain a cooling demand sequence within the data center and environmental factor characteristic data of the data center at the current moment. The cooling demand sequence is formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. Then, the cooling demand sequence and the environmental factor characteristic data of the current moment are input into a data center cooling demand prediction model. The model predicts future cooling demand to determine the future cooling demand of the data center. Finally, the current temperature, current humidity, and future cooling demand of the data center are input into a data center temperature control model to obtain an air conditioning control strategy. Compared to existing technologies that rely solely on fixed settings to adjust air conditioning, which are unable to adapt to the complex and dynamic environment of the computer room and result in untimely temperature control, thus affecting equipment lifespan, this application predicts future cooling demand based on historical cooling demand sequences and current environmental data. This avoids the limitations of existing technologies that only respond to the current environmental state. The computer room temperature control model combines current temperature and humidity data with future cooling demand to generate an air conditioning control strategy, achieving a shift from passive response to proactive prediction. This avoids the problem of delayed temperature control caused by fixed settings, ensuring that the temperature and humidity in the computer room are always maintained within the safe operating range of the equipment, thereby extending the equipment's lifespan. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 A flowchart illustrating a method for controlling and regulating the temperature of a computer room provided in this application is shown. Figure 2 A schematic flowchart of a computer room temperature control method according to an embodiment of this application is shown; Figure 3 A flowchart illustrating a method for controlling and regulating the temperature of a computer room according to another embodiment of this application is shown; Figure 4 A flowchart illustrating a method for controlling and regulating the temperature of a computer room according to another embodiment of this application is shown; Figure 5 A flowchart illustrating the process of constructing a data center cooling demand forecasting model is shown. Figure 6 A schematic diagram of the structure of the computer room temperature control and regulation device provided in this application is shown. Figure 7This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0014] The features and exemplary embodiments of various aspects of this application will be described in detail below. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are intended only to explain this application and not to limit it. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples.

[0015] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes the element.

[0016] It should be noted that the acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0017] First, let me explain the terms used in this application: U-position: Rack Unit Usage Status (U-position) is a standardized unit of measurement commonly used in data centers to measure the height occupied by hardware such as servers and network equipment in the rack.

[0018] Dew point temperature refers to the temperature at which air is cooled to just reach saturation, i.e., when water vapor in the air begins to condense into liquid water, under the condition that the water vapor content and air pressure in the air remain constant. It is a key indicator reflecting the actual water vapor content in the air.

[0019] Humidity ratio: refers to the mass of water vapor contained in a unit mass of dry air, directly reflecting the absolute moisture content in the air.

[0020] Enthalpy is a physical quantity that measures the total energy contained in air. It includes the energy generated by changes in the temperature of dry air, i.e., sensible heat, and the energy generated by changes in the state of water vapor in the air, i.e., latent heat.

[0021] Currently, air conditioning control strategies are based on fixed settings. For example, when a temperature sensor detects that the current temperature has reached a preset threshold, the air conditioner is turned on; or when the temperature sensor detects that the current temperature has reached the desired temperature, the air conditioner is turned off. However, existing technologies rely solely on a single fixed setting to adjust the air conditioning control strategy, which cannot adapt to the complex and dynamic environment within a computer room. Furthermore, the air conditioning control strategy is a passive response based on actual temperature changes. This means that even after the temperature in the computer room has reached the preset threshold, there is still a period of time between the air conditioner's adjustment and the temperature reaching the desired temperature. Therefore, the existing temperature control strategies result in untimely temperature regulation in the computer room, affecting the lifespan of the equipment.

[0022] Based on this, this application provides a method for regulating the temperature of a computer room. By using the historical cooling demand sequence in the computer room and the environmental factor characteristic data at the current moment, the method can predict the future cooling demand in advance, so as to realize the proactive adjustment of the air conditioning control strategy. Then, based on the predicted future cooling demand, the output cooling capacity of the air conditioner is adjusted in advance, so that the temperature of the computer room can be effectively regulated before reaching the threshold. This avoids the problem of control lag caused by the passive response in the prior art, and can always ensure that the temperature and humidity in the computer room are within the safe operating range of the equipment, thereby extending the service life of the equipment.

[0023] The following is a description of a computer room temperature control method provided in the embodiments of this application.

[0024] Figure 1 The flowchart of a data center temperature control method provided in this application is shown, including: S101, constructing a digital twin model of the data center; S102, modeling the real-time performance and alarms of the data center; S103, constructing a data center cooling demand prediction model; S104, the data center temperature control model dynamically adjusts the control strategy according to the cooling demand results at future moments; S105, modeling the prediction feedback of the data center air conditioning control strategy.

[0025] Specifically, the digital twin model of the computer room constructed in step S101 can be used to simulate changes in the environment inside the computer room in real time, including key parameters such as temperature, humidity, and airflow distribution. The digital twin model of the computer room can reflect the real physical structure, equipment layout, heat source distribution, etc. of the computer room to ensure the accuracy of the simulation results. Specifically, in step S102, the real-time performance and alarm modeling of the computer room can be used to acquire real-time signal data from temperature and humidity sensors, visualize the data, and display different temperatures and humidity levels in real time using different colors. This enables the real-time display of temperature cloud maps or humidity cloud maps for the air-cooled computer room. At the same time, temperature alarms or humidity alarms are specially marked, thereby accurately grasping the temperature and humidity of the computer room and determining whether alarm limits are exceeded. Furthermore, when displaying the temperature cloud map, the usage status of U-positions in the rack, real-time signals of equipment load, and real-time data of air conditioning operation can be displayed simultaneously to achieve monitoring and linkage analysis of the computer room's operating status. Specifically, the data center cooling demand prediction model constructed in step S103 can be used to predict the temperature and humidity change trends at multiple future moments, predict cooling demand in advance, and change the control strategy from "passive response" to "active control". Specifically, in step S104, the computer room temperature control adjustment model dynamically adjusts the control strategy based on the future cooling demand results. This can be used to dynamically adjust the control strategy based on the future cooling demand results, thereby achieving the optimal control effect and ensuring that the temperature and humidity in the computer room are always within the safe operating range of the equipment. Specifically, in step S105, the prediction feedback model of the air conditioning control strategy in the computer room can be used to accurately feed back the air conditioning control strategy and the changes in the computer room temperature under the strategy, accurately control the temperature of the computer room, and determine whether the predicted effect can be achieved in actual control.

[0026] Figure 2 This application provides a schematic flowchart of a computer room temperature control method according to one embodiment. Figure 2 As shown, the method includes the following steps: S201: Obtain the cooling demand sequence in the computer room and obtain the environmental factor characteristic data of the computer room at the current moment. The cooling demand sequence is the cooling demand sequence formed by the cooling demand at the current moment and the cooling demand at multiple historical moments.

[0027] In this embodiment of the application, the cooling demand sequence in the computer room is a cooling demand sequence formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. For example, the cooling demand sequence is the cooling demand corresponding to each hour in the past day, or the cooling demand corresponding to each minute in the past hour. The cooling demand sequence can reflect the trend of cooling changes and provide a data basis for predicting future cooling demand.

[0028] In one example, the environmental factor characteristic data at the current moment is the factor that affects the dynamic changes in cooling demand. In order to accurately predict future cooling demand, the current environmental factor characteristic data includes, but is not limited to: temperature characteristic data, humidity characteristic data, dew point temperature characteristic data, moisture content characteristic data, enthalpy characteristic data, temperature threshold of temperature sensor, humidity threshold of humidity sensor, device power consumption change rate, device moving average power consumption, device power consumption fluctuation, and time characteristic data.

[0029] In particular, due to frequent personnel entering and exiting the computer room entrance area, outdoor air rushes in when the door is opened, resulting in unstable ventilation. Therefore, the temperature threshold of the temperature sensor at the computer room entrance needs to be set separately, with a higher temperature threshold than that of temperature sensors in other areas, to avoid frequent alarms caused by irrelevant interference. Furthermore, the computer room is laid out in a hot and cold aisle configuration to efficiently dissipate heat from the equipment. The cold aisle is the area where air is directly supplied by the air conditioning vents and is the main source of cool air for the equipment. The hot aisle is the area where air is collected from the equipment exhaust vents and is specifically designed to collect hot air, which is then drawn away by the air conditioning return vents. Therefore, the temperature threshold of the temperature sensor in the hot aisle is higher than that in the cold aisle.

[0030] When the current environmental factor characteristic data includes: dew point temperature characteristic data, moisture content characteristic data, and enthalpy characteristic data, the environmental factor characteristic data obtained at the current moment in the computer room includes: Based on temperature and humidity characteristic data, dew point temperature characteristic data are calculated.

