Liquid cooling adaptive control method and system based on single-phase immersion liquid cooling data center

Abstract

The application discloses a liquid cooling adaptive control method and system based on a single-phase immersion liquid cooling data center, relates to the technical field of liquid cooling control, and comprises the following steps: reading real-time power consumption data of the data center, synchronously starting a temperature monitoring sensor to read chip temperature data and environment temperature data; configuring a short-time predictor by using a historical load sequence, performing information gain ratio and prediction uncertainty analysis of a preset candidate granularity; performing matching evaluation of the single-phase immersion liquid cooling and the data center by using a calibrated prediction granularity; if a trigger threshold is met, performing adaptive optimization under local-global control coordination constraints of single-phase immersion liquid cooling control, and outputting a liquid cooling response scheme. The technical problems of low matching degree of cooling capacity of single-phase immersion liquid cooling and load demand, response lag and insufficient energy consumption control in the prior art are solved, adaptive optimization control based on real-time and prediction data is achieved, and the technical effects of reducing energy consumption, improving heat dissipation efficiency and temperature control precision are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of liquid cooling control, and particularly relates to a liquid cooling adaptive control method and system based on a single-phase immersion liquid cooling data center. BACKGROUND

[0002] With the development of cloud computing and big data, the computing density of data centers continues to increase, and the power consumption of server chips increases significantly. The traditional air cooling system gradually exposes limitations in terms of heat dissipation capacity, energy consumption control, and noise. Single-phase immersion liquid cooling technology can achieve efficient heat exchange and uniform temperature distribution by completely immersing IT equipment in insulating cooling liquid, significantly reducing the energy consumption required for heat dissipation, and having higher heat dissipation efficiency and lower operating noise. However, in actual application, the load of the data center changes significantly and dynamically and is uncertain. The power consumption of the server and the chip temperature will frequently change with the fluctuation of task scheduling and computing demand. The liquid cooling system cannot respond in time to the changes in load and temperature, resulting in a mismatch between cooling capacity and load demand, causing insufficient heat dissipation or energy waste, and a lag in control strategy, leading to the generation of local hot spots, affecting the stability and life of the chip, and lacking comprehensive analysis of historical load patterns and future trends, which cannot effectively balance energy consumption and temperature control accuracy.

[0003] Therefore, in the related art, there are technical problems of low matching degree between the cooling capacity of single-phase immersion liquid cooling and the load demand, response lag, and insufficient energy consumption control. SUMMARY

[0004] The present application provides a liquid cooling adaptive control method and system based on a single-phase immersion liquid cooling data center, which solves the technical problems of low matching degree between the cooling capacity of single-phase immersion liquid cooling and the load demand, response lag, and insufficient energy consumption control in the prior art, achieves adaptive optimization control based on real-time and predicted data, and reduces energy consumption, improves heat dissipation efficiency and temperature control accuracy.

[0005] The application provides a liquid cooling adaptive control method based on a single-phase immersion liquid cooling data center, which comprises the following steps: after reading real-time power consumption data of the data center, synchronously starting a temperature monitoring sensor to read chip temperature data and environment temperature data, and constructing a temperature data set; after reading a historical load sequence, configuring a short-time predictor by using the historical load sequence, taking the real-time power consumption data, the temperature data set and task scheduling data as input data, performing information gain ratio and prediction uncertainty analysis of each granularity under a preset candidate granularity set, and determining a calibration prediction granularity; after configuring a sliding window by using the calibration prediction granularity, performing matching evaluation of the single-phase immersion liquid cooling and the data center in the sliding window, and establishing a matching evaluation result; if the matching evaluation result meets a trigger threshold, performing adaptive optimization under local-global control coordination constraints of single-phase immersion liquid cooling control in the sliding window, and outputting a liquid cooling response scheme.

[0006] In a possible implementation, the liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center further performs the following processing: taking liquid cooling flow, circulating pump rotating speed, heat exchanger inlet and outlet temperature difference and cold liquid regeneration interval as multi-dimensional control variables, performing variable space analysis of the multi-dimensional control variables at a load change rate corresponding to the calibration prediction granularity, and setting a solution space; establishing an optimization objective function, wherein the optimization objective function comprises a thermal management target, an energy efficiency target and a cold liquid life target; after performing key local positioning based on joint identification of a temperature rise inflection point and a liquid cooling life attenuation threshold in the sliding window, performing segmented response optimization of a local layer-global layer in the solution space by using the key local positioning result, and performing synthetic decision evaluation by using the optimization objective function, adaptive optimization is completed according to the synthetic decision evaluation result.

[0007] In a possible implementation, the liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center further performs the following processing: establishing a local subarea by using the key local positioning result; performing adaptive optimization of local control under solution space constraints in the local subarea, and establishing a local solution set; performing local solution set coordination control optimization of the global layer, establishing a first set of real solution sets by using the coordination control optimization result; performing fitness evaluation of the first set of real solution sets according to the optimization objective function, and generating optimization feedback by using the fitness evaluation result; after feedback optimization of the local layer-global layer is performed by using the optimization feedback, performing optimization iteration, and completing adaptive optimization.

[0008] In a possible implementation, the liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center further performs the following processing: after the preset candidate granularity is selected, the input data is resampled and aggregated according to the selected preset candidate granularity; the information gain ratio of the feature to the target variable is calculated for the resampling and aggregation result; the multiple predictions based on the resampling and aggregation result are performed based on the short-time predictor, and the prediction uncertainty is generated according to the variance of the prediction result; after the information gain ratio and the prediction uncertainty corresponding to the preset candidate granularity in the preset candidate granularity set are obtained, the granularity screening is performed to determine the calibration prediction granularity.

