A liquid cooling system dynamic control method and system based on thermal resistance and flow resistance analysis

Abstract

The application discloses a kind of liquid cooling system dynamic control method and system based on thermal resistance flow resistance analysis, and it is related to liquid cooling thermal management field.The method comprises the following steps: S1: collecting multi-source heterogeneous sensor array parameters, calculating dynamic confidence weight, and generating system state observation value;S2: based on system state observation value, establish thermal resistance flow resistance coupling model framework and online identification and real-time update pre-built key parameters, obtain updated thermal resistance flow resistance coupling model;S3: by the model predictive control parameter of preestablished, construct multi-objective optimization problem, based on thermal resistance flow resistance coupling model, solve current optimal control instruction using predictive control algorithm;S4: collect historical operation data to establish long-term evolution mode factor, optimize model predictive control parameter to generate correction execution parameter.Through real-time optimization thermal resistance flow resistance model and control strategy, combined with deep learning realizes intelligent pre-adjustment, significantly improves the heat dissipation efficiency of liquid cooling system, reduces energy consumption and ensures stable operation.

Description

Technical Field

[0001] This invention relates to the field of liquid cooling thermal management, specifically to a dynamic control method and system for liquid cooling systems based on thermal resistance and flow resistance analysis. Background Technology

[0002] With the rapid development of modern technology, the power density and heat generation in electronic devices, new energy vehicles, data centers, and other fields are constantly increasing, and heat dissipation has become a key bottleneck restricting performance improvement. Liquid cooling technology, due to its efficient heat conduction capabilities, has gradually become one of the mainstream technologies for solving high power density heat dissipation problems. However, the thermal resistance and flow resistance issues in liquid cooling systems severely affect heat dissipation efficiency and system reliability.

[0003] In liquid cooling systems, thermal resistance refers to the resistance encountered during heat transfer, directly affecting the efficiency of heat transfer from the heat source to the coolant. Thermal resistance mainly originates from the contact thermal resistance between the coolant and the heat source, the thermal conductivity resistance of the coolant itself, and the thermal resistance of the flow channel walls. The presence of thermal resistance prevents heat from being transferred to the coolant in a timely manner, thus affecting the heat dissipation effect.

[0004] Flow resistance is the resistance encountered by coolant when it flows through the flow channels, directly affecting the flow rate and volume of the coolant. Excessive flow resistance can lead to poor coolant circulation, resulting in localized heat buildup and reduced overall heat dissipation efficiency. Increased flow resistance not only reduces heat dissipation efficiency but also increases system energy consumption, reduces operating efficiency, and may even cause system failure in extreme cases.

[0005] Therefore, optimizing the thermal and flow resistance in liquid cooling systems and proposing effective control strategies can not only improve the system's heat dissipation efficiency but also enhance its stability and reliability. In-depth research into the formation mechanisms, online estimation, and real-time control methods of thermal and flow resistance is of great significance for promoting technological advancements in liquid cooling systems and the development of related industries. Summary of the Invention

[0006] In view of the shortcomings of the prior art described above, the purpose of this invention is to provide a dynamic control method and system for liquid cooling systems based on thermal resistance and flow resistance analysis, so as to solve the above-mentioned technical problems.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis, comprising: S1: Collect parameters of the multi-source heterogeneous sensor array, calculate dynamic confidence weights, and generate system state observations; S2: Based on the system state observations, establish a thermal resistance-flow resistance coupling model framework and identify and update the pre-built key parameters online and in real time to obtain the updated thermal resistance-flow resistance coupling model. S3: Construct a multi-objective optimization problem using preset model predictive control parameters, and solve the current optimal control command using a predictive control algorithm based on a thermal resistance-current resistance coupling model. S4: Collect historical operating data to establish long-term evolutionary model factors, optimize model prediction control parameters, and generate corrected execution parameters.

[0008] The present invention is further configured such that S1 includes: Multi-source heterogeneous sensor arrays are deployed at each key node of the liquid cooling system flow channel to collect raw data from the data source, including the temperature, pressure and flow rate of the current data source node. Based on the original data, outlier data were removed and corrected data was generated using variance statistics and consistency tests. The stability parameter of the data source is obtained by calculating the variance of the corrected data, and the volatility parameter of the data source is obtained by calculating the rate of change of the corrected data. Calculate the dynamic confidence weight for each data source based on its stability and volatility. An autoregressive sliding calibration algorithm is used to smooth the corrected data, suppressing short-term fluctuations and generating fused data. Based on the fused data, high-confidence system state observations are generated.

[0009] The present invention is further configured such that S2 includes: a framework and interface definition unit and an online identification and update unit; The framework and interface definition units include: The key parameters are initialized and preset to initial values. The key parameters include: dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometry change coefficient. Obtain a pre-stored set of inherent system attribute parameters, which includes the following parameters determined by the system design and materials: heat source contact area, contact interface material properties, flow channel material, flow channel wall thickness, original surface roughness of the flow channel, flow channel geometry, and flow channel bending angle.

