A heat exchanger adaptive regulation method, system and medium

Abstract

This invention relates to the field of heat exchanger technology, and in particular to an adaptive control method, system, and medium for heat exchangers. The method includes: constructing a thermodynamic entropy production matrix to identify the dominant dissipation mechanism and its spatial distribution weights; adjusting the spatial pose of the flow guiding components according to the distribution weights to optimize the field cooperation angle and reconstruct the flow resistance characteristics; using the reconstructed flow resistance characteristics and physical property parameters to drive a digital twin for transient simulation, predicting the spatiotemporal evolution trend of the equivalent fouling thermal resistance; introducing the evolution trend as a penalty factor into a multi-objective optimization model to solve for the flow ratio and phase change interface control strategy that minimizes global entropy production; outputting control commands based on the control strategy, and using actual operating data to online correct the boundary conditions of the digital twin and the configuration function of the optimization model, thereby achieving closed-loop control of energy loss tracing, active flow field intervention, fouling prediction, and model self-evolution, improving the energy efficiency and operational reliability of the heat exchanger throughout its entire life cycle.

Description

Technical Field

[0001] This invention relates to the field of heat exchanger technology, and in particular to a heat exchanger adaptive control method, system, and medium. Background Technology

[0002] As core unit equipment in process industries such as chemical, pharmaceutical, and power plants, the energy efficiency and reliability of heat exchangers directly affect the overall performance of the entire production system. Traditional heat exchanger control methods are mostly based on preset operating points using PID regulation or simple logic control, which are insufficient to cope with complex variable operating conditions caused by factors such as raw material batch variations, ambient temperature fluctuations, and equipment aging. While some control strategies based on data acquisition and feedback have emerged in existing technologies, they are essentially reactive, intervening only after deviations from the setpoint, failing to address energy quality degradation at its source. Furthermore, these methods typically treat heat exchange efficiency and fouling thermal resistance as two independent control objectives, failing to consider their strong coupling relationship, leading to inconsistencies in the control strategy and hindering the achievement of global optimization. At a deeper level, existing technologies lack the ability to identify the internal energy dissipation mechanisms of heat exchangers (such as viscous dissipation and irreversible heat conduction losses) in real time, making it impossible to accurately pinpoint the root cause of efficiency losses, let alone utilize this mechanistic information to actively intervene in fluid flow patterns to optimize field-coordinated relationships. Furthermore, traditional methods for assessing the evolution of fouling on heat exchange surfaces largely rely on periodic shutdowns for cleaning or empirical judgments based on pressure differentials. These methods lack the ability to predict the spatiotemporal distribution and growth trends of fouling, leading to delayed maintenance decisions and often resulting in action only after performance has significantly deteriorated. Therefore, there is an urgent need for an adaptive control method for heat exchangers that addresses the thermodynamic essence and achieves synergistic optimization of energy efficiency and reliability. This would solve the problems of response lag, lack of understanding of underlying mechanisms, insufficient predictive capabilities, and fragmented control objectives in existing technologies. Summary of the Invention

[0003] This invention overcomes the shortcomings of the prior art and provides an adaptive control method, system and medium for heat exchangers.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention discloses an adaptive control method for a heat exchanger, comprising the following steps: Real-time acquisition of temperature and pressure field distribution data across the entire heat exchanger domain; construction of the fluid's thermodynamic entropy production matrix; quantitative identification of the dominant dissipation mechanisms leading to energy quality loss and their spatial distribution weights. Based on the distribution weight of the dominant dissipation mechanism, the spatial pose of the flow guiding components in the internal flow channel of the heat exchanger is adjusted to change the field cooperation angle of the fluid, optimize the cooperation between the velocity field and the temperature gradient, and simultaneously reconstruct the flow resistance characteristics. By utilizing the reconstructed flow resistance characteristics and real-time acquired physical property parameters, transient simulation of the heat exchanger digital twin is performed to predict the spatiotemporal evolution trend of the equivalent fouling thermal resistance of the heat exchange surface under the current operating boundary. The predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is used as a penalty factor and introduced into a multi-objective optimization model with thermodynamic perfection as the goal. The flow ratio of the hot and cold fluid channels and the control strategy of the phase change interface that minimizes global entropy production are solved in real time. Based on the aforementioned control strategy, control commands are output to adjust the actuator, and the boundary conditions of the digital twin and the configuration function of the multi-objective optimization model are corrected online using the actual operating data after adjustment.

[0005] Furthermore, real-time acquisition of temperature and pressure field distribution data across the entire heat exchanger domain is used to construct a thermodynamic entropy production matrix for the fluid, quantitatively identifying the dominant dissipation mechanisms leading to energy quality loss and their spatial distribution weights, specifically: Based on real-time acquired global temperature and pressure field distribution data of the heat exchanger, a three-dimensional temperature gradient field and a three-dimensional velocity vector field are generated through spatial interpolation and gradient reconstruction. The three-dimensional temperature gradient field and the three-dimensional velocity vector field are subjected to tensor dot product operation, and combined with the local physical properties of the fluid, a local entropy yield distribution field characterizing the intensity of irreversible loss at various spatial locations inside the heat exchanger is constructed. By integrating the local entropy production rate distribution field along the three-dimensional spatial coordinates of the heat exchanger, a total entropy production matrix covering viscous entropy production and thermal conduction entropy production is obtained. The total entropy production matrix is ​​orthogonally decomposed and spectral analyzed. By separating the entropy production contribution values ​​corresponding to different dissipation mechanisms, the dominant dissipation mechanisms that lead to energy quality loss are identified. Based on the proportion of each dominant dissipation mechanism in the total entropy production, the spatial distribution weight coefficient of the dominant dissipation mechanism is determined.

[0006] Furthermore, based on the distribution weight of the dominant dissipation mechanism, the spatial pose of the flow guiding components in the internal flow channel of the heat exchanger is adjusted to change the field cooperation angle of the fluid, optimize the cooperation between the velocity field and the temperature gradient, and simultaneously reconstruct the flow resistance characteristics, specifically: Based on the distribution weight of the dominant dissipation mechanism, the spatial coordinate set with local entropy production exceeding a preset threshold is extracted as the key control area, and a streamline curvature gradient distribution map is generated based on the velocity vector field within the area. The streamline curvature gradient distribution map is compared point by point with the reference streamline profile pre-generated based on the principle of minimum entropy production, and the required streamline deflection angle field at each control position is calculated. Based on the streamline deflection angle field, and combined with the kinematic mapping relationship between the local velocity vector at the installation position of the guide component and the adjustable parameters of the component, the target angle of attack adjustment amount and execution sequence of each guide component are determined, and the guide component pose control command is formed. The orientation control command of the flow guide component is sent to the actuator to drive the flow guide component to change its spatial orientation and induce a controllable secondary flow in the flow channel, thereby adjusting the direction of the velocity vector so that the field coordination angle between the velocity field and the temperature gradient field approaches the theoretical optimal value. Real-time acquisition of local pressure field data after adjustment; by comparing the pressure difference and flow rate relationship between the inlet and outlet of the flow channel before and after adjustment, calculation of the change in flow resistance coefficient of each flow section; and online correction of the global flow resistance characteristic curve of the heat exchanger based on the change in flow resistance coefficient, thus completing the reconstruction of the flow resistance characteristics.