[0009] In a possible implementation, the liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center further performs the following processing: the data acquisition of the cooling liquid is performed, the cold liquid health index is established according to the data acquisition result, and the data acquisition result includes the viscosity, the conductivity, the particulate matter concentration and the chemical stability data; it is judged whether the cold liquid health index meets the health threshold; if the cold liquid health index cannot meet the health threshold, the high flow rate limitation constraint is established; the adaptive optimization is completed by taking the high flow rate limitation constraint as the optimization constraint condition.

[0010] In a possible implementation, the liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center further performs the following processing: if the cold liquid health index meets the health threshold, the high-performance liquid cooling strategy bias constraint is established; the adaptive optimization is performed according to the high-performance liquid cooling strategy bias constraint.

[0011] In a possible implementation, the liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center further performs the following processing: the ladder compensation node is set according to the response period of the liquid cooling response scheme; the control state verification of the data center is performed at the ladder compensation node, and the verification deviation is established; the node adaptive control compensation management is performed based on the corresponding verification deviation at each ladder compensation node.

[0012] The application also provides a liquid cooling adaptive control system based on a single-phase immersion liquid cooling data center, which comprises: a temperature data monitoring module, which is used to start a temperature monitoring sensor to read chip temperature data and environmental temperature data simultaneously after reading real-time power consumption data of the data center, and to construct a temperature data set; a calibration prediction granularity determination module, which is used to configure a short-time predictor by using a historical load sequence after reading the historical load sequence, to take the real-time power consumption data, the temperature data set and task scheduling data as input data, to perform information gain ratio and prediction uncertainty analysis of each granularity under a preset candidate granularity set, and to determine a calibration prediction granularity; a matching evaluation module, which is used to perform matching evaluation of the single-phase immersion liquid cooling and the data center in a sliding window after configuring the sliding window by using the calibration prediction granularity, and to establish a matching evaluation result; and a liquid cooling response scheme output module, which is used to perform adaptive optimization under local-global control coordination constraints of single-phase immersion liquid cooling control in the sliding window if the matching evaluation result meets a trigger threshold, and to output a liquid cooling response scheme.

[0013] The liquid cooling adaptive control method and system based on a single-phase immersion liquid cooling data center provided by the application are used to read real-time power consumption data of a data center, start a temperature monitoring sensor to read chip temperature data and environmental temperature data simultaneously, configure a short-time predictor by using a historical load sequence, perform information gain ratio and prediction uncertainty analysis of a preset candidate granularity, perform matching evaluation of the single-phase immersion liquid cooling and the data center by using a calibration prediction granularity, and perform adaptive optimization under local-global control coordination constraints of single-phase immersion liquid cooling control in a sliding window if a trigger threshold is met, and output a liquid cooling response scheme. The technical problems of low matching degree of cooling capacity of single-phase immersion liquid cooling and load demand, response lag and insufficient energy consumption control in the prior art are solved, adaptive optimization control based on real-time and prediction data is achieved, and the technical effects of reducing energy consumption, improving heat dissipation efficiency and temperature control precision are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. The flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application in the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously according to needs. Meanwhile, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 The liquid cooling adaptive control method based on a single-phase immersion liquid cooling data center provided by the embodiments of the present application is shown in the flowchart.

[0016] Figure 2A liquid cooling adaptive control system structure schematic diagram based on a single-phase immersion liquid cooling data center is provided for an embodiment of the application.

[0017] Reference signs: temperature data monitoring module 10, calibration prediction granularity determination module 20, matching evaluation module 30, and liquid cooling response scheme output module 40. DETAILED DESCRIPTION

[0018] The above description is only a summary of the technical solutions of the application. In order to enable the technical means of the application to be more clearly understood, and to be implemented according to the content of the description, and in order to enable the above and other purposes, characteristics and advantages of the application to be more apparent and easy to understand, the following specific embodiments of the application are described.

[0019] In order to make the purposes, technical solutions and advantages of the application more clear, the following will further describe the application in combination with the drawings. The described embodiments should not be regarded as limiting the application. All other embodiments obtained by those skilled in the art without making creative efforts fall within the scope of protection of the application.

[0020] In the following description, “some embodiments” are related to a subset of all possible embodiments, but it can be understood that “some embodiments” can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict. The term “first\second” is only to distinguish similar objects, and does not represent a specific order of the objects. The terms “include” and “have” and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art to which the application belongs. The terms used herein are only for the purpose of describing the embodiments of the application.

[0021] The embodiments of the application provide a liquid cooling adaptive control method based on a single-phase immersion liquid cooling data center, as shown in the method. Figure 1 The method comprises the following steps:

[0022] In step S100, after reading the real-time power consumption data of the data center, the temperature monitoring sensor is started to read the chip temperature data and the environmental temperature data synchronously, and a temperature data set is constructed.

[0023] Preferably, the power consumption monitoring device such as the power consumption sensor on the smart meter, PDU monitoring unit or server motherboard is used to collect the power consumption data of the IT equipment such as servers and storage devices in the data center in real time, obtain real-time power consumption data, reflect the current computing load, and indirectly reflect the trend of heat generation; at the same time of collecting the power consumption data, the temperature monitoring sensor is started to monitor the temperature to ensure the time consistency of the monitoring data, wherein the temperature monitoring sensor includes temperature probes arranged on the server chips such as CPU and GPU built-in temperature sensors, and data center environment temperature monitoring sensors for monitoring the air temperature or cooling liquid temperature in the cabinet; then the chip temperature data and the environment temperature data are read respectively, wherein the chip temperature data reflect the actual working temperature of the core components such as processors and graphics cards, and are used to judge the heat dissipation effect and overheating risk; the environment temperature data include the cooling liquid temperature in the cabinet and the liquid cooling tank, or the air temperature in the computer room, and are used to evaluate the external cooling condition; finally, the collected chip temperature data and environment temperature data are integrated according to time to form a temperature data set for load prediction and liquid cooling control optimization.

