Liquid cooling server flow dynamic adjusting system based on genetic algorithm

Abstract

The invention discloses a liquid cooling server flow dynamic adjusting system based on a genetic algorithm, and relates to the technical field of liquid cooling flow data processing. The liquid cooling server flow dynamic adjusting system based on the genetic algorithm comprises a liquid cooling flow heat energy data collecting and preprocessing module used for collecting and preprocessing liquid cooling flow heat energy operation data; the genetic algorithm-based controllable baffle opening degree optimization module is used for generating a controllable baffle opening degree combination scheme, evaluating the flow heat energy efficiency of the scheme, executing genetic algorithm optimization and outputting a final controllable baffle opening degree combination scheme; the controllable baffle execution control and opening correction processing module is used for adjusting the position of the controllable baffle, executing controllable baffle opening correction calculation and adjusting the actual opening; and the operation equilibrium evaluation and self-adaptive optimization module is used for evaluating the liquid cooling flow thermal equilibrium state and adjusting subsequent genetic algorithm parameters. The problem that the cooling efficiency of a high-load chip area is insufficient due to the fact that the flow direction of cooling liquid cannot be adjusted according to real-time loads in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of liquid cooling flow data processing technology, specifically to a liquid cooling server flow dynamic adjustment system based on a genetic algorithm. Background Technology

[0002] In large-scale data centers and high-performance computing scenarios, as processor power density continues to increase, liquid cooling technology is gradually replacing traditional air cooling as the mainstream heat dissipation method. However, most existing liquid cooling systems generally adopt fixed controllable baffle flow guiding structures or simple flow regulation methods, lacking the ability to dynamically adapt to the real-time load differences of different chips. This easily leads to insufficient cooling of high-load chip areas and the inability to suppress local temperature hotspots in a timely manner. Therefore, how to achieve intelligent allocation and rapid optimization of multi-path flow in liquid cooling systems has become a key technical direction for improving overall heat dissipation performance and operational stability.

[0003] For example, application CN115666088B provides an electronic device and a liquid cooling flow control method thereof. The electronic device includes electronic components, a cold plate, a flow regulation unit, an operating parameter acquisition unit, and a control unit. The cold plate is arranged adjacent to the electronic components and is used to circulate coolant for heat dissipation. The cold plate includes an inlet and an outlet. The flow regulation unit is located at the inlet and is used to regulate the flow rate of the coolant flowing into the cold plate. The operating parameter acquisition unit is used to acquire the operating parameters of the electronic components. The control unit is used to acquire the operating parameters of the electronic components acquired by the operating parameter acquisition unit and, at least based on the acquired operating parameters, controls the flow regulation unit to adjust the flow rate of the coolant flowing into the cold plate.

[0004] For example, application 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 monitored object in the liquid cooling computing device to obtain multiple first operating parameters of each monitored object at a first moment; wherein the first operating parameters can characterize the temperature of the monitored object, the monitored 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 monitored object is one of all monitored objects; and controlling the cooling unit in the liquid cooling computing device by means of the BMC of the first monitored object, based on the first operating parameters of each monitored object, to adjust the flow rate and temperature of the coolant.

[0005] Although the above technical solutions can achieve flow regulation of liquid cooling systems to a certain extent, they still have problems such as a single optimization target and a long response cycle for parameter adjustment. This is mainly manifested in the fact that they only perform single-path flow control based on a small number of temperature or flow parameters, lack the ability to optimize the global distribution of coolant across multiple paths, and cannot quickly generate and execute the optimal flow distribution scheme when there are significant load differences between different chips. As a result, the cooling of high-load areas is insufficient and the coolant utilization rate in low-load areas is low, making it difficult to cope with the real-time cooling requirements caused by load differences between multiple nodes.

[0006] Therefore, to address the above issues, there is an urgent need for a liquid-cooled server dynamic flow adjustment system based on genetic algorithms. Summary of the Invention

[0007] Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm, which solves the problem of insufficient cooling efficiency in high-load chip areas caused by the inability of existing technologies to adjust the coolant flow direction according to real-time load. Technical solution

[0008] To achieve the above objectives, this invention provides the following technical solution: a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm, comprising: a liquid-cooled 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. The liquid-cooled flow thermal energy data acquisition and preprocessing module is used to acquire liquid-cooled flow thermal energy operation data and perform time base unification, noise suppression, physical consistency verification, data structure unification, and scale standardization processing on the liquid-cooled flow thermal energy operation data to obtain preprocessed liquid-cooled flow thermal energy operation data. The controllable baffle opening optimization module based on a genetic algorithm is used to optimize the flow based on the preprocessed liquid-cooled flow thermal energy operation data. The system uses liquid cooling flow and thermal energy operation data to generate controllable baffle opening combination schemes, evaluates the flow and thermal energy efficiency of the schemes, executes a genetic algorithm to optimize based on the evaluation results, and outputs 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 baffles based on the final controllable baffle opening combination scheme, and performs controllable baffle opening correction calculations based on real-time liquid cooling flow and 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 baffles are adjusted, quantitatively evaluate the stability of liquid cooling flow distribution and chip temperature balance, and adjust subsequent genetic algorithm parameters and controllable baffle opening combination schemes.

