Intelligent temperature control system and method based on chip heat energy guide

Abstract

The invention provides an intelligent temperature control system and method based on chip heat energy guiding. The method comprises the steps of obtaining running state information of a target chip set; the working mode of the loop heat pipe is determined according to the operation state information of the target chip set, and stepped energy recovery is conducted on conduction heat of the loop heat pipe through the recovery heat pump; thermal state information of the target chipset and energy recovery information of the recovery heat pump in different historical time windows are obtained, correlation verification analysis is carried out according to the thermal state information and the energy recovery information, and an intermittent boiling threshold value of the loop heat pipe is determined; acquiring operation instruction stream data of the chip, and performing dynamic power consumption prediction according to the operation instruction stream data to obtain a dynamic power consumption prediction value; when the dynamic power consumption predicted value is higher than the intermittent boiling threshold value, the liquid return rate of the loop heat pipe is adjusted based on the dynamic power consumption predicted value, the liquid return rate in the heat transfer process can be controlled according to dynamic power consumption prediction, and the heat energy guiding and recycling efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of chip thermal energy guidance, and more specifically, to an intelligent temperature control system and method based on chip thermal energy guidance. Background Art

[0002] With the continued development of technologies such as artificial intelligence, high-performance computing, and big data processing, chip performance continues to improve, integration levels are increasing, and power consumption per unit area continues to rise. The resulting high heat flux problem has become a major factor limiting chip stability, reliability, and lifespan. Currently, chip systems deployed in data centers, edge computing devices, and high-computing terminals convert over 90% of the electrical energy consumed during operation into heat. If heat is not promptly dissipated, the chip temperature will rapidly increase, leading to frequency degradation, system instability, and even device damage.

[0003] Chip cooling technologies primarily include air cooling, liquid cooling, and heat pipe cooling. These technologies primarily focus on discharging chip heat from the system as quickly as possible to ensure safe operation. However, with the advent of green computing, cooling methods that simply discard heat have gradually exposed two significant problems: first, low energy efficiency, as a large amount of heat released by the chip is directly consumed and dissipated by the cooling system, resulting in energy waste; second, the cooling system itself consumes additional power, further increasing the overall system's energy burden and carbon emissions. Therefore, how to conduct thermal management during chip operation to improve the efficiency of chip waste heat utilization has become an urgent issue. Summary of the Invention

[0004] The present application provides an intelligent temperature control system and method based on chip thermal energy guidance, which can control the liquid return rate of the heat transfer process according to dynamic power consumption prediction, thereby improving the thermal energy guidance recovery efficiency.

[0005] In a first aspect, the present application provides an intelligent temperature control method based on chip thermal energy guidance, which can be executed by a network device, or by a chip configured in the network device, and the present application does not limit this.

[0006] Specifically, the method includes:

[0007] Get the target chipset's operating status information;

[0008] A loop heat pipe is used as a heat conduction structure to conduct heat to the target chipset, and an operating mode of the loop heat pipe is determined according to the operating status information of the target chipset, and heat conducted by the loop heat pipe is recovered in a cascaded manner through a recovery heat pump;

[0009] Acquiring thermal state information of the target chipset and energy recovery information of the recovery heat pump within different historical time windows, performing correlation verification analysis based on the thermal state information and the energy recovery information, and determining an intermittent boiling threshold of the loop heat pipe;

[0010] Obtain the chip's operating instruction stream data, perform dynamic power consumption prediction based on the operating instruction stream data, and obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, adjust the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value.

[0011] In combination with the first aspect, in certain implementations of the first aspect, in the process of adjusting the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value, the liquid return rate of the loop heat pipe is adjusted by controlling the opening of the liquid return passage valve.

[0012] In combination with the first aspect, in some implementations of the first aspect, the operating status information of the target chipset includes a core operating temperature value and a dynamic real-time power consumption value of the target chipset.

[0013] In conjunction with the first aspect, in certain implementations of the first aspect, determining the operating mode of the loop heat pipe according to the operating status information of the target chipset specifically includes:

[0014] Obtaining the core operating temperature value and the dynamic real-time power consumption value in the operating status information, and using threshold judgment logic to make an operating mode decision;

[0015] When the core operating temperature value is lower than a first temperature threshold and the dynamic real-time power consumption value is lower than a first power consumption threshold, determining that the loop heat pipe is in a low power standby mode;

[0016] When the core operating temperature value is between a first temperature threshold and a second temperature threshold, and the dynamic real-time power consumption value is between a first power consumption threshold and a second power consumption threshold, determining that the loop heat pipe is in a normal heat transfer mode;

[0017] When the core operating temperature value is higher than the second temperature threshold or the dynamic real-time power consumption value is higher than the second power consumption threshold, it is determined that the loop heat pipe is in the enhanced heat transfer mode.

[0018] In conjunction with the first aspect, in certain implementations of the first aspect, performing correlation verification analysis based on the thermal state information and the energy recovery information to determine the intermittent boiling threshold of the loop heat pipe specifically includes:

[0019] Performing heat pump efficiency detection on the thermal state information of the target chipset and the energy recovery information of the recovery heat pump in different historical time windows to obtain heat pump equivalent efficiencies corresponding to the different historical time windows;

[0020] Time series fitting is performed based on the heat pump equivalent efficiency corresponding to different historical windows to obtain a heat pump efficiency trend function, and thermal state information corresponding to the function inflection point of the heat pump efficiency trend function is determined. Mapping is performed based on the thermal state information to determine the intermittent boiling threshold of the loop heat pipe.

[0021] In conjunction with the first aspect, in certain implementations of the first aspect, performing dynamic power consumption prediction based on the running instruction stream data to obtain the dynamic power consumption prediction value specifically includes:

[0022] Obtaining the running instruction stream data within different historical time windows and performing feature vector conversion to obtain the running feature vectors corresponding to the different historical time windows. The different dimensions of the running feature vectors represent the number of instructions per cycle, floating-point operation instruction ratio, instruction emission rate, and memory access frequency within the corresponding historical time window.