[0024] In step S200, after reading the historical load sequence, the short-time predictor is configured using the historical load sequence, the real-time power consumption data, the temperature data set and the task scheduling data are taken as input data, the information gain ratio and the prediction uncertainty of each granularity in the preset candidate granularity set are analyzed, and the calibration prediction granularity is determined.

[0025] Preferably, the load change record in the past period of time is called from the operation monitoring data recorded by the data center, such as the monitoring log of server operation or the task scheduling record, which may include CPU usage, GPU occupancy, task execution quantity, historical power data or power consumption curve, etc., to generate a historical load sequence, and then the short-time predictor is configured using the historical load sequence, wherein the short-time predictor is a prediction model based on ARIMA, LSTM, moving average model, etc., which is used to predict the load change in the future short time, specifically, the historical load sequence is taken as input data to train and initialize the prediction model, and the short-time predictor is obtained, which can predict the load trend in the next short period of time according to the recent data.

[0026] Preferably, the real-time power consumption data, the temperature data set and the task scheduling data are taken as input data, wherein the task scheduling data includes a task allocation plan, a predicted resource occupation and the like, that is, according to historical load data, the real-time power consumption data, the temperature data set and the task scheduling data are combined to perform information gain ratio and prediction uncertainty analysis of each granularity under a preset candidate granularity set by using a short-time predictor. Specifically, the preset candidate granularity set includes a plurality of time resolution candidate values for prediction, such as data center load after 1 minute, 5 minutes and 10 minutes. The information gain ratio analysis is used to evaluate the contribution of input data to the prediction result under different prediction granularities. The more reasonable the prediction granularity is, the higher the utilization rate of input information is. The prediction uncertainty analysis is used to measure the stability and reliability of the prediction result under the granularity, for example, variance or confidence interval is used to evaluate the prediction fluctuation. After analyzing all candidate granularities, the granularity with high information gain ratio and low prediction uncertainty is selected as the best prediction time scale, that is, the calibration prediction granularity is determined, which directly affects the subsequent liquid cooling adjustment strategy, such as pre-cooling 2 minutes in advance or pre-cooling 10 minutes in advance.

[0027] Further, the step S200 further includes the following steps: in step S210, after the preset candidate granularity is selected, the input data is resampled and aggregated according to the selected preset candidate granularity; in step S220, the information gain ratio of the feature to the target variable is calculated for the resampling and aggregation result; in step S230, a plurality of predictions based on the resampling and aggregation result are performed based on the short-time predictor, and the prediction uncertainty is generated according to the variance of the prediction result; in step S240, after the information gain ratio and the prediction uncertainty corresponding to the preset candidate granularity in the preset candidate granularity set are obtained, the granularity screening is performed to determine the calibration prediction granularity.

[0028] Preferably, a preset candidate granularity is selected from a preset candidate granularity set, and a preset prediction time step, such as 1 minute, 5 minutes, 10 minutes, etc., is selected, and then real-time power consumption data, temperature data, task scheduling data, etc. are resampled and aggregated according to the selected preset candidate granularity, including re-segmenting according to the selected preset candidate granularity and aggregating within each segment, such as calculating the average, maximum, sum, etc. Calculate the feature of the resampled and aggregated result to determine the information gain ratio of power consumption, chip temperature, environmental temperature, task scheduling quantity, etc. to the target variable, wherein the target variable refers to the object of prediction, which may include power consumption, load level, temperature change in the future period of time, and the information gain ratio is used to measure the prediction value of the feature to the target variable at this granularity. The greater the information gain ratio, the more useful the feature, i.e. to judge the contribution of the input feature to the prediction task at this time granularity; then input the resampled and aggregated result into the short-term predictor for multiple predictions to obtain multiple prediction results, and calculate the variance of the multiple prediction results to represent the prediction uncertainty, wherein the smaller the variance, the more stable the prediction, indicating high reliability; otherwise, the greater the variance, the greater the prediction fluctuation, indicating low reliability; finally, information gain ratio calculation and prediction uncertainty calculation are performed on all preset candidate granularities in the preset candidate granularity set, wherein the information gain ratio is high, the feature utilization rate is high, the prediction uncertainty is low, and the prediction is stable and reliable, and granularity screening is performed, i.e. comparing the information gain ratio and the prediction uncertainty of each preset candidate granularity, selecting the preset candidate granularity with high information gain ratio and low prediction uncertainty as the calibration prediction granularity, and using it as the prediction time step for liquid cooling control strategy optimization.

[0029] Preferably, the resampled and aggregated result is discretized or interval divided with the target variable, i.e. equidistant division, equal frequency division or entropy optimal division, to divide it into a plurality of intervals, then calculate the information entropy of the target variable in the resampled and aggregated result, and then calculate the information entropy of the target variable in each interval after the feature is divided and weighted sum to obtain the conditional entropy, then calculate the information gain, i.e. the degree of reduction of uncertainty of the target variable after using the feature to divide, and finally normalize by introducing the information entropy of the feature itself, i.e. calculate the ratio of information gain to the information entropy of the feature as the information gain ratio, and then obtain the information gain ratio corresponding to the plurality of resampled and aggregated results.

