A multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method

Abstract

The application discloses a kind of multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method, relating to liquid cooling energy efficiency optimization technical field.The multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method include: S1, acquisition and preprocessing two-phase liquid cooling control data;S2, periodically assesses each regional heat load level, constructs discrete time linear state space model, extracts heat load change rate;S3, periodically quantifies the cold quantity distribution of each region, calculates pump speed and valve opening target value, and generates adjustment control instruction;S4, identify high heat risk and low heat stable area, assess the demand for cold compensation, output cold quantity enhancement and recovery distribution value, and generate corresponding control instruction.The existing two-phase liquid cooling system generally only relies on temperature feedback for adjustment, resulting in cold quantity distribution response lag, high heat area prone to short-term overheating phenomenon, while low heat area exists simultaneously cold quantity waste problem.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of liquid cooling energy efficiency optimization, and particularly relates to a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method. BACKGROUND

[0002] With the rapid development of high-performance computing, big data and artificial intelligence and the like, the server load of a data center is significantly improved, the heat flow density is increasingly increased, and higher requirements are put forward for the heat dissipation efficiency and energy efficiency control capability of a cooling system. The traditional air cooling technology is limited by the heat conduction efficiency and spatial layout, and has been difficult to meet the heat dissipation demand of modern high-density servers. The liquid cooling technology, especially the two-phase liquid cooling, is widely regarded as the next generation of high-efficiency heat management scheme due to the use of the latent heat transfer capability brought by the evaporation of the cooling liquid.

[0003] For example, the application with the announcement number CN114126365B discloses a data center high-energy-efficiency liquid cooling method and system, which comprises the following steps: determining a key heat generation parameter x1 of a computing hardware causing heat generation and a key heat generation parameter x2 of a memory hardware causing heat generation; obtaining water temperature T, flow rate v and computing hardware temperature T' of the key heat generation parameter x1 at different values, obtaining a first data set of the computing hardware and performing fitting to obtain a computing hardware temperature function T'=F1(T, v, x1); obtaining a memory hardware temperature change rate δ of the current hardware temperature Td, water temperature T and the key heat generation parameter x2 at different values, obtaining a second data set of the memory hardware and performing fitting to obtain a memory hardware temperature change rate function δ=F2(T, Td, x2); and solving the target water temperature and flow rate to be controlled according to the safety temperature of each hardware. According to the heat dissipation demand of each hardware, the target water temperature and flow rate are adjusted to minimize the refrigeration energy consumption under the premise of meeting the heat dissipation demand of each hardware.

[0004] For example, the application with the announcement number CN114390859B discloses a liquid cooling server cabinet system and a liquid cooling server cabinet, which comprises the following: a cold water distribution pipe and a hot water collection pipe connected with a server; a refrigerant distribution unit, one end of the refrigerant distribution unit is connected with the hot water collection pipe, and the other end of the refrigerant distribution unit is connected with the cold water distribution pipe; a collection unit connected with the refrigerant distribution unit; a liquid cooling assembly connected with the collection unit; and a switch connected with the collection unit; wherein the collection unit is configured to obtain the environmental indicators of the cabinet detected by the liquid cooling assembly, and manage the refrigerant distribution unit.

[0005] However, although the above scheme has achieved certain results in improving heat dissipation efficiency or optimizing refrigerant distribution, there are still the following shortcomings: firstly, the existing method is mainly based on static adjustment parameter fitting or preset threshold, and lacks the ability of rolling prediction and feedback control facing the dynamic change of complex heat load; secondly, the cooling resource allocation is mainly based on the single-point index of the whole machine or cabinet level, and cannot fully consider the thermal coupling characteristics between multiple nodes and the load fluctuation characteristics of the regional level, so it is difficult to realize the collaborative scheduling of cold energy across regions; thirdly, the response granularity of the control strategy is relatively coarse, and it is difficult to realize the fine allocation of cold energy and the rapid compensation of local thermal disturbance at the cycle level, resulting in adjustment lag or energy consumption redundancy.

[0006] Therefore, in view of the above problems, there is an urgent need for a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method. SUMMARY

[0007] Technical problems to be solved

[0008] In view of the shortcomings of the prior art, the present application provides a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method, which solves the problem that the existing two-phase liquid cooling system generally only relies on temperature feedback for adjustment, resulting in a response lag of cold energy distribution, and the phenomenon of short-term overheating in high-heat areas, while there is also a waste of cold energy in low-heat areas.

[0009] Technical scheme

[0010] To achieve the above purpose, the present application is implemented by the following technical scheme: a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method, comprising: S1, setting a control cycle to collect two-phase liquid cooling control data, performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data; S2, based on the preprocessed two-phase liquid cooling control data, periodically evaluating the heat load level of each region, constructing a discrete time linear state space model, and outputting a heat load prediction sequence; comparing the prediction sequence with the current heat load state, extracting the heat load change rate, and generating a cold energy distribution input data set; S3, based on the cold energy distribution input data set, periodically quantifying the cold energy distribution amount of each region, combining the pump type characteristics and flow resistance parameters to construct a control amount conversion relationship, calculating the pump speed and valve opening target value, and generating an adjustment control instruction to realize partitioned cold energy allocation control; S4, analyze the heat load prediction trend, identify high-heat risk and low-heat stable regions, evaluate the cold energy compensation demand, output the cold energy enhancement and recovery distribution value, and generate the corresponding control instruction to realize the dynamic adjustment and compensation of cold energy and the rapid response of thermal disturbance.

[0011] Further, the specific steps of collecting two-phase liquid cooling control data in a control period, performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data are as follows: a control period of a fixed time window is designed, two-phase liquid cooling control data is collected, and the two-phase liquid cooling control data includes server power consumption, cooling liquid inlet temperature, cooling liquid outlet temperature, vapor flow rate, cooling liquid mass flow rate, cooling liquid specific heat capacity, and cooling liquid density; through constructing a time sequence synchronization method based on multi-channel time stamp alignment and sampling interval reconstruction, cross-module time alignment and data frame unified processing are performed on the two-phase liquid cooling control data; through introducing a window sliding mean analysis and a first-order derivative mutation recognition edge disturbance judgment algorithm, short-period fluctuation detection and high-frequency noise removal processing are performed on the two-phase liquid cooling control data; by establishing a physical consistency criterion based on interval change range and logical boundary condition, abnormal offset identification and dynamic correction compensation processing are performed on the two-phase liquid cooling control data; a unit conversion matrix for physical quantity unification and an input field mapping rule are constructed, and the two-phase liquid cooling control data is standardized and normalized.

[0012] Further, based on the preprocessed two-phase liquid cooling control data, the specific steps of periodically evaluating the thermal load level of each region are as follows: based on the server structure and the cooling path, the control components with independent monitoring and control capabilities are defined as control nodes, and the control nodes are divided into regions according to whether the cooling path structures are the same; based on the preprocessed two-phase liquid cooling control data, the thermal load level of each region in the control period is evaluated: the cooling liquid mass flow rate is multiplied by the cooling liquid specific heat capacity, and then multiplied by the difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature to obtain the cooling liquid sensible heat transfer term; the cooling liquid mass flow rate is multiplied by the cooling liquid specific heat capacity, and then multiplied by the first-order derivative of the cooling liquid outlet temperature with respect to time to obtain the temperature dynamic response term; the cooling liquid sensible heat transfer term, the server power consumption, and the temperature dynamic response term are sequentially added to obtain the control period thermal load evaluation value.

[0013] Further, the specific steps of constructing a discrete-time linear state-space model and outputting a heat load prediction sequence are as follows: constructing a discrete-time linear state-space model with two-phase liquid cooling control data as an input vector and a control period heat load evaluation value as a state variable; establishing a difference mapping relationship between the input vector and the state variable; performing parameter estimation on the two-phase liquid cooling control data of a fixed control period by using a least squares method to complete model parameter training; based on the trained discrete-time linear state-space model, performing multi-step rolling prediction on the heat load state variable in each control period to obtain a heat load prediction sequence of multiple future periods; performing time point corresponding comparison between the heat load state variable prediction sequence and the heat load evaluation value at the current time, constructing a heat load change slope sequence, extracting the heat load change rate between each control period, and combining the two-phase liquid cooling control data of the current control period to construct a cold quantity distribution input data set.