[0009] Further, liquid cooling flow thermal energy operation data is collected, and the data undergoes time base unification, noise suppression, physical consistency verification, data structure unification, and scale standardization to obtain preprocessed liquid cooling flow thermal energy operation data. The specific steps are as follows: Liquid cooling flow thermal energy operation data is collected, 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; The liquid cooling flow thermal energy operation data is then processed using global clock synchronization and sampling phase calibration methods. The data undergoes time benchmark unification and sampling interval matching; noise suppression and disturbance compensation are performed on the liquid cooling flow thermal energy operation data through a multi-parameter filtering method that integrates frequency domain analysis and adaptive weights; a physical consistency verification mechanism is established based on the thermodynamic conservation equation and the fluid continuity equation to identify and remove abnormal data from the liquid cooling flow thermal energy operation data; data structure unification and interface format conversion are performed on the liquid cooling flow thermal energy operation data through standardized feature mapping and multi-protocol encapsulation methods; and unit conversion and feature scale standardization are performed on the liquid cooling flow thermal energy operation data through dimensional mapping and interval adaptive normalization methods.

[0010] Furthermore, based on the preprocessed liquid cooling flow thermal energy operation data, the specific steps for generating controllable baffle opening combination schemes are as follows: According to the opening range of each controllable baffle, generate a set of allocation schemes containing multiple controllable baffle opening combination schemes, each allocation scheme corresponding to a liquid cooling flow allocation method; represent each allocation scheme as a set of ordered parameter sequences according to the encoding rules of the genetic algorithm, and record the ordered parameter sequences as chromosomes for genetic operations.

[0011] Furthermore, the specific steps for evaluating the flow-heat efficiency of the scheme are as follows: Multiply the coolant density, coolant flow rate, and coolant specific heat capacity together, and 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 before and after the controllable baffle 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, and 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.

[0012] Further, the specific steps for optimizing the genetic algorithm based on the evaluation results and outputting the final controllable baffle opening combination scheme are as follows: Record the heat flow efficiency evaluation value as the genetic fitness of the corresponding chromosome in the genetic algorithm; sort all chromosomes in descending order of genetic fitness; retain the chromosomes with the top K genetic fitness rankings for the next round of genetic operations, where K is a natural number less than or equal to the total number of chromosomes; perform a crossover operation on the retained chromosomes, exchanging the controllable baffle opening parameters at corresponding positions between two chromosomes according to their parameter positions, generating a new chromosome; and perform a mutation operation on the new chromosome. The calculation involves performing numerical adjustment operations on the controllable baffle opening parameters at a specified location to form a new opening combination. The new chromosome is then merged with the allocation schemes corresponding to the retained chromosomes to form a new set of allocation schemes, which are then re-input into the heat flow efficiency evaluation process to calculate the corresponding genetic fitness. The chromosomes generated after crossover and mutation are then 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 then selected as the final controllable baffle opening combination scheme.

[0013] Furthermore, based on the final controllable baffle opening combination scheme, the actuator is driven to adjust the position of the controllable baffles, and the controllable baffle opening correction calculation is performed in conjunction with real-time liquid cooling thermal energy operation data. The specific steps for adjusting the actual opening of the actuator are as follows: The final controllable baffle opening combination scheme is received, parsed into target opening commands for each controllable baffle, and transmitted to the controllable baffle actuator, driving the actuator to adjust the controllable baffles to the target position; during the controllable baffle adjustment process, the actual opening data fed back by the controllable baffle position sensor is collected in real time and compared with the target position. The target opening is compared with the actual opening. When the absolute value of the difference between the actual opening and the target opening exceeds the deviation threshold, an opening correction calculation based on real-time liquid cooling flow thermal energy operation data is performed to obtain the controllable baffle opening correction amount. The original actual opening and the controllable baffle opening correction amount are arithmetically calculated to generate a new controllable baffle opening value, which is then sent to the controllable baffle actuator as the target opening command to drive the actuator to perform the opening adjustment action. The opening correction calculation and opening adjustment action are repeated until the absolute value of the difference between the actual opening and the target opening does not exceed the deviation threshold.