[0010] The present invention is further configured to collect and construct a real-time operating condition parameter set through sensors, the real-time operating condition parameter set including: coolant flow rate, coolant flow rate and coolant temperature; The hydraulic diameter of the flow channel is calculated based on the channel geometry, and the real-time viscosity and density of the coolant are calculated based on the temperature observed in the system state.

[0011] The present invention is further configured to calculate physical property parameters within a preset time window based on the temperature parameters in the system state observation values, wherein the physical property parameters include: temperature change rate, thermal conductivity, and specific heat capacity; Based on the system's inherent attribute parameter set and the real-time operating condition parameter set, the contact thermal resistance, the flow channel wall thermal resistance, and the coolant thermal conduction resistance are calculated respectively, and the total system thermal resistance is obtained by summing them. A flow resistance model is established based on the original surface roughness of the flow channel, the geometric dimensions of the flow channel, the tortuosity of the flow channel, the real-time viscosity and density of the coolant, and the flow velocity of the coolant. Using coolant flow rate as the key variable, the total thermal resistance of the system is correlated with the flow resistance model, and correction interfaces are reserved for key parameters in both models to construct a thermal resistance-flow resistance coupled model.

[0012] The present invention is further configured such that the online identification and update unit includes: Based on real-time system state observations and physical property parameters, the values ​​of dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometric change coefficient in the thermal resistance-flow resistance coupling model are estimated and updated in real time using the recursive least squares method. The recursive least squares method adaptively adjusts key parameters by minimizing the error between the thermal resistance and flow resistance values ​​predicted by the model and the actual thermal resistance and flow resistance values ​​inverted from the observations.

[0013] The present invention is further configured such that S3 includes: With the optimization objectives of minimizing total system power consumption and temperature deviation at key points, a multi-objective optimization problem is constructed by predicting control parameters using a pre-set model. Set the feasible range of system flow, pressure, and control variables as constraints for the objective optimization problem; Based on real-time updated system state observations and an online-updated thermal resistance-current resistance coupling model, the optimal control command sequence in the finite time domain is solved. Apply immediate control commands from the optimal control command sequence to the actuators of the liquid cooling system.

[0014] The present invention is further configured such that S4 includes: Regularly collect and store timestamp sequences, external operating condition data, and system performance indicators from the long-term historical operating data of the system. The external operating condition data includes: ambient temperature, ambient humidity, and business load cycle. The system performance indicators include: average power consumption, average temperature of key points, and average flow rate. Seasonal and periodic features are extracted from the timestamp sequence and combined with external operating condition data and system performance indicators to form a training feature set; Based on the training feature set, the long-term evolution of the system's operating mode is learned through a long short-term memory network model.

[0015] The present invention is further configured to extract the state vector of the hidden layer of the long short-term memory network model as a long-term evolution mode factor; Establish a mapping network between long-term evolutionary model factors and model prediction control parameters; Long-term evolutionary model factors are converted into specific model prediction control parameter preset values ​​through a mapping relationship network; The model predictive control parameters are pre-tuned and downloaded to the main controller. The parameters of the model predictive control module are pre-tuned to optimize the control strategy, generate corrected execution parameters, and adjust the operating parameters of the actuators of the liquid cooling system.

[0016] The present invention also provides a dynamic control system for a liquid cooling system based on thermal resistance and flow resistance analysis, the system comprising: Data sensing and fusion module: collects parameters of multi-source heterogeneous sensor arrays, calculates dynamic confidence weights, and generates system state observations; Online thermal resistance and flow resistance estimation module: Based on system state observations, a thermal resistance and flow resistance coupling model framework is established and the pre-built key parameters are identified and updated in real time to obtain the updated thermal resistance and flow resistance coupling model. Model Predictive Control Module: Constructs a multi-objective optimization problem using preset model predictive control parameters, and solves the current optimal control command based on a thermal resistance-current resistance coupling model using predictive control algorithms; Intelligent optimization and long-term learning module: Collect historical operating data to establish long-term evolutionary model factors, optimize model prediction control parameters, and generate corrected execution parameters.

[0017] This invention provides a dynamic control method and system for a liquid cooling system based on thermal resistance and flow resistance analysis. The method comprises: S1: acquiring parameters of a multi-source heterogeneous sensor array, calculating dynamic confidence weights, and generating system state observations; S2: establishing a thermal resistance and flow resistance coupling model framework based on the system state observations, identifying and updating pre-built key parameters online and in real time to obtain an updated thermal resistance and flow resistance coupling model; S3: constructing a multi-objective optimization problem using preset model predictive control parameters, and solving the current optimal control command based on the thermal resistance and flow resistance coupling model using a predictive control algorithm; S4: acquiring historical operating data to establish a long-term evolutionary model factor, optimizing the model predictive control parameters, and generating corrected execution parameters. The beneficial effects include: Improving the heat dissipation efficiency and stability of liquid cooling systems: By accurately constructing and updating a thermal resistance-flow resistance coupling model in real time, the thermal resistance and flow resistance parameters in the liquid cooling system can be dynamically optimized, effectively improving the fluidity and heat transfer efficiency of the coolant, thereby significantly enhancing the system's heat dissipation performance. This method can adapt to different operating conditions and environmental changes, ensuring high-efficiency heat dissipation performance under various working conditions and enhancing the long-term stability of the system.