[0031] In this embodiment of the application, since computer equipment in the computer room is sensitive to changes in humidity, if the air temperature drops below the dew point temperature, condensation will occur on the surface of the equipment, which may lead to short circuits, corrosion and other malfunctions. Therefore, by calculating the dew point temperature characteristic data, the risk of condensation on the equipment can be judged in advance, providing a basis for the dehumidification operation of the air conditioner.

[0032] Among them, the dew point temperature characteristic data can be calculated based on the Magnus-Tetens approximation formula. The dew point temperature characteristic data is indirectly derived through temperature and humidity characteristic data. First, the intermediate variables are calculated. Based on Solve for the dew point temperature. The formula is: (1) in, It serves as an intermediate variable, used only in the calculation process of connecting temperature characteristic data, humidity characteristic data, and dew point temperature characteristic data; Represents the natural logarithm; Represents temperature characteristic data; Represents humidity characteristic data; This represents the characteristic data of dew point temperature.

[0033] Based on temperature characteristic data, humidity characteristic data, and atmospheric pressure, the moisture content characteristic data is calculated.

[0034] In this embodiment of the application, since the moisture content characteristic data is an absolute humidity index, it directly reflects the absolute content of moisture in the air and is not affected by temperature changes. Therefore, it is a key parameter for calculating the cooling capacity required during the dehumidification process of air conditioning. If the moisture content in the computer room is too high, the air conditioner needs to consume more cooling capacity to remove the excess moisture.

[0035] The moisture content characteristic data can be derived by first calculating the saturated water vapor pressure using temperature characteristic data, then calculating the actual water vapor partial pressure using humidity characteristic data, and finally deriving the moisture content characteristic data based on atmospheric pressure. The calculation formula is as follows: Saturated water vapor pressure can be used with Arden The Arden Buck Equation (Buck) is calculated using the following formula: (2) in, Indicates saturated water vapor pressure; This represents the natural exponential function, used to convert linear relationships into nonlinear relationships that conform to the laws governing water vapor pressure changes; This represents temperature characteristic data.

[0036] Actual water vapor partial pressure refers to the pressure exerted by the actual water vapor present in the air. It directly reflects the absolute moisture content. The formula for calculating actual water vapor partial pressure is: (3) in, This represents the actual partial pressure of water vapor; Indicates saturated water vapor pressure; This represents humidity characteristic data.

[0037] The moisture content is obtained by combining the difference between the actual water vapor partial pressure and atmospheric pressure with the molar mass ratio. The calculation formula is as follows: (4) in, This represents the characteristic data of moisture content; This represents the actual partial pressure of water vapor; Indicates atmospheric pressure; It represents the ratio of the molar mass of water to the average molar mass of dry air.

[0038] Based on temperature and moisture content characteristic data, enthalpy characteristic data are calculated.

[0039] In this embodiment, the core function of the air conditioner is to remove excess heat from the air in the computer room. The enthalpy characteristic data directly corresponds to the total heat that needs to be removed. The higher the enthalpy characteristic data, the more total energy the air contains, and the greater the required cooling capacity. Therefore, the enthalpy characteristic data is one of the strongest relevant features for predicting the required cooling capacity. Furthermore, the enthalpy characteristic data directly provides the computer room cooling capacity demand prediction model with information on the total heat that needs to be removed, greatly simplifying the difficulty for the computer room cooling capacity demand prediction model to learn and fit the total cooling capacity demand from multiple dispersed features such as temperature characteristic data, humidity characteristic data, and equipment power consumption.

[0040] The formula for calculating enthalpy characteristic data is: (5) in, This represents enthalpy characteristic data; Represents temperature characteristic data; This represents the characteristic data of moisture content.

[0041] The current environmental factor characteristic data includes: equipment power consumption change, equipment moving average power consumption, and equipment power consumption standard deviation. The environmental factor characteristic data obtained at the current moment within the computer room includes: Obtain the device power consumption sequence, and calculate the change in device power consumption based on the device power consumption at each time point in the device power consumption sequence. The device power consumption sequence is the device power consumption sequence formed by the device power consumption at the current time and the device power consumption at multiple historical times.

[0042] In this embodiment of the application, a device power consumption sequence is obtained. The device power consumption sequence is formed by the device power consumption at the current moment and the device power consumption at multiple historical moments. Based on the device power consumption at each moment in the device power consumption sequence, the change in device power consumption is calculated.

[0043] In one example, the change in device power consumption is calculated as follows: (6) in, This indicates the change in device power consumption; This represents the device power consumption at the current time t; Indicates the past time t. Device power consumption after a certain period.

[0044] In another example, the change in device power consumption is calculated based on the device power consumption at all historical moments in the device power consumption sequence, and the calculation formula is as follows: (7) in, This indicates the change in device power consumption; This indicates the device's power consumption at the current moment; This indicates the number of historical moments in the device power consumption sequence; Indicates the device power consumption at each time point; index parameter Used to determine the sort position in the device power consumption sequence. .

[0045] Different time windows are selected, and based on the device power consumption sequence, the device power consumption at multiple historical moments corresponding to different time windows is extracted. For each time window, the average device power consumption at multiple historical moments corresponding to each time window is calculated, and finally the average device power consumption corresponding to multiple time windows is obtained.

[0046] In this embodiment, different time windows are selected, and the average power consumption of the device corresponding to each time window is calculated. In one example, the time window can be the past 5 minutes or the past 1 hour, etc. When the time window is the past 5 minutes, the average device power consumption is calculated based on the device power consumption of multiple historical moments corresponding to the past 5 minutes; when the time window is the past 1 hour, the average device power consumption is calculated based on the device power consumption of multiple historical moments corresponding to the past 1 hour. Finally, the average device power consumption within 5 minutes and the average device power consumption within the past 1 hour can be obtained. By obtaining the average device power consumption corresponding to different time windows, short-term fluctuations can be smoothed out, long-term trends can be highlighted, and large adjustments in cooling demand caused by short-term spikes in device power consumption due to instantaneous tasks can be avoided.

[0047] The standard deviation of device power consumption is calculated based on the device power consumption at each time point in the device power consumption sequence.

[0048] In this embodiment of the application, the standard deviation of device power consumption is calculated based on the device power consumption at each moment in the device power consumption sequence. The larger the standard deviation of device power consumption, the more violent the fluctuation of device power consumption. High volatility means that the heat output of the device is unstable. In order to avoid temperature runaway caused by sudden high heat, more cooling capacity needs to be reserved to cope with high temperature demand.

[0049] When the current environmental factor characteristic data includes time-periodic characteristic data, the environmental factor characteristic data obtained at the current moment in the computer room includes: Obtain environmental factor characteristic data and time information corresponding to the cooling demand sequence, and encode the time information according to the periodic sine and cosine to obtain time periodic characteristic data.

[0050] In this embodiment, the cooling demand and environmental factors of the data center exhibit significant periodicity. For example, during the workday, equipment load is high, leading to high cooling demand, while during off-hours, equipment load is low, resulting in low cooling demand. Similarly, during the week, cooling demand is high on weekdays and low on off-weeks. If traditional discrete numerical representations are used, such as representing hours with integers from 0-23 and days of the week with integers from 1-7, the difference between 0:00 and 23 is significant (23), even though they are adjacent in time. Using discrete numerical representations fails to help the model learn the periodicity of cooling demand over time. Therefore, periodic sine and cosine encoding is applied to the hourly information to make the distance between the 23rd hour and the 0th hour in the feature space similar, and periodic sine and cosine encoding is applied to the day of the week to make Monday and Sunday similar in the feature space, thus reflecting the actual situation.

[0051] In one example, time information corresponding to hourly environmental factor characteristic data and cooling demand sequences is encoded based on periodic sine and cosine coding. The calculation formula is as follows: (8) in, This represents the sinusoidal component in hourly periodic indicators; Indicates hourly information; This represents the cosine component in the hourly periodic index.

[0052] In another example, time information corresponding to environmental factor characteristic data with a weekday cycle and cooling demand sequence is encoded based on periodic sine and cosine coding. The calculation formula is as follows: (9) in, This represents the sine component of the weekday periodicity indicator. Indicates the day of the week; This represents the cosine component of the weekday periodicity indicator.

[0053] In this embodiment, the device power consumption change, moving average power consumption, and standard deviation of device power consumption calculated through the device power consumption sequence can dynamically capture the load trend and stability of computer equipment, avoiding the problem that relying solely on instantaneous power consumption data cannot reflect trends. Furthermore, by encoding environmental factor characteristic data and the time information corresponding to the cooling demand sequence using periodic sine and cosine encoding, time-periodic characteristic data is obtained. This enables the data center cooling demand prediction model to accurately identify the daily, weekly, and other periodic variations in cooling demand, making the cooling demand prediction results more closely reflect the actual periodic fluctuations.

