Temperature regulation and control method and device, computer program product and electronic equipment

Abstract

The invention relates to a temperature regulation and control method and device, a computer program product and electronic equipment, and relates to the technical field of thermal management, and the method comprises the steps: obtaining historical load data of computing power equipment and operation state data of refrigeration equipment, according to the historical load data and the operation state data, predicting to obtain a thermal load of the computing power equipment; determining sharing data of the refrigeration equipment for the thermal load, and predicting the temperature difference of the refrigeration equipment according to the sharing data, the operation parameters of the refrigeration equipment and the thermal load; according to the temperature difference of the refrigeration equipment, the output temperature of the refrigeration equipment is obtained, the preset reference temperature and the preset reference humidity are obtained, the output temperature is adjusted based on the preset reference temperature and the preset reference humidity, and the target temperature is obtained; and controlling the temperature of the refrigeration equipment based on the target temperature. The energy consumption of the refrigeration equipment is reduced, and active regulation and control of the temperature of the refrigeration equipment are achieved.

Description

Technical Field

[0001] This disclosure relates to the field of thermal management technology, and more specifically, to a temperature control method and apparatus, a computer program product, and an electronic device. Background Technology

[0002] Currently, the rapid development of next-generation information technologies, represented by artificial intelligence, big data, and cloud computing, has driven an explosive growth in computing power. As the core of computing infrastructure, the power density of IT equipment within intelligent computing centers, particularly graphics processing unit (GPU) servers used for high-performance computing (HPC) and artificial intelligence training, is increasing at an unprecedented rate. This results in a high concentration of heat generated by the equipment. Traditional air-cooling solutions, which use air as the cooling medium, are gradually becoming inadequate in terms of heat dissipation efficiency, energy consumption, and space occupation, making it difficult to meet the heat dissipation requirements of next-generation high-density intelligent computing centers.

[0003] Against this backdrop, liquid cooling technology, which uses liquid as a cooling medium, has become an inevitable choice for addressing the challenges of high heat density due to its superior thermophysical properties (high specific heat capacity and high thermal conductivity). In practical deployments, to balance cost and heat dissipation efficiency, intelligent computing centers generally adopt a "hybrid cooling" mode that combines air cooling and liquid cooling.

[0004] However, existing control strategies for hybrid cooling scenarios generally suffer from the following technical bottlenecks: First, lag in response: Traditional control logic is mostly passive feedback control, which lags behind adjusting the supply temperature or flow rate of the cooling distribution unit based on the monitored return temperature or chip temperature. Second, energy redundancy: In practice, maintenance personnel often adopt a conservative strategy, setting the secondary supply temperature of the cooling distribution unit to a safe value far below actual needs. This "one-size-fits-all" static setting cannot be dynamically adjusted according to real-time fluctuations in IT load, resulting in huge and unnecessary energy waste during periods of low load.

[0005] Therefore, there is an urgent need to provide a new method for temperature control.

[0006] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] The purpose of this disclosure is to provide a temperature control method, a temperature control device, a computer program product, and an electronic device, thereby overcoming, to at least a certain extent, the problems of response lag and energy redundancy in temperature control of refrigeration equipment in related technologies.

[0008] According to one aspect of this disclosure, a temperature control method is provided, comprising: Acquire historical load data of computing power equipment and operating status data of cooling equipment, and predict the thermal load of computing power equipment based on the historical load data and the operating status data. The data on the heat load distribution of the refrigeration equipment is determined, and the temperature difference of the refrigeration equipment is predicted based on the distribution data, the operating parameters of the refrigeration equipment, and the heat load. Based on the temperature difference of the refrigeration equipment, the output temperature of the refrigeration equipment is obtained, a preset reference temperature and a preset reference humidity are acquired, and the output temperature is adjusted based on the preset reference temperature and preset reference humidity to obtain the target temperature; The temperature of the refrigeration equipment is controlled based on the target temperature.

[0009] According to one aspect of this disclosure, a temperature control device is provided, comprising: The heat load prediction module is used to acquire historical load data of computing equipment and operating status data of cooling equipment, and predict the heat load of computing equipment based on the historical load data and the operating status data. A refrigeration equipment temperature difference prediction module is used to determine the heat load sharing data of the refrigeration equipment, and predict the temperature difference of the refrigeration equipment based on the sharing data, the operating parameters of the refrigeration equipment, and the heat load. The output temperature compensation module is used to obtain the output temperature of the refrigeration equipment based on the temperature difference of the refrigeration equipment, acquire a preset reference temperature and a preset reference humidity, and adjust the output temperature based on the preset reference temperature and preset reference humidity to obtain the target temperature. A temperature control module is used to control the temperature of the refrigeration equipment based on the target temperature.

[0010] According to one aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the temperature control method described in any of the exemplary embodiments above.

[0011] According to one aspect of this disclosure, an electronic device is provided, comprising: A processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the temperature control method according to any of the exemplary embodiments described above by executing the executable instructions.

[0012] This disclosure provides a temperature control method that involves acquiring historical load data of a computing device and operating status data of a cooling device; predicting the heat load of the computing device based on the historical load data and the operating status data; determining the heat load sharing data of the cooling device; predicting the temperature difference of the cooling device based on the sharing data, the operating parameters of the cooling device, and the heat load; obtaining the output temperature of the cooling device based on the temperature difference; acquiring a preset reference temperature and a preset reference humidity; adjusting the output temperature based on the preset reference temperature and preset reference humidity to obtain a target temperature; and controlling the temperature of the cooling device based on the target temperature. On the one hand, the heat load of the computing equipment is predicted based on historical load data and operating status data of the cooling equipment. After obtaining the computing equipment's heat load, it is decomposed into data shared by the cooling equipment. Based on this shared data, the operating parameters of the cooling equipment, and the heat load, the temperature difference of the cooling equipment is predicted. This achieves the time-series superposition of heat load prediction and cooling equipment response, estimating the cooling capacity of the cooling equipment while predicting the heat load, contributing to a two-way balance between heat and cold, which is crucial for precise subsequent cooling supply. On the other hand, the output temperature of the cooling equipment is obtained based on the temperature difference. The output temperature is adjusted based on preset reference temperature and humidity to obtain the target temperature. By obtaining the temperature difference and ensuring the safety of the strategy, the output temperature is adjusted, enabling the temperature adjustment strategy to adaptively compensate for changes in the external environment, solving the energy redundancy problem in related technologies and reducing the energy consumption of the cooling equipment. Furthermore, after obtaining the target temperature, the temperature of the cooling equipment is directly controlled using the target temperature, solving the problem of lag in the response of the cooling equipment in related technologies and achieving active temperature control of the cooling equipment.