[0007] Furthermore, using the reconstructed flow resistance characteristics and real-time acquired physical property parameters, transient simulations are performed on the digital twin of the heat exchanger to predict the spatiotemporal evolution trend of the equivalent fouling thermal resistance on the heat exchange surface under the current operating boundary conditions, specifically: Based on the reconstructed flow resistance characteristics and real-time acquired physical property parameters, the instantaneous flow distribution data of each flow channel of the current heat exchanger and the fluid inlet and outlet temperature data are used as boundary conditions and loaded into the pre-constructed digital twin of the heat exchanger. The digital twin is embedded with a transient fouling deposition kinetic model based on the heat and mass transfer mechanism. By using the transient fouling deposition kinetic model and combining the wall shear force distribution data reflected by the flow resistance characteristics, the adhesion probability and erosion rate of particulate matter in the fluid on the heat exchange surface are determined, and the initial distribution density of point-like fouling nuclei at each spatial location on the heat exchange surface under the current operating boundary is generated. Taking the initial distribution density of the point-like dirt cores as the starting point of evolution, the crystal growth dynamics module in the digital twin is invoked, and the real-time collected wall temperature and near-wall fluid concentration are used as the driving potential to iteratively calculate the thickness growth rate of the dirt layer in the direction perpendicular to the wall and the coverage expansion rate in the direction along the wall. Based on the iterative calculation results of the thickness growth rate and coverage expansion rate, and combined with the physical property parameter library of the fouling layer, the equivalent fouling thermal resistance value of each grid node on the heat exchange surface is dynamically updated to generate a non-uniformly distributed thermal resistance network composed of fouling layers of different thicknesses and densities. Spatial integration and temporal recursion are performed on the thermal resistance network along the heat exchange surface to obtain the equivalent spatiotemporal evolution trend of fouling thermal resistance covering the entire process from initial formation to stable growth of fouling. The spatiotemporal evolution trend is presented in the form of a thermal resistance spatiotemporal evolution cloud map, which is used to characterize the thermal resistance change trajectory of each region of the heat exchange surface within a future preset time period.

[0008] Furthermore, the predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is used as a penalty factor and introduced into a multi-objective optimization model with thermodynamic perfection as the objective. This allows for the real-time solution of the flow ratio of the hot and cold fluid channels and the control strategy of the phase change interface that minimizes global entropy production. Specifically: The predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is spatially convolved with the real-time collected heat flux field distribution data to generate a dynamic penalty function matrix that varies with the position of the grid nodes on the heat exchange surface. The dynamic penalty function matrix is ​​embedded as a penalty term into an optimization model with the goal of minimizing total entropy yield, and a thermodynamic perfection evaluation function that includes the influence of fouling thermal resistance is constructed. The sensitivity analysis of the thermodynamic perfection evaluation function with respect to the valve opening degree and phase change interface height of the hot and cold fluid channels was performed using the adjoint variable method, and the descent gradient field of each operating variable with respect to the objective function was generated. Based on the descent gradient field, a non-dominated solution search based on the Pareto front is performed in the feasible region of the operational variable space to obtain a set of candidate control parameters that simultaneously make the total entropy yield and the integral value of the dynamic penalty function approach optimal. The candidate control parameter set is input into the heat exchanger digital twin for transient simulation verification. The temperature fluctuation characteristic value and pressure pulsation amplitude corresponding to each candidate parameter are extracted, and the inferior solutions that induce local thermal stress concentration or flow field instability are eliminated. The parameter combination that minimizes the spatiotemporal integral value of the dynamic penalty function matrix and the global entropy yield is selected from the remaining candidate parameters and used as the final cold and hot fluid channel flow ratio and phase change interface control strategy.

[0009] Furthermore, based on the aforementioned control strategy, control commands are output to adjust the actuators, and the boundary conditions of the digital twin and the configuration function of the multi-objective optimization model are corrected online using the adjusted actual operating data, specifically as follows: Based on the aforementioned control strategy, the flow ratio of the hot and cold fluid channels and the phase change interface control strategy are parsed into an executable control command sequence, and the control command sequence is sent down to the underlying actuator for driving. Real-time acquisition of updated temperature field and updated pressure field distribution data of the entire heat exchanger after regulation, and input of the updated temperature field and updated pressure field distribution data into the heat exchanger digital twin, and perform point-by-point residual calculation with the predicted output data of the digital twin under the current regulation strategy to generate a residual field containing the temperature deviation value and pressure deviation value of each spatial node. The residual field is correlated with the valve opening change and phase change interface displacement corresponding to the control strategy. The parameter correction gradient of the internal physical mechanism model of the digital twin is calculated by the backpropagation algorithm. Based on the parameter correction gradient, the wall heat transfer coefficient distribution map and the flow resistance coefficient distribution map in the digital twin are updated online to complete the adaptive calibration of the boundary conditions of the digital twin. The calibrated digital twin is re-run under the current operating conditions, and the newly predicted heat flux field distribution data is extracted. The newly predicted heat flux field distribution data is then compared with the actual heat flux field distribution data collected in real time after regulation, and the model mismatch error vector caused by the change in regulation strategy is calculated. Based on the model mismatch error vector, the weight coefficient allocation rule of the thermodynamic perfection evaluation function embedded in the multi-objective optimization model is modified to generate an updated configuration function, so that the configuration function can accurately reflect the actual thermodynamic characteristics of the current heat exchanger under the updated boundary conditions.

[0010] The control command sequence includes the valve position opening setting value of each regulating valve and the displacement setting value of the phase change interface height adjuster.

[0011] The physical properties include the fluid's density, dynamic viscosity, thermal conductivity, specific heat capacity, surface tension, diffusion coefficient, and latent heat of vaporization.

[0012] The second aspect of the present invention discloses a heat exchanger adaptive control system, including a memory and a processor. The memory stores a heat exchanger adaptive control method program. When the heat exchanger adaptive control method program is executed by the processor, any one of the steps of the heat exchanger adaptive control method is implemented.

[0013] The third aspect of the present invention discloses a computer-readable storage medium storing a heat exchanger adaptive control method program, wherein when the heat exchanger adaptive control method program is executed by a processor, any of the steps of the heat exchanger adaptive control method are implemented.

[0014] This invention addresses the technical deficiencies in the prior art and has the following beneficial effects: This invention achieves precise location of energy loss sources within heat exchangers, fundamentally identifying the causes and locations of inefficiencies and providing a physical basis for targeted regulation. Through proactive intervention in the flow field morphology, it improves the synergistic matching between the velocity field and temperature gradient, enhancing heat exchange efficiency without increasing additional power consumption. Furthermore, it possesses the ability to predict the fouling growth process on heat exchange surfaces, transforming traditional passive response maintenance into mechanism-based proactive suppression, significantly delaying performance degradation. In addition, by incorporating the future evolution trend of fouling thermal resistance into optimization decisions, it effectively balances the conflict between immediate energy efficiency and long-term anti-fouling, minimizing overall operating costs. The system has self-learning and self-evolution capabilities based on actual operating data, dynamically updating its internal models and decision logic as equipment ages and operating conditions change, ensuring the output of optimal regulation strategies throughout its entire lifecycle, thus improving the energy efficiency, safety, and reliability of heat exchanger operation. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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 embodiments can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the adaptive control method for this heat exchanger; Figure 2 This is a framework diagram of the adaptive control system for this heat exchanger. Detailed Implementation

[0017] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0018] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0019] like Figure 1 As shown, the first aspect of this invention discloses an adaptive control method for a heat exchanger, comprising the following steps: Real-time acquisition of temperature and pressure field distribution data across the entire heat exchanger domain; construction of the fluid's thermodynamic entropy production matrix; quantitative identification of the dominant dissipation mechanisms leading to energy quality loss and their spatial distribution weights. Based on the distribution weight of the dominant dissipation mechanism, the spatial pose of the flow guiding components in the internal flow channel of the heat exchanger is adjusted to change the field cooperation angle of the fluid, optimize the cooperation between the velocity field and the temperature gradient, and simultaneously reconstruct the flow resistance characteristics. By utilizing the reconstructed flow resistance characteristics and real-time acquired physical property parameters, transient simulation of the heat exchanger digital twin is performed to predict the spatiotemporal evolution trend of the equivalent fouling thermal resistance of the heat exchange surface under the current operating boundary. The predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is used as a penalty factor and introduced into a multi-objective optimization model with thermodynamic perfection as the goal. The flow ratio of the hot and cold fluid channels and the control strategy of the phase change interface that minimizes global entropy production are solved in real time. Based on the aforementioned control strategy, control commands are output to adjust the actuator, and the boundary conditions of the digital twin and the configuration function of the multi-objective optimization model are corrected online using the actual operating data after adjustment.

