Intelligent heat dissipation control method and system of server

Abstract

The invention discloses an intelligent heat dissipation control method and system of a server, and relates to the technical field of server heat dissipation control, and the intelligent heat dissipation control method and system of the server comprises the steps: obtaining multi-source data, generating fusion data, reconstructing temperature distribution, predicting a thermal behavior, and dynamically generating a control instruction. The problems of response lag and energy waste of a traditional heat dissipation technology are solved, multi-source data can be fused, thermal behavior changes can be predicted, and a heat dissipation control instruction is dynamically generated, so that the heat dissipation efficiency and the energy utilization rate are improved, local hot spots are avoided, and the operation stability of a server is improved.

Description

Technical Field

[0001] This application relates to the field of server heat dissipation control technology, and in particular to intelligent heat dissipation control methods and systems for servers. Background Technology

[0002] In modern data centers and high-performance computing environments, thermal management of server systems is becoming increasingly prominent. With continuous upgrades to processor architectures and the increasing complexity of computing tasks, the heat generated during server operation is growing exponentially. Traditional cooling technologies primarily rely on forced fan ventilation or liquid cooling circulation systems. These solutions typically employ static threshold control mechanisms, such as triggering fan acceleration or increasing cooling medium flow when a temperature sensor detects a fixed temperature point. However, such fixed strategies lack the ability to perceive the dynamic evolution of the internal thermal field of the server and cannot adapt to real-time fluctuations in workload. In high-density deployment scenarios, such as cloud computing data centers, server clusters often face sudden high-load tasks, leading to the rapid formation of local hotspots. Because traditional methods rely solely on single temperature readings for decision-making, failing to integrate multi-dimensional data such as pressure distribution and power consumption changes, they are prone to delayed thermal response or overcooling. This not only leads to server performance degradation or even downtime risks but also significantly increases energy consumption, as the cooling system operates at high power continuously when not in use. Furthermore, the complex layout of internal server components and the significant structural influence on heat conduction paths make it difficult for fixed strategies to accurately match actual thermal behavior, further reducing cooling efficiency.

[0003] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main purpose of this application is to provide an intelligent heat dissipation control method and system for servers, which aims to improve heat dissipation efficiency and energy utilization.

[0005] To achieve the above objectives, this application proposes an intelligent heat dissipation control method for servers, the method comprising: The system acquires temperature sensor data, pressure sensor data, and power consumption measurement data from the server system, and inputs these data into a data fusion unit for processing to generate multi-source fused data. Based on the temperature sensor data, the current temperature distribution data of the server is generated through a thermal field reconstruction algorithm; The current temperature distribution data, the pressure sensor data, and the power consumption measurement data are input into the digital twin model, and the server workload data is input into the digital twin model. The predicted thermal behavior data of the server is generated through the digital twin model. The predicted thermal behavior data, the current temperature distribution data, and the multi-source fusion data are input into the control strategy generation unit, and a heat dissipation control instruction set is generated through the control strategy generation unit. The heat dissipation control instruction set is sent to the heat dissipation actuator for execution in order to control the heat dissipation of the server.

[0006] In one embodiment, the steps of acquiring temperature sensor data, pressure sensor data, and power consumption measurement data of a server system, and inputting the temperature sensor data, pressure sensor data, and power consumption measurement data into a data fusion unit for processing to generate multi-source fused data include: Temperature sensor data from the server is collected via a temperature sensor network. The system's pressure sensor data is acquired through a pressure sensor array; Power consumption data of multiple components are collected using a power consumption monitoring device; The collected temperature sensor data, pressure sensor data, and power consumption measurement data are input into the data fusion unit for time alignment and data cleaning to generate the multi-source fused data.

[0007] In one embodiment, the step of generating the current temperature distribution data of the server based on the temperature sensor data using a thermal field reconstruction algorithm includes: Extract multiple spatially distributed temperature measurement values ​​from the temperature sensor data; Based on a preset physical model of heat conduction, the temperature measurement values ​​are spatially interpolated to generate initial temperature field data. Based on the internal structure distribution of the server, the initial temperature field data is subjected to thermal field correction processing to generate the current temperature distribution data.

[0008] In one embodiment, the steps of inputting the current temperature distribution data, the pressure sensor data, and the power consumption measurement data into a digital twin model, inputting server workload data into the digital twin model, and generating predicted thermal behavior data of the server through the digital twin model include: The current temperature distribution data, the pressure sensor data, and the power consumption measurement data are input into the digital twin model; The server workload data is input into the digital twin model; Based on the digital twin model, thermodynamic simulation calculations are performed on the current temperature distribution data, the pressure sensor data, the power consumption measurement data, and the server workload data to generate current thermal state data; Based on the current thermal state data, time series prediction processing is performed to generate the predicted thermal behavior data.

[0009] In one embodiment, the step of performing time series prediction processing based on the current thermal state data to generate the predicted thermal behavior data includes: Based on the digital twin model, the current thermal state data is subjected to multi-step forward simulation to generate predicted temperature data at multiple time points. Based on the changing trend of the server workload data, the predicted temperature data is weighted and adjusted. The weighted predicted temperature data is fused with historical thermal behavior data to generate the predicted thermal behavior data.

[0010] In one embodiment, the step of inputting the predicted thermal behavior data, the current temperature distribution data, and the multi-source fusion data into a control strategy generation unit, and generating a heat dissipation control instruction set through the control strategy generation unit includes: Based on the predicted thermal behavior data, temperature stability analysis is performed to generate temperature stability assessment data. Based on the power consumption measurement data in the multi-source fusion data, energy efficiency analysis is performed to generate cooling energy efficiency evaluation data; Based on the current temperature distribution data, a collaborative analysis of the actuators is performed to generate collaborative control evaluation data; The temperature stability assessment data, the cooling energy efficiency assessment data, and the collaborative control assessment data are comprehensively evaluated to generate comprehensive assessment data; Based on the comprehensive evaluation data, fan control command data, cooling medium control command data, and valve control command data are generated. The fan control command data, the cooling medium control command data, and the valve control command data are combined into the heat dissipation control command set.

[0011] In one embodiment, the step of generating fan control command data, cooling medium control command data, and valve control command data based on the comprehensive evaluation data includes: Based on the comprehensive evaluation data, fan control parameter data, cooling medium control parameter data, and valve control parameter data are extracted. The fan control parameter data is converted into specific fan speed command data, the cooling medium control parameter data is converted into specific cooling medium flow command data, and the valve control parameter data is converted into specific valve opening command data; The fan speed command data, the cooling medium flow command data, and the valve opening command data are used as the fan control command data, the cooling medium control command data, and the valve control command data, respectively.

[0012] In one embodiment, the method further includes: Acquire heat dissipation control effect data, which includes actual response data after the heat dissipation actuator executes the heat dissipation control instruction set and new sensor-collected data; The heat dissipation control effect data is input into the learning and updating unit, and the learning and updating unit generates learning and updating data. Based on the learned update data, the digital twin model and the control strategy generation unit are updated.