[0013] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0014] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure. It is obvious that the drawings described below are merely some embodiments of this disclosure, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0015] Figure 1 The flowchart illustrates a temperature control method according to an exemplary embodiment of the present disclosure.

[0016] Figure 2 This diagram schematically illustrates an overall architecture of temperature control according to an exemplary embodiment of the present disclosure.

[0017] Figure 3 The diagram illustrates the predictive performance of a computing load prediction model according to an example embodiment of the present disclosure.

[0018] Figure 4 This diagram illustrates the predictive performance of a temperature rise prediction model for a refrigeration device according to an example embodiment of the present disclosure.

[0019] Figure 5 This diagram schematically illustrates the probability distribution of the secondary side return liquid temperature of a liquid-cooled distribution unit according to an exemplary embodiment of the present disclosure.

[0020] Figure 6 A schematic diagram of a reinforcement learning environment according to an exemplary embodiment of the present disclosure is shown.

[0021] Figure 7 The flowchart illustrates a temperature control method according to an exemplary embodiment of the present disclosure.

[0022] Figure 8 A block diagram of a temperature control device according to an exemplary embodiment of the present disclosure is shown schematically.

[0023] Figure 9 An electronic device for implementing the temperature control method described above is illustrated according to an example embodiment of the present disclosure. Detailed Implementation

[0024] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.

[0025] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0026] Currently, the world is undergoing a profound digital transformation, with the rapid development of a new generation of information technologies, represented by artificial intelligence, big data, and cloud computing, driving an explosive growth in computing power. As the core of computing infrastructure, the IT equipment within intelligent computing centers, especially graphics processing unit (GPU) servers used for high-performance computing (HPC) and artificial intelligence training, is experiencing an unprecedented rate of increase in power density. This results in a high concentration of heat generated by computing equipment. Traditional air-cooling solutions, which use air as the cooling medium, are gradually becoming inadequate in terms of heat dissipation efficiency, energy consumption, and space occupation, making it difficult to meet the heat dissipation requirements of the new generation of high-density intelligent computing centers.

[0027] Against this backdrop, liquid cooling technology, using liquid as a cooling medium, has become an inevitable choice for addressing the challenges of high heat density due to its superior thermophysical properties (high specific heat capacity, high thermal conductivity). Direct-to-Chip liquid cooling technology, in particular, can deliver coolant directly to the most important heat-generating components inside the server (such as the CPU / GPU), achieving efficient and precise heat removal. In practical deployments, to balance cost and heat dissipation efficiency, intelligent computing centers generally adopt a "hybrid cooling" mode that combines air cooling and liquid cooling. In this mode, the liquid cooling system (usually composed of cooling distribution units (CDUs), piping, and cold plates) is responsible for handling the vast majority of the heat generated by high heat flux density components such as chips, while traditional air-cooled equipment such as CRACs or IRACs are responsible for removing heat from other components (such as memory and hard drives) within the rack, as well as residual ambient heat.

[0028] However, existing control strategies for hybrid heat dissipation scenarios generally suffer from the following technical bottlenecks: Lagging response poses a risk of thermal runaway: Traditional control logic is mostly passive feedback control, which adjusts the CDU's supply temperature or flow rate laggingly based on the monitored return liquid temperature or chip temperature. When the computing load (such as deep learning training tasks) increases instantaneously and non-linearly in a short period of time, heat will be generated rapidly. Due to the delay in the issuance of adjustment commands by the cooling system and the response of mechanical components, the cooling capacity cannot match the heat generation in time, which may cause a momentary "overshoot" of the chip temperature. This can range from mild overheating protection causing frequency throttling and affecting computing performance to severe hardware damage and reliability risks.

[0029] Redundant energy consumption and high operating costs: To mitigate the aforementioned risks, maintenance personnel often adopt a conservative strategy in practice, setting the secondary-side liquid supply temperature of the CDU to a safe value far below actual requirements. This means that primary-side refrigeration equipment, such as chillers providing cooling for the CDU, needs to operate under high load for extended periods to maintain excessively low liquid supply temperatures. This static, one-size-fits-all approach cannot be dynamically adjusted based on real-time fluctuations in IT load, resulting in significant and unnecessary energy waste during periods of low load, directly increasing the overall power efficiency (PUE) and operating costs (OPEX) of the intelligent computing center.

[0030] Lack of synergy leads to poor overall efficiency: In hybrid cooling architectures, air-cooling and liquid-cooling systems are typically managed and controlled as two independent subsystems, lacking effective information exchange and collaborative optimization. The system cannot take a holistic view, dynamically and economically allocating cooling tasks between air and liquid cooling based on real-time total computing load, indoor and outdoor temperature and humidity, and other conditions. For example, in seasons with lower outdoor temperatures, the natural cooling capacity of air cooling could be utilized more effectively, but the isolated liquid-cooling system fails to recognize this and continues to operate at high intensity, thus missing out on opportunities for overall energy savings.