[0023] The operating feature vectors corresponding to the historical time windows are clustered to obtain the operating power consumption values ​​corresponding to different historical time windows; a moving average autoregressive model of the operating power consumption values ​​corresponding to different historical time windows is constructed, and the dynamic power consumption prediction value of the next cycle is predicted.

[0024] In conjunction with the first aspect, in certain implementations of the first aspect, performing cascade energy recovery on the heat conducted by the loop heat pipe by using a recovery heat pump specifically includes:

[0025] A primary heat pump system is provided that is coupled to the hot surface of the loop heat pipe condenser to absorb and increase the temperature of the waste heat conducted by the loop heat pipe and transfer the heat to the first target heat load;

[0026] A secondary heat pump is provided which is connected to the discharge end of the primary heat pump system to recover the waste heat which is not fully recovered by the primary heat pump at low temperature.

[0027] In a second aspect, the present application provides an intelligent temperature control system based on chip thermal energy guidance, which includes a heat conduction control unit, and the heat conduction control unit includes:

[0028] Chip status acquisition module, used to obtain the operating status information of the target chipset;

[0029] a heat recovery module, configured to conduct heat to the target chipset using a loop heat pipe as a heat conduction structure, determine an operating mode of the loop heat pipe based on operating status information of the target chipset, and perform cascade energy recovery of the conducted heat of the loop heat pipe using a recovery heat pump;

[0030] a liquid return control module, configured to obtain thermal status information of the target chipset and energy recovery information of the recovery heat pump within different historical time windows, perform correlation verification analysis based on the thermal status information and the energy recovery information, and determine an intermittent boiling threshold of the loop heat pipe;

[0031] The liquid return control module is also used to obtain the chip's operating instruction stream data, perform dynamic power consumption prediction based on the operating instruction stream data, and obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, adjust the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value.

[0032] In a third aspect, the present application provides a computer terminal device, which includes a memory and a processor, wherein the memory stores code, and the processor is configured to obtain the code and execute the above-mentioned intelligent temperature control method based on chip thermal energy guidance.

[0033] In a fourth aspect, the present application provides a computer-readable storage medium, which stores at least one computer program, and the computer program is loaded and executed by a processor to implement the operations performed by the above-mentioned intelligent temperature control method based on chip thermal energy guidance.

[0034] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0035] In an intelligent temperature control system and method based on chip thermal energy guidance provided by the present application, the operating status information of the target chipset is first obtained; a loop heat pipe is used as a heat conduction structure to conduct heat to the target chipset, and the working mode of the loop heat pipe is determined according to the operating status information of the target chipset, and the conducted heat of the loop heat pipe is recovered in a step-by-step manner through a recovery heat pump; the thermal status information of the target chipset and the energy recovery information of the recovery heat pump in different historical time windows are obtained, and a correlation verification analysis is performed based on the thermal status information and the energy recovery information to determine the intermittent boiling threshold of the loop heat pipe; the operating instruction stream data of the chip is obtained, and dynamic power consumption prediction is performed based on the operating instruction stream data to obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, the return liquid rate of the loop heat pipe is adjusted based on the dynamic power consumption prediction value.

[0036] Therefore, it can be seen that the present application determines the intermittent boiling threshold of the loop heat pipe according to the abnormal change range of the heat recovery efficiency through the correlation analysis of historical thermal states and energy recovery data, that is, the safety limit of the power consumption prediction value and the heat transfer limit. By dynamically predicting the power consumption of the operating instruction stream data, the feedforward regulation based on the power consumption prediction is combined with the closed-loop control of the return liquid rate, which significantly shortens the response time of the thermal control system, reduces thermal fluctuations and temperature peaks, and improves the stability of the chip working environment, thereby breaking through the time lag limitation of traditional passive temperature feedback regulation, enabling the system to actively make an adjustment response before the heat load changes, and adjust the return liquid rate of the liquid in the loop heat pipe according to the dynamic power consumption prediction value, avoiding temperature fluctuations inside the heat pipe, resulting in a decrease in the energy recovery efficiency of the heat pump, and improving the thermal energy guidance recovery efficiency.

[0037] In summary, the present application can control the liquid return rate of the heat transfer process according to the dynamic power consumption prediction, thereby improving the heat energy guidance recovery efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is an exemplary flow chart of an intelligent temperature control method based on chip thermal energy guidance according to some embodiments of the present application;

[0039] Figure 2 is a schematic structural diagram of a heat conduction control unit according to some embodiments of the present application;

[0040] Figure 3 This is a structural diagram of a computer terminal device that implements an intelligent temperature control method based on chip thermal energy guidance according to some embodiments of the present application. DETAILED DESCRIPTION

[0041] The present application obtains the operating status information of the target chipset; uses a loop heat pipe as a heat conduction structure to conduct heat to the target chipset, and determines the working mode of the loop heat pipe according to the operating status information of the target chipset, and performs cascade energy recovery of the conducted heat of the loop heat pipe through a recovery heat pump; obtains the thermal status information of the target chipset and the energy recovery information of the recovery heat pump in different historical time windows, performs correlation verification analysis based on the thermal status information and the energy recovery information, and determines the intermittent boiling threshold of the loop heat pipe; obtains the operating instruction stream data of the chip, and performs dynamic power consumption prediction based on the operating instruction stream data to obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, the return liquid rate of the loop heat pipe is adjusted based on the dynamic power consumption prediction value, which can control the return liquid rate of the heat transfer process according to the dynamic power consumption prediction, thereby improving the thermal energy guidance recovery efficiency.

[0042] In order to better understand the above technical solution, the following will be combined with the accompanying drawings and specific implementation methods to describe the above technical solution in detail. Figure 1 , which is an exemplary flow chart of an intelligent temperature control method based on chip thermal energy guidance according to some embodiments of the present application. The intelligent temperature control method 100 based on chip thermal energy guidance mainly includes the following steps:

[0043] In step S101, the operating status information of the target chipset is obtained.