[0018] Adaptive optimization control strategy reduces energy consumption: Based on real-time estimated parameters such as dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometry change coefficient, combined with model predictive control algorithm, the system power consumption can be minimized and the energy consumption of the liquid cooling system can be reduced by optimizing the control of coolant flow rate and pump power. In addition, the multi-objective optimization method can take into account power consumption and key point temperature deviation, ensuring that heat dissipation efficiency is guaranteed without wasting excessive energy.

[0019] Intelligent long-term evolutionary learning and adaptive adjustment: By establishing long-term evolutionary pattern factors and combining them with deep learning algorithms such as long short-term memory networks, the system can predict future load changes based on historical operating data and external operating conditions, and pre-adjust model predictive control parameters in real time. This adaptive learning capability not only enables the liquid cooling system to cope with long-term changing loads, but also optimizes operating strategies under seasonal changes and external environmental fluctuations, extending system life and maintaining long-term high-efficiency operation.

[0020] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A flowchart illustrating a dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis, as an exemplary embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the structure of a dynamic control system for a liquid cooling system based on thermal resistance and flow resistance analysis, as an exemplary embodiment of the present invention. Detailed Implementation

[0022] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention and not for limiting the scope of protection of the present invention.

[0023] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0024] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention. Example 1

[0025] A dynamic control method for liquid cooling systems based on thermal resistance and flow resistance analysis, such as Figure 1 As shown, it includes: S1: Collect parameters of the multi-source heterogeneous sensor array, calculate dynamic confidence weights, and generate system state observations; S2: Based on the system state observations, establish a thermal resistance-flow resistance coupling model framework and identify and update the pre-built key parameters online and in real time to obtain the updated thermal resistance-flow resistance coupling model. S3: Construct a multi-objective optimization problem using preset model predictive control parameters, and solve the current optimal control command using a predictive control algorithm based on a thermal resistance-current resistance coupling model. S4: Collect historical operating data to establish long-term evolutionary model factors, optimize model prediction control parameters, and generate corrected execution parameters.

[0026] The present invention is further configured such that S1 includes: Multi-source heterogeneous sensor arrays are deployed at each key node of the liquid cooling system flow channel to collect raw data from the data source, including the temperature, pressure and flow rate of the current data source node. Based on the original data, outlier data were removed and corrected data was generated using variance statistics and consistency tests. The stability parameter of the data source is obtained by calculating the variance of the corrected data, and the volatility parameter of the data source is obtained by calculating the rate of change of the corrected data. Calculate the dynamic confidence weight for each data source based on its stability and volatility. An autoregressive sliding calibration algorithm is used to smooth the corrected data, suppressing short-term fluctuations and generating fused data. Based on the fused data, high-confidence system state observations are generated. Specifically, a multi-source heterogeneous sensor array is deployed at key nodes in the liquid cooling system flow channel to collect raw data of parameters such as temperature, pressure, and flow rate in real time. Subsequently, consistency testing methods, such as the Raida criterion, are used to analyze the data from multiple sensors at the same measuring point, eliminating outliers that significantly deviate from the group data, and obtaining a set of reliable corrected data. Two indicators are calculated for the corrected data of each sensor: 1. Stability parameter: The stability parameter reflects the consistency and stability of the sensor's output data within a given time period. It is calculated by calculating the variance of the corrected data of the sensor over a period of time, and then setting the variance as the stability parameter. The smaller the variance, the more stable the sensor. 2. Volatility parameter: The volatility parameter represents the degree of drastic change in sensor data over time, reflecting the short-term trend of data change. It is calculated by calculating the rate of change of the corrected data of the sensor at adjacent moments, and then setting the rate of change as the volatility parameter. The smaller the rate of change, the smaller the instantaneous volatility. Subsequently, a weighted scoring method is used to calculate a comprehensive score for each sensor based on its stability and volatility parameters. This score is then normalized to obtain the dynamic confidence weight for that sensor. The dynamic confidence weight is calculated based on the stability and volatility of the data source and is used to calibrate the contribution of each data source to the final system state; the higher the stability and the lower the volatility of the data source, the greater the confidence weight. First, a first-order hysteresis filtering algorithm is used to smooth the corrected data sequence of each sensor to suppress high-frequency fluctuations caused by random noise. Then, the smoothed sensor data for the same physical quantity are weighted and averaged according to their corresponding dynamic confidence weights to generate fused data. The fused data is obtained by integrating the correction results of multi-source data, and after smoothing, outlier removal, and weight adjustment, the final data is obtained. Finally, the weighted average result is used as the high-confidence system state observation value for that physical quantity, including observed temperature, observed pressure, and observed flow rate, and output to the subsequent control module.