[0031] To address one or more of the aforementioned issues, this exemplary embodiment first provides a temperature control method, which can be applied to a server. Of course, those skilled in the art can also run the method of this invention on other platforms as needed, and this exemplary embodiment does not impose any special limitations on this. (Reference) Figure 1 As shown, the temperature control method may include steps S110-S140: Step S110. Obtain historical load data of the computing power device and operating status data of the cooling device, and predict the thermal load of the computing power device based on the historical load data and the operating status data; Step S120. Determine the heat load sharing data of the refrigeration equipment, and predict the temperature difference of the refrigeration equipment based on the sharing data, the operating parameters of the refrigeration equipment, and the heat load; Step S130. Based on the temperature difference of the refrigeration equipment, obtain the output temperature of the refrigeration equipment, acquire a preset reference temperature and a preset reference humidity, and adjust the output temperature based on the preset reference temperature and preset reference humidity to obtain the target temperature; Step S140. Control the temperature of the refrigeration equipment based on the target temperature.

[0032] The above temperature control method involves acquiring historical load data of the computing device and operating status data of the cooling device; predicting the heat load of the computing device based on the historical load data and the operating status data; determining the heat load sharing data of the cooling device; predicting the temperature difference of the cooling device based on the sharing data, the operating parameters of the cooling device, and the heat load; obtaining the output temperature of the cooling device based on the temperature difference; acquiring a preset reference temperature and a preset reference humidity; adjusting the output temperature based on the preset reference temperature and preset reference humidity to obtain a target temperature; and controlling the temperature of the cooling device based on the target temperature. On the one hand, the heat load of the computing equipment is predicted based on historical load data and operating status data of the cooling equipment. After obtaining the computing equipment's heat load, it is decomposed into data shared by the cooling equipment. Based on this shared data, the operating parameters of the cooling equipment, and the heat load, the temperature difference of the cooling equipment is predicted. This achieves the time-series superposition of heat load prediction and cooling equipment response, estimating the cooling capacity of the cooling equipment while predicting the heat load, contributing to a two-way balance between heat and cold, which is crucial for precise subsequent cooling supply. On the other hand, the output temperature of the cooling equipment is obtained based on the temperature difference. The output temperature is adjusted based on preset reference temperature and humidity to obtain the target temperature. By obtaining the temperature difference and ensuring the safety of the strategy, the output temperature is adjusted, enabling the temperature adjustment strategy to adaptively compensate for changes in the external environment, solving the energy redundancy problem in related technologies and reducing the energy consumption of the cooling equipment. Furthermore, after obtaining the target temperature, the temperature of the cooling equipment is directly controlled using the target temperature, solving the problem of lag in the response of the cooling equipment in related technologies and achieving active temperature control of the cooling equipment.

[0033] The following provides a detailed explanation and description of each step involved in the temperature control method of the exemplary embodiments of this disclosure.

[0034] First, the overall architecture of the embodiments disclosed herein will be explained and described. Specifically, refer to... Figure 2As shown, the system includes an intelligent computing center, a control system, and a controller. The intelligent computing center includes computing equipment, environmental parameters, a liquid cooling distribution unit, and inter-row air conditioning. The control system includes a data acquisition module, a prediction module, a decision optimization module, and a control execution module. The data acquisition module collects data from the intelligent computing center in real time, including the power of the computing equipment, the operating status of the liquid cooling distribution unit, the operating status of the inter-row air conditioning, and indoor and outdoor environmental parameters. The prediction module predicts the heat load of the computing equipment at a preset time based on the data collected by the data acquisition module. After obtaining the heat load, this heat load, along with the data collected by the data acquisition module, is input to the decision optimization module, which generates a target temperature. After generating the target temperature, the target temperature is sent to the control execution module, which issues control commands to the physical world controller. This controller includes a liquid cooling distribution unit controller and an air conditioning controller. The controller changes the state of the physical world according to the target temperature, and the new state data is collected, thus forming a complete and continuously optimized data flow closed loop.

[0035] Next, steps S110-S140 will be further explained and described.

[0036] In step S110, historical load data of the computing device and operating status data of the cooling device are obtained, and the thermal load of the computing device is predicted based on the historical load data and the operating status data.

[0037] The computing power equipment can be IT equipment included in the intelligent computing center. In this example embodiment, the specific IT equipment is not limited. The historical load data can be the load data of the computing power equipment within a preset historical time period. The time granularity of this preset time period can be 5 minutes or 10 minutes; in this example embodiment, it is not specifically limited. The load data can be the total power consumption of the computing power equipment. When the computing power equipment is a GPU server cluster, the load data can be the total power consumption of the GPU server cluster. The cooling equipment can include a liquid cooling distribution unit and an air cooling unit. The operating status of the cooling equipment can include the operating status data of the liquid cooling distribution unit and the operating status data of the air cooling unit. The operating status data of the liquid cooling distribution unit can include the secondary side liquid supply temperature of the liquid cooling distribution unit. Secondary side return liquid temperature Secondary side mass flow rate The operating status data of the liquid-cooled distribution unit, such as secondary side filter pressure and water pump speed, are not specifically limited in this example embodiment. The operating status data of the air-cooled equipment may include the set temperature, actual supply air temperature, actual return air temperature, and air volume, etc., and are not specifically limited in this example embodiment.

[0038] After obtaining historical load data of the computing power equipment and operating status data of the cooling equipment, the obtained historical load data and operating status data can be input into the computing power load prediction model to obtain the thermal load of the computing power equipment. The step of predicting the thermal load of the computing power equipment based on the historical load data and the operating status data includes: The total power consumption of the computing device, the operating status data of the liquid cooling distribution unit, the operating status data of the air cooling equipment, and the environmental data are input into the computing load prediction model to obtain the thermal load of the computing device.

