Environment-friendly continuous flow high-pressure reaction method and system based on supercritical fluid

Abstract

The invention relates to the technical field of industry, and discloses an environment-friendly continuous flow high-pressure reaction method based on supercritical fluid, which is characterized in that a set of complete supercritical fluid reaction intelligent control system is constructed through organic combination of the eight steps. According to the system, full-process automatic management from data acquisition to control execution is realized, and core control tasks such as accurate perception of a process state, accurate prediction of a future trend, optimization decision of a control strategy, coordinated configuration of execution actions and the like are completed. Through the combination of model prediction and real-time calibration, the system can accurately grasp complex physical property change and reaction dynamic characteristics of the supercritical fluid; through cooperation of feed-forward compensation and feedback adjustment, the system effectively suppresses the influence of various disturbances on the process stability.

Description

Technical Field

[0001] This invention relates to the field of industrial technology, specifically to an environmentally friendly continuous flow high-pressure reaction method and system based on supercritical fluid. Background Technology

[0002] In the subfield of continuous reaction process control, supercritical fluid high-pressure reaction systems represent the pinnacle of control difficulty. These systems need to maintain stable operation under extreme conditions, with pressure-flow coordinated control being a core element in ensuring consistent product quality. Existing digital twin applications mostly remain at the level of offline simulation or condition monitoring, failing to truly achieve a multi-physics real-time prediction model integrating fluid dynamics calculations, phase thermodynamic descriptions, and chemical reaction kinetics. Furthermore, they lack a closed-loop architecture that directly translates prediction results into control commands.

[0003] The control schemes commonly used in industry today suffer from three major flaws. First, the models are oversimplified, reducing the complex three-dimensional flow field to a one-dimensional lumped parameter model, failing to capture uneven velocity distribution and local density fluctuations in the pipeline. Second, the physical property calculations are lagging, as the density and viscosity data used by the controller are based on the temperature and pressure conditions of the previous moment, rather than the current actual conditions. Third, multivariable decoupled control ignores the direct impact of pressure in the supercritical region on flow measurement; when pressure fluctuates, the flowmeter readings produce spurious deviations, and the controller's adjustment based on these deviations actually amplifies the disturbance. This architecture can barely maintain operation under steady-state conditions, but it is slow to respond and prone to oscillations when faced with dynamic disturbances.

[0004] Therefore, we propose an environmentally friendly continuous flow high-pressure reaction method and system based on supercritical fluids to solve the problems mentioned above. Summary of the Invention

[0005] The purpose of this invention is to provide an environmentally friendly continuous flow high-pressure reaction method and system based on supercritical fluids, in order to solve the problem of oversimplification of the models proposed in the background art, which simplifies the complex three-dimensional flow field to a one-dimensional lumped parameter model and fails to capture the uneven velocity distribution and local density fluctuations in the pipeline.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid, the specific steps of which are as follows: S1. Distributed data acquisition: Pressure sensors, temperature sensors, flow meters, and density meters are respectively arranged at the feed end, reaction section, and outlet section of the reaction device. The sensors acquire data in a high-frequency manner and transmit it to the data processing unit after filtering. The local fluid density and phase information are calculated through soft measurement algorithms. S2. Simplified Model Establishment: Feature extraction is performed on the offline three-dimensional flow field simulation results to construct a simplified fluid dynamics model containing the main flow characteristics. This model is combined with the supercritical fluid state equation to calculate real-time physical property parameters and coupled with the reaction rate calculation module so that a single prediction calculation meets the real-time control requirements. S3, Model Calibration Synchronization: The process parameters measured in real time are used as input to drive the simplified model to run. The allowable deviation range between the predicted value and the measured value is set. When the deviation exceeds the range, the heat transfer coefficient, resistance coefficient and catalyst activity parameters inside the model are automatically adjusted. The best estimate of the system operating status is obtained through data fusion technology. S4. Multivariable Optimization Control: Using the calibrated model as a tool for predicting future states, a prediction time window is set. Under the premise of satisfying constraints such as pressure fluctuation, flow deviation and temperature change rate, the optimal adjustment scheme of back pressure valve opening, feed pump speed and heater power is calculated. After solving the optimization problem, the first step of control action is executed. S5. Disturbance compensation mechanism: For observable feed fluctuations or composition changes, the impact on the system is calculated in advance based on the model's response characteristics, and a compensation signal is generated. At the same time, the deviation-based feedback adjustment function is maintained to handle uncertainties. The two adjustment signals are adaptively weighted according to the current working conditions and then act together on the actuator. S6. Actuator Coordination Management: Before issuing control commands, check the action capability boundaries of each actuator. When an actuator reaches its adjustment limit, allocate some control tasks to other available equipment. Allocate control actions reasonably according to the cumulative workload of each actuator and set safety limits to prevent operation from exceeding the limits. S7. Experience accumulation and application: Establish a historical operation record library covering different raw material characteristics, catalyst state and environmental conditions, summarize past operating conditions into several typical categories, identify the current category through feature comparison during operation and adopt the corresponding mature control scheme, and continuously incorporate new successful experiences into the record library to improve the control strategy. S8. Equipment Status Prediction: Track the changing trends of model parameters to establish a predictive relationship for equipment performance degradation, judge the development process of conditions such as catalyst activity decline, seal aging or heat transfer surface contamination, and when it is expected to reach a level that requires attention in the short term, fine-tune the operating parameters in advance to compensate, and remind the operators of the appropriate maintenance time.

[0007] Preferably, in step S1, the specific steps of distributed data acquisition are as follows: S1.1 A pressure transmitter and a flow meter are installed at the feed end of the reaction device. Temperature measuring points are arranged along the flow direction in the reaction section. A density meter and a pressure sensor are installed at the outlet section. The sensors acquire process parameters by continuous sampling and transmit them to the data processing unit via the industrial communication network. The transmitted signals are digitally filtered to eliminate interference and vibration effects. S1.2 The data processing unit calculates the flow resistance based on the pressure measurement value and pipeline parameters, estimates the heat transfer parameters based on the temperature distribution and heat flux density, obtains the fluid density and viscosity using the pressure and temperature values ​​through physical property correlation formulas or table lookups, organizes the measured data and estimated parameters into a time-stamped data sequence and transmits it to the model calculation unit, and simultaneously establishes a rolling window to store recent data.

[0008] Preferably, in step S2, the simplified model establishment steps are as follows: S2.1 Perform modal decomposition on the three-dimensional flow field simulation data, extract the main spatial modes and time evolution coefficients, reconstruct and simplify the fluid dynamics model based on the dominant modes, retain the velocity distribution and pressure drop characteristics and reduce the calculation dimension, convert the three-dimensional mesh into a one-dimensional pipeline model and a quasi-two-dimensional cross-section model, and control the number of degrees of freedom through the modal truncation threshold; S2.2 Establish a data interface between the simplified fluid dynamics model and the supercritical fluid state equation. At each calculation moment, calculate the fluid density and viscosity values ​​based on the temperature and pressure values ​​and update them to the constitutive relation of the flow field model. At the same time, embed the reaction dynamics module, calculate the reaction rate based on the local temperature and concentration as the source term and introduce it into the equation to complete the iterative calculation of the flow field, physical property parameters and reaction rate.

[0009] Preferably, in step S3, the specific steps for model calibration synchronization are as follows: S3.1 Input the real-time parameters obtained in step S1 into the simplified model established in step S2 to drive it to perform calculations. The model outputs the predicted values ​​of pressure, temperature and flow at each key location. The predicted values ​​are compared with the corresponding sensor measurements to calculate the deviation. Deviation thresholds are set for different parameters. When the deviation of the measuring point exceeds the threshold in multiple consecutive cycles, it is determined that the model is mismatched with the actual state and the deviation characteristics are recorded. S3.2. Based on the deviation characteristics in step S3.1, start the parameter identification program, analyze the deviation amplitude and location to determine the correction parameter category. When the temperature deviation is the main factor, adjust the heat transfer coefficient; when the pressure deviation is the main factor, adjust the drag coefficient; when the conversion rate deviation is the main factor, adjust the catalyst activity coefficient. Use the least squares method or gradient optimization algorithm to solve for the parameter value that minimizes the sum of squares of the deviations and update the model. Apply Kalman filtering to fuse the updated prediction results with the measurement data to generate the optimal estimates of system pressure, temperature and flow rate, and pass them to step S4.

[0010] Preferably, in step S4, the specific steps of multivariate optimization control are as follows: S4.1 Using the state estimate obtained in step S3 as the initial condition, the system's evolution trajectory within the prediction window is deduced using the calibrated model. The prediction window length is set to cover the main dynamic response characteristics. The back pressure valve opening, pump speed, and heating power are set as adjustable operating variables, and their variation amplitude and rate limits are set. Constraints are set for pressure fluctuations, flow deviations, and temperature change rates. The constraints are expressed in the form of a system of inequalities, and a constrained optimization mathematical model is established. S4.2 Establish a weighted objective function that comprehensively reflects the deviation of the controlled variable and the change of the operating variable. Use a sequential quadratic programming algorithm or gradient optimization algorithm to solve the operating variable adjustment trajectory that minimizes the objective function under the constraints of step S4.1. This trajectory includes the valve opening, pump speed and heating power values ​​at each time. Extract the first set of values ​​as a control command and send it to the actuator. In the next cycle, the prediction and optimization process is repeated with the state estimate updated in step S3 as the initial condition.

