Liquid-cooled server flow dynamic adjustment system based on genetic algorithm

Abstract

The application discloses a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm, and relates to the technical field of liquid-cooled flow data processing. The liquid-cooled server flow dynamic adjustment system based on the genetic algorithm comprises the following modules: a liquid-cooled flow-heat energy data acquisition and preprocessing module, which is used for acquiring and preprocessing liquid-cooled flow-heat energy operation data; a controllable baffle opening optimization module based on the genetic algorithm, which is used for generating a controllable baffle opening combination scheme, evaluating the flow-heat energy efficiency of the scheme, executing genetic algorithm optimization, and outputting a final controllable baffle opening combination scheme; a controllable baffle execution control and opening correction processing module, which is used for adjusting the position of the controllable baffle, executing controllable baffle opening correction calculation, and adjusting the actual opening; and a running balance evaluation and adaptive optimization module, which is used for evaluating the liquid-cooled flow heat balance state and adjusting subsequent genetic algorithm parameters. The application solves the problem that the prior art cannot adjust the cooling liquid flow direction according to real-time load, thereby causing the problem of insufficient cooling efficiency of the high-load chip area.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of liquid cooling flow data processing, in particular to a liquid cooling server flow dynamic adjustment system based on a genetic algorithm. BACKGROUND

[0002] In large-scale data centers and high-performance computing scenarios, with the continuous improvement of processor power density, liquid cooling technology gradually replaces traditional air cooling methods and becomes the mainstream heat dissipation method. However, most existing liquid cooling systems generally use fixed controllable baffle flow guide structures or simple flow adjustment methods, lack dynamic adaptation capabilities to different chip real-time load differences, and are prone to cause problems such as insufficient cooling of high-load chip areas and inability to timely suppress local temperature hotspots. Therefore, how to realize intelligent allocation and rapid optimization of multi-path flow in the liquid cooling system has become a key technical direction to improve overall heat dissipation performance and operation stability.

[0003] For example, the application with publication number CN115666088B provides an electronic device and a liquid cooling heat dissipation flow control method thereof. The electronic device includes electronic devices, a cold plate, a flow regulation unit, a running parameter acquisition unit, and a control unit. The cold plate is arranged adjacent to the electronic devices, and is used to circulate cooling liquid to dissipate heat for the electronic devices. The cold plate includes an inlet and an outlet. The flow regulation unit is arranged at the inlet, and is used to regulate the flow of the cooling liquid flowing into the cold plate. The running parameter acquisition unit is used to acquire the running parameters of the electronic devices. The control unit is used to obtain the running parameters of the electronic devices acquired by the running parameter acquisition unit, and to control the flow regulation unit to regulate the flow of the cooling liquid flowing into the cold plate according to at least the acquired running parameters.

[0004] For example, the application with publication number CN118444758A discloses a liquid cooling control method and a liquid cooling computing device. The method includes: using a baseboard management controller (BMC) of a first monitoring object in a liquid cooling computing device to obtain a plurality of first running parameters of each monitoring object at a first time; wherein the first running parameters can represent the temperature of the monitoring object, the monitoring object includes at least one server chassis in the liquid cooling computing device and at least one server located on each server chassis, and the first monitoring object is one of all monitoring objects; through the BMC of the first monitoring object, based on the first running parameters of each monitoring object, the cooling unit in the liquid cooling computing device is controlled to adjust the flow size and temperature of the cooling liquid.

[0005] Although the above technical solutions can achieve flow regulation of the liquid cooling system to some extent, there are still problems of single optimization target and long parameter adjustment response period, mainly manifested as single-path flow control based on only a small number of temperature or flow parameters, lack of global optimization capability for multi-path cooling liquid distribution, and inability to quickly generate and execute an optimal flow distribution scheme when there is a significant load difference between different chips, thereby leading to insufficient cooling in high-load areas, low utilization of cooling liquid in low-load areas, and difficulty in coping with real-time cooling demand caused by multi-node load difference.

[0006] Therefore, in view of the above problems, there is an urgent need for a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm. SUMMARY

[0007] Technical problems to be solved

[0008] In view of the deficiencies of the prior art, the present application provides a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm, which solves the problem that the prior art cannot adjust the cooling liquid flow direction according to the real-time load, resulting in insufficient cooling efficiency in high-load chip areas,

[0009] Technical solutions

[0010] To achieve the above object, the present application is implemented by the following technical solutions: a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm, comprising: a liquid cooling flow-thermal energy data acquisition and preprocessing module, a controllable baffle opening optimization module based on a genetic algorithm, a controllable baffle execution control and opening correction processing module, and a running balance evaluation and adaptive optimization module, wherein: the liquid cooling flow-thermal energy data acquisition and preprocessing module is used to acquire liquid cooling flow-thermal energy operation data, and perform time reference unification, noise suppression, physical consistency verification, data structure unification and scale standardization processing on the liquid cooling flow-thermal energy operation data to obtain preprocessed liquid cooling flow-thermal energy operation data; the controllable baffle opening optimization module based on a genetic algorithm is used to generate a controllable baffle opening combination scheme based on the preprocessed liquid cooling flow-thermal energy operation data, and evaluate the flow-thermal energy efficiency of the scheme, and execute genetic algorithm optimization and output the final controllable baffle opening combination scheme according to the evaluation result; the controllable baffle execution control and opening correction processing module is used to drive the execution mechanism to adjust the controllable baffle position based on the final controllable baffle opening combination scheme, and execute controllable baffle opening correction calculation combined with real-time liquid cooling flow-thermal energy operation data to adjust the actual opening of the execution mechanism; the running balance evaluation and adaptive optimization module is used to construct a running monitoring data set after the controllable baffle is adjusted, and quantitatively evaluate the liquid cooling flow distribution stability and chip temperature balance, and adjust the subsequent genetic algorithm parameters and controllable baffle opening combination scheme.

[0011] Further, the specific steps of collecting liquid cooling flow heat energy operation data, performing time reference unification, noise suppression, physical consistency verification, data structure unification and scale standardization processing on the liquid cooling flow heat energy operation data are as follows: collecting liquid cooling flow heat energy operation data, the liquid cooling flow heat energy operation data including cooling liquid density, cooling liquid specific heat capacity, chip temperature, cooling liquid inlet temperature, cooling liquid outlet temperature, cooling liquid flow rate, controllable baffle front and back pressure difference, controllable baffle opening, pump power, target cooling liquid flow rate and target chip temperature; performing time reference unification and sampling interval matching processing on the liquid cooling flow heat energy operation data through a global clock synchronization and sampling phase calibration method; performing noise suppression and disturbance compensation processing on the liquid cooling flow heat energy operation data through a multi-parameter filtering method combining frequency domain analysis and adaptive weight; establishing a physical consistency verification mechanism based on the thermodynamic conservation equation and the fluid continuity equation to identify and eliminate abnormal data of the liquid cooling flow heat energy operation data; performing data structure unification and interface format conversion processing on the liquid cooling flow heat energy operation data through a standardized feature mapping and multi-protocol packaging method; and performing unit conversion and feature scale standardization processing on the liquid cooling flow heat energy operation data through a dimension mapping and interval adaptive normalization method.