[0027] The present invention is further configured such that S2 includes: a framework and interface definition unit and an online identification and update unit; The framework and interface definition units include: The key parameters are initialized and preset to initial values. The key parameters include: dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometry change coefficient. The system acquires a pre-stored set of inherent system attribute parameters, which includes parameters determined by system design and materials: heat source contact area, contact interface material properties, channel material, channel wall thickness, initial channel surface roughness, channel geometry, and channel tortuosity. Specifically, key parameters are those used in the liquid cooling system's thermal resistance / flow resistance coupling model to describe thermodynamic and fluid dynamic characteristics. Initializing these key parameters is crucial because they affect the overall system's thermal performance and flow efficiency. Specifically, these include the following key parameters: the dynamic temperature response coefficient reflects the impact of temperature changes on thermal and flow resistance. As system temperature changes, the coolant's thermal conductivity, fluidity, and other physical properties also change, thus affecting thermal and flow resistance. The adaptive viscosity correction factor considers changes in coolant viscosity, which is affected by temperature and flow rate; adjusting the viscosity helps to more accurately predict flow resistance. The channel geometry variation coefficient describes the impact of channel geometry, such as diameter, roughness, and tortuosity, on flow resistance. The channel geometry directly affects the coolant's fluidity. The specific initialization implementation approach is as follows: During system initialization, initial values ​​are set for each key parameter. The default initial value is 1 or other reasonable initial estimates, which can also be set based on previous experimental data or system design specifications. These initial values ​​provide the basis for subsequent online updates and optimizations. The inherent properties of the system are provided by design and material specifications and are usually determined during system design. These parameters are loaded into the system through a database or manual input. Among these, the heat source contact area represents the area of ​​contact between the coolant and the heat source; this parameter affects heat transfer efficiency and is usually related to the size and shape of the heat source. The properties of the contact interface material determine the heat exchange capacity between the coolant and the heat source. The material of the flow channel directly affects the fluid's flowability and heat transfer efficiency; different materials have different thermal conductivity and coefficients of friction. The channel wall thickness and initial surface roughness affect the flow resistance of the coolant, thus further affecting the flow resistance. The channel geometry and channel bending angle directly affect the coolant's flow path and velocity, thus affecting the flow resistance.

[0028] The present invention is further configured to collect and construct a real-time operating condition parameter set through sensors, the real-time operating condition parameter set including: coolant flow rate, coolant flow rate and coolant temperature; The hydraulic diameter of the flow channel is calculated based on its geometry, and the real-time viscosity and density of the coolant are calculated based on the temperature observed in the system state. Specifically, velocity sensors such as vortex flow meters and ultrasonic flow meters are used to measure the flow velocity of the coolant in the flow channel in real time; flow meters such as mass flow meters and orifice plate flow meters are used to measure the volume of coolant passing through the flow channel per unit time; and temperature sensors such as thermocouples and RTD sensors are used to monitor the real-time temperature changes of the coolant. The collected parameters are then integrated to construct a real-time operating condition parameter set. The hydraulic diameter of a flow channel is an equivalent geometric parameter characterizing the flow capacity of the channel. For non-circular channels, it describes the flow characteristics more accurately than the physical diameter. The implementation approach is to use the corresponding hydraulic diameter calculation formula based on the channel geometry, taking the channel geometry as input to calculate the hydraulic diameter. Real-time coolant viscosity represents the viscosity of the coolant at the current temperature and is a key physical property parameter for calculating flow resistance. Viscosity decreases significantly with increasing temperature. The specific implementation approach is to obtain the viscosity by interpolation from a database of coolant properties based on the coolant type and the real-time measured temperature. Real-time coolant density represents the mass per unit volume of the coolant at the current temperature. The implementation approach is similar to viscosity, obtaining the density by looking up a table based on the coolant type and the real-time temperature.