[0011] Preferably, in step S5, the disturbance compensation mechanism comprises the following steps: S5.1 Real-time monitoring of feed flow and composition data. When the flow rate changes by a step beyond the threshold or the composition deviates from the normal range, it is determined to be a measurable disturbance. Using the model established in step S2, the disturbance is simulated on pressure, temperature and flow based on the current working conditions. The peak value, arrival time and decay characteristics of the influence trajectory are analyzed. The valve opening, pump speed and heating power adjustment required to offset the influence are calculated in reverse and organized into a feedforward compensation signal. S5.2 Maintain the operation of the feedback control loop, collect the measured values ​​of pressure, temperature and flow and calculate their deviation from the set value. Generate a feedback control signal based on the deviation and the rate of change using the proportional-integral-derivative algorithm. Set the initial weighting coefficients of the feedforward and feedback signals. Increase the feedforward weight when the disturbance amplitude is large and the model error is small. Increase the feedback weight when there is unmodeled dynamic or large noise. Adjust the weights according to the operating conditions and control effect. Superimpose the weighted and fused comprehensive signal with the prediction command of step S4 and transmit it to step S6.

[0012] Preferably, in step S6, the actuator coordination management involves the following specific steps: S6.1 Receive the integrated control signal output in step S5, query the current status of actuators such as back pressure valve opening, pump speed and heater power, calculate the available adjustment margin of each actuator to the physical limit, start task reallocation when the required adjustment exceeds the available margin, determine the alternative actuator combination based on the coupling relationship of pressure, flow and temperature, calculate the action amount of each mechanism in the alternative combination and reallocate the control task according to the adjustment capacity ratio. S6.2 Establish and update the cumulative number of actions, runtime and response time records of each actuator, calculate the health assessment index, and when multiple actuators can meet the requirements, prioritize the actuator with high health and low load. Before issuing the command, perform three-level verification to verify whether the position parameters are within the safe range, whether the adjustment rate exceeds the allowable value, and whether the coordinated action causes a sudden change. If a risk is detected, limit the amplitude or extend the execution interval. Issue the verified command to the actuator and transmit the action record and health status data to step S7.

[0013] Preferably, in step S7, the specific steps for applying experience accumulation are as follows: S7.1 Record process parameters, control actions and system response data under different raw material batches, catalyst usage stages and environmental conditions, extract operating condition feature vectors including average pressure, temperature distribution, flow stability and conversion rate level, use clustering algorithm to divide historical operating conditions into several typical operating modes, establish feature parameter ranges and corresponding effective control parameter combinations for each mode and store them in the historical database in a structured form. S7.2 During runtime, extract the feature vector of the current operating condition, calculate its distance or similarity with the feature center of each typical mode, select the mode with the closest distance or the highest similarity and retrieve its control parameters as the initial configuration for steps S4 and S5. After completing the control cycle, evaluate performance indicators such as pressure stability, flow fluctuation and temperature deviation. When the standard is met, label the operating condition data with the mode category and add it incrementally to the database. For new operating conditions with similarity below the threshold, select the relatively optimal mode parameters for initialization and record them separately. Transfer the updated database and mode information to step S8.

[0014] Preferably, in step S8, the specific steps for predicting the device status are as follows: S8.1 Continuously monitor the historical sequence of model parameters obtained in step S3, extract the time evolution curves of heat transfer coefficient, resistance coefficient and catalyst activity coefficient, use the sliding window method to calculate the rate of change and fluctuation of each parameter, compare the rate of change with the normal range, and determine that when the heat transfer coefficient continues to decrease and exceeds the threshold, it is determined to be surface contamination; when the resistance coefficient continues to increase, it is determined to be channel blockage; when the catalyst activity continues to decline, it is determined to be deactivation. Combine the historical data in step S7 to establish a mapping model between the parameter change trend and the equipment performance degradation state. S8.2 Based on the mapping model in step S8.1, predict the development trajectory of each parameter in the future time period. When it is predicted that a certain parameter will reach the level of concern within a certain period, start pre-compensation. Increase heating power in advance for the decrease in heat transfer coefficient, adjust pump speed or valve opening for the increase in resistance coefficient, and adjust reaction temperature or residence time for the decrease in catalyst activity. Add the compensation amount to the control command in step S4. Generate maintenance suggestions based on the degree of parameter degradation and remaining time. Send a cleaning prompt when the degree of contamination is expected to reach the cleaning threshold, and send a replacement prompt when the activity is expected to drop to the replacement standard. Output the status assessment and maintenance suggestions to the interactive interface.

[0015] This application also provides an environmentally friendly continuous flow high-pressure reaction system based on supercritical fluid, including a reaction device module, a data acquisition module, a model calculation module, an optimization control module, an execution adjustment module, a historical data module, a state prediction module, and a human-machine interaction module; The reaction unit module is used to carry out the supercritical fluid reaction process. The feed end receives raw materials, the catalyst in the reaction section completes the chemical conversion, and the outlet section outputs products. The overall pressure bearing capacity meets the supercritical operating conditions. The data acquisition module is used to monitor the reaction process status in real time. It collects pressure, temperature, flow rate and density parameters through sensors placed at key locations. After filtering to eliminate interference, the data is transmitted to the computing unit. The model calculation module is used to predict the system's operating state. It uses a simplified fluid dynamics model combined with state equations to calculate real-time physical property parameters, couples reaction kinetics to obtain the reaction rate, and automatically corrects internal parameters through a calibration function to make the prediction results consistent with the actual operation. The optimization control module is used to generate the optimal control scheme, and to deduce the future evolution trend of the system based on the calibration model. Under the premise of meeting safety constraints, it calculates the coordinated adjustment scheme of the back pressure valve, feed pump and heater, and integrates multiple control strategies such as prediction, compensation and feedback. The execution adjustment module is used to implement control commands. According to the commands of the optimization control module, it drives the back pressure valve to adjust the pressure, the feed pump to adjust the flow rate, and the heater to adjust the temperature. It also monitors the operating status and cumulative load of each mechanism and feeds back the execution status to the control loop. The historical data module is used to store operational experience and identify operating conditions. It categorizes and records operational data under different raw material, catalyst states, and environmental conditions, establishes a typical operating condition mode library, quickly matches the current state during operation and calls verified control schemes, and continuously absorbs new experience to enrich the knowledge base. The condition prediction module is used to predict the trend of equipment performance changes, monitor the long-term evolution of model parameters, identify performance degradation signs such as catalyst decay, heat transfer efficiency decline and flow resistance increase, predict the development trend and start compensation adjustment in advance, and generate maintenance timing suggestions. The human-machine interface module connects operators with the control system, displays current operating parameters, control actions, equipment health assessments and maintenance prompts, receives process settings and mode selections, and supports visual management of the system and necessary manual intervention.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This method constructs a complete intelligent control system for supercritical fluid reactions through the organic combination of eight steps. This system realizes fully automated management of the entire process from data acquisition to control execution, and completes core control tasks such as accurate perception of process status, accurate prediction of future trends, optimized decision-making of control strategies, and coordinated configuration of execution actions.

[0017] 2. By combining model prediction with real-time calibration, the system can accurately grasp the complex physical property changes and dynamic characteristics of supercritical fluids. Through the synergy of feedforward compensation and feedback regulation, the system effectively suppresses the impact of various disturbances on process stability. Through actuator coordination and safety verification, the system ensures reliable execution of control actions and inherent process safety. Through experience learning and state prediction, the system achieves continuous optimization of control performance and proactive management of equipment maintenance. The combined effect of these functions enables precise and stable control of key parameters such as pressure, temperature, and flow rate in supercritical fluid reaction processes, improving product quality consistency, reducing raw material and energy consumption, extending equipment lifespan, reducing unplanned downtime losses, and reducing reliance on operator experience. Attached Figure Description

[0018] Figure 1 This is a diagram illustrating the method steps of the present invention.