[0014] Further, the specific steps of periodically quantifying the cold quantity distribution amount of each region based on the cold quantity distribution input data set are as follows: multiplying the cooling liquid mass flow rate by the specific heat capacity of the cooling liquid, and then multiplying the absolute value of the first derivative of the cooling liquid outlet temperature with respect to time to obtain a cooling liquid temperature dynamic change term; multiplying the heat load change rate by the server power consumption, and dividing by the product of the cooling liquid density and the steam flow rate to obtain a load adjustment compensation term; adding the cooling liquid temperature dynamic change term and the load adjustment compensation term to obtain a basic cold quantity distribution evaluation value; dividing the difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature by the sum of the cooling liquid inlet temperature and a minimum term to obtain a temperature difference correction factor; multiplying the basic cold quantity distribution evaluation value by the result obtained by adding one to the temperature difference correction factor to obtain a cold quantity distribution amount evaluation value.

[0015] Further, the specific steps of constructing a control quantity conversion relationship combined with pump type characteristics and flow resistance parameters are as follows: taking the cold quantity distribution amount evaluation value as a target value, combining the physical response characteristics of the flow regulation component in the cooling liquid driving path and the shunt control component in the distribution path, and establishing a conversion relationship between the cold quantity adjustment output and the driving set value; wherein, a functional mapping relationship between the cooling liquid instantaneous flow rate and the pump rotation speed is constructed according to the pump type performance curve; based on the cooling circuit pressure drop and the flow resistance parameter, a functional mapping relationship between the branch valve opening and the local flow distribution ratio is constructed.

[0016] Further, the pump speed and the valve opening target value are calculated, and the adjustment control instruction is generated, and the specific steps of realizing the partition cold quantity deployment control are as follows: the pump speed adjustment target value and the valve opening adjustment target value corresponding to the current cold quantity distribution evaluation value are calculated with the conversion relationship as the constraint condition; the pump speed adjustment target value is taken as the instruction generation basis to construct the pump speed control instruction; the valve opening adjustment target value is written into the opening instruction structure according to the branch number sequence to generate the valve opening control instruction; the pump speed control instruction and the valve opening control instruction are written into the pump control buffer and the valve control buffer respectively, and the control parameter refreshing and the adjustment instruction issuing are completed according to the current control cycle time base; under the driving of the valve opening control instruction, the nozzle control unit is activated in linkage, the nozzle performs the directional spraying operation according to the branch flow configuration corresponding to the current cycle valve opening, the two-phase cooling liquid is supplemented to the server peripheral heat dissipation area, and the partition cold quantity enhanced flow field is formed.

[0017] Further, the specific steps of analyzing the heat load prediction trend, identifying the high heat risk and low heat stable area, and evaluating the cold quantity compensation demand are as follows: the heat load state variable prediction sequence of each area is extracted, and the heat load change amount per unit time is calculated in combination with the heat load change rate sequence of the historical control cycle; if the heat load change amount is positive, and the heat load change amount of the last three control cycles exceeds the load increment threshold, the corresponding area is marked as a high heat risk area; if the heat load change amount is negative, and the heat load change amount of the last three control cycles exceeds the load attenuation threshold, the corresponding area is marked as a low heat stable area; according to the heat disturbance risk area of the heat load change amount of each node, in combination with the two-phase liquid cooling control data, the cold quantity compensation demand of each area in the next control cycle is evaluated: the absolute value of the second derivative of the absolute value of the heat load change amount with respect to time is taken, to obtain the heat disturbance acceleration term; the absolute value of the second derivative of the cooling liquid outlet temperature with respect to time is taken, and multiplied by the heat disturbance adjustment coefficient, to obtain the temperature disturbance acceleration term; the heat disturbance acceleration term and the temperature disturbance acceleration term are added, to obtain the heat temperature disturbance response base value; the cooling liquid mass flow rate is multiplied by the specific heat capacity of the cooling liquid, and then divided by the product of the cooling liquid density and the steam flow rate, to obtain the cooling liquid adjustment amount corresponding to a unit disturbance; the cooling liquid outlet temperature standard deviation is divided by the sum of the cooling liquid outlet temperature average value and the minimum term, and the ratio is added to the constant one, to obtain the temperature fluctuation correction factor; the heat temperature disturbance response base value, the cooling liquid adjustment amount corresponding to a unit disturbance, and the temperature fluctuation correction factor are multiplied, to obtain the cold quantity distribution compensation evaluation value.

[0018] Further, the output cold energy enhancement and recovery distribution values are generated, and corresponding control instructions are generated to realize the specific steps of dynamic adjustment compensation of cold energy and rapid response of thermal disturbance as follows: the cold energy distribution compensation evaluation value of each node in the high heat risk area is superimposed on the corresponding cold energy distribution amount evaluation value to form a cold energy enhancement distribution value; the cold energy distribution amount evaluation value of each node in the low heat stable area is subtracted by the corresponding cold energy distribution compensation evaluation value to form a cold energy recovery distribution value; the cold energy enhancement distribution value and the cold energy recovery distribution value are respectively input into a conversion relationship to obtain the pump speed adjustment target value and the valve opening degree adjustment target value corresponding to each node; the pump speed adjustment target value and the valve opening degree adjustment target value are written into an opening degree instruction structure to generate corresponding pump speed adjustment instructions, branch valve opening degree instructions and nozzle partition spraying instructions; the control parameters are refreshed according to the current control period time base, and the instructions are issued to realize the rapid compensation response of cold energy in the high heat area and the recovery adjustment of cold energy in the low heat area.

[0019] The second aspect of the application provides a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system, comprising: a data acquisition preprocessing module, a heat load modeling and prediction module, a cold energy distribution control execution module and a disturbance compensation dynamic adjustment module, wherein: the data acquisition preprocessing module is used for setting a control period to acquire two-phase liquid cooling control data, and performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data; the heat load modeling and prediction module is used for periodically evaluating the heat load level of each area based on the preprocessed two-phase liquid cooling control data, constructing a discrete time linear state space model, and outputting a heat load prediction sequence; comparing the prediction sequence with the current heat load state, extracting the heat load change rate, and generating a cold energy distribution input data set; the cold energy distribution control execution module is used for periodically quantifying the cold energy distribution amount of each area based on the cold energy distribution input data set, constructing a control amount conversion relationship combined with pump type characteristics and flow resistance parameters, calculating pump speed and valve opening degree target values, and generating adjustment control instructions to realize partitioned cold energy deployment control; the disturbance compensation dynamic adjustment module is used for analyzing the heat load prediction trend, identifying high heat risk and low heat stable areas, evaluating cold energy compensation demand, outputting cold energy enhancement and recovery distribution values, and generating corresponding control instructions to realize dynamic adjustment compensation of cold energy and rapid response of thermal disturbance.

[0020] Advantages

[0021] The application has the following advantages:

[0022] (1) The multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method can identify the trend of thermal load changes in the future control period in advance by constructing a thermal load dynamic modeling method based on a discrete-time linear state space model and combining a multi-cycle thermal load rolling prediction mechanism. Compared with the traditional adjustment method that only relies on chip or cooling liquid temperature feedback, it has stronger state prediction ability and regulation advance, realizes the transformation of cooling capacity distribution from passive response to active prediction, significantly shortens the response delay time in high heat load area, and effectively prevents local overheating.

[0023] (2) The multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method dynamically quantifies the thermal load fluctuation by a second derivative calculation method, further combines the disturbance adjustment coefficient and the temperature fluctuation correction factor, generates a quantifiable cooling capacity distribution compensation evaluation value, and is used for pre-response regulation of high heat risk area. This mechanism actively improves the cooling capacity supply level before the actual temperature rise occurs, realizes the advance configuration and rapid response of partition cooling capacity, and effectively suppresses the risk of local thermal runaway of chips or components.