[0039] Specifically, the total power consumption of computing devices, the operating status data of liquid-cooled distribution units, the operating status data of air-cooled devices, and environmental data within a historical preset time period are input into the computing load prediction model to obtain the thermal load of computing devices within a future preset time period. The environmental data includes both internal and external environmental data of the intelligent computing center. Internal environmental data can include the temperature and humidity of key monitoring points within the intelligent computing center; external environmental data includes the dry-bulb temperature and humidity of the outdoor environment. In this example embodiment, no specific limitations are made on the internal and external environmental data of the intelligent computing center.

[0040] For example, the total power consumption of the computing device and the secondary-side liquid supply temperature of the liquid cooling distribution unit within the past 10 minutes will be obtained. Secondary side return liquid temperature Secondary side mass flow rate The secondary filter pressure, water pump speed, set temperature, actual supply air temperature, actual return air temperature, and air volume of the air-cooled equipment, as well as environmental data inside and outside the intelligent computing center, are input into the computing load prediction model to obtain the thermal load of the computing equipment in the next 5 minutes.

[0041] In one exemplary embodiment, the computing load prediction model is a deep learning model based on long short-term memory networks and attention mechanisms; The method further includes: when the frequency at which the average absolute error of the computing load prediction model exceeds the first preset threshold reaches a preset frequency, the computing load prediction model is updated.

[0042] Specifically, the computing load prediction model is a deep learning model that integrates a Long Short-Term Memory (LSTM) network and an attention mechanism. This model takes into account the time-series data of the total power of computing devices over a past time window (e.g., 10 sampling points, or 50 minutes), captures its long-term dependencies through an LSTM layer, and then automatically identifies and weights key time node features using the attention mechanism to predict the thermal load of computing devices within one or more future time steps (e.g., the next 10 minutes). To ensure model accuracy, this computing load prediction model incorporates a dynamic retraining mechanism: when the model's mean absolute error (MAE) exceeds a first preset threshold at a preset frequency, an online model update is triggered. The first preset threshold can be 5 kW, and the preset frequency can be two consecutive occurrences exceeding the preset threshold. In this example embodiment, the first preset threshold and the preset frequency are not specifically limited.

[0043] Figure 3 The graph illustrates the predictive performance of the computing load prediction model. Taking GPU real-time power prediction as an example, the horizontal axis represents the time step, and the vertical axis represents the real-time power (kW). The graph contains two types of curves: actual values ​​(represented by solid lines) and predicted values ​​(represented by dashed lines). It can be seen from the graph that the predicted values ​​perform well during the load increase phase, ensuring the safety of the control strategy based on this prediction, and the model's average error is less than 5kW.

[0044] In step S120, the heat load sharing data of the refrigeration equipment is determined, and the temperature difference of the refrigeration equipment is predicted based on the sharing data, the operating parameters of the refrigeration equipment, and the heat load.

[0045] The thermal load of the computing equipment is obtained. Then, based on the thermal balance model and dynamic thermal correlation analysis, the heat load sharing data of the refrigeration equipment can be determined. After obtaining the sharing data, the sharing data, the operating parameters of the refrigeration equipment, and the heat load of the computing equipment can be input into the prediction model to obtain the temperature difference of the refrigeration equipment.

[0046] In one exemplary embodiment, determining the data on the heat load sharing of the refrigeration device includes: The heat load is decomposed into the liquid cooling load of the liquid cooling distribution unit and the air cooling load of the air cooling equipment.

[0047] Specifically, after obtaining the thermal load of the computing equipment, the thermal load of the computing equipment can be dynamically decomposed to obtain the liquid cooling load of the liquid cooling distribution unit. and air-cooled refrigeration load of air-cooled equipment ;in, , , ;in, These are the heat dissipation efficiency of the liquid-cooled distribution unit and the heat dissipation efficiency of the air-cooled equipment, respectively. The liquid-cooled cooling load of this liquid-cooled distribution unit... and air-cooled refrigeration load of air-cooled equipment This refers to the data on the heat load sharing of refrigeration equipment. By calculating the dynamic correlation between the return air temperature and return liquid temperature of air-cooled and liquid-cooled equipment and changes in heat load, the heat load sharing of the two types of refrigeration equipment at a certain time period can be deduced. This dynamic correlation analysis provides a simple, feasible, and adaptive air-liquid synergy mode, realizing the synergy between the liquid-cooled distribution unit and the air-cooled equipment, and improving overall efficiency.

[0048] In one exemplary embodiment, predicting the temperature difference of the refrigeration equipment based on the sharing data, the operating parameters of the refrigeration equipment, and the heat load includes: Heat transfer characteristics are generated based on the shared data. The heat transfer characteristics, the operating parameters of the refrigeration equipment, and the heat load are input into the temperature rise prediction model of the refrigeration equipment to obtain the temperature difference of the refrigeration equipment.

[0049] Specifically, after obtaining the heat transfer data, a heat transfer characteristic can be generated based on the data. That is, for liquid-cooled refrigeration equipment, a strongly correlated heat transfer characteristic can be constructed based on the liquid cooling heat dissipation calculation formula. The heat transfer characteristics, operating parameters of the refrigeration equipment, and heat load of the computing equipment are input into the refrigeration equipment temperature rise prediction model to obtain the temperature difference of the refrigeration equipment. By introducing this heat transfer characteristic, the model can better understand the physical process of heat transfer, thus improving the prediction accuracy. It should be noted that when predicting the temperature difference of the refrigeration equipment using the refrigeration equipment temperature rise prediction model, data corresponding to the liquid-cooled refrigeration unit can be input into the model to obtain the temperature difference of the liquid-cooled refrigeration unit; data corresponding to the air-cooled equipment can be input into the model to obtain the temperature difference of the air-cooled equipment. The operating parameters of the refrigeration equipment may include the inlet temperature, flow rate, pressure, pump speed, etc., but in this example embodiment, the operating parameters of the refrigeration equipment are not specifically limited.