[0012] Further, based on the preprocessed liquid cooling flow heat energy operation data, the specific steps of generating a controllable baffle opening combination scheme are as follows: generating a distribution scheme set containing multiple controllable baffle opening combination schemes according to the opening range of each controllable baffle, each distribution scheme corresponding to a liquid cooling flow distribution mode; expressing each distribution scheme as a set of ordered parameter sequences according to the coding rules of the genetic algorithm, and recording the ordered parameter sequences as chromosomes for genetic operation.

[0013] Further, the specific steps of evaluating the flow heat energy efficiency of the scheme are as follows: multiplying the cooling liquid density, cooling liquid flow rate and cooling liquid specific heat capacity, then multiplying the difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature to obtain a heat exchange power term; taking the absolute value of the difference between the actual pressure difference before and after the controllable baffle and the design pressure difference before and after the controllable baffle, then dividing by the design pressure difference before and after the controllable baffle, and adding one to obtain a flow resistance deviation coefficient; multiplying the pump power and the time unit conversion coefficient, then multiplying the flow resistance deviation coefficient to obtain an energy consumption flow resistance term; and dividing the heat exchange power term by the energy consumption flow resistance term to obtain a flow heat energy efficiency evaluation value.

[0014] Further, the specific steps of executing genetic algorithm optimization and outputting the final controllable baffle opening combination scheme according to the evaluation result are as follows: record the flow-thermal energy efficiency evaluation value as the genetic fitness of the corresponding chromosome in the genetic algorithm, sort all chromosomes in descending order according to the genetic fitness, retain the chromosomes ranked in the top K positions to enter the next round of genetic operation, and K is a natural number less than or equal to the total number of chromosomes; perform a crossover operation on the retained chromosomes, exchange the controllable baffle opening parameters at the corresponding positions between two chromosomes in the order of parameter positions to generate new chromosomes; perform a mutation operation on the new chromosomes, perform a numerical adjustment operation on the controllable baffle opening parameters at the specified positions to form a new opening combination form; combine the new chromosomes with the allocation schemes corresponding to the retained chromosomes to form a new set of allocation schemes, and input them into the flow-thermal energy efficiency evaluation process to calculate the corresponding genetic fitness; decode the chromosomes generated after the crossover and mutation into controllable baffle opening combination schemes, and repeat the processes of evaluation, sorting, retention, crossover, and mutation until the genetic fitness does not improve in consecutive fixed evaluation periods, and select the allocation scheme with the highest genetic fitness as the final controllable baffle opening combination scheme.

[0015] Further, based on the final controllable baffle opening combination scheme, the driving actuator adjusts the controllable baffle position, and the specific steps of adjusting the actual opening of the actuator by combining the real-time liquid cooling flow-thermal energy operation data and performing controllable baffle opening correction calculation are as follows: receive the final controllable baffle opening combination scheme, and parse it into target opening instructions for each controllable baffle and transmit it to the controllable baffle actuator to drive the actuator to adjust the controllable baffle to the target position; during the controllable baffle adjustment process, real-time acquisition of actual opening data feedback from the controllable baffle position sensor is performed, and the target opening is compared, when the absolute value of the difference between the actual opening and the target opening exceeds the deviation threshold, the opening correction calculation based on the real-time liquid cooling flow-thermal energy operation data is performed to obtain the controllable baffle opening correction amount; perform arithmetic operation on the original actual opening and the controllable baffle opening correction amount to generate a new controllable baffle opening value and issue it as a target opening instruction to the controllable baffle actuator to drive the actuator to perform opening adjustment action; repeat the opening correction calculation and opening adjustment action until the absolute value of the difference between the actual opening and the target opening does not exceed the deviation threshold.

[0016] Further, the specific steps of performing the opening degree correction calculation based on the real-time liquid cooling flow heat energy operation data are as follows: subtracting the target chip temperature from the chip temperature and dividing by the target chip temperature to obtain a temperature deviation term; subtracting the target cooling liquid flow from the cooling liquid flow and dividing by the target cooling liquid flow to obtain a flow deviation term; subtracting the cooling liquid inlet temperature from the cooling liquid outlet temperature and dividing by the cooling liquid inlet temperature to obtain a temperature difference ratio term; multiplying the flow deviation term by the temperature difference ratio term to obtain a flow temperature difference coupling term; subtracting the designed pressure difference before and after the controllable baffle from the actual pressure difference before and after the controllable baffle, taking the absolute value, and then dividing by the designed pressure difference before and after the controllable baffle to obtain a pressure difference deviation term; adding the temperature deviation term and the flow temperature difference coupling term, and then subtracting the pressure difference deviation term to obtain the controllable baffle opening degree correction amount.

[0017] Further, after the controllable baffle adjustment is completed, the specific steps of constructing the operation monitoring data set are as follows: during the operation of the liquid cooling system, continuously collecting the liquid cooling flow heat energy operation data, the controllable baffle actual opening degree data and the flow heat energy efficiency evaluation value of each evaluation period, and extracting the maximum chip temperature and the minimum chip temperature during the operation from the chip temperature data, calculating the average chip temperature during the operation, and recording in time sequence as the operation monitoring data set.

[0018] Further, the specific steps of quantitatively evaluating the liquid cooling flow distribution stability and the chip temperature uniformity are as follows: based on the operation monitoring data set, dividing the flow heat energy efficiency evaluation value of the i-th evaluation period by the flow heat energy efficiency evaluation value of the previous evaluation period, subtracting one, squaring to obtain the energy efficiency change rate square term of the i-th evaluation period; summing all the energy efficiency change rate square terms and dividing by the total number of evaluation periods to obtain the energy efficiency fluctuation average term; subtracting the minimum chip temperature during the operation from the maximum chip temperature during the operation and dividing by the average chip temperature during the operation to obtain the temperature difference ratio term; adding the energy efficiency fluctuation average term and the temperature difference ratio term to obtain the liquid cooling flow heat balance evaluation value.

[0019] Further, the specific steps of adjusting the subsequent genetic algorithm parameters and controllable baffle opening degree combination scheme are as follows: comparing the liquid cooling flow heat balance evaluation value and the balance threshold in real time, when the liquid cooling flow heat balance evaluation value is not greater than the balance threshold, maintaining the existing controllable baffle opening degree combination scheme to the next evaluation period; when the liquid cooling flow heat balance evaluation value is greater than the balance threshold, triggering the adaptive optimization process: adjusting the crossover ratio, mutation ratio and iteration termination condition of the genetic algorithm based on the current operation monitoring data, generating a new controllable baffle opening degree combination scheme and outputting to the controllable baffle execution control and opening degree correction processing module.