[0054] In this embodiment, by acquiring multi-source environmental factor characteristic data, comprehensive and accurate input features are provided for cooling demand forecasting, significantly improving the adaptability of the data center cooling demand forecasting model to complex data center environments. For example, by acquiring temperature characteristic data, humidity characteristic data, and dew point temperature characteristic data, moisture content characteristic data, and enthalpy characteristic data derived from temperature and humidity characteristic data, the energy state and moisture content of air can be completely characterized from a thermodynamic perspective. Dew point temperature characteristic data can quantify the risk of condensation on equipment, moisture content characteristic data can stably reflect absolute humidity to accurately assess dehumidification needs, and enthalpy characteristic data directly integrates sensible heat and latent heat to quantify total heat, avoiding the one-sidedness of describing environmental information from single temperature and humidity characteristic data. By acquiring the above-mentioned environmental factor characteristic data, the accuracy of cooling demand forecasting is improved, providing a precise data foundation for subsequent air conditioning control strategies.

[0055] S202: Input the cooling demand sequence and the current environmental factor characteristic data into the computer room cooling demand prediction model, and use the computer room cooling demand prediction model to predict the cooling demand at future times, and determine the future cooling demand of the computer room.

[0056] In this embodiment, a data center cooling demand prediction model is used to integrate the cooling demand sequence and current environmental factor characteristic data to output the cooling demand for future moments. The data center cooling demand prediction model simultaneously captures the periodic patterns in the cooling demand sequence and the influence of current environmental factor characteristic data, accurately predicting the cooling demand for future moments.

[0057] S203: Input the current temperature, current humidity and future cooling demand of the computer room into the computer room temperature control model, and obtain the air conditioning control strategy through the computer room temperature control model.

[0058] In this embodiment, the data center temperature control model takes the current temperature, current humidity, and future cooling demand as inputs, and outputs an air conditioning control strategy to meet future cooling demand. In one example, the air conditioning control strategy includes: the air conditioner's start / stop status, the air conditioner's supply air temperature, and the air conditioner's fan speed.

[0059] In one example, to accurately obtain the air conditioning control strategy through the data center temperature control model, a training process for the data center temperature control model is as follows: Obtain training data for computer room temperature control. The training data for computer room temperature control includes multiple training samples for computer room temperature control. Each training sample for computer room temperature control includes: current temperature sample, current humidity sample, and future cooling demand sample.

[0060] In this embodiment, multiple training samples for data center temperature control are obtained. Each training sample includes: a current temperature sample, a current humidity sample, and a sample of future cooling demand. These diverse samples provide a comprehensive learning foundation for the data center temperature control model to be trained, ensuring that the trained model can adapt to complex and ever-changing data center environments.

[0061] In one example, the future cooling demand samples provide constraints for the training process of the computer room temperature control model. That is, during the training process of the computer room temperature control model, instead of blindly searching in the huge parameter space, the model searches for air conditioning operating parameters that can meet the future cooling demand samples, so that the finally determined air conditioning operating parameters can adapt to future environmental changes.

[0062] For each training sample of temperature control adjustment in the computer room, perform the following steps: Input the training samples of the computer room temperature control into the computer room temperature control model to be trained, and obtain the air conditioning control strategy of the training samples of the computer room temperature control.

[0063] In this embodiment, training samples for computer room temperature control are input into the computer room temperature control model to be trained. The computer room temperature control model to be trained will output an air conditioning control strategy based on the training samples. In one example, the air conditioning control strategy includes, but is not limited to: air conditioning start / stop status, air conditioning set supply air temperature, air conditioning set supply air speed, air conditioning set supply air frequency, and air conditioning set supply air volume.

[0064] Based on the air conditioning control strategy of the training samples, the reward function value is calculated. The reward function value is calculated based on the air conditioning energy consumption value, temperature value, humidity value, ideal temperature value, ideal humidity value and their respective weights corresponding to the air conditioning control strategy of the training samples. In this embodiment, a reward function value is calculated based on the air conditioning control strategy of the training samples. The reward function value is calculated based on the air conditioning energy consumption value, temperature value, humidity value, ideal temperature value, ideal humidity value, and their respective weights corresponding to the air conditioning control strategy of the training samples. The reward function calculation formula is as follows: (10) in, This indicates the weight value corresponding to the air conditioner's energy consumption value; This represents the weight value corresponding to the difference between the temperature value and the ideal temperature value under the air conditioning control strategy. This represents the weight value corresponding to the difference between the humidity value and the ideal humidity value under the air conditioning control strategy; Indicates air conditioner The corresponding energy consumption value; This indicates the temperature value under the air conditioning control strategy; Indicates the ideal temperature value; This indicates the humidity value under the air conditioning control strategy. This represents the ideal humidity value; N represents the number of air conditioners.

[0065] Based on the reward function value, determine whether the reward function value meets the preset training stopping condition. The preset training stopping condition is that the reward function value is minimized.

[0066] In this embodiment of the application, the preset training stop condition is determined based on the reward function value. When the reward function value is the minimum, the selected air conditioning control strategy can meet the requirement of the lowest air conditioning energy consumption, and the temperature and humidity values ​​achieved based on the air conditioning control strategy are closest to the ideal temperature and humidity values.

[0067] If the conditions are not met, adjust the model parameters of the computer room temperature control model to be trained, and continue training the computer room temperature control model using the computer room temperature control training set until the preset training stop condition is met, and obtain the trained computer room temperature control model.

[0068] In this embodiment of the application, if the reward function calculated based on the air conditioning control strategy does not meet the training stop condition, the model parameters of the computer room temperature control model to be trained will continue to be adjusted to select a new round of air conditioning control strategy. In one example, the model parameters of the computer room temperature control model to be trained can be adjusted based on particle swarm optimization algorithm, genetic algorithm, ant colony algorithm, bee algorithm, simulated annealing algorithm, etc., to select a new round of air conditioning control strategy.

[0069] In this embodiment, training data for data center temperature control is acquired. This training data includes multiple training samples for data center temperature control, enabling the data center temperature control model to learn diverse environmental characteristics and avoid overfitting in a single scenario. The reward function then comprehensively considers air conditioning energy consumption and deviations from ideal temperature and humidity values ​​to ensure that the air conditioning control strategy output by the data center temperature control model minimizes energy consumption while maintaining temperature and humidity within ideal ranges. Furthermore, during training, the data center temperature control model is trained by inputting samples of future cooling demand. The training objective is transformed into: how to accurately generate future cooling demand with minimal air conditioning energy consumption in the next time period, while ensuring that the temperature and humidity within the data center are within ideal ranges. This allows for the early generation of future air conditioning control strategies, ensuring that the temperature and humidity within the data center remain within ideal ranges.

[0070] In this embodiment, the cooling demand sequence in the computer room and the environmental factor characteristic data of the computer room at the current moment are obtained. The cooling demand sequence is formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. Then, the cooling demand sequence and the environmental factor characteristic data at the current moment are input into the computer room cooling demand prediction model. The computer room cooling demand prediction model is used to predict the cooling demand at future moments to determine the future cooling demand of the computer room. Finally, the current temperature, current humidity and future cooling demand of the computer room are input into the computer room temperature control model to obtain the air conditioning control strategy. Compared to existing technologies that rely solely on fixed settings to adjust air conditioning, which are unable to adapt to the complex and dynamic environment of the computer room and result in untimely temperature control, thus affecting equipment lifespan, this application predicts future cooling demand based on historical cooling demand sequences and current environmental data. This avoids the limitations of existing technologies that only respond to the current environmental state. The computer room temperature control model combines current temperature and humidity data with future cooling demand to generate an air conditioning control strategy, achieving a shift from passive response to proactive prediction. This avoids the problem of delayed temperature control caused by fixed settings, ensuring that the temperature and humidity in the computer room are always maintained within the safe operating range of the equipment, thereby extending the equipment's lifespan.

[0071] Figure 3 This illustration shows a flowchart of a computer room temperature control method according to another embodiment of this application, as shown below. Figure 3 As shown above, in the above Figure 2 Based on the illustrated embodiment, one specific implementation of step S202 is as follows: S301: Input the cooling demand sequence into the first computer room cooling demand prediction model, and use the first computer room cooling demand prediction model to predict the cooling demand at future times, and determine the first cooling demand of the computer room at future times. The first computer room cooling demand prediction model is a time series model.