[0029] The present invention is further configured to calculate physical property parameters within a preset time window based on the temperature parameters in the system state observation values, wherein the physical property parameters include: temperature change rate, thermal conductivity, and specific heat capacity; Based on the system's inherent attribute parameter set and the real-time operating condition parameter set, the contact thermal resistance, the flow channel wall thermal resistance, and the coolant thermal conduction resistance are calculated respectively, and the total system thermal resistance is obtained by summing them. A flow resistance model is established based on the original surface roughness of the flow channel, the geometric dimensions of the flow channel, the tortuosity of the flow channel, the real-time viscosity and density of the coolant, and the flow velocity of the coolant. Using coolant flow rate as the key variable, the total thermal resistance and flow resistance models of the system are correlated, and correction interfaces are reserved for key parameters in both models to construct a coupled thermal and flow resistance model. Specifically, the rate of temperature change represents the amount of temperature change per unit time, characterizing the trend and speed of temperature change; positive values ​​indicate heating up, and negative values ​​indicate cooling down; it is calculated using existing numerical differentiation / first-order difference methods or moving average filtering algorithms. Thermal conductivity is a physical quantity used to measure a material's ability to conduct heat; the higher the thermal conductivity, the faster the heat is conducted in the coolant; it is obtained using existing physical property databases and linear interpolation algorithms. Specific heat capacity measures the amount of heat required to raise the temperature of a unit mass of material by a unit; the larger the specific heat capacity, the stronger the coolant's heat storage capacity, and the slower the temperature rise; similar to thermal conductivity, it is obtained using existing physical property databases and linear interpolation algorithms. Thermal resistance is the resistance encountered during heat transfer; its reciprocal is the heat dissipation capacity. The total thermal resistance is composed of multiple individual thermal resistances connected in series. These individual thermal resistances include: 1. Contact thermal resistance: This refers to the resistance to heat transfer between the coolant and the heat source interface. Its magnitude is affected by the material properties of the interface and the contact area. 2. Channel wall thermal resistance: This is the resistance to heat transfer between the coolant and the channel wall. It is affected by factors such as the channel material, channel wall thickness, and surface roughness. The thinner the channel wall and the higher the thermal conductivity, the lower the thermal resistance. 3. Coolant thermal conduction resistance: This is the thermal resistance within the coolant caused by temperature differences. It is affected by factors such as the coolant type, flow rate, density, and viscosity. The total system thermal resistance is the sum of these three individual thermal resistances. The higher the total thermal resistance, the lower the heat dissipation efficiency. The contact thermal resistance is calculated by obtaining the heat source contact area and the material properties of the contact interface from the system's inherent property parameter set. Based on the cross-sectional heat transfer theory, the reciprocal of the contact thermal conductivity is calculated and then divided by the contact area to obtain the contact thermal resistance. The flow channel wall thermal resistance is calculated by obtaining the flow channel wall thickness, the thermal conductivity of the flow channel material, and the flow channel contact area from the system's inherent property parameter set. Based on Fourier's law of heat conduction, the wall thickness is divided by the material's thermal conductivity multiplied by the contact area. The coolant thermal conduction resistance is calculated by obtaining the coolant flow velocity from the real-time operating condition parameter set, the thermal conductivity from the physical property parameters, and the hydraulic diameter of the flow channel from the inherent properties. First, using correlations such as the Darcy-Weisbach formula and the Dittus-Belt formula, the convective heat transfer coefficient is calculated based on the flow velocity, physical properties, and geometric dimensions. The coolant thermal conduction resistance is the reciprocal of the product of the convective heat transfer coefficient and the heat source contact area. Finally, based on the principle of series thermal resistance superposition, the three component thermal resistances are added to obtain the total thermal resistance. Regarding the construction of the flow resistance model, firstly, the friction coefficient and local resistance coefficient are calculated based on the flow channel geometry and bending angle.Then, combining the real-time viscosity and density of the coolant with the coolant flow rate, the Reynolds number at the current flow rate is calculated to determine the flow regime. Next, based on the surface roughness and Reynolds number, the friction factor is calculated using a Chamoody diagram or other appropriate formulas, such as the Körbrook-White formula. Then, the local resistance is obtained based on the standard of the flow channel bend angle. Subsequently, the friction pressure drop is calculated using the Darcy-Weisbach formula, and the bend pressure drop is calculated using the local resistance formula. The two are added together to obtain the total flow channel pressure drop at the current flow rate. Finally, this process is repeated at different flow rate points to plot the "pressure drop-flow rate" curve, which is the flow resistance model. In the modeling software, both the total thermal resistance and flow resistance models of the system are linked to the variable "coolant flow rate." A model object is created in the software, and then adjustable coefficient positions are set for three key parameters—dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometry change coefficient—in the mathematical expressions for calculating thermal and flow resistance. Specifically, the process of calculating flow resistance is encapsulated as a function, and a unified model object is created in the software. Adjustable coefficient positions are reserved for the three key parameters in the calculation expressions for thermal and flow resistance: in the calculation of thermal resistance, the calculated convective heat transfer coefficient is multiplied by the dynamic temperature response coefficient; in the calculation of flow resistance, the calculated friction factor is multiplied by the adaptive viscosity correction factor and the flow channel geometry change coefficient. During initialization, the default values ​​for the three key parameters—dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometry change coefficient—are all set to 1, or other initial values ​​verified by expert experiments; this provides an interface for subsequent online updates of the key parameters. Finally, the thermal resistance model and the flow resistance model that are well-connected and have interfaces are integrated into a unified model framework to form a thermal resistance and flow resistance coupled model.