[0014] Furthermore, the specific steps for performing the opening correction calculation based on real-time liquid cooling flow thermal energy operation data are as follows: Subtract the target chip temperature from the chip temperature and divide by the target chip temperature to obtain the temperature deviation term; Subtract the target coolant flow rate from the coolant flow rate and divide by the target coolant flow rate to obtain the flow rate deviation term; Subtract the coolant inlet temperature from the coolant outlet temperature and divide by the coolant inlet temperature to obtain the temperature difference ratio term; Multiply the flow rate deviation term by the temperature difference ratio term to obtain the flow rate-temperature difference coupling term; 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 then divide by the design pressure difference before and after the controllable baffle to obtain the pressure difference deviation term; Add the temperature deviation term and the flow rate-temperature difference coupling term, and then subtract the pressure difference deviation term to obtain the controllable baffle opening correction amount.

[0015] Furthermore, after the controllable baffle is adjusted, the specific steps for constructing the operation monitoring dataset are as follows: During the operation of the liquid cooling system, continuously collect liquid cooling thermal energy operation data, actual opening data of the controllable baffle, and thermal energy efficiency evaluation values ​​for each evaluation cycle. Extract the maximum and minimum chip temperatures during operation from the chip temperature data, calculate the average chip temperature during operation, and record them in chronological order as the operation monitoring dataset.

[0016] Furthermore, 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, divide 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, subtract one, and then square it to obtain the squared term of the energy efficiency change rate of the i-th evaluation period; sum all the squared terms of energy efficiency change rates and divide them by the total number of evaluation periods to obtain the average energy efficiency fluctuation term; subtract the minimum chip temperature during operation from the maximum chip temperature during operation and divide it by the average chip temperature during operation to obtain the temperature difference ratio term; add the average energy efficiency fluctuation term and the temperature difference ratio term to obtain the liquid cooling flow thermal uniformity evaluation value.

[0017] Furthermore, the specific steps for adjusting the subsequent genetic algorithm parameters and controllable baffle opening combination scheme are as follows: Real-time comparison of the liquid cooling flow thermal equilibrium evaluation value and the equilibrium threshold. When the liquid cooling flow thermal equilibrium evaluation value is not greater than the equilibrium threshold, the existing controllable baffle opening combination scheme is maintained until the next evaluation cycle. When the liquid cooling flow thermal equilibrium evaluation 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.

[0018] Beneficial effects The present invention has the following beneficial effects: (1) The liquid cooling server flow dynamic adjustment system based on genetic algorithm encodes the opening range of each controllable baffle and uses genetic algorithm to generate and optimize a variety of controllable baffle opening combination schemes. During the operation of the liquid cooling system, the flow direction and flow distribution ratio of coolant can be dynamically adjusted according to the real-time load of the chip, heat distribution and cooling demand. This achieves priority liquid supply to the high-load chip area, so that the high heat flux area can be fully cooled, thereby quickly reducing the local temperature, avoiding hot spot accumulation, ensuring stable chip operation, and taking into account the cooling demand of the low-load area, thus achieving the optimal allocation of overall flow resources.

[0019] (2) The liquid-cooled server flow dynamic adjustment system based on genetic algorithm comprehensively considers thermodynamic parameters reflecting heat exchange capacity, such as coolant density, coolant flow rate, coolant specific heat capacity and coolant inlet and outlet temperature difference, and introduces fluid dynamic parameters reflecting flow resistance and energy consumption, such as pump power, actual pressure difference before and after the controllable baffle and design pressure difference. It establishes a comprehensive evaluation formula that integrates heat exchange efficiency and flow resistance loss, which can comprehensively reflect the advantages and disadvantages of the controllable baffle opening combination scheme in both thermal management performance and energy utilization efficiency, avoids the problem of single optimization target and one-sided evaluation in the existing technology, and provides a scientific fitness basis for genetic algorithm.

[0020] (3) The liquid cooling 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 thermal energy operation data. This correction amount integrates the chip temperature deviation term, the coolant flow temperature difference coupling term, and the pressure difference deviation term, and can reflect the comprehensive influence of temperature distribution, flow rate change, and flow resistance deviation on the cooling effect. By calculating the correction amount with the original actual opening and sending it to the actuator in real time, the deviation caused by mechanical delay, external disturbance, or change in operating conditions can be quickly compensated, ensuring that the system is always in the optimal operating state.

[0021] (4) This liquid-cooled server flow dynamic adjustment system based on genetic algorithm quantifies the stability of liquid-cooled flow distribution and chip temperature uniformity by constructing an operation monitoring dataset and calculating the liquid-cooled flow thermal balance evaluation value. When the thermal balance evaluation value exceeds the balance threshold, an adaptive optimization process is automatically triggered. Based on the current operation monitoring data, the crossover ratio, mutation ratio, and iteration termination condition of the genetic algorithm are dynamically adjusted to optimize the subsequent controllable baffle opening combination scheme generation strategy. This adaptive mechanism enables the system to actively optimize the control strategy according to changes in operating status, improve flow distribution stability, extend equipment life, and reduce the risk of performance degradation caused by uneven temperature. Attached Figure Description

[0022] Figure 1This is a structural diagram of a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm. Figure 2 A bar chart showing the descending order of chromosome flow heat efficiency assessment values; Figure 3 This is a structural diagram of the controllable baffle inside the liquid cooling system.