[0020] The physical properties include the fluid's density, dynamic viscosity, thermal conductivity, specific heat capacity, surface tension, diffusion coefficient, and latent heat of vaporization.

[0021] Furthermore, real-time acquisition of temperature and pressure field distribution data across the entire heat exchanger domain is used to construct a thermodynamic entropy production matrix for the fluid, quantitatively identifying the dominant dissipation mechanisms leading to energy quality loss and their spatial distribution weights, specifically: Based on real-time acquired global temperature and pressure field distribution data of the heat exchanger, a three-dimensional temperature gradient field and a three-dimensional velocity vector field are generated through spatial interpolation and gradient reconstruction. The three-dimensional temperature gradient field and the three-dimensional velocity vector field are subjected to tensor dot product operation, and combined with the local physical properties of the fluid, a local entropy yield distribution field characterizing the intensity of irreversible loss at various spatial locations inside the heat exchanger is constructed. By integrating the local entropy production rate distribution field along the three-dimensional spatial coordinates of the heat exchanger, a total entropy production matrix covering viscous entropy production and thermal conduction entropy production is obtained. The total entropy production matrix is ​​orthogonally decomposed and spectral analyzed. By separating the entropy production contribution values ​​corresponding to different dissipation mechanisms, the dominant dissipation mechanisms that lead to energy quality loss are identified. Based on the proportion of each dominant dissipation mechanism in the total entropy production, the spatial distribution weight coefficient of the dominant dissipation mechanism is determined.

[0022] In practical implementation, a distributed sensor array deployed at key sections of the heat exchanger is used to capture raw temperature and pressure signals across the entire area in real time. Through spatial interpolation algorithms, discrete measurement data are reconstructed into a continuously distributed three-dimensional temperature gradient field and a three-dimensional velocity vector field. This process essentially establishes a high-fidelity digital image of the complex flow and heat transfer state inside the heat exchanger. Based on this, the second law of thermodynamics is introduced to couple the aforementioned three-dimensional field data with the local physical properties of the fluid. Specifically, by calculating the tensor dot product between the temperature gradient field and the velocity vector field, and combining this with the fluid's thermal conductivity and dynamic viscosity, the intensity of irreversible losses within each tiny control volume can be quantified, thereby constructing a local entropy yield distribution field covering the entire heat exchange space. It is worth emphasizing that this distribution field is not a single numerical value, but a vector field that includes contributions from both viscous dissipation and thermal conduction dissipation.

[0023] To separate the contributions of different physical mechanisms from the mixed dissipation signals, the total entropy production matrix obtained by integrating the local entropy production rate distribution field is decomposed into a series of mutually orthogonal characteristic modes, each corresponding to a specific energy dissipation mode. Since viscous entropy production and thermal conduction entropy production have different characteristic spectra in the frequency domain or spatial structure, analyzing the energy proportion and spatial distribution of these orthogonal modes can automatically identify the core mechanism dominating energy quality loss under the current operating conditions. For example, if the characteristic modes in the high-frequency or near-wall region are dominant, viscous dissipation can be identified as the main factor; if the characteristic modes in the region with a severe temperature gradient in the mainstream region contribute more, thermal conduction dissipation is dominant. Finally, based on the contribution ratio of each dominant dissipation mechanism to the total entropy production, a set of quantified spatial distribution weight coefficients can be output. These coefficients directly indicate which region of the heat exchanger and which dissipation mechanism should be targeted for intervention, thus providing a precise physical basis for the adaptive adjustment of the flow channel configuration.

[0024] Furthermore, based on the distribution weight of the dominant dissipation mechanism, the spatial pose of the flow guiding components in the internal flow channel of the heat exchanger is adjusted to change the field cooperation angle of the fluid, optimize the cooperation between the velocity field and the temperature gradient, and simultaneously reconstruct the flow resistance characteristics, specifically as follows: Based on the distribution weight of the dominant dissipation mechanism, the spatial coordinate set with local entropy production exceeding a preset threshold is extracted as the key control area, and a streamline curvature gradient distribution map is generated based on the velocity vector field within the area. The streamline curvature gradient distribution map is compared point by point with the reference streamline profile pre-generated based on the principle of minimum entropy production, and the required streamline deflection angle field at each control position is calculated. Based on the streamline deflection angle field, and combined with the kinematic mapping relationship between the local velocity vector at the installation position of the guide component and the adjustable parameters of the component, the target angle of attack adjustment amount and execution sequence of each guide component are determined, and the guide component pose control command is formed. The orientation control command of the flow guide component is sent to the actuator to drive the flow guide component to change its spatial orientation and induce a controllable secondary flow in the flow channel, thereby adjusting the direction of the velocity vector so that the field coordination angle between the velocity field and the temperature gradient field approaches the theoretical optimal value. Real-time acquisition of local pressure field data after adjustment; by comparing the pressure difference and flow rate relationship between the inlet and outlet of the flow channel before and after adjustment, calculation of the change in flow resistance coefficient of each flow section; and online correction of the global flow resistance characteristic curve of the heat exchanger based on the change in flow resistance coefficient, thus completing the reconstruction of the flow resistance characteristics.

[0025] It should be noted that, based on the distribution weight of the dominant dissipation mechanism, a local entropy production rate threshold is set. For example, three times the average local entropy production rate across the entire field can be used as the trigger threshold. All spatial coordinates exceeding this threshold are extracted, and the set of these coordinates is defined as the key control zone, representing the flow channel region where energy dissipation is most concentrated and requires intervention. For the velocity vector field within this region, a streamline curvature gradient distribution map is generated by calculating the rate of change of the streamline tangent direction, visually reflecting the degree and direction of fluid deviation from the ideal flow state within the key control zone. To evolve the flow state towards a more thermodynamically optimal direction, the system pre-stores a set of baseline streamline profiles generated offline based on the principle of minimum entropy production. These profiles represent the ideal streamline shape with the theoretically lowest energy loss under the current boundary conditions. By comparing the measured streamline curvature gradient distribution map with this baseline profile point by point in space, the angle and direction of actual streamline deflection at each spatial location are calculated, thus constructing a streamline deflection angle field covering the key control zone.

[0026] After obtaining the ideal angle field, the key is how to implement it onto specific mechanical components. This method solves this problem through a pre-defined kinematic mapping relationship: this mapping relationship is essentially a lookup table or transformation matrix that records the correspondence between the geometric pose (such as angle of attack and penetration depth) of each adjustable flow guide component in the heat exchanger at its installation position and the resulting local velocity vector deflection. Based on this, the target deflection value of the corresponding component installation point in the streamline deflection angle field is used as input, and the target angle of attack adjustment amount required for each flow guide component can be accurately calculated through inverse kinematic mapping. At the same time, according to the entropy weight of each point in the key control area, the execution order of the components is automatically arranged to ensure that the components located in the highest entropy production area are adjusted first to suppress the most important energy loss source as quickly as possible, thereby forming a complete flow guide component pose control command.