[0030] In step S300, a sliding window is configured using the calibration prediction granularity, and a single-phase immersion liquid cooling and data center matching evaluation is performed in the sliding window to establish a matching evaluation result.

[0031] Preferably, the sliding window is configured by using the calibration prediction granularity, including setting the window length as the calibration prediction granularity multiplied by a multiple, such as 3-5 times, and setting the step length of the window each time as the calibration prediction granularity, so as to continuously analyze the data in the recent period of time, and then perform the matching evaluation of the single-phase immersion liquid cooling and the data center in the sliding window, that is, to judge whether the actual cooling capacity of the liquid cooling system matches the current heat dissipation demand of the data center, specifically, to monitor and obtain the inlet temperature, outlet temperature, flow rate, pump power, heat exchange efficiency and the like of the cooling liquid of the liquid cooling system, and then perform the matching evaluation with the server power consumption, chip temperature, environment temperature, load prediction value and the like of the data center, including energy consumption comparison, temperature control effect verification, efficiency index and prediction matching, wherein the energy consumption comparison is to evaluate whether the cooling capacity provided by the liquid cooling system is greater than or equal to the heat generated by the load, the temperature control effect verification is to verify whether the chip temperature is maintained within a safe range, the efficiency index is the cooling capacity per unit energy consumption, and the prediction matching is to evaluate whether the future predicted load trend will exceed the liquid cooling capacity, and finally to comprehensively output the matching evaluation result, that is, the matching degree of the cooling capacity of the liquid cooling system and the heat dissipation demand of the data center, for adjusting the liquid cooling operation strategy.

[0032] In step S400, if the matching evaluation result meets the trigger threshold, adaptive optimization under the local-global control coordination constraint of the single-phase immersion liquid cooling control is performed in the sliding window, and a liquid cooling response scheme is output.

[0033] Preferably, the trigger threshold is a critical condition set in advance, which is used to judge whether the liquid cooling strategy needs to be adjusted, such as the matching degree being less than 0.8, the chip temperature being close to the upper limit, or the future predicted load exceeding the liquid cooling capacity by 10%, if the matching evaluation result meets the trigger threshold, it means that the matching condition is deteriorated to a certain extent, and optimization is triggered, that is, adaptive optimization under the local-global control coordination constraint of the single-phase immersion liquid cooling control is performed in the sliding window, so as to avoid energy waste caused by frequent adjustment, specifically, the local control is the adjustment of the liquid cooling parameters of a specific device or cabinet, such as adjusting the speed of a pump, the opening degree of a valve, the flow rate of cooling liquid of a single server and the like, to quickly respond to the heat spot; the global control is the optimization of the operation strategy of the whole data center liquid cooling system, such as the overall flow distribution, the working mode of the heat exchanger, the total temperature setting of the cooling liquid and the like, to ensure the optimal energy efficiency and resource allocation; the local-global control coordination constraint of the single-phase immersion liquid cooling control means that the local adjustment cannot destroy the global optimization target, and the global strategy also needs to consider the local hotspot demand, including dynamically adjusting the liquid cooling control parameters according to the real-time and predicted data, and then determining the optimal cooling scheme under the current conditions as the liquid cooling response scheme, which may include multiple liquid cooling adjustment instructions, such as the speed setting value of each pump, the cooling liquid flow distribution, the cooling liquid inlet temperature target value, the heat exchanger start-stop strategy, and the local cooling strategy of the cabinet or server level and the like, so as to ensure to reduce the energy consumption, improve the heat dissipation efficiency and temperature control precision.

[0034] Further, step S400 further comprises step S410, taking the liquid cooling flow rate, circulating pump speed, heat exchanger inlet and outlet temperature difference, and cold liquid regeneration interval as multi-dimensional control variables, performing variable space analysis of the multi-dimensional control variables at the load change rate corresponding to the calibration prediction granularity, setting the solution space; step S420, establishing an optimization objective function, the optimization objective function including a thermal management objective, an energy efficiency objective, and a cold liquid life objective; step S430, after key local positioning based on joint identification of the temperature rise inflection point and the liquid cooling life attenuation threshold in the sliding window, using the key local positioning result to perform segmented response optimization in the local layer and the global layer in the solution space, and performing synthetic decision evaluation through the optimization objective function, and completing adaptive optimization according to the synthetic decision evaluation result.

[0035] Preferably, the liquid cooling flow rate, circulating pump speed, heat exchanger inlet and outlet temperature difference, and cold liquid regeneration interval are taken as multi-dimensional control variables, wherein the liquid cooling flow rate refers to the circulating amount of cooling liquid per unit time, directly affecting the heat carried away; the circulating pump speed affects the flow rate and system energy consumption, and adjusts the response speed; the heat exchanger inlet and outlet temperature difference reflects the heat exchange efficiency, which can be indirectly controlled by adjusting the liquid temperature difference; the cold liquid regeneration interval is the time period for replacing or regenerating the cooling liquid, affecting the cooling liquid life and the stability of the cooling effect. Then, variable space analysis of the multi-dimensional control variables is performed at the load change rate corresponding to the calibration prediction granularity, that is, according to the current and predicted load change rate, the value range of the multi-dimensional control variables that meet the heat dissipation demand, equipment safety, energy efficiency requirements, etc. is analyzed, and all feasible value ranges of the control variables are combined to obtain the solution space, including all possible control parameter combinations.