[0013] In one embodiment, the step of updating the digital twin model and the control policy generation unit based on the learned update data includes: The heat dissipation control effect data is compared with the expected control effect data to generate error evaluation data; Based on the error assessment data, the internal parameters of the digital twin model are updated to generate an updated digital twin model; Based on the heat dissipation control effect data, the control parameters of the control strategy generation unit are updated to generate an updated control strategy generation unit.

[0014] In addition, to achieve the above objectives, this application also proposes an intelligent heat dissipation control system for a server, the intelligent heat dissipation control system for the server comprising: a memory, a processor, and an intelligent heat dissipation control program for the server stored in the memory and executable on the processor, the intelligent heat dissipation control program for the server being configured to implement the steps of the intelligent heat dissipation control method for the server.

[0015] The intelligent heat dissipation control method and system for servers proposed in this application solves the problems of slow response and energy waste in traditional heat dissipation technologies by acquiring multi-source data, generating fused data, reconstructing temperature distribution, predicting thermal behavior, and dynamically generating control commands. It can fuse multi-source data, predict changes in thermal behavior, and dynamically generate heat dissipation control commands, thereby improving heat dissipation efficiency and energy utilization, avoiding local hot spots, and enhancing the stability of server operation. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating an embodiment of the intelligent heat dissipation control method for the server in this application; Figure 2 This is a schematic diagram of a structural embodiment of the intelligent heat dissipation control system for the server in this application.

[0019] Explanation of icon numbers: 10. Memory; 20. Processor.

[0020] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of this application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0022] It should be understood that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] In the current server technology field, as server performance continues to improve, the heat generated also increases dramatically. Traditional cooling methods often suffer from low cooling efficiency and high energy consumption when facing high-load, high-density server clusters. Furthermore, traditional cooling control methods are usually based on fixed strategies and cannot be dynamically adjusted according to the actual operating status and thermal behavior of the server, which to some extent restricts the improvement of server performance and stable operation.

[0024] Based on this, embodiments of this application provide an intelligent heat dissipation control method for a server, referring to... Figure 1The intelligent heat dissipation control method for the server includes steps S100 to S500, wherein: Step S100: Obtain temperature sensor data, pressure sensor data, and power consumption measurement data of the server system, and input the temperature sensor data, pressure sensor data, and power consumption measurement data into the data fusion unit for processing to generate multi-source fused data; Step S200: Based on the temperature sensor data, generate the current temperature distribution data of the server using a thermal field reconstruction algorithm; Step S300: Input the current temperature distribution data, the pressure sensor data, and the power consumption measurement data into the digital twin model; input the server workload data into the digital twin model; and generate predicted thermal behavior data of the server through the digital twin model. Step S400: Input the predicted thermal behavior data, the current temperature distribution data, and the multi-source fusion data into the control strategy generation unit, and generate a heat dissipation control instruction set through the control strategy generation unit; In step S500, the heat dissipation control instruction set is sent to the heat dissipation actuator for execution to control the heat dissipation of the server.

[0025] In this embodiment, temperature sensor data refers to real-time temperature measurements collected by temperature sensors deployed inside or outside the server system, reflecting the temperature conditions in different areas of the server. Pressure sensor data refers to pressure measurements collected by a pressure sensor array inside or outside the system, such as coolant pressure or airflow pressure. This data helps assess the operating status of the cooling system. Power consumption measurement data is real-time power consumption data of various components or the server as a whole collected by a power consumption monitoring device, reflecting the server's current workload and heat generation. The data fusion unit is a processing module whose function is to receive and integrate heterogeneous data from different sensors and monitoring devices, perform preprocessing such as time alignment and data cleaning, to generate unified and reliable multi-source fused data.

[0026] In this embodiment, multi-source fusion data refers to comprehensive data that integrates various information such as temperature, pressure, and power consumption after processing by the data fusion unit, providing comprehensive input for subsequent analysis and decision-making. The thermal field reconstruction algorithm is a computational method that, based on limited temperature sensor data and combined with the server's physical structure and thermal conductivity characteristics, infers and generates a continuous temperature distribution map inside the server, i.e., the current temperature distribution data. The current temperature distribution data refers to the real-time temperature status of various areas inside the server obtained through the thermal field reconstruction algorithm, which can intuitively reflect the server's hotspot locations and overall thermal load. The digital twin model is a virtual mapping of the server; it can synchronize the server's physical state and behavior in real time and perform simulation calculations based on the physical model and historical data to predict the server's future thermodynamic behavior.

[0027] In this embodiment, server workload data refers to data reflecting the server's computing load and resource consumption, such as the amount of tasks currently being processed, CPU utilization, memory usage, and network I / O throughput. Predicted thermal behavior data refers to the predictions made by the digital twin model based on the current state and workload regarding the server's temperature change trends, hotspot migration, and other thermodynamic behaviors over a future period. The control strategy generation unit is a decision-making module that receives predicted thermal behavior data, current temperature distribution data, and multi-source fusion data, and generates specific heat dissipation control instructions based on preset optimization goals and rules. The heat dissipation control instruction set refers to a series of specific operation commands output by the control strategy generation unit, used to adjust the working state of the heat dissipation actuators, such as fan speed, coolant flow rate, or valve opening. The heat dissipation actuators refer to the hardware devices that actually perform heat dissipation operations, such as cooling fans inside the server, cooling pumps in the liquid cooling system, and valves. They operate according to the heat dissipation control instruction set to adjust the server's heat dissipation effect.

[0028] In this embodiment, the intelligent heat dissipation control method for the server first acquires temperature sensor data, pressure sensor data, and power consumption measurement data from the server system. This data is then input into a data fusion unit for processing to generate multi-source fused data. Specifically, temperature sensor data can be collected by deploying thermistors or thermocouples at key locations within the server; pressure sensor data can be obtained by installing pressure sensors in the coolant circulation pipes or airflow channels; and power consumption measurement data can be measured by installing current / voltage sensors at the server's power input or on the power supply lines of major components. This raw data is then aggregated into the data fusion unit, which can be an independent processor or a functional module on the server's main control chip. This unit performs preliminary format conversion and verification of the received data to ensure its usability.

[0029] Furthermore, this embodiment generates the current temperature distribution data of the server based on the temperature sensor data using a thermal field reconstruction algorithm. For example, finite element analysis or interpolation methods based on empirical formulas can be used to extend the discrete temperature sensor measurements to the entire internal space of the server. This thermal field reconstruction algorithm can be pre-stored in the server's memory and run on the processor. It generates a two-dimensional or three-dimensional temperature field data representing the temperature at various points inside the server by processing the sensor data.

[0030] Based on this, this embodiment inputs the current temperature distribution data, the pressure sensor data, and the power consumption measurement data into the digital twin model, along with the server workload data. The digital twin model then generates predicted thermal behavior data for the server. This digital twin model can be a complex simulation model built based on physical equations and machine learning algorithms, capable of simulating the server's thermodynamic response under different workloads. Server workload data may include CPU utilization, memory usage, network I / O throughput, etc., which can be obtained through the API interface of the operating system or virtualization platform. After receiving these inputs, the digital twin model performs real-time thermodynamic simulation to predict the server's temperature change trend and potential hotspots over a future period.