[0050] For example, taking a liquid-cooled refrigeration unit as an example, the liquid supply temperature on the secondary side of the liquid-cooled refrigeration unit can be used as a reference. Secondary side return liquid temperature And the formula for calculating liquid cooling heat dissipation. Construct heat transfer characteristics, in which, This represents the coefficient of performance (COP) of the liquid-cooled refrigeration unit. After obtaining the heat transfer characteristics, these characteristics, the heat load of the computing device, and the operating parameters of the refrigeration equipment can be input into the refrigeration equipment temperature rise prediction model to obtain the temperature difference between the supply and return liquid temperatures on the secondary side of the liquid-cooled refrigeration unit. .

[0051] In one exemplary embodiment, the temperature rise prediction model for the refrigeration equipment is constructed based on a random forest regression algorithm. To improve the prediction accuracy of the model, it can be updated when the root mean square error of the model exceeds a second preset threshold. The second preset threshold can be 8%, but this example embodiment does not impose a specific limitation on it.

[0052] Figure 4 The figure illustrates the predictive performance of a temperature rise prediction model for refrigeration equipment. Taking the temperature rise prediction of a liquid-cooled distribution unit as an example, the horizontal axis represents the time step, and the vertical axis represents the temperature rise (degrees Celsius). The figure contains two curves: the actual value (solid line) and the predicted value (dashed line). As can be seen from the figure, the predicted value curve and the actual value curve highly overlap, proving the accuracy of the model. The performance indicators in the figure (MSE (mean squared error): 0.2392, RMSE (root mean squared error): 0.4891) are... (Determination coefficient: 0.8918) also quantitatively demonstrates the superior performance of the model.

[0053] In this embodiment, a dual-model superposition prediction mechanism is used, namely, the computing load prediction model and the refrigeration equipment temperature rise prediction model are superimposed in time according to the control frequency. This enables the subsequent control of the refrigeration equipment to have the predictability of computing load and the predictability of refrigeration equipment response. This achieves the orderly superposition of heat load prediction and refrigeration response. While predicting the heat load, it also estimates the supply of cooling capacity, which helps to achieve a two-way balance between heat and cold and is the key to accurate subsequent refrigeration supply.

[0054] In step S130, the output temperature of the refrigeration equipment is obtained based on the temperature difference of the refrigeration equipment, a preset reference temperature and a preset reference humidity are obtained, and the output temperature is adjusted based on the preset reference temperature and the preset reference humidity to obtain the target temperature.

[0055] After obtaining the temperature difference of the refrigeration equipment, the output temperature of the refrigeration equipment can be obtained based on the baseline control strategy of thermodynamic mechanism according to the temperature difference of the refrigeration equipment.

[0056] In one exemplary embodiment, obtaining the output temperature of the refrigeration device based on the temperature difference of the refrigeration device includes: The historical feedback temperature of the refrigeration equipment is obtained, and the historical feedback temperature is statistically analyzed to obtain the safe temperature of the refrigeration equipment. The output temperature of the refrigeration equipment is obtained based on the safe temperature and the temperature difference.

[0057] Specifically, when the temperature difference of the refrigeration equipment is obtained Then, the output temperature of the refrigeration equipment can be pre-calculated based on the offline prediction model. ,in, The safe temperature is determined based on historical feedback temperatures from the refrigeration equipment. Specifically, when the refrigeration equipment is air-cooled, the feedback temperature can be the return air temperature; when the refrigeration equipment is a liquid-cooled distribution unit, the feedback temperature can be the secondary side return liquid temperature. Statistical analysis of the historical feedback temperatures of the refrigeration equipment yields its safe temperature.

[0058] For example, a probability distribution can be generated based on the historical feedback temperature of the refrigeration equipment, and a safe temperature can be determined based on this probability distribution. (Reference) Figure 5 The probability distribution of the secondary side return liquid temperature of the liquid-cooled distribution unit shown shows that the probability of the secondary side return liquid temperature being in the range of [35℃, 37℃] exceeds 50%. Therefore, we can take the average value and determine 36℃ as the safe temperature.

[0059] In one exemplary embodiment, the step of acquiring a preset reference temperature and a preset reference humidity, and adjusting the output temperature based on the preset reference temperature and preset reference humidity to obtain a target temperature, includes: The temperature difference is obtained based on the outdoor temperature of the intelligent computing center where the computing power equipment is located and the preset reference temperature; The humidity difference is obtained based on the humidity of the outdoor environment of the intelligent computing center where the computing power device is located and the preset benchmark humidity. The compensated temperature of the output temperature is obtained based on the shared data, the first correction factor corresponding to the shared data, the temperature difference and the second correction factor corresponding to the temperature difference, the humidity difference and the third correction factor corresponding to the humidity difference; The target temperature is obtained based on the output temperature and the compensation temperature.

[0060] Specifically, to further improve the precision of temperature control, the final output temperature sent to the refrigeration equipment can be fine-tuned using an online reinforcement learning-based environment compensation model. First, a preset reference temperature is obtained. Preset baseline humidity The outdoor temperature and humidity of the intelligent computing center are based on a preset baseline temperature. The temperature difference between the outdoor environment and the ambient temperature of the intelligent computing center is obtained, based on a preset baseline humidity. The humidity difference is obtained from the humidity of the outdoor environment of the intelligent computing center. Among these, the preset reference temperature... It can be set to 25℃ and preset reference humidity. The value can be 60%, but this is not specifically limited in this example embodiment. Next, based on the first correction factor corresponding to the shared data, the second correction factor corresponding to the temperature difference, and the third correction factor corresponding to the humidity difference obtained through reinforcement learning, the target temperature of the refrigeration equipment is obtained based on the output temperature, the first correction factor, the shared data, the temperature difference, the second correction factor, the humidity difference, and the third correction factor. Wherein, when the refrigeration equipment is a liquid-cooled distribution unit, the target temperature can be expressed as... When the refrigeration equipment is air-cooled, .in, The first correction factor for the liquid cooling distribution unit. The second correction factor for the liquid cooling distribution unit. This is the third correction factor for the liquid cooling distribution unit; The first correction factor for air-cooled equipment. The second correction factor for air-cooled equipment. This is the third correction factor for air-cooled equipment.