[0044] It should be noted that the target chipset described in this application is a combination of one or more integrated circuit chip units that serve as the main heat source or computing unit in the current application scenario and require temperature monitoring, power consumption prediction and thermal management control. Preferably, in some embodiments, the operating status information of the target chipset includes the core operating temperature value and dynamic real-time power consumption value of the target chipset; in specific implementation, the core operating temperature value can be obtained by detecting multiple digital temperature sensors installed on the target chipset. In some other embodiments, other parameters and data that can represent the chip operating status can also be collected as the operating status information of the target chipset, and this application does not limit this.

[0045] Specifically, the target chipset is equipped with multiple temperature collection points, which use the digital temperature sensor inside the chip to collect core temperature information in real time; at the same time, the power supply voltage, current and clock frequency of the chip are collected through the power consumption management unit; the current thread information of the chip and the instruction execution statistics are combined to construct an operation status data set to form the operation status information, and it is regularly uploaded to the control module for dynamic judgment of thermal load status and prediction of thermal risks.

[0046] In step S102, a loop heat pipe is used as a heat conduction structure to conduct heat to the target chipset, and the working mode of the loop heat pipe is determined according to the operating status information of the target chipset, and the heat conducted by the loop heat pipe is recovered in a stepwise manner through a recovery heat pump.

[0047] It should be noted that the loop heat pipe is a passive and efficient heat transfer device based on capillary force-driven working fluid circulation. It is mainly composed of an evaporator, a compensation chamber, a condenser, a steam pipeline and a liquid pipeline. The evaporator is used to absorb the heat generated by the heat source and provide continuous capillary suction through the internal capillary wick structure to drive the working fluid to evaporate; the compensation chamber is connected to the evaporator, which is used to store the working fluid and maintain pressure stability during operation; the condenser is thermally coupled with the external cooling environment or the heat pump heat exchange module to condense the steam and release heat; the steam pipeline and the liquid pipeline transmit steam and liquid respectively, and the gas-liquid channels of the two are separated, which can effectively reduce shear interference during the circulation process. Compared with the traditional heat pipe structure, the loop heat pipe has superior performance in long-distance heat transfer capacity, directional adaptability and operation stability, and is suitable for high heat flux density chips and heat energy recovery scenarios.

[0048] The operating mode of the loop heat pipe in this application refers to the adjustment combination of parameters such as the liquid return path inside the loop heat pipe, the activation degree of the evaporation zone, and the pressure state of the compensation chamber, so as to achieve different chipset heat transfer requirements. Preferably, in some embodiments, determining the operating mode of the loop heat pipe based on the operating status information of the target chipset specifically includes:

[0049] Obtaining the core operating temperature value and the dynamic real-time power consumption value in the operating status information, and using threshold judgment logic to make an operating mode decision;

[0050] When the core operating temperature value is lower than a first temperature threshold and the dynamic real-time power consumption value is lower than a first power consumption threshold, determining that the loop heat pipe is in a low power standby mode;

[0051] When the core operating temperature value is between a first temperature threshold and a second temperature threshold, and the dynamic real-time power consumption value is between a first power consumption threshold and a second power consumption threshold, determining that the loop heat pipe is in a normal heat transfer mode;

[0052] When the core operating temperature value is higher than the second temperature threshold or the dynamic real-time power consumption value is higher than the second power consumption threshold, it is determined that the loop heat pipe is in the enhanced heat transfer mode.

[0053] Among them, the temperature threshold and power consumption threshold are experimentally calibrated based on the rated power and chip parameters of the target chipset. This application does not limit this. Taking a low-power experimental chipset as an example, Qualcomm Snapdragon8Gen 2SoC is used as the target chipset experimental object, including an octa-core CPU, Adreno 740GPU and a mobile terminal chipset with an integrated neural network processing unit. When the chip operating power consumption continues to be higher than 6W and the temperature rise rate exceeds 1°C / s, the system controls the loop heat pipe to enter the enhanced heat transfer mode and switch to the strong + bypass condensation working mode of the loop heat pipe. At the same time, the heat pump primary loop is turned on and the negative pressure of the compensation chamber is adjusted to increase the steam flow rate and liquid return rate.

[0054] In some specific embodiments of the present application, the evaporator section of the loop heat pipe is arranged close to the high heat flux area of ​​the chipset, the real-time heat load status of the chip is obtained through the chip operation status judgment module, different heat conduction working modes are selected according to the judgment results, and corresponding control strategies are adopted according to different working modes. In specific implementation, in the enhanced heat transfer mode, the loop heat pipe activates the main condensation path and the secondary channel, and adjusts the capillary resistance to the first resistance value. In the low power standby mode, the loop heat pipe closes the secondary channel, the main condensation path maintains a low thermal conductivity state, and adjusts the capillary resistance to the second resistance value ( Greater than the first resistance value) to limit the liquid reflux velocity, achieve energy saving and reduce heat transfer efficiency; in normal heat transfer mode, the loop heat pipe activates the main condensation path, closes the secondary channel, and adjusts the capillary resistance to the third resistance value (between the first resistance value and the second resistance value) to maintain a stable liquid reflux velocity and achieve balanced heat transfer performance and energy consumption; further, the heat pump condenser is coupled with the condensing end face of the loop heat pipe to construct a first-level evaporation loop of the heat pump to achieve heat energy extraction. The multi-stage heat pump series connection can achieve cascade energy utilization, and ultimately the excess heat energy is introduced into the heating platform or heat storage equipment.

