A Smart Temperature Control Method for Fine Chemical Reactors Based on Adaptive Search

Abstract

This invention relates to the field of reactor temperature control technology, and discloses an intelligent temperature control method for fine chemical reactors based on adaptive search. The method includes: S1, data acquisition, collecting reactor operating data, including reactor internal temperature, jacket inlet and outlet temperatures, fluid flow rate, and valve opening signals; S2, heat balance calculation, establishing a heat balance equation based on the collected data, and calculating the reaction heat release and equivalent heat transfer coefficient; S3, model update; S4, predictive control solution; S5, adaptive search; S6, safety constraints; S7, stage identification; and S8, control execution. Through the linkage between the heat balance calculation module and the model update module, the reaction heat release and equivalent heat transfer coefficient are calculated in real time, and the recursive least squares algorithm is used to achieve online parameter identification, enabling the thermal model to be dynamically corrected according to changes in operating conditions. This ensures that the predicted model is consistent with the actual reaction process, significantly improving the accuracy of temperature prediction and the reliability of control response.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of reaction kettle temperature control, in particular to a fine chemical reaction kettle temperature intelligent control method based on adaptive search. BACKGROUND

[0002] The fine chemical reaction kettle is an important equipment for chemical reaction, heat exchange and mass transfer process, and is widely used in resin synthesis, pharmaceutical intermediates, fine chemical additives and other production links. The chemical reaction in the reaction kettle is mostly strong exothermic or endothermic reaction, and the temperature change directly affects the reaction rate and product quality. In order to ensure the stable operation of the reaction, a jacket or coil is usually arranged outside the reaction kettle for heating or cooling medium to adjust the reaction temperature. Based on this, the reaction kettle temperature control technology becomes the key link of fine chemical automation control, especially in the multi-stage, nonlinear and strong coupling chemical reaction process, higher requirements are put forward for the accuracy and stability of the control algorithm, and the temperature control of the fine chemical reaction kettle currently mostly adopts model predictive control or traditional PID control mode based on fixed parameters

[0003] However, in the current technology, the model parameters are not corrected in real time with the change of working conditions, which is easy to cause temperature lag, overshoot or fluctuation, and even causes the risk of heat runaway. SUMMARY

[0004] In view of the defects of the prior art, the present application provides a fine chemical reaction kettle temperature intelligent control method based on adaptive search, which solves the problem that the model parameters are not corrected in real time with the change of working conditions, resulting in temperature regulation lag.

[0005] In order to achieve the above purpose, the present application realizes the following technical scheme: a fine chemical reaction kettle temperature intelligent control method based on adaptive search, comprising:

[0006] S1, data acquisition, collecting reaction kettle operation data, including kettle temperature, jacket inlet and outlet temperature, fluid flow and valve opening signal;

[0007] S2, heat balance calculation, based on the collected data to establish heat balance equation and calculate reaction heat and equivalent heat transfer coefficient;

[0008] S3, model updating, using the calculated reaction heat and heat transfer coefficient to update the prediction model, and generating the predicted value of the future change of kettle temperature;

[0009] S4, prediction control solving, according to the prediction model output, executing model predictive control, and solving the control input candidate;

[0010] S5, adaptive search, introducing an adaptive search algorithm to the control instruction candidate, iteratively searching the prediction time domain length, control time domain length, weight coefficient and temperature setting trajectory parameters to generate a new parameter candidate;

[0011] S6, safety constraint, setting the temperature range, heating rate range and control instruction range as safety guardrails during the adaptive search process, and screening the new parameter candidate through the trust region to obtain a control instruction candidate that meets the constraint condition;

[0012] S7, phase identification, identifying the reaction process into heating stage, constant temperature stage and cooling stage, and switching the search variables and optimization objectives according to the stages;

[0013] S8, control execution, applying the control input screened by the constraint to the reaction kettle to complete the temperature control process.