[0027] Finally, to ensure the control effect can be quantified and used for subsequent model updates, this step also includes closed-loop verification of the flow resistance characteristic reconstruction: after the actuator changes the component pose according to the command, the adjusted local pressure field data is immediately collected. By comparing the inlet and outlet pressure difference and flow rate relationship of the same flow channel before and after adjustment, the change in flow resistance coefficient of each flow segment is calculated using the flow resistance coefficient calculation formula (e.g., the ratio of pressure difference to dynamic head). This measured change is used to correct the global flow resistance characteristic curve of the heat exchanger online: specifically, the initial flow resistance characteristic curve stored in the system is usually obtained through theoretical calculation or factory calibration, but it will deviate due to factors such as changes in the medium and processing deviations during actual operation. This method uses the newly measured change in flow resistance coefficient as a calibration point to interpolate or fit the corresponding flow segment of the original curve, so that the internal model of the system reflects the real flow resistance state in real time, providing accurate boundary conditions for the next cycle of digital twin simulation.

[0028] Furthermore, using the reconstructed flow resistance characteristics and real-time acquired physical property parameters, transient simulations are performed on the digital twin of the heat exchanger to predict the spatiotemporal evolution trend of the equivalent fouling thermal resistance on the heat exchange surface under the current operating boundary conditions, specifically: Based on the reconstructed flow resistance characteristics and real-time acquired physical property parameters, the instantaneous flow distribution data of each flow channel of the current heat exchanger and the fluid inlet and outlet temperature data are used as boundary conditions and loaded into the pre-constructed digital twin of the heat exchanger. The digital twin is embedded with a transient fouling deposition kinetic model based on the heat and mass transfer mechanism. By using the transient fouling deposition kinetic model and combining the wall shear force distribution data reflected by the flow resistance characteristics, the adhesion probability and erosion rate of particulate matter in the fluid on the heat exchange surface are determined, and the initial distribution density of point-like fouling nuclei at each spatial location on the heat exchange surface under the current operating boundary is generated. Taking the initial distribution density of the point-like dirt cores as the starting point of evolution, the crystal growth dynamics module in the digital twin is invoked, and the real-time collected wall temperature and near-wall fluid concentration are used as the driving potential to iteratively calculate the thickness growth rate of the dirt layer in the direction perpendicular to the wall and the coverage expansion rate in the direction along the wall. Based on the iterative calculation results of the thickness growth rate and coverage expansion rate, and combined with the physical property parameter library of the fouling layer, the equivalent fouling thermal resistance value of each grid node on the heat exchange surface is dynamically updated to generate a non-uniformly distributed thermal resistance network composed of fouling layers of different thicknesses and densities. Spatial integration and temporal recursion are performed on the thermal resistance network along the heat exchange surface to obtain the equivalent spatiotemporal evolution trend of fouling thermal resistance covering the entire process from initial formation to stable growth of fouling. The spatiotemporal evolution trend is presented in the form of a thermal resistance spatiotemporal evolution cloud map, which is used to characterize the thermal resistance change trajectory of each region of the heat exchange surface within a future preset time period.

[0029] It should be noted that the resistance coefficients of each flow segment reflected by the flow resistance characteristic curve obtained from the previous cycle reconstruction, combined with the real-time collected fluid density and viscosity and other physical properties, are used as the current boundary conditions and loaded into the digital twin. At the same time, the instantaneous flow distribution data of each flow channel and the measured inlet and outlet temperatures of the hot and cold fluids are used as the driving boundary of the transient simulation to ensure that the simulation conditions are synchronized with the actual operating conditions.

[0030] To clearly define the determination methods of adhesion probability and erosion rate, wall shear force is introduced as a key coupling parameter. Since the flow resistance characteristic curve directly reflects the resistance characteristics of fluid flowing through the heat exchange surface, by solving this curve in conjunction with the current flow rate data, the local velocity gradient in different regions of the heat exchange surface can be calculated, thereby obtaining the spatial distribution data of wall shear force. In the fouling deposition kinetic model, adhesion probability is defined as the possibility that particles overcome the near-wall lift and contact the wall surface. It is negatively correlated with wall shear force: higher shear force means that the fluid's ability to carry particles is enhanced, and the probability of particles breaking through the boundary layer and reaching the wall surface is reduced. Erosion rate, on the other hand, characterizes the rate at which deposited fouling is washed away from the wall by the fluid. It is positively correlated with wall shear force and usually considers the age effect of the fouling layer, i.e., newly deposited soft fouling is more easily eroded. By substituting the calculated wall shear force distribution data into the aforementioned adhesion probability function and erosion rate function, the model can calculate the particle adhesion probability value and erosion rate value at each spatial grid node at the current moment. The dynamic balance between the two determines whether fouling nuclei are generated at that location and their initial density, thereby generating an initial distribution density of point-like fouling nuclei covering the entire heat exchange surface. This distribution density is not a uniform field, but reflects the modulation effect of flow resistance characteristics and flow state on the location of fouling formation. For example, high flow velocity regions have a lower initial distribution density due to large shear force, while flow stagnation regions or low shear force regions become the preferred nucleation sites for fouling nuclei.

[0031] The initial distribution density, serving as the spatial starting point for subsequent crystal growth evolution, is transmitted to the crystal growth kinetics module. Using real-time acquired wall temperature as the thermal driving potential and near-wall fluid concentration as the mass transport potential, the crystal growth equation is iteratively solved to calculate the thickening rate of the fouling layer perpendicular to the wall and its spreading rate along the wall. After each iteration, the thickness and coverage are updated, and the built-in fouling layer property parameter library is immediately invoked. This library contains thermal conductivity and porosity data for fouling with different densities and chemical compositions, converting the geometric growth into a change in thermal resistance, dynamically updating the equivalent fouling thermal resistance value of each grid node. As the iteration progresses, the originally discrete fouling nuclei gradually grow and connect, forming a non-uniform thermal resistance network with varying thicknesses and densities in space. This network reflects the real-time impact of the fouling layer on heat transfer performance. By performing spatial integration (statistical total thermal resistance burden) and temporal recursion (extrapolating the evolution state at future moments) along the heat exchange surface of the thermal resistance network, an equivalent fouling thermal resistance spatiotemporal evolution trend map containing time and spatial dimensions can be output. The trend map intuitively shows in the form of a cloud map which areas of the heat exchange surface will have significantly increased thermal resistance and which areas will remain stable in the next few hours.

[0032] Furthermore, the predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is used as a penalty factor and introduced into a multi-objective optimization model with thermodynamic perfection as the objective. This allows for the real-time solution of the flow ratio of the hot and cold fluid channels and the control strategy of the phase change interface that minimizes global entropy production. Specifically: The predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is spatially convolved with the real-time collected heat flux field distribution data to generate a dynamic penalty function matrix that varies with the position of the grid nodes on the heat exchange surface. The dynamic penalty function matrix is ​​embedded as a penalty term into an optimization model with the goal of minimizing total entropy yield, and a thermodynamic perfection evaluation function that includes the influence of fouling thermal resistance is constructed. The sensitivity analysis of the thermodynamic perfection evaluation function with respect to the valve opening degree and phase change interface height of the hot and cold fluid channels was performed using the adjoint variable method, and the descent gradient field of each operating variable with respect to the objective function was generated. Based on the descent gradient field, a non-dominated solution search based on the Pareto front is performed in the feasible region of the operational variable space to obtain a set of candidate control parameters that simultaneously make the total entropy yield and the integral value of the dynamic penalty function approach optimal. The candidate control parameter set is input into the heat exchanger digital twin for transient simulation verification. The temperature fluctuation characteristic value and pressure pulsation amplitude corresponding to each candidate parameter are extracted, and the inferior solutions that induce local thermal stress concentration or flow field instability are eliminated. The parameter combination that minimizes the spatiotemporal integral value of the dynamic penalty function matrix and the global entropy yield is selected from the remaining candidate parameters and used as the final cold and hot fluid channel flow ratio and phase change interface control strategy.