[0072] In this embodiment of the application, since the cooling demand has a time periodicity, the periodicity of the cooling demand sequence is analyzed by a time series model to predict the first cooling demand in the future. That is, the cooling demand sequence is input into the first data center cooling demand prediction model, and the cooling demand is predicted by the first data center cooling demand prediction model to obtain the first cooling demand at the future time.

[0073] In one example, a time-series model is used to analyze and predict time-series data, where a time series refers to a series of data arranged in chronological order and strongly correlated with time. In this embodiment, the cooling demand sequence is a sequence of cooling demand formed by the current cooling demand and the cooling demand at multiple historical moments, and it has a time periodicity. Therefore, the future cooling demand at the first moment can be predicted using a time-series model, namely the first data center cooling demand prediction model.

[0074] In another example, the cooling demand prediction model for the first data center can be a traditional time series algorithm or a deep learning time series algorithm. Traditional time series algorithms include: Moving Average (MA), Exponential Smoothing (ES), and Autoregressive Integrated Moving Average (ARIMA). Deep learning time series algorithms include: Multi-Layer Perceptron (MLP), Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), and Long Short-Term Time-series Network (LSTNet).

[0075] S302: Input the current environmental factor characteristic data into the second computer room cooling demand prediction model, and use the second computer room cooling demand prediction model to predict the future cooling demand, and determine the future cooling demand of the computer room. The second computer room cooling demand prediction model is a regression model.

[0076] In this embodiment of the application, the cooling demand is not only affected by historical cooling data, but also by real-time environmental factors. For example, a surge in equipment power consumption will lead to an increase in cooling demand, and a high temperature and high humidity environment will increase the cooling load. Therefore, the impact of environmental factors on the cooling demand prediction is predicted by a regression model. That is, by inputting the environmental factor characteristic data at the current moment into the second computer room cooling demand prediction model, the second cooling demand at the future moment is obtained.

[0077] In one example, regression models are predictive modeling techniques used to reveal the quantitative relationship between dependent and independent variables. They perform predictive analysis, causal inference, and variable selection by constructing mathematical equations. In this embodiment, environmental factor characteristic data is a key factor affecting cooling demand. Therefore, by using regression models to predict the impact of environmental factor characteristic data on cooling demand, future cooling demand can be predicted from a causal relationship perspective.

[0078] In another example, the cooling demand prediction model for the second data center can be a regression algorithm, such as: Linear Regression, Lasso Regression, Ridge Regression, Support Vector Regression (SVR), Random Forest Regression (RF), XGBoost Regression, and LightGBM Regression.

[0079] S303: The future cooling demand is calculated based on the first and second future cooling demand.

[0080] In this embodiment of the application, the future cooling demand can be calculated based on the first future cooling demand and the second future cooling demand. In one example, the future cooling demand can be obtained by adding the first future cooling demand and the second future cooling demand.

[0081] In this embodiment, the cooling demand sequence is input into a first data center cooling demand prediction model. This model predicts future cooling demand, determining the first cooling demand for the data center in the future. This fully leverages the time-varying patterns of cooling demand, providing a historical reference for future cooling demand prediction and ensuring that future cooling demand aligns with the natural fluctuations in cooling demand over time. Then, current environmental factor characteristic data is input into a second data center cooling demand prediction model. This model predicts future cooling demand, determining the second cooling demand for the data center in the future. This integrates the causal relationship between environmental factor characteristic data and cooling demand, allowing future cooling demand results to quickly respond to changes in environmental factors within the data center. Based on the first and second future cooling demand, the final future cooling demand is calculated. This ensures that the final future cooling demand adapts to both the periodic changes in cooling demand within the data center and the impact of environmental factor characteristic data on future cooling demand, thereby improving the accuracy of future cooling demand prediction.

[0082] In one example, to accurately predict future cooling demand using a data center cooling demand prediction model, the training process for the model is as follows: First, the training process for the cooling demand prediction model for the first computer room is as follows: Obtain a training sample set of cooling demand for the first data center. The training sample set of cooling demand for the first data center includes multiple training samples of cooling demand for the first data center. Each training sample of cooling demand for the first data center includes: a cooling demand sequence sample and a sample of actual cooling demand at the next moment corresponding to the cooling demand sequence. The cooling demand sequence sample refers to the cooling demand sequence sample formed by cooling demand at multiple moments. The sample of actual cooling demand at the next moment corresponding to the cooling demand sequence refers to the actual cooling demand at the next moment after the last moment of the cooling demand sequence.

[0083] In this embodiment of the application, multiple training samples of cooling demand for first data centers are obtained. Each training sample of cooling demand for first data centers includes: a cooling demand sequence and the actual cooling demand at the next moment corresponding to the cooling demand sequence. The cooling demand sequence is composed of cooling demand corresponding to multiple moments. For example, the cooling demand sequence is the cooling demand recorded every hour from 9:00 to 12:00. The actual cooling demand at the next moment corresponding to the cooling demand sequence refers to the actual cooling demand at the next moment immediately following the moment corresponding to the last cooling demand in the cooling demand sequence. For example, if the cooling demand sequence is the cooling demand recorded every hour from 9:00 to 12:00, the actual cooling demand at the next moment corresponding to the cooling demand sequence is the actual cooling demand at 13:00.

[0084] In one example, continuous cooling demand records can be extracted from the historical operational data of the data center, and samples can be generated using a sliding window approach. For instance, a cooling demand sequence can be created with each 24-hour window as a window, and the actual cooling demand in the first hour after the window ends can be used as the actual cooling demand for the next moment. Samples can be generated in batches to cover different time scenarios, such as weekdays, weekends, and different seasons.

[0085] Based on the training sample set of cooling demand of the first computer room, the model for predicting cooling demand of the first computer room is trained to obtain the trained model for predicting cooling demand of the first computer room.

[0086] In this embodiment, the training sample set of cooling demand for the first data center is input into the first data center cooling demand prediction model to be trained for model training, resulting in a trained first data center cooling demand prediction model. In one example, the first data center cooling demand prediction model to be trained is an ARIMA model. The ARIMA model can effectively capture the autoregressive, differencing, and moving average characteristics in the cooling demand sequence to predict future cooling demand. In another example, the first data center cooling demand prediction model to be trained can also be a Seasonal Autoregressive Integrated Moving Average Model (SARIMA). SARIMA retains the ability of the ARIMA model to capture time-series features while adding modeling of seasonal sequence data, thereby improving the prediction accuracy for data containing seasonal components. Here, data containing seasonal components refers to time-series data that exhibits periodic fluctuations with seasons, months, or weeks.

[0087] In this embodiment, multiple training samples of cooling demand for the first data center are obtained. These training samples include cooling demand sequences, which contain historical data on cooling demand changes over time. This provides a data foundation for the first data center cooling demand prediction model to learn time dependencies, enabling the model to capture the periodic patterns of cooling demand. By repeatedly learning the correlation between the cooling demand sequence and the actual cooling demand at the next moment corresponding to the sequence, the model continuously adjusts parameters to reduce errors. Ultimately, the trained first data center cooling demand prediction model can extract time patterns from historical cooling demand sequences and output reliable predictions of cooling demand at the first future moment, providing an accurate benchmark for subsequent future cooling demand predictions.

[0088] Secondly, the training process for the cooling demand model of the second computer room is as follows: Obtain a training sample set for the cooling demand of the second data center. The training sample set for the cooling demand of the second data center includes multiple training samples for the cooling demand of the second data center. Each training sample for the cooling demand of the second data center includes: environmental factor feature data samples at the current moment and cooling demand residual samples at the next moment. The cooling demand residual samples at the next moment are calculated based on the actual cooling demand and the predicted cooling demand at the next moment. The predicted cooling demand at the next moment is predicted based on the trained cooling demand prediction model of the first data center.

[0089] In this embodiment, multiple training samples of cooling demand for second data centers are obtained. Each training sample includes: environmental factor characteristic data samples at the current moment and cooling demand residuals at the next moment. The cooling demand residuals at the next moment are calculated based on the actual cooling demand and the predicted cooling demand at the next moment. The predicted cooling demand at the next moment is predicted based on the trained cooling demand prediction model for the first data center. The cooling demand residuals at the next moment represent the cooling deviations that the first data center cooling demand prediction model cannot explain. For example, if the first data center cooling demand prediction model does not consider "sudden high load on equipment," the predicted cooling demand at the next moment may be lower than the actual cooling demand at the next moment. In this case, the residual is positive, indicating that this "underestimated cooling" is caused by the environmental factor characteristic data.