[0030] The present invention is further configured such that the online identification and update unit includes: Based on real-time system state observations and physical property parameters, the dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometry variation coefficient in the thermal resistance-flow resistance coupling model are estimated and updated in real time using the recursive least squares method. The recursive least squares method adaptively adjusts key parameters by minimizing the error between the model-predicted thermal resistance and flow resistance values ​​and the actual thermal resistance and flow resistance values ​​derived from the observations. Specifically, the recursive least squares method is a commonly used algorithm for estimating system parameters. This method estimates and updates key parameters in the thermal resistance-flow resistance coupling model by minimizing the error between the model-predicted thermal resistance and flow resistance values ​​and the actual observed values. The specific steps are as follows: First, calculate the actual thermal resistance and actual flow resistance based on the temperature information in the system state observations. The actual thermal resistance is calculated according to the fundamental laws of thermodynamics by subtracting the difference between the system state observation temperature and the coolant inlet temperature, and then dividing by the power. Both the coolant inlet temperature and power can be measured by sensors. The actual flow resistance is mainly reflected in the pressure drop, which can be directly obtained by measuring the pressure difference between the inlet and outlet of the flow channel. Existing technologies such as thermal resistance calculation based on the first law of thermodynamics and pressure sensor measurement can be used. Second, combine the current system state observations, physical property parameters, and other known quantities in the model into a regression vector. This regression vector describes the current state of the system. Then, using an existing thermal resistance and flow resistance model, input the regression vector into the thermal resistance and flow resistance coupled model for prediction, obtaining the predicted thermal resistance and flow resistance values. Finally, calculate the error between the predicted thermal resistance and flow resistance values ​​and the actual observations. Finally, the error and regression vector are input together into the core algorithm of recursive least squares. This algorithm first calculates a gain vector based on the internal state, then uses this gain vector to weight the prediction error, generating parameter adjustments to update the estimates of key parameters. Simultaneously, the algorithm recursively updates its internal state, repeating this process in each control cycle to continuously reduce the prediction error and optimize the model's accuracy.

[0031] The present invention is further configured such that S3 includes: With the optimization objectives of minimizing total system power consumption and temperature deviation at key points, a multi-objective optimization problem is constructed by predicting control parameters using a pre-set model. Set the feasible range of system flow, pressure, and control variables as constraints for the objective optimization problem; Based on real-time updated system state observations and an online-updated thermal resistance-current resistance coupling model, the optimal control command sequence in the finite time domain is solved. The immediate control commands from the optimal control command sequence are applied to the actuators of the liquid cooling system. Specifically, the model predictive control parameters are used to configure the optimization problem, including weighting coefficients, prediction time domain, and control time domain. The weighting coefficients balance the relative importance of the conflicting objectives of "system power consumption" and "temperature deviation." For example, a higher weight for temperature deviation means the controller is more likely to rapidly cool down the system and may ignore energy consumption; the default value is 1.0. The prediction time domain is used to predict the future time step of the optimization problem. A longer time domain provides better predictability but also increases computational complexity; the default value is 10 control cycles, approximately 10 minutes. The control time domain is used to determine the length of the future control command sequence to be optimized, and is typically shorter than the prediction time domain; the default value is 5 control cycles. A comprehensive objective function is constructed, whose value is the weighted sum of the squares of the total system power consumption and the deviations of the key point temperature from the setpoint. The weighting coefficients in the model predictive control parameters are invoked here to set the weights of the two objectives. The optimization problem must satisfy the following constraints: 1. Process constraints: the system flow rate must be within the safe operating range of the pump, and the system pressure must not exceed the pressure limits of the pipelines and components; 2. Control variable constraints: control commands, such as pump speed setpoints, must be within their executable range. For example, the pump speed cannot be negative and cannot exceed the maximum speed. The above multi-objective optimization problem can be constructed using existing technology: multi-objective optimization theory, which transforms multiple performance indicators into a single objective function through a weighted sum method. Then, under the given constraints, the sequence of future control actions that minimizes the objective function is sought. The specific implementation process is as follows: the current high-confidence system state observations provided by module S1 are read as the initial conditions for optimization. Then, the thermal resistance and flow resistance coupling model provided by module S2, after online parameter updates, is used to predict how the temperature and pressure in the system state observations will evolve under different control commands in the future prediction time domain. Finally, numerical optimization algorithms, such as quadratic programming (QP) or interior point methods, are used to automatically search for the sequence of future control commands that minimizes the objective function value, provided that all constraints are satisfied. This process takes place within a finite prediction time domain and can be solved using existing technologies such as Model Predictive Control (MMCC) to obtain the optimal control command sequence. Finally, the first command, the immediate control command, is extracted from the obtained optimal control command sequence. This command value is then sent to the corresponding actuator, such as a frequency converter-driven pump or regulating valve, via an industrial communication protocol. The actuator adjusts its output according to the command, changing the coolant flow rate and completing one control loop.

[0032] The present invention is further configured such that S4 includes: Regularly collect and store timestamp sequences, external operating condition data, and system performance indicators from the long-term historical operating data of the system. The external operating condition data includes: ambient temperature, ambient humidity, and business load cycle. The system performance indicators include: average power consumption, average temperature of key points, and average flow rate. Seasonal and periodic features are extracted from the timestamp sequence and combined with external operating condition data and system performance indicators to form a training feature set; Based on the training feature set, a Long Short-Term Memory (LSTM) network model is used to learn the long-term evolution patterns of the system's operating modes. Specifically, the system periodically collects and stores long-term historical operating data, including timestamp sequences, external operating condition data, and system performance indicators. External operating condition data includes ambient temperature, ambient humidity, and business load cycles; system performance indicators include average power consumption, average temperature at key points, and average flow rate. Seasonal and periodic features are then extracted from the timestamp sequences using time-series databases or feature engineering methods. These seasonal and periodic features help identify patterns in system changes over time, such as seasonal temperature variations or periodic load fluctuations. These features enable the identification of the liquid cooling system's behavior patterns under different environmental conditions and operating cycles. The extracted seasonal and periodic features, along with the external operating condition data and system performance indicators, constitute the training feature set. This processed training feature set is then input into a LSM network model for training. The LSM network model learns the sequence relationships in historical data, automatically capturing the long-term evolution patterns between load, environment, and system performance.