[0024] (3) The multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method can reduce the cooling capacity distribution target automatically by identifying the cooling redundancy and constructing a cooling capacity recovery regulation strategy to release liquid cooling resources for high load area. Combined with the functional mapping relationship of pump speed and valve opening, the local flow and cooling intensity are precisely adjusted, so as to maximize the reduction of unnecessary energy consumption under the premise of ensuring system thermal safety, and improve the overall cooling capacity utilization efficiency and energy efficiency ratio of two-phase liquid cooling system.

[0025] (4) The multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method can identify the trend and rate of thermal disturbance by predicting the sequence of thermal load state variables and historical change rate data, dynamically evaluate the thermal disturbance response base value combined with the fluctuation characteristics of the cooling liquid outlet temperature, and construct a regulation target function with cooling liquid flow as the constraint to form an implementable cooling capacity compensation value. This scheme has good thermal disturbance decoupling ability in a multi-region, multi-node parallel working environment, significantly improves the sensing accuracy and regulation sensitivity of the two-phase liquid cooling system to sudden load changes and local thermal shock. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 It is a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control method flow chart;

[0027] Figure 2 It is a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system structure diagram;

[0028] Figure 3 It is a distribution diagram of cooling capacity distribution evaluation value of each region;

[0029] Figure 4 A multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system schematic diagram.

[0030] In the figure, 1, a cooling liquid reservoir; 2, a cooling tower; 3, a CDU distribution device; 4, a pump; 5, a dryness sensor; 6, a temperature sensor; 7, a server; 8, a nozzle; 9, an electrically controlled valve; 10, a pipeline. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying 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 of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0032] Please refer to Figures 1-4 The embodiments of the present application provide a technical solution: a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system and method, comprising: S1, setting a control cycle to collect two-phase liquid cooling control data, performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data; S2, based on the pre-processed two-phase liquid cooling control data, periodically evaluating the thermal load level of each region, constructing a discrete time linear state space model, and outputting a thermal load prediction sequence; comparing the prediction sequence with the current thermal load state, extracting the thermal load change rate, and generating a cold quantity distribution input data set; S3, based on the cold quantity distribution input data set, periodically quantifying the cold quantity distribution amount of each region, combining the pump type characteristics and flow resistance parameters to construct a control amount conversion relationship, calculating the pump speed and valve opening target value, and generating adjustment control instructions to realize partitioned cold quantity deployment control; S4, analyzing the thermal load prediction trend, identifying high thermal risk and low thermal stable regions, evaluating the cold quantity compensation demand, outputting the cold quantity enhancement and recovery distribution value, and generating the corresponding control instructions to realize dynamic adjustment compensation and rapid response of thermal disturbance of cold quantity.

[0033] Specifically, the two-phase liquid cooling control data is collected in a control cycle, and the specific steps of time alignment, noise removal, abnormal correction and standard normalization processing of the two-phase liquid cooling control data are as follows: a fixed time window control cycle is designed, a synchronous collection signal is sent by the central coordinator, and the data sampling operation of each measuring point is started at the same starting time to collect the two-phase liquid cooling control data; the two-phase liquid cooling control data includes server power consumption, cooling liquid inlet temperature, cooling liquid outlet temperature, steam flow rate, cooling liquid mass flow rate, cooling liquid specific heat capacity, cooling liquid density, cooling liquid inlet temperature and cooling liquid outlet temperature are obtained by temperature sensor 6 arranged at the inlet and outlet positions of the server 7 cooling channel respectively, the steam flow rate is continuously measured by the dryness sensor 5 installed at the outlet port of the CDU distribution device, the cooling liquid mass flow rate is obtained by joint calculation of the data path between the flow meter and the pump 4 arranged on the flow regulation path, and the cooling liquid specific heat capacity and the cooling liquid density are matched and called in real time by the preset cooling liquid physical property parameter library. Through the construction of a time sequence synchronization method based on multi-channel time stamp alignment and sampling interval reconstruction, the two-phase liquid cooling control data is processed across the path time alignment and data frame unification, wherein the time stamp alignment mechanism accurately aligns the sampling time of each collection port based on the clock driving signal of the central coordinator, the sampling interval reconstruction process converts non-equidistant sampling data into continuous and unified data frame structure through interpolation restoration and pseudo-synchronization strategy, and ensures that all cooling parameter data collected from the server 7 mainboard, pump 4 body driving section, CDU distribution device 3, electric control valve 9 execution port and central coordinator input interface can be called synchronously in the control cycle. Through the introduction of edge disturbance judgment algorithm of window sliding mean analysis and first-order derivative mutation recognition, the variables directly related to the thermal flow fluctuation characteristics in the two-phase liquid cooling control data are detected and processed, and the data dimensions such as cooling liquid inlet temperature, cooling liquid outlet temperature and chip temperature are mainly processed. The sliding window mechanism is used to capture the temperature fluctuation critical point, and the derivative change gradient is combined to detect the local disturbance of the cooling channel, so as to realize the dynamic smoothing and denoising enhancement of the non-continuous heat flux change signal caused by the short-time load surge of the server 7. Through the establishment of physical consistency criterion based on interval change range and logical boundary condition, the parameters related to the cooling liquid flow state in the two-phase liquid cooling control data are identified and dynamically corrected and compensated, including the cooling liquid mass flow rate of the pump 4 outlet section, the cooling liquid density near the nozzle 8, and the steam flow rate at the outlet of the CDU distribution device 3, etc. Through setting the flow parameter boundary threshold value and combining the coupling relationship between the cooling liquid heat capacity parameters, the rule deviation judgment and dynamic callback correction are performed, so as to ensure that each measurement data maintains coordination in physical sense, and maintains the real-time closed-loop response between the cold energy output of the nozzle 8, the chip heat load of the server 7 and the cooling liquid flow state.A unit conversion matrix and input field mapping rule for physical quantity unification are constructed, and the physical units of two-phase liquid cooling control data are standardized and normalized, including the fields of cooling liquid density, cooling liquid specific heat capacity, cooling liquid inlet temperature, cooling liquid outlet temperature, and vapor flow rate. The physical quantity conversion is completed by setting a unified unit reference dimension, and the field remapping and normalization are implemented by combining the numbering rule of the input field and the sampling partition attribution relationship, thereby providing input standard unified data support for the central coordinator to perform the multi-parameter fusion MPC calculation process.

[0034] In the embodiment, by constructing a multi-channel synchronous acquisition mechanism based on the central coordinator driving, the cooling liquid inlet temperature, cooling liquid outlet temperature, cooling liquid density, cooling liquid specific heat capacity, vapor flow rate, and cooling liquid mass flow rate are ensured to complete cross-path time alignment and timing alignment within a unified control period, thereby improving the timeliness and structural consistency of the cold energy scheduling input data. The window sliding mean analysis and first-order derivative mutation recognition algorithm are introduced, which effectively enhances the short-period disturbance detection capability of the cooling liquid temperature and the chip temperature, and improves the anti-interference stability of the thermal disturbance response. By constructing a physical consistency criterion based on the logical boundary condition and the interval change range, a dynamic linkage correction mechanism among the cooling liquid flow rate, density, and vapor flow rate is realized, thereby ensuring the coordination and reliability of the cooling channel thermal-flow data. The unit conversion matrix and field mapping rule are used to complete the physical standard unification and feature normalization of the input parameters, thereby improving the model recognition accuracy of the multi-parameter fusion MPC control strategy for the input variables, and providing a high-quality data basis for the fine and dynamic adjustment response of the cold energy distribution control.