[0090] Based on the training sample set of cooling demand for the second computer room, the model to be trained for predicting the cooling demand for the second computer room is trained to obtain the trained model.

[0091] In this embodiment, the training sample set of cooling demand for the second data center is input into the second data center cooling demand prediction model to be trained for model training, resulting in a trained second data center cooling demand prediction model. In one example, the second data center cooling demand prediction model to be trained is an XGB model or an LGB model. By training the second data center cooling demand prediction model to be trained, the training model can learn the correlation between environmental factor characteristic data and cooling demand residuals, thereby achieving correction of cooling demand.

[0092] In this embodiment, multiple training samples of cooling demand for the second data center are obtained. These training samples include environmental factor characteristic data samples at the current moment and cooling demand residuals at the next moment. This provides a data foundation for the second data center cooling demand prediction model to learn the correlation between environmental factor characteristic data and cooling demand residuals. This enables the second data center cooling demand prediction model to accurately grasp the correlation between environmental factor characteristic data and cooling demand residuals, and ultimately to accurately predict the cooling demand residuals based on real-time environmental factor characteristic data, i.e., the second cooling demand at the future moment. This accurately corrects the deviation of the first data center cooling demand prediction model, thereby improving the accuracy of the cooling demand prediction value.

[0093] Figure 4 This illustration shows a flowchart of a computer room temperature control method according to another embodiment of this application, as shown below. Figure 4 As shown above, in the above Figure 2 Based on the illustrated embodiment, after step S203, the method further includes: S401: Construct a digital twin model of the computer room corresponding to the computer room. The digital twin model of the computer room is used to simulate the physical environment parameters and air conditioning operation status in the computer room.

[0094] In this embodiment, a digital twin model of the data center is constructed based on its physical spatial structure and characteristics to simulate the physical environmental parameters and air conditioning operation status within the data center. In one example, the digital twin model can be constructed based on the data center floor plan, server rack placement, temperature sensor locations, humidity sensor locations, and air conditioning unit locations to accurately simulate physical environmental parameters such as temperature and humidity changes, as well as the air conditioning operation status. In another example, when constructing the digital twin model, non-critical data parameters such as decorative details in some areas of the data center, like wall trim and slight unevenness in the floor, are reasonably simplified to significantly reduce the amount of data that needs to be processed during the construction of the digital twin model, thus accelerating its construction. Furthermore, since their impact on the overall airflow, temperature, and humidity distribution is negligible, they ultimately have no effect on the temperature field, humidity field, or air conditioning control effect.

[0095] S402: Obtain the influence relationship model of the air conditioner temperature sensor in the computer room. The influence relationship model of the air conditioner temperature sensor is used to represent the correlation between the change of the air conditioner output cooling capacity and the change of the temperature sensor detection value.

[0096] In this embodiment, the impact of changes in cooling output of each air conditioner on different sensors varies. For example, the temperature sensor near the air outlet of a certain air conditioner is mainly affected by that air conditioner, while the other air conditioners have almost no impact on the temperature sensor. Therefore, by obtaining the influence relationship model of the air conditioner temperature sensors in the computer room, the temperature change can be accurately predicted, providing a quantitative basis for the simulation of the computer room digital twin model.

[0097] In one example, to accurately calculate the temperature change of each temperature sensor after being affected by the air conditioning control strategy, the influence relationship model of the air conditioning temperature sensors in the computer room is obtained as follows: The system acquires the operating status sequences of multiple air conditioners and the temperature sequences of multiple temperature sensors within the computer room. The operating status sequence of each air conditioner includes the operating status at multiple times, and the temperature sequence of each temperature sensor includes the temperature at multiple times.

[0098] In this embodiment, the operating status sequences of multiple air conditioners and the temperature sequences of multiple temperature sensors within the computer room are acquired. Each air conditioner's operating status sequence includes operating statuses at multiple times, and each temperature sensor's temperature sequence includes temperatures at multiple times. In one example, historical operating status data and temperature data from all air conditioners and temperature sensors within the computer room are collected and integrated. Then, using a unified timestamp as a reference, all historical operating status data and temperature data are aligned to ensure that at any given time, the operating status data of all air conditioners and the temperature data of the temperature sensors can be obtained. Finally, the collected historical operating status data and temperature data are divided into multiple operating status sequences and temperature sequences.

[0099] In one example, the sequence of operating states includes, but is not limited to: start / stop state, temperature state, air supply speed state, and operating power state.

[0100] Multiple different combinations of air conditioning temperature sensors are identified, each combination including an air conditioner and a temperature sensor.

[0101] In this embodiment of the application, based on all temperature sensors and air conditioners in the computer room, multiple different combinations of air conditioner temperature sensors are determined. Each combination of air conditioner temperature sensors includes an air conditioner and a temperature sensor. The control of the air conditioner in the combination of air conditioners will affect the temperature sensor.

[0102] For each air conditioner temperature sensor combination, select sequence data belonging to the same time period from the operating state sequence and temperature sequence of the air conditioner temperature sensor combination, and calculate the first correlation coefficient of the air conditioner temperature sensor combination.

[0103] In this embodiment of the application, for each air conditioner temperature sensor combination, sequence data belonging to the same time period are selected from the operating state sequence and temperature sequence of the air conditioner temperature sensor combination, and the first correlation coefficient of the air conditioner temperature sensor combination is calculated. The first correlation coefficient represents the degree of correlation between the air conditioner control strategy and the temperature detected by the temperature sensor. The higher the absolute value of the first correlation coefficient, the stronger the influence of the air conditioner control strategy on the temperature detected by the temperature sensor.

[0104] For example: the running state sequence is , The corresponding time is , The corresponding time is , The corresponding time is The temperature sequence is , The corresponding time is , The corresponding time is , The corresponding time is Then calculate and The correlation coefficient is the first correlation coefficient between the air conditioner and the temperature sensor.

[0105] In one example, multiple correlation coefficients can be calculated based on multiple operating status sequences and temperature sequences belonging to the same time period, and then the first correlation coefficient can be calculated based on the average of the multiple correlation coefficients.

[0106] In another example, the first correlation coefficient between the running state sequence and the temperature sequence can be calculated based on the Pearson correlation coefficient algorithm or the Spearman rank correlation coefficient algorithm.

[0107] For each air conditioner temperature sensor combination, the sequence data belonging to different time periods are selected multiple times from the operating state sequence and temperature sequence of the air conditioner temperature sensor combination according to different lag times. A second correlation coefficient is calculated for each selected sequence data. The lag time corresponding to the largest absolute value among the multiple calculated second correlation coefficients is determined as the delay time of the air conditioner sensor combination.

[0108] In this embodiment, for each air conditioner temperature sensor combination, sequence data belonging to different time periods are selected multiple times from the operating state sequence and temperature sequence of the air conditioner temperature sensor combination, according to different lag times. Selecting sequence data belonging to different time periods according to different lag times means that the selected operating state sequence and temperature sequence belong to different time periods. The lag time refers to the offset of the time period corresponding to the temperature sequence compared to the time period corresponding to the operating state sequence. For example, the time period offset refers to the difference between the start time or the end time of the time period corresponding to the temperature sequence and the time period corresponding to the operating state sequence. Then, a second correlation coefficient is calculated for the air conditioner temperature sensor combination based on the selected sequence data each time. The lag time corresponding to the largest absolute value among the multiple calculated second correlation coefficients is determined as the delay time of the air conditioner sensor combination. The delay time represents the time difference between the implementation of the air conditioner control strategy and the detection of temperature change by the temperature sensor, that is, it represents the lag in the temperature response after air conditioner control.

[0109] For example: the lag time is At that time, the corresponding running state sequence is: , The corresponding time is , The corresponding time is , The corresponding time is The corresponding temperature sequence is: , The corresponding time is , The corresponding time is , The corresponding time is .

[0110] Lag time is At that time, the corresponding running state sequence is: , The corresponding time is , The corresponding time is , The corresponding time is The corresponding temperature sequence is: , The corresponding time is , The corresponding time is , The corresponding time is .

[0111] Lag time is At that time, the corresponding running state sequence is: , The corresponding time is , The corresponding time is , The corresponding time is The corresponding temperature sequence is: , The corresponding time is , The corresponding time is , The corresponding time is .

[0112] Calculate the second correlation coefficients for the operating state sequence and temperature sequence at different lag times, respectively. hour, and The second correlation coefficient, in hour, and The second correlation coefficient, in hour, and The second correlation coefficient of the above three factors was used. Finally, the lag time corresponding to the largest absolute value of the second correlation coefficient of the above three factors was selected as the delay time.