[0036] Preferably, the thermal management objective, the energy efficiency objective, and the cold liquid life objective are taken as optimization objectives, and a weighted sum is performed to establish an optimization objective function, wherein the thermal management objective requires that the chip temperature be kept within a safe range and temperature fluctuations be reduced; the energy efficiency objective is to minimize the total energy consumption of the liquid cooling system, including pump power consumption, heat exchanger power consumption, etc.; the cold liquid life objective refers to extending the effective use time of the cooling liquid to avoid performance degradation of the cooling liquid due to high temperature and rapid chemical reaction; then, key local positioning based on joint identification of the temperature rise inflection point and the liquid cooling life attenuation threshold is performed in the sliding window, wherein the temperature rise inflection point refers to a point on the temperature curve where the change rate changes significantly, such as the starting point of rapid temperature rise; the liquid cooling life attenuation threshold refers to a sign that the cooling liquid performance has dropped to a certain critical value, such as a 10% reduction in thermal conductivity; both the temperature rise inflection point and the cooling liquid life critical value are detected, and if both are close to the critical state, they are marked as key local, and the key area corresponding to the key time is determined in the sliding window to complete key local positioning, and the key local positioning result is obtained.

[0037] Preferably, the key local positioning result is used to perform local layer-global layer segmentation response optimization in the solution space, i.e., fast local response is ensured first, and then global adjustment is performed to make the overall liquid cooling system enter a high-efficiency balanced state. For example, for a key local part such as a server with rapidly rising temperature, the parameters are quickly adjusted in a small range, such as instantaneously increasing the flow rate; then the global energy efficiency distribution of the entire liquid cooling system is considered, such as overall reducing the liquid temperature and optimizing the pump group speed; then the synthesized decision evaluation is performed through the optimization objective function, i.e., the results after local and global adjustment are substituted into the optimization objective function for comprehensive scoring, and if the new scheme has a higher comprehensive score, the new scheme is adopted and executed, thereby determining the synthesized decision evaluation result, and finally completing adaptive optimization.

[0038] Further, step S430 further includes step S431 of establishing a local partition by using the key local positioning result; step S432 of performing adaptive optimization of local control under solution space constraint in the local partition to establish a local solution set; step S433 of performing local solution set coordination control optimization of the global layer to establish a first set of real solutions; step S434 of performing fitness evaluation of the first set of real solutions according to the optimization objective function, and generating optimization feedback by using the fitness evaluation result; and step S435 of performing optimization iteration after feedback optimization of the local layer-global layer through the optimization feedback, thereby completing adaptive optimization.

[0039] Preferably, the key local positioning result is used to establish a local partition, i.e., to divide a physical region related to the key local positioning result in the liquid cooling system, such as a certain cabinet, a certain server cluster, or a certain liquid cooling loop, and then adaptive optimization of local control under solution space constraint is performed in the local partition. Specifically, in the multi-dimensional control variable solution space, a subset that meets the local region operation condition is taken out, for example, the flow rate range allowed by the cabinet, the pump speed limit, and the like, and adaptive adjustment of liquid cooling parameters is performed according to real-time and predicted data of the partition, such as local flow rate distribution, local pump speed, and local heat exchanger working state. Local gradient search and small-scale particle swarm are used for optimization to obtain local optimal control parameters to form a local solution set. Then, local solution set coordination control optimization of the global layer is performed, i.e., all local solution sets of the partitions are taken as inputs, and operation constraints of the entire liquid cooling system are considered, such as total flow rate, total energy consumption, and liquid temperature distribution balance, to perform re-optimization on the global layer, so that the optimization results of different partitions do not conflict with each other, and the overall energy efficiency is optimal, thereby obtaining a control parameter combination after global coordination as a first set of real solutions.

[0040] Preferably, the first set of real solution set fitness evaluation is performed according to the optimization target function, that is, each set of control parameter combination in the first set of real solution set is substituted into the optimization target function to calculate the comprehensive fitness score, and a high score indicates that the comprehensive performance of the control parameter combination is good. Then, the optimization feedback is generated using the fitness evaluation result, including outputting a feedback signal according to the fitness evaluation result to indicate that the control parameters need to be increased, decreased, secondarily optimized, or re-distributed in terms of flow rate / pump speed, etc. Finally, the local layer-global layer feedback optimization is performed through the optimization feedback, including adjusting the local solution set according to the feedback, such as reducing the flow rate upper limit in the hot spot area, increasing the pump speed, etc. The global layer re-coordinates the resource allocation of each sub-area according to the feedback, such as adjusting more flow rate from a low-load sub-area to a high-load sub-area. The optimization iteration is performed until the convergence condition is met, such as the fitness improvement being less than a certain threshold or the number of iterations reaching an upper limit, and finally the adaptive optimization is completed to obtain the optimal liquid cooling control scheme, which can cope with the current load and temperature control demand and also maintain efficient and stable operation under the predicted future load.

[0041] Further, step S400 further includes step S440 of performing data acquisition of the cooling liquid, step S450 of judging whether the cooling liquid health index meets the health threshold, step S460 of establishing a high flow rate limitation constraint if the cooling liquid health index fails to meet the health threshold, and step S470 of taking the high flow rate limitation constraint as an optimization constraint condition to complete the adaptive optimization.