[0023] In the diagram: 1. Controllable baffle movable axis; 2. Controllable baffle; 3. Central processing unit; 4. Memory module; 5. RAID card; 6. Stand plate; 7. Graphics card board. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figures 1-3 This invention provides a technical solution: a liquid-cooled server flow dynamic adjustment system based on a genetic algorithm, comprising: a liquid-cooled 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. The liquid-cooled flow thermal energy data acquisition and preprocessing module is used to acquire liquid-cooled flow thermal energy operation data and perform time base unification, noise suppression, physical consistency verification, data structure unification, and scale standardization processing on the liquid-cooled flow thermal energy operation data to obtain preprocessed liquid-cooled flow thermal energy operation data. The controllable baffle opening optimization module based on a genetic algorithm is used to optimize the preprocessed liquid-cooled flow thermal energy operation data... The system generates controllable baffle opening combination schemes based on data, evaluates the thermal efficiency of the schemes, and optimizes them using a genetic algorithm based on the evaluation results, outputting the final controllable baffle opening combination scheme. The controllable baffle execution control and opening correction processing module drives the actuator to adjust the position of controllable baffle 2 based on the final controllable baffle opening combination scheme, and performs controllable baffle opening correction calculations based on real-time liquid cooling thermal energy operation data to adjust the actual opening of the actuator. The operation balance evaluation and adaptive optimization module constructs an operation monitoring dataset after the controllable baffle 2 is adjusted, quantitatively evaluates the stability of liquid cooling flow distribution and chip temperature balance, and adjusts subsequent genetic algorithm parameters and controllable baffle opening combination schemes.

[0026] Specifically, the process involves collecting liquid cooling thermal energy operation data and performing time base unification, noise suppression, physical consistency verification, data structure unification, and scale standardization on the preprocessed liquid cooling thermal energy operation data. The specific steps for obtaining preprocessed liquid cooling thermal energy operation data are as follows: The liquid cooling thermal energy operation data includes 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. The coolant mass per unit volume is measured in real time using an online density sensor. The following steps are taken to obtain the coolant density: Coolant density; specific heat capacity of the coolant through experimental calibration and by referring to a table using coolant formula parameters; chip temperature by collecting real-time temperature data from the core area of ​​the chip using distributed temperature sensors; coolant inlet temperature by measuring the coolant temperature entering the cold plate in real-time using a temperature sensor installed on the inlet pipe; coolant outlet temperature by measuring the coolant temperature leaving the cold plate in real-time using a temperature sensor installed on the outlet pipe; coolant flow rate by detecting the volume of coolant flowing through the cold plate per unit time using a flow meter; pressure difference across the controllable baffle by synchronously acquiring the instantaneous pressure difference between the front and rear sides of the controllable baffle 2 using a differential pressure sensor; controllable baffle opening by real-time detection of the current opening position of the controllable baffle actuator using a position sensor; pump power by collecting instantaneous power consumption data from the pump motor using a power sensor; target coolant flow rate by determining the optimal coolant flow rate under specific load conditions using a thermal balance algorithm; and target chip temperature by determining the optimal operating temperature under specific operating conditions through thermal reliability design calculations.Then, through global clock synchronization and sampling phase calibration methods, the liquid cooling fluid thermal energy operation data is processed to unify the time base and match the sampling interval, aligning the timestamps of each data sampling point and eliminating phase deviations between different acquisition channels. A multi-parameter filtering method incorporating frequency domain analysis and adaptive weights is used to perform noise suppression and disturbance compensation processing on the liquid cooling fluid thermal energy operation data. The noise suppression process employs adaptive cutoff frequency filtering for high-frequency random noise, while the disturbance compensation process corrects transient deviations based on the operating state characteristic curve. Finally, a physical consistency verification mechanism is established based on the thermodynamic conservation equation and the fluid continuity equation to identify and remove abnormal data in the liquid cooling fluid thermal energy operation data, ensuring... The relationship between the heat transfer power, pump power, and flow resistance parameters calculated from coolant density, coolant flow rate, coolant specific heat capacity, and temperature difference conforms to the conditions of energy conservation and flow conservation. Through standardized feature mapping and multi-protocol encapsulation methods, the liquid cooling flow thermal energy operation data undergoes data structure unification and interface format conversion, ensuring complete consistency in structure, field order, and encoding format for data from different sources and communication protocols. Furthermore, through dimensional mapping and interval adaptive normalization methods, the liquid cooling flow thermal energy operation data undergoes unit conversion and feature scale standardization, ensuring that parameters such as coolant density, coolant specific heat capacity, and coolant flow rate have directly comparable and computationally consistent scales in subsequent calculations and optimizations.