[0014] Preferably, the heat balance calculation comprises:

[0015] According to the inlet and outlet temperatures and flow rates of the jacket, the heat exchange amount between the jacket and the reaction kettle is calculated;

[0016] The heat exchange amount and the kettle temperature change rate are substituted into the heat balance equation to obtain the reaction heat release amount, and the equivalent heat transfer coefficient is estimated by the recursive least squares method.

[0017] Preferably, the model updating comprises:

[0018] Based on the reaction heat release amount and the equivalent heat transfer coefficient, the heat transfer coefficient and the heat release term in the prediction model are corrected to obtain an updated prediction model;

[0019] The future change value of the reaction kettle temperature in the given prediction time domain is calculated by the updated prediction model.

[0020] Preferably, the prediction control solving comprises:

[0021] The updated prediction model is used as the control prediction model, with the reaction kettle temperature as the controlled variable and the jacket valve opening or the refrigerant inlet temperature as the control variable;

[0022] The upper and lower temperature constraints, the temperature heating rate constraints and the control variable value range constraints are set in the prediction time domain;

[0023] According to the constraint conditions, an optimization problem is constructed, and the control input candidate that meets the constraint conditions is obtained by solving in the control time domain.

[0024] Preferably, the adaptive search comprises:

[0025] An adaptive search algorithm is introduced to the control instruction candidate to determine the parameters to be optimized, including the prediction time domain length, the control time domain length, the weight coefficient and the temperature setting trajectory parameter;

[0026] A new parameter candidate is generated by using the adaptive search algorithm, and the new parameter candidate is input into the prediction control solving process;

[0027] The prediction control problem is solved again according to the new parameter candidate to obtain the corresponding control instruction candidate.

[0028] Preferably, the safety constraint includes:

[0029] The temperature range, the temperature change rate range and the control instruction range are set as the safety guardrail in the adaptive search process;

[0030] A trust domain is established on the basis of the safety guardrail to limit the value range of the parameter candidate;

[0031] The parameter candidates generated by the adaptive search are screened, the candidates exceeding the safety guardrail or the trust domain are eliminated, and the control instruction candidates meeting the conditions are retained.

[0032] Preferably, the stage identification includes:

[0033] The running data of the reaction process are obtained, and it is judged that the reaction is in the heating stage, the constant temperature stage or the cooling stage based on the running data;

[0034] When it is judged that the reaction is in the heating stage, the temperature setting trajectory parameter and the prediction time domain are selected as the search variables;

[0035] When it is judged that the reaction is in the constant temperature stage, the weight coefficient is selected as the search variable;

[0036] When it is judged that the reaction is in the cooling stage, the control time domain and the temperature setting trajectory parameter are selected as the search variables.

[0037] Preferably, the control execution includes:

[0038] The control instruction screened by the safety constraint is issued to the execution mechanism;

[0039] The execution mechanism includes a valve for adjusting the jacket fluid or a heat exchange device for adjusting the refrigerant temperature;

[0040] The control instruction is applied to the reaction kettle through the execution mechanism.

[0041] Preferably, the running data are monitored during the execution process;

[0042] When the operation data is monitored to be out of the preset range, the control parameter is switched to a safe parameter set, and a control instruction is generated based on the safe parameter set and acts on the reaction kettle.

[0043] An intelligent control system for fine chemical reaction kettle temperature based on adaptive search, the system comprises:

[0044] A data acquisition module acquires reaction kettle operation data, including in-kettle temperature, jacket inlet and outlet temperature, fluid flow and valve opening signal;

[0045] A heat balance calculation module, based on operation data, establishes a heat balance equation, and calculates reaction heat and equivalent heat transfer coefficient;

[0046] A model updating module, by reaction heat and equivalent heat transfer coefficient, parameter correction is carried out on the prediction model based on the heat balance equation, and the prediction value of the future change of the reaction kettle temperature is generated;

