A reactor temperature control system and a temperature control method thereof

Abstract

This invention relates to a reactor temperature control system, comprising a macroscopic model prediction and control unit, a local temperature field analysis and dynamic compensation unit, a multi-source data fusion sensing unit, an adaptive safety decision-making unit, and a control command fusion execution unit. This invention enables simultaneous and precise tracking of the overall reactor temperature and high spatial homogenization of the internal temperature field, while embedding active safety protection throughout the entire process, thereby significantly improving the product quality, consistency, and operational safety of complex chemical reactions.

Description

Technical Field

[0001] This invention relates to the field of chemical process control technology, specifically to a reaction vessel temperature control system and its temperature control method, and particularly to an intelligent temperature control system that combines macroscopic process control with microscopic temperature field uniformity regulation. Background Technology

[0002] Reactors are core equipment for chemical reactions in fields such as fine chemicals, pharmaceuticals, and materials synthesis. The ability to achieve precise, stable, and uniform temperature control within the reactor directly determines the selectivity, conversion rate, product quality (such as molecular weight distribution and crystal morphology), and production safety. Many chemical reactions exhibit strong exothermic or endothermic characteristics, and the heat transfer process involves hysteresis and nonlinearity, placing extremely high demands on the performance of the control system.

[0003] Most existing methods separate macroscopic overall temperature control from microscopic local temperature field uniformity regulation, or while related, the coupling is shallow. They either emphasize overall tracking while neglecting local uniformity, or attempt local compensation but lack coordination and foresight with the overall process. Furthermore, the safety mechanisms in existing solutions are mostly used as optimization constraints or ex-post fallback strategies, failing to permeate the entire control decision-making process. Therefore, how to provide a reactor temperature control system and method that deeply integrates macroscopic process control and microscopic temperature field regulation, achieves dynamic adaptation and forward-looking compensation, and incorporates intelligent active safety protection, is an unsolved problem. Summary of the Invention

[0004] To address the shortcomings of existing technologies, the present invention aims to provide a reactor temperature control system that can simultaneously achieve precise tracking of the overall temperature of the reactor and high uniformity of the internal temperature field, while embedding active safety protection throughout the process, thereby significantly improving the product quality, consistency, and operational safety of complex chemical reactions.

[0005] One of the above-mentioned objectives of this invention is achieved through the following technical solution: A temperature control system for a reaction vessel, characterized in that it comprises: The macroscopic model predictive control unit is used to generate macroscopic reference control commands by using the model predictive control algorithm to solve for the optimal jacket medium flow rate or temperature setpoint based on the overall thermal balance model of the reactor, with the average temperature inside the reactor or the temperature at key points as the controlled variable. The local temperature field analysis and dynamic compensation unit is used to receive reactant concentration, viscosity, stirring speed and temperature data of multiple regions in the reactor provided by the multi-source data fusion sensing unit, construct a dynamic temperature field prediction model that integrates reaction exothermic kinetics and space heat transfer, identify temperature non-uniform regions at the current and future times, and calculate dynamic heat flow compensation commands for specific local regions. The multi-source data fusion sensing unit is used to collect and process multi-dimensional data from the reactor and auxiliary systems in real time, including multi-point temperature inside the reactor, jacket inlet and outlet temperature and flow rate, stirrer power / speed, online component analysis signal, and provide consistent data input for the macro model prediction control unit and the local temperature field analysis and dynamic compensation unit. The adaptive safety decision unit is used to set dynamic safety boundaries for macro-based baseline control commands and dynamic heat flux compensation commands respectively, and to arbitrate and correct conflicting or boundary-crossing commands based on the current reaction stage, historical control effects and predicted risks, and output a fused set of safety constraint control commands. The control command fusion execution unit is used to parse and distribute the safety constraint control command set to the corresponding actuators, including the jacketed circulating pump frequency converter, multiple zone regulating valves, and auxiliary heating / cooling units, so as to collaboratively achieve the composite control objective of macro temperature tracking and local temperature uniformity.