[0020] Beneficial effects

[0021] The present application has the following beneficial effects:

[0022] (1) The liquid-cooled server flow dynamic adjustment system based on genetic algorithm can dynamically adjust the cooling liquid flow direction and flow distribution ratio during the operation of the liquid cooling system according to the real-time load, heat distribution and cooling demand of the chip, realize the preferential liquid supply of the high-load chip area, make the high-heat-flux area fully cooled, thereby quickly reduce the local temperature and avoid the accumulation of hot spots, ensure the stable operation of the chip, and take into account the cooling demand of the low-load area, and realize the optimal allocation of overall flow resources.

[0023] (2) The liquid-cooled server flow dynamic adjustment system based on genetic algorithm comprehensively considers the thermodynamic parameters reflecting the heat exchange capacity such as the cooling liquid density, cooling liquid flow, cooling liquid specific heat capacity and cooling liquid inlet and outlet temperature difference, and introduces the fluid mechanics parameters reflecting the flow resistance and energy consumption such as the pump power, actual pressure difference before and after the controllable baffle and design pressure difference, establishes a comprehensive evaluation formula combining heat exchange efficiency and flow resistance loss, which can comprehensively reflect the advantages and disadvantages of the controllable baffle opening combination scheme in the two dimensions of thermal management performance and energy utilization efficiency, avoid the problem of single optimization target and one-sided evaluation in the prior art, and provide a scientific fitness basis for the genetic algorithm.

[0024] (3) The liquid-cooled server flow dynamic adjustment system based on genetic algorithm introduces a real-time opening correction mechanism. When the absolute value of the difference between the actual controllable baffle opening and the target controllable baffle opening exceeds the deviation threshold, the controllable baffle opening correction amount is calculated based on the real-time liquid cooling flow and heat energy operation data. The correction amount comprehensively considers the chip temperature deviation term, the cooling liquid flow temperature difference coupling term and the pressure difference deviation term, and can reflect the comprehensive influence of temperature distribution, flow change and flow resistance deviation on the cooling effect. By calculating the correction amount and the original actual opening and real-time issuing to the actuator, the deviation caused by mechanical delay, external disturbance or changes in operating conditions can be quickly compensated to ensure that the system is always in the optimal operating state.

[0025] (4) The liquid-cooled server flow dynamic adjustment system based on genetic algorithm constructs an operation monitoring data set and calculates a liquid cooling flow heat balance evaluation value to quantify the stability of the liquid cooling flow distribution and the uniformity of the chip temperature. When the heat balance evaluation value exceeds the balance threshold, an adaptive optimization process is automatically triggered to dynamically adjust the crossover ratio, mutation ratio and iteration termination condition of the genetic algorithm based on the current operation monitoring data, thereby optimizing the subsequent controllable baffle opening combination scheme generation strategy. The adaptive mechanism enables the system to actively optimize the control strategy according to the changes in the operating state, improve the stability of the flow distribution, prolong the service life of the equipment, and reduce the risk of performance degradation caused by temperature unevenness. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1A structure diagram of a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm;

[0027] Figure 2 A column chart of chromosome flow heat energy efficiency evaluation value descending arrangement;

[0028] Figure 3 A structure diagram of a controllable baffle inside a liquid cooling system.

[0029] In the figure, 1 is a controllable baffle movable shaft, 2 is a controllable baffle, 3 is a central processing unit, 4 is a memory bank, 5 is a RAID card, 6 is a vertical plate, and 7 is a graphics card plate. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the present application.

[0031] Please refer to Figures 1-3 The embodiments of the present application provide a technical solution: a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm, comprising: a liquid-cooled flow heat energy data acquisition and preprocessing module, a controllable baffle opening degree optimization module based on a genetic algorithm, a controllable baffle execution control and opening degree correction processing module, and a running balance evaluation and adaptive optimization module, wherein: the liquid-cooled flow heat energy data acquisition and preprocessing module is used to acquire liquid-cooled flow heat energy running data, and perform time reference unification, noise suppression, physical consistency verification, data structure unification and scale standardization processing on the liquid-cooled flow heat energy running data to obtain preprocessed liquid-cooled flow heat energy running data; the controllable baffle opening degree optimization module based on a genetic algorithm is used to generate a controllable baffle opening degree combination scheme based on the preprocessed liquid-cooled flow heat energy running data, and evaluate the flow heat energy efficiency of the scheme, and execute genetic algorithm optimization and output the final controllable baffle opening degree combination scheme according to the evaluation result; the controllable baffle execution control and opening degree correction processing module is used to drive an execution mechanism to adjust the position of the controllable baffle 2 based on the final controllable baffle opening degree combination scheme, and execute controllable baffle opening degree correction calculation combined with real-time liquid-cooled flow heat energy running data to adjust the actual opening degree of the execution mechanism; the running balance evaluation and adaptive optimization module is used to construct a running monitoring data set after the controllable baffle 2 is adjusted, quantitatively evaluate the liquid-cooled flow distribution stability and chip temperature balance, and adjust subsequent genetic algorithm parameters and controllable baffle opening degree combination schemes.

[0032] Specifically, the specific steps of collecting liquid cooling flow heat energy operation data, performing time reference unification, noise suppression, physical consistency verification, data structure unification and scale standardization processing on the liquid cooling flow heat energy operation data are as follows: collecting liquid cooling flow heat energy operation data, the liquid cooling flow heat energy operation data including cooling liquid density, cooling liquid specific heat capacity, chip temperature, cooling liquid inlet temperature, cooling liquid outlet temperature, cooling liquid flow, controllable baffle front and back pressure difference, controllable baffle opening, pump power, target cooling liquid flow and target chip temperature; wherein the cooling liquid density is obtained by measuring the mass per unit volume of the cooling liquid in real time through an online density sensor; the cooling liquid specific heat capacity is obtained by experiment calibration and table lookup combined with cooling liquid formula parameters; the chip temperature is obtained by collecting the instant temperature of the chip core area through a distributed temperature sensor; the cooling liquid inlet temperature is obtained by measuring the cooling liquid temperature entering the cold plate in real time through a temperature sensor installed on the inlet pipeline; the cooling liquid outlet temperature is obtained by measuring the cooling liquid temperature leaving the cold plate in real time through a temperature sensor installed on the outlet pipeline; the cooling liquid flow is obtained by detecting the cooling liquid volume flowing through the cold plate per unit time through a flowmeter; the controllable baffle front and back pressure difference is obtained by synchronously acquiring the instant pressure difference value of the controllable baffle 2 on both sides through a pressure difference sensor. The controllable baffle opening is obtained by detecting the current opening position of the controllable baffle actuator in real time through a position sensor; the pump power is obtained by collecting the instant power consumption of the pump motor through a power sensor; the target cooling liquid flow is obtained by determining the optimal cooling liquid flow under a specific load condition through a heat balance algorithm; and the target chip temperature is obtained by determining the optimal operating temperature under a specific working condition through a thermal reliability design calculation.Then, through the global clock synchronization and sampling phase calibration method, the liquid cooling flow thermal energy operation data is processed for time reference unification and sampling interval matching, so that the timestamps of each data sampling point are aligned and the phase deviation between different collection channels is eliminated; through the fusion of frequency domain analysis and adaptive weight multi-parameter filtering method, the liquid cooling flow thermal energy operation data is processed for noise suppression and disturbance compensation, wherein the noise suppression process adopts adaptive cutoff frequency filtering for high-frequency random noise, and the disturbance compensation process is based on the running state characteristic curve to correct transient deviation; based on the thermodynamic conservation equation and the fluid continuity equation, a physical consistency checking mechanism is established to identify and eliminate abnormal data of the liquid cooling flow thermal energy operation data, so as to ensure that the relationship between the heat exchange power calculated by the cooling liquid density, cooling liquid flow rate, cooling liquid specific heat capacity and temperature difference and the pump power and flow resistance parameters meets the energy conservation and flow conservation conditions; through the standardized feature mapping and multi-protocol packaging method, the liquid cooling flow thermal energy operation data is processed for data structure unification and interface format conversion, so that the data of different sources and different communication protocols are completely consistent in structure, field order and coding format; through the dimension mapping and interval adaptive normalization method, the liquid cooling flow thermal energy operation data is processed for unit conversion and feature scale standardization, so as to ensure that the cooling liquid density, cooling liquid specific heat capacity, cooling liquid flow rate and other parameters have the same scale consistency for direct comparison and operation in the subsequent calculation and optimization process.