[0019] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: Please refer to Figure 1 An environmentally friendly continuous flow high-pressure reaction method based on supercritical fluids, the specific steps of which are as follows: S1. Distributed data acquisition: Pressure sensors, temperature sensors, flow meters, and density meters are respectively arranged at the feed end, reaction section, and outlet section of the reaction device. The sensors acquire data in a high-frequency manner and transmit it to the data processing unit after filtering. The local fluid density and phase information are calculated through soft measurement algorithms. S2. Simplified Model Establishment: Feature extraction is performed on the offline three-dimensional flow field simulation results to construct a simplified fluid dynamics model containing the main flow characteristics. This model is combined with the supercritical fluid state equation to calculate real-time physical property parameters and coupled with the reaction rate calculation module so that a single prediction calculation meets the real-time control requirements. S3, Model Calibration Synchronization: The process parameters measured in real time are used as input to drive the simplified model operation. The allowable deviation range between the predicted value and the measured value is set. When the deviation exceeds the range, the heat transfer coefficient, resistance coefficient and catalyst activity parameters inside the model are automatically adjusted. The best estimate of the system operating status is obtained through data fusion technology. S4. Multivariable Optimization Control: Using the calibrated model as a tool for predicting future states, a prediction time window is set. Under the premise of satisfying constraints such as pressure fluctuation, flow deviation and temperature change rate, the optimal adjustment scheme of back pressure valve opening, feed pump speed and heater power is calculated. After solving the optimization problem, the first step of control action is executed. S5. Disturbance compensation mechanism: For observable feed fluctuations or composition changes, the impact on the system is calculated in advance based on the model's response characteristics, and a compensation signal is generated. At the same time, the deviation-based feedback adjustment function is maintained to handle uncertainties. The two adjustment signals are adaptively weighted according to the current working conditions and then act together on the actuator. S6. Actuator Coordination Management: Before issuing control commands, check the action capability boundaries of each actuator. When an actuator reaches its adjustment limit, allocate some control tasks to other available equipment. Allocate control actions reasonably according to the cumulative workload of each actuator and set safety limits to prevent operation from exceeding the limits. S7. Experience accumulation and application: Establish a historical operation record library covering different raw material characteristics, catalyst state and environmental conditions, summarize past operating conditions into several typical categories, identify the current category through feature comparison during operation and adopt the corresponding mature control scheme, and continuously incorporate new successful experiences into the record library to improve the control strategy. S8. Equipment Status Prediction: Track the changing trends of model parameters to establish a predictive relationship for equipment performance degradation, judge the development process of conditions such as catalyst activity decline, seal aging or heat transfer surface contamination, and when it is expected to reach a level that requires attention in the short term, fine-tune the operating parameters in advance to compensate, and remind the operators of the appropriate maintenance time.

[0022] In this embodiment, by arranging pressure, temperature, flow rate, and density sensors at the feed end, reaction section, and outlet section of the reaction device in step S1, full-process monitoring of the supercritical fluid reaction is achieved. High-frequency acquisition combined with filtering effectively eliminates measurement noise, ensuring data reliability. The soft measurement algorithm overcomes the technical challenge of directly measuring some parameters under supercritical conditions, calculating local density and phase information from known parameters, providing a complete state data foundation for the control system, and avoiding control deviations caused by insufficient information.

[0023] Step S2 transforms the offline simulation results into a simplified model that can run in real time, solving the problem that detailed simulation calculations are too time-consuming to be used for online control. The simplified model retains the core elements of flow characteristics, combines the supercritical fluid equation of state to achieve dynamic calculation of physical parameters, and couples the reaction rate module to form a comprehensive model covering flow, heat transfer, and reaction. The calculation speed of this model meets the requirements of real-time control, providing a fast and accurate basis for subsequent predictive control and optimization decisions.

[0024] The real-time calibration mechanism established in step S3 ensures the model's predictive accuracy during actual operation. By inputting measurement data into the model and comparing prediction deviations, the system automatically updates internal parameters such as heat transfer coefficient, drag coefficient, and catalyst activity, enabling the model to adapt to actual conditions such as changes in feedstock, catalyst degradation, and equipment aging. Data fusion technology integrates model predictions and sensor measurement information, generating an optimal estimate of the system state while considering noise and uncertainties, providing a more reliable basis for control decisions.

[0025] Step S4 achieves coordinated optimization and regulation of the back pressure valve, feed pump, and heater, overcoming the problem of mutual interference between loops in traditional single-loop control. The setting of the predictive time window gives the control system foresight, enabling it to respond to system changes in advance. Solving for the optimal solution while meeting safety constraints such as pressure fluctuations, flow deviations, and temperature change rates ensures both process safety and stability while optimizing control performance. The rolling optimization mechanism allows the control strategy to be continuously adjusted according to actual conditions, improving its adaptability to changes in operating conditions.

[0026] Step S5 establishes a control architecture combining active compensation and feedback correction. For measurable feed fluctuations and composition changes, the feedforward stage generates compensation signals in advance based on model predictions, taking control measures before disturbances affect the system, significantly reducing deviations caused by disturbances. The feedback stage handles uncertainties that the model cannot predict, ensuring the robustness of the control system. The adaptive weighting mechanism dynamically adjusts the weights of the two signals according to operating conditions, allowing the advantages of the control actions to complement each other and improving the system's adaptability to complex operating conditions.

[0027] Step S6 addresses the constraint handling and load distribution issues in the coordinated operation of multiple actuators. Before issuing commands, the available adjustment range of each actuator is checked. When an actuator approaches its limit, the task is automatically distributed to other equipment, preventing control failure due to saturation. Control actions are rationally allocated based on accumulated load, extending equipment lifespan and reducing maintenance frequency. Hierarchical safety verification validates control commands at three levels: position, rate, and system response, effectively preventing unsafe operations and ensuring process safety.

[0028] Step S7 establishes a learning mechanism that extracts knowledge from historical data to guide current control. By classifying operational data under different raw material, catalyst states, and environmental conditions, a structured operating condition knowledge base is formed. During runtime, feature comparison quickly identifies the current operating condition category and calls upon verified control parameters, shortening parameter tuning time and improving response speed under new operating conditions. Continuous experience accumulation allows the control strategy to gradually improve over time, continuously enhancing control performance and achieving system self-optimization.

[0029] Step S8 achieves a shift from a passive response to a proactive prevention-based maintenance model. By tracking the long-term trends of model parameters, a correlation between parameter evolution and equipment performance degradation is established, enabling early identification of deterioration phenomena such as catalyst deactivation, seal aging, and heat transfer surface contamination. Predictive compensation adjusts operating parameters before performance degradation impacts normal operation, extending equipment uptime. Maintenance timing alerts provide a basis for planned maintenance, avoiding unplanned shutdowns caused by sudden failures and improving the continuity and economic efficiency of plant operation.

[0030] This method, through the organic combination of the above eight steps, constructs a complete intelligent control system for supercritical fluid reactions. This system achieves fully automated management of the entire process from data acquisition to control execution, completing core control tasks such as precise perception of process status, accurate prediction of future trends, optimized decision-making of control strategies, and coordinated configuration of execution actions. Through the combination of model prediction and real-time calibration, the system can accurately grasp the complex physical property changes and dynamic characteristics of supercritical fluids; through the synergy of feedforward compensation and feedback regulation, the system effectively suppresses the impact of various disturbances on process stability; through actuator coordination and safety verification, the system ensures reliable execution of control actions and inherent process safety; through experience learning and state prediction, the system achieves continuous optimization of control performance and proactive management of equipment maintenance. The combined effect of these functions enables precise and stable control of key parameters such as pressure, temperature, and flow rate in supercritical fluid reaction processes, improving product quality consistency, reducing raw material and energy consumption, extending equipment lifespan, reducing unplanned downtime losses, and reducing reliance on operator experience.

[0031] Example 2: Please refer to Figure 1 In step S1, the specific steps of distributed data acquisition are as follows: S1.1 A pressure transmitter and a flow meter are installed at the feed end of the reaction device. Temperature measuring points are arranged along the flow direction in the reaction section. A density meter and a pressure sensor are installed at the outlet section. The sensors acquire process parameters by continuous sampling and transmit them to the data processing unit via the industrial communication network. The transmitted signals are digitally filtered to eliminate interference and vibration effects. S1.2 The data processing unit calculates the flow resistance based on the pressure measurement value and pipeline parameters, estimates the heat transfer parameters based on the temperature distribution and heat flux density, obtains the fluid density and viscosity using the pressure and temperature values ​​through physical property correlation formulas or table lookups, organizes the measured data and estimated parameters into a time-stamped data sequence and transmits it to the model calculation unit, and simultaneously establishes a rolling window to store recent data.

[0032] In this embodiment: By rationally arranging multiple types of sensors at key locations in the reaction device in step S1.1, comprehensive coverage of the inlet conditions, reaction zone, and outlet status of the supercritical fluid reaction process is achieved. Continuous sampling ensures real-time acquisition of process parameters, and the application of industrial communication networks improves the reliability and speed of data transmission. Digital filtering effectively eliminates the influence of electromagnetic interference and mechanical vibration commonly found in industrial settings on the measurement signals, ensuring the authenticity and stability of the sensor data. This provides high-quality data input for subsequent model calculations and control decisions, avoiding misjudgments and control fluctuations caused by measurement noise.

[0033] The parameter extrapolation mechanism established in step S1.2 compensates for the shortcomings of direct measurement methods. It uses measured parameters to calculate key physical properties that are difficult to measure directly, such as flow resistance, heat transfer parameters, fluid density, and viscosity, through physical correlation. The measured data and extrapolated parameters are organized into a structured data sequence in chronological order, facilitating state estimation and trend analysis by the model calculation unit. The scrolling window setting enables effective storage and rapid retrieval of recent historical data, providing a data foundation for parameter change trend identification, disturbance detection, and model parameter identification, thereby improving the control system's ability to perceive the dynamic characteristics of the process.