[0006] Preferably, the local temperature field analysis and dynamic compensation unit includes: The dynamic parameter identification module is used to correct the reaction exothermic kinetic parameters and the effective heat transfer coefficient of the fluid online based on the real-time acquired reaction exothermic rate precursor signal and stirring power. The spatiotemporal fusion prediction module is used to discretize the reactor space into multiple interconnected control volume units, establish the energy balance equation for each unit, and include the reaction exothermic term provided by the dynamic parameter identification module, as well as the heat transfer and jacket heat exchange term from adjacent units. Based on the current state, it predicts the spatiotemporal temperature evolution of each unit over a future period of time. The non-uniformity diagnosis and compensation module is used to calculate the current and predicted future non-uniformity index of the temperature field inside the vessel based on the output of the spatiotemporal fusion prediction module. When the index exceeds the adaptive threshold, the main non-uniform region is located, and the local compensation command required to offset the non-uniformity is calculated in combination with the predicted thermal state of the region.

[0007] Preferably, in the spatiotemporal fusion prediction module, the energy balance equation for each control volume unit is: ,in, , , , Let represent the density, specific heat capacity, volume, and temperature of the i-th element, respectively. The reaction exothermic rate is based on the current concentration c and the parameter k corrected by the dynamic parameter identification module; For the heat transfer term with adjacent unit j; This refers to the heat exchange term of the jacket for this unit.

[0008] Preferably, the adaptive security decision-making unit includes: The layered safety boundary setting module is used to set hard constraint boundaries based on process safety for macro-based baseline control commands and soft constraint boundaries based on equipment capacity and thermal shock risk for dynamic heat flux compensation commands. The soft constraint boundaries are dynamically adjusted according to the current reaction intensity and temperature field stability. The instruction conflict arbitration module is used to arbitrate when macro heating instructions and local cooling compensation instructions exist simultaneously and may cancel each other out, or when the demand for multiple local compensation instructions exceeds the total execution capacity, based on a preset priority strategy, and generate an executable instruction combination. The forward-looking risk mitigation module is used to identify potential future overheating or severe temperature gradient risk points based on the future temperature field distribution output by the spatiotemporal fusion prediction module, and to apply damping or reverse correction to the control commands of the current cycle in advance.

[0009] Preferably, in the layered safety boundary setting module, the dynamic adjustment rule of the soft constraint boundary is as follows: when the reaction exothermic rate exceeds the set threshold or the temperature field non-uniformity index rises rapidly, the allowable adjustment range of the local compensation command is reduced; when the reaction enters the stable isothermal stage, the allowable adjustment range is relaxed to optimize uniformity.

[0010] Preferably, a method for controlling the temperature of a reaction vessel is characterized by comprising the following steps: S1: Real-time acquisition of multi-dimensional operating data of the reactor through a multi-source data fusion sensing unit; S2: The macroscopic model prediction control unit is based on the overall thermal balance model of the reactor and the set temperature trajectory, and solves the macroscopic reference control command in a rolling manner. S3: The local temperature field analysis and dynamic compensation unit, based on the operating data of step S1, uses the dynamic temperature field prediction model to perform spatiotemporal prediction, diagnose temperature non-uniformity, and generate dynamic heat flow compensation commands. S4: The adaptive safety decision unit receives macro-based baseline control instructions and dynamic heat flow compensation instructions, performs safety boundary verification, conflict arbitration and risk suppression, and outputs a fused set of safety constraint control instructions. S5: The control instruction fusion execution unit executes the safety constraint control instruction set to drive the corresponding actuators to move.

[0011] Preferably, in step S3, the specific process of constructing and applying the dynamic temperature field prediction model includes: S31: Using current time and recent historical data, the reaction exothermic kinetic parameters and the equivalent heat transfer coefficient between units are identified online by recursive least squares method or Kalman filter algorithm. S32: Taking each discretized control volume unit as the object, substitute the parameters identified in step S31, solve the coupled energy balance equation set, and predict the distribution of the three-dimensional temperature field inside the vessel in the next few control cycles. S33: Based on the temperature field distribution predicted in step S32, calculate the global and zone temperature non-uniformity index. If the index exceeds the limit, locate the hot or cold spot area, and calculate the compensation command for the specific zone valve or auxiliary component according to the predicted temperature deviation and change trend of the area.

[0012] Preferably, in step S4, the specific rules for conflict arbitration by the instruction conflict arbitration module include: if a macro instruction requires overall heating while a local compensation instruction requires cooling of a certain area, the local cooling compensation instruction is executed first, while the macro heating instructions for other areas are proportionally reduced to maintain overall thermal balance; if the total demand of multiple local compensation instructions exceeds the total capacity of the actuator, the compensation demand of the area with the largest temperature deviation or the fastest rate of change is satisfied first.