[0031] Subsequently, in this embodiment, the predicted thermal behavior data, the current temperature distribution data, and the multi-source fusion data are input into the control strategy generation unit, which generates a set of heat dissipation control instructions. This control strategy generation unit can be a decision module based on a rule engine or optimization algorithm. For example, a series of heat dissipation rules can be preset, such as increasing the fan speed when the predicted temperature exceeds a certain threshold; or selecting the lowest energy consumption heat dissipation scheme while meeting heat dissipation requirements, based on energy consumption optimization goals. This unit comprehensively considers the server's current thermal state, future thermal behavior predictions, and comprehensive information provided by the multi-source fusion data to generate specific control instructions for different heat dissipation actuators.

[0032] Finally, in this embodiment, the heat dissipation control command set is sent to the heat dissipation actuators for execution to control the server's heat dissipation. The heat dissipation actuators may include cooling fans inside the server, cooling pumps in the liquid cooling system, flow control valves, etc. After receiving the command set, these actuators adjust their operating states according to the command content, such as increasing or decreasing fan speed, increasing or decreasing coolant flow, and adjusting valve opening, thereby achieving dynamic and precise control of the server's heat dissipation.

[0033] In this embodiment, this application achieves refined and intelligent management of server heat dissipation through multi-source data fusion, thermal field reconstruction, digital twin model prediction, and dynamic control strategy generation. This method can dynamically adjust the heat dissipation strategy based on the server's real-time operating status and future thermal behavior trends, effectively addressing the heat dissipation challenges of high-load, high-density server clusters in large-scale server deployment scenarios such as data centers, improving heat dissipation efficiency, reducing energy consumption, and thus ensuring stable server operation and performance.

[0034] In one feasible implementation, the steps of acquiring temperature sensor data, pressure sensor data, and power consumption measurement data of a server system, and inputting the temperature sensor data, pressure sensor data, and power consumption measurement data into a data fusion unit for processing to generate multi-source fused data include: acquiring temperature sensor data of the server through a temperature sensor network; acquiring pressure sensor data of the system through a pressure sensor array; acquiring power consumption measurement data of multiple components through a power consumption monitoring device; and inputting the acquired temperature sensor data, pressure sensor data, and power consumption measurement data into the data fusion unit for time alignment processing and data cleaning processing to generate the multi-source fused data.

[0035] In this embodiment, temperature sensor data from the server is collected via a temperature sensor network. This network consists of multiple distributed temperature sensors capable of real-time monitoring of temperatures at different locations within the server. These sensors can be thermistors, thermocouples, or infrared sensors, and are connected to the data acquisition module or server management controller (BMC) via wired (e.g., I2C, SPI, Modbus) or wireless (e.g., ZigBee, Wi-Fi, LoRa) connections. This networked deployment can cover key heat-generating components such as the central processing unit (CPU), graphics processing unit (GPU), memory, hard drive, and power supply module, as well as points along airflow paths such as air inlets and outlets, thereby obtaining detailed temperature distribution information within the server. The sampling frequency can be configured according to system requirements, such as once per second or more frequently.

[0036] In this embodiment, pressure sensor data of the system is acquired through a pressure sensor array. The pressure sensor array consists of multiple pressure sensors arranged in a specific layout to monitor airflow or liquid pressure at key points inside the server or the cooling system. These sensors can be microelectromechanical systems (MEMS) pressure sensors or piezoresistive sensors, typically deployed at fan inlets and outlets, heatsink channels, liquid cooling pipes, etc. By using an array, pressure differences in different areas can be obtained, thereby assessing the resistance, velocity, and uniformity of airflow or liquid flow. For example, monitoring the pressure difference before and after a fan can determine the fan's operating status and the blockage of cooling channels; monitoring the pressure at different points in the liquid cooling system can assess the coolant circulation efficiency.

[0037] In this embodiment, power consumption data for multiple components is collected using a power monitoring device. This power monitoring device is used to measure the real-time power consumption of various key components within the server (such as the CPU, GPU, memory, hard drive, and power module). This can be achieved through a power management chip (PMIC) integrated on the server motherboard, independent power sensors (such as current sensors and voltage sensors), or software report data obtained through server management interfaces (such as IPMI and Redfish). For each component, its instantaneous power consumption, average power consumption, or cumulative power consumption can be collected. For example, by using shunt current detection or a power analyzer, real-time power consumption data for high-power components such as the CPU core, GPU memory, and DRAM module can be accurately obtained; this data directly reflects the heat generated by the component.

[0038] In this embodiment, the collected temperature sensor data, pressure sensor data, and power consumption measurement data are input into a data fusion unit for time alignment and data cleaning to generate the multi-source fused data. Time alignment aims to synchronize all data to a unified time reference, as different sensors and monitoring devices may have different sampling frequencies and timestamps. This can be achieved through interpolation (such as linear interpolation or spline interpolation) or resampling techniques, mapping data points onto a common time axis. Data cleaning aims to identify and correct noise, outliers, missing values, or redundant information contained in the original collected data. For example, statistical methods (such as Z-score or IQR) can be used to detect and remove outliers; interpolation techniques (such as mean interpolation or regression interpolation) can be used to fill missing values; and filtering algorithms (such as Kalman filtering or moving average) can be used to smooth noisy data. Furthermore, data deduplication and format standardization can be performed to improve data quality and usability.

[0039] In this embodiment, the above-described technical solution enables comprehensive and detailed acquisition of temperature, pressure, and power consumption data within the server system, overcoming the limitations of insufficient coverage or incomplete information from a single sensor. Furthermore, inputting this heterogeneous data into the data fusion unit for time alignment and data cleaning effectively solves problems such as inconsistent sampling frequencies from different data sources, timestamp discrepancies, and noise, outliers, and missing values ​​in the original data. This ensures the consistency of multi-source fused data in the time dimension and the reliability of data quality, thus providing high-quality, high-precision input for subsequent thermal field reconstruction, thermodynamic simulation calculations of digital twin models, and control strategy generation. Ultimately, this significantly improves the accuracy of the server's intelligent heat dissipation control system in perceiving the server's thermal state, the precision of its predictions, and the effectiveness of its heat dissipation control commands, avoiding misjudgments and inefficient heat dissipation caused by data quality issues, and ensuring stable server operation and energy efficiency optimization.

[0040] In one feasible implementation, the step of generating the current temperature distribution data of the server based on the temperature sensor data and using a thermal field reconstruction algorithm includes: extracting multiple spatially distributed temperature measurement values ​​from the temperature sensor data; performing spatial interpolation processing on the temperature measurement values ​​based on a preset heat conduction physical model to generate initial temperature field data; and performing thermal field correction processing on the initial temperature field data based on the internal structural distribution of the server to generate the current temperature distribution data.

[0041] In this embodiment, extracting multiple spatially distributed temperature measurements from the temperature sensor data refers to identifying and separating the temperature values ​​measured by each temperature sensor at a specific physical location from the raw temperature sensor data stream acquired from the server system. Servers typically have multiple temperature sensors deployed internally, such as in key locations like the central processing unit (CPU), graphics processing unit (GPU), memory modules, power supply units, and air inlets / outlets. This step aims to obtain these discrete temperature readings with defined spatial coordinates as the foundational input for subsequent thermal field reconstruction.