[0047] A predictive control solving module, according to the prediction result and the set temperature range, the temperature rising rate range and the valve opening range, an optimization problem is constructed, and the control instruction candidate is solved and transmitted to the adaptive search module;

[0048] An adaptive search module, the control instruction candidate is executed by adaptive search, the parameters to be optimized are determined, and the new parameter candidate is generated and fed back to the predictive control solving module for re-solution;

[0049] A safety constraint module, based on the generated new parameter candidate, a safety guardrail and a trust domain are established according to the temperature range, the temperature rising rate range and the control instruction range, the parameter candidate is screened, and the control instruction meeting the constraint condition is output;

[0050] A phase identification module, the operation data is monitored and the reaction process is determined to be in the temperature rising stage, the constant temperature stage or the temperature falling stage, and the phase determination result is transmitted to the adaptive search module to determine the corresponding search variable;

[0051] A control execution module, the control instruction is issued to the execution mechanism to adjust the jacket valve or the refrigerant temperature, so as to realize the temperature regulation of the reaction kettle;

[0052] An abnormality rollback module, in the control execution process, the operation data is monitored, when the operation data is out of the preset range, the control parameter is switched to the safe parameter set, a new control instruction is generated and transmitted to the control execution module.

[0053] The present application provides an intelligent control method for fine chemical reaction kettle temperature based on adaptive search. The method has the following advantages:

[0054] 1、The present application realizes real-time calculation of reaction heat release and equivalent heat transfer coefficient through the linkage of heat balance calculation module and model updating module, and realizes online parameter identification by using recursive least squares algorithm, so that the heat model can be dynamically corrected with the change of working conditions, ensuring that the prediction model is consistent with the actual reaction process, and significantly improving the temperature prediction accuracy and the reliability of control response.

[0055] 2、The present application introduces a parameter optimization algorithm with safety constraints in the adaptive search module, and uses the trust region to constrain the parameter search space. This mechanism realizes the balance of control performance optimization and process safety, prevents search from causing out-of-limit operation, and thus realizes sustainable and safe intelligent control.

[0056] 3、The present application realizes real-time judgment of the reaction process in the heating, constant temperature or cooling stage through the stage identification module, and dynamically adjusts the adaptive search variable and optimization target. This design realizes the self-matching of control strategy and process stage, makes the heating more stable, the constant temperature more accurate, and the cooling more controllable, and improves the adaptability and process adaptability of the reaction kettle temperature control as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 The flow chart of the present application is a fine chemical reaction kettle temperature intelligent control method based on adaptive search.

[0058] Figure 2 The architecture diagram of the present application is a fine chemical reaction kettle temperature intelligent control system based on adaptive search. DETAILED DESCRIPTION

[0059] The technical solutions of the present application will be described in detail below with reference to the drawings of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0060] Please refer to the accompanying Figure 1 The present application provides a fine chemical reaction kettle temperature intelligent control method based on adaptive search, which comprises:

[0061] S1, data acquisition, collecting reaction kettle operation data, including kettle temperature, jacket inlet and outlet temperature, fluid flow and valve opening signal;

[0062] Specifically, data acquisition is achieved by setting temperature, flow and valve opening and other conventional measuring devices on the reactor and its pipeline, each measuring signal is collected and transmitted to the subsequent calculation link in real time through the control system, providing necessary input information for heat balance calculation and predictive control, ensuring that the subsequent control steps are based on accurate and continuous operation state.

[0063] S2, heat balance calculation, based on the collected data to establish heat balance equation, and calculate the reaction heat and equivalent heat transfer coefficient;

[0064] Heat balance calculation includes:

[0065] According to the jacket inlet and outlet temperature and flow, the heat transfer between the jacket and the reactor is calculated;

[0066] The heat transfer and the rate of change of the kettle temperature are substituted into the heat balance equation to obtain the reaction heat, and the equivalent heat transfer coefficient is estimated by recursive least squares method.