[0034] Step S1, through the combined action of two smaller steps, achieves comprehensive acquisition and in-depth processing of the state information of the supercritical fluid reaction process. Traditional control systems typically rely on only a few measuring points and directly measured parameters, resulting in an insufficient grasp of the process state and difficulty in adapting to the rapid changes in the physical properties of supercritical fluids. This step expands the measurement range through distributed sensor deployment, improves measurement accuracy through digital filtering, and enriches the dimensions of state information through parameter extrapolation, forming a complete dataset containing both directly measured and indirectly extrapolated values. This multi-layered data acquisition and processing approach enables the control system to accurately capture the drastic changes in fluid properties and the complex dynamic characteristics of the reaction process under supercritical conditions, providing rich and reliable information support for model prediction, optimized control, and fault diagnosis. Compared to traditional methods, this step improves the comprehensiveness and accuracy of process state perception and reduces the risk of control performance degradation due to missing information or measurement errors.

[0035] Example 3: Please refer to Figure 1 In step S2, the specific steps for simplifying model building are as follows: S2.1 Perform modal decomposition on the three-dimensional flow field simulation data, extract the main spatial modes and time evolution coefficients, reconstruct and simplify the fluid dynamics model based on the dominant modes, retain the velocity distribution and pressure drop characteristics and reduce the calculation dimension, convert the three-dimensional mesh into a one-dimensional pipeline model and a quasi-two-dimensional cross-section model, and control the number of degrees of freedom through the modal truncation threshold; S2.2 Establish a data interface between the simplified fluid dynamics model and the supercritical fluid state equation. At each calculation moment, calculate the fluid density and viscosity values ​​based on the temperature and pressure values ​​and update them to the constitutive relation of the flow field model. At the same time, embed the reaction dynamics module, calculate the reaction rate based on the local temperature and concentration as the source term and introduce it into the equation to complete the iterative calculation of the flow field, physical property parameters and reaction rate.

[0036] In this embodiment: Step S2.1 uses modal decomposition to extract the main spatial modes and temporal evolution coefficients from the three-dimensional flow field simulation data, identifying the dominant modes that have the greatest impact on flow characteristics, thus achieving an effective conversion from a high-dimensional complex simulation model to a low-dimensional simplified model. By reasonably setting the modal truncation threshold, the number of degrees of freedom of the model is significantly reduced while retaining core flow characteristics such as velocity distribution and pressure drop, simplifying the three-dimensional mesh into a one-dimensional pipe model or a quasi-two-dimensional cross-sectional model. This dimensionality reduction reduces the computational load by several orders of magnitude, shortening the computation time from several hours in offline simulation to milliseconds, meeting the response speed requirements of real-time control, while maintaining the ability to accurately describe the main characteristics of the flow field.

[0037] The data interface established in step S2.2 enables tight coupling between the simplified fluid dynamics model and the supercritical fluid equation of state. This allows the model to update fluid properties such as density and viscosity in real time based on temperature and pressure at each calculation moment, accurately reflecting the dramatic changes in properties under supercritical conditions. The embedded reaction kinetics module calculates the reaction rate based on local temperature and concentration and introduces it as a source term into the governing equation, achieving cyclic coupling calculations of the flow field, physical properties, and reaction rate. This multiphysics integration approach allows the simplified model to simultaneously capture fluid dynamics, thermodynamics, and chemical reaction effects, improving the model's accuracy in describing complex reaction processes.

[0038] Step S2 resolves the conflict between computational accuracy and speed in supercritical fluid reaction process modeling through two complementary steps. Traditional methods either employ detailed 3D simulation models, which are computationally too time-consuming for real-time control, or rely on empirical correlations, which lack sufficient accuracy to reflect complex property changes. This step innovatively extracts the dominant flow mode through modal decomposition, reducing the computational dimensionality while preserving core flow characteristics. By coupling with the equations of state and reaction kinetics, it achieves dynamic updates of property parameters and collaborative calculations of multiphysics fields, constructing a simplified model that meets both real-time requirements and possesses sufficient prediction accuracy.

[0039] Example 4: Please refer to Figure 1 In step S3, the specific steps for model calibration synchronization are as follows: S3.1 Input the real-time parameters obtained in step S1 into the simplified model established in step S2 to drive it to perform calculations. The model outputs the predicted values ​​of pressure, temperature and flow at each key location. The predicted values ​​are compared with the corresponding sensor measurements to calculate the deviation. Deviation thresholds are set for different parameters. When the deviation of the measuring point exceeds the threshold in multiple consecutive cycles, it is determined that the model is mismatched with the actual state and the deviation characteristics are recorded. S3.2. Based on the deviation characteristics in step S3.1, start the parameter identification program, analyze the deviation amplitude and location to determine the correction parameter category. When the temperature deviation is the main factor, adjust the heat transfer coefficient; when the pressure deviation is the main factor, adjust the drag coefficient; when the conversion rate deviation is the main factor, adjust the catalyst activity coefficient. Use the least squares method or gradient optimization algorithm to solve for the parameter value that minimizes the sum of squares of the deviations and update the model. Apply Kalman filtering to fuse the updated prediction results with the measurement data to generate the optimal estimates of system pressure, temperature and flow rate, and pass them to step S4.

[0040] In this embodiment, the model validation mechanism established in step S3.1 simplifies the continuous monitoring of model prediction performance. Real-time measurement parameters are input into the model to drive its calculations, and the current prediction accuracy of the model is evaluated by comparing the deviation between the predicted values ​​and the actual measured values. Reasonable deviation thresholds are set for different process parameters to avoid false alarms triggered by occasional measurement noise, and the continuous multi-cycle threshold exceeding criterion improves the reliability of model mismatch identification. Detailed records of deviation characteristics include information such as deviation amplitude, duration, and spatial distribution, providing accurate diagnostic basis for subsequent parameter identification, enabling the system to promptly detect deviations between the model and the actual state and initiate calibration procedures.

[0041] The parameter identification mechanism established in step S3.2 enables adaptive adjustment of the model's internal parameters. Based on the analysis of deviation characteristics, the categories of parameters requiring correction are determined. The correspondence between temperature deviation and heat transfer coefficient, pressure deviation and drag coefficient, and conversion rate deviation and catalyst activity coefficient is established, achieving precise orientation of parameter correction. The least squares method or gradient optimization algorithm is used to solve for the optimal parameter values, ensuring the mathematical optimality of the identification results. Kalman filtering technology integrates the updated model predictions and sensor measurements, generating the optimal estimate of the system state while considering model uncertainty and measurement noise, providing the controller with more accurate state information than using measured or predicted values ​​alone.

[0042] Step S3 establishes a closed-loop mechanism for model calibration and data fusion through two smaller steps, solving the problem of decreased prediction accuracy of simplified models due to changes in operating conditions and equipment aging during long-term operation. Traditional open-loop model methods lack adaptive capabilities; once model parameters are determined, they are not updated. As raw material characteristics fluctuate, catalyst activity declines, and equipment conditions change, model prediction errors gradually accumulate, ultimately leading to deterioration in control performance. This step innovatively establishes a three-layer calibration system of deviation monitoring, parameter identification, and data fusion, enabling the model to continuously adjust its internal parameters based on actual operating data, maintaining its ability to accurately describe the true state of the process.

[0043] This online adaptive mechanism significantly improves the long-term reliability and applicability of the model compared to traditional offline modeling methods, enabling the same model to adapt to various operating conditions, such as different batches of raw materials, different stages of catalyst use, and different operating states of equipment. The application of data fusion technology fully utilizes the dynamic predictive capabilities of the model and the real-time measurement information from sensors, generating optimal state estimates even when measurement noise and model errors coexist, providing high-quality state feedback for subsequent predictive control.

[0044] Example 5: Please refer to Figure 1 In step S4, the specific steps of multivariate optimization control are as follows: S4.1 Using the state estimate obtained in step S3 as the initial condition, the system's evolution trajectory within the prediction window is deduced using the calibrated model. The prediction window length is set to cover the main dynamic response characteristics. The back pressure valve opening, pump speed, and heating power are set as adjustable operating variables, and their variation amplitude and rate limits are set. Constraints are set for pressure fluctuations, flow deviations, and temperature change rates. The constraints are expressed in the form of a system of inequalities, and a constrained optimization mathematical model is established. S4.2 Establish a weighted objective function that comprehensively reflects the deviation of the controlled variable and the change of the operating variable. Use a sequential quadratic programming algorithm or gradient optimization algorithm to solve the operating variable adjustment trajectory that minimizes the objective function under the constraints of step S4.1. This trajectory includes the valve opening, pump speed and heating power values ​​at each time. Extract the first set of values ​​as a control command and send it to the actuator. In the next cycle, the prediction and optimization process is repeated with the state estimate updated in step S3 as the initial condition.