[0033] In the embodiment, by specifying the specific acquisition methods of cooling liquid density, cooling liquid specific heat capacity, chip temperature, cooling liquid inlet temperature, cooling liquid outlet temperature, cooling liquid flow rate, controllable baffle front and back pressure difference, controllable baffle opening, pump power, target cooling liquid flow rate, and target chip temperature, the source of the liquid cooling flow thermal energy operation data in the acquisition link is ensured to be reliable and traceable, so that the subsequent time reference unification, noise suppression, physical consistency checking, data structure unification, and scale standardization processing processes have a solid data foundation, thereby improving the accuracy of flow thermal energy efficiency evaluation calculation and the scientificity of genetic algorithm optimization of controllable baffle opening combination scheme.

[0034] Specifically, based on the pre-processed liquid cooling flow heat energy operation data, the specific steps of generating the controllable baffle opening degree combination scheme are as follows: according to the opening degree range of each controllable baffle 2, combining the structural parameters and cooling liquid flow characteristics of each cooling circuit in the liquid cooling system, a distribution scheme set containing multiple controllable baffle opening degree combination schemes is generated, each distribution scheme corresponds to a liquid cooling flow distribution mode and can cover the cooling requirements of high-load chip area, low-load chip area and medium-load chip area; each distribution scheme is expressed as a set of ordered parameter sequences according to the coding rules of the genetic algorithm, wherein the coding rules of the genetic algorithm are: the baffle opening degree of each controllable baffle 2 is quantitatively and discretely processed within the allowed opening degree range, and is arranged in sequence according to the physical position of the controllable baffle 2 in the cooling circuit of the liquid cooling system to form a parameter sequence, so that each position in the parameter sequence corresponds to the actual installation position of the corresponding controllable baffle 2; the above ordered parameter sequence is recorded as the chromosome of genetic operation, providing input basis for subsequent flow heat energy efficiency evaluation and genetic operation optimization.

[0035] In the present embodiment, by combining the opening degree range of each controllable baffle 2, the structural parameters and cooling liquid flow characteristics of each cooling circuit in the liquid cooling system, a plurality of distribution scheme sets are generated and converted into ordered parameter sequences in accordance with the coding rules of the genetic algorithm, so that the controllable baffle opening degree combination scheme can be accurately mapped to the actual controllable baffle 2 position, thereby ensuring the integrity and accuracy of the input of the optimization process in the subsequent genetic operation, improving the reliability of the flow heat energy efficiency evaluation result, and laying a foundation for realizing differentiated liquid cooling flow distribution for different load chip areas.

[0036] Specifically, the specific steps of evaluating the flow-heat energy efficiency of the scheme are as follows: multiplying the cooling liquid density, the cooling liquid flow rate and the specific heat capacity of the cooling liquid, and then multiplying the difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature, to obtain a heat exchange power term that can reflect the heat removal capacity of the cooling liquid from the chip area per unit time; taking the absolute value of the difference between the actual pressure difference before and after the controllable baffle and the design pressure difference before and after the controllable baffle, and then dividing the design pressure difference before and after the controllable baffle, and then adding 1, to obtain a flow resistance deviation coefficient that can represent the deviation degree of the flow resistance caused by the structure of the controllable baffle 2; multiplying the pump power and the time unit conversion coefficient, and then multiplying the flow resistance deviation coefficient, to obtain an energy consumption flow resistance term for quantifying the additional energy consumption caused by the flow resistance deviation in the cooling liquid transportation process; dividing the heat exchange power term by the energy consumption flow resistance term, to obtain a flow-heat energy efficiency evaluation value that can comprehensively reflect the performance of the controllable baffle opening combination scheme in terms of the balance between heat transfer efficiency and energy consumption loss. The design pressure difference before and after the controllable baffle refers to the reference value of the pressure difference before and after the controllable baffle calculated according to the pipeline and the controllable baffle 2 structure design under the condition of the rated flow rate of the liquid cooling system; the time unit conversion coefficient is a positive real number calculated by a time unit conversion formula using the flow rate record and the time record in the liquid cooling flow-heat energy operation data to realize the conversion between seconds, minutes or hours.

[0037] The specific calculation formula of the flow-heat energy efficiency evaluation value is as follows:

[0038] ;

[0039] In the formula, represents the flow-heat energy efficiency evaluation value, represents the cooling liquid density, represents the cooling liquid flow rate, represents the specific heat capacity of the cooling liquid, represents the cooling liquid outlet temperature, represents the cooling liquid inlet temperature, represents the pump power, represents the actual pressure difference before and after the controllable baffle, represents the design pressure difference before and after the controllable baffle, represents the time unit conversion coefficient.