[0067] Specifically, first, according to the jacket inlet and outlet temperature and flow information, the heat transfer between the jacket and the reactor is calculated, which is based on the principle of energy conservation, through the monitoring of the temperature and flow change of the jacket side, the instantaneous heat transfer intensity of the system is obtained, which is used to characterize the heat transfer level between the material and the jacket, providing the basis input for the subsequent reaction heat calculation, reflecting the heat transfer state of the system, so that the reactor heat behavior calculation is established on the basis of quantifiable heat transfer information, thereby supporting the dynamic solution of the heat balance model.

[0068] Then, the heat transfer and the rate of change of the kettle temperature are substituted into the heat balance equation to obtain the reaction heat, this step analyzes the change trend of the kettle temperature signal, relates the energy change to the reaction heat effect, realizes the real-time estimation of the reaction heat intensity, that is, the energy change is calculated by using the heat capacity parameter and temperature change rate of the material in the kettle, and the reaction heat is obtained by combining the heat transfer, to characterize the heat release dynamics of the reaction process, introduce the reaction heat information into the system heat analysis framework, form a real-time updated reaction heat load, provide dynamic heat source basis for the subsequent control algorithm.

[0069] Finally, the equivalent heat transfer coefficient is estimated online by recursive least squares method, that is, the heat transfer relationship is identified model: ;

[0070] It is the heat transfer obtained in step one (calculated from the jacket side measurement), The heat transfer area of the equipment (known value or nameplate parameter of the device), The temperature in the kettle (collected by the kettle temperature sensor), The average temperature of the jacket (calculated by the jacket inlet temperature And the outlet temperature , for example, take For the equivalent heat transfer coefficient to be estimated online, the above equation is written in the regression form of recursive least square (RLS) The RLS update is adopted: wherein is the covariance matrix, is a forgetting factor (a constant less than or equal to 1, used to weaken the influence of old data), so that the heat transfer coefficient is continuously updated with the change of working conditions and is sent into the model updating link in real time, maintaining the consistency of the heat model with the field state when the heat exchange condition changes with time, thereby supporting the stable operation of subsequent predictive control and adaptive search.

[0071] S3, model updating, using the calculated reaction heat release and heat transfer coefficient to update the prediction model, and generating the predicted value of the future change of the kettle temperature;

[0072] Further, the model updating includes:

[0073] Based on the reaction heat release and the equivalent heat transfer coefficient, the heat transfer coefficient and the heat release term in the prediction model are corrected to obtain the updated prediction model;

[0074] The future change value of the reaction kettle temperature is calculated in the given prediction time domain through the updated prediction model.

[0075] Specifically, in the implementation of the model updating, first, based on the reaction heat release and the equivalent heat transfer coefficient obtained in the previous stage, the heat release term and the heat transfer parameter in the prediction model are corrected, and a dynamic energy equation is established according to the heat balance relationship of the reaction kettle:

[0076] is the material density, is the specific heat capacity of the material at constant pressure, is the reaction volume, is the kettle temperature, is the average temperature of the jacket, is the heat exchange area, is the reaction heat release calculated by heat balance, is the equivalent heat transfer coefficient estimated in real time by the recursive least square method, and is obtained by and ​​​​​​​​​The latest estimated value of the model replaces the corresponding fixed parameter in the model, so that the prediction model automatically updates the thermal dynamic and heat transfer characteristics in each control period, thereby realizing model adaptive correction in synchronization with the actual working condition to maintain the accurate expression of the prediction model to the thermal state of the system and ensure that the subsequent prediction calculation and control solution are based on real-time effective thermal characteristic parameters;

[0077] After completing the parameter correction, the updated prediction model is used to calculate the future change value of the reactor temperature in a given prediction time domain, i.e., the current reactor temperature and the jacket temperature are taken as initial conditions, the updated model is called for multi-step prediction, the discrete integral or numerical iteration method is used to generate the temperature prediction sequence, and the future temperature change trend is output.