[0055] It should be noted that the present application uses a heat pump to perform cascade energy recovery on the heat conducted by the loop heat pipe, which can effectively reduce the overall energy consumption of the system, improve the heat utilization efficiency of the heating platform or other heat application areas, and meet the heating needs of different temperature levels in multiple scenarios. Optionally, in some embodiments, the heat conducted by the loop heat pipe is recovered in a cascade energy recovery manner by a recovery heat pump, specifically including:

[0056] A primary heat pump system is provided that is tightly coupled to the hot surface of the loop heat pipe condenser, for absorbing and raising the temperature of waste heat conducted by the loop heat pipe, and transferring the heat to a first target heat load;

[0057] A secondary heat pump is provided which is connected to the discharge end of the primary heat pump system to recover the waste heat which is not fully recovered by the primary heat pump at low temperature;

[0058] According to the operating status and heat load requirements of the target chipset, a fuzzy control strategy is adopted to dynamically adjust the operating parameters of each level of heat pump.

[0059] In step S103, the thermal state information of the target chipset and the energy recovery information of the recovery heat pump in different historical time windows are obtained, and an association verification analysis is performed based on the thermal state information and the energy recovery information to determine the intermittent boiling threshold of the loop heat pipe.

[0060] Optionally, in some embodiments, the size of the historical time window may be calibrated according to the temperature fluctuation rate of the target chipset. For example, the size of the historical time window may be calibrated to 30S.

[0061] Preferably, in some embodiments, the thermal state information includes: the average temperature value, temperature change rate and chip operating power consumption of the target chipset in different historical time windows, and the energy recovery information of the recovery heat pump includes: the average heat absorption power of the recovery heat pump in different historical time windows, wherein the heat absorption power is the heat power absorbed by the heat pump from the loop heat pipe condenser, which is calculated by the heat pump inlet temperature, flow rate and heat capacity. For example, the heat absorption power, that is, the heat absorbed by the heat pump per unit time = the specific heat capacity of the circulating fluid × the density of the working fluid × the flow rate of the working fluid per unit time × the temperature difference between the inlet and outlet of the heat pump. It should be noted that the heat absorption power value described in this application is used to reflect the heat energy transfer efficiency between the heat pump and the loop heat pipe, and serves as an important basis for judging whether intermittent boiling occurs in the system; in actual applications, the original data of the above parameters in the corresponding time window can be obtained through the real-time acquisition module, and a standardized indicator set can be generated through data preprocessing and averaging to support subsequent correlation analysis and boiling threshold identification. This application will not go into details about this.

[0062] Preferably, in some embodiments, performing correlation verification analysis based on the thermal state information and the energy recovery information to determine the intermittent boiling threshold of the loop heat pipe specifically includes:

[0063] Performing heat pump efficiency detection on the thermal state information of the target chipset and the energy recovery information of the recovery heat pump in different historical time windows to obtain heat pump equivalent efficiencies corresponding to the different historical time windows;

[0064] Time series fitting is performed based on the heat pump equivalent efficiency corresponding to different historical windows to obtain a heat pump efficiency trend function, and thermal state information corresponding to the function inflection point of the heat pump efficiency trend function is determined. Mapping is performed based on the thermal state information to determine the intermittent boiling threshold of the loop heat pipe.

[0065] In a specific implementation, the multidimensional data vector composed of the thermal state information can be used, and different dimensions of the data vector respectively represent the average temperature value, temperature change rate and chip operating power consumption of the target chipset in the historical time window. The multidimensional data vector is input into a single hidden layer neural network for clustering scoring, and the corresponding loop heat pipe energy score is output from the clustering result. In a specific implementation, the multidimensional thermal state data vector is input into the single hidden layer neural network model, and a plurality of activation function nodes are provided in the hidden layer of the neural network. The ReLU activation function is used to realize nonlinear mapping capability, which is used for high-dimensional feature expression and cluster classification training of the input thermal state sample. The output layer of the single hidden layer neural network outputs the clustering result to which the current thermal state vector belongs, and the clustering result is mapped to the loop heat pipe energy through a predefined thermal conduction efficiency grading system. Score, which is used to reflect the overall heat transfer performance of the loop heat pipe system under the current working conditions. In order to improve the reliability of the clustering score, a training set sample containing a large amount of historical thermal state data and manually evaluated and annotated energy scores is prepared in advance as the training data set of the neural network. Classification training is performed by inputting multiple historical thermal state samples, and the classification results are compared with the corresponding manual energy score labels for error evaluation. When the matching degree (similarity coefficient) between the clustering result and the manual score result is lower than the set threshold, the back propagation algorithm is used to perform gradient optimization adjustment on the activation parameters in the hidden layer of the neural network until the matching degree between the classification result and the manual energy score meets the preset standard. After the training is completed, the system can obtain the loop heat pipe energy score corresponding to different historical time windows through the thermal state data corresponding to different historical time windows.

[0066] Optionally, in some embodiments, in order to quantify the heat energy recovery capacity of the heat pump under different operating conditions, the ratio between the energy score of the loop heat pipe and the heat absorption power in different historical time windows is used as the heat pump equivalent efficiency corresponding to different historical time windows. The equivalent efficiency index takes into account the thermal conductivity performance of the loop heat pipe under the current operating conditions (reflected in the energy score) and the actual heat recovery amount (heat absorption power) of the heat pump, and can be used to dynamically judge the current operating matching degree and thermal energy response capability of the heat pump. When the equivalent efficiency of multiple consecutive time windows shows a downward trend, it can be regarded as the system may enter the intermittent boiling risk zone. In specific implementation, the heat pump equivalent efficiency corresponding to different historical windows can be used for time series fitting to obtain the heat pump efficiency trend function to determine the heat pump efficiency. The system analyzes the first-order derivative and the second-order derivative of the thermal state information corresponding to the inflection point of the rate trend function to extract the inflection point characteristics in the function, that is, the inflection point where the heat pump equivalent efficiency drops significantly or tends to be unstable. The time window corresponding to the inflection point is regarded as the dividing point where the system energy efficiency enters the critical fluctuation zone from the high-efficiency stable zone. The thermal state information corresponding to the inflection point time window is further extracted, and the threshold interval of the thermal state information and the preset thermal state mapping table are mapped to the corresponding intermittent boiling threshold. In specific implementation, the thermal state mapping table can be calibrated by the mapping relationship between the chip dynamic operating power consumption and the chip thermal state information detected under different experimental conditions. This application will not go into details about this.