[0013] Preferably, after step S5 is executed, a feedback learning step S6 is further included: comparing the actual temperature field change with the temperature field predicted in step S32, calculating the prediction error; and using the prediction error, fine-tuning the parameters or compensation strategy model of the dynamic temperature field prediction model online through machine learning algorithms or model parameter adaptive laws.

[0014] Preferably, in the feedback learning step S6, a recursive learning algorithm with a forgetting factor is used, so that the model can gradually adapt to slow time-varying processes such as reactor scaling and catalyst activity changes, while not overly relying on recent accidental perturbations.

[0015] In summary, the beneficial technical effects of this invention are as follows: Through a collaborative architecture comprising a "macroscopic model prediction control unit" and a "local temperature field analysis and dynamic compensation unit," the system simultaneously achieves high-precision tracking of the overall temperature of the reactor and high homogenization of the internal three-dimensional temperature field. This solves the problem of traditional lumped control struggling to handle spatial temperature gradients, significantly improving the synthesis quality and consistency of high-end chemicals. By incorporating a "dynamic temperature field prediction model" integrating reaction exothermic kinetics and spatial heat transfer, along with online parameter identification, the system enables forward-looking prediction and adaptive compensation of local temperature non-uniformity trends, making control actions more predictable and precise, and eliminating reliance on fixed thresholds and delayed responses. Furthermore, the system features a multi-level, proactive "adaptive safety decision unit," which sets dynamic safety boundaries for all control commands, intelligently arbitrates command conflicts, and proactively mitigates risks based on predictions. This deeply integrates safety assurance into the control loop, greatly enhancing the system's robustness and intrinsic safety while pursuing high performance. (See attached figures.) Figure 1 This is an overall architecture block diagram of the reactor temperature control system provided in an embodiment of the present invention; Figure 2This is a diagram showing the internal module composition and data flow of the local temperature field analysis and dynamic compensation unit in this embodiment of the invention. Figure 3 This is a flowchart illustrating the internal module composition and workflow of the adaptive security decision-making unit in this embodiment of the invention. Figure 4 This is a flowchart of the steps of the reaction vessel temperature control method provided in the embodiments of the present invention.

[0016] In the diagram, 1. Macroscopic model prediction and control unit; 2. Local temperature field analysis and dynamic compensation unit; 21. Dynamic parameter identification module; 22. Spatiotemporal fusion prediction module; 23. Non-uniformity diagnosis and compensation module; 3. Multi-source data fusion sensing unit; 4. Adaptive safety decision-making unit; 41. Hierarchical safety boundary setting module; 42. Command conflict arbitration module; 43. Forward-looking risk suppression module; 5. Control command fusion execution unit. Detailed Implementation

[0017] The present invention will be further described in detail below with reference to the accompanying drawings.

[0018] Reference Figures 1-3 A reactor temperature control system includes: a macroscopic model predictive control unit (1), which is used to generate macroscopic reference control commands by using a model predictive control algorithm to solve for the optimal jacket medium flow rate or temperature setpoint based on the overall thermal balance model of the reactor, with the average temperature inside the reactor or the temperature at key points as the controlled variable. The local temperature field analysis and dynamic compensation unit (2) is used to receive reactant concentration, viscosity, stirring speed and temperature data of multiple regions in the reactor provided by the multi-source data fusion sensing unit (3), construct a dynamic temperature field prediction model that integrates reaction exothermic dynamics and space heat transfer, identify temperature non-uniform regions at the current and future times, and calculate dynamic heat flow compensation commands for specific local regions. The multi-source data fusion sensing unit (3) is used to collect and process multi-dimensional data from the reactor and auxiliary system in real time, including multi-point temperature inside the reactor, jacket inlet and outlet temperature and flow rate, stirrer power / speed, online component analysis signal, and provide consistent data input for the macro model prediction control unit (1) and the local temperature field analysis and dynamic compensation unit (2); The adaptive safety decision unit (4) is used to set dynamic safety boundaries for macro-based control commands and dynamic heat flow compensation commands respectively, and to arbitrate and correct conflicting or out-of-bounds commands based on the current reaction stage, historical control effects and predicted risks, and output the fused safety constraint control command set. The control command fusion execution unit (5) is used to parse and distribute the safety constraint control command set to the corresponding actuators, including the jacketed circulating pump frequency converter, multiple zone regulating valves, and auxiliary heating / cooling units, so as to achieve the composite control objective of macro temperature tracking and local temperature uniformity in a coordinated manner.