[0040] In this embodiment, Table 1 is a flow-heat energy efficiency evaluation value data table, which lists the liquid cooling flow-heat energy operation data indicators and the corresponding calculated flow-heat energy efficiency evaluation values of the five chromosomes in the liquid cooling flow distribution optimization process. The specific data are as follows: in the chromosome R1, the cooling liquid density is 1035 kg / m³, the cooling liquid flow rate is 0.035 m³ / s, the specific heat capacity of the cooling liquid is 3850 J / (kg·K), the cooling liquid outlet temperature is 35℃, the cooling liquid inlet temperature The temperature is 30℃, the pump power is 1200W, the actual pressure difference before and after the controllable baffle is 18000Pa, the design pressure difference before and after the controllable baffle is 20000Pa, the time unit conversion factor is 60s, and the calculated flow-heat efficiency evaluation value is 8.72; in chromosome R2, the coolant density is 1030kg / m³, the coolant flow rate is 0.040m³ / s, the coolant specific heat capacity is 3860J / (kg·K), the coolant outlet temperature is 36℃, the coolant inlet temperature is 31℃, the pump power is 1250W, and the pressure difference before and after the controllable baffle is... The actual pressure difference is 18000 Pa, the design pressure difference before and after the controllable baffle is 20000 Pa, the time unit conversion factor is 60 s, and the calculated flow-heat efficiency evaluation value is 10.10. In chromosome R3, the coolant density is 1020 kg / m³, the coolant flow rate is 0.038 m³ / s, the coolant specific heat capacity is 3840 J / (kg·K), the coolant outlet temperature is 34℃, the coolant inlet temperature is 29℃, the pump power is 1180 W, the actual pressure difference before and after the controllable baffle is 17500 Pa, and the design pressure difference before and after the controllable baffle is... The pressure difference is 20000 Pa, the time unit conversion factor is 60 s, and the calculated heat transfer efficiency evaluation value is 9.34. In chromosome R4, the coolant density is 1028 kg / m³, the coolant flow rate is 0.036 m³ / s, the coolant specific heat capacity is 3855 J / (kg·K), the coolant outlet temperature is 37℃, the coolant inlet temperature is 32℃, the pump power is 1230 W, the actual pressure difference before and after the controllable baffle is 18500 Pa, the design pressure difference before and after the controllable baffle is 20000 Pa, and the time unit conversion factor is... For 60 seconds, the calculated heat transfer efficiency evaluation value is 8.99. In chromosome R5, the coolant density is 1023 kg / m³, the coolant flow rate is 0.039 m³ / s, the coolant specific heat capacity is 3852 J / (kg·K), the coolant outlet temperature is 33℃, the coolant inlet temperature is 28℃, the pump power is 1190 W, the actual pressure difference before and after the controllable baffle is 17800 Pa, the design pressure difference before and after the controllable baffle is 20000 Pa, and the time unit conversion factor is 60 s. The calculated heat transfer efficiency evaluation value is 9.70.

[0041] Table 1. Data Table of Thermal Efficiency Evaluation Values

[0042]

[0043] like Figure 2 The bar chart shown is a descending order of chromosome heat efficiency assessment values, displaying the results of heat efficiency assessment for five different chromosomes, arranged from highest to lowest. The horizontal axis represents the chromosome number, and the vertical axis represents the corresponding heat efficiency assessment value. The results show that chromosome R2 has the highest heat efficiency assessment value, followed by R5, R3, and R4, while R1 has the lowest assessment value.Figure 2 The relative advantages and disadvantages of different chromosomes in the liquid cooling flow distribution optimization can be intuitively reflected, and a reference basis is provided for genetic algorithm screening and optimization.

[0044] In the embodiment, by comprehensively calculating the coolant density, the coolant flow rate, the specific heat capacity of the coolant, the coolant outlet temperature, the coolant inlet temperature, the pump power, the time unit conversion coefficient, the actual pressure difference before and after the controllable baffle, and the design pressure difference before and after the controllable baffle, the contributions of the heat exchange power term and the energy consumption flow resistance term can be quantified in the same evaluation system, and it is ensured that the obtained flow-heat energy efficiency evaluation value can comprehensively reflect the overall performance of the controllable baffle opening combination scheme in terms of thermal management performance and energy utilization efficiency, thereby providing an accurate, stable and physically consistent fitness reference for genetic algorithm optimization.

[0045] Specifically, the specific steps of executing genetic algorithm optimization according to the evaluation results and outputting the final controllable baffle opening combination scheme are as follows: the flow-heat energy efficiency evaluation value obtained by calculation is accurately recorded as the genetic fitness of the corresponding chromosome in the genetic algorithm, so that the thermal management performance and energy utilization efficiency of each chromosome are quantitatively reflected; all chromosomes are sorted in descending order of genetic fitness, and the chromosomes ranked in the top K positions are retained for the next round of genetic operation, wherein K is a natural number less than or equal to the total number of chromosomes and is set according to the system running scale and optimization accuracy requirement; the retained chromosomes are subjected to cross operation, the controllable baffle opening parameters at corresponding positions are exchanged one by one in a fixed parameter position order between two chromosomes to generate offspring chromosomes that can inherit the characteristics of the parents and have new combination characteristics; the new chromosomes are subjected to mutation operation, the controllable baffle opening parameters at the specified positions are subjected to numerical adjustment operation with limited amplitude to increase the diversity of the population and avoid falling into local optimum; the distribution schemes corresponding to the new chromosomes and the retained chromosomes are combined to form a new set of distribution schemes including the parents and the offspring, and are re-input into the flow-heat energy efficiency evaluation process to calculate the corresponding genetic fitness to form a new optimization basis; the chromosomes generated after cross and mutation are decoded into controllable baffle opening combination schemes that can be directly used for control execution, and the cycle process of evaluation, sorting, retention, cross and mutation is repeatedly executed until the genetic fitness does not increase in consecutive fixed evaluation periods, and the distribution scheme with the highest genetic fitness is selected as the final controllable baffle opening combination scheme and enters the subsequent control execution link. The evaluation period is a time period in which the liquid cooling system performs running data acquisition, performance calculation and control decision once in a fixed time interval.

[0046] In this embodiment, by accurately associating the flow-thermal energy efficiency evaluation value with the genetic fitness of the genetic algorithm, and performing parameter position order crossover operation and amplitude limited numerical adjustment mutation operation on the basis of preserving high fitness chromosomes, it is ensured that the new chromosomes generated have optimization potential in terms of thermal management performance and energy utilization efficiency; by merging the decoded and distributed schemes of the chromosomes after crossover and mutation, and combining the fitness change trend in a continuous fixed evaluation period to judge the optimization termination condition, the final output controllable baffle opening combination scheme can approach the optimal solution in the global search space, thereby improving the flow distribution accuracy and system heat dissipation stability driven by liquid cooling flow-thermal energy operation data.

[0047] Specifically, based on the final controllable baffle opening combination scheme, the actuator is driven to adjust the position of the controllable baffle 2, and the controllable baffle opening correction calculation is performed in combination with the real-time liquid cooling flow-thermal energy operation data. The specific steps of adjusting the actual opening of the actuator are as follows: receiving the final controllable baffle opening combination scheme and parsing it into target opening instructions for each controllable baffle 2 and transmitting it to the controllable baffle actuator through the industrial bus protocol, driving the actuator to adjust the controllable baffle 2 to the target position according to the target opening instruction; during the adjustment of the controllable baffle 2, the actual opening data fed back by the controllable baffle position sensor is collected in real time, and is compared with the target opening stored in the controller cache one by one, when the absolute value of the difference between the actual opening and the target opening exceeds the deviation threshold, the opening correction calculation module based on real-time liquid cooling flow-thermal energy operation data is called, and the controllable baffle opening correction amount is calculated using parameters such as chip temperature, target chip temperature, cooling liquid flow, target cooling liquid flow, cooling liquid outlet temperature, cooling liquid inlet temperature, actual pressure difference before and after the controllable baffle, and design pressure difference before and after the controllable baffle, wherein the deviation threshold is the maximum absolute deviation range allowed between the actual opening and the target opening, and the maximum absolute deviation range is determined according to the heat dissipation performance requirement and control accuracy requirement of the liquid cooling system; the original actual opening collected and the controllable baffle opening correction amount calculated are arithmetically operated to generate a new controllable baffle opening value, and the new controllable baffle opening value is issued as a target opening instruction to the controllable baffle actuator, driving the actuator to perform fine opening adjustment action; the cycle of opening correction calculation and opening adjustment action is repeated until the absolute value of the difference between the actual opening of all controllable baffles 2 and the corresponding target opening does not exceed the deviation threshold, thereby ensuring that the controllable baffle opening accuracy meets the dynamic control requirements driven by liquid cooling flow-thermal energy operation data.