[0113] Based on the first correlation coefficient and delay time of multiple different combinations of air conditioning temperature sensors, an influence relationship model of air conditioning temperature sensors is constructed.

[0114] In this embodiment, an influence relationship model of air conditioning temperature sensors is constructed based on the first correlation coefficient and delay time of multiple different combinations of air conditioning temperature sensors.

[0115] In one example, after calculating the correlation coefficient and delay time of each air conditioner and each temperature sensor, an air conditioner-temperature sensor influence relationship model can be constructed. The air conditioner-temperature sensor influence relationship model includes an influence weight matrix and a delay time matrix. The influence weight matrix is ​​an M x N matrix W, where M is the number of temperature sensors and N is the number of air conditioners. The element W(i, j) in the matrix represents the influence weight of air conditioner j on temperature sensor i, that is, the correlation coefficient between air conditioner j and temperature sensor i. The delay time matrix is ​​also an M x N matrix P, where M is the number of temperature sensors and N is the number of air conditioners. The element P(i, j) in the matrix represents the delay time of air conditioner j on temperature sensor i.

[0116] In this embodiment, by acquiring the operating status sequences of multiple air conditioners and the temperature sequences of multiple temperature sensors in the computer room, and then determining multiple different combinations of air conditioner temperature sensors, for each combination, sequence data belonging to the same time period are selected from the operating status sequence and temperature sequence of that combination, and the first correlation coefficient of that combination is calculated to quantify the impact of the air conditioner on the temperature sensor readings. By calculating the second correlation coefficient between each air conditioner and each temperature sensor at different lag times, the lag time corresponding to the maximum absolute value of the second correlation coefficient is determined as the delay time between each air conditioner and each temperature sensor, thus determining the lag effect of the air conditioner on temperature, clarifying the time difference of temperature response after the air conditioner's operating status changes, thereby improving the accuracy of temperature prediction, providing a scientific basis for temperature control strategies, and improving the precision of regulation.

[0117] S403: Based on the air conditioning control strategy, predict the output cooling capacity of each air conditioner.

[0118] In this application embodiment, the output cooling capacity of each air conditioner is predicted based on the air conditioning control strategy. In one example, the output cooling capacity of each air conditioner is calculated based on the parameters in the air conditioning control strategy and the performance curve of the air conditioner.

[0119] S404: Input the output cooling capacity of each air conditioner into the air conditioner temperature sensor influence relationship model to obtain the temperature change of each temperature sensor.

[0120] In this embodiment, the output cooling capacity of each air conditioner is input into the air conditioner temperature sensor influence relationship model to obtain the temperature change of each temperature sensor, providing basic data on temperature change for the simulation of the digital twin model of the computer room.

[0121] S405: Input the air conditioning control strategy, the output cooling capacity of each air conditioner, and the temperature change of each temperature sensor into the data center digital twin model, and simulate the dynamic temperature change process of the data center after the implementation of the air conditioning control strategy based on the data center digital twin model.

[0122] In this embodiment, the air conditioning control strategy, the output cooling capacity of the air conditioner, and the temperature changes of each temperature sensor are input into the data center digital twin model. The data center digital twin model will dynamically simulate the temperature change process in the data center over time after the control strategy is implemented based on the digital space. The air conditioning control strategy is no longer a black box for users, and they do not need to passively wait for the final feedback of the air conditioning control strategy, so that users can know the control result of the control strategy in advance.

[0123] In this embodiment, by constructing a digital twin model of the computer room corresponding to the computer room, and combining the influence relationship model of the air conditioning temperature sensor to quantify the correlation between cooling capacity and temperature, the output cooling capacity of the air conditioner is first predicted based on the control strategy. Then, the sensor temperature change of the temperature sensor is obtained through the influence relationship model of the air conditioning temperature sensor. Finally, the digital twin model is input to simulate the dynamic temperature change after the strategy is implemented. The computer room temperature under the strategy can be known in advance, and it can be judged whether the control strategy can achieve the prediction effect. There is no need to actually operate the air conditioner, which greatly reduces the trial and error cost and improves the implementation efficiency of the air conditioning control strategy.

[0124] Figure 5 The flowchart illustrates the process of constructing a data center cooling demand prediction model, including: S501, collecting historical cooling demand sequences and current environmental factor characteristic data; S502, preprocessing the historical cooling demand sequences and current environmental factor characteristic data; S503, preprocessing the historical cooling demand sequences and current environmental factor characteristic data; S504, performing feature transformation and feature filtering on the current environmental factor characteristic data; and S505, constructing the data center cooling demand prediction model using regression and time series models. Specifically, step S501 is used to collect historical cooling demand sequences and environmental factor characteristic data at the current moment; Specifically, in step S502, the historical cooling demand sequence and the environmental factor characteristic data at the current moment are preprocessed, including but not limited to: missing value handling and outlier detection. For example, the missing ratio of samples and features is statistically analyzed. If the missing ratio is too large, the sample or variable is considered to be removed. If the missing ratio is reasonable, different methods can be used to fill the missing values, such as filling with statistical values ​​such as mean or median, filling with data forward and backward after sorting, or filling with interpolation.

[0125] Specifically, in step S503, anomaly detection is performed on the historical cooling demand sequence and the environmental factor characteristic data at the current moment. Detection methods include statistical analysis methods such as quantile analysis, 3-Sigma analysis, and box plot analysis. Anomaly detection algorithms such as Isolation Forest (IF), Local Outlier Factor (LOF), and One-Class Support Vector Machine (SVM-one-class) can also be used. For identified abnormal data, direct removal can be considered, or empirical values ​​or statistical values ​​can be used to fill in the gaps. Specifically, in step S504, feature transformation and feature filtering are performed on the environmental factor feature data at the current moment. This is mainly used to normalize or standardize numerical features and to perform feature filtering using variance analysis, stepwise regression, tree model feature importance analysis, etc. Specifically, step S505 is used to construct a data center cooling demand prediction model using regression and time series models.

[0126] Figure 6 A schematic diagram of the computer room temperature control and regulation device provided in this application is shown. Figure 6 As shown, the computer room temperature control and regulation device 600 provided in this application includes: The acquisition module 601 is used to acquire the cooling demand sequence in the computer room and the environmental factor characteristic data of the computer room at the current moment. The cooling demand sequence is the cooling demand sequence formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. Prediction module 602 is used to input the cooling demand sequence and the environmental factor characteristic data at the current moment into the computer room cooling demand prediction model, and to predict the cooling demand at the future moment through the computer room cooling demand prediction model to determine the cooling demand at the future moment of the computer room. The control module 603 is used to input the current temperature and humidity of the computer room and the cooling demand at future moments into the computer room temperature control model, and obtain the air conditioning control strategy through the computer room temperature control model.

[0127] In one example, the computer room temperature control device 600 may also include: The first acquisition submodule is used to acquire the training set data of the computer room temperature control and regulation. The training set data of the computer room temperature control and regulation includes multiple training samples of computer room temperature control and regulation. Each training sample of computer room temperature control and regulation includes: current temperature sample, current humidity sample and future cooling demand sample. For each training sample of temperature control adjustment in the computer room, perform the following steps: The first input submodule is used to input the training samples of the computer room temperature control into the computer room temperature control model to be trained, so as to obtain the air conditioning control strategy of the training samples of the computer room temperature control. The first calculation submodule is used to calculate the reward function value based on the air conditioning control strategy of the training samples. The reward function value is calculated based on the air conditioning energy consumption value, temperature value, humidity value, ideal temperature value, ideal humidity value and their respective weights corresponding to the air conditioning control strategy of the training samples. The first judgment submodule is used to determine whether the reward function value meets the preset training stopping condition based on the reward function value. The preset training stopping condition is that the reward function value is the minimum. The first adjustment submodule is used to adjust the model parameters of the computer room temperature control model to be trained when the conditions are not met, and to continue training the computer room temperature control model using the computer room temperature control training set until the preset training stop condition is met, so as to obtain the trained computer room temperature control model.

[0128] In one example, prediction module 602 includes: The second input submodule is used to input the cooling demand sequence into the first computer room cooling demand prediction model, and to predict the cooling demand at future time through the first computer room cooling demand prediction model to determine the first cooling demand at future time of the computer room. The first computer room cooling demand prediction model is a time series model. The third input submodule is used to input the current environmental factor characteristic data into the second computer room cooling demand prediction model, and to predict the future cooling demand through the second computer room cooling demand prediction model to determine the future second cooling demand of the computer room. The second computer room cooling demand prediction model is a regression model. The second calculation submodule is used to calculate the future cooling demand based on the first future cooling demand and the second future cooling demand.