[0042] Preferably, a viscosity meter and a conductivity sensor are installed in the liquid cooling circulation pipeline to respectively acquire the viscosity and conductivity of the cooling liquid. A laser particle counter is used to monitor the number of particles in different particle size ranges in the liquid in real time through laser scattering to obtain the particulate matter concentration of the cooling liquid. A pH meter or a redox potential sensor is used to acquire the pH value or redox potential of the cooling liquid to evaluate the chemical stability data of the cooling liquid. The viscosity of the cooling liquid reflects the flow resistance, and an excessively high viscosity will increase the pump load, while an excessively low viscosity may cause a decrease in heat transfer efficiency. The conductivity reflects the ion concentration of the cooling liquid, and an excessively high conductivity will increase the risk of corrosion or electric leakage. The particulate matter concentration refers to the solid impurity content, and an excessively high particulate matter concentration will block the pipeline or damage the pump. The chemical stability data includes pH value change, redox potential, etc., and reflects whether the cooling liquid has undergone chemical deterioration. The cooling liquid health index is established according to the data acquisition results, that is, the viscosity, conductivity, particulate matter concentration, and chemical stability data are normalized and weighted to calculate a comprehensive index to determine the cooling liquid health index, wherein the weights of the indexes are set according to the importance.

[0043] Preferably, the health threshold is set to 0.8 according to actual needs and historical data, and it is judged whether the coolant health index meets the health threshold. If the coolant health index cannot meet the health threshold, a high flow rate limit constraint is established, that is, the maximum value of the circulating flow rate of the cooling liquid is determined, wherein the high flow rate may accelerate particle wear, chemical reaction, oxidation and cooling liquid deterioration, and affect the service life of the cooling liquid. Then the high flow rate limit constraint is added as an optimization constraint condition to the solution space, and adaptive optimization is performed to ensure that the control parameter combination cannot exceed the high flow rate limit constraint, so as to ensure that the optimized liquid cooling operation scheme can meet the temperature control demand and will not further damage the cooling liquid.

[0044] Further, step S450 further includes step S451 of establishing a high-performance liquid cooling strategy bias constraint if the coolant health index meets the health threshold; and step S452 of performing adaptive optimization according to the high-performance liquid cooling strategy bias constraint.

[0045] Preferably, if the coolant health index meets the health threshold, it indicates that the physical and chemical states of the cooling liquid are within the normal range, and there is no significant degradation, pollution or corrosion risk, so it is not necessary to be overly conservative in energy saving or prolong the service life of the cooling liquid, and then a high-performance liquid cooling strategy bias constraint is established, and the optimization target is more inclined to improve the heat dissipation capacity, reduce the chip temperature and support high-load operation. Specifically, the upper limit of the flow rate is increased, the circulating pump is allowed to operate at a higher speed, and the flow rate can be close to the maximum rated value of the system; heat exchange efficiency is prioritized, cooling capacity of the heat exchanger is preferentially allocated, and the temperature difference between the inlet and outlet of the heat exchanger is reduced; response speed is accelerated, and the cooling intensity is increased more quickly in response to load changes; energy efficiency weight is reduced, and the weight of the thermal management target is increased while the weight of the energy efficiency target is reduced when the optimization algorithm weighs thermal management, energy efficiency and cooling liquid life; local supercooling is allowed, and local temperature is allowed to drop below the safe temperature margin to support higher computing performance. Finally, adaptive optimization is performed according to the high-performance liquid cooling strategy bias constraint, that is, when performing local-global coordinated optimization search in the solution space, the high-power area of the solution space is relaxed according to the high-performance liquid cooling strategy bias constraint, and the weight of the optimization objective function is reset, and the optimization result is inclined to maximize the computing performance on the basis of meeting the temperature control demand, that is, higher power consumption and heat are allowed, and finally the liquid cooling response scheme with high heat dissipation, high flow rate and high heat exchange performance is output.

[0046] Further, step S400 further includes step S480 of setting a ladder compensation node according to a response period of the liquid cooling response scheme; step S490 of performing control state verification of the data center at the ladder compensation node and establishing a verification deviation; and step S4100 of performing node adaptive control compensation management based on the corresponding verification deviation at each ladder compensation node.

[0047] Preferably, the response period of the liquid cooling response scheme, i.e. the expected adjustment period, such as 5 minutes from the current flow to the target flow, is set according to the response period, i.e. the entire response process is divided into several state stages, such as 0%→33%→66%→100% phased flow increase, and the end point of each stage is the step compensation node, which avoids temperature fluctuations, pump impact or excessive energy consumption caused by one-time large-scale adjustment; then the control state verification of the data center is carried out at the step compensation node, specifically, after each step compensation node reaches the predetermined adjustment value, the actual running state is detected and verified, including verifying whether the chip temperature reaches the expectation, whether the environment temperature remains stable, whether the liquid flow and pressure meet the control value, whether the power consumption and energy efficiency indicators meet the standards, then calculating the difference between the actual state value and the predicted value in the liquid cooling response scheme to obtain the verification deviation; finally, based on the corresponding verification deviation, the node adaptive control compensation management is carried out at each step compensation node, if the verification deviation occurs at a certain step compensation node, compensation management is immediately carried out at this stage, including referring to the direction and size of the current deviation, calculating the required correction amplitude, for example, if the temperature is higher than expected, increase the flow or pump speed, reduce the liquid inlet temperature; if the temperature is lower than expected, reduce the flow to avoid excessive cooling and waste energy consumption; if the flow is insufficient, check the pump control signal and valve opening and correct; if the energy efficiency is too low, optimize the balance between pump speed and heat exchange efficiency, so as to ensure that the temperature, flow, power consumption and other indicators more accurately meet the optimization target.

[0048] In the foregoing, with reference to Figure 1 The liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center according to the embodiments of the present application is described in detail. Next, with reference to Figure 2 The liquid cooling adaptive control system based on the single-phase immersion liquid cooling data center according to the embodiments of the present application is described.