[0019] Preferably, the local temperature field analysis and dynamic compensation unit (2) includes: The dynamic parameter identification module (21) is used to correct the reaction heat release kinetic parameters and the effective heat transfer coefficient of the fluid online based on the real-time acquired reaction heat release rate precursor signal and stirring power. The spatiotemporal fusion prediction module (22) is used to discretize the reactor space into multiple interconnected control volume units, establish the energy balance equation of each unit, and include the reaction exothermic term provided by the dynamic parameter identification module (21), as well as the heat transfer and jacket heat exchange term of the adjacent units, and predict the spatiotemporal temperature evolution of each unit in the future based on the current state. The non-uniformity diagnosis and compensation module (23) is used to calculate the current and predicted future non-uniformity index of the temperature field inside the vessel based on the output of the spatiotemporal fusion prediction module (22). When the index exceeds the adaptive threshold, the main non-uniform region is located, and the local compensation command required to offset the non-uniformity is calculated in combination with the predicted thermal state of the region.

[0020] Preferably, in the spatiotemporal fusion prediction module (22), the energy balance equation for each control volume unit is: ,in, , , , Let represent the density, specific heat capacity, volume, and temperature of the i-th element, respectively. The reaction exothermic rate is based on the current concentration c and the parameter k corrected by the dynamic parameter identification module (21); For the heat transfer term with adjacent unit j; This refers to the heat exchange term of the jacket for this unit.

[0021] Preferably, the adaptive security decision unit (4) includes: The layered safety boundary setting module (41) is used to set hard constraint boundaries based on process safety for macro-baseline control commands and soft constraint boundaries based on equipment capacity and thermal shock risk for dynamic heat flow compensation commands. The soft constraint boundaries are dynamically adjusted according to the current reaction intensity and temperature field stability. The instruction conflict arbitration module (42) is used to arbitrate according to a preset priority strategy when macro heating instructions and local cooling compensation instructions exist at the same time and may cancel out the control effect, or when the demand of multiple local compensation instructions exceeds the total execution capacity, and generate an executable instruction combination. The forward-looking risk suppression module (43) is used to identify potential future overheating or severe temperature gradient risk points based on the future temperature field distribution output by the spatiotemporal fusion prediction module (22), and apply damping or reverse correction to the control command of the current cycle in advance.

[0022] Preferably, in the layered safety boundary setting module (41), the dynamic adjustment rule of the soft constraint boundary is as follows: when the reaction exothermic rate exceeds the set threshold or the temperature field non-uniformity index rises rapidly, the allowable adjustment range of the local compensation command is reduced; when the reaction enters the stable isothermal stage, the allowable adjustment range is relaxed to optimize uniformity.

[0023] Reference Figure 4 Preferably, a method for controlling the temperature of a reaction vessel includes the following steps: S1: Real-time acquisition of multi-dimensional operating data of the reactor through the multi-source data fusion sensing unit (3); S2: Macroscopic model prediction control unit (1) Based on the overall thermal balance model of the reactor and the set temperature trajectory, it solves the macroscopic reference control command in a rolling manner. S3: Local temperature field analysis and dynamic compensation unit (2) Based on the running data of step S1, use the dynamic temperature field prediction model to make spatiotemporal predictions, diagnose temperature non-uniformity and generate dynamic heat flow compensation commands. S4: Adaptive safety decision unit (4) receives macro-based baseline control instructions and dynamic heat flow compensation instructions, performs safety boundary verification, conflict arbitration and risk suppression, and outputs the fused safety constraint control instruction set; S5: Control instruction fusion execution unit (5) executes the safety constraint control instruction set and drives the corresponding actuator to move.

[0024] Preferably, in step S3, the specific process of constructing and applying the dynamic temperature field prediction model includes: S31: Using current time and recent historical data, the reaction exothermic kinetic parameters and the equivalent heat transfer coefficient between units are identified online by recursive least squares method or Kalman filter algorithm. S32: Taking each discretized control volume unit as the object, substitute the parameters identified in step S31, solve the coupled energy balance equation set, and predict the distribution of the three-dimensional temperature field inside the vessel in the next few control cycles. S33: Based on the temperature field distribution predicted in step S32, calculate the global and zone temperature non-uniformity index. If the index exceeds the limit, locate the hot or cold spot area, and calculate the compensation command for the specific zone valve or auxiliary component according to the predicted temperature deviation and change trend of the area.