[0129] In one example, the computer room temperature control device 600 may also include: The second acquisition submodule is used to acquire the first computer room cooling demand training sample set. The first computer room cooling demand training sample set includes multiple first computer room cooling demand training samples. Each first computer room cooling demand training sample includes: cooling demand sequence and the actual cooling demand at the next moment corresponding to the cooling demand sequence. The first training submodule is used to train the first computer room cooling demand prediction model based on the first computer room cooling demand training sample set, so as to obtain the trained first computer room cooling demand prediction model.

[0130] In one example, the computer room temperature control device 600 may also include: The third acquisition submodule is used to acquire the training sample set of the cooling demand of the second computer room. The training sample set of the cooling demand of the second computer room includes multiple training samples of the cooling demand of the second computer room. Each training sample of the cooling demand of the second computer room includes: the environmental factor feature data sample at the current moment and the cooling demand residual at the next moment. The cooling demand residual at the next moment is calculated based on the actual cooling demand at the next moment and the predicted cooling demand at the next moment. The predicted cooling demand at the next moment is predicted based on the trained cooling demand prediction model of the first computer room. The second training submodule is used to train the second computer room cooling demand prediction model based on the training sample set of the second computer room cooling demand, so as to obtain the trained second computer room cooling demand prediction model.

[0131] In one example, the computer room temperature control device 600 may also include: The building module is used to build a digital twin model of the data center corresponding to the data center. The digital twin model of the data center is used to simulate the physical environment parameters and air conditioning operation status in the data center. The fourth acquisition submodule is used to acquire the influence relationship model of the air conditioner temperature sensor in the computer room. The influence relationship model of the air conditioner temperature sensor is used to represent the correlation between the change of the air conditioner output cooling capacity and the change of the temperature sensor detection value. The first prediction submodule is used to predict the output cooling capacity of each air conditioner based on the air conditioning control strategy; The fourth input submodule is used to input the output cooling capacity of each air conditioner into the air conditioner temperature sensor influence model to obtain the temperature change of each temperature sensor. The fifth input submodule is used to input the air conditioning control strategy, the output cooling capacity of each air conditioner, and the temperature change of each temperature sensor into the data center digital twin model, and simulate the dynamic temperature change process of the data center after the implementation of the air conditioning control strategy based on the data center digital twin model.

[0132] In one example, the computer room temperature control device 600 may also include: The fifth acquisition submodule is used to acquire the operating status sequence of multiple air conditioners and the temperature sequence of multiple temperature sensors in the computer room. The operating status sequence of each air conditioner includes the operating status at multiple times, and the temperature sequence of each temperature sensor includes the temperature at multiple times. The first determining submodule is used to determine multiple different combinations of air conditioning temperature sensors, each combination of air conditioning temperature sensors including an air conditioner and a temperature sensor. The third calculation submodule is used to select sequence data belonging to the same time period from the operating status sequence and temperature sequence of each air conditioner temperature sensor combination, and calculate the first correlation coefficient of the air conditioner temperature sensor combination. The fourth calculation submodule is used to select sequence data belonging to different time periods multiple times from the operating status sequence and temperature sequence of each air conditioner temperature sensor combination, according to different lag times. Each selected sequence data corresponds to a second correlation coefficient of the air conditioner temperature sensor combination. The lag time corresponding to the largest absolute value among the multiple calculated second correlation coefficients is determined as the delay time of the air conditioner sensor combination. The first construction submodule is used to construct an influence relationship model of air conditioning temperature sensors based on the first correlation coefficient and delay time of multiple different combinations of air conditioning temperature sensors.

[0133] In one example, obtaining module 601 includes: The fifth acquisition submodule is used to acquire multiple types of characteristic data, including temperature, humidity, dew point temperature, moisture content, enthalpy, temperature threshold of temperature sensor, humidity threshold of humidity sensor, device power consumption change, device moving average power consumption, device power consumption fluctuation, and time periodicity characteristic data.

[0134] In one example, obtaining module 601 includes: The eighth calculation submodule is used to obtain the device power consumption sequence and calculate the change in device power consumption based on the device power consumption at each time in the device power consumption sequence. The device power consumption sequence is the device power consumption sequence formed by the device power consumption at the current time and the device power consumption at multiple historical times. The ninth calculation submodule is used to select different time windows and, based on the device power consumption sequence, extract the device power consumption at multiple historical moments corresponding to different time windows. For each time window, it calculates the average device power consumption at multiple historical moments corresponding to each time window, and finally obtains the average device power consumption corresponding to multiple time windows. The tenth calculation submodule is used to calculate the standard deviation of device power consumption based on the device power consumption at each time point in the device power consumption sequence; The sixth acquisition submodule is used to acquire environmental factor characteristic data and time information corresponding to the cooling demand sequence, and encode the time information according to the periodic sine and cosine to obtain time periodic characteristic data.

[0135] Figure 7 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application.

[0136] An electronic device may include a processor 701 and a memory 702 storing computer program instructions.

[0137] Specifically, the processor 701 may include a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.

[0138] Memory 702 may include mass storage for data or instructions. For example, and not limitingly, memory 702 may include a hard disk drive (HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or Universal Serial Bus (USB) drive, or a combination of two or more of these. In one instance, memory 702 may include removable or non-removable (or fixed) media, or memory 702 may be non-volatile solid-state memory. Memory 702 may be internal or external to the integrated gateway disaster recovery device.

[0139] In one instance, memory 702 may be read-only memory (ROM). In one instance, the ROM may be a mask-programmed ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), an electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these.

[0140] Memory 702 may include read-only memory (ROM), random access memory (RAM), disk storage media device, optical storage media device, flash memory device, electrical, optical, or other physical / tangible memory storage device. Therefore, generally, memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to one aspect of this disclosure.

[0141] The processor 701 reads and executes computer program instructions stored in the memory 702 to implement a computer room temperature control method in the above embodiment.

[0142] In one example, the electronic device may also include a communication interface 703 and a bus 704. Wherein, as... Figure 7 As shown, the processor 701, memory 702, and communication interface 703 are connected through bus 704 and complete communication with each other.

[0143] The communication interface 703 is mainly used to realize communication between various modules, devices, units and / or equipment in the embodiments of this application.

[0144] Bus 704 includes hardware, software, or both, that couples components of an online data traffic metering device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, bus 704 may include one or more buses. Although specific buses are described and illustrated in embodiments of this application, this application contemplates any suitable bus or interconnect.

[0145] In addition, in conjunction with the computer room temperature control method described in the above embodiments, this application embodiment can provide a computer storage medium for implementation. The computer storage medium stores computer program instructions; when these computer program instructions are executed by a processor, they implement any of the computer room temperature control methods described in the above embodiments.

[0146] This application also provides a computer program product, including a computer program, which, when executed, implements any of the computer room temperature control methods described in the above embodiments.

[0147] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.

[0148] The functional blocks shown in the above block diagram can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, they can be, for example, electronic circuits, application-specific integrated circuits (ASICs), appropriate firmware, plug-ins, function cards, etc. When implemented in software, the elements of this application are programs or code segments used to perform the required tasks. Programs or code segments can be stored on a machine-readable medium or transmitted over a transmission medium or communication link via data signals carried on a carrier wave. "Machine-readable medium" can include any medium capable of storing or transmitting information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, read-only memory (ROM), flash memory, erasable read-only memory (EROM), floppy disks, compact disc read-only memory (CD-ROM), optical disks, hard disks, fiber optic media, radio frequency (RF) links, etc. Code segments can be downloaded via computer networks such as the Internet, intranets, etc.

[0149] It should also be noted that the exemplary embodiments mentioned in this application describe methods or systems based on a series of steps or apparatus. However, this application is not limited to the order of the above steps; that is, the steps can be performed in the order mentioned in the embodiments, or in a different order, or several steps can be performed simultaneously.

[0150] The aspects of this disclosure have been described above with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block in the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that these instructions, executable via the processor of the computer or other programmable data processing apparatus, enable the implementation of the functions / actions specified in one or more blocks of the flowchart illustrations and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It is also understood that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by special-purpose hardware performing the specified functions or actions, or can be implemented by a combination of special-purpose hardware and computer instructions.

[0151] The above are merely specific embodiments of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.

Claims

1. A method for temperature control and regulation in a computer room, characterized in that, include: Obtain the cooling demand sequence in the computer room, and obtain the environmental factor characteristic data of the computer room at the current moment. The cooling demand sequence is a cooling demand sequence formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. The cooling demand sequence and the environmental factor characteristic data at the current moment are input into the data center cooling demand prediction model. The data center cooling demand prediction model is used to predict the cooling demand at future moments to determine the cooling demand of the data center at future moments. The current temperature and humidity of the computer room, as well as the future cooling demand, are input into the computer room temperature control model, and the air conditioning control strategy is obtained through the computer room temperature control model.