[0049] The liquid cooling adaptive control system based on the single-phase immersion liquid cooling data center according to the embodiments of the present application is used to solve the technical problems of low matching degree of cooling capacity and load demand, response lag and insufficient energy consumption control of the single-phase immersion liquid cooling in the prior art, and achieves adaptive optimization control based on real-time and predicted data, reduces energy consumption, improves heat dissipation efficiency and temperature control precision. As Figure 2 shown, the liquid cooling adaptive control system based on the single-phase immersion liquid cooling data center includes a temperature data monitoring module 10, a calibration and prediction granularity determination module 20, a matching evaluation module 30 and a liquid cooling response scheme output module 40.

[0050] The temperature data monitoring module 10 is configured to start the temperature monitoring sensor to read the chip temperature data and the environmental temperature data synchronously after reading the real-time power consumption data of the data center, and to construct a temperature data set; the calibration prediction granularity determination module 20 is configured to read a historical load sequence, configure a short-time predictor by using the historical load sequence, take the real-time power consumption data, the temperature data set and task scheduling data as input data, perform information gain ratio and prediction uncertainty analysis of each granularity in a preset candidate granularity set, and determine a calibration prediction granularity; the matching evaluation module 30 is configured to configure a sliding window by using the calibration prediction granularity, perform matching evaluation of the single-phase immersion liquid cooling and the data center in the sliding window, and establish a matching evaluation result; and the liquid cooling response scheme output module 40 is configured to perform adaptive optimization of the single-phase immersion liquid cooling control under the local-global control coordination constraint in the sliding window if the matching evaluation result meets a trigger threshold, and output a liquid cooling response scheme.

[0051] In the following, the specific configuration of the liquid cooling response scheme output module 40 will be described in detail. The liquid cooling response scheme output module 40 further comprises: taking the liquid cooling flow, the circulating pump rotating speed, the heat exchanger inlet and outlet temperature difference and the cold liquid regeneration interval as multi-dimensional control variables, performing variable space analysis of the multi-dimensional control variables on the load change rate corresponding to the calibration prediction granularity, and setting a solution space; establishing an optimization objective function, the optimization objective function comprising a thermal management target, an energy efficiency target and a cold liquid life target; after performing key local positioning based on joint identification of the temperature rise inflection point and the liquid cooling life attenuation threshold in the sliding window, performing segmented response optimization of the local layer-global layer in the solution space by using the key local positioning result, and performing synthetic decision evaluation by the optimization objective function, completing adaptive optimization according to the synthetic decision evaluation result.

[0052] In the following, the specific configuration of the liquid cooling response scheme output module 40 will be described in detail. The liquid cooling response scheme output module 40 further comprises: establishing a local partition by using the key local positioning result; performing adaptive optimization of the local control under the solution space constraint in the local partition, establishing a local solution set; performing local solution set coordination control optimization of the global layer, establishing a first set of real solution set by the coordination control optimization result; performing fitness evaluation of the first set of real solution set according to the optimization objective function, and generating optimization feedback by using the fitness evaluation result; after feedback optimization of the local layer-global layer by the optimization feedback, performing optimization iteration, and completing adaptive optimization.

[0053] Below, the specific configuration of the calibration prediction granularity determination module 20 will be described in detail. The calibration prediction granularity determination module 20 further comprises: after selecting the preset candidate granularity, resampling and aggregating the input data according to the selected preset candidate granularity; calculating the information gain ratio of the feature to the target variable on the resampling and aggregation result; performing multiple predictions based on the resampling and aggregation result based on the short-time predictor, and generating a prediction uncertainty according to the variance of the prediction result; after obtaining the information gain ratio and the prediction uncertainty corresponding to the preset candidate granularity in the preset candidate granularity set, performing granularity screening to determine the calibration prediction granularity.

[0054] Below, the specific configuration of the liquid cooling response scheme output module 40 will be described in detail. The liquid cooling response scheme output module 40 further comprises: performing data acquisition of the cooling liquid, establishing a cooling liquid health index according to the data acquisition result, and the data acquisition result includes viscosity, conductivity, particulate matter concentration, and chemical stability data; judging whether the cooling liquid health index meets the health threshold; if the cooling liquid health index cannot meet the health threshold, establishing a high flow rate limit constraint; completing adaptive optimization by taking the high flow rate limit constraint as an optimization constraint condition.

[0055] Below, the specific configuration of the liquid cooling response scheme output module 40 will be described in detail. The liquid cooling response scheme output module 40 further comprises: if the cooling liquid health index meets the health threshold, establishing a high-performance liquid cooling strategy bias constraint; and performing adaptive optimization according to the high-performance liquid cooling strategy bias constraint.

[0056] Below, the specific configuration of the liquid cooling response scheme output module 40 will be described in detail. The liquid cooling response scheme output module 40 further comprises: setting a ladder compensation node according to the response period of the liquid cooling response scheme; performing control state verification of the data center at the ladder compensation node, and establishing a verification deviation; and performing node adaptive control compensation management based on the corresponding verification deviation at each ladder compensation node.

[0057] The liquid cooling adaptive control system based on the single-phase immersion liquid cooling data center provided in the embodiments of the present application can perform the liquid cooling adaptive control method based on the single-phase immersion liquid cooling data center provided in any of the embodiments of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0058] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server, and each unit and module included is only divided according to the functional logic, but is not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of each functional unit are only for the convenience of mutual differentiation, and do not limit the protection scope of the present application.

[0059] The above detailed description does not limit the scope of the application. Various modifications, combinations and equivalents thereof will become apparent to those skilled in the art in view of the foregoing description. Any modifications, equivalents and alternatives falling within the spirit and principles of the application should be included in the scope of the application.