[0035] Specifically, based on the pre-processed two-phase liquid cooling control data, the specific steps of periodically evaluating the heat load level of each region are as follows: based on the structure of the server 7 and the cooling path, the cooling liquid flow passage between the server 7 motherboard and the CDU distribution device 3 is determined, the control component with independent monitoring and control capability is defined as a control node, the control node includes a server power consumption acquisition unit, a chip temperature acquisition unit, a cooling liquid inlet temperature sensor, a cooling liquid outlet temperature sensor, a flow detection unit and a cooling liquid mass flow monitoring device, and the control nodes are divided into regions according to whether the cooling path structure is the same, and the geographical topology and path topology information of each control node are recorded by the central coordinator; based on the pre-processed two-phase liquid cooling control data, the heat load level of each region in the control period is evaluated: the cooling liquid mass flow and the specific heat capacity of the cooling liquid are multiplied, and then multiplied by the difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature, to obtain the cooling liquid sensible heat transfer term, which reflects the heat transfer capacity of the cooling liquid in the control period; the cooling liquid mass flow and the specific heat capacity of the cooling liquid are multiplied, and multiplied by the first derivative of the cooling liquid outlet temperature with respect to time, to obtain the temperature dynamic response term, which is used to quantify the influence of the rate of change of the cooling liquid outlet temperature with respect to time on the heat load state; the cooling liquid sensible heat transfer term, the server power consumption and the temperature dynamic response term are added in turn to obtain the control period heat load evaluation value, which is used as the input reference value of the subsequent discrete time linear state space modeling and MPC regulation model, to support the accurate generation of cooling quantity allocation instructions and the dynamic linkage of spray execution rhythm.

[0036] wherein the specific calculation formula of the control period heat load evaluation value is:

[0037] ;

[0038] In the formula, represents the control period heat load evaluation value, represents the cooling liquid mass flow, represents the specific heat capacity of the cooling liquid, represents the cooling liquid inlet temperature, represents the cooling liquid outlet temperature, represents the server power consumption.

[0039] In the embodiment, by fusing and calculating multiple physical quantities such as the cooling liquid mass flow rate of each region in the control period, the specific heat capacity of the cooling liquid, the cooling liquid inlet temperature, the cooling liquid outlet temperature, the first derivative of the cooling liquid outlet temperature with respect to time, and the server power consumption, the sensible heat transfer term of the cooling liquid, the temperature dynamic response term, and the control period heat load evaluation value are quantitatively obtained, which can accurately reflect the real change state of the current heat load level of each control node, improve the timeliness and accuracy of heat load modeling, enhance the dynamic adaptability of the heat load prediction model, provide high-credibility input support for subsequent MPC regulation based on the discrete-time linear state space, and effectively improve the cold energy allocation accuracy and energy efficiency control accuracy of the two-phase liquid cooling system under different regions.

[0040] Specifically, a discrete-time linear state space model is constructed to output a heat load prediction sequence; the prediction sequence is compared with the current heat load state, the heat load change rate is extracted, and the specific steps of generating a cold energy allocation input data set are as follows: the input vector is constructed by using the multi-dimensional time sequence variables such as the server power consumption, the cooling liquid inlet temperature, the cooling liquid outlet temperature, the vapor flow rate, the cooling liquid mass flow rate, the cooling liquid specific heat capacity, and the cooling liquid density collected in the two-phase liquid cooling control data; the control period heat load evaluation value obtained by fusing the sensible heat transfer term of the cooling liquid, the temperature dynamic response term, and the server power consumption is set as the state variable, and a discrete-time linear state space model with the control period as the time base is constructed; in the model construction process, based on the control dependency relationship between the input vector and the state variable, a difference mapping structure based on the state increment is established, and the least square method algorithm is used to identify and train the parameters of the historical two-phase liquid cooling control data of the fixed control period, so as to improve the stability and accuracy of the state prediction; based on the trained discrete-time linear state space model, the multi-step rolling prediction of the heat load state variable is performed in each control period, and the heat load prediction sequence corresponding to multiple future control periods is output; by aligning the heat load state variable prediction sequence with the actual observed heat load evaluation value at the current time point, the heat load change slope sequence reflecting the heat load change trend is constructed, and the heat load change rate in each control period is calculated; combined with the liquid cooling control data such as the server power consumption, the cooling liquid mass flow rate, and the cooling liquid specific heat capacity collected in the current control period, a cold energy allocation input data set for supporting the optimization of cold energy allocation decision is generated.

[0041] In this embodiment, by constructing a discrete-time linear state-space model with server power consumption, cooling liquid inlet temperature, cooling liquid outlet temperature, steam flow rate, cooling liquid mass flow rate, cooling liquid specific heat capacity, and cooling liquid density as input vectors, the least squares method is used to train the parameters of the two-phase liquid cooling control data in the historical control period, forming a rolling prediction structure that can dynamically reflect the trend of the thermal load state variable changes, and realizing the accurate prediction of the thermal load state of multiple future control periods; combined with the current control period thermal load evaluation value and the prediction sequence, the thermal load change rate is extracted and the cold distribution input data set is constructed, which significantly improves the real-time, trend and dynamic of thermal load evaluation, and provides a high reliability and strong forward-looking data foundation support for subsequent cold distribution evaluation and partition adjustment control.

[0042] Specifically, based on the cold distribution input data set, the specific steps of quantifying the cold distribution amount of each region in the period are as follows: multiply the cooling liquid mass flow rate by the cooling liquid specific heat capacity, and then multiply the absolute value of the first derivative of the cooling liquid outlet temperature with respect to time, to obtain the cooling liquid temperature dynamic change term, which is used to describe the temperature response strength of the cooling liquid under the action of the thermal load per unit time, and reflects the local dynamic heat exchange capacity caused by thermal inertia; multiply the thermal load change rate by the server power consumption, and divide by the product of the cooling liquid density and the steam flow rate, to obtain the load adjustment compensation term, which is used to represent the dynamic compensation demand of the cold supply to the thermal load disturbance, and embodies the influence of the strength change of the energy flow disturbance on the cold configuration decision; add the cooling liquid temperature dynamic change term and the load adjustment compensation term to obtain the basic cold distribution evaluation value, which is used as the preliminary response index of the cold demand; divide the difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature by the sum of the cooling liquid inlet temperature and the minimum term, to obtain the temperature difference correction factor, which is used to quantitatively adjust the nonlinear difference of the heat driving ability in different branch cooling paths; wherein, the minimum term is a minimum positive constant used to prevent the denominator from being zero or close to zero, which is used to enhance the numerical stability and robustness of the model calculation, and ensure that the control logic can also operate stably under extreme conditions; multiply the basic cold distribution evaluation value by the result of the temperature difference correction factor plus one, to obtain the cold distribution amount evaluation value, which realizes the comprehensive quantification of the current period cold demand level of each control node, and provides the target basis for the subsequent cold adjustment instruction.

[0043] wherein, the specific calculation formula of the cold distribution amount evaluation value is:

[0044] ;

[0045] In the formula, represents the cold distribution amount evaluation value, represents the cooling liquid mass flow rate, represents the cooling liquid specific heat capacity, represents the cooling liquid inlet temperature, represents a cooling liquid outlet temperature, represents a heat load change rate, represents a server power consumption, represents a cooling liquid density, represents a vapor flow rate, represents a minimum term.

[0046] In the present embodiment, Table 1 is a cooling capacity distribution evaluation value data table, which lists the cooling liquid mass flow rate, the cooling liquid specific heat capacity, the heat load change rate, the server power consumption, the cooling liquid density, the vapor flow rate, the cooling liquid outlet temperature, the cooling liquid inlet temperature, and the cooling capacity distribution evaluation value calculated based on the above variables for five regions in the current control period. The specific data is explained as follows: in region 1, the cooling liquid mass flow rate is 1.5, the cooling liquid specific heat capacity is 4.2, the heat load change rate is 0.15, the server power consumption is 180, the cooling liquid density is 997, the vapor flow rate is 0.30, the cooling liquid outlet temperature is 50, the cooling liquid inlet temperature is 45, and the calculated cooling capacity distribution evaluation value is 0.42; in region 2, the cooling liquid mass flow rate is 1.8, the cooling liquid specific heat capacity is 4.2, the heat load change rate is 0.07, the server power consumption is 210, the cooling liquid density is 997, the vapor flow rate is 0.35, the cooling liquid outlet temperature is 52, the cooling liquid inlet temperature is 45, and the calculated cooling capacity distribution evaluation value is 0.72; in region 3, the cooling liquid mass flow rate is 2.0, the cooling liquid specific heat capacity is 4.2, the heat load change rate is 0.06, the server power consumption is 250, the cooling liquid density is 997, the vapor flow rate is 0.32, the cooling liquid outlet temperature is 51, the cooling liquid inlet temperature is 45, and the calculated cooling capacity distribution evaluation value is 0.68; in region 4, the cooling liquid mass flow rate is 1.6, the cooling liquid specific heat capacity is 4.2, the heat load change rate is 0.08, the server power consumption is 200, the cooling liquid density is 997, the vapor flow rate is 0.34, the cooling liquid outlet temperature is 53, the cooling liquid inlet temperature is 45, and the calculated cooling capacity distribution evaluation value is 0.71; in region 5, the cooling liquid mass flow rate is 1.7, the cooling liquid specific heat capacity is 4.2, the heat load change rate is 0.05, the server power consumption is 190, the cooling liquid density is 997, the vapor flow rate is 0.33, the cooling liquid outlet temperature is 50, the cooling liquid inlet temperature is 45, and the calculated cooling capacity distribution evaluation value is 0.48.