[0027] This implementation plan clarifies the specific methods for acquiring 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. This ensures that the source of liquid cooling flow thermal energy operation data is reliable and traceable during the acquisition process. This provides a solid data foundation for subsequent processes such as unified time reference, noise suppression, physical consistency verification, unified data structure, and standardized scale processing. As a result, it improves the accuracy of flow thermal energy efficiency assessment calculations and the scientific validity of the genetic algorithm-optimized controllable baffle opening combination scheme.

[0028] Specifically, based on the preprocessed liquid cooling flow thermal energy operation data, the specific steps for generating controllable baffle opening combination schemes are as follows: According to the opening range of each controllable baffle 2, combined with the structural parameters of each cooling circuit in the liquid cooling system and the flow characteristics of the coolant, a set of allocation schemes containing multiple controllable baffle opening combination schemes is generated. Each allocation scheme corresponds to a liquid cooling flow distribution method and can cover the cooling needs of high-load chip areas, low-load chip areas, and medium-load chip areas. Each allocation scheme is represented as an ordered parameter sequence according to the encoding rules of a genetic algorithm. The encoding rules of the genetic algorithm are as follows: the baffle opening of each controllable baffle 2 is quantized and discretized within the allowable opening range, and arranged sequentially according to the physical position order of the controllable baffles 2 in the cooling circuit of the liquid cooling system to form a parameter sequence, so that each position in the parameter sequence uniquely corresponds to the corresponding controllable baffle 2 in its actual installation position. The above ordered parameter sequence is recorded as the chromosome of the genetic operation, providing an input basis for subsequent flow thermal energy efficiency evaluation and genetic operation optimization.

[0029] In this implementation scheme, by combining the opening range of each controllable baffle 2, the structural parameters of each cooling circuit in the liquid cooling system, and the flow characteristics of the coolant, a set of multiple allocation schemes is generated and transformed into an ordered parameter sequence that conforms to the encoding rules of the genetic algorithm. This allows the controllable baffle opening combination scheme to be accurately mapped to the actual position of the controllable baffle 2, thereby ensuring the integrity and accuracy of the input to the optimization process in subsequent genetic operations, improving the reliability of the thermal efficiency evaluation results, and laying the foundation for realizing differentiated liquid cooling flow allocation for different load chip regions.

[0030] Specifically, the evaluation steps for the flow-thermal efficiency of the scheme are as follows: Multiply the coolant density, coolant flow rate, and coolant specific heat capacity, and then multiply by the difference between the coolant outlet temperature and the coolant inlet temperature to obtain the heat transfer power term, which reflects the coolant's ability to remove heat from the chip area per unit time; Take 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, divide by the design pressure difference before and after the controllable baffle, and add one to obtain the flow resistance deviation coefficient, which characterizes the degree of deviation of the controllable baffle 2 structure from the flow resistance; Multiply the pump power by the time unit conversion factor, and then multiply by the flow resistance deviation coefficient to obtain the energy consumption flow resistance term, which is used to quantify the additional energy consumption caused by the flow resistance deviation during coolant transportation; Divide the heat transfer power term by the energy consumption flow resistance term to obtain the flow-thermal efficiency evaluation value, which comprehensively reflects the performance of the controllable baffle opening combination scheme in terms of heat transfer efficiency and energy loss balance. Among them, 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 structural design of the pipeline and the controllable baffle 2 under the rated flow conditions of the liquid cooling system; the time unit conversion factor is a positive real number calculated by using the flow record and time record in the liquid cooling flow thermal energy operation data and the time unit conversion formula to realize the conversion between seconds, minutes or hours.

[0031] The specific formula for calculating the heat flow efficiency evaluation value is as follows: ; In the formula, This indicates the thermal efficiency evaluation value. Indicates the density of the coolant. Indicates coolant flow rate. Indicates the specific heat capacity of the coolant. Indicates the coolant outlet temperature. Indicates the coolant inlet temperature. Indicates pump power. This indicates the actual pressure difference before and after the controllable baffle. This indicates the design pressure difference before and after the controllable baffle. Indicates the conversion factor for time units.

[0032] In this embodiment, Table 1 is a data table of liquid cooling flow efficiency evaluation values, listing the liquid cooling flow efficiency operation data indicators and corresponding calculated flow efficiency evaluation values ​​for the five chromosomes during the liquid cooling flow rate allocation optimization process. Specific data are as follows: In chromosome R1, the coolant density is 1035 kg / m³, the coolant flow rate is 0.035 m³ / s, the coolant specific heat capacity is 3850 J / (kg·K), the coolant outlet temperature is 35°C, and the coolant inlet temperature is... 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.

[0033] Table 1. Data Table of Thermal Efficiency Evaluation Values

[0034] 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 2It can intuitively reflect the relative advantages and disadvantages of different chromosomes in the optimization of liquid cooling flow rate allocation, and provide a reference for the screening and optimization of genetic algorithms.