[0042] In this embodiment, based on a preset heat conduction physical model, the temperature measurements are spatially interpolated to generate initial temperature field data. This involves using mathematical algorithms and physical principles to estimate the temperature at locations within the server where no sensors are installed, based on known discrete temperature measurements, thereby constructing a continuous, preliminary temperature distribution field. The preset heat conduction physical model may include Fourier's law of heat conduction, convective heat transfer models, etc., which define the basic laws governing heat transfer in different media. Spatial interpolation techniques, such as Kriging interpolation, inverse distance weighted (IDW) interpolation, or radial basis function (RBF) interpolation, combined with these physical models, can more reasonably infer the temperature distribution within the entire server space, forming a preliminary, continuous temperature field data.

[0043] In this embodiment, based on the internal structural distribution of the server, the initial temperature field data undergoes thermal field correction processing to generate the current temperature distribution data. This means that, based on the initially generated temperature field data, the actual physical structure and component layout inside the server are further considered to refine and optimize the temperature field. The internal structural distribution of the server includes various heat-generating components (such as CPU and GPU), heat sinks, fans, heat conduction paths, airflow channels, and the physical location and thermal properties of different materials. Thermal field correction processing can utilize this detailed structural information, such as by superimposing heat source models of components, considering the impact of airflow convection heat transfer, or adjusting the local temperature gradient according to the thermal conductivity of different materials, thereby correcting any deviations that may exist in the initial temperature field, making it more accurately reflect the true temperature distribution inside the server, and ultimately obtaining highly accurate current temperature distribution data.

[0044] In this embodiment, the above-described technical solution first accurately extracts spatially distributed temperature measurements from the original sensor data, ensuring the accuracy of the input data. Then, based on a preset thermal conduction physical model, these discrete measurements are spatially interpolated, overcoming the limitations of the number and location of sensors to generate a continuous initial temperature field that conforms to physical laws. Furthermore, by combining detailed structural distribution information within the server to correct the initial temperature field, complex factors such as local heat sources, heat dissipation paths, and airflow organization are fully considered, thereby generating highly accurate and detailed current temperature distribution data. This refined temperature distribution data not only compensates for the spatial coverage limitations of discrete sensor data but also, by combining the physical model and structural information, reveals potential hotspots and temperature gradients within the server, providing a more realistic and reliable input for subsequent thermodynamic simulations of the digital twin model. This significantly improves the accuracy of server thermal state perception, resulting in more accurate predicted thermal behavior data. Ultimately, this optimizes the generation of the heat dissipation control instruction set, ensuring that the heat dissipation strategy can more accurately respond to the actual heat load of the server, effectively avoiding overheating risks while improving heat dissipation efficiency.

[0045] In one feasible implementation, the steps of inputting the current temperature distribution data, the pressure sensor data, and the power consumption measurement data into a digital twin model, inputting the server workload data into the digital twin model, and generating predicted thermal behavior data of the server through the digital twin model include: inputting the current temperature distribution data, the pressure sensor data, and the power consumption measurement data into the digital twin model; inputting the server workload data into the digital twin model; performing thermodynamic simulation calculations on the current temperature distribution data, the pressure sensor data, the power consumption measurement data, and the server workload data based on the digital twin model to generate current thermal state data; and performing time series prediction processing on the current thermal state data to generate the predicted thermal behavior data.

[0046] In this embodiment, inputting current temperature distribution data, pressure sensor data, and power consumption measurement data into the digital twin model means using various sensor data acquired from the physical server system in real-time or near real-time as input parameters for the digital twin model. The digital twin model, as a virtual copy of the physical server, can receive and parse this data to reflect the current operating status of the physical server. For example, current temperature distribution data can provide the digital twin model with the temperature boundary conditions or initial temperature field of various regions within the server; pressure sensor data can reflect the operating status or airflow distribution of the cooling system (such as fans or liquid cooling circulation); and power consumption measurement data is directly related to the heat generation rate of various components within the server. These data collectively provide the digital twin model with the real-time physical environment information required for accurate thermodynamic analysis.

[0047] In this embodiment, inputting server workload data into the digital twin model refers to using information about the computing tasks currently being executed or about to be executed by the server as another key input to the digital twin model. Server workload data may include metrics such as CPU utilization, GPU utilization, memory usage, and network I / O throughput. These workload metrics are closely related to the power consumption and heat generation of various components within the server. By inputting workload data into the digital twin model, the model can dynamically adjust its internal heat source model to simulate the actual heat generation of various components under different workloads, thereby making the thermodynamic simulation more closely resemble the actual operating scenario.

[0048] In this embodiment, based on the digital twin model, thermodynamic simulation calculations are performed on the current temperature distribution data, pressure sensor data, power consumption measurement data, and server workload data to generate current thermal state data. This means that after receiving all the above input data, the digital twin model uses its built-in physical model and simulation algorithms to simulate and calculate the heat generation, transfer, and dissipation processes inside the server. This may involve computational fluid dynamics (CFD) simulation, finite element analysis (FEA), or solving other thermodynamic equations related to heat conduction, convection, and radiation. By comprehensively considering the current temperature distribution, cooling system status, component power consumption, and heat generation from the workload, the digital twin model can generate comprehensive and high-precision current thermal state data. This data not only includes the precise temperature of each component but may also include detailed information such as internal airflow velocity, pressure distribution, and heat flux density, thereby providing a deep characterization of the server's current thermal environment.

[0049] In this embodiment, time-series prediction processing is performed based on the current thermal state data to generate the predicted thermal behavior data. This means that after obtaining accurate current thermal state data, the digital twin model or its integrated prediction module uses this data as a starting point, combined with historical thermal behavior data and expected workload change trends, and applies time-series analysis or machine learning algorithms (e.g., recurrent neural networks, long short-term memory networks, ARIMA models, etc.) to predict the future thermal behavior of the server. The goal of the prediction is to generate predicted thermal behavior data, which includes predicted values ​​of key thermal indicators such as temperature and heat flow of various server components over a future period. This prediction provides forward-looking information, enabling the thermal control system to anticipate potential overheating risks and take preventative measures.

[0050] In this embodiment, the aforementioned technical solution comprehensively and systematically inputs the server's current temperature distribution data, pressure sensor data, power consumption measurement data, and server workload data into the digital twin model, enabling the digital twin model to obtain complete real-time status information of the server's operation. Based on this, precise thermodynamic simulation calculations are performed using the digital twin model, generating highly accurate current thermal state data. This data not only reflects the actual situation measured by the sensors but also incorporates the heat generated by the workload, thus providing a deep understanding of the server's internal thermal environment. Furthermore, time-series prediction processing based on this accurate current thermal state data can effectively predict the server's future thermal behavior trends, overcoming the limitations of traditional methods that rely solely on real-time data for passive control. This method, combining physical simulation and time-series prediction, significantly improves the accuracy and foresight of predicted thermal behavior data, providing a solid foundation for the subsequent control strategy generation unit to formulate more precise, efficient, and predictive heat dissipation control instruction sets, thereby effectively avoiding the risk of server overheating and optimizing heat dissipation efficiency.