[0047] Table 1 Cooling capacity distribution evaluation value data table

[0048]

[0049] As Figure 3As shown, the figure shows the distribution of the cold distribution evaluation value of five different areas in the same control cycle, reflecting the actual demand degree of each area for cooling resources. As can be seen from the figure: the cold distribution evaluation value of region 2, region 3 and region 4 is relatively high, which is 0.72, 0.68 and 0.71 respectively, indicating that the cooling demand of these three regions is large. The cold distribution evaluation value of region 1 and region 5 is relatively low, which is 0.42 and 0.48 respectively, indicating that the corresponding cooling pressure is relatively small. All the cold distribution evaluation values are between 0.4 and 0.75, which is within the reasonable distribution range. Figure 3 It intuitively shows the distribution difference of cooling resources among regions, which can provide a basis for subsequent cold quantity regulation, pump speed setting and valve adjustment strategy.

[0050] In this embodiment, by introducing the cooling liquid temperature dynamic change term, the load adjustment compensation term and the temperature difference correction factor, a cold distribution evaluation mechanism is constructed for the response characteristics of thermal disturbance, which can accurately describe the coupling relationship between the dynamic heat transfer capacity of the cooling liquid and the response ability of the server power consumption in each control cycle, and improve the adaptability of the cold distribution input data set under different load change situations. The mechanism effectively couples the temperature change rate, energy flow disturbance intensity and path heat driving difference, which helps to realize the dynamic quantitative expression of the cold distribution evaluation value to the local cooling demand of the control node, provides clear, continuous and identifiable target guidance for subsequent pump speed and valve opening adjustment, and improves the cold distribution accuracy and response efficiency.

[0051] Specifically, the specific steps of establishing the conversion relationship of the control quantity combining the pump type characteristics and the flow resistance parameters are as follows: taking the cold distribution evaluation value as the target value, combining the physical response characteristics of the flow regulation components in the cooling liquid driving path and the shunt control components in the distribution path, the conversion relationship between the cold regulation output and the driving setting is established; wherein, according to the pump type performance curve, the functional mapping relationship between the instantaneous flow of the cooling liquid and the pump speed is constructed, the pump type performance curve is the standard characteristic curve provided by the manufacturer of the pump 4, which describes the flow variation law of the pump 4 under different speed and pressure conditions; at the same time, the nonlinear response expression between pump work and flow is established combining the cooling liquid density and the pressure difference before and after the pump; based on the cooling circuit pressure drop and the flow resistance parameters, the functional mapping relationship between the branch valve opening and the local flow distribution ratio is constructed, wherein the valve opening is adjusted and executed by the electric control valve 9, the branch flow distribution ratio is adjusted and executed by the central coordinator in real time, combining the cooling liquid return pressure, the steam flow rate and the valve execution delay parameter in the control cycle, the valve physical response model is further improved, so that in the actual adjustment process, the response behavior of the cooling liquid flow can accurately reflect the set cold distribution target value, supporting the high-precision conversion control strategy of pump speed adjustment and valve opening adjustment under the multi-region collaborative control.

[0052] In this embodiment, by taking the cold distribution amount evaluation value as the target value, combining the two-phase liquid cooling control data, and establishing the accurate conversion relationship between the cold adjustment output and the drive setting amount, the pump speed adjustment and the valve opening adjustment response can effectively match the actual flow state of the cooling liquid, realizing high consistency driving matching of the cold distribution target in different control periods, enhancing the cold adjustment sensitivity under multi-region collaborative cooling, improving the linkage efficiency of the cooling liquid driving path and the distribution path under dynamic load, ensuring that the setting instructions generated by the central coordinator have high precision and low deviation execution capability at the pump 4 and the electric control valve 9 response level, thereby improving the cold delivery adaptability and energy efficiency control stability of the two-phase liquid cooling system in the rapid thermal disturbance scene.

[0053] Specifically, the pump speed and valve opening target values are calculated, and the adjustment control instructions are generated, and the specific steps of the partitioned cold distribution control are as follows: taking the conversion relationship as the constraint condition, combining the cold distribution amount evaluation value, the pump type performance curve, the cooling liquid instantaneous flow, the pump speed, the flow resistance parameter, the valve opening, the cooling liquid return pressure, and the cooling liquid mass flow of each branch, the pump speed adjustment target value and the valve opening adjustment target value corresponding to the current cold distribution amount evaluation value are calculated; the pump speed adjustment target value is taken as the instruction generation basis, the pump speed control instruction is constructed based on the central coordinator control instruction structure, and synchronous binding with the pump speed feedback channel is ensured; the valve opening adjustment target value is written into the opening instruction structure according to the branch number sequence, and the physical position number corresponding to each branch is labeled, and the valve opening control instruction is generated; the opening instruction structure is a control field set constructed according to the region number mapping relationship, used to store the valve opening adjustment target values corresponding to multiple branch channels, has structured indexing and parallel scheduling capability, and supports independent issuance of valve setting instructions according to regions in the current control period; the pump speed control instruction and the valve opening control instruction are written into the pump control buffer and the valve control buffer respectively, the pump control buffer and the valve control buffer are double-channel instruction buffer areas arranged in the central coordinator, used to temporarily store the control instructions issued to the pump 4 and the electric control valve 9 in the current period, have the functions of time alignment, frame header marking and delay synchronous delivery of multiple channel setting values, and complete control parameter refreshing and adjustment instruction issuance according to the time base of the current control period, so that the pump 4 and the electric control valve 9 complete linkage response under the unified beat control of the central coordinator; under the driving of the valve opening control instruction, the nozzle control unit is activated, the nozzle control unit includes the nozzle 8 and the control channel established between the nozzle 8 and the electric control valve 9 and the central coordinator, and has precise response capability to the start-stop, direction and atomization intensity of the spray; the nozzle 8 performs directional spraying operation according to the branch flow corresponding to the current period valve opening, and quantitatively guides the two-phase cooling liquid to the server 7 peripheral heat dissipation area through high-precision atomization control, forming a targeted partitioned cold distribution enhanced flow field, and providing rapid response local cooling support for each high-heat region.

[0054] In the embodiment, by specifying the calculation process of the pump speed adjustment target value and the valve opening adjustment target value, based on the cooling capacity distribution amount evaluation value and the conversion relationship, a linkage mechanism between the pump speed control instruction and the valve opening control instruction is constructed, and the dynamic matching of the cooling liquid instantaneous flow and the branch flow in the unified control cycle is realized. By synchronously refreshing the pump control buffer and the valve control buffer according to the control beat, combined with the directional spraying operation of the nozzle 8, a high-precision, partitioned cooling capacity enhanced flow field is formed around the server 7, thereby significantly improving the accuracy, response speed and spatial resolution of the partitioned cooling capacity allocation control, and providing continuous and stable thermal management support for high-heat areas.