[0035] In this implementation scheme, by combining the coolant density, coolant flow rate, coolant specific heat capacity, coolant outlet temperature, coolant inlet temperature, pump power, time unit conversion factor, actual pressure difference before and after the controllable baffle, and design pressure difference before and after the controllable baffle for comprehensive calculation, the contribution of heat transfer power and energy consumption flow resistance can be quantified simultaneously in the same evaluation system. This ensures 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 both thermal management performance and energy utilization efficiency, and provides an accurate, stable and physically consistent fitness reference for genetic algorithm optimization.

[0036] Specifically, the steps for optimizing the genetic algorithm based on the evaluation results and outputting the final controllable baffle opening combination scheme are as follows: The calculated thermal efficiency evaluation value is accurately recorded as the genetic fitness of the corresponding chromosome in the genetic algorithm, ensuring 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, where K is a natural number less than or equal to the total number of chromosomes and is set according to the system's operational scale and optimization accuracy requirements; crossover is performed on the retained chromosomes, exchanging the controllable baffle opening parameters one by one between two chromosomes according to a fixed parameter position order, generating offspring chromosomes that can inherit parental characteristics and possess new combination characteristics; The new chromosome undergoes mutation operations, performing amplitude-limited numerical adjustments to the controllable baffle opening parameters at designated locations to increase population diversity and avoid getting trapped in local optima. The new chromosome is then merged with the allocation schemes corresponding to the retained chromosomes, forming a new set of allocation schemes including both parent and offspring generations. This set is then re-input into the heat transfer efficiency evaluation process to calculate the corresponding genetic fitness, forming a new optimization basis. The chromosomes generated after crossover and mutation are re-decoded into controllable baffle opening combination schemes that can be directly used for execution control. The cycle of evaluation, sorting, retention, crossover, and mutation is repeated until the genetic fitness does not improve within a continuous fixed evaluation period. The allocation scheme with the highest genetic fitness is selected as the final controllable baffle opening combination scheme and enters the subsequent execution control stage. The evaluation period is the time interval during which the liquid cooling system performs one operation data acquisition, performance calculation, and control decision within a fixed time interval.

[0037] In this implementation scheme, the thermal efficiency evaluation value is precisely correlated with the genetic fitness of the genetic algorithm. Based on the retention of high-fitness chromosomes, parameter position order crossover and amplitude-limited numerical adjustment mutation operations are performed to ensure that the generated new chromosomes have optimization potential in terms of thermal management performance and energy utilization efficiency. By decoding and merging the chromosomes after crossover and mutation, and combining the fitness change trend within a continuous fixed evaluation period to determine the optimization termination condition, the final output controllable baffle opening combination scheme can approach the optimum in the global search space, thereby improving the accuracy of flow allocation driven by liquid cooling thermal energy operation data and the system heat dissipation stability.

[0038] Specifically, based on the final controllable baffle opening combination scheme, the actuator is driven to adjust the position of controllable baffle 2, and the controllable baffle opening correction calculation is performed in conjunction with real-time liquid cooling flow thermal energy operation data. The specific steps for adjusting the actual opening of the actuator are as follows: The final controllable baffle opening combination scheme is received and parsed into target opening instructions for each controllable baffle 2, which are then transmitted to the controllable baffle actuator via the industrial bus protocol. The actuator is driven to adjust the controllable baffle 2 to the target position according to the target opening instructions. 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 one by one with the target opening stored in the controller cache. 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, utilizing chip temperature, target chip temperature, coolant flow rate, target coolant flow rate, and coolant outlet... The controllable baffle opening correction amount is calculated based on parameters such as inlet temperature, coolant inlet temperature, actual pressure difference before and after the controllable baffle, and design pressure difference before and after the controllable baffle. The deviation threshold is the maximum absolute deviation range between the allowable actual opening and the target opening, determined based on the heat dissipation performance requirements and control accuracy requirements of the liquid cooling system. The original actual opening and the calculated controllable baffle opening correction amount are arithmetically calculated to generate a new controllable baffle opening value. This new value is then sent as the target opening command to the controllable baffle actuator, driving it to perform a fine-tuned opening adjustment. This cycle of opening correction calculation and adjustment is repeated until the absolute value of the difference between the actual opening and the corresponding target opening of all controllable baffles 2 does not exceed the deviation threshold, thus ensuring that the controllable baffle opening accuracy meets the dynamic control requirements driven by liquid cooling thermal energy operation data.