[0051] In one feasible implementation, the step of performing time series prediction processing based on the current thermal state data to generate the predicted thermal behavior data includes: performing multi-step forward simulation on the current thermal state data based on the digital twin model to generate predicted temperature data at multiple time points; adjusting the weights of the predicted temperature data based on the changing trend of the server workload data; and fusing the weighted predicted temperature data with historical thermal behavior data to generate the predicted thermal behavior data.

[0052] In this embodiment, a multi-step forward simulation is first performed based on a digital twin model to generate predicted temperature data for multiple time points. Specifically, the digital twin model utilizes its built-in physical models (e.g., heat conduction, convection, and radiation equations) and / or machine learning algorithms, using the current thermal state data as initial conditions, to simulate the change in heat within the server over time. This simulation process iteratively calculates the temperature distribution of key locations or regions within the server at a series of discrete future time points, thus providing a dynamic sequence prediction of future thermal behavior, rather than a prediction of a single time point.

[0053] Secondly, the predicted temperature data is weighted and adjusted based on the changing trends of server workload data. Server workload data, such as CPU utilization, memory usage, and I / O throughput, directly affect its heat generation. The system analyzes historical or real-time server workload data to identify its rising, falling, or stable trends. Based on a preset weighting function or the load-temperature correlation obtained through machine learning model training, the predicted temperature data obtained from the aforementioned multi-step forward simulation is corrected. For example, if the predicted workload will increase significantly, the corresponding predicted temperature value will be adjusted upwards to more accurately reflect the expected heat generation.

[0054] Finally, the weighted predicted temperature data is fused with historical thermal behavior data to generate predicted thermal behavior data. Historical thermal behavior data includes actual temperature records of the server under different workloads and cooling conditions. Fusion processing can employ various techniques, such as Kalman filtering, weighted averaging, or machine learning-based fusion algorithms. By combining workload-adjusted predicted data with actual historical operating experience, potential biases in model predictions can be effectively corrected, enhancing the robustness of the prediction results and their adaptability to real-world operating environments, thereby obtaining more accurate and reliable predicted thermal behavior data.

[0055] In this embodiment, this application utilizes a digital twin model for multi-step forward simulation, generating predicted temperature data for multiple future time points, thereby providing a more comprehensive and dynamic view of server thermal behavior prediction. Given the significant impact of server workload dynamism on thermal behavior, weighting the predicted temperature data based on workload data trends allows the prediction results to more accurately reflect changes in heat generation under actual operating conditions. Furthermore, fusing the weighted predicted temperature data with historical thermal behavior data effectively compensates for potential biases in pure model predictions and enhances the robustness and adaptability of the predictions to real-world operating environments. This prediction mechanism, combining physical simulation, dynamic load adaptation, and historical experience, significantly improves the accuracy and reliability of server thermal behavior prediction, providing a more precise and forward-looking basis for subsequent heat dissipation control strategies. This avoids insufficient or excessive heat dissipation due to inaccurate predictions, thereby optimizing heat dissipation efficiency and server operational stability.

[0056] In one feasible implementation, the step of inputting the predicted thermal behavior data, the current temperature distribution data, and the multi-source fusion data into a control strategy generation unit, and generating a heat dissipation control instruction set through the control strategy generation unit includes: performing temperature stability analysis based on the predicted thermal behavior data to generate temperature stability assessment data; performing energy efficiency analysis based on power consumption measurement data in the multi-source fusion data to generate cooling energy efficiency assessment data; performing actuator coordination analysis based on the current temperature distribution data to generate coordination control assessment data; comprehensively evaluating the temperature stability assessment data, the cooling energy efficiency assessment data, and the coordination control assessment data to generate comprehensive assessment data; generating fan control instruction data, cooling medium control instruction data, and valve control instruction data based on the comprehensive assessment data; and combining the fan control instruction data, the cooling medium control instruction data, and the valve control instruction data into the heat dissipation control instruction set.

[0057] In this embodiment, temperature stability analysis is first performed based on predicted thermal behavior data to generate temperature stability assessment data. This analysis aims to proactively assess the server's thermal state over a future period, identifying potential overheating risks or temperature fluctuation trends. Predicted thermal behavior data provides the expected trajectory of server temperature changes over time. By comparing this data with preset temperature thresholds, safe ranges, or historical stability patterns, the stability of the server's thermal environment can be quantified. For example, the magnitude, duration, or probability of predicted temperature fluctuations exceeding safe thresholds can be calculated to generate temperature stability assessment data reflecting future thermal risks, such as a stability index or risk level.

[0058] Secondly, based on power consumption metering data from multi-source fusion data, energy efficiency analysis is performed to generate cooling energy efficiency assessment data. This step focuses on evaluating the energy efficiency of the current cooling strategy. Power consumption metering data reflects the real-time energy consumption of various server components, while energy efficiency analysis measures the economics of cooling by calculating the relationship between the energy consumed by the cooling system and the actual cooling needs of the server. For example, it can calculate the energy consumption required per unit of heat dissipation or assess whether the cooling system is over-operating under the current load. Cooling energy efficiency assessment data can be an energy efficiency ratio, an energy cost indicator, or an optimization potential assessment, guiding the cooling system to minimize energy consumption while meeting temperature requirements.

[0059] Furthermore, based on the current temperature distribution data, a collaborative analysis of the actuators is performed to generate collaborative control evaluation data. This analysis aims to optimize the coordination between different heat dissipation actuators. The current temperature distribution data provides a detailed temperature map of each area inside the server, revealing specific local hotspots. The actuator collaborative analysis plans the optimal collaborative heat dissipation scheme based on the location and intensity of these hotspots, as well as the physical location and cooling capacity of each heat dissipation actuator (such as fans, liquid cooling modules, valves, etc. in different areas). For example, if the temperature in a specific processor area is too high, the collaborative control evaluation data will indicate how to adjust the fan speed, liquid cooling medium flow rate, or the opening of relevant valves near that area to achieve precise and efficient local heat dissipation and avoid unnecessary global heat dissipation actions.

[0060] Subsequently, the temperature stability assessment data, cooling energy efficiency assessment data, and collaborative control assessment data are comprehensively evaluated to generate integrated assessment data. This step is crucial for decision-making, as it integrates and weighs the assessment results from the aforementioned three dimensions. Since temperature stability, energy efficiency, and collaborative control may have conflicting objectives (e.g., extreme temperature stability may sacrifice energy efficiency), the integrated assessment uses pre-defined strategies, priorities, or machine learning models to weight, fuse, or perform multi-objective optimization of these assessment data, thereby forming a comprehensive and guiding integrated assessment. This data reflects the most significant problems currently facing server cooling and the optimal solutions.

[0061] Based on the comprehensive evaluation data, fan control command data, cooling medium control command data, and valve control command data are generated. This step transforms the abstract comprehensive evaluation data into specific, executable control commands. For example, if the comprehensive evaluation data indicates "high heat risk and low energy efficiency, requiring enhanced local heat dissipation," then the appropriate increase in fan speed, cooling medium flow rate, and valve opening will be determined based on the evaluation results. These command data are precise operating parameters for different types of heat dissipation actuators.