[0033] It should be noted that spatial convolution is performed between the massive node data contained in the evolution trend map and the real-time collected heat flux field data. This operation is equivalent to using the predicted distribution of fouling thermal resistance as a weight template, superimposed on the current heat flux distribution, to generate a dynamic penalty function matrix that varies with different locations on the heat transfer surface. The physical meaning of this matrix is ​​that areas predicted to experience severe fouling and a significant increase in thermal resistance are given higher penalty weights, meaning that in subsequent optimizations, it is preferable to sacrifice a small amount of heat transfer efficiency to avoid exacerbating fouling in these areas; while clean areas are given lower penalties, allowing for the pursuit of higher heat transfer efficiency.

[0034] Then, a composite thermodynamic perfection evaluation function is constructed. This function does not simply pursue the minimum entropy production, but instead superimposes the aforementioned dynamic penalty function matrix as a penalty term after the original total entropy production rate objective term. This gives the optimization model predictive capabilities: when a candidate control strategy, while reducing current entropy production, leads to a sharp increase in the thermal resistance penalty weight in certain regions in the future, the overall evaluation value of that strategy will not be optimal, and it will be automatically excluded by the model. To efficiently solve this complex optimization problem, the adjoint variable method is introduced for sensitivity analysis. The adjoint variable method is essentially an efficient gradient calculation method, particularly suitable for physical field optimization problems with many input variables and complex constraints. By performing a reverse solution of the thermodynamic perfection evaluation function in a digital twin, all operational variables—namely, the valve openings and phase change interface heights of each hot and cold fluid channel—can be obtained at once, along with the descending gradient field of the objective function. This gradient field indicates which valve to adjust and by how much to raise or lower the phase change interface in the current state to most effectively reduce the overall evaluation value.

[0035] Guided by gradient information, a non-dominated solution search based on the Pareto front is performed within the feasible region of the manipulated variables. The feasible region is defined by hard constraints such as valve physical stroke and the safety height of the phase change interface. The Pareto search aims to address the potentially conflicting objectives of "reducing entropy production" and "reducing fouling penalty," finding all compromise solutions that cannot improve one objective without worsening the other, thus forming a set of candidate control parameters. However, the mathematically optimal solution is not necessarily the safe solution in engineering. Therefore, this step sets up a safety verification barrier: the candidate parameter set is sent back to the heat exchanger digital twin for high-precision transient simulation verification, focusing on observing the temperature fluctuation characteristics and pressure pulsation amplitude under each parameter. If a candidate parameter causes an excessively large local temperature gradient (e.g., the rate of temperature change at a certain point exceeds the allowable threshold for thermal stress) or causes severe oscillations in the flow field (pressure pulsation amplitude exceeds the stable operating limit), it is determined that it will induce local thermal stress concentration or flow field instability and is directly eliminated as a suboptimal solution. Finally, from the remaining candidate parameters that have passed safety verification, the parameter combination that minimizes the spatiotemporal integral value of the dynamic penalty function matrix and the global entropy production rate is selected as the final control strategy output. This strategy ensures optimal energy efficiency at present while minimizing the negative impact of future fouling and posing no threat to equipment safety.

[0036] Furthermore, based on the aforementioned control strategy, control commands are output to adjust the actuators, and the boundary conditions of the digital twin and the configuration function of the multi-objective optimization model are corrected online using the adjusted actual operating data, specifically as follows: Based on the aforementioned control strategy, the flow ratio of the hot and cold fluid channels and the phase change interface control strategy are parsed into an executable control command sequence, and the control command sequence is sent down to the underlying actuator for driving. Real-time acquisition of updated temperature field and updated pressure field distribution data of the entire heat exchanger after regulation, and input of the updated temperature field and updated pressure field distribution data into the heat exchanger digital twin, and perform point-by-point residual calculation with the predicted output data of the digital twin under the current regulation strategy to generate a residual field containing the temperature deviation value and pressure deviation value of each spatial node. The residual field is correlated with the valve opening change and phase change interface displacement corresponding to the control strategy. The parameter correction gradient of the internal physical mechanism model of the digital twin is calculated by the backpropagation algorithm. Based on the parameter correction gradient, the wall heat transfer coefficient distribution map and the flow resistance coefficient distribution map in the digital twin are updated online to complete the adaptive calibration of the boundary conditions of the digital twin. The calibrated digital twin is re-run under the current operating conditions, and the newly predicted heat flux field distribution data is extracted. The newly predicted heat flux field distribution data is then compared with the actual heat flux field distribution data collected in real time after regulation, and the model mismatch error vector caused by the change in regulation strategy is calculated. Based on the model mismatch error vector, the weight coefficient allocation rule of the thermodynamic perfection evaluation function embedded in the multi-objective optimization model is modified to generate an updated configuration function, so that the configuration function can accurately reflect the actual thermodynamic characteristics of the current heat exchanger under the updated boundary conditions. The control command sequence includes the valve position opening setting value of each regulating valve and the displacement setting value of the phase change interface height adjuster.

[0037] It should be noted that the control strategy (i.e. the flow ratio of the hot and cold fluid channels and the phase change interface height) obtained from the previous cycle is parsed into a sequence of control commands that can be recognized by the underlying actuator. This sequence includes the valve position opening setting value of each regulating valve and the displacement setting value of the phase change interface height adjuster. These commands are then sent to the actuator via fieldbus or industrial Ethernet to complete the drive and achieve physical-level adjustment of the heat exchanger's operating status.

[0038] Once the control command is executed, the model calibration process is immediately initiated. At this time, updated temperature and pressure field distribution data for the entire heat exchanger domain after control are collected in real time. These two sets of data represent the actual response of the heat exchanger to the control strategy in the real physical world. After inputting these data into the digital twin, the predicted temperature and pressure fields output by the digital twin at the same moment based on the boundary conditions of the previous cycle are extracted and compared point by point to generate a residual field covering all grid nodes of the entire heat exchange space. This residual field quantifies the deviation between the model prediction and physical reality. To trace the source of this deviation, the residual field is correlated with the valve opening change and phase change interface displacement corresponding to the current control strategy. Simply put, by analyzing which change in the executed action caused an increase in the deviation in which spatial region, it is possible to deduce which physical parameters in the model are no longer accurately described. Using this correlation, the backpropagation algorithm is invoked, and the residual field is used as a loss signal to propagate back along the computational graph of the digital twin. The required correction gradients for the key physical field parameters in the model are calculated, and the distribution maps of the wall heat transfer coefficient and the flow resistance coefficient are updated online accordingly, so that the boundary conditions of the digital twin are realigned with the actual heat exchanger.