[0078] S4, prediction control solving, according to the output of the prediction model, performing model predictive control, and solving to obtain a control input candidate;

[0079] Further, the prediction control solving includes:

[0080] The updated prediction model is taken as a control prediction model, and the reactor temperature is taken as a controlled variable, and the jacket valve opening or the refrigerant inlet temperature is taken as a control variable;

[0081] Upper and lower limit constraints of the temperature, temperature rising rate constraints, and control variable value range constraints are set in the prediction time domain;

[0082] An optimization problem is constructed according to the constraint conditions, and a control input candidate satisfying the constraint conditions is solved in the control time domain.

[0083] Specifically, first, the updated prediction model in S3 is taken as a control prediction model, in which the reactor temperature is taken as a controlled variable, and the jacket valve opening or the refrigerant inlet temperature is taken as a control variable, a prediction equation is established through the updated energy balance model, the control calculation is based on the latest heat conduction and heat release characteristics for rolling optimization, and the calculation process of the prediction control is consistent with the working condition change by updating the model parameters in each sampling period;

[0084] Subsequently, constraint conditions are set in the prediction time domain, including upper and lower limit constraints of the reactor temperature, temperature rising rate constraints, and value range constraints of the control variable (valve opening or refrigerant temperature), i.e., boundary conditions are introduced in the control optimization process to ensure that the control solution does not produce instructions that exceed the safety or equipment allowable range, and the constraint conditions are generally embedded in the optimization problem in the form of linear inequalities, which are commonly represented as

[0085]

[0086] wherein, T is the reactor temperature at the prediction time, is the temperature change amount, To control the variable (valve opening or refrigerant temperature), the subscripts min and max represent the allowable range, so that the safety and equipment constraints are explicitly introduced into the control optimization structure, preventing the control result from exceeding the limit and ensuring the safe and stable operation of the process.

[0087] Finally, the optimization problem of model predictive control is constructed according to the constraint conditions, and the control input candidate that satisfies the constraint conditions is obtained by solving the control time domain. The optimization objective is to minimize the weighted sum of the temperature deviation and the control amount change, and the optimization objective function is expressed as: ;

[0088] Wherein, is the performance index, is the prediction horizon length, is the control horizon length, and is the weight coefficient, is the temperature set value, is the control amount increment. By solving the optimization problem, the optimal control input sequence of the current control period can be obtained, and the first control instruction is extracted as the execution value to realize temperature optimization control under the conditions of multiple constraints and multiple objectives. On the premise of ensuring safety constraints, temperature fluctuations are reduced and system control accuracy is improved.

[0089] S5, adaptive search, introducing an adaptive search algorithm to the control instruction candidate, iteratively searching for the prediction horizon length, the control horizon length, the weight coefficient, and the temperature setting trajectory parameters to generate new parameter candidates;

[0090] Further, the adaptive search includes:

[0091] An adaptive search algorithm is introduced to the control instruction candidate to determine the parameters to be optimized, including the prediction horizon length, the control horizon length, the weight coefficient, and the temperature setting trajectory parameters;

[0092] A new parameter candidate is generated using the adaptive search algorithm, and the new parameter candidate is input into the predictive control solving process;

[0093] The predictive control problem is re-solved according to the new parameter candidate to obtain the corresponding control instruction candidate.

[0094] Specifically, first, an adaptive search algorithm is introduced to the control instruction candidate obtained by solving the predictive control to determine the control parameters to be optimized. The prediction horizon length , the control horizon length , the weight coefficient , , and the temperature setting trajectory parameters are taken as optimization variables, and the control performance index function is taken as the target:

[0095] ;

[0096] wherein, T is the reactor temperature at the prediction time, T is the temperature setpoint, is the control variable change, the algorithm adjusts the value of to minimize the objective function , so as to obtain the optimal parameter combination of the control performance, and realize the self-learning and dynamic setting of the parameters with the change of the working condition.