2. The method according to claim 1, characterized in that, Before inputting the current temperature, current humidity, and future cooling demand of the computer room into the computer room temperature control model, and obtaining the air conditioning control strategy through the computer room temperature control model, the following steps are also included: Acquire training data for computer room temperature control and regulation. The training data for computer room temperature control and regulation includes multiple training samples for computer room temperature control and regulation. Each training sample for computer room temperature control and regulation includes: current temperature sample, current humidity sample, and future cooling demand sample. For each of the aforementioned computer room temperature control training samples, perform the following steps: The training samples of the computer room temperature control are input into the computer room temperature control model to be trained to obtain the air conditioning control strategy of the training samples of the computer room temperature control. Based on the air conditioning control strategy of the training samples, a reward function value is calculated. The reward function value is calculated based on the air conditioning energy consumption value, temperature value, humidity value, ideal temperature value, ideal humidity value and their respective weights corresponding to the air conditioning control strategy of the training samples. Based on the reward function value, determine whether the reward function value meets the preset training stopping condition, wherein the preset training stopping condition is that the reward function value is minimized; If the conditions are not met, adjust the model parameters of the computer room temperature control model to be trained, and continue training the computer room temperature control model using the computer room temperature control training set until the preset training stop condition is met, thus obtaining the trained computer room temperature control model.

3. The method according to claim 1, characterized in that, The step of inputting the cooling demand sequence and the current environmental factor characteristic data into the data center cooling demand prediction model, and using the data center cooling demand prediction model to predict the future cooling demand, and determining the future cooling demand of the data center, includes: The cooling demand sequence is input into the first data center cooling demand prediction model. The cooling demand at future times is predicted by the first data center cooling demand prediction model to determine the first cooling demand at future times of the data center. The first data center cooling demand prediction model is a time series model. The environmental factor characteristic data at the current moment is input into the second computer room cooling demand prediction model. The cooling demand at future moments is predicted through the second computer room cooling demand prediction model to determine the second cooling demand of the computer room at future moments. The second computer room cooling demand prediction model is a regression model. The future cooling demand is calculated based on the first future cooling demand and the second future cooling demand.

4. The method according to claim 3, characterized in that, Before inputting the cooling demand sequence into the first data center cooling demand prediction model, the method further includes: Obtain a training sample set of cooling demand for a first data center. The training sample set of cooling demand for a first data center includes multiple training samples of cooling demand for a first data center. Each training sample of cooling demand for a first data center includes: a cooling demand sequence sample and a sample of actual cooling demand at the next moment corresponding to the cooling demand sequence. The cooling demand sequence sample refers to a cooling demand sequence sample formed by cooling demand at multiple moments. The sample of actual cooling demand at the next moment corresponding to the cooling demand sequence refers to the actual cooling demand at the next moment after the last moment of the cooling demand sequence. Based on the training sample set of the first data center cooling demand, the first data center cooling demand prediction model to be trained is trained to obtain the trained first data center cooling demand prediction model.

5. The method according to claim 4, characterized in that, Before inputting the current environmental factor characteristic data into the second computer room cooling demand prediction model, the following steps are also included: Obtain a training sample set of cooling demand for the second data center. The training sample set of cooling demand for the second data center includes multiple training samples of cooling demand for the second data center. Each training sample of cooling demand for the second data center includes: environmental factor feature data sample at the current moment and cooling demand residual sample at the next moment. The cooling demand residual sample at the next moment is calculated based on the actual cooling demand at the next moment and the predicted cooling demand at the next moment. The predicted cooling demand at the next moment is predicted based on the trained cooling demand prediction model of the first data center. Based on the training sample set of the cooling demand of the second data center, the model to be trained for predicting the cooling demand of the second data center is trained to obtain the trained model.

6. The method according to claim 1, characterized in that, After inputting the current temperature, current humidity, and future cooling demand of the computer room into the computer room temperature control model, and obtaining the air conditioning control strategy through the computer room temperature control model, the following steps are also included: Construct a digital twin model of the computer room corresponding to the computer room. The digital twin model of the computer room is used to simulate the physical environment parameters and air conditioning operation status in the computer room. Obtain the influence relationship model of the air conditioner temperature sensor in the computer room. The influence relationship model of the air conditioner temperature sensor is used to represent the correlation between the change of the air conditioner output cooling capacity and the change of the temperature sensor detection value. Based on the aforementioned air conditioning control strategy, predict the cooling output of each air conditioner; The output cooling capacity of each air conditioner is input into the air conditioner temperature sensor influence model to obtain the temperature change of each temperature sensor. The air conditioning control strategy, the output cooling capacity of each air conditioner, and the temperature change of each temperature sensor are input into the data center digital twin model. Based on the data center digital twin model, the dynamic temperature change process of the data center after the implementation of the air conditioning control strategy is simulated.

7. The method according to claim 6, characterized in that, The process of obtaining the influence relationship model of the air conditioning temperature sensor in the computer room includes: The system acquires the operating status sequences of multiple air conditioners and the temperature sequences of multiple temperature sensors within the computer room. Each air conditioner's operating status sequence includes operating statuses at multiple times, and each temperature sensor's temperature sequence includes temperatures at multiple times. A plurality of different air conditioner temperature sensor combinations are determined, each air conditioner temperature sensor combination including one air conditioner and one temperature sensor; For each of the air conditioner temperature sensor combinations, select sequence data belonging to the same time period from the operating state sequence and the temperature sequence of the air conditioner temperature sensor combination, and calculate the first correlation coefficient of the air conditioner temperature sensor combination. For each air conditioner temperature sensor combination, sequence data belonging to different time periods are selected multiple times from the operating state sequence and temperature sequence of the air conditioner temperature sensor combination according to different lag times. A second correlation coefficient is calculated for the air conditioner temperature sensor combination for each selected sequence data. The lag time corresponding to the largest absolute value among the multiple calculated second correlation coefficients is determined as the delay time of the air conditioner sensor combination. Based on the first correlation coefficient and the delay time of the multiple different combinations of air conditioning temperature sensors, an influence relationship model of the air conditioning temperature sensors is constructed.

8. The method according to claim 1, characterized in that, The environmental factor characteristic data at the current moment includes multiple data such as temperature characteristic data, humidity characteristic data, dew point temperature characteristic data, moisture content characteristic data, enthalpy characteristic data, temperature threshold of temperature sensor, humidity threshold of humidity sensor, device power consumption change, device moving average power consumption, device power consumption fluctuation, and time periodic characteristic data.

9. The method according to claim 8, characterized in that, The acquisition of environmental factor characteristic data in the computer room at the current moment includes: Obtain the device power consumption sequence, and calculate the change in device power consumption based on the device power consumption at each time point in the device power consumption sequence, wherein the device power consumption sequence is a sequence formed by the device power consumption at the current time and the device power consumption at multiple historical times; Different time windows are selected, and based on the device power consumption sequence, the device power consumption at multiple historical moments corresponding to different time windows is extracted. The average device power consumption at multiple historical moments corresponding to each time window is calculated to obtain the average device power consumption corresponding to multiple time windows. Calculate the standard deviation of device power consumption based on the device power consumption at each moment in the device power consumption sequence; The environmental factor characteristic data and the time information corresponding to the cooling demand sequence are obtained, and the time information is encoded according to the periodic sine and cosine to obtain the time periodic characteristic data.

10. A computer room temperature control and regulation device, characterized in that, include: The acquisition module is used to acquire the cooling demand sequence in the computer room and the environmental factor characteristic data of the computer room at the current moment. The cooling demand sequence is a cooling demand sequence formed by the cooling demand at the current moment and the cooling demand at multiple historical moments. The prediction module is used to input the cooling demand sequence and the environmental factor characteristic data at the current moment into the computer room cooling demand prediction model, and to predict the cooling demand at future moments through the computer room cooling demand prediction model to determine the cooling demand at future moments of the computer room. The control module is used to input the current temperature and humidity of the computer room and the future cooling demand into the computer room temperature control model, and obtain the air conditioning control strategy through the computer room temperature control model.

11. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement a computer room temperature control method as described in any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement a computer room temperature control method as described in any one of claims 1-9.

13. A computer program product, characterized in that, When the instructions in the computer program product are executed by the processor of the electronic device, the electronic device performs a computer room temperature control method as described in any one of claims 1-9.

Images