[0062] Finally, the fan control command data, cooling medium control command data, and valve control command data are combined into a heat dissipation control command set. This command set contains all the control commands that need to be sent to the heat dissipation actuators, ensuring the synchronization and coordination of all heat dissipation actions, forming a complete and unified heat dissipation control scheme.

[0063] In this embodiment, by performing temperature stability analysis on predicted thermal behavior data using the above technical solution, the future thermal risks of the server can be proactively assessed, thereby achieving preventative heat dissipation control. By performing energy efficiency analysis on power consumption measurement data from multi-source fusion data, it can be ensured that the heat dissipation strategy meets temperature requirements while also considering energy consumption, improving overall operating efficiency. Simultaneously, by performing collaborative analysis of actuators based on current temperature distribution data, the coordination between different heat dissipation actuators can be optimized, avoiding localized overcooling or overheating, and achieving precise and efficient localized heat dissipation. Comprehensive evaluation of these assessment data allows for a holistic balance of multiple objectives, including temperature stability, energy efficiency, and collaborative control, thereby generating a more intelligent, coordinated, and efficient heat dissipation control instruction set. This comprehensive evaluation mechanism effectively solves the problem of difficulty in effectively integrating multi-source data to generate optimal control strategies, ensuring that the server maintains optimal heat dissipation under different workloads, while reducing energy consumption and extending equipment lifespan.

[0064] In one feasible implementation, the step of generating fan control command data, cooling medium control command data, and valve control command data based on the comprehensive evaluation data includes: extracting fan control parameter data, cooling medium control parameter data, and valve control parameter data based on the comprehensive evaluation data; converting the fan control parameter data into specific fan speed command data, converting the cooling medium control parameter data into specific cooling medium flow command data, and converting the valve control parameter data into specific valve opening command data; and using the fan speed command data, the cooling medium flow command data, and the valve opening command data as the fan control command data, the cooling medium control command data, and the valve control command data, respectively.

[0065] In this embodiment, the macro-level guidance information on overall heat dissipation performance and energy efficiency contained in the aforementioned comprehensive evaluation data is parsed and refined into specific control parameters for different types of heat dissipation actuators (such as fans, cooling medium circulation systems, and fluid control valves). The comprehensive evaluation data may encompass diverse information such as the server's overall heat load, local hotspot distribution, and expected energy consumption targets. Therefore, it is necessary to intelligently identify and extract control elements directly related to the functions of each actuator. For example, when the comprehensive evaluation data indicates a need to improve overall heat dissipation capacity and optimize local hotspots, the system will extract corresponding fan speed parameters, cooling medium flow rate parameters, and valve opening parameters for specific areas. This extraction process can be achieved through preset mapping rules, decision logic based on expert experience, or training and prediction using machine learning models (such as regression or classification models) to ensure that the extracted parameters accurately reflect the intent of the comprehensive evaluation data.

[0066] In this embodiment, abstract fan control parameter data is converted into physical quantity commands that the fan actuator can directly recognize and execute. Fan control parameter data is typically a relative value or a level, such as a percentage from 0 to 100, or discrete levels like "low," "medium," and "high." To drive an actual fan, this needs to be converted into a specific physical quantity, namely, fan speed command data. This conversion can be dynamically adjusted using predefined function curves, lookup tables, or based on a PID (proportional-integral-derivative) control algorithm. For example, a fan control parameter of 70% might be converted into a fan speed command of 2800 RPM (revolutions per minute), or a specific PWM (pulse width modulation) signal duty cycle that can directly control the fan motor speed.

[0067] In this embodiment, abstract cooling medium control parameter data is transformed into physical quantity commands that the cooling medium circulation system (such as a pump) can directly recognize and execute. Cooling medium control parameter data may indicate the circulation intensity or cooling capacity requirement of the cooling medium. Converting this into specific cooling medium flow command data, such as a flow rate value in liters per minute (L / min), is crucial for achieving precise cooling. This conversion can also utilize preset conversion curves, lookup tables, or calculations based on fluid dynamics models. For example, when the cooling medium control parameters indicate the need for enhanced cooling, the system may convert this into a flow rate command of 5 liters per minute to ensure sufficient cooling medium flows through the heat source area.

[0068] In this embodiment, abstract valve control parameter data is transformed into physical quantity commands that the fluid control valve can directly recognize and execute. Valve control parameter data is typically used to regulate the distribution or flow direction of the cooling medium in different areas. Converting this data into specific valve opening command data, such as expressed in angles (e.g., 0-90 degrees) or percentage openings (e.g., 0-100%), enables fine-grained control of the cooling medium path. This conversion can be achieved through the valve's mechanical characteristic curves, lookup tables, or stepper motor control algorithms. For example, when valve control parameters indicate the need to direct the cooling medium to a specific hot spot area, the system may convert this into a 75% valve opening command to precisely regulate the flow rate of the cooling medium through that area.

[0069] In this embodiment, the specific physical quantity commands obtained after the above conversion are clearly defined, namely, fan speed command data, cooling medium flow rate command data, and valve opening command data. These constitute the directly executable control command data that is finally sent to their respective actuators. These commands are low-level, operable instructions, ensuring that the heat dissipation actuators can accurately receive and execute the intentions of the control center.

[0070] In this embodiment, the high-level comprehensive evaluation data is broken down into control parameters for the fan, cooling medium, and valves, as described above. These parameters are then precisely converted into specific physical commands that can directly drive the actuators, such as fan speed commands, cooling medium flow commands, and valve opening commands. This ensures the accuracy and executability of the heat dissipation control commands, enabling different types of heat dissipation actuators to respond collaboratively and efficiently to the server's heat dissipation needs. It avoids control lag or inefficiency caused by ambiguous or incompatible commands, thereby improving the overall refined management level and response speed of the heat dissipation system and effectively coping with the complex and ever-changing thermal environment inside the server.

[0071] In one feasible implementation, the method further includes: acquiring heat dissipation control effect data, the heat dissipation control effect data including actual response data after the heat dissipation actuator executes the heat dissipation control instruction set and new sensor acquisition data; inputting the heat dissipation control effect data into a learning update unit, generating learning update data through the learning update unit; and updating the digital twin model and the control strategy generation unit based on the learning update data.

[0072] In this embodiment, acquiring heat dissipation control effect data aims to collect actual feedback information after the execution of heat dissipation control commands, in order to evaluate the effectiveness of the control commands and the system response. The heat dissipation control effect data may include the actual operating parameters of the heat dissipation actuator, such as the actual fan speed, the actual coolant flow rate, and the actual valve opening. This data is typically acquired through sensors or feedback interfaces integrated into the actuator. It also includes newly acquired sensor data, namely, temperature sensor data, pressure sensor data, and power consumption measurement data re-acquired by the system after the heat dissipation control command is executed. This data reflects the actual thermal and operational status of the server after heat dissipation control. This data can be acquired in real-time from the actuator and sensor network via the data acquisition module and timestamped to ensure that the data is associated with the corresponding control command and time point.