[0067] It should be noted that the intermittent boiling threshold described in this application is a chipset power consumption threshold determined based on multi-dimensional thermal state information, which means that under the current thermal management state, if the chip operating power consumption exceeds the intermittent boiling threshold, the high heat load will cause temperature fluctuations in the loop heat pipe, and then the temperature fluctuations caused by intermittent boiling will lead to a decrease in the heat recovery efficiency of the heat pump.

[0068] In step S104, the chip's operating instruction stream data is obtained, and dynamic power consumption prediction is performed based on the operating instruction stream data to obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, the liquid return rate of the loop heat pipe is adjusted based on the dynamic power consumption prediction value.

[0069] It should be noted that the heat-generating power consumption of a chip mainly consists of three parts: peripheral circuit power consumption, static power consumption, and dynamic power consumption. Among them, peripheral circuit power consumption and static power consumption are relatively stable and change little with chip performance, while dynamic power consumption is mainly affected by the chip's operating frequency and voltage. The chip's operating frequency is usually dynamically adjusted according to the current computing load and system requirements. When the data processing volume is large, the chip will automatically increase the operating frequency to meet performance requirements, resulting in a significant increase in dynamic power consumption, which in turn generates more heat, affecting the chip's temperature state and overall work efficiency. The chip's running instruction stream data reflects the instruction sequence currently executed by the chip and its complexity, and directly reflects the chip's load intensity and operating status. Specifically, information such as the number, type, distribution, and execution rate of the running instruction stream is the core driving factor for the generation of dynamic power consumption. Based on the instruction stream data, the chip's real-time computing needs and resource usage can be captured, and then by establishing a power consumption model, real-time prediction of dynamic power consumption can be achieved.

[0070] Preferably, in some embodiments, the execution instruction stream data can be collected through a driver interface provided by the chip manufacturer, including but not limited to a performance monitoring unit, an on-chip scheduler data interface, or an operating system-level performance analysis tool interface. The execution instruction stream data includes key operational characteristics such as the number of instructions per cycle within the chip, branch prediction hit rate, cache access count, and floating-point / integer operation ratio, which are used to characterize the current real-time load of the chip and the characteristics of the instruction structure distribution.

[0071] Optionally, in some embodiments, performing dynamic power consumption prediction based on the running instruction stream data to obtain the dynamic power consumption prediction value specifically includes:

[0072] Obtaining the running instruction stream data within different historical time windows and performing feature vector conversion to obtain the running feature vectors corresponding to the different historical time windows. The different dimensions of the running feature vectors represent the number of instructions per cycle, floating-point operation instruction ratio, instruction emission rate, and memory access frequency within the corresponding historical time window.

[0073] The operating feature vectors corresponding to the historical time windows are clustered to obtain the operating power consumption values ​​corresponding to different historical time windows; a moving average autoregressive model of the operating power consumption values ​​corresponding to different historical time windows is constructed, and the dynamic power consumption prediction value of the next cycle is predicted.

[0074] In specific implementation, the operating instruction stream data within different historical time windows is obtained, and the original instruction data is standardized, preprocessed, and dimensionally reconstructed to convert it into a structured operating feature vector. A historical power consumption sample library is constructed based on the above operating feature vector, and the K-means clustering algorithm is used to perform cluster analysis on the historical operating characteristics. In combination with the actual measured or estimated chip power consumption data, the typical operating power consumption states represented by various cluster centers are calibrated. The following is a specific embodiment of the present application using the K-means clustering algorithm to perform feature clustering on the operating feature vectors corresponding to the historical time windows to obtain the operating power consumption values ​​corresponding to different historical time windows:

[0075] Construct a sample set for cluster analysis. The sample set is derived from the operating status of the target chipset under different operating instruction stream data conditions, collected in an experimental environment. The samples include: an operating feature vector within a running time window, i.e., preprocessing and vectorizing the operating instruction stream data. The different dimensions of the feature vector include: instructions per cycle (IPC), floating-point instruction ratio, instruction emission rate, memory access frequency, etc.; and dynamic power consumption values ​​(in watts (W)) measured by thermocouples or on-chip thermal sensors, corresponding to the operating feature vectors. Cluster analysis of the above sample set is performed using the K-means clustering algorithm. Specifically, all historical samples are divided into K categories using the operating feature vector as the input feature to form several clusters with similar operating states. Each cluster represents a type of operating behavior characteristics. After the clustering is completed, for the samples within each cluster, the corresponding dynamic power consumption measured values ​​are aggregated to calculate the average operating power consumption value of the cluster as the representative power consumption level of this type of operating state. In the actual prediction stage, the operating feature vector of the current time window is matched with the above-mentioned historical clustering model to identify the cluster to which it belongs, and the representative average power consumption value of the cluster is directly referenced as the operating power consumption prediction value at the current moment.

[0076] A specific embodiment of determining the dynamic power consumption prediction value of the present application is given below:

[0077] The preset prediction period is 5 minutes. During this period, the system can obtain historical operating feature vector data collected in units of 30 seconds in the past 3 hours based on the instruction stream analysis mechanism of the selected chip, and summarize the corresponding historical dynamic power consumption values ​​to form an operating power consumption sequence.

[0078] The operating power consumption time series is then plotted as a time series diagram, with the horizontal axis representing the time step and the vertical axis representing the corresponding dynamic power consumption value. To improve the stability of the model, the power consumption time series can be processed for stationarity, such as by using logarithmic transformation or differencing to eliminate trend terms or fluctuation non-uniformity.