[0045] In this embodiment: the prediction framework established in step S4.1 enables forward-looking projection of the system's future state. Using the optimal state estimate provided in step S3 as initial conditions, a simplified model calibrated in real-time is used to simulate the system's dynamic response within the prediction window, giving the control system the ability to anticipate the future. The reasonable setting of the prediction window length ensures coverage of the system's main dynamic characteristics while avoiding over-computation. Back pressure valve opening, pump speed, and heating power are set as adjustable operating variables, and restrictions are imposed on their variation amplitude and rate, considering both the physical constraints of the actuators and preventing excessively drastic operational adjustments. The constraints set for pressure fluctuations, flow deviations, and temperature change rates are expressed in the form of a system of inequalities, constructing a rigorous mathematical framework for the optimization problem and providing a mathematical foundation for solving safe and feasible control schemes.

[0046] The optimization mechanism established in step S4.2 enables the automatic generation of a multivariable collaborative control strategy. The weighted objective function, which comprehensively considers the deviation of the controlled variable and the changes in the manipulated variable, balances the requirements of control accuracy and operational stability, avoiding frequent adjustments caused by solely pursuing control accuracy. Sequential quadratic programming or gradient optimization algorithms are used to solve the constrained optimization problem, obtaining the adjustment trajectory of the manipulated variable that minimizes the objective function while satisfying all safety constraints, thus achieving collaborative optimization of the back pressure valve, feed pump, and heater. The rolling optimization mechanism, which extracts the first set of values ​​as the current control command and repeats the execution in the next cycle, allows the control strategy to be continuously adjusted according to the actual system response, forming a dynamic feedback correction capability.

[0047] Step S4, through the combined action of two smaller steps, achieves model-based prediction-based multivariable optimization control, solving the complex control problem of multiple coupled manipulated variables and simultaneous multiple constraints in supercritical fluid reactions. Traditional single-loop control methods decompose multivariable systems into several independent single-variable loops. The lack of coordination between these loops leads to mutual interference of control actions, making it difficult to achieve overall optimality while satisfying all constraints. This step innovatively uses a predictive model to extrapolate future states, comprehensively considering the coordinated adjustment of multiple manipulated variables within an optimization framework. It solves for the globally optimal control scheme while satisfying multiple constraints such as pressure fluctuations, flow deviations, and temperature change rates, significantly improving the overall performance of the control system.

[0048] This predictive optimization control method offers significant advantages in terms of foresight and systematic approach compared to traditional feedback control. It can anticipate dynamic changes in the system, avoiding overshoot and oscillations caused by delayed responses. The rolling optimization mechanism endows the system with dynamic adaptability, enabling the control strategy to continuously adjust optimization objectives and constraints based on actual operating conditions. This step is particularly suitable for industrial processes with complex dynamic responses, strict operational constraints, and strong coupling of multiple variables. It optimizes control performance while ensuring safe and stable process operation, improving product quality stability and resource utilization efficiency in the reaction process.

[0049] Example 6: Please refer to Figure 1 In step S5, the specific steps of the disturbance compensation mechanism are as follows: S5.1 Real-time monitoring of feed flow and composition data. When the flow rate changes by a step beyond the threshold or the composition deviates from the normal range, it is determined to be a measurable disturbance. Using the model established in step S2, the disturbance is simulated on pressure, temperature and flow based on the current working conditions. The peak value, arrival time and decay characteristics of the influence trajectory are analyzed. The valve opening, pump speed and heating power adjustment required to offset the influence are calculated in reverse and organized into a feedforward compensation signal. S5.2 Maintain the operation of the feedback control loop, collect the measured values ​​of pressure, temperature and flow and calculate their deviation from the set value. Generate a feedback control signal based on the deviation and the rate of change using the proportional-integral-derivative algorithm. Set the initial weighting coefficients of the feedforward and feedback signals. Increase the feedforward weight when the disturbance amplitude is large and the model error is small. Increase the feedback weight when there is unmodeled dynamic or large noise. Adjust the weights according to the operating conditions and control effect. Superimpose the weighted and fused comprehensive signal with the prediction command of step S4 and transmit it to step S6.

[0050] In this embodiment, the feedforward compensation mechanism established in step S5.1 achieves active suppression of measurable disturbances. Real-time monitoring of the feed flow rate and composition data allows for timely identification of step changes or deviations from the normal range based on set thresholds, avoiding the lag problem of disturbances being detected only after they propagate to the reaction system. A calibrated simplified model simulates the impact of disturbances on key system parameters, and by analyzing the peak value, arrival time, and decay characteristics of the impact trajectory, the law governing the disturbance's effect is accurately grasped. The required adjustment amount of the manipulated variable to counteract the disturbance's impact is calculated in reverse, generating a feedforward compensation signal and implementing control measures before the disturbance affects the system. This significantly reduces the deviation amplitude caused by the disturbance on the controlled variable, improving the stability of the process operation.

[0051] The feedback regulation and adaptive weighting mechanism established in step S5.2 enhances the robustness and flexibility of the control system. The feedback loop continuously collects the deviation between the measured value and the setpoint, and generates a control signal through a proportional-integral-derivative algorithm. This effectively handles unpredictable unmodeled dynamics and random disturbances that the feedforward loop cannot predict, ensuring the basic stability of the control system. The adaptive weighting strategy dynamically adjusts the weighting coefficients of the feedforward and feedback signals based on operating condition characteristics such as disturbance amplitude, model error, and measurement noise. When disturbances are significant and the model is accurate, the feedforward role is emphasized to leverage its rapid response advantage; when uncertainties exist, the feedback role is emphasized to ensure stability. This achieves complementary advantages and synergistic cooperation between the two control actions, improving the control system's adaptability to complex operating conditions.

[0052] Step S5 establishes a disturbance suppression system combining active compensation and passive correction through two sub-steps, solving the problem of the impact of external disturbances such as feed fluctuations and composition changes on system stability during supercritical fluid reactions. Traditional feedback control uses a deviation-driven adjustment method, which requires waiting until the disturbance has affected the controlled variable and caused a deviation before taking control measures, resulting in inherent lag and leading to large overshoot and long settling time. This step innovatively combines model-predictive feedforward compensation with deviation feedback-based correction adjustment. The feedforward stage utilizes the measurability of the disturbance and the predictive ability of the model to calculate the compensation amount in advance, while the feedback stage handles uncertainties to ensure robustness. The adaptive weighting mechanism dynamically adjusts the ratio of the two control actions according to the actual operating conditions, achieving an organic unity of speed and robustness.

[0053] Compared with single feedback control, the two-layer compensation control method significantly improves the system's ability to suppress disturbances, reducing the amplitude of fluctuations in the controlled variable caused by disturbances and shortening the recovery time. Combined with the predictive optimization control in step S4, it forms a comprehensive control strategy integrating prediction, compensation, and correction, comprehensively enhancing the anti-interference capability and control quality of the supercritical fluid reaction process.

[0054] Example 7: Please refer to Figure 1 In step S6, the specific steps of actuator coordination management are as follows: S6.1 Receive the integrated control signal output in step S5, query the current status of actuators such as back pressure valve opening, pump speed and heater power, calculate the available adjustment margin of each actuator to the physical limit, start task reallocation when the required adjustment exceeds the available margin, determine the alternative actuator combination based on the coupling relationship of pressure, flow and temperature, calculate the action amount of each mechanism in the alternative combination and reallocate the control task according to the adjustment capacity ratio. S6.2 Establish and update the cumulative number of actions, runtime and response time records of each actuator, calculate the health assessment index, and when multiple actuators can meet the requirements, prioritize the actuator with high health and low load. Before issuing the command, perform three-level verification to verify whether the position parameters are within the safe range, whether the adjustment rate exceeds the allowable value, and whether the coordinated action causes a sudden change. If a risk is detected, limit the amplitude or extend the execution interval. Issue the verified command to the actuator and transmit the action record and health status data to step S7.

[0055] In this embodiment, the saturation constraint problem of actuators is solved through the execution capability assessment and task redistribution mechanism established in step S6.1. After receiving the comprehensive control signal output in step S5, the current operating status of actuators such as the back pressure valve, feed pump, and heater is immediately queried, and the available adjustment margin of each actuator to its physical limit position is calculated to identify potential saturation risks in advance. When the required adjustment exceeds the available margin of a certain actuator, the task redistribution program is automatically started, and alternative actuator combinations that can produce equivalent control effects are determined based on the physical coupling relationship between pressure, flow rate, and temperature. The control tasks are redistributed according to the adjustment capacity ratio of each actuator, which not only avoids control failure caused by the saturation of a single actuator, but also makes full use of the flexibility of multi-actuator coordinated adjustment, improving the control system's ability to cope with extreme operating conditions.