[0048] As Figure 3As shown, the controllable baffle movable shaft 1 is used for rotating connection with the vertical plate 6 to support the controllable baffle 2 to realize angle adjustment; the controllable baffle 2 changes the flow path of the cooling liquid in different hardware regions and adjusts the flow distribution ratio under the driving of the controllable baffle actuator, so as to realize the dynamic distribution of the liquid cooling flow in cooperation with the controllable baffle opening combination scheme generated based on the genetic algorithm. The central processing unit 3 is provided with a temperature sensor at the back, which is used for collecting the chip temperature; the memory bank 4 is used for storing and reading and writing data at high speed; the RAID card 5 is used for data management and transmission of the multi-hard disk array; the vertical plate 6 is used for fixing and supporting various hardware components and providing a rotating installation position for the controllable baffle 2; and the graphics card plate 7 is used for graphic processing and accelerated calculation. During the operation of the liquid cooling system, the heat generated by the central processing unit 3, the memory bank 4, the RAID card 5 and the graphics card plate 7 is taken away by the cooling liquid, and the chip temperature collected by the temperature sensor is input into the controllable baffle execution control and opening correction processing module as part of the liquid cooling flow and heat energy operation data, which is used to guide the real-time adjustment and optimization of the controllable baffle opening.

[0049] In the embodiment, by collecting the actual opening data feedback by the controllable baffle position sensor in real time during the adjustment of the controllable baffle 2 and comparing with the target opening, when the absolute value of the difference between the actual opening and the target opening exceeds the deviation threshold, the controllable baffle opening correction amount is calculated by using the liquid cooling flow and heat energy operation data, so that when the mechanical delay of the actuator exists, the running condition changes, and the external disturbance affects the opening accuracy, the opening deviation can be quickly corrected, the actual opening of the controllable baffle is continuously close to the target opening, and the stability of the liquid cooling system and the consistency of the heat dissipation performance are maintained.

[0050] Specifically, the specific steps of performing the opening correction calculation based on the real-time liquid cooling flow heat energy operation data are as follows: subtracting the target chip temperature from the collected chip temperature and dividing by the target chip temperature to obtain a temperature deviation term for quantifying the relative deviation of the current chip actual operation temperature from the target chip temperature; subtracting the target cooling liquid flow from the real-time collected cooling liquid flow and dividing by the target cooling liquid flow to obtain a flow deviation term for reflecting the relative change amplitude of the actual liquid supply flow of the current liquid cooling system from the target cooling liquid flow; subtracting the cooling liquid inlet temperature from the real-time collected cooling liquid outlet temperature and dividing by the cooling liquid inlet temperature to obtain a temperature difference ratio term for characterizing the relative temperature rise degree of the current cooling liquid in the heat exchange process; multiplying the flow deviation term by the temperature difference ratio term to obtain a flow temperature difference coupling term for reflecting the influence of the combined action of flow change and heat exchange temperature difference on the cooling performance; subtracting the absolute value of the actual pressure difference before and after the controllable baffle from the design pressure difference before and after the controllable baffle, and then dividing by the design pressure difference before and after the controllable baffle to obtain a pressure difference deviation term for measuring the relative deviation of the actual flow resistance state from the design flow resistance state; adding the temperature deviation term and the flow temperature difference coupling term, and then subtracting the pressure difference deviation term to obtain the controllable baffle opening correction amount, which serves as the basis for calculating the controllable baffle opening adjustment.

[0051] wherein the specific calculation formula of the controllable baffle opening correction amount is:

[0052] ;

[0053] In the formula, represents the controllable baffle opening correction amount, represents the chip temperature, represents the target chip temperature, represents the cooling liquid flow, represents the target cooling liquid flow, represents the cooling liquid outlet temperature, represents the cooling liquid inlet temperature, represents the actual pressure difference before and after the controllable baffle, represents the design pressure difference before and after the controllable baffle.

[0054] In the present embodiment, by comprehensively introducing the temperature deviation term, the flow deviation term, the temperature difference ratio term, the flow temperature difference coupling term and the pressure difference deviation term in the calculation process, the simultaneous quantitative analysis of the chip temperature change, the cooling liquid flow change, the cooling liquid inlet and outlet temperature difference change and the pressure difference change before and after the controllable baffle is realized, which can accurately reflect the operation deviation state of the liquid cooling system under the dual dimensions of thermodynamics and fluid mechanics, thereby providing a scientific basis for generating the controllable baffle opening correction amount, improving the accuracy and response efficiency of the controllable baffle opening adjustment, and ensuring the high-efficiency and stable cooling performance of the liquid cooling system under different operating conditions.

[0055] Specifically, after the controllable baffle 2 adjustment is completed, the specific steps of constructing the operation monitoring data set are as follows: during the operation of the liquid cooling system, the liquid cooling flow heat energy operation data, the controllable baffle actual opening data and the flow heat energy efficiency evaluation value of each evaluation period are continuously collected, and the time synchronization mechanism is used to ensure that the time stamps of various types of data are consistent; the maximum chip temperature and the minimum chip temperature during operation are extracted from the chip temperature data, the extraction process covers the complete operation time and excludes extreme value deviation caused by sensor instantaneous abnormality, and statistical operation is performed on the chip temperature sequence to obtain the average chip temperature during operation; all the above data are recorded in time sequence as the operation monitoring data set.

[0056] In the embodiment, by continuously collecting the liquid cooling flow heat energy operation data, the controllable baffle actual opening data and the flow heat energy efficiency evaluation value of each evaluation period during the operation of the liquid cooling system, and using the time synchronization mechanism to ensure that the time stamps of different types of data are consistent, the accurate alignment of multi-source operation data is realized; when the maximum chip temperature and the minimum chip temperature during operation are extracted, the complete operation time is ensured and the extreme value deviation caused by sensor instantaneous abnormality is excluded, and the average chip temperature during operation is obtained by statistical operation on the chip temperature sequence, which ensures the accuracy and representativeness of the temperature characteristic data; finally, all the operation data are recorded in time sequence as the operation monitoring data set, which provides a high-precision, time-consistent and complete data basis for subsequent quantitative evaluation of the liquid cooling flow distribution stability and the chip temperature uniformity.