Claims

1. A liquid cooling adaptive control method based on single-phase immersion liquid cooling data center, characterized in that, The method comprises: After reading real-time power consumption data of the data center, synchronously starting temperature monitoring sensor to read chip temperature data and environment temperature data, and constructing temperature data set; After reading historical load sequence, configuring short-time predictor by using the historical load sequence, taking the real-time power consumption data, the temperature data set and task scheduling data as input data, performing information gain ratio and prediction uncertainty analysis of each granularity under a preset candidate granularity set, and determining calibration prediction granularity; After configuring sliding window by using the calibration prediction granularity, performing matching evaluation of single-phase immersion liquid cooling and the data center in the sliding window, and establishing matching evaluation result; If the matching evaluation result meets trigger threshold, performing adaptive optimization under local-global control coordination constraint of single-phase immersion liquid cooling control in the sliding window, and outputting liquid cooling response scheme.

2. The liquid cooling adaptive control method for single-phase immersion liquid cooling data center of claim 1, wherein, The adaptive optimization under local-global control coordination constraint of single-phase immersion liquid cooling control in the sliding window comprises: Taking liquid cooling flow, circulating pump rotating speed, heat exchanger inlet and outlet temperature difference and cold liquid regeneration interval as multi-dimensional control variables, performing variable space analysis of the multi-dimensional control variables under load change rate corresponding to the calibration prediction granularity, and setting solution space; Establishing optimization objective function, wherein the optimization objective function comprises thermal management target, energy efficiency target and cold liquid life target; After performing key local positioning based on joint identification of temperature rise inflection point and liquid cooling life attenuation threshold in the sliding window, performing segmented response optimization of local layer-global layer in the solution space by using the key local positioning result, performing synthetic decision evaluation by using the optimization objective function, and completing adaptive optimization according to the synthetic decision evaluation result.

3. The liquid cooling adaptive control method for single-phase immersion liquid cooling data center of claim 2, wherein, The adaptive optimization under local-global control coordination constraint of single-phase immersion liquid cooling control in the sliding window comprises: Establishing local subarea by using the key local positioning result; Performing adaptive optimization of local control under solution space constraint in the local subarea, and establishing local solution set; Performing local solution set coordination control optimization of the global layer, establishing first group of real solution set by using the coordination control optimization result; Performing fitness evaluation of the first group of real solution set according to the optimization objective function, and generating optimization feedback by using the fitness evaluation result; After feedback optimization of the local layer-global layer by using the optimization feedback, performing optimization iteration, and completing adaptive optimization.

4. The liquid cooling adaptive control method for single-phase immersion liquid cooling data center of claim 1, wherein, The information gain ratio and prediction uncertainty analysis of each granularity under the preset candidate granularity set comprises: After selecting preset candidate granularity, performing resampling and aggregation of input data according to the selected preset candidate granularity; Calculating information gain ratio of the feature to the target variable for the resampling and aggregation result; Performing multiple predictions based on the resampling and aggregation result based on the short-time predictor, and generating prediction uncertainty according to variance of the prediction result; After obtaining information gain ratio and prediction uncertainty corresponding to the preset candidate granularity in the preset candidate granularity set, performing granularity screening, and determining calibration prediction granularity.

5. The single-phase liquid immersion data center based liquid cooling adaptive control method of claim 1, wherein, The adaptive optimization under the local-global control coordination constraint of the single-phase immersion liquid cooling control performed in the sliding window comprises: Data acquisition of the cooling liquid is performed, and a cooling liquid health index is established according to the data acquisition result, wherein the data acquisition result includes viscosity, conductivity, particulate matter concentration, and chemical stability data; It is judged whether the cooling liquid health index meets a health threshold value; If the cooling liquid health index cannot meet the health threshold value, a high flow rate limiting constraint is established; The high flow rate limiting constraint is used as an optimization constraint condition, and adaptive optimization is completed.

6. The single-phase liquid immersion data center based liquid cooling adaptive control method of claim 5, wherein, The judgment of whether the cooling liquid health index meets the health threshold value comprises: If the cooling liquid health index meets the health threshold value, a high-performance liquid cooling strategy biasing constraint is established; Adaptive optimization is performed according to the high-performance liquid cooling strategy biasing constraint.

7. The single-phase liquid immersion datacenter-based liquid cooling adaptive control method of claim 1, wherein, After the liquid cooling response scheme is output, the following steps are included: A ladder compensation node is set according to a response period of the liquid cooling response scheme; Control state verification of the data center is performed at the ladder compensation node, and a verification deviation is established; Node adaptive control compensation management is performed based on the corresponding verification deviation at each ladder compensation node.

8. A liquid cooling adaptive control system for a single-phase immersion liquid-cooled data center, comprising: The system is used to implement the liquid cooling adaptive control method of the single-phase immersion liquid cooling data center according to any one of claims 1 to 7, and the system comprises: A temperature data monitoring module is configured to start a temperature monitoring sensor to read chip temperature data and environmental temperature data simultaneously after reading real-time power consumption data of a data center, and to construct a temperature data set; A calibration prediction granularity determination module is configured to read a historical load sequence, configure a short-time predictor using the historical load sequence, use real-time power consumption data, the temperature data set, and task scheduling data as input data, perform information gain ratio and prediction uncertainty analysis of each granularity in a preset candidate granularity set, and determine a calibration prediction granularity; A matching evaluation module is configured to configure a sliding window using the calibration prediction granularity, perform matching evaluation of single-phase immersion liquid cooling and a data center in the sliding window, and establish a matching evaluation result; A liquid cooling response scheme output module is configured to perform adaptive optimization of single-phase immersion liquid cooling control under a local-global control coordination constraint in the sliding window if the matching evaluation result meets a trigger threshold value, and output a liquid cooling response scheme.

Images