[0039] After completing the above calibration, the digital twin can accurately reproduce the current heat exchange state. However, the ultimate goal of this step is to correct the optimization model used for decision-making. Therefore, the calibrated digital twin is run again under the current operating conditions, and the newly predicted heat flux field distribution data is extracted. This data is then compared a second time with the actual heat flux field distribution data collected in real time after adjustment. This comparison is no longer used to correct the physical model, but rather to solve a higher-level deviation. Due to the slow drift of the actual thermodynamic characteristics of the heat exchanger, the allocation rules of the weight coefficients in the original thermodynamic perfection evaluation function are no longer optimal. By statistically analyzing the deviation distribution between the newly predicted heat flux field and the actual heat flux field, the least squares identification algorithm is used to calculate the model mismatch error vector. This vector directly indicates which objective terms in the evaluation function are overestimated and which are underestimated. Based on this error vector, the weight coefficient allocation rules of the thermodynamic perfection evaluation function embedded in the multi-objective optimization model are corrected, generating an updated configuration function. This correction enables the optimization model to accurately reflect the true thermodynamic characteristics of the heat exchanger under updated boundary conditions when performing strategy optimization in the next cycle. For example, if actual operating data shows that the impact of fouling is more serious than expected, the configuration function will automatically increase the weight of the fouling penalty term, thereby paying more attention to suppressing fouling growth in subsequent regulation and control, and ultimately achieving synchronous evolution of the decision model and the physical entity.

[0040] In addition, this method may also include the following steps in specific applications: The temperature fluctuation time sequence, pressure pulsation spectrum and electrochemical mobility of fouling ions in the micro-region near the wall of the heat exchanger channel are collected in real time to construct a phase space reconstruction vector characterizing the dynamic adsorption state of the solid-liquid interface. The phase space reconstruction vector is input into the tight-binding approximation model based on density functional theory. The relationship between the total energy of the system and the atomic configuration is solved iteratively to calculate the adiabatic potential energy surface distribution of the interaction between fluid molecules and wall material under the current working condition. Perform surface topology analysis on the adiabatic potential energy surface distribution to locate local maxima and local minima on the potential energy surface, and generate a potential energy ridge network characterizing the molecular adsorption path by connecting the fastest descent paths between each extreme point. Graph theory analysis is performed on the potential energy ridge network to identify accessible adsorption channels that connect adjacent potential wells and whose saddle point height is lower than the preset nucleation energy barrier threshold. The wall regions corresponding to the accessible adsorption channels are marked as preferential nucleation regions for dirt molecules. The potential well depth and saddle height values ​​of each grid node in the preferred nucleation region are weighted and summed to generate the charge injection intensity coefficient required for each node. Suppression charges are injected into the corresponding region according to the charge injection intensity coefficient required for each node, and the potential energy surface morphology of the solid-liquid interface is reconstructed to block the passable adsorption channels.

[0041] It should be noted that for some high-value-added materials, highly fouling media, or pharmaceutical and fine chemical applications with extremely high hygiene requirements, once fouling nuclei form and grow on the wall surface, subsequent cleaning is not only energy-intensive and costly, but may even affect product quality due to residual cleaning media. To address this pain point, this embodiment, based on the aforementioned macroscopic entropy production control and mesoscopic fouling evolution prediction, further introduces a microscopic solid-liquid interface control mechanism, aiming to block the initial adsorption pathway of fouling molecules at the source.

[0042] In practical applications, microelectrode arrays and electrochemical sensors are deployed in the near-wall micro-regions of the heat exchanger flow channels to capture the dynamic behavior characteristics at the fluid-wall interface in real time. Unlike macroscopic temperature and pressure monitoring, these sensors collect temperature fluctuation time series (reflecting the disturbance of the near-wall thermal boundary layer), pressure pulsation spectrum (reflecting the impact frequency of turbulent bursts on the wall), and electrochemical mobility of fouling ions (reflecting the mobility of charged particles in the electric double layer). By reconstructing the phase space of these three types of characteristic parameters reflecting the dynamic adsorption state of the interface, a high-dimensional vector that uniquely describes the current energy state of the solid-liquid interface can be generated. The significance of this vector lies in its transformation of macroscopically measurable fluid dynamics and electrochemical signals into the boundary conditions required for microscopic molecular dynamics simulations.

[0043] The built-in tight-binding approximation model based on density functional theory is invoked to calculate the potential energy surface at the molecular level. Using the aforementioned phase space reconstruction vector as input parameters, the total energy of the system consisting of fluid molecules and metal atoms on the wall in the near-wall micro-region is iteratively solved. Specifically, this process is equivalent to simulating in a computer the entire process of a fluid molecule tentatively approaching the wall and attempting to find a stable adsorption site. By solving for the total energy of the system at different atomic spacings, an adiabatic potential energy surface distribution curve describing the change of interaction forces with distance is obtained, which is represented as an undulating potential energy surface topography map in three-dimensional space. Topological analysis of this potential energy surface automatically identifies the peaks (local maxima, i.e., energy barriers that molecules cannot easily overcome) and valleys (local minima, i.e., adsorption sites where molecules may reside). By connecting the fastest descending paths, a network of potential energy ridges is generated, running through the valleys and saddles, to reveal all possible paths for molecules to move from the free state near the wall to the adsorbed state on the wall.

[0044] The potential energy ridge network is abstracted as a graph theory model, with each adsorption site considered as a node, the ridge connecting the nodes as edges, and the saddle point height as the edge weight. A graph search algorithm identifies all edges that can connect adjacent potential wells and whose saddle point height is below a preset nucleation energy barrier threshold. The physical paths corresponding to these edges are the adsorption channels that molecules can actually pass through, and the wall regions they cover are the preferential nucleation regions for dirt molecules, thus accurately pinpointing the initial formation location and path of dirt. Based on the identified preferential nucleation regions, the energy intensity required for intervention is further quantified. By weighted summing of the potential well depth (representing the energy released after molecule adsorption, with deeper depth indicating stronger adsorption) and saddle point height (representing the energy barrier to overcome to enter the site, with lower height indicating easier entry) of each grid node within the region, the required charge injection intensity coefficient for each node is calculated. Based on this coefficient, a specific waveform and intensity of charge are applied to the wall surface of the corresponding region, reconstructing the double layer structure and potential energy surface morphology of the solid-liquid interface. This raises saddle points that were originally below the nucleation energy barrier and fills in previously deeper potential wells, thereby physically blocking the passable adsorption channels. This process is equivalent to actively altering the topography of the landing field at the moment a molecule is about to land on the wall, preventing the molecule from finding a stable landing point and causing it to be drawn back into the main fluid. This represents a technological leap from passive cleaning after macroscopic thermal resistance prediction to active inhibition before molecule nucleation.

[0045] In addition, this method may also include the following steps in specific applications: Real-time acquisition of near-wall fluid temperature and velocity gradient distribution at the flow channel wall of heat exchanger; combined with fluid composition and physical property parameters under current operating conditions, constructing force field parameter boundary conditions characterizing intermolecular interaction forces at the solid-liquid interface. The force field parameters and boundary conditions are input into a pre-built molecular dynamics simulator. The number density distribution curve of fluid molecules in the direction of the wall normal is used as the objective function. The arrangement configuration of fluid molecules at the solid-liquid interface that makes the number density distribution curve match the actual collected data is deduced in reverse through an iterative optimization algorithm. The arrangement configuration includes the number of adsorption layers, molecular orientation angle, and the characteristic peak position of the intermolecular radial distribution function. Based on the number of adsorption layers and molecular orientation angle in the arrangement configuration, the average molecular residence time and jumping frequency, which characterize the degree of restriction of fluid molecule movement near the wall, are extracted, and the average molecular residence time and jumping frequency are mapped to the slip length and apparent viscosity near the wall at the solid-liquid interface. Based on the slip length and the near-wall apparent viscosity, by solving the Navier-Stokes equations coupled with the interface slip boundary conditions, near-wall micro-constraints are generated to make the current arrangement configuration evolve into a preset ideal configuration. The micro-constraints are specifically manifested as the normal vibration frequency and tangential shear amplitude at the wall. The normal vibration frequency and the tangential shear amplitude are applied to a piezoelectric micro-actuator array arranged on the back side of the flow channel wall. Dynamic normal displacement and tangential standing wave field are induced at the wall through the piezoelectric effect, so as to adjust the arrangement configuration of fluid molecules at the solid-liquid interface.