[0079] Subsequently, the autocorrelation coefficient and partial autocorrelation coefficient plots of the operating power consumption value series are plotted to analyze its time-dependent characteristics. The autocorrelation coefficient plot is used to identify the linear correlation of the power consumption series at different time lags. If the autocorrelation coefficient decays sharply to zero after a certain order and remains within the confidence interval, the order of the autoregressive component can be preliminarily determined. The partial autocorrelation coefficient plot is used to analyze the net impact of each lag order on the current value after controlling for the influence of the intermediate variable. When the partial autocorrelation value decreases rapidly and converges to zero after a certain lag period, the order of the moving average component can be determined.

[0080] For example, when the autocorrelation coefficient graph shows that it decays rapidly after the third order, and the partial autocorrelation coefficient graph begins to approach zero after the second order, it can be preliminarily determined that the operating power consumption series is suitable for an autoregressive moving average model of order (3,2).

[0081] In specific implementation, the least squares method can be used to estimate the parameters of the autoregressive coefficient and the moving average coefficient in the model, and further significance test can be carried out. In some embodiments, the test level is set to 0.05. Finally, the information quantity criteria such as the Schwarz Bayes criterion can be used to compare multiple groups of models, and the optimal fitting model can be selected as the operating power consumption prediction model; the constructed operating power consumption prediction model is applied to the operating feature data collected in the current time window, and the dynamic power consumption prediction value in the next prediction period can be output.

[0082] It should be noted that the heat transfer performance of the loop heat pipe is affected by many factors such as shape, size, and physical properties of the working fluid. Under the influence of these factors, a specific loop heat pipe has a fixed maximum heat load. If the heat load of the heat source exceeds the maximum heat load, the loop heat pipe will not work normally and may even be damaged. Preferably, in some embodiments, dynamic power consumption prediction is performed based on the running instruction stream data. After obtaining the dynamic power consumption prediction value, the method further includes: obtaining the maximum heat transfer limit (in watts W) of the target loop heat pipe. The maximum heat transfer limit is determined based on the operating limits of the loop heat pipe under experimental conditions, such as the carrying limit, capillary limit, sonic limit, boiling limit, and viscosity limit. In some other embodiments, the maximum heat transfer limit can also be calibrated based on the factory test of the loop heat pipe; and then, based on the threshold range of the maximum heat transfer limit, the maximum dynamic power consumption prediction value of the target chipset is determined by linear mapping; when the dynamic power consumption prediction value obtained based on the running instruction stream prediction exceeds the maximum dynamic power consumption prediction value, a temperature control alarm message is issued.

[0083] The loop heat pipe relies on the heat absorption and vaporization of the working fluid to transfer heat, and usually operates under a stable phase change mechanism such as thin liquid film boiling or nucleate boiling. Although the heat load is not exceeded, when it exceeds a certain local boiling critical heat flux, that is, the chip power consumption corresponding to the intermittent boiling threshold in this application, it will cause the following problems: the working fluid vaporizes violently in a short time on the evaporation surface; the local liquid film is completely evaporated and temporarily fails (dry area), and then is replenished with liquid to form a vaporization cycle manifested as periodic bubble bursts and evaporation unevenness, which is called intermittent boiling. Intermittent boiling will cause temperature fluctuations in the loop heat pipe, and then under the temperature fluctuations caused by intermittent boiling, the efficiency of the heat pump condensing section will decrease, which is equivalent to a decrease in the heat pump COP (coefficient of performance), affecting the energy recovery efficiency of the chip thermal energy guidance.

[0084] Preferably, in some embodiments, the return liquid rate of the loop heat pipe is regulated by a liquid circuit structure with an electrically controlled microchannel valve. By adjusting the opening of the microchannel valve, the effective cross-sectional area of ​​the return liquid channel is changed, thereby finely controlling the rate at which liquid from the condensing section returns to the evaporating section. The electrically controlled microchannel valve can receive return liquid rate control instructions through closed-loop PID control logic and adjust the valve opening in real time to match changes in the heat load.

[0085] In some other embodiments, the liquid return rate can also be adjusted by electrically regulating the internal pressure of the compensation chamber. By controlling the gas volume within the compensation chamber or increasing or decreasing the pressure of the compensation chamber using an external micro-vacuum pump / compression module, the capillary pressure differential between the compensation chamber and the two ends of the capillary wick can be adjusted. This capillary pressure differential directly affects the driving force for the return of the working fluid within the capillary structure. When the liquid return rate needs to be accelerated to cope with high heat loads or the risk of intermittent boiling, the compensation chamber pressure can be reduced to increase capillary suction. Conversely, when the heat load decreases, the compensation chamber pressure can be appropriately increased to suppress the liquid return rate and avoid flow disturbances.

[0086] Furthermore, in certain high-precision temperature control scenarios, the system can also be combined with a micro flow meter or liquid return rate sensor to monitor the liquid return rate in real time, forming a liquid return closed-loop control system, realizing a dynamic liquid return adjustment response mechanism based on the heat load prediction value, and improving the real-time performance and stability of the thermal control system.

[0087] Optionally, in some embodiments, when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, in the process of adjusting the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value, the liquid return rate of the loop heat pipe is adjusted by controlling the opening of the liquid return passage valve, wherein the opening control signal of the liquid return passage valve is determined according to the following formula:

[0088]

[0089] Among them, u(t) is the opening control signal of the return liquid passage valve, K p , Ki , K d are proportional, integral, and differential coefficients, respectively, which are calibrated to constants based on experience. e(t) is the error signal, which is represented in this application as the difference signal between the dynamic power consumption prediction value and the intermittent boiling threshold. T s is the control cycle sampling time, △e(t) is the change value of the error signal per unit time, dt is the time signal differential, t0 is the start time of the current control cycle, and t is the current moment.