[0056] The health assessment and safety verification mechanism established in step S6.2 achieves load balancing and operational safety assurance for the actuators. Historical operational data, such as the cumulative number of actions, runtime, and response time of each actuator, are continuously recorded to calculate a comprehensive evaluation index reflecting the equipment's health status. When multiple actuators can meet control requirements, the actuator with high health and low cumulative load is prioritized to perform the control task, achieving a balanced distribution of equipment wear and extending the overall service life. The three-level safety verification implemented before issuing commands verifies the safety of control commands from three dimensions: position limitation, rate limitation, and system response. When potential risks are detected, the adjustment amplitude is automatically limited or the execution interval is extended, effectively preventing unsafe operations. Commands that pass verification are issued to the actuators, and the action records and health status data are simultaneously transferred to step S7 for subsequent analysis.

[0057] Step S6 establishes an intelligent coordination and management system for actuators through two smaller steps, resolving key issues such as saturation constraint handling, load balancing, and safety risk prevention in multi-actuator systems. Traditional control systems typically employ fixed actuator configurations. When an actuator reaches its adjustment limit, control performance often drops sharply or even fails, and uneven usage frequencies among actuators lead to premature damage to some equipment. This step innovatively establishes a real-time availability margin assessment and dynamic task reallocation mechanism, initiating alternative solutions before actuators reach saturation, ensuring the continuous effectiveness of the control system under various operating conditions. Health assessment and load balancing strategies result in a more rational distribution of equipment wear, reducing maintenance costs and spare parts consumption.

[0058] This intelligent coordination and management method significantly improves system reliability and flexibility compared to traditional fixed-configuration control. A three-level safety verification mechanism verifies the feasibility and safety of control commands from multiple dimensions, effectively preventing safety accidents such as equipment damage or process instability caused by improper control commands, thus enhancing the system's inherent safety level. The transmission of action records and health status data provides a data foundation for subsequent equipment status prediction, forming a closed-loop management system of execution, monitoring, evaluation, and prediction. This step is particularly suitable for industrial processes with a large number of actuators, limited adjustment range, and high safety requirements, extending equipment life and reducing operating costs while ensuring control performance.

[0059] Example 8: Please refer to Figure 1 In step S7, the specific steps for applying accumulated experience are as follows: S7.1 Record process parameters, control actions and system response data under different raw material batches, catalyst usage stages and environmental conditions, extract operating condition feature vectors including average pressure, temperature distribution, flow stability and conversion rate level, use clustering algorithm to divide historical operating conditions into several typical operating modes, establish feature parameter ranges and corresponding effective control parameter combinations for each mode and store them in the historical database in a structured form. S7.2 During runtime, extract the feature vector of the current operating condition, calculate its distance or similarity with the feature center of each typical mode, select the mode with the closest distance or the highest similarity and retrieve its control parameters as the initial configuration for steps S4 and S5. After completing the control cycle, evaluate performance indicators such as pressure stability, flow fluctuation and temperature deviation. When the standard is met, label the operating condition data with the mode category and add it incrementally to the database. For new operating conditions with similarity below the threshold, select the relatively optimal mode parameters for initialization and record them separately. Transfer the updated database and mode information to step S8.

[0060] In this embodiment: the operating condition knowledge base construction mechanism established in step S7.1 realizes the structured storage and systematic management of operating experience. Process parameters, control actions, and system response data under different raw material batches, catalyst usage stages, and environmental conditions are recorded, accumulating a complete dataset covering multiple operating scenarios. Operating condition feature vectors containing average pressure, temperature distribution, flow stability, and conversion rate levels are extracted, transforming complex multidimensional operating data into feature expressions that are easy to compare and analyze. A clustering algorithm is used to divide historical operating conditions into several typical operating modes, identifying operating condition categories with similar characteristics, and establishing feature parameter ranges and practically verified effective control parameter combinations for each mode. This information is stored in a structured form in the historical database, transforming scattered operating experience into searchable and accessible knowledge resources, providing a reference for quickly responding to similar operating conditions.

[0061] The online operating condition identification and experience retrieval mechanism established in step S7.2 enables historical knowledge to guide current control. During runtime, the feature vector of the current operating condition is extracted and its distance or similarity to the feature centers of each typical mode is calculated, quickly identifying the category to which the current operating condition belongs. The mode with the highest similarity is selected, and its control parameters are retrieved as the initial configuration for steps S4 and S5, significantly shortening the controller parameter tuning time and improving the response speed when facing new batches of raw materials or operating condition changes. After completing the control cycle, performance indicators such as pressure stability, flow fluctuation, and temperature deviation are evaluated. When quality standards are met, the verified and valid operating condition data increments are added to the database, continuously enriching the knowledge base. New operating conditions with significantly different characteristics are recorded and marked separately, accumulating data for the subsequent formation of new typical modes. The updated database and mode information are transferred to step S8 to support long-term trend analysis of equipment status.

[0062] Step S7, through the combined action of two smaller steps, establishes a self-learning mechanism that extracts knowledge from historical data to guide current operations. This solves the problem that control parameter settings in supercritical fluid reaction processes rely heavily on operator experience and are difficult to adapt to changes in operating conditions. Traditional control system parameter tuning primarily relies on operator experience and repeated trials. When encountering new raw material batches or operating conditions, it often requires a lengthy period of trial and error, and the adjustment results vary from person to person. This step innovatively clusters operational data under different operating conditions to form a typical pattern library. Through feature comparison, it achieves rapid identification of operating conditions and automatic recall of mature control schemes, transforming implicit operational experience into explicit knowledge resources. Continuous experience accumulation allows the knowledge base to continuously improve over time, gradually expanding the applicability of control strategies and continuously enhancing the system's adaptability to new operating conditions.

[0063] This data-driven knowledge accumulation method significantly reduces the skill requirements for operators compared to traditional experience-based operations, and improves the intelligence level of the control system. The online operating condition identification and parameter calling mechanism enables the system to quickly reach a stable operating state during operating condition changes, reducing product quality fluctuations and raw material losses during the transition phase.

[0064] Example 9: Please refer to Figure 1 In step S8, the specific steps for predicting the device status are as follows: S8.1 Continuously monitor the historical sequence of model parameters obtained in step S3, extract the time evolution curves of heat transfer coefficient, resistance coefficient and catalyst activity coefficient, use the sliding window method to calculate the rate of change and fluctuation of each parameter, compare the rate of change with the normal range, and determine that when the heat transfer coefficient continues to decrease and exceeds the threshold, it is determined to be surface contamination; when the resistance coefficient continues to increase, it is determined to be channel blockage; when the catalyst activity continues to decline, it is determined to be deactivation. Combine the historical data in step S7 to establish a mapping model between the parameter change trend and the equipment performance degradation state. S8.2 Based on the mapping model in step S8.1, predict the development trajectory of each parameter in the future time period. When it is predicted that a certain parameter will reach the level of concern within a certain period, start pre-compensation. Increase heating power in advance for the decrease in heat transfer coefficient, adjust pump speed or valve opening for the increase in resistance coefficient, and adjust reaction temperature or residence time for the decrease in catalyst activity. Add the compensation amount to the control command in step S4. Generate maintenance suggestions based on the degree of parameter degradation and remaining time. Send a cleaning prompt when the degree of contamination is expected to reach the cleaning threshold, and send a replacement prompt when the activity is expected to drop to the replacement standard. Output the status assessment and maintenance suggestions to the interactive interface.

[0065] In this embodiment: the parameter trend monitoring and equipment degradation diagnosis mechanism established in step S8.1 enables early identification of equipment performance degradation. Continuous monitoring of the historical sequence of model parameters provided in step S3 extracts the time evolution curves of heat transfer coefficient, drag coefficient, and catalyst activity coefficient, revealing the long-term changing patterns of the equipment state. The sliding window method is used to calculate the rate of change and fluctuation amplitude of each parameter, and comparison with normal ranges determines whether parameter changes are abnormal. When the heat transfer coefficient continuously decreases beyond a threshold, it is determined to be heat transfer surface contamination; when the drag coefficient continuously increases, it is determined to be flow channel blockage; and when the catalyst activity continuously declines, it is determined to be catalyst deactivation, establishing a correspondence between parameter change characteristics and equipment degradation types. Combining the historical data accumulated in step S7, a mapping model between parameter change trends and equipment performance degradation status is established, providing a quantitative tool for predicting future equipment status and developing maintenance plans.

[0066] The predictive compensation and maintenance alert mechanism established in step S8.2 achieves a shift from passive response to proactive prevention. Based on the mapping model established in step S8.1, the development trajectory of each parameter over a future time period is predicted. When a parameter is predicted to reach a level of concern affecting normal operation within several control cycles, performance compensation is initiated in advance. For a decrease in heat transfer coefficient, heating power is increased in advance to maintain heat transfer efficiency; for an increase in resistance coefficient, pump speed or valve opening is adjusted to maintain system pressure drop; for a decrease in catalyst activity, reaction temperature or residence time is adjusted to maintain conversion rate. The compensation amount is superimposed on the control command in step S4 to achieve feedforward compensation for performance degradation. Equipment maintenance recommendations are generated based on the degree of parameter degradation and the predicted remaining available time. A cleaning alert is sent when the contamination level is expected to reach the cleaning threshold, and a replacement alert is sent when the catalyst activity is expected to drop to the replacement standard, providing operators with a basis for planned maintenance decisions.