[0073] In this embodiment, the heat dissipation control effect data is input into the learning and update unit, which generates learning and update data. The learning and update unit is a processing module whose function is to analyze the heat dissipation control effect data and extract valuable information to guide subsequent model and strategy updates. The learning and update unit can preprocess the heat dissipation control effect data, such as data cleaning, feature extraction, and anomaly detection. The generated learning and update data may include, but is not limited to: control error (the deviation between actual temperature and target temperature), energy efficiency indicators (the ratio of cooling power consumption to heat dissipation effect), system response delay, and signs of wear or aging of actuators. These data are crucial for evaluating current control performance and guiding future optimization. The learning and update unit can use statistical analysis, machine learning algorithms (such as regression analysis and classification algorithms), or rule-based expert systems to generate this data.

[0074] In this embodiment, the digital twin model and the control strategy generation unit are updated based on the learned update data. This is the core of the entire feedback loop, using the learned update data to correct and optimize the digital twin model and the control strategy generation unit, making them better adaptable to the actual operating state and environmental changes of the server. The digital twin model typically includes physical parameters, heat conduction models, fluid dynamics models, etc. Based on the learned update data (e.g., the deviation between the actual temperature and the model's predicted temperature), methods such as model parameter identification, Kalman filtering, and adaptive control can be used to adjust the model's internal parameters, making its prediction results closer to reality, such as adjusting parameters like heat transfer coefficient and radiator efficiency. Simultaneously, the update of the control strategy generation unit aims to optimize its decision logic, enabling it to generate more efficient and energy-saving heat dissipation control instruction sets. Based on the learned update data (e.g., control error, energy efficiency assessment), methods such as reinforcement learning, adaptive PID control, fuzzy logic control, and neural networks can be used to adjust the parameters or rules of the control strategy, such as adjusting the mapping relationship between fan speed and temperature threshold, or optimizing the weights in a multi-objective optimization algorithm. Updates can be performed periodically or triggered when a significant performance degradation or environmental change is detected.

[0075] In this embodiment, through the above technical solution, this application constructs a closed-loop adaptive intelligent heat dissipation control system. This system can monitor the actual effects of heat dissipation control commands in real time and continuously optimize the predictive accuracy of the digital twin model and the decision-making ability of the control strategy using this feedback information. This effectively solves the problem of decreased model and strategy accuracy caused by dynamic changes in the server operating environment, hardware status, or workload, ensuring that the heat dissipation control system can maintain optimal performance over a long period, achieving more accurate, efficient, and stable server heat dissipation management, thereby extending equipment lifespan and reducing operating energy consumption. This adaptive learning mechanism, combined with basic predictive control, enables the system not only to anticipate future thermal behavior but also to continuously correct and improve its cognition and decision-making based on actual feedback, thus maintaining optimal heat dissipation control performance in dynamically changing and complex environments.

[0076] In one feasible implementation, the step of updating the digital twin model and the control strategy generation unit based on the learning update data includes: comparing the heat dissipation control effect data with the expected control effect data to generate error evaluation data; updating the internal parameters of the digital twin model based on the error evaluation data to generate an updated digital twin model; and updating the control parameters of the control strategy generation unit based on the heat dissipation control effect data to generate an updated control strategy generation unit.

[0077] In this embodiment, the heat dissipation control effect data is compared with the expected control effect data to generate error evaluation data. This aims to quantify the difference between the actual system state and the expected state after the heat dissipation control is executed. The heat dissipation control effect data typically includes the actual response data of the heat dissipation actuator after executing the heat dissipation control command set, such as actual measured server internal temperature, fan speed, cooling medium flow rate, and new sensor data. The expected control effect data can be derived from the prediction results of the digital twin model under a given control command, or the performance targets set by the control strategy generation unit. The comparison process can employ various mathematical methods, such as calculating absolute error, relative error, root mean square error (RMSE), or mean absolute error (MAE), to generate error evaluation data that reflects the magnitude and direction of the deviation. This error evaluation data forms the basis for subsequent model and strategy updates.

[0078] Based on this, the internal parameters of the digital twin model are updated using the error assessment data, generating an updated digital twin model. The internal parameters of the digital twin model typically include thermal conductivity, convective heat transfer coefficient, material heat capacity, and component power consumption model parameters, which directly affect the model's prediction accuracy. The update process can utilize optimization algorithms, such as gradient descent, Kalman filtering, particle filtering, or machine learning-based adaptive algorithms, to adjust these internal parameters by minimizing the error assessment data. For example, if the actual temperature is higher than the model's prediction, it may be necessary to adjust the parameters related to the heat source or heat dissipation path in the model to more accurately reflect the actual physical characteristics. In this way, the digital twin model can continuously learn and adapt to changes in the server's operating environment, improving its accuracy in predicting server thermal behavior.

[0079] Simultaneously, based on the heat dissipation control effect data, the control parameters of the control strategy generation unit are updated, generating an updated control strategy generation unit. The control parameters of the control strategy generation unit can be weighting coefficients, thresholds, PID controller gains, or neural network connection weights in the generation logic of fan control command data, cooling medium control command data, and valve control command data. The update process can directly utilize the heat dissipation control effect data as feedback, optimizing the performance of the control strategy through methods such as reinforcement learning, adaptive control, or heuristic rule adjustment. For example, if the heat dissipation control effect data shows large temperature fluctuations under a specific load, the control strategy generation unit may adjust the control parameters of the fan response speed or cooling medium flow rate to improve temperature stability. This update mechanism enables the control strategy to adaptively optimize based on actual operational feedback, thereby generating a more effective and accurate heat dissipation control command set.

[0080] In this embodiment, by comparing the actual heat dissipation control effect data with the expected control effect data using the above technical solution, the deviation between the actual performance of the system after executing heat dissipation commands and the ideal target can be quantified, thereby generating accurate error assessment data. Based on this error assessment data, the system can selectively update the internal parameters of the digital twin model, making it more accurately reflect the actual thermal behavior of the server and improving the accuracy of predictions. Simultaneously, the control parameters of the control strategy generation unit are directly adjusted using the heat dissipation control effect data, enabling the control strategy to adaptively optimize based on actual operational feedback, thereby generating a more effective and precise heat dissipation control command set. This dual update mechanism ensures the continuous optimization and co-evolution of the digital twin model and the control strategy, significantly improving the adaptive capability, robustness, and overall performance of server heat dissipation control, effectively avoiding the risk of low heat dissipation efficiency or overheating due to inaccurate models or strategies, and achieving more stable and energy-efficient server operation.

[0081] In the embodiments of this application, the intelligent heat dissipation control method for servers solves the problems of slow response and energy waste in traditional heat dissipation technologies by acquiring multi-source data, generating fused data, reconstructing temperature distribution, predicting thermal behavior, and dynamically generating control commands. It can fuse multi-source data, predict changes in thermal behavior, and dynamically generate heat dissipation control commands, thereby improving heat dissipation efficiency and energy utilization, avoiding local hot spots, and enhancing the stability of server operation.