[0090] It should be noted that loop heat pipes exhibit a certain degree of response lag during the actual heat transfer process. Their evaporation-condensation-liquid return cycle has a time constant, making it unable to respond in real time to changes in heat source power consumption. Passive adjustments based solely on the current temperature state can easily lead to intense local evaporation or delayed liquid return, triggering intermittent boiling and causing dramatic temperature fluctuations within the heat pipe. In severe cases, this can even cause air blockage or a sudden drop in heat transfer efficiency, impacting system stability and chip reliability. To overcome these issues, this application proposes dynamic power consumption prediction based on the operating instruction stream. This allows for early perception and response to impending changes in the chip's thermal load, thereby driving the system to adjust the liquid return rate in advance, effectively reducing the risk of thermal inertia shock. Specifically, when the predicted value approaches the intermittent boiling threshold, a PID or fuzzy logic controller fine-tunes the microchannel valve opening or compensates for the negative pressure in the chamber to preemptively increase the liquid return rate. This not only avoids heat transfer instability caused by sudden phase changes in the energy source, but also maintains heat exchange stability in the heat pump condenser section, improving overall heat recovery efficiency.

[0091] In addition, in another aspect of the present application, in some embodiments, the present application provides an intelligent temperature control system based on chip thermal energy guidance, the system includes a heat conduction control unit, reference Figure 2 , which is a schematic diagram of exemplary hardware and / or software structure of a heat conduction control unit according to some embodiments of the present application. The heat conduction control unit 200 includes: a chip status acquisition module 201, a heat recovery module 202, and a liquid return control module 204, which are described as follows:

[0092] Chip status acquisition module 201, used to obtain the operating status information of the target chipset;

[0093] a heat recovery module 202 for conducting heat to the target chipset using a loop heat pipe as a heat conduction structure, determining an operating mode of the loop heat pipe based on operating status information of the target chipset, and performing cascade energy recovery of the heat conducted by the loop heat pipe using a recovery heat pump;

[0094] The liquid return control module 203 is configured to obtain thermal status information of the target chipset and energy recovery information of the recovery heat pump within different historical time windows, perform correlation verification analysis based on the thermal status information and the energy recovery information, and determine the intermittent boiling threshold of the loop heat pipe;

[0095] The liquid return control module 203 is also used to obtain the chip's operating instruction stream data, perform dynamic power consumption prediction based on the operating instruction stream data, and obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, adjust the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value.

[0096] The above describes in detail an example of an intelligent temperature control system and method based on chip thermal energy guidance provided in an embodiment of the present application. It can be understood that in order to achieve the above functions, the corresponding device includes hardware structures and / or software modules corresponding to executing each function.

[0097] Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function in the application is executed in hardware or in a computer software-driven hardware manner depends on the specific application and design constraints of the technical solution. Therefore, professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0098] In addition, the present application also provides a computer terminal device, which includes a memory and a processor, the memory stores code, and the processor is configured to obtain the code and execute the above-mentioned intelligent temperature control method based on chip thermal energy guidance.

[0099] In some embodiments, reference Figure 3 , which is a schematic diagram of the structure of a computer terminal device that implements an intelligent temperature control method based on chip thermal energy guidance according to some embodiments of the present application. In the above embodiment, an intelligent temperature control method based on chip thermal energy guidance can be achieved by Figure 3 The computer terminal device 300 shown in FIG. 1 is implemented as shown in FIG. 1 , and the computer terminal device 300 includes at least one communication bus 301 , a communication interface 302 , a processor 303 and a memory 304 .

[0100] The processor 303 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more processors for controlling the execution of an intelligent temperature control method based on chip thermal energy guidance in this application.

[0101] The communication bus 301 may include a path for transmitting information between the aforementioned components.

[0102] The memory 304 may be a read-only memory (ROM) or other type of static storage device that can store static information and instructions, a random access memory (RAM) or other type of dynamic storage device that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, an optical disc storage (including a compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a magnetic disk or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 304 may exist independently and be connected to the processor 303 via the communication bus 301. The memory 304 may also be integrated with the processor 303.

[0103] The memory 304 is used to store program code for executing the solution of the present application, and is controlled by the processor 303 for execution. The processor 303 is used to execute the program code stored in the memory 304. The program code may include one or more software modules. The determination of the intermittent boiling threshold in the above embodiment can be implemented by the processor 303 and one or more software modules in the program code in the memory 304.

[0104] The communication interface 302 uses any transceiver or other device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area network (WLAN), etc.

[0105] Optionally, the computer terminal device 300 may further include a power supply 305 for providing power to various devices or circuits in the real-time computer terminal device.

[0106] In a specific implementation, as an embodiment, a computer terminal device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. The processor herein may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0107] The computer terminal device can be a general-purpose computer terminal device or a dedicated computer terminal device. In a specific implementation, the computer terminal device can be a desktop computer, a portable computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. The embodiments of the present application do not limit the type of computer terminal device.

[0108] In addition, other aspects of the present application further provide a computer-readable storage medium, which stores at least one computer program, and the computer program is loaded and executed by a processor to implement the operations performed by the above-mentioned intelligent temperature control method based on chip thermal energy guidance.

[0109] In summary, in an intelligent temperature control system and method based on chip thermal energy guidance disclosed in an embodiment of the present application, the operating status information of the target chipset is obtained; a loop heat pipe is used as a heat conduction structure to conduct heat to the target chipset, and the working mode of the loop heat pipe is determined according to the operating status information of the target chipset, and the conducted heat of the loop heat pipe is recovered in a step-by-step manner through a recovery heat pump; the thermal status information of the target chipset and the energy recovery information of the recovery heat pump in different historical time windows are obtained, and correlation verification analysis is performed based on the thermal status information and the energy recovery information to determine the intermittent boiling threshold of the loop heat pipe; the operating instruction stream data of the chip is obtained, and dynamic power consumption prediction is performed based on the operating instruction stream data to obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, the return liquid rate of the loop heat pipe is adjusted based on the dynamic power consumption prediction value, which can control the return liquid rate of the heat transfer process according to the dynamic power consumption prediction, thereby improving the thermal energy guidance recovery efficiency.