[0033] The present invention is further configured to extract the state vector of the hidden layer of the long short-term memory network model as a long-term evolution mode factor; Establish a mapping network between long-term evolutionary model factors and model prediction control parameters; Long-term evolutionary model factors are converted into specific model prediction control parameter preset values ​​through a mapping relationship network; The pre-tuned values ​​of the model predictive control parameters are downloaded to the main controller to pre-tune the parameters of the model predictive control module, thereby optimizing the control strategy, generating corrected execution parameters, and adjusting the operating parameters of the actuators in the liquid cooling system. Specifically, a Long Short-Term Memory (LSTM) network model is trained. After training, the state vectors of the hidden layers of the LSM model are extracted. These state vectors are highly condensed summaries of the system's long-term behavioral characteristics, i.e., long-term evolutionary pattern factors. The hidden layer state vectors contain historical information about the system's operation, which helps to capture long-term dependencies in the system's operation. Through these long-term evolutionary pattern factors, the possible future operating states of the system can be predicted. Subsequently, a mapping network between the long-term evolutionary pattern factors and the model predictive control parameters is established. The role of this network is to transform the long-term evolutionary pattern factors generated by the LSM model into specific pre-tuned values ​​of the model predictive control parameters. This mapping network uses deep learning methods to process the long-term evolutionary factors and transform them into optimized control parameters. These control parameters provide a basis for subsequent system optimization control strategies. Finally, the long-term evolutionary pattern factors are transformed into pre-tuned values ​​of control parameters through the mapping network, and these pre-tuned values ​​are downloaded to the main controller. The main controller pre-tunes the parameters of the model predictive control module based on the preset values, optimizes the control strategy, and generates corrected execution parameters. Based on these corrected execution parameters, the actuators of the liquid cooling system, such as pumps and valves, will adjust their operating parameters accordingly to ensure the system maintains optimal performance under various operating conditions. The entire implementation process utilizes time-series data analysis, deep learning techniques such as long short-term memory networks, and neural network mapping models to achieve data-driven long-term performance optimization. Example 2

[0034] Please see Figure 2 This exemplary dynamic control system for a liquid cooling system based on thermal resistance and flow resistance analysis includes: Data sensing and fusion module: collects parameters of multi-source heterogeneous sensor arrays, calculates dynamic confidence weights, and generates system state observations; Online thermal resistance and flow resistance estimation module: Based on system state observations, a thermal resistance and flow resistance coupling model framework is established and the pre-built key parameters are identified and updated in real time to obtain the updated thermal resistance and flow resistance coupling model. Model Predictive Control Module: Constructs a multi-objective optimization problem using preset model predictive control parameters, and solves the current optimal control command based on a thermal resistance-current resistance coupling model using predictive control algorithms; Intelligent optimization and long-term learning module: Collect historical operating data to establish long-term evolutionary model factors, optimize model prediction control parameters, and generate corrected execution parameters.

[0035] It should be noted that the liquid cooling system dynamic control system based on thermal resistance and flow resistance analysis provided in the above embodiments and the liquid cooling system dynamic control method based on thermal resistance and flow resistance analysis provided in the above embodiments belong to the same concept. The specific operation methods of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the liquid cooling system dynamic control system based on thermal resistance and flow resistance analysis provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0036] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis, characterized in that, include: S1: Collect parameters of the multi-source heterogeneous sensor array, calculate dynamic confidence weights, and generate system state observations; S2: Based on the system state observations, establish a thermal resistance-flow resistance coupling model framework and identify and update the pre-built key parameters online and in real time to obtain the updated thermal resistance-flow resistance coupling model. S3: Construct a multi-objective optimization problem using preset model predictive control parameters, and solve the current optimal control command using a predictive control algorithm based on a thermal resistance-current resistance coupling model. S4: Collect historical operating data to establish long-term evolutionary model factors, optimize model prediction control parameters, and generate corrected execution parameters.

2. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 1, characterized in that, S1 includes: Multi-source heterogeneous sensor arrays are deployed at each key node of the liquid cooling system flow channel to collect raw data from the data source, including the temperature, pressure and flow rate of the current data source node. Based on the original data, outlier data were removed and corrected data was generated using variance statistics and consistency tests. The stability parameter of the data source is obtained by calculating the variance of the corrected data, and the volatility parameter of the data source is obtained by calculating the rate of change of the corrected data. Calculate the dynamic confidence weight for each data source based on its stability and volatility. An autoregressive sliding calibration algorithm is used to smooth the corrected data, suppressing short-term fluctuations and generating fused data. Based on the fused data, high-confidence system state observations are generated.

3. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 1, characterized in that, S2 includes: a framework and interface definition unit and an online identification and update unit; The framework and interface definition units include: The key parameters are initialized and preset to initial values. The key parameters include: dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometry change coefficient. Obtain a pre-stored set of inherent system attribute parameters, which includes the following parameters determined by the system design and materials: heat source contact area, contact interface material properties, flow channel material, flow channel wall thickness, original surface roughness of the flow channel, flow channel geometry, and flow channel bending angle.

4. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 3, characterized in that, A real-time operating condition parameter set is constructed by collecting data through sensors. The real-time operating condition parameter set includes: coolant flow rate, coolant flow rate and coolant temperature. The hydraulic diameter of the flow channel is calculated based on the channel geometry, and the real-time viscosity and density of the coolant are calculated based on the temperature observed in the system state.

5. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 4, characterized in that, Based on the temperature parameters in the system state observation values, the physical property parameters within a preset time window are calculated. The physical property parameters include: temperature change rate, thermal conductivity, and specific heat capacity. Based on the system's inherent attribute parameter set and the real-time operating condition parameter set, the contact thermal resistance, the flow channel wall thermal resistance, and the coolant thermal conduction resistance are calculated respectively, and the total system thermal resistance is obtained by summing them. A flow resistance model is established based on the original surface roughness of the flow channel, the geometric dimensions of the flow channel, the tortuosity of the flow channel, the real-time viscosity and density of the coolant, and the flow velocity of the coolant. Using coolant flow rate as the key variable, the total thermal resistance of the system is correlated with the flow resistance model, and correction interfaces are reserved for key parameters in both models to construct a thermal resistance-flow resistance coupled model.

6. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 5, characterized in that, The online identification and update unit includes: Based on real-time system state observations and physical property parameters, the values ​​of dynamic temperature response coefficient, adaptive viscosity correction factor, and flow channel geometric change coefficient in the thermal resistance-flow resistance coupling model are estimated and updated in real time using the recursive least squares method. The recursive least squares method adaptively adjusts key parameters by minimizing the error between the thermal resistance and flow resistance values ​​predicted by the model and the actual thermal resistance and flow resistance values ​​inverted from the observations.

7. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 1, characterized in that, S3 includes: With the optimization objectives of minimizing total system power consumption and temperature deviation at key points, a multi-objective optimization problem is constructed by predicting control parameters using a pre-set model. Set the feasible range of system flow, pressure, and control variables as constraints for the objective optimization problem; Based on real-time updated system state observations and an online-updated thermal resistance-current resistance coupling model, the optimal control command sequence in the finite time domain is solved. Apply immediate control commands from the optimal control command sequence to the actuators of the liquid cooling system.

8. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 1, characterized in that, S4 includes: Regularly collect and store timestamp sequences, external operating condition data, and system performance indicators from the long-term historical operating data of the system. The external operating condition data includes: ambient temperature, ambient humidity, and business load cycle. The system performance indicators include: average power consumption, average temperature of key points, and average flow rate. Seasonal and periodic features are extracted from the timestamp sequence and combined with external operating condition data and system performance indicators to form a training feature set; Based on the training feature set, the long-term evolution of the system's operating mode is learned through a long short-term memory network model.

9. The dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis according to claim 8, characterized in that, Extract the state vectors of the hidden layers of the Long Short-Term Memory network model as long-term evolution mode factors; Establish a mapping network between long-term evolutionary model factors and model prediction control parameters; Long-term evolutionary model factors are converted into specific model prediction control parameter preset values ​​through a mapping relationship network; The model predictive control parameters are pre-tuned and downloaded to the main controller. The parameters of the model predictive control module are pre-tuned to optimize the control strategy, generate corrected execution parameters, and adjust the operating parameters of the actuators of the liquid cooling system.

10. A dynamic control system for a liquid cooling system based on thermal resistance and flow resistance analysis, used to implement the dynamic control method for a liquid cooling system based on thermal resistance and flow resistance analysis as described in any one of claims 1-9, characterized in that, include: Data sensing and fusion module: collects parameters of multi-source heterogeneous sensor arrays, calculates dynamic confidence weights, and generates system state observations; Online thermal resistance and flow resistance estimation module: Based on system state observations, a thermal resistance and flow resistance coupling model framework is established and the pre-built key parameters are identified and updated in real time to obtain the updated thermal resistance and flow resistance coupling model. Model Predictive Control Module: Constructs a multi-objective optimization problem using preset model predictive control parameters, and solves the current optimal control command based on a thermal resistance-current resistance coupling model using predictive control algorithms; Intelligent optimization and long-term learning module: Collect historical operating data to establish long-term evolutionary model factors, optimize model prediction control parameters, and generate corrected execution parameters.

Images