[0046] It should be noted that in applications such as high-purity pharmaceuticals and precision chemicals, where extremely high heat transfer efficiency and surface cleanliness are required, the adsorption orientation, density of arrangement, and residence time of fluid molecules near the wall directly determine the magnitude of the wall thermal resistance and the probability of fouling nuclei formation. Therefore, this embodiment utilizes a miniature thermal film sensor and a particle image velocimetry system deployed on the flow channel wall to capture the temperature distribution and velocity gradient distribution of the fluid within the near-wall micro-region in real time. These two sets of data, along with the fluid composition (such as molecular chain length and type of polar groups) and physical properties (such as viscosity and surface tension) under the current operating conditions, constitute the force field parameter boundary conditions describing the dynamic characteristics of the solid-liquid interface. Their physical significance lies in converting macroscopically measurable flow and heat transfer signals into the interaction potential field input required for molecular simulation.

[0047] Then, the aforementioned boundary conditions are loaded into a pre-constructed molecular dynamics simulator. Unlike traditional molecular simulations that use a preset force field for forward calculations, this method employs a reverse inference strategy: using the experimentally measured number density distribution curve of fluid molecules along the wall normal as the objective function, an iterative optimization algorithm is used to adjust the initial molecular configuration and arrangement in the simulation until the simulated number density distribution matches the actual collected data. This process is equivalent to calculating the actual arrangement configuration of fluid molecules at the solid-liquid interface under the current conditions by "inferring how molecules are arranged from the known density distribution," specifically including the number of adsorption layers formed near the wall, the orientation angle of the molecular long axis relative to the wall, and the characteristic peak position of the intermolecular radial distribution function.

[0048] Based on the calculated arrangement configuration, kinetic parameters reflecting the degree of restriction on molecular motion are extracted. A greater number of adsorption layers and a molecular orientation angle closer to parallel to the wall mean a longer average residence time and a lower jumping frequency near the wall. This directly leads to a decrease in slip length and an increase in apparent viscosity near the wall, i.e., increased flow resistance and thermal resistance at the wall. After mapping the average residence time and jumping frequency to slip length and apparent viscosity, these two parameters are used as new constraints and substituted into the Navier-Stokes equations coupled with interfacial slip boundary conditions for solution. The goal of the solution is to generate a set of near-wall micro-constraints that cause the current arrangement configuration to evolve towards a predetermined ideal configuration. These constraints are specifically manifested as the normal vibrational frequency and tangential shear amplitude that need to be applied at the wall.

[0049] The calculated frequency and amplitude parameters are converted into electrical signals and applied to an array of piezoelectric microactuators pre-attached to the back of the flow channel wall. Through the inverse piezoelectric effect, each unit can generate precise normal displacement and tangential expansion within nanoseconds, inducing dynamic normal displacement waves and tangential standing wave fields at the wall surface. These directly act on the fluid molecules in the near-wall micro-region: appropriate normal vibrations can interfere with the vertical adsorption tendency of molecules on the wall surface, reducing their residence time; precisely controlled tangential standing waves can change the molecular orientation angle, transforming them from a parallel arrangement that is prone to heat transfer obstruction to an inclined or vertical arrangement that is conducive to heat transport. Through this active intervention at the microscale, this method achieves a technological leap from passively adapting to the arrangement of fluid molecules to actively guiding the molecular arrangement configuration, optimizing the momentum and heat transfer characteristics of the interface from the source before the fouling molecules form stable adsorption nuclei.

[0050] like Figure 2 As shown, the second aspect of the present invention discloses a heat exchanger adaptive control system, including a memory and a processor. The memory stores a heat exchanger adaptive control method program. When the heat exchanger adaptive control method program is executed by the processor, any one of the steps of the heat exchanger adaptive control method is implemented.

[0051] The third aspect of the present invention discloses a computer-readable storage medium storing a heat exchanger adaptive control method program, wherein when the heat exchanger adaptive control method program is executed by a processor, any of the steps of the heat exchanger adaptive control method are implemented.

[0052] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A heat exchanger adaptive control method, characterized in that, Includes the following steps: Real-time acquisition of temperature and pressure field distribution data across the entire heat exchanger domain; construction of the fluid's thermodynamic entropy production matrix; quantitative identification of the dominant dissipation mechanisms leading to energy quality loss and their spatial distribution weights. Based on the distribution weight of the dominant dissipation mechanism, the spatial pose of the flow guiding components in the internal flow channel of the heat exchanger is adjusted to change the field cooperation angle of the fluid, optimize the cooperation between the velocity field and the temperature gradient, and simultaneously reconstruct the flow resistance characteristics. By utilizing the reconstructed flow resistance characteristics and real-time acquired physical property parameters, transient simulation of the heat exchanger digital twin is performed to predict the spatiotemporal evolution trend of the equivalent fouling thermal resistance of the heat exchange surface under the current operating boundary. The predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is used as a penalty factor and introduced into a multi-objective optimization model with thermodynamic perfection as the goal. The flow ratio of the hot and cold fluid channels and the control strategy of the phase change interface that minimizes global entropy production are solved in real time. Based on the aforementioned control strategy, control commands are output to adjust the actuator, and the boundary conditions of the digital twin and the configuration function of the multi-objective optimization model are corrected online using the actual operating data after adjustment.

2. The adaptive control method for a heat exchanger according to claim 1, characterized in that, Real-time acquisition of temperature and pressure field distribution data across the entire heat exchanger domain; construction of a fluid thermodynamic entropy production matrix; quantitative identification of the dominant dissipation mechanisms leading to energy quality loss and their spatial distribution weights, specifically: Based on real-time acquired global temperature and pressure field distribution data of the heat exchanger, a three-dimensional temperature gradient field and a three-dimensional velocity vector field are generated through spatial interpolation and gradient reconstruction. The three-dimensional temperature gradient field and the three-dimensional velocity vector field are subjected to tensor dot product operation, and combined with the local physical properties of the fluid, a local entropy yield distribution field characterizing the intensity of irreversible loss at various spatial locations inside the heat exchanger is constructed. By integrating the local entropy production rate distribution field along the three-dimensional spatial coordinates of the heat exchanger, a total entropy production matrix covering viscous entropy production and thermal conduction entropy production is obtained. The total entropy production matrix is ​​orthogonally decomposed and spectral analyzed. By separating the entropy production contribution values ​​corresponding to different dissipation mechanisms, the dominant dissipation mechanisms that lead to energy quality loss are identified. Based on the proportion of each dominant dissipation mechanism in the total entropy production, the spatial distribution weight coefficient of the dominant dissipation mechanism is determined.