[0055] Specifically, the specific steps of analyzing the heat load prediction trend, identifying the high heat risk and low heat stable region, and evaluating the specific cold compensation demand are as follows: the heat load state variable prediction sequence of each region is extracted, and the heat load change amount per unit time is calculated by combining the heat load change rate sequence of the historical control period; wherein the heat load state variable is obtained based on the previous control period heat load evaluation value and the prediction model rolling output, and reflects the instantaneous heat load response level of the server 7 structure and the cooling path corresponding to the control node. If the heat load change amount is positive, and the heat load change amount of the last three control periods exceeds the load increment threshold, the central coordinator updates the state of the control node, and marks the corresponding region as a high heat risk region; if the heat load change amount is negative, and the heat load change amount of the last three control periods exceeds the load attenuation threshold, the region is marked as a low heat stable region as a candidate region for subsequent cold recovery adjustment. According to the heat disturbance risk region of the heat load change amount of each control node, combined with the two-phase liquid cooling control data, the cold compensation demand of each region in the next control period is evaluated: the absolute value of the second derivative of the absolute value of the heat load change amount with respect to time is taken, to obtain the heat disturbance acceleration term, which reflects the heat mutation trend; the absolute value of the second derivative of the cooling liquid outlet temperature with respect to time is taken, and multiplied by the heat disturbance adjustment coefficient to obtain the temperature disturbance acceleration term, which quantifies the cooling liquid temperature change rate caused by the heat load fluctuation; wherein the heat disturbance adjustment coefficient is based on the multi-period sampling data of the cooling liquid outlet temperature and the heat load change amount in the historical control period, the disturbance acceleration sequence is constructed by calculating the time second derivative respectively, and the disturbance response error minimum criterion is used to estimate the value by a fitting optimization algorithm, the value range is 0.5 to 2.0. Add the heat disturbance acceleration term and the temperature disturbance acceleration term to obtain the heat temperature disturbance response base value; multiply the cooling liquid mass flow rate and the specific heat capacity of the cooling liquid, and then divide by the product of the cooling liquid density and the steam flow rate to obtain the cooling liquid adjustment amount corresponding to each unit disturbance, which reflects the cooling liquid response capability required for each unit heat disturbance; divide the cooling liquid outlet temperature standard deviation by the sum of the cooling liquid outlet temperature average value and the minimum term, and add a constant one to obtain the temperature fluctuation correction factor; multiply the heat temperature disturbance response base value, the cooling liquid adjustment amount corresponding to each unit disturbance, and the temperature fluctuation correction factor to obtain the cold distribution compensation evaluation value, which provides an accurate basis for the subsequent revision of the pump speed and valve opening target value, and ensures that the feedforward response capability and disturbance adaptive level of the cold quantity scheduling are enhanced synchronously.

[0056] wherein the specific calculation formula of the cold distribution compensation evaluation value is:

[0057] ;

[0058] In the formula, represents the cold distribution compensation evaluation value, represents the heat load change amount, represents the cooling liquid outlet temperature, represents the cooling liquid mass flow rate, represents the cooling liquid specific heat capacity, represents the cooling liquid density, represents the vapor flow rate, represents the cooling liquid outlet temperature standard deviation, represents the cooling liquid outlet temperature average value, represents the minimum term, represents the thermal disturbance adjustment coefficient.

[0059] In the embodiment, the dynamic identification of the high thermal risk area and the low thermal stability area is realized through the thermal disturbance identification and compensation mechanism constructed based on the thermal load state variable prediction sequence and the thermal load change rate sequence, and the cold energy distribution compensation evaluation value is calculated based on the thermal disturbance acceleration term, the temperature disturbance acceleration term, the cooling liquid adjustment amount corresponding to the unit disturbance and the temperature fluctuation correction factor, so that the feedforward adjustment of the cold energy scheduling and the disturbance response are synergistically enhanced, the thermal load response accuracy and the cold energy distribution adaptability under the two-phase liquid cooling control data driving are effectively improved, and the consistency between the key physical variables such as the cooling liquid mass flow rate, the cooling liquid specific heat capacity, the cooling liquid density, the cooling liquid outlet temperature and the vapor flow rate is ensured to be regulated and controlled, and the real-time performance and the robustness of the cold energy deployment within the control period are improved.

[0060] Specifically, the output cold energy enhancement and recovery distribution values are generated, and the corresponding control instructions are generated to realize the specific steps of dynamic adjustment compensation of cold energy and rapid response of thermal disturbance as follows: the cold energy distribution compensation evaluation value of each node in the high heat risk area is superimposed on the corresponding cold energy distribution amount evaluation value to form a cold energy enhancement distribution value, so as to ensure compensation coverage under the driving of variables such as cooling liquid mass flow rate, cooling liquid specific heat capacity, heat load change rate, and cooling liquid outlet temperature standard deviation; the cold energy distribution amount evaluation value of each node in the low heat stable area is subtracted by the corresponding cold energy distribution compensation evaluation value to form a cold energy recovery distribution value, so as to avoid energy waste caused by redundant deployment of cooling liquid resources; the cold energy enhancement distribution value and the cold energy recovery distribution value are respectively input into a control amount conversion relationship, and the pump speed adjustment target value and the valve opening adjustment target value corresponding to each node are calculated according to the pump type performance curve and the cooling circuit flow resistance parameter, so as to ensure that the response characteristics of the flow regulation component and the shunt control component remain dynamically coupled; the pump speed adjustment target value and the valve opening adjustment target value are written into an opening instruction structure to construct and generate the corresponding pump speed adjustment instruction, branch valve opening instruction, and nozzle partition spraying instruction, which are written into the pump control buffer area and the valve control buffer area through the central coordinator, and the control parameters are refreshed according to the current control period time base to complete the distribution and issuance of the adjustment instruction, so as to drive the electric control valve 9, the pump 4, and the nozzle 8 to form corresponding adjustment actions, realize rapid compensation response of cold energy in the high heat area and recovery adjustment of cold energy in the low heat area, and strengthen the dynamic adjustment capability and extreme load response efficiency of the two-phase liquid cooling cold energy control link.

[0061] In the embodiment, by constructing a cold energy enhancement distribution value and a cold energy recovery distribution value calculation method with a cold energy distribution compensation evaluation value as the core, differential dynamic adjustment of cooling resources in the high heat risk area and the low heat stable area is realized, key control amounts such as cooling liquid mass flow rate, cooling liquid specific heat capacity, heat load change rate, and cooling liquid outlet temperature standard deviation are used, the pump speed adjustment target value and the valve opening adjustment target value are accurately determined in combination with the pump type performance curve and the cooling circuit flow resistance parameter, the pump speed adjustment instruction, the branch valve opening instruction, and the nozzle partition spraying instruction are constructed and generated synchronously with the control period, the adjustment response precision and the cold energy allocation efficiency are improved, and the rapid response capability and the cold energy resource utilization efficiency of the two-phase liquid cooling system under the multi-area heterogeneous heat load working condition are significantly enhanced.

[0062] As Figure 2As shown, the second aspect of the present application provides a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control system, comprising: a data acquisition preprocessing module, a heat load modeling and prediction module, a cold quantity distribution control execution module and a disturbance compensation dynamic adjustment module, wherein: the data acquisition preprocessing module is used for setting a control cycle to acquire two-phase liquid cooling control data, performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data; the heat load modeling and prediction module is used for periodically evaluating the heat load level of each region based on the preprocessed two-phase liquid cooling control data, constructing a discrete time linear state space model, and outputting a heat load prediction sequence; comparing the prediction sequence with the current heat load state, extracting the heat load change rate, and generating a cold quantity distribution input data set; the cold quantity distribution control execution module is used for periodically quantifying the cold quantity distribution amount of each region based on the cold quantity distribution input data set, constructing a control amount conversion relationship combined with pump type characteristics and flow resistance parameters, calculating pump speed and valve opening target values, and generating adjustment control instructions to realize partitioned cold quantity deployment control; the disturbance compensation dynamic adjustment module is used for analyzing heat load prediction trends, identifying high heat risk and low heat stable regions, evaluating cold quantity compensation demand, outputting cold quantity enhancement and recovery distribution values, and generating corresponding control instructions to realize dynamic adjustment compensation of cold quantity and rapid response to heat disturbance.