[0057] Specifically, the specific steps of quantitatively evaluating the liquid cooling flow distribution stability and the chip temperature uniformity are as follows: based on the operation monitoring data set, the flow heat energy efficiency evaluation value of the i-th evaluation period is divided by the flow heat energy efficiency evaluation value of the previous evaluation period, then subtracted by one, squared to obtain the energy efficiency change rate square item of the i-th evaluation period, wherein the selection of the evaluation period needs to ensure that the stable operation stage is covered and the fluctuation interference of the start-up or shutdown stage is avoided; the sum of all energy efficiency change rate square items is divided by the total number of evaluation periods to obtain the energy efficiency fluctuation mean value item, which can quantitatively reflect the stability level of the liquid cooling flow heat energy operation data in different periods; the maximum chip temperature during operation is subtracted from the minimum chip temperature during operation, and then divided by the average chip temperature during operation to obtain the temperature difference proportion item, which needs to ensure that the temperature data comes from the time-synchronized chip temperature sequence and excludes instantaneous abnormal values during the calculation process; the energy efficiency fluctuation mean value item and the temperature difference proportion item are added to obtain the liquid cooling flow heat balance evaluation value.

[0058] The specific calculation formula of the liquid cooling flow heat balance evaluation value is as follows:

[0059] ;

[0060] In the formula, represents the liquid cooling flow heat balance evaluation value, represents the flow heat energy efficiency evaluation value of the ith evaluation period, represents the flow heat energy efficiency evaluation value of the i-1th evaluation period, and n represents the total number of evaluation periods, represents the maximum chip temperature during operation, represents the minimum chip temperature during operation, represents the average value of the chip temperature during operation.

[0061] In the embodiment, by calculating the energy efficiency fluctuation mean term and the temperature difference proportion term on the basis of the operation monitoring data set, and combining the two to obtain the liquid cooling flow heat balance evaluation value, the integrated quantitative evaluation of the liquid cooling flow distribution stability and the chip temperature balance is realized, which can provide accurate characterization for the state change of the liquid cooling flow heat energy operation data in the long-term operation process, thereby providing scientific and reliable judgment basis for the adaptive optimization process and improving the stability and balance of the liquid cooling system operation.

[0062] Specifically, the specific steps of adjusting the subsequent genetic algorithm parameters and controllable baffle opening degree combination scheme are as follows: real-time comparison of the liquid cooling flow heat balance evaluation value and the balance threshold value, and ensuring the accuracy of the evaluation value and the threshold value comparison through the periodic operation monitoring mechanism; when the liquid cooling flow heat balance evaluation value is not greater than the balance threshold value, the current executed controllable baffle opening degree combination scheme is directly maintained to the next evaluation period, and the liquid cooling flow heat energy operation data is continuously collected in the evaluation period for subsequent analysis; when the liquid cooling flow heat balance evaluation value is greater than the balance threshold value, the adaptive optimization process is immediately triggered, and based on all variables including the cooling liquid density, the cooling liquid flow, the cooling liquid specific heat capacity, the chip temperature, the cooling liquid inlet temperature, the cooling liquid outlet temperature, the actual pressure difference before and after the controllable baffle, the design pressure difference before and after the controllable baffle, the pump power and the controllable baffle opening degree in the current operation monitoring data, the crossover ratio, the mutation ratio and the iteration termination condition of the genetic algorithm are dynamically adjusted to generate a new controllable baffle opening degree combination scheme that meets the current operation state, and the combination scheme is directly output to the controllable baffle execution control and opening correction processing module for execution.

[0063] In the embodiment, the accuracy of the comparison result is ensured by introducing a periodic operation monitoring mechanism in the real-time comparison of the liquid cooling flow heat balance evaluation value and the balance threshold value, and when the liquid cooling flow heat balance evaluation value is greater than the balance threshold value, the cooling liquid density, cooling liquid flow, cooling liquid specific heat capacity, chip temperature, cooling liquid inlet temperature, cooling liquid outlet temperature, actual pressure difference before and after the controllable baffle, designed pressure difference before and after the controllable baffle, pump power and controllable baffle opening degree and other multi-dimensional parameters included in the operation monitoring data are combined to dynamically adjust the genetic algorithm crossover ratio, mutation ratio and iteration termination condition, so as to generate a controllable baffle opening combination scheme highly matched with the current operation state and timely issue and execute, realizing the rapid optimization of the flow distribution strategy and the stability improvement of the system operation.

[0064] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0065] The above disclosed preferred embodiments of the present application are only used to help explain the present application. The preferred embodiments do not describe all the details and do not limit the present application to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of the present specification. The present specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present application, so that those skilled in the art can well understand and utilize the present application. The present application is limited only by the claims and their full scope and equivalents.

Claims

1. A liquid-cooled server flow dynamic adjustment system based on genetic algorithm, characterized in that, include: The system comprises a liquid cooling flow thermal energy data acquisition and preprocessing module, a controllable baffle opening optimization module based on a genetic algorithm, a controllable baffle execution control and opening correction processing module, and an operation balance evaluation and adaptive optimization module, among which: The liquid cooling flow thermal energy data acquisition and preprocessing module is used to acquire liquid cooling flow thermal energy operation data and perform time base unification, noise suppression, physical consistency verification, data structure unification and scale standardization on the liquid cooling flow thermal energy operation data to obtain preprocessed liquid cooling flow thermal energy operation data. The controllable baffle opening optimization module based on genetic algorithm is used to generate controllable baffle opening combination schemes based on preprocessed liquid cooling flow thermal energy operation data, evaluate the flow thermal energy efficiency of the schemes, perform genetic algorithm optimization based on the evaluation results, and output the final controllable baffle opening combination scheme. The controllable baffle execution control and opening correction processing module is used to drive the actuator to adjust the position of the controllable baffle (2) based on the final controllable baffle opening combination scheme, and to perform controllable baffle opening correction calculation in combination with real-time liquid cooling flow thermal energy operation data to adjust the actual opening of the actuator; The operation balance evaluation and adaptive optimization module is used to construct an operation monitoring dataset after the controllable baffle (2) is adjusted, to quantitatively evaluate the stability of liquid cooling flow distribution and chip temperature balance, and to adjust the subsequent genetic algorithm parameters and controllable baffle opening combination scheme.

2. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for collecting liquid cooling flow thermal energy operation data and performing time base unification, noise suppression, physical consistency verification, data structure unification, and scale standardization on the liquid cooling flow thermal energy operation data to obtain preprocessed liquid cooling flow thermal energy operation data are as follows: Collect liquid cooling flow thermal energy operation data, including coolant density, coolant specific heat capacity, chip temperature, coolant inlet temperature, coolant outlet temperature, coolant flow rate, pressure difference before and after the controllable baffle, controllable baffle opening, pump power, target coolant flow rate, and target chip temperature. By using global clock synchronization and sampling phase calibration methods, the time base of liquid cooling thermal energy operation data is unified and the sampling interval is matched. By integrating frequency domain analysis and adaptive weighting multi-parameter filtering methods, noise suppression and disturbance compensation are performed on liquid cooling flow thermal energy operation data; a physical consistency verification mechanism is established based on thermodynamic conservation equations and fluid continuity equations to identify and remove abnormal data from liquid cooling flow thermal energy operation data. By using standardized feature mapping and multi-protocol encapsulation methods, the liquid cooling flow thermal energy operation data is processed to unify the data structure and convert the interface format; by using dimensional mapping and interval adaptive normalization methods, the liquid cooling flow thermal energy operation data is processed to convert units and standardize the feature scale.

3. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for generating a controllable baffle opening combination scheme based on the pre-processed liquid cooling thermal energy operation data are as follows: Based on the opening range of each controllable baffle (2), a set of allocation schemes containing multiple controllable baffle opening combination schemes is generated. Each allocation scheme corresponds to a liquid cooling flow allocation method. Each allocation scheme is represented as a set of ordered parameter sequences according to the encoding rules of the genetic algorithm, and the ordered parameter sequences are recorded as chromosomes of the genetic operation.

4. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for evaluating the thermal efficiency of the aforementioned scheme are as follows: Multiply the coolant density, coolant flow rate, and coolant specific heat capacity together, then multiply by the difference between the coolant outlet temperature and the coolant inlet temperature to obtain the heat transfer power term. Take the absolute value of the difference between the actual pressure difference and the design pressure difference before and after the controllable baffle, divide it by the design pressure difference before and after the controllable baffle, and add one to obtain the flow resistance deviation coefficient. Multiply the pump power by the time unit conversion factor, then multiply by the flow resistance deviation coefficient to obtain the energy consumption flow resistance term. Divide the heat transfer power term by the energy consumption flow resistance term to obtain the flow heat efficiency evaluation value.

5. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for performing genetic algorithm optimization based on the evaluation results and outputting the final controllable baffle opening combination scheme are as follows: The thermal efficiency evaluation value is recorded as the genetic fitness of the corresponding chromosome in the genetic algorithm. All chromosomes are sorted in descending order of genetic fitness, and the chromosomes with the top K genetic fitness rankings are retained to enter the next round of genetic operation, where K is a natural number less than or equal to the total number of chromosomes. Crossover operation is performed on the retained chromosomes, and the controllable baffle opening parameters at corresponding positions are exchanged between two chromosomes according to the parameter position order to generate a new chromosome. Mutation operation is performed on the new chromosome, and the controllable baffle opening parameters are numerically adjusted at specified positions to form a new opening combination. The allocation schemes corresponding to the new chromosomes and the retained chromosomes are merged to form a new set of allocation schemes, which are then re-input into the thermal efficiency evaluation process to calculate the corresponding genetic fitness. The chromosomes generated after crossover and mutation are re-decoded into controllable baffle opening combination schemes, and the process of evaluation, sorting, retention, crossover and mutation is repeated until the genetic fitness does not improve in a continuous fixed evaluation cycle. The allocation scheme with the highest genetic fitness is selected as the final controllable baffle opening combination scheme.

6. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for adjusting the actual opening degree of the actuator based on the final controllable baffle opening combination scheme are as follows: The actuator is driven to adjust the position of the controllable baffle, and the controllable baffle opening degree correction calculation is performed in conjunction with real-time liquid cooling thermal energy operation data. The final controllable baffle opening combination scheme is received and parsed into target opening instructions for each controllable baffle (2) and transmitted to the controllable baffle actuator to drive the actuator to adjust the controllable baffle (2) to the target position. During the adjustment of the controllable baffle (2), the actual opening data fed back by the controllable baffle position sensor is collected in real time and compared with the target opening. When the absolute value of the difference between the actual opening and the target opening exceeds the deviation threshold, the opening correction calculation based on the real-time liquid cooling flow thermal energy operation data is performed to obtain the controllable baffle opening correction amount. The original actual opening degree and the controllable baffle opening degree correction amount are arithmetically calculated to generate a new controllable baffle opening degree value, which is then sent to the controllable baffle actuator as the target opening degree command to drive the actuator to perform the opening degree adjustment action; the opening degree correction calculation and opening degree adjustment action are repeated until the absolute value of the difference between the actual opening degree and the target opening degree does not exceed the deviation threshold.

7. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 6, characterized in that: The specific steps for performing the opening correction calculation based on real-time liquid cooling flow thermal energy operation data are as follows: The following steps are performed to obtain the following terms:

1. Subtract the target chip temperature from the chip temperature and divide by the target chip temperature to obtain the temperature deviation term; 2. Subtract the target coolant flow rate from the coolant flow rate and divide by the target coolant flow rate to obtain the flow deviation term; 3. Subtract the coolant inlet temperature from the coolant outlet temperature and divide by the coolant inlet temperature to obtain the temperature difference ratio term; 4. Multiply the flow deviation term by the temperature difference ratio term to obtain the flow-temperature difference coupling term; 5. Subtract the design pressure difference before and after the controllable baffle from the actual pressure difference before and after the controllable baffle, take the absolute value, and divide by the design pressure difference before and after the controllable baffle to obtain the pressure difference deviation term; 6. Add the temperature deviation term and the flow-temperature difference coupling term, and then subtract the pressure difference deviation term to obtain the controllable baffle opening correction amount.

8. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for constructing the operation monitoring dataset after the controllable baffle adjustment is completed are as follows: During the operation of the liquid cooling system, liquid cooling thermal energy operation data, actual opening data of controllable baffles, and thermal energy efficiency evaluation values ​​for each evaluation period are continuously collected. The maximum and minimum chip temperatures during operation are extracted from the chip temperature data, the average chip temperature during operation is calculated, and the data are recorded in chronological order as an operation monitoring dataset.

9. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for quantitatively evaluating the stability of liquid cooling flow distribution and chip temperature uniformity are as follows: Based on the operational monitoring dataset, the thermal efficiency evaluation value of the i-th evaluation period is divided by the thermal efficiency evaluation value of the previous evaluation period, minus one, and then squared to obtain the squared term of the efficiency change rate of the i-th evaluation period. The sum of all squared terms of efficiency change rate is divided by the total number of evaluation periods to obtain the average term of efficiency fluctuation. The maximum chip temperature during operation is subtracted from the minimum chip temperature during operation and then divided by the average chip temperature during operation to obtain the temperature difference ratio term. The average term of efficiency fluctuation and the temperature difference ratio term are added to obtain the liquid cooling flow thermal balance evaluation value.

10. The liquid-cooled server flow dynamic adjustment system based on genetic algorithm according to claim 1, characterized in that: The specific steps for adjusting the combination scheme of subsequent genetic algorithm parameters and controllable baffle opening are as follows: The liquid cooling flow rate thermal equilibrium assessment value and the equilibrium threshold are compared in real time. When the liquid cooling flow rate thermal equilibrium assessment value is not greater than the equilibrium threshold, the existing controllable baffle opening combination scheme is maintained until the next assessment cycle. When the liquid cooling flow rate thermal equilibrium assessment value is greater than the equilibrium threshold, the adaptive optimization process is triggered: based on the current operation monitoring data, the crossover ratio, mutation ratio and iteration termination condition of the genetic algorithm are adjusted to generate a new controllable baffle opening combination scheme and output to the controllable baffle execution control and opening correction processing module.

Images