[0067] Step S8, through the combined action of two smaller steps, establishes a complete system for equipment condition monitoring, performance prediction, and proactive maintenance. This addresses the issues of gradual equipment performance degradation affecting process stability and unplanned shutdowns caused by sudden failures in supercritical fluid reactors. Traditional maintenance strategies primarily employ periodic maintenance or post-failure maintenance. Periodic maintenance is often overly conservative, leading to high costs and potential unnecessary shutdowns, while post-failure maintenance results in production losses and safety risks due to equipment failure. This step innovatively identifies early signs of equipment performance degradation through long-term trend analysis of model parameters, establishing a quantitative correlation between parameter evolution and equipment degradation status. This enables accurate diagnosis of equipment condition and reliable prediction of future trends. The predictive compensation mechanism compensates for equipment performance degradation by adjusting operating parameters before it impacts product quality, extending the equipment's usability in its degraded state. The maintenance reminder function provides optimal timing suggestions for planned maintenance.

[0068] This predictive maintenance approach significantly improves equipment utilization and maintenance efficiency compared to traditional maintenance strategies. By identifying equipment degradation trends in advance and taking compensatory measures, it avoids product quality fluctuations and increased energy consumption caused by equipment performance decline, and extends uptime between maintenance sessions. Maintenance recommendations based on actual conditions make maintenance activities more precise, avoiding resource waste from over-maintenance and the risk of failure from under-maintenance. This approach is particularly suitable for industrial processes with significant gradual equipment performance degradation, high maintenance costs, and substantial downtime losses, effectively reducing unplanned shutdowns, lowering maintenance costs, and improving the continuity and economic efficiency of plant operation.

[0069] This application also provides an environmentally friendly continuous flow high-pressure reaction system based on supercritical fluids. Please refer to [link / reference]. Figure 2It includes a reaction device module, a data acquisition module, a model calculation module, an optimization control module, an execution adjustment module, a historical data module, a state prediction module, and a human-computer interaction module; The reaction unit module is used to carry out the supercritical fluid reaction process. The feed end receives raw materials, the catalyst in the reaction section completes the chemical conversion, and the outlet section outputs products. The overall pressure bearing capacity meets the supercritical operating conditions. The data acquisition module is used to monitor the reaction process status in real time. It collects pressure, temperature, flow rate and density parameters through sensors placed at key locations. After filtering to eliminate interference, the data is transmitted to the computing unit. The model calculation module is used to predict the system's operating state. It uses a simplified fluid dynamics model combined with state equations to calculate real-time physical property parameters, couples reaction kinetics to obtain the reaction rate, and automatically corrects internal parameters through a calibration function to make the prediction results consistent with the actual operation. The optimization control module is used to generate the optimal control scheme, and to deduce the future evolution trend of the system based on the calibration model. Under the premise of meeting safety constraints, it calculates the coordinated adjustment scheme of the back pressure valve, feed pump and heater, and integrates multiple control strategies such as prediction, compensation and feedback. The execution adjustment module is used to implement control commands. According to the commands of the optimization control module, it drives the back pressure valve to adjust the pressure, the feed pump to adjust the flow rate, and the heater to adjust the temperature. It also monitors the operating status and cumulative load of each mechanism and feeds back the execution status to the control loop. The historical data module is used to store operational experience and identify operating conditions. It categorizes and records operational data under different raw material, catalyst states, and environmental conditions, establishes a typical operating condition mode library, quickly matches the current state during operation and calls verified control schemes, and continuously absorbs new experience to enrich the knowledge base. The condition prediction module is used to predict the trend of equipment performance changes, monitor the long-term evolution of model parameters, identify performance degradation signs such as catalyst decay, heat transfer efficiency decline and flow resistance increase, predict the development trend and start compensation adjustment in advance, and generate maintenance timing suggestions. The human-machine interface module connects operators with the control system, displays current operating parameters, control actions, equipment health assessments and maintenance prompts, receives process settings and mode selections, and supports visual management of the system and necessary manual intervention.

[0070] The contents not described in detail in this specification are existing technologies known to those skilled in the art.

[0071] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid, characterized in that: The specific steps are as follows: S1. Distributed data acquisition: Pressure sensors, temperature sensors, flow meters, and density meters are respectively arranged at the feed end, reaction section, and outlet section of the reaction device. The sensors acquire data in a high-frequency manner and transmit it to the data processing unit after filtering. The local fluid density and phase information are calculated through soft measurement algorithms. S2. Simplified Model Establishment: Feature extraction is performed on the offline three-dimensional flow field simulation results to construct a simplified fluid dynamics model containing the main flow characteristics. This model is combined with the supercritical fluid state equation to calculate real-time physical property parameters and coupled with the reaction rate calculation module so that a single prediction calculation meets the real-time control requirements. S3, Model Calibration Synchronization: The process parameters measured in real time are used as input to drive the simplified model to run. The allowable deviation range between the predicted value and the measured value is set. When the deviation exceeds the range, the heat transfer coefficient, resistance coefficient and catalyst activity parameters inside the model are automatically adjusted. The best estimate of the system operating status is obtained through data fusion technology. S4. Multivariable Optimization Control: Using the calibrated model as a tool for predicting future states, a prediction time window is set. Under the premise of satisfying constraints such as pressure fluctuation, flow deviation and temperature change rate, the optimal adjustment scheme of back pressure valve opening, feed pump speed and heater power is calculated. After solving the optimization problem, the first step of control action is executed. S5. Disturbance compensation mechanism: For observable feed fluctuations or composition changes, the impact on the system is calculated in advance based on the model's response characteristics, and a compensation signal is generated. At the same time, the deviation-based feedback adjustment function is maintained to handle uncertainties. The two adjustment signals are adaptively weighted according to the current working conditions and then act together on the actuator. S6. Actuator Coordination Management: Before issuing control commands, check the action capability boundaries of each actuator. When an actuator reaches its adjustment limit, allocate some control tasks to other available equipment. Allocate control actions reasonably according to the cumulative workload of each actuator and set safety limits to prevent operation from exceeding the limits. S7. Experience accumulation and application: Establish a historical operation record library covering different raw material characteristics, catalyst state and environmental conditions, summarize past operating conditions into several typical categories, identify the current category through feature comparison during operation and adopt the corresponding mature control scheme, and continuously incorporate new successful experiences into the record library to improve the control strategy. S8. Equipment Status Prediction: Track the changing trends of model parameters to establish a predictive relationship for equipment performance degradation, judge the development process of conditions such as catalyst activity decline, seal aging or heat transfer surface contamination, and when it is expected to reach a level that requires attention in the short term, fine-tune the operating parameters in advance to compensate, and remind the operators of the appropriate maintenance time.

2. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 1, characterized in that: In step S1, the specific steps of distributed data acquisition are as follows: S1.1 A pressure transmitter and a flow meter are installed at the feed end of the reaction device. Temperature measuring points are arranged along the flow direction in the reaction section. A density meter and a pressure sensor are installed at the outlet section. The sensors acquire process parameters by continuous sampling and transmit them to the data processing unit via the industrial communication network. The transmitted signals are digitally filtered to eliminate interference and vibration effects. S1.2 The data processing unit calculates the flow resistance based on the pressure measurement value and pipeline parameters, estimates the heat transfer parameters based on the temperature distribution and heat flux density, obtains the fluid density and viscosity using the pressure and temperature values ​​through physical property correlation formulas or table lookups, organizes the measured data and estimated parameters into a time-stamped data sequence and transmits it to the model calculation unit, and simultaneously establishes a rolling window to store recent data.

3. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 2, characterized in that: In step S2, the specific steps for simplifying the model establishment are as follows: S2.1 Perform modal decomposition on the three-dimensional flow field simulation data, extract the main spatial modes and time evolution coefficients, reconstruct and simplify the fluid dynamics model based on the dominant modes, retain the velocity distribution and pressure drop characteristics and reduce the calculation dimension, convert the three-dimensional mesh into a one-dimensional pipeline model and a quasi-two-dimensional cross-section model, and control the number of degrees of freedom through the modal truncation threshold; S2.2 Establish a data interface between the simplified fluid dynamics model and the supercritical fluid state equation. At each calculation moment, calculate the fluid density and viscosity values ​​based on the temperature and pressure values ​​and update them to the constitutive relation of the flow field model. At the same time, embed the reaction dynamics module, calculate the reaction rate based on the local temperature and concentration as the source term and introduce it into the equation to complete the iterative calculation of the flow field, physical property parameters and reaction rate.

4. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 3, characterized in that: In step S3, the specific steps for model calibration synchronization are as follows: S3.1 Input the real-time parameters obtained in step S1 into the simplified model established in step S2 to drive it to perform calculations. The model outputs the predicted values ​​of pressure, temperature and flow at each key location. The predicted values ​​are compared with the corresponding sensor measurements to calculate the deviation. Deviation thresholds are set for different parameters. When the deviation of the measuring point exceeds the threshold in multiple consecutive cycles, it is determined that the model is mismatched with the actual state and the deviation characteristics are recorded. S3.

2. Based on the deviation characteristics in step S3.1, start the parameter identification program, analyze the deviation amplitude and location to determine the correction parameter category. When the temperature deviation is the main factor, adjust the heat transfer coefficient; when the pressure deviation is the main factor, adjust the drag coefficient; when the conversion rate deviation is the main factor, adjust the catalyst activity coefficient. Use the least squares method or gradient optimization algorithm to solve for the parameter value that minimizes the sum of squares of the deviations and update the model. Apply Kalman filtering to fuse the updated prediction results with the measurement data to generate the optimal estimates of system pressure, temperature and flow rate, and pass them to step S4.

5. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 4, characterized in that: In step S4, the specific steps of multivariate optimization control are as follows: S4.1 Using the state estimate obtained in step S3 as the initial condition, the system's evolution trajectory within the prediction window is deduced using the calibrated model. The prediction window length is set to cover the main dynamic response characteristics. The back pressure valve opening, pump speed, and heating power are set as adjustable operating variables, and their variation amplitude and rate limits are set. Constraints are set for pressure fluctuations, flow deviations, and temperature change rates. The constraints are expressed in the form of a system of inequalities, and a constrained optimization mathematical model is established. S4.2 Establish a weighted objective function that comprehensively reflects the deviation of the controlled variable and the change of the operating variable. Use a sequential quadratic programming algorithm or gradient optimization algorithm to solve the operating variable adjustment trajectory that minimizes the objective function under the constraints of step S4.

1. This trajectory includes the valve opening, pump speed and heating power values ​​at each time. Extract the first set of values ​​as a control command and send it to the actuator. In the next cycle, the prediction and optimization process is repeated with the state estimate updated in step S3 as the initial condition.

6. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 5, characterized in that: In step S5, the specific steps of the disturbance compensation mechanism are as follows: S5.1 Real-time monitoring of feed flow and composition data. When the flow rate changes by a step beyond the threshold or the composition deviates from the normal range, it is determined to be a measurable disturbance. Using the model established in step S2, the disturbance is simulated on pressure, temperature and flow based on the current working conditions. The peak value, arrival time and decay characteristics of the influence trajectory are analyzed. The valve opening, pump speed and heating power adjustment required to offset the influence are calculated in reverse and organized into a feedforward compensation signal. S5.2 Maintain the operation of the feedback control loop, collect the measured values ​​of pressure, temperature and flow and calculate their deviation from the set value. Generate a feedback control signal based on the deviation and the rate of change using the proportional-integral-derivative algorithm. Set the initial weighting coefficients of the feedforward and feedback signals. Increase the feedforward weight when the disturbance amplitude is large and the model error is small. Increase the feedback weight when there is unmodeled dynamic or large noise. Adjust the weights according to the operating conditions and control effect. Superimpose the weighted and fused comprehensive signal with the prediction command of step S4 and transmit it to step S6.

7. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 6, characterized in that: In step S6, the specific steps of actuator coordination management are as follows: S6.1 Receive the integrated control signal output in step S5, query the current status of actuators such as back pressure valve opening, pump speed and heater power, calculate the available adjustment margin of each actuator to the physical limit, start task reallocation when the required adjustment exceeds the available margin, determine the alternative actuator combination based on the coupling relationship of pressure, flow and temperature, calculate the action amount of each mechanism in the alternative combination and reallocate the control task according to the adjustment capacity ratio. S6.2 Establish and update the cumulative number of actions, runtime and response time records of each actuator, calculate the health assessment index, and when multiple actuators can meet the requirements, prioritize the actuator with high health and low load. Before issuing the command, perform three-level verification to verify whether the position parameters are within the safe range, whether the adjustment rate exceeds the allowable value, and whether the coordinated action causes a sudden change. If a risk is detected, limit the amplitude or extend the execution interval. Issue the verified command to the actuator and transmit the action record and health status data to step S7.

8. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 7, characterized in that: In step S7, the specific steps for applying accumulated experience are as follows: S7.1 Record process parameters, control actions and system response data under different raw material batches, catalyst usage stages and environmental conditions, extract operating condition feature vectors including average pressure, temperature distribution, flow stability and conversion rate level, use clustering algorithm to divide historical operating conditions into several typical operating modes, establish feature parameter ranges and corresponding effective control parameter combinations for each mode and store them in the historical database in a structured form. S7.2 During runtime, extract the feature vector of the current operating condition, calculate its distance or similarity with the feature center of each typical mode, select the mode with the closest distance or the highest similarity and retrieve its control parameters as the initial configuration for steps S4 and S5. After completing the control cycle, evaluate performance indicators such as pressure stability, flow fluctuation and temperature deviation. When the standard is met, label the operating condition data with the mode category and add it incrementally to the database. For new operating conditions with similarity below the threshold, select the relatively optimal mode parameters for initialization and record them separately. Transfer the updated database and mode information to step S8.

9. The environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid according to claim 8, characterized in that: In step S8, the specific steps for predicting the device status are as follows: S8.1 Continuously monitor the historical sequence of model parameters obtained in step S3, extract the time evolution curves of heat transfer coefficient, resistance coefficient and catalyst activity coefficient, use the sliding window method to calculate the rate of change and fluctuation of each parameter, compare the rate of change with the normal range, and determine that when the heat transfer coefficient continues to decrease and exceeds the threshold, it is determined to be surface contamination; when the resistance coefficient continues to increase, it is determined to be channel blockage; when the catalyst activity continues to decline, it is determined to be deactivation. Combine the historical data in step S7 to establish a mapping model between the parameter change trend and the equipment performance degradation state. S8.2 Based on the mapping model in step S8.1, predict the development trajectory of each parameter in the future time period. When it is predicted that a certain parameter will reach the level of concern within a certain period, start pre-compensation. Increase heating power in advance for the decrease in heat transfer coefficient, adjust pump speed or valve opening for the increase in resistance coefficient, and adjust reaction temperature or residence time for the decrease in catalyst activity. Add the compensation amount to the control command in step S4. Generate maintenance suggestions based on the degree of parameter degradation and remaining time. Send a cleaning prompt when the degree of contamination is expected to reach the cleaning threshold, and send a replacement prompt when the activity is expected to drop to the replacement standard. Output the status assessment and maintenance suggestions to the interactive interface.

10. An environmentally friendly continuous flow high-pressure reaction system based on supercritical fluid, characterized in that: The environmentally friendly continuous flow high-pressure reaction system based on supercritical fluid is used to execute the environmentally friendly continuous flow high-pressure reaction method based on supercritical fluid as described in any one of claims 1 to 9. The system includes a reaction device module, a data acquisition module, a model calculation module, an optimization control module, an execution adjustment module, a historical data module, a state prediction module, and a human-computer interaction module. The reaction unit module is used to carry out the supercritical fluid reaction process. The feed end receives raw materials, the catalyst in the reaction section completes the chemical conversion, and the outlet section outputs products. The overall pressure bearing capacity meets the supercritical operating conditions. The data acquisition module is used to monitor the reaction process status in real time. It collects pressure, temperature, flow rate and density parameters through sensors placed at key locations. After filtering to eliminate interference, the data is transmitted to the computing unit. The model calculation module is used to predict the system's operating state. It uses a simplified fluid dynamics model combined with state equations to calculate real-time physical property parameters, couples reaction kinetics to obtain the reaction rate, and automatically corrects internal parameters through a calibration function to make the prediction results consistent with the actual operation. The optimization control module is used to generate the optimal control scheme, and to deduce the future evolution trend of the system based on the calibration model. Under the premise of meeting safety constraints, it calculates the coordinated adjustment scheme of the back pressure valve, feed pump and heater, and integrates multiple control strategies such as prediction, compensation and feedback. The execution adjustment module is used to implement control commands. According to the commands of the optimization control module, it drives the back pressure valve to adjust the pressure, the feed pump to adjust the flow rate, and the heater to adjust the temperature. It also monitors the operating status and cumulative load of each mechanism and feeds back the execution status to the control loop. The historical data module is used to store operational experience and identify operating conditions. It categorizes and records operational data under different raw material, catalyst states, and environmental conditions, establishes a typical operating condition mode library, quickly matches the current state during operation and calls verified control schemes, and continuously absorbs new experience to enrich the knowledge base. The condition prediction module is used to predict the trend of equipment performance changes, monitor the long-term evolution of model parameters, identify performance degradation signs such as catalyst decay, heat transfer efficiency decline and flow resistance increase, predict the development trend and start compensation adjustment in advance, and generate maintenance timing suggestions. The human-machine interface module connects operators with the control system, displays current operating parameters, control actions, equipment health assessments and maintenance prompts, receives process settings and mode selections, and supports visual management of the system and necessary manual intervention.

Images