[0039] like Figure 3As shown, the controllable baffle movable shaft 1 is used to rotatably connect with the upright plate 6 to support the controllable baffle 2 to achieve angle adjustment. Under the drive of the controllable baffle actuator, the controllable baffle 2 changes the flow path of the coolant in different hardware areas and adjusts the flow distribution ratio to achieve dynamic distribution of liquid cooling flow in conjunction with the controllable baffle opening combination scheme generated based on the genetic algorithm. The back of the central processing unit 3 is equipped with a temperature sensor for collecting chip temperature; the memory module 4 is used for storing and high-speed reading and writing data; the RAID card 5 is used for data management and transmission of multi-hard drive arrays; the upright plate 6 is used to fix and support the various hardware components and provide a rotation mounting position for the controllable baffle 2; the graphics card board 7 is used for graphics processing and accelerated computing. During the operation of the liquid cooling system, the heat generated by the central processing unit 3, memory module 4, RAID card 5 and graphics card board 7 is carried away by the coolant. The chip temperature collected by the temperature sensor is used as part of the liquid cooling flow thermal energy operation data and input to the controllable baffle execution control and opening correction processing module to guide the real-time adjustment and optimization of the controllable baffle opening.

[0040] In this implementation scheme, the actual opening data fed back by the position sensor of the controllable baffle 2 is collected in real time during the adjustment process 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 amount of the controllable baffle is calculated using the liquid cooling thermal energy operation data. In this way, when there is mechanical delay in the actuator, changes in operating conditions, or external disturbances affecting the opening accuracy, the opening deviation can be quickly corrected to ensure that the actual opening of the controllable baffle continues to be close to the target opening, and to maintain the stability of the thermal energy distribution and the consistency of the heat dissipation performance of the liquid cooling system.

[0041] Specifically, the steps for performing the opening correction calculation based on real-time liquid cooling flow thermal energy operation data are as follows: Subtract the predetermined target chip temperature from the collected chip temperature and divide by the target chip temperature to obtain a temperature deviation term, which quantifies the relative deviation of the current actual chip operating temperature from the target chip temperature; Subtract the target coolant flow rate from the real-time collected coolant flow rate and divide by the target coolant flow rate to obtain a flow rate deviation term, which reflects the relative change in the current liquid cooling system's actual supply flow rate relative to the target coolant flow rate; Subtract the coolant inlet temperature from the real-time collected coolant outlet temperature and divide by the coolant inlet temperature to obtain the temperature difference. The proportional term characterizes the relative temperature rise of the coolant during the heat exchange process. Multiplying the flow deviation term by the temperature difference proportional term yields the flow-temperature difference coupling term, which reflects the combined effect of flow rate change and heat exchange temperature difference on cooling performance. Subtracting the design pressure difference before and after the actual pressure difference before and after the controllable baffle from the design pressure difference before and after the actual pressure difference yields the pressure difference deviation term, which measures 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, yields the controllable baffle opening correction amount, which serves as the basis for calculating the controllable baffle opening adjustment.

[0042] The specific formula for calculating the controllable baffle opening correction is as follows: ; In the formula, This indicates the correction amount for the controllable baffle opening. Indicates chip temperature. Indicates the temperature of the target chip. Indicates coolant flow rate. Indicates the target coolant flow rate. Indicates the coolant outlet temperature. Indicates the coolant inlet temperature. This indicates the actual pressure difference before and after the controllable baffle. This indicates the design pressure difference before and after the controllable baffle.

[0043] In this implementation scheme, by comprehensively introducing temperature deviation, flow rate deviation, temperature difference ratio, flow rate-temperature difference coupling, and pressure difference deviation terms during the calculation process, synchronous quantitative analysis of chip temperature changes, coolant flow rate changes, coolant inlet-outlet temperature difference changes, and pressure difference changes before and after the controllable baffle is achieved. This can accurately reflect the operating deviation state of the liquid cooling system under both thermodynamic and fluid dynamic dimensions, thereby providing a scientific basis for generating controllable baffle opening correction, improving the accuracy and response efficiency of controllable baffle opening adjustment, and ensuring that the liquid cooling system maintains efficient and stable cooling performance under different operating conditions.

[0044] Specifically, after the controllable baffle 2 is adjusted, the specific steps for constructing the operation monitoring dataset are as follows: During the operation of the liquid cooling system, continuously collect liquid cooling thermal energy operation data, actual opening data of the controllable baffle, and thermal energy efficiency evaluation values ​​for each evaluation cycle, and ensure that the timestamps of various types of data are consistent through a time synchronization mechanism; extract the maximum and minimum chip temperatures during operation from the chip temperature data, ensuring that the extraction process covers the complete operating time and eliminates extreme value deviations caused by instantaneous sensor anomalies, and perform statistical calculations on the chip temperature sequence to obtain the average chip temperature during operation; record all the above data in chronological order as the operation monitoring dataset.

[0045] In this implementation scheme, by continuously collecting liquid cooling flow thermal energy operation data, actual controllable baffle opening data, and flow thermal efficiency evaluation values ​​for each evaluation period during the operation of the liquid cooling system, and using a time synchronization mechanism to ensure the consistency of timestamps for different types of data, precise alignment of multi-source operation data is achieved. When extracting the maximum and minimum chip temperatures during operation, it is ensured that the entire operation duration is covered and extreme value deviations caused by instantaneous sensor anomalies are eliminated. At the same time, statistical calculations are performed on the chip temperature sequence to obtain the average chip temperature during operation, ensuring the accuracy and representativeness of temperature characteristic data. Finally, all operation data are recorded in chronological order as an operation monitoring dataset, providing a high-precision, time-consistent, and complete data foundation for subsequent quantitative evaluation of the stability of liquid cooling flow distribution and chip temperature uniformity.