[0025] Preferably, in step S4, the specific rules for conflict arbitration by the instruction conflict arbitration module (42) include: if a macro instruction requires overall heating while a local compensation instruction requires cooling of a certain area, the local cooling compensation instruction is executed first, and the macro heating instructions of other areas are proportionally reduced to maintain overall thermal balance; if the total demand of multiple local compensation instructions exceeds the total capacity of the actuator, the compensation demand of the area with the largest temperature deviation or the fastest rate of change is satisfied first.

[0026] Preferably, after step S5 is executed, a feedback learning step S6 is further included: comparing the actual temperature field change with the temperature field predicted in step S32, calculating the prediction error; and using the prediction error, fine-tuning the parameters or compensation strategy model of the dynamic temperature field prediction model online through machine learning algorithms or model parameter adaptive laws.

[0027] Preferably, in the feedback learning step S6, a recursive learning algorithm with a forgetting factor is used, so that the model can gradually adapt to slow time-varying processes such as reactor scaling and catalyst activity changes, while not overly relying on recent accidental perturbations.

[0028] The embodiments described herein are preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape, and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A temperature control system for a reaction vessel, characterized in that, include: The macroscopic model predictive control unit (1) is used to generate macroscopic reference control commands by using the model predictive control algorithm to solve the optimal jacket medium flow rate or temperature setpoint based on the overall thermal balance model of the reactor, with the average temperature inside the reactor or the temperature at key points as the controlled variable. The local temperature field analysis and dynamic compensation unit (2) is used to receive reactant concentration, viscosity, stirring speed and temperature data of multiple regions in the reactor provided by the multi-source data fusion sensing unit (3), construct a dynamic temperature field prediction model that integrates reaction exothermic dynamics and space heat transfer, identify temperature non-uniform regions at the current and future times, and calculate dynamic heat flow compensation commands for specific local regions. The multi-source data fusion sensing unit (3) is used to collect and process multi-dimensional data from the reactor and auxiliary system in real time, including multi-point temperature inside the reactor, jacket inlet and outlet temperature and flow rate, stirrer power / speed, online component analysis signal, and provide consistent data input for the macro model prediction control unit (1) and the local temperature field analysis and dynamic compensation unit (2); The adaptive safety decision unit (4) is used to set dynamic safety boundaries for macro-based control commands and dynamic heat flow compensation commands respectively, and to arbitrate and correct conflicting or out-of-bounds commands based on the current reaction stage, historical control effects and predicted risks, and output the fused safety constraint control command set. The control command fusion execution unit (5) is used to parse and distribute the safety constraint control command set to the corresponding actuators, including the jacketed circulating pump frequency converter, multiple zone regulating valves, and auxiliary heating / cooling units, so as to achieve the composite control objective of macro temperature tracking and local temperature uniformity in a coordinated manner.

2. A reactor temperature control system according to claim 1, characterized in that, The local temperature field analysis and dynamic compensation unit (2) includes: The dynamic parameter identification module (21) is used to correct the reaction heat release kinetic parameters and the effective heat transfer coefficient of the fluid online based on the real-time acquired reaction heat release rate precursor signal and stirring power. The spatiotemporal fusion prediction module (22) is used to discretize the reactor space into multiple interconnected control volume units, establish the energy balance equation of each unit, and include the reaction exothermic term provided by the dynamic parameter identification module (21), as well as the heat transfer and jacket heat exchange term of the adjacent units, and predict the spatiotemporal temperature evolution of each unit in the future based on the current state. The non-uniformity diagnosis and compensation module (23) is used to calculate the current and predicted future non-uniformity index of the temperature field inside the vessel based on the output of the spatiotemporal fusion prediction module (22). When the index exceeds the adaptive threshold, the main non-uniform region is located, and the local compensation command required to offset the non-uniformity is calculated in combination with the predicted thermal state of the region.

3. A reactor temperature control system according to claim 2, characterized in that, In the spatiotemporal fusion prediction module (22), the energy balance equations for each control volume unit are: ,in, , , , Let represent the density, specific heat capacity, volume, and temperature of the i-th element, respectively. The reaction exothermic rate is based on the current concentration c and the parameter k corrected by the dynamic parameter identification module (21); For the heat transfer term with adjacent unit j; This refers to the heat exchange term of the jacket for this unit.