[0063] As Figure 4As shown, the structure and flow path layout of the two-phase immersion liquid cooling system based on the present application are shown, and the functions of each structural component are described as follows: The cooling liquid reservoir 1 is used to store the two-phase cooling liquid, providing a basis for the circulation of each stage of the system. The cooling liquid reservoir 1 is equipped with a liquid level sensor and a backflow section pressure sensor to continuously monitor the change in cooling liquid level and backflow pressure, supporting the stability of the liquid path and the dynamic adjustment of the system heat balance. The cooling tower 2 undertakes the task of heat exchange condensation, and the recovered vapor is liquefied by natural cooling or air cooling condensation, and the output cooling liquid returns to the cooling liquid reservoir 1, providing closed-loop support for the cooling liquid circulation path, enhancing the system energy efficiency and cooling capacity. The CDU distribution device 3 is located in the middle of the system, serving as the center of two-phase cooling liquid distribution and rectification, and realizing the branch guidance of the required cooling liquid flow of each regional server 7. The CDU distribution device 3 integrates a dryness sensor 5 and other detection components to assist in judging the heat exchange efficiency and phase change integrity of the system. The pump 4 is used to drive the circulation of the two-phase cooling liquid in the system, and its speed is controlled by the central coordinator adjustment strategy. According to the cooling capacity distribution evaluation value and the conversion relationship, the pump speed adjustment target value is output, and the instantaneous flow control is realized by combining the pump type performance curve, thereby supporting the dynamic scheduling demand of the heat load. The dryness sensor 5 is installed on the outlet side of the CDU distribution device 3 to detect the dryness information of the recovered vapor after the nozzle 8 sprays. The dryness sensor output signal is used to evaluate the phase change efficiency and heat exchange saturation degree in the liquid cooling system, providing key feedback for subsequent heat disturbance trend identification and cooling capacity compensation adjustment. The temperature sensor 6 is arranged at both ends of the cooling liquid inlet and outlet path to obtain the cooling liquid temperature information at the server 7. The temperature data is used to calculate the cooling capacity distribution input data set, including the sensible heat transfer term, temperature dynamic response term, and temperature difference correction factor. The server 7 is the target object to be cooled, and the heat load generated during its operation is absorbed by the two-phase cooling liquid and removed through evaporation phase change. The server 7 is provided with a power consumption acquisition unit and a chip temperature monitoring unit to provide calculation basis for heat load evaluation and heat disturbance identification. The nozzle 8 is installed above the heat dissipation area of the server 7 and is controlled by the electrically controlled valve 9 to perform directional spraying. The nozzle 8 forms a local cooling capacity enhancement flow field that matches the current cooling capacity distribution evaluation value in each control period, realizing quantitative liquid supplement and efficient heat exchange in high-heat areas. The electrically controlled valve 9 is arranged on the branch path between the CDU distribution device 3 and the nozzle 8, and is used to dynamically adjust the cooling liquid partition flow. The electrically controlled valve opening adjustment target value is calculated from the cooling capacity distribution input data set and the control quantity conversion relationship, and is used to form the branch valve opening control instruction, ensuring that the cooling capacity is distributed as needed in each region.Pipeline 10: build a closed loop flow path of two-phase cooling liquid, connect the cooling liquid reservoir 1, pump 4, CDU distribution device 3, server 7, cooling tower 2 and nozzle 8, realize the functions of driving liquid supply, hot area spraying, steam recovery, etc., support the stable operation of the system heat-flow-pressure coupling path.

[0064] In this embodiment, by building a multi-parameter MPC two-phase liquid cooling energy efficiency optimization distribution control structure including data acquisition preprocessing module, thermal load modeling prediction module, cold distribution control execution module and disturbance compensation dynamic adjustment module, around the key variables of cooling liquid mass flow, cooling liquid specific heat capacity, cooling liquid inlet temperature, cooling liquid outlet temperature, steam flow rate, server power consumption and chip temperature, the system realizes unified collection and high-quality preprocessing of two-phase liquid cooling control data, builds a discrete-time linear state space model to predict and analyze the thermal load state variables, and extracts the thermal load change rate to generate cold distribution input data set, further builds the conversion relationship through pump type performance curve and flow resistance parameter, accurately calculates the pump speed adjustment target value and valve opening adjustment target value, and generates adjustment instructions in the control period; On this basis, dynamically identify high heat risk areas and low heat stable areas, output cold energy enhancement and recovery distribution values, and timely issue cold energy compensation control instructions to realize the whole process dynamic response of cold energy deployment behavior, significantly improve the energy efficiency distribution level and heat disturbance adaptability of the two-phase liquid cooling system.

[0065] It should be noted that, in this text, 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 that there is any such actual relationship or order between these entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0066] The preferred embodiments of the application disclosed above are only used to help explain the application. The preferred embodiments do not describe all the details and do not limit the application to the specific embodiments described. Obviously, many modifications and changes can be made according to the content of the specification. The specification selects and describes these embodiments in order to better explain the principles and practical applications of the application, so that those skilled in the art can well understand and utilize the application. The application is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method, characterized in that, The method comprises the following steps: S1, setting a control period to collect two-phase liquid cooling control data, performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data; S2, based on the pre-processed two-phase liquid cooling control data, periodically evaluating the heat load level of each region, constructing a discrete time linear state space model, and outputting a heat load prediction sequence; Compare the prediction sequence with the current heat load state, extract the heat load change rate, and generate a cold quantity allocation input data set; S3, based on the cold quantity allocation input data set, periodically quantifying the cold quantity allocation of each region, combining the pump type characteristics and flow resistance parameters to construct a control quantity conversion relationship, calculating the pump speed and valve opening target value, and generating adjustment control instructions to realize partitioned cold quantity allocation control; S4, analyze the heat load prediction trend, identify high heat risk and low heat stable regions, evaluate the cold quantity compensation demand, output the cold quantity enhancement and recovery allocation value, and generate the corresponding control instructions to realize dynamic adjustment compensation and rapid response to heat disturbance of cold quantity.

2. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 1, characterized in that: The specific steps of setting a control period to collect two-phase liquid cooling control data, performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data are as follows: A fixed time window control period is designed to collect two-phase liquid cooling control data, which includes server power consumption, cooling liquid inlet temperature, cooling liquid outlet temperature, vapor flow rate, cooling liquid mass flow rate, cooling liquid specific heat capacity, and cooling liquid density. Through the construction of a time sequence synchronization method based on multi-channel timestamp alignment and sampling interval reconstruction, cross-module time alignment and data frame unified processing are performed on the two-phase liquid cooling control data. Through the introduction of edge disturbance judgment algorithm of window sliding mean analysis and first-order derivative mutation identification, short-period fluctuation detection and high-frequency noise removal processing are performed on the two-phase liquid cooling control data; By establishing a physical consistency criterion based on interval change range and logical boundary condition, abnormal deviation identification and dynamic correction compensation processing are performed on the two-phase liquid cooling control data; A unit conversion matrix and input field mapping rule for physical quantity unification are constructed to standardize and normalize the physical units and characteristics of the two-phase liquid cooling control data.

3. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 1, characterized in that: The specific steps of periodically evaluating the heat load level of each region based on the pre-processed two-phase liquid cooling control data are as follows: Based on the structure and cooling path of the server (7), the control components with independent monitoring and control capabilities are defined as control nodes, and the control nodes are divided into regions according to whether the cooling path structure is the same; Based on the pre-processed two-phase liquid cooling control data, the heat load level of each region in the control period is evaluated: the cooling liquid mass flow rate is multiplied by the cooling liquid specific heat capacity, and then multiplied by the difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature to obtain the cooling liquid sensible heat transfer term; The cooling liquid mass flow rate is multiplied by the cooling liquid specific heat capacity, and then multiplied by the first-order derivative of the cooling liquid outlet temperature with respect to time to obtain the temperature dynamic response term; The cooling liquid sensible heat transfer term, server power consumption and temperature dynamic response term are added in turn to obtain the control period heat load evaluation value.

4. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 1, characterized in that: The specific steps of constructing the discrete time linear state space model, outputting the heat load prediction sequence, comparing the prediction sequence with the current heat load state, extracting the heat load change rate, and generating the cold distribution input data set are as follows: A discrete time linear state space model is constructed with two-phase liquid cooling control data as an input vector and a control period heat load evaluation value as a state variable. A difference mapping relationship between the input vector and the state variable is established, and the least squares method is used to estimate the parameters of the two-phase liquid cooling control data for a fixed control period to complete the model parameter training. Based on the trained discrete time linear state space model, the heat load state variable is predicted in each control period, and a heat load prediction sequence for multiple future periods is obtained. The heat load state variable prediction sequence is compared with the heat load evaluation value at the current time point, a heat load change slope sequence is constructed, the heat load change rate between each control period is extracted, and the two-phase liquid cooling control data of the current control period is combined to construct a cold distribution input data set.

5. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 1, characterized in that: The specific steps of quantifying the cold distribution amount of each region based on the cold distribution input data set are as follows: The cooling liquid mass flow rate is multiplied by the specific heat capacity of the cooling liquid, and then multiplied by the absolute value of the first derivative of the cooling liquid outlet temperature with respect to time to obtain a cooling liquid temperature dynamic change term. The heat load change rate is multiplied by the server power consumption and divided by the product of the cooling liquid density and the steam flow rate to obtain a load adjustment compensation term. The cooling liquid temperature dynamic change term and the load adjustment compensation term are added to obtain a basic cold distribution evaluation value. The difference between the cooling liquid outlet temperature and the cooling liquid inlet temperature is divided by the sum of the cooling liquid inlet temperature and a minimum term to obtain a temperature difference correction factor. The basic cold distribution evaluation value is multiplied by the result obtained by adding one to the temperature difference correction factor to obtain a cold distribution amount evaluation value.

6. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 1, characterized in that: The specific steps of constructing the control amount conversion relationship combined with the pump type characteristics and the flow resistance parameters are as follows: The cold distribution amount evaluation value is taken as the target value, and the conversion relationship between the cold distribution output and the driving set value is established by combining the physical response characteristics of the flow regulation components in the cooling liquid driving path and the flow control components in the distribution path. According to the pump type performance curve, a function mapping relationship between the cooling liquid instantaneous flow rate and the pump speed is constructed. Based on the cooling circuit pressure drop and the flow resistance parameters, a function mapping relationship between the branch valve opening and the local flow distribution ratio is constructed.

7. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 6, characterized in that: The specific steps of calculating the pump speed and valve opening target values and generating adjustment control instructions to realize partitioned cold distribution control are as follows: With the conversion relationship as a constraint condition, a pump speed adjustment target value and a valve opening degree adjustment target value corresponding to the current cold quantity distribution evaluation value are calculated; the pump speed adjustment target value is taken as a basis for generating an instruction to construct a pump speed control instruction; the valve opening degree adjustment target value is written into an opening degree instruction structure in sequence according to the branch numbers to generate a valve opening degree control instruction; the pump speed control instruction and the valve opening degree control instruction are written into a pump control buffer and a valve control buffer respectively, and control parameter refreshing and adjustment instruction issuing are completed according to a current control cycle time base; under the drive of the valve opening degree control instruction, a nozzle control unit is activated in linkage, and the nozzle (8) performs directional spraying operation according to the branch flow configuration corresponding to the current cycle valve opening degree to supplement the two-phase cooling liquid to the server (7) peripheral heat dissipation area, forming a partitioned cold quantity enhanced flow field.

8. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 1, characterized in that: The specific steps of the analysis of the heat load prediction trend, identification of the high heat risk and low heat stable area, and evaluation of the cold quantity compensation demand are as follows: The heat load state variable prediction sequence of each area is extracted, and the heat load change amount per unit time is calculated in combination with the heat load change rate sequence of the historical control cycle; if the heat load change amount is positive and the heat load change amount of the last three control cycles exceeds the load increment threshold, the corresponding area is marked as a high heat risk area; if the heat load change amount is negative and the heat load change amount of the last three control cycles exceeds the load attenuation threshold, the corresponding area is marked as a low heat stable area; According to the heat load change amount of each node in the heat disturbance risk area, in combination with the two-phase liquid cooling control data, the cold quantity compensation demand of each area that needs to be responded in advance in the next control cycle is evaluated: the absolute value of the second derivative of the absolute value of the heat load change amount with respect to time is taken to obtain a heat disturbance acceleration term; the absolute value of the second derivative of the cooling liquid outlet temperature with respect to time is taken, and multiplied by a heat disturbance adjustment coefficient to obtain a temperature disturbance acceleration term; the heat disturbance acceleration term and the temperature disturbance acceleration term are added to obtain a heat temperature disturbance response base value; the cooling liquid mass flow rate is multiplied by the specific heat capacity of the cooling liquid, and then divided by the product of the cooling liquid density and the steam flow rate to obtain the cooling liquid adjustment amount corresponding to a unit disturbance; the cooling liquid outlet temperature standard deviation is divided by the sum of the cooling liquid outlet temperature average value and the minimum term, and the ratio is added to the constant one to obtain a temperature fluctuation correction factor; The cold quantity distribution compensation evaluation value is obtained by multiplying the heat temperature disturbance response base value, the cooling liquid adjustment amount corresponding to a unit disturbance, and the temperature fluctuation correction factor.

9. The multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control method according to claim 1, characterized in that: The specific steps of outputting the cold quantity enhancement and recovery distribution value and generating the corresponding control instruction to realize the dynamic adjustment compensation of the cold quantity and the rapid response of the heat disturbance are as follows: The cold quantity distribution compensation evaluation value of each node in the high heat risk area is superimposed on the corresponding cold quantity distribution evaluation value to form a cold quantity enhancement distribution value; the cold quantity distribution evaluation value of each node in the low heat stable area is subtracted by the corresponding cold quantity distribution compensation evaluation value to form a cold quantity recovery distribution value; The cold energy enhancement allocation value and the cold energy recovery allocation value are respectively input into the conversion relationship to obtain the pump speed adjustment target value and the valve opening adjustment target value corresponding to each node; the pump speed adjustment target value and the valve opening adjustment target value are written into the opening instruction structure to generate corresponding pump speed adjustment instructions, branch valve opening instructions and nozzle partition spraying instructions; the control parameters are refreshed according to the current control cycle time base, and the instruction issuing is completed to realize the fast compensation response of cold energy in the high heat area and the recovery regulation of cold energy in the low heat area.

10. A multi-parameter MPC two-phase liquid cooling energy efficiency optimized distribution control system, characterized in that: Comprise: The data acquisition preprocessing module, the heat load modeling prediction module, the cold energy allocation control execution module and the disturbance compensation dynamic adjustment module, wherein: The data acquisition preprocessing module is used for setting a control cycle to collect two-phase liquid cooling control data, performing time alignment, noise removal, abnormal correction and standard normalization processing on the two-phase liquid cooling control data; The heat load modeling prediction module is used for periodically evaluating the heat load level of each region based on the preprocessed two-phase liquid cooling control data, constructing a discrete time linear state space model, and outputting a heat load prediction sequence; The prediction sequence is compared with the current heat load state, the heat load change rate is extracted, and a cold energy allocation input data set is generated; The cold energy allocation control execution module is used for periodically quantifying the cold energy allocation amount of each region based on the cold energy allocation input data set, constructing a control amount conversion relationship combined with pump type characteristics and flow resistance parameters, calculating pump speed and valve opening target values, and generating adjustment control instructions to realize partitioned cold energy allocation control; The disturbance compensation dynamic adjustment module is used for analyzing the heat load prediction trend, identifying high heat risk and low heat stable regions, evaluating cold energy compensation demand, outputting cold energy enhancement and recovery allocation values, and generating corresponding control instructions to realize dynamic adjustment compensation of cold energy and fast response to heat disturbance.

Images