[0061] In one exemplary embodiment, the first correction factor, the second correction factor, and the third correction factor are determined based on reinforcement learning.

[0062] For details, please refer to Figure 6 The reinforcement learning environment shown employs a Proximal Policy Optimization (PPO) algorithm as its core, interacting with the intelligent computing center environment and learning based on a reward and punishment mechanism. The state space is a multi-dimensional vector representing the environmental state observed by the agent at each decision-making moment. It includes: GPU power, the current state of the liquid cooling distribution unit (inlet temperature, flow rate, pressure, pump speed, etc.), the state of air-cooled equipment such as the inter-row air conditioning system (on / off status, set temperature, set humidity, return air temperature, return air humidity, etc.), and the outdoor ambient temperature. Outdoor humidity etc.; Action space: The actions that the agent can perform are used as a correction factor. , , Assign a value, the value of which is... , , Reward function: The basis for the agent's learning and optimization.

[0063] In one exemplary embodiment, the reward function of the reinforcement learning is determined based on the instantaneous power consumption of the cooling device and the degree of deviation between the actual temperature and the safe temperature of the computing device.

[0064] Specifically, in order to achieve the best balance between ensuring the safety of refrigeration equipment and realizing energy conservation and consumption reduction, the reward function is designed as a negative penalty function:

[0065] The first item is the temperature penalty item. The first item indicates the degree of deviation between the actual temperature of the chip in the computing device and the safe temperature; the second item is the energy consumption penalty. This represents the instantaneous total power consumption of the liquid-cooled distribution unit and the air-cooled equipment. and These are weighting coefficients that can be dynamically adjusted by operations and maintenance personnel based on actual needs, used to balance the two objectives of "safety" and "energy saving." The goal of the intelligent agent is to learn a strategy that maximizes long-term cumulative rewards through continuous trial and error.

[0066] Furthermore, before deployment, intelligent simulation interaction can be selectively performed to accelerate the convergence of the reinforcement learning model. If a simulation environment is unavailable, and since the reinforcement learning model is primarily used to correct baseline policies—meaning the policy's safety and energy efficiency are already guaranteed to a certain extent—a fallback strategy can be set to allow the model to learn and converge online. In actual operation, the decision-making module calculates the optimal liquid supply temperature of the liquid cooling distribution unit and the air supply temperature setpoint of the air-cooling equipment through the policy network based on real-time collected status information and the output of the prediction module. For example, when a significant increase in future load is predicted, the setpoints of the cooling equipment can be adjusted in advance to utilize the pre-cooling capacity of the cooling equipment, thereby smoothing out fluctuations in chip temperature.

[0067] In step S140, the temperature of the refrigeration device is controlled based on the target temperature.

[0068] Once the target temperature of the refrigeration equipment is obtained, the target temperature can be sent to the refrigeration equipment via control commands, so that the refrigeration equipment can adjust its liquid delivery or air delivery temperature according to the received target temperature.

[0069] The following will combine Figure 7The temperature control method of this application is further explained and described. The thermal load of the computing equipment is obtained based on the computing load prediction model. This thermal load is input into the cooling load distribution model, which obtains the liquid cooling load of the liquid cooling distribution unit and the air cooling load of the air-cooled equipment corresponding to the thermal load. The liquid cooling load of the liquid cooling distribution unit, the air cooling load of the air-cooled equipment, and the thermal load are input into the temperature rise prediction model of the refrigeration equipment to obtain the temperature difference of the refrigeration equipment. A baseline control strategy based on thermodynamic mechanisms determines the safe temperature of the refrigeration equipment. Based on the safe temperature and the temperature difference of the refrigeration equipment, the output temperature of the refrigeration equipment is obtained. After obtaining the output temperature, the output temperature can be adjusted based on a reinforcement learning-based environmental compensation correction strategy to obtain the target temperature. The correction factor involved in the environmental compensation correction strategy is obtained based on online reinforcement learning.

[0070] The temperature control method provided by the exemplary embodiments of this disclosure has at least the following advantages: On the one hand, the heat load of the computing device is predicted based on the historical load data of the computing device and the operating status data of the cooling device. After obtaining the computing device, the heat load is decomposed into the sharing data of the cooling device. Based on the sharing data, the operating parameters of the cooling device, and the heat load, the temperature difference of the cooling device is predicted. This realizes the time-series superposition of heat load prediction and cooling device response. While predicting the heat load, the cooling capacity of the cooling device is also estimated, which helps to achieve a two-way balance between heat and cold and is the key to subsequent accurate cooling supply. On the other hand, the output temperature of the cooling device is obtained based on the temperature difference of the cooling device. The output temperature is adjusted based on a preset reference temperature and a preset reference humidity to obtain the target temperature. By obtaining the temperature difference, the output temperature is adjusted under the premise of ensuring the safety of the strategy. This enables the temperature adjustment strategy to adaptively compensate for changes in the external environment, solves the problem of energy redundancy in related technologies, and reduces the energy consumption of the cooling device. Furthermore, after obtaining the target temperature, the temperature of the cooling device is directly controlled by the target temperature, which solves the problem of lag in the response of the cooling device in related technologies and realizes active control of the temperature of the cooling device.