[0082] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the intelligent heat dissipation control method of the server in this application. Any simple modifications based on this technical concept are within the protection scope of this application.

[0083] This application also provides an intelligent heat dissipation control system for a server, as shown in the reference. Figure 2 The intelligent heat dissipation control system of the server includes: a memory 10, a processor 20, and an intelligent heat dissipation control program of the server stored on the memory 10 and executable on the processor 20. The intelligent heat dissipation control program of the server is configured to implement the steps of the intelligent heat dissipation control method of the server.

[0084] The intelligent heat dissipation control system for servers provided in this application adopts the intelligent heat dissipation control method for servers in the above embodiments, which can improve heat dissipation efficiency and energy utilization. Compared with the prior art, the beneficial effects of the intelligent heat dissipation control system for servers provided in this application are the same as the beneficial effects of the intelligent heat dissipation control method for servers provided in the above embodiments, and other technical features in the intelligent heat dissipation control system for servers are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0085] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0086] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. All equivalent structural transformations made under the technical concept of this application using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included within the scope of patent protection of this application.

Claims

1. A method for intelligent heat dissipation control of a server, characterized in that, The method comprises: obtaining temperature sensor data, pressure sensor data and power consumption metering data of a server system, inputting the temperature sensor data, the pressure sensor data and the power consumption metering data into a data fusion unit for processing to generate multi-source fusion data; based on the temperature sensor data, generating current temperature distribution data of the server through a thermal field reconstruction algorithm; inputting the current temperature distribution data, the pressure sensor data and the power consumption metering data into a digital twin model, inputting server workload data into the digital twin model, and generating predicted thermal behavior data of the server through the digital twin model; inputting the predicted thermal behavior data, the current temperature distribution data and the multi-source fusion data into a control strategy generation unit to generate a set of heat dissipation control instructions through the control strategy generation unit; sending the set of heat dissipation control instructions to a heat dissipation execution mechanism for execution to control the heat dissipation of the server. 2.The intelligent heat dissipation control method of a server according to claim 1, wherein, The step of obtaining temperature sensor data, pressure sensor data and power consumption metering data of a server system, inputting the temperature sensor data, the pressure sensor data and the power consumption metering data into a data fusion unit for processing to generate multi-source fusion data comprises: collecting temperature sensor data of the server through a temperature sensor network; collecting pressure sensor data of the system through a pressure sensor array; collecting power consumption metering data of multiple components through a power consumption monitoring device; inputting the collected temperature sensor data, pressure sensor data and power consumption metering data into a data fusion unit for time alignment processing and data cleaning processing to generate the multi-source fusion data. 3.The intelligent heat dissipation control method of a server of claim 1, wherein, The step of generating current temperature distribution data of the server based on the temperature sensor data through a thermal field reconstruction algorithm comprises: extracting a plurality of spatially distributed temperature measurements from the temperature sensor data; based on a pre-set thermal conduction physical model, performing spatial interpolation processing on the temperature measurements to generate initial temperature field data; based on the internal structure distribution of the server, performing thermal field correction processing on the initial temperature field data to generate the current temperature distribution data. 4.The intelligent heat dissipation control method of a server according to claim 1, wherein, The step of inputting the current temperature distribution data, the pressure sensor data and the power consumption metering data into a digital twin model, inputting server workload data into the digital twin model, and generating predicted thermal behavior data of the server through the digital twin model comprises: inputting the current temperature distribution data, the pressure sensor data and the power consumption metering data into a digital twin model; inputting the server workload data into the digital twin model; based on the digital twin model, performing thermodynamic simulation calculation on the current temperature distribution data, the pressure sensor data, the power consumption metering data and the server workload data to generate current thermal state data; based on the current thermal state data, performing time series prediction processing to generate the predicted thermal behavior data. 5.The intelligent heat dissipation control method of a server according to claim 4, wherein, The step of performing time series prediction processing based on the current thermal state data to generate the predicted thermal behavior data comprises: Based on the digital twin model, the current thermal state data is simulated forward for multiple steps to generate predicted temperature data at multiple time points; Based on the trend of the server workload data, the predicted temperature data is subjected to weight adjustment processing; The predicted temperature data after weight adjustment is fused with historical thermal behavior data to generate the predicted thermal behavior data. 6.The intelligent heat dissipation control method of a server according to claim 1, wherein, The step of inputting the predicted thermal behavior data, the current temperature distribution data, and the multi-source fusion data into a control strategy generation unit to generate a set of heat dissipation control instructions by the control strategy generation unit comprises: Based on the predicted thermal behavior data, temperature stability analysis is performed to generate temperature stability evaluation data; Based on the power consumption metering data in the multi-source fusion data, energy efficiency analysis is performed to generate cooling energy efficiency evaluation data; Based on the current temperature distribution data, execution mechanism coordination analysis is performed to generate coordination control evaluation data; The temperature stability evaluation data, the cooling energy efficiency evaluation data, and the coordination control evaluation data are comprehensively evaluated to generate comprehensive evaluation data; Based on the comprehensive evaluation data, fan control instruction data, cooling medium control instruction data, and valve control instruction data are generated; The fan control instruction data, the cooling medium control instruction data, and the valve control instruction data are combined into the set of heat dissipation control instructions. 7.The intelligent heat dissipation control method of a server according to claim 6, wherein, The step of generating fan control instruction data, cooling medium control instruction data, and valve control instruction data based on the comprehensive evaluation data comprises: Based on the comprehensive evaluation data, fan control parameter data, cooling medium control parameter data, and valve control parameter data are extracted; The fan control parameter data is converted into specific fan speed instruction data, the cooling medium control parameter data is converted into specific cooling medium flow instruction data, and the valve control parameter data is converted into specific valve opening degree instruction data; The fan speed instruction data, the cooling medium flow instruction data, and the valve opening degree instruction data are taken as the fan control instruction data, the cooling medium control instruction data, and the valve control instruction data. 8.The intelligent heat dissipation control method of a server according to claim 1, wherein, The method further comprises: Obtaining heat dissipation control effect data, which includes actual response data after the heat dissipation execution mechanism executes the set of heat dissipation control instructions and new sensor acquisition data; Inputting the heat dissipation control effect data into a learning update unit to generate learning update data by the learning update unit; Based on the learning update data, updating the digital twin model and the control strategy generation unit. 9.The intelligent heat dissipation control method of a server according to claim 8, wherein, The step of updating the digital twin model and the control strategy generation unit based on the learning update data comprises: Comparing the heat dissipation control effect data with expected control effect data to generate error evaluation data; Based on the error evaluation data, updating the internal parameters of the digital twin model to generate an updated digital twin model; Based on the heat dissipation control effect data, a control parameter of the control strategy generation unit is updated to generate an updated control strategy generation unit.

10. An intelligent heat dissipation control system of a server, characterized in that, The intelligent heat dissipation control system of the server comprises a memory, a processor, and an intelligent heat dissipation control program of the server stored in the memory and capable of running on the processor. The intelligent heat dissipation control program of the server is configured to implement the steps of the intelligent heat dissipation control method of the server according to any one of claims 1 to 9.

Images