[0097] Subsequently, a new parameter candidate is generated according to the adaptive search algorithm, and the algorithm evaluates the performance of the historical search results, predicts the improvement potential of the current parameter combination, and selects a new search point to improve the overall performance.

[0098] Finally, the new parameter candidate is input into the predictive control solving process, the control input is recalculated, and a new control instruction candidate is generated, which can be executed in a loop during the running process, realizing the closed-loop iteration of search, optimization and control, so that the control system can continuously maintain the optimal regulation performance under different reaction stages and working conditions.

[0099] S6, safety constraints, set the temperature range, temperature rise rate range and control instruction range as safety fences during the adaptive search process, and screen the new parameter candidate through the trust domain to obtain the control instruction candidate meeting the constraint condition;

[0100] Further, the safety constraints include:

[0101] The temperature range, temperature change rate range and control instruction range are set as safety fences during the adaptive search process;

[0102] The trust domain is established on the basis of the safety fence to limit the value range of the parameter candidate;

[0103] The parameter candidate generated by the adaptive search is screened, the candidate exceeding the safety fence or the trust domain is removed, and the control instruction candidate meeting the condition is retained.

[0104] Specifically, the safety constraints of S6 set safety fences and trust domains for the adaptive search process itself during implementation, which are used to prevent the parameter search from causing the control instruction to exceed the safety range. The working principle is to limit the allowed interval of temperature, temperature rise rate and control instruction during the search stage, and use these boundaries as the algorithm screening conditions, rather than only as the constraints of control optimization. After the adaptive search generates a parameter candidate, it is first determined whether the response of the parameter combination under the model prediction is located within the set safety fence, and if it meets the requirement, it enters the next round of optimization, otherwise it is removed;

[0105] On this basis, the parameter update range is further limited by a trust region method. The trust region takes the current optimal parameter as the center and defines an acceptable search radius by the condition: ;

[0106] wherein is the trust region radius, used to constrain the search step to prevent rapid parameter changes that cause control fluctuations. The model prediction is combined to safely screen the candidate parameters, and adaptive search is realized to iteratively optimize under the premise of process safety.

[0107] Thus, the safety constraint is extended from the control layer to the algorithm layer, so that parameter optimization is performed in the controlled space, preventing search algorithms from causing temperature out-of-bounds due to misjudgment or excessive adjustment, thereby realizing a safe and adaptive parameter optimization process.

[0108] S7, stage identification, identifying the stages of the reaction process, including the warming-up stage, the constant-temperature stage, and the cooling-down stage, and switching the search variables and optimization objectives according to the stages;

[0109] Further, the stage identification includes:

[0110] obtaining the running data of the reaction process, and determining whether the reaction is in the warming-up stage, the constant-temperature stage, or the cooling-down stage based on the running data;

[0111] when it is determined that the reaction is in the warming-up stage, selecting the temperature setting trajectory parameters and the prediction time domain as the search variables;

[0112] when it is determined that the reaction is in the constant-temperature stage, selecting the weight coefficient as the search variable;

[0113] when it is determined that the reaction is in the cooling-down stage, selecting the control time domain and the temperature setting trajectory parameters as the search variables.

[0114] Specifically, the running data of the reaction process is obtained and the trend of the kettle temperature is analyzed to determine whether the reaction is in the warming-up stage, the constant-temperature stage, or the cooling-down stage, wherein a continuous temperature rise is determined as the warming-up stage, a small temperature fluctuation is determined as the constant-temperature stage, and a continuous temperature drop is determined as the cooling-down stage; the system automatically switches the variables and optimization objectives of the adaptive search according to the different stages, optimizes the temperature setting trajectory parameters and the prediction time domain in the warming-up stage, adjusts the weight coefficient in the constant-temperature stage, and optimizes the control time domain and the temperature trajectory parameters in the cooling-down stage. The role of this step is to match the control strategy with the reaction stage, and the effect is to realize adaptive adjustment of temperature control in different stages, taking into account the smooth warming-up, accurate constant-temperature, and controllable cooling-down.