[0071] This disclosure also provides a temperature control device, with reference to... Figure 8 As shown, it may include: a heat load prediction module 810, a refrigeration equipment temperature difference prediction module 820, an output temperature compensation module 830, and a temperature regulation module 840. Wherein: The heat load prediction module 810 is used to acquire historical load data of computing equipment and operating status data of cooling equipment, and predict the heat load of computing equipment based on the historical load data and the operating status data. The refrigeration equipment temperature difference prediction module 820 is used to determine the heat load sharing data of the refrigeration equipment, and predict the temperature difference of the refrigeration equipment based on the sharing data, the operating parameters of the refrigeration equipment and the heat load. The output temperature compensation module 830 is used to obtain the output temperature of the refrigeration equipment based on the temperature difference of the refrigeration equipment, acquire a preset reference temperature and a preset reference humidity, and adjust the output temperature based on the preset reference temperature and preset reference humidity to obtain a target temperature. Temperature regulation module 840 is used to control the temperature of the refrigeration equipment based on the target temperature.

[0072] The specific details of each module in the above-mentioned temperature control device have been described in detail in the corresponding temperature control method, so they will not be repeated here.

[0073] In one exemplary embodiment of this disclosure, the historical load data is the total power consumption of the computing device, and the operating status data of the cooling device includes the operating status data of the liquid cooling distribution unit and the operating status data of the air cooling device.

[0074] In one exemplary embodiment of this disclosure, the heat load prediction module includes: The data acquisition and calculation module is used to input the total power consumption of the computing power device, the operating status data of the liquid cooling distribution unit, the operating status data of the air cooling device, and environmental data into the computing power load prediction model to obtain the thermal load of the computing power device.

[0075] In one exemplary embodiment of this disclosure, the computing load prediction model is a deep learning model based on long short-term memory networks and attention mechanisms, and the hot load prediction module includes: The model update module is used to trigger an update of the computing load prediction model when the frequency at which the average absolute error of the prediction model exceeds a first preset threshold reaches a preset frequency.

[0076] In one exemplary embodiment of this disclosure, the environmental data includes the temperature and humidity of key measuring points in the intelligent computing center where the computing power device is located, as well as the temperature and humidity of the outdoor environment of the intelligent computing center.

[0077] In one exemplary embodiment of this disclosure, the refrigeration equipment temperature difference prediction module includes: The heat load decomposition module is used to decompose the heat load into the liquid cooling load of the liquid cooling distribution unit and the air cooling load of the air cooling equipment.

[0078] In one exemplary embodiment of this disclosure, the refrigeration equipment temperature difference prediction module includes: The temperature difference prediction module is used to generate heat transfer characteristics based on the shared data, and input the heat transfer characteristics, the operating parameters of the refrigeration equipment, and the heat load into the temperature rise prediction model of the refrigeration equipment to obtain the temperature difference of the refrigeration equipment.

[0079] In one exemplary embodiment of this disclosure, the temperature rise prediction model for the refrigeration equipment is constructed based on the random forest regression algorithm.

[0080] In one exemplary embodiment of this disclosure, the output temperature compensation module includes: The safe temperature determination module is used to acquire the historical feedback temperature of the refrigeration equipment, perform statistical analysis on the historical feedback temperature, and obtain the safe temperature of the refrigeration equipment. The output temperature determination module is used to obtain the output temperature of the refrigeration equipment based on the safe temperature and the temperature difference.

[0081] In one exemplary embodiment of this disclosure, the output temperature compensation module includes: The temperature difference determination module is used to obtain the temperature difference based on the outdoor temperature of the intelligent computing center where the computing power device is located and a preset reference temperature; The humidity difference determination module is used to obtain the humidity difference based on the humidity of the outdoor environment of the intelligent computing center where the computing power device is located and a preset benchmark humidity. The compensation temperature calculation module is used to obtain the compensation temperature of the output temperature based on the liquid cooling load, a first correction factor corresponding to the liquid cooling load, the temperature difference and a second correction factor corresponding to the temperature difference, the humidity difference and a third correction factor corresponding to the humidity difference. The target temperature determination module is used to obtain the target temperature based on the output temperature and the compensation temperature.

[0082] In one exemplary embodiment of this disclosure, the first correction factor, the second correction factor, and the third correction factor are determined based on reinforcement learning.

[0083] In one exemplary embodiment of this disclosure, the reward function of the reinforcement learning is determined based on the instantaneous power consumption of the cooling device and the degree of deviation between the actual temperature and the safe temperature of the computing device.

[0084] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0085] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0086] In an exemplary embodiment of this disclosure, an electronic device capable of implementing the above-described method is also provided.

[0087] Those skilled in the art will understand that various aspects of this disclosure can be implemented as a system, method, or program product. Therefore, various aspects of this disclosure can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, collectively referred to herein as a "circuit," "module," or "system."

[0088] The following reference Figure 9 To describe an electronic device 900 according to such an embodiment of the present disclosure. Figure 9 The electronic device 900 shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments disclosed herein.

[0089] like Figure 9 As shown, the electronic device 900 is manifested in the form of a general-purpose computing device. The components of the electronic device 900 may include, but are not limited to: at least one processing unit 910, at least one storage unit 920, a bus 930 connecting different system components (including storage unit 920 and processing unit 910), and a display unit 940.

[0090] The storage unit stores program code that can be executed by the processing unit 910, causing the processing unit 910 to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure. For example, the processing unit 910 can perform actions such as... Figure 1The steps shown are as follows: S110: Obtain historical load data of the computing power device and operating status data of the cooling device; predict the heat load of the computing power device based on the historical load data and the operating status data; S120: Determine the heat load sharing data of the cooling device; predict the temperature difference of the cooling device based on the sharing data, the operating parameters of the cooling device, and the heat load; S130: Obtain the output temperature of the cooling device based on the temperature difference; obtain a preset reference temperature and a preset reference humidity; adjust the output temperature based on the preset reference temperature and preset reference humidity to obtain a target temperature; S140: Control the temperature of the cooling device based on the target temperature.

[0091] Storage unit 920 may include readable media in the form of volatile storage units, such as random access memory (RAM) 9201 and / or cache memory 9202, and may further include read-only memory (ROM) 9203.