[0110] The above description is merely an embodiment of the present application. Common knowledge such as the specific technical solutions or features of the solutions is not described in detail herein. It should be noted that those skilled in the art may make various modifications and improvements without departing from the technical solution of the present application, and these modifications and improvements should also be considered within the scope of protection of the present application. These modifications and improvements will not affect the effectiveness of the implementation of the present application or the practical application of the patent.

[0111] The scope of protection claimed by this application shall be determined by the content of the claims. The specific embodiments and other descriptions in the specification may be used to interpret the content of the claims. Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of the invention. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application is intended to include such modifications and variations.

Claims

1. An intelligent temperature control method based on chip thermal energy guidance, characterized in that: include: Get the target chipset's operating status information; A loop heat pipe is used as a heat conduction structure to conduct heat to the target chipset, and an operating mode of the loop heat pipe is determined according to the operating status information of the target chipset, and heat conducted by the loop heat pipe is recovered in a cascaded manner through a recovery heat pump; Acquiring thermal state information of the target chipset and energy recovery information of the recovery heat pump within different historical time windows, performing correlation verification analysis based on the thermal state information and the energy recovery information, and determining an intermittent boiling threshold of the loop heat pipe; Obtain the chip's operating instruction stream data, perform dynamic power consumption prediction based on the operating instruction stream data, and obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, adjust the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value.

2. The method according to claim 1, wherein In the process of adjusting the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value, the liquid return rate of the loop heat pipe is adjusted by controlling the opening of the liquid return passage valve.

3. The method according to claim 1, wherein The operating status information of the target chipset includes a core operating temperature value and a dynamic real-time power consumption value of the target chipset.

4. The method according to claim 1, wherein Determining the operating mode of the loop heat pipe according to the operating status information of the target chipset specifically includes: Obtaining the core operating temperature value and the dynamic real-time power consumption value in the operating status information, and using threshold judgment logic to make an operating mode decision; When the core operating temperature value is lower than a first temperature threshold and the dynamic real-time power consumption value is lower than a first power consumption threshold, determining that the loop heat pipe is in a low power standby mode; When the core operating temperature value is between a first temperature threshold and a second temperature threshold, and the dynamic real-time power consumption value is between a first power consumption threshold and a second power consumption threshold, determining that the loop heat pipe is in a normal heat transfer mode; When the core operating temperature value is higher than the second temperature threshold or the dynamic real-time power consumption value is higher than the second power consumption threshold, it is determined that the loop heat pipe is in the enhanced heat transfer mode.

5. The method according to claim 1, wherein Performing correlation verification analysis based on the thermal state information and the energy recovery information to determine the intermittent boiling threshold of the loop heat pipe specifically includes: Performing heat pump efficiency detection on the thermal state information of the target chipset and the energy recovery information of the recovery heat pump in different historical time windows to obtain heat pump equivalent efficiencies corresponding to the different historical time windows; Time series fitting is performed based on the heat pump equivalent efficiency corresponding to different historical windows to obtain a heat pump efficiency trend function, and thermal state information corresponding to the function inflection point of the heat pump efficiency trend function is determined. Mapping is performed based on the thermal state information to determine the intermittent boiling threshold of the loop heat pipe.

6. The method according to claim 1, wherein Dynamic power consumption prediction is performed based on the running instruction stream data, and the dynamic power consumption prediction value obtained specifically includes: Obtaining the running instruction stream data within different historical time windows and performing feature vector conversion to obtain the running feature vectors corresponding to the different historical time windows. The different dimensions of the running feature vectors represent the number of instructions per cycle, floating-point operation instruction ratio, instruction emission rate, and memory access frequency within the corresponding historical time window. The operating feature vectors corresponding to the historical time windows are clustered to obtain the operating power consumption values ​​corresponding to different historical time windows; a moving average autoregressive model of the operating power consumption values ​​corresponding to different historical time windows is constructed, and the dynamic power consumption prediction value of the next cycle is predicted.

7. The method according to claim 1, wherein The heat conducted by the loop heat pipe is recovered in a cascaded manner by using a recovery heat pump, specifically comprising: A primary heat pump system is provided that is coupled to the hot surface of the loop heat pipe condenser to absorb and increase the temperature of the waste heat conducted by the loop heat pipe and transfer the heat to the first target heat load; A secondary heat pump is provided which is connected to the discharge end of the primary heat pump system to recover the waste heat which is not fully recovered by the primary heat pump at low temperature.

8. An intelligent temperature control system based on chip thermal energy guidance, comprising a heat conduction control unit, wherein the heat conduction control unit is configured to execute the intelligent temperature control method based on chip thermal energy guidance according to any one of claims 1 to 7, characterized in that: The heat conduction control unit comprises: Chip status acquisition module, used to obtain the operating status information of the target chipset; a heat recovery module, configured to conduct heat to the target chipset using a loop heat pipe as a heat conduction structure, determine an operating mode of the loop heat pipe based on operating status information of the target chipset, and perform cascade energy recovery of the conducted heat of the loop heat pipe using a recovery heat pump; a liquid return control module, configured to obtain thermal status information of the target chipset and energy recovery information of the recovery heat pump within different historical time windows, perform correlation verification analysis based on the thermal status information and the energy recovery information, and determine an intermittent boiling threshold of the loop heat pipe; The liquid return control module is also used to obtain the chip's operating instruction stream data, perform dynamic power consumption prediction based on the operating instruction stream data, and obtain a dynamic power consumption prediction value; when the dynamic power consumption prediction value is higher than the intermittent boiling threshold, adjust the liquid return rate of the loop heat pipe based on the dynamic power consumption prediction value.

9. A computer terminal device, characterized in that: The computer terminal device includes a memory and a processor, the memory stores code, and the processor is configured to obtain the code and execute the intelligent temperature control method based on chip thermal energy guidance according to any one of claims 1 to 7.

10. A computer-readable storage medium storing at least one computer program, characterized in that: The computer program is loaded and executed by a processor to implement the operations performed by the intelligent temperature control method based on chip thermal energy guidance as claimed in any one of claims 1 to 7.

Images