[0115] S8, control execution, applying the control input screened by the constraint to the reaction kettle to complete the temperature control process.

[0116] Further, the control execution includes:

[0117] The control instruction screened by the safety constraint is issued to the actuator;

[0118] The actuator includes a valve for adjusting the jacket fluid or a heat exchange device for adjusting the coolant temperature;

[0119] The control instruction is applied to the reactor by the actuator.

[0120] Specifically, the control instruction screened by the safety constraint is issued to the actuator, and the actual adjustment of the reactor temperature is completed by the actuator. The control result obtained by the comprehensive optimization of model predictive control and adaptive search is converted into a physical action signal. The valve controls the jacket fluid flow or the heat exchange device adjusts the coolant temperature to realize direct control of the heat exchange process in the reactor. The control instruction is constantly updated according to real-time feedback to form a closed-loop regulation, so as to keep the reaction temperature stable within the set range. Thus, the calculation optimization result is converted into a specific execution action to realize the final implementation of temperature regulation, ensure that the calculation result of the control strategy accurately reflects the actual process, and keep the reactor temperature stable and controllable to meet the process requirements.

[0121] Further, the running data is monitored during the execution;

[0122] When it is monitored that the running data exceeds the preset range, the control parameters are switched to the safety parameter set, and the control instruction is generated based on the safety parameter set and applied to the reactor.

[0123] Specifically, the running data of the reactor is monitored in real time to determine whether the system is in a safe range. By collecting signals such as the temperature in the reactor, the jacket temperature, the valve opening, and the fluid flow, and comparing them with the preset upper and lower limits, when any signal exceeds the allowed range, it is determined as an abnormal state. The system triggers the safety switching logic immediately after detecting the abnormality, and switches the control parameters to the safety parameter set set in advance. This parameter set usually includes conservative prediction time domain, control time domain, and control weight value, which is used to reduce the system response strength and prevent the temperature from deviating from the target interval. Then, the control instruction is regenerated based on the safety parameter set and applied to the reactor, and the actuator enters the safe operation mode, so as to ensure that the control system can still maintain stability in the case of sudden disturbance or abnormal working condition. The effect is to prevent temperature overshoot, loss of control, or equipment overrunning.

[0124] Please refer to the attached Figure 2 An intelligent control system for fine chemical reactor temperature based on adaptive search, the system comprises:

[0125] A data acquisition module acquires the running data of the reactor, including the temperature in the reactor, the inlet and outlet temperatures of the jacket, the fluid flow, and the valve opening signal;

[0126] a heat balance calculation module, based on the operation data, establishes a heat balance equation, calculates the reaction heat release and the equivalent heat transfer coefficient;

[0127] a model updating module, through the reaction heat release and the equivalent heat transfer coefficient, parameter correction is performed on the prediction model based on the heat balance equation, and a prediction value of future change of the temperature of the reaction kettle is generated;

[0128] a predictive control solving module, according to the prediction result and the set temperature range, the temperature rise rate range and the valve opening range, an optimization problem is constructed, a control instruction candidate is solved and is transmitted to the adaptive search module;

[0129] an adaptive search module, adaptive search is performed on the control instruction candidate, parameters to be optimized are determined, new parameter candidates are generated and are fed back to the predictive control solving module for re-solution;

[0130] a safety constraint module, based on the generated new parameter candidate, a safety guardrail and a trust domain are established according to the temperature range, the temperature rise rate range and the control instruction range, the parameter candidate is screened, and a control instruction meeting the constraint condition is output;

[0131] a stage identification module, monitors the operation data and judges whether the reaction process is in a temperature rise stage, a constant temperature stage or a temperature drop stage, and transmits the stage determination result to the adaptive search module to determine corresponding search variables;

[0132] a control execution module, the control instruction is transmitted to an execution mechanism to adjust a jacket valve or a refrigerant temperature, so as to realize temperature regulation of the reaction kettle;

[0133] an abnormality rollback module, monitors the operation data during the control execution process, when the operation data exceeds a preset range, the control parameters are switched to a safety parameter set, a new control instruction is generated and is transmitted to the control execution module.