[0092] The storage unit 920 may also include a program / utility 9204 having a set (at least one) program module 9205, such program module 9205 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0093] Bus 930 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0094] Electronic device 900 can also communicate with one or more external devices 1000 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 900, and / or with any device that enables electronic device 900 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 950. Furthermore, electronic device 900 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 960. As shown, network adapter 960 communicates with other modules of electronic device 900 via bus 930. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 900, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0095] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0096] In exemplary embodiments of this disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the methods described above is stored. In some possible implementations, various aspects of this disclosure may also be implemented as a program product including program code that, when the program product is run on a terminal device, causes the terminal device to perform the steps of the various exemplary embodiments of this disclosure described in the "Exemplary Methods" section above.

[0097] The program product for implementing the above-described method according to embodiments of the present disclosure may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0098] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0099] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in conjunction with an instruction execution system, apparatus, or device.

[0100] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0101] Program code for performing the operations of this disclosure can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0102] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of this disclosure and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0103] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

Claims

1. A temperature control method, characterized in that, The method includes: Acquire historical load data of computing power equipment and operating status data of cooling equipment, and predict the thermal load of computing power equipment based on the historical load data and the operating status data. The data on the heat load distribution of the refrigeration equipment is determined, and the temperature difference of the refrigeration equipment is predicted based on the distribution data, the operating parameters of the refrigeration equipment, and the heat load. Based on the temperature difference of the refrigeration equipment, the output temperature of the refrigeration equipment is obtained, a preset reference temperature and a preset reference humidity are acquired, and the output temperature is adjusted based on the preset reference temperature and preset reference humidity to obtain the target temperature; The temperature of the refrigeration equipment is controlled based on the target temperature.

2. The method according to claim 1, characterized in that, The historical load data is the total power consumption of the computing device, and the operating status data of the cooling device includes the operating status data of the liquid cooling distribution unit and the operating status data of the air cooling device.

3. The method according to claim 2, characterized in that, The step of predicting the thermal load of the computing device based on the historical load data and the operating status data includes: The total power consumption of the computing device, the operating status data of the liquid cooling distribution unit, the operating status data of the air cooling equipment, and the environmental data are input into the computing load prediction model to obtain the thermal load of the computing device.

4. The method according to claim 3, characterized in that, The computing load prediction model is a deep learning model based on long short-term memory networks and attention mechanisms. The method further includes: when the frequency at which the average absolute error of the computing load prediction model exceeds the first preset threshold reaches a preset frequency, the computing load prediction model is updated.

5. The method according to claim 3, characterized in that, The environmental data includes the temperature and humidity of key measuring points in the intelligent computing center where the computing equipment is located, as well as the temperature and humidity of the outdoor environment of the intelligent computing center.

6. The method according to claim 2, characterized in that, The determination of the heat load sharing data of the refrigeration equipment includes: The heat load is decomposed into the liquid cooling load of the liquid cooling distribution unit and the air cooling load of the air cooling equipment.

7. The method according to claim 1, characterized in that, The step of predicting the temperature difference of the refrigeration equipment based on the sharing data, the operating parameters of the refrigeration equipment, and the heat load includes: Heat transfer characteristics are generated based on the shared data. The heat transfer characteristics, the operating parameters of the refrigeration equipment, and the heat load are input into the temperature rise prediction model of the refrigeration equipment to obtain the temperature difference of the refrigeration equipment.

8. The method according to claim 7, characterized in that, The temperature rise prediction model for the refrigeration equipment is constructed based on the random forest regression algorithm.

9. The method according to claim 1, characterized in that, The step of obtaining the output temperature of the refrigeration equipment based on the temperature difference of the refrigeration equipment includes: The historical feedback temperature of the refrigeration equipment is obtained, and the historical feedback temperature is statistically analyzed to obtain the safe temperature of the refrigeration equipment. The output temperature of the refrigeration equipment is obtained based on the safe temperature and the temperature difference.

10. The method according to claim 1, characterized in that, The process of obtaining a preset reference temperature and a preset reference humidity, and adjusting the output temperature based on the preset reference temperature and preset reference humidity to obtain a target temperature, includes: The temperature difference is obtained based on the outdoor temperature of the intelligent computing center where the computing power equipment is located and the preset reference temperature; The humidity difference is obtained based on the humidity of the outdoor environment of the intelligent computing center where the computing power device is located and the preset benchmark humidity. The compensated temperature for the output temperature is obtained based on the liquid cooling load, the first correction factor corresponding to the liquid cooling load, the temperature difference and the second correction factor corresponding to the temperature difference, the humidity difference and the third correction factor corresponding to the humidity difference. The target temperature is obtained based on the output temperature and the compensation temperature.

11. The method according to claim 10, characterized in that, The first correction factor, the second correction factor, and the third correction factor are determined based on reinforcement learning.

12. The method according to claim 11, characterized in that, The reward function for reinforcement learning is determined based on the instantaneous power consumption of the cooling device and the degree of deviation between the actual temperature and the safe temperature of the computing device.

13. A temperature control device, characterized in that, The device includes: The heat load prediction module is used to acquire historical load data of computing equipment and operating status data of cooling equipment, and predict the heat load of computing equipment based on the historical load data and the operating status data. A refrigeration equipment temperature difference prediction module is used to determine the heat load sharing data of the refrigeration equipment, and predict the temperature difference of the refrigeration equipment based on the sharing data, the operating parameters of the refrigeration equipment, and the heat load. The output temperature compensation module is used to obtain the output temperature of the refrigeration equipment based on the temperature difference of the refrigeration equipment, acquire a preset reference temperature and a preset reference humidity, and adjust the output temperature based on the preset reference temperature and preset reference humidity to obtain the target temperature. A temperature control module is used to control the temperature of the refrigeration equipment based on the target temperature.

14. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method according to any one of claims 1 to 12.

15. An electronic device, characterized in that, include: processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method of any one of claims 1-12 by executing the executable instructions.

Images