3. The adaptive control method for a heat exchanger according to claim 1, characterized in that, Based on the distribution weight of the dominant dissipation mechanism, the spatial orientation of the flow guiding components in the internal flow channel of the heat exchanger is adjusted to change the field cooperation angle of the fluid, optimize the cooperation between the velocity field and the temperature gradient, and simultaneously reconstruct the flow resistance characteristics, specifically: Based on the distribution weight of the dominant dissipation mechanism, the spatial coordinate set with local entropy production exceeding a preset threshold is extracted as the key control area, and a streamline curvature gradient distribution map is generated based on the velocity vector field within the area. The streamline curvature gradient distribution map is compared point by point with the reference streamline profile pre-generated based on the principle of minimum entropy production, and the required streamline deflection angle field at each control position is calculated. Based on the streamline deflection angle field, and combined with the kinematic mapping relationship between the local velocity vector at the installation position of the guide component and the adjustable parameters of the component, the target angle of attack adjustment amount and execution sequence of each guide component are determined, and the guide component pose control command is formed. The orientation control command of the flow guide component is sent to the actuator to drive the flow guide component to change its spatial orientation and induce a controllable secondary flow in the flow channel, thereby adjusting the direction of the velocity vector so that the field coordination angle between the velocity field and the temperature gradient field approaches the theoretical optimal value. Real-time acquisition of local pressure field data after adjustment; by comparing the pressure difference and flow rate relationship between the inlet and outlet of the flow channel before and after adjustment, calculation of the change in flow resistance coefficient of each flow section; and online correction of the global flow resistance characteristic curve of the heat exchanger based on the change in flow resistance coefficient, thus completing the reconstruction of the flow resistance characteristics.

4. The adaptive control method for a heat exchanger according to claim 1, characterized in that, Using the reconstructed flow resistance characteristics and real-time acquired physical property parameters, transient simulations are performed on the digital twin of the heat exchanger to predict the spatiotemporal evolution trend of the equivalent fouling thermal resistance on the heat exchange surface under the current operating boundary conditions. Specifically: Based on the reconstructed flow resistance characteristics and real-time acquired physical property parameters, the instantaneous flow distribution data of each flow channel of the current heat exchanger and the fluid inlet and outlet temperature data are used as boundary conditions and loaded into the pre-constructed digital twin of the heat exchanger. The digital twin is embedded with a transient fouling deposition kinetic model based on the heat and mass transfer mechanism. By using the transient fouling deposition kinetic model and combining the wall shear force distribution data reflected by the flow resistance characteristics, the adhesion probability and erosion rate of particulate matter in the fluid on the heat exchange surface are determined, and the initial distribution density of point-like fouling nuclei at each spatial location on the heat exchange surface under the current operating boundary is generated. Taking the initial distribution density of the point-like dirt cores as the starting point of evolution, the crystal growth dynamics module in the digital twin is invoked, and the real-time collected wall temperature and near-wall fluid concentration are used as the driving potential to iteratively calculate the thickness growth rate of the dirt layer in the direction perpendicular to the wall and the coverage expansion rate in the direction along the wall. Based on the iterative calculation results of the thickness growth rate and coverage expansion rate, and combined with the physical property parameter library of the fouling layer, the equivalent fouling thermal resistance value of each grid node on the heat exchange surface is dynamically updated to generate a non-uniformly distributed thermal resistance network composed of fouling layers of different thicknesses and densities. Spatial integration and temporal recursion are performed on the thermal resistance network along the heat exchange surface to obtain the equivalent spatiotemporal evolution trend of fouling thermal resistance covering the entire process from initial formation to stable growth of fouling. The spatiotemporal evolution trend is presented in the form of a thermal resistance spatiotemporal evolution cloud map, which is used to characterize the thermal resistance change trajectory of each region of the heat exchange surface within a future preset time period.

5. The adaptive control method for a heat exchanger according to claim 1, characterized in that, The predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is used as a penalty factor and introduced into a multi-objective optimization model with thermodynamic perfection as the goal. This model is then used to solve in real-time the flow ratio of the hot and cold fluid channels and the control strategy of the phase change interface that minimizes global entropy production. Specifically: The predicted spatiotemporal evolution trend of the equivalent fouling thermal resistance is spatially convolved with the real-time collected heat flux field distribution data to generate a dynamic penalty function matrix that varies with the position of the grid nodes on the heat exchange surface. The dynamic penalty function matrix is ​​embedded as a penalty term into an optimization model with the goal of minimizing total entropy yield, and a thermodynamic perfection evaluation function that includes the influence of fouling thermal resistance is constructed. The sensitivity analysis of the thermodynamic perfection evaluation function with respect to the valve opening degree and phase change interface height of the hot and cold fluid channels was performed using the adjoint variable method, and the descent gradient field of each operating variable with respect to the objective function was generated. Based on the descent gradient field, a non-dominated solution search based on the Pareto front is performed in the feasible region of the operational variable space to obtain a set of candidate control parameters that simultaneously make the total entropy yield and the integral value of the dynamic penalty function approach optimal. The candidate control parameter set is input into the heat exchanger digital twin for transient simulation verification. The temperature fluctuation characteristic value and pressure pulsation amplitude corresponding to each candidate parameter are extracted, and the inferior solutions that induce local thermal stress concentration or flow field instability are eliminated. The parameter combination that minimizes the spatiotemporal integral value of the dynamic penalty function matrix and the global entropy yield is selected from the remaining candidate parameters and used as the final cold and hot fluid channel flow ratio and phase change interface control strategy.

6. The adaptive control method for a heat exchanger according to claim 1, characterized in that, Based on the aforementioned control strategy, control commands are output to adjust the actuators, and the boundary conditions of the digital twin and the configuration function of the multi-objective optimization model are corrected online using the adjusted actual operating data. Specifically: Based on the aforementioned control strategy, the flow ratio of the hot and cold fluid channels and the phase change interface control strategy are parsed into an executable control command sequence, and the control command sequence is sent down to the underlying actuator for driving. Real-time acquisition of updated temperature field and updated pressure field distribution data of the entire heat exchanger after regulation, and input of the updated temperature field and updated pressure field distribution data into the heat exchanger digital twin, and perform point-by-point residual calculation with the predicted output data of the digital twin under the current regulation strategy to generate a residual field containing the temperature deviation value and pressure deviation value of each spatial node. The residual field is correlated with the valve opening change and phase change interface displacement corresponding to the control strategy. The parameter correction gradient of the internal physical mechanism model of the digital twin is calculated by the backpropagation algorithm. Based on the parameter correction gradient, the wall heat transfer coefficient distribution map and the flow resistance coefficient distribution map in the digital twin are updated online to complete the adaptive calibration of the boundary conditions of the digital twin. The calibrated digital twin is re-run under the current operating conditions, and the newly predicted heat flux field distribution data is extracted. The newly predicted heat flux field distribution data is then compared with the actual heat flux field distribution data collected in real time after regulation, and the model mismatch error vector caused by the change in regulation strategy is calculated. Based on the model mismatch error vector, the weight coefficient allocation rule of the thermodynamic perfection evaluation function embedded in the multi-objective optimization model is modified to generate an updated configuration function, so that the configuration function can accurately reflect the actual thermodynamic characteristics of the current heat exchanger under the updated boundary conditions.

7. The adaptive control method for a heat exchanger according to claim 6, characterized in that: The control command sequence includes the valve position opening setting value of each regulating valve and the displacement setting value of the phase change interface height adjuster.

8. The adaptive control method for a heat exchanger according to claim 1, characterized in that: The physical properties include the fluid's density, dynamic viscosity, thermal conductivity, specific heat capacity, surface tension, diffusion coefficient, and latent heat of vaporization.

9. A heat exchanger adaptive control system, characterized in that, The device includes a memory and a processor. The memory stores a heat exchanger adaptive control method program. When the heat exchanger adaptive control method program is executed by the processor, it implements the steps of the heat exchanger adaptive control method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a heat exchanger adaptive control method program, which, when executed by a processor, implements the steps of the heat exchanger adaptive control method as described in any one of claims 1 to 8.

Images