[0046] Specifically, the steps for quantitatively evaluating the stability of liquid cooling flow distribution and chip temperature uniformity are as follows: Based on the operational monitoring dataset, divide the flow thermal efficiency evaluation value of the i-th evaluation period by the flow thermal efficiency evaluation value of the previous evaluation period, subtract one, and then square the result to obtain the squared term of the energy efficiency change rate for the i-th evaluation period. The selection of the evaluation period must ensure coverage of the stable operation phase and avoid fluctuations during startup or shutdown. Sum all squared terms of energy efficiency change rate and divide by the total number of evaluation periods to obtain the average energy efficiency fluctuation term. The average energy efficiency fluctuation term can quantitatively reflect the stability level of the liquid cooling flow thermal energy operation data in different periods. Subtract the minimum chip temperature during operation from the maximum chip temperature during operation and divide by the average chip temperature during operation to obtain the temperature difference ratio term. During the calculation, it is necessary to ensure that the temperature data comes from the chip temperature sequence after time synchronization and exclude instantaneous outliers. Add the average energy efficiency fluctuation term and the temperature difference ratio term to obtain the liquid cooling flow thermal uniformity evaluation value.

[0047] The specific formula for calculating the liquid cooling flow rate thermal equilibrium assessment value is as follows: ; In the formula, This represents the liquid cooling flow rate as an assessment value for thermal equilibrium. This represents the thermal efficiency evaluation value for the i-th evaluation period. This represents the thermal efficiency evaluation value for the (i-1)th evaluation period, where n represents the total number of evaluation periods. Indicates the maximum chip temperature during operation. Indicates the minimum chip temperature during operation. This represents the average chip temperature during operation.

[0048] In this implementation scheme, by calculating the average energy efficiency fluctuation term and the temperature difference ratio term based on the operation monitoring dataset, and merging the two to obtain the liquid cooling flow thermal balance evaluation value, an integrated quantitative evaluation of the stability of liquid cooling flow distribution and chip temperature balance is achieved. This can provide an accurate characterization of the state changes of liquid cooling flow thermal energy operation data in the long-term operation process, thereby providing a scientific and reliable basis for the adaptive optimization process and improving the stability and balance of liquid cooling system operation.

[0049] Specifically, the steps for adjusting the subsequent genetic algorithm parameters and the controllable baffle opening combination scheme are as follows: Real-time comparison of the liquid cooling flow thermal equilibrium evaluation value and the equilibrium threshold; ensuring the accuracy of the comparison through a periodic operation monitoring mechanism; when the liquid cooling flow thermal equilibrium evaluation value is not greater than the equilibrium threshold, the currently executed controllable baffle opening combination scheme is directly maintained until the next evaluation cycle, and liquid cooling flow thermal energy operation data is continuously collected during the evaluation cycle for subsequent analysis; when the liquid cooling flow thermal equilibrium evaluation value is greater than the equilibrium threshold, an adaptive optimization process is immediately triggered. Based on all variables included in the current operation monitoring data, such as coolant density, coolant flow rate, coolant specific heat capacity, chip temperature, coolant inlet temperature, coolant outlet temperature, actual pressure difference before and after the controllable baffle, design pressure difference before and after the controllable baffle, pump power, and controllable baffle opening, the crossover ratio, mutation ratio, and iteration termination condition of the genetic algorithm are dynamically adjusted to generate a new controllable baffle opening combination scheme that conforms to the current operating state. The combination scheme is then directly output to the controllable baffle execution control and opening correction processing module for execution.

[0050] In this implementation scheme, a periodic operation monitoring mechanism is introduced into the real-time comparison of the liquid cooling flow thermal equilibrium assessment value and the equilibrium threshold to ensure the accuracy of the comparison results. When the liquid cooling flow thermal equilibrium assessment value is greater than the equilibrium threshold, the genetic algorithm dynamically adjusts the crossover ratio, mutation ratio, and iteration termination condition by combining multi-dimensional parameters including coolant density, coolant flow rate, coolant specific heat capacity, chip temperature, coolant inlet temperature, coolant outlet temperature, actual pressure difference before and after the controllable baffle, design pressure difference before and after the controllable baffle, pump power, and controllable baffle opening. This generates a controllable baffle opening combination scheme that highly matches the current operating state and issues it for execution in a timely manner, thereby achieving rapid optimization of the flow distribution strategy and improving the stability of system operation.

[0051] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0052] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention 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