4. A reactor temperature control system according to claim 1, characterized in that, The adaptive security decision unit (4) includes: The layered safety boundary setting module (41) is used to set hard constraint boundaries based on process safety for macro-baseline control commands and soft constraint boundaries based on equipment capacity and thermal shock risk for dynamic heat flow compensation commands. The soft constraint boundaries are dynamically adjusted according to the current reaction intensity and temperature field stability. The instruction conflict arbitration module (42) is used to arbitrate according to a preset priority strategy when macro heating instructions and local cooling compensation instructions exist at the same time and may cancel out the control effect, or when the demand of multiple local compensation instructions exceeds the total execution capacity, and generate an executable instruction combination. The forward-looking risk suppression module (43) is used to identify potential future overheating or severe temperature gradient risk points based on the future temperature field distribution output by the spatiotemporal fusion prediction module (22), and apply damping or reverse correction to the control command of the current cycle in advance.

5. A reactor temperature control system according to claim 4, characterized in that, In the layered safety boundary setting module (41), the dynamic adjustment rule of the soft constraint boundary is as follows: when the reaction exothermic rate exceeds the set threshold or the temperature field non-uniformity index rises rapidly, the allowable adjustment range of the local compensation command is reduced; when the reaction enters the stable isothermal stage, the allowable adjustment range is relaxed to optimize uniformity.

6. A method for controlling the temperature of a reactor applied to any one of the systems of claims 1 to 5, characterized in that, Includes the following steps: S1: Real-time collection of multi-dimensional operating data of the reactor through the multi-source data fusion sensing unit (3); S2: Macro model prediction control unit (1) Based on the overall thermal balance model of the reactor and the set temperature trajectory, it solves the macro benchmark control command in a rolling manner; S3: Local temperature field analysis and dynamic compensation unit (2) Based on the running data of step S1, use the dynamic temperature field prediction model to make spatiotemporal predictions, diagnose temperature non-uniformity and generate dynamic heat flow compensation commands. S4: Adaptive safety decision unit (4) receives macro-based baseline control instructions and dynamic heat flow compensation instructions, performs safety boundary verification, conflict arbitration and risk suppression, and outputs the fused safety constraint control instruction set; S5: Control instruction fusion execution unit (5) executes the safety constraint control instruction set and drives the corresponding actuator to move.

7. A method for controlling the temperature of a reaction vessel according to claim 6, characterized in that, Step S3, the specific process of constructing and applying the dynamic temperature field prediction model includes: S31: Using current time and recent historical data, the reaction exothermic kinetic parameters and the equivalent heat transfer coefficient between units are identified online by recursive least squares method or Kalman filter algorithm. S32: Taking each discretized control volume unit as the object, substitute the parameters identified in step S31, solve the coupled energy balance equation set, and predict the distribution of the three-dimensional temperature field inside the vessel in the next few control cycles. S33: Based on the temperature field distribution predicted in step S32, calculate the global and zone temperature non-uniformity index. If the index exceeds the limit, locate the hot or cold spot area, and calculate the compensation command for the specific zone valve or auxiliary component according to the predicted temperature deviation and change trend of the area.

8. A method for controlling the temperature of a reaction vessel according to claim 6, characterized in that, In step S4, the specific rules for conflict arbitration by the instruction conflict arbitration module (42) include: if the macro instruction requires overall heating, while the local compensation instruction requires cooling of a certain area, the local cooling compensation instruction shall be executed first, and the macro heating instructions of other areas shall be reduced proportionally to maintain overall thermal balance; if the total demand of multiple local compensation instructions exceeds the total capacity of the actuator, the compensation demand of the area with the largest temperature deviation or the fastest rate of change shall be met first.

9. A method for controlling the temperature of a reaction vessel according to claim 6, characterized in that, After step S5 is executed, a feedback learning step S6 is also included: comparing the actual temperature field change with the temperature field predicted in step S32, calculating the prediction error; and using the prediction error, fine-tuning the parameters or compensation strategy model of the dynamic temperature field prediction model online through machine learning algorithms or model parameter adaptive laws.

10. A method for controlling the temperature of a reaction vessel according to claim 9, characterized in that, In the feedback learning step S6, a recursive learning algorithm with a forgetting factor is used to enable the model to gradually adapt to slow time-varying processes such as reactor scaling and catalyst activity changes, while not overly relying on recent accidental perturbations.

Images