[0134] Specifically, the fine chemical reaction kettle temperature intelligent control system based on adaptive search acquires the reaction kettle operation information in real time through the data acquisition module, establishes the heat balance equation and calculates the reaction heat and the equivalent heat transfer coefficient through the heat balance calculation module, the model updating module corrects the prediction model according to the parameters and generates the future temperature change trend, the predictive control solving module calculates the control instruction candidate combined with the constraint condition, the adaptive search module iteratively optimizes the control parameters in the parameter space, the safety constraint module screens the control instruction meeting the safety condition through the safety guardrail and the trust domain, the stage identification module judges the reaction stage according to the operation data and adjusts the search variable, the control execution module issues the instruction to the execution mechanism to adjust the jacket or the refrigerant to realize the temperature control, the abnormal rollback module switches to the safety parameter set to regenerate the control instruction when the operation data is monitored to be out of limit, realizes the intelligent, adaptive and safety control of the reaction kettle temperature, and the effect is to ensure the stability, precision and safety of the temperature regulation under complex working conditions.

[0135] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. An intelligent control method for fine chemical reactor temperature based on adaptive search, characterized in that, Comprise: S1, data acquisition, collect the reactor running data, including the temperature in the kettle, jacket inlet and outlet temperature, fluid flow and valve opening signal; S2, heat balance calculation, based on the collected data to establish heat balance equation, and calculate the reaction heat and equivalent heat transfer coefficient; S3, model update, the calculated reaction heat and heat transfer coefficient for updating the prediction model, and generate the prediction value of the future change of kettle temperature; S4, solving of predictive control, according to the prediction model output, execute model predictive control, solve to get control input candidate; S5, adaptive search, introduce adaptive search algorithm to control instruction candidate, iterative search for prediction time domain length, control time domain length, weight coefficient and temperature setting trajectory parameters, generate new parameter candidate; The adaptive search comprises: introducing adaptive search algorithm to control instruction candidate, determining the parameters to be optimized, the parameters including prediction time domain length, control time domain length, weight coefficient and temperature setting trajectory parameters; Using adaptive search algorithm to generate new parameter candidate, and inputting new parameter candidate into predictive control solving process; According to the new parameter candidate, the predictive control problem is solved again to obtain the corresponding control instruction candidate; S6, safety constraint, set temperature range, temperature change rate range and control instruction range as safety fence in adaptive search process, and screen the new parameter candidate through trust region to obtain control instruction candidate meeting the constraint condition; The safety constraint comprises: setting temperature range, temperature change rate range and control instruction range as safety fence in adaptive search process; Establishing trust region based on safety fence to limit the value range of parameter candidate; Screening the parameter candidate generated by adaptive search, eliminating the candidate beyond safety fence or trust region, and retaining the control instruction candidate meeting the condition; S7, stage identification, stage identification is carried out for the reaction process, including temperature rising stage, constant temperature stage and temperature falling stage, and searching variable and optimization target are changed according to stage; S8, control execution, applying the control input screened by constraint to the reactor to complete the temperature control process. 2.The fine intelligent control method for temperature of chemical reaction kettle based on adaptive search according to claim 1, characterized in that, The heat balance calculation comprises: According to the jacket inlet and outlet temperature and flow, the heat transfer amount between the jacket and the reactor is calculated; The heat transfer amount and the kettle temperature change rate are substituted into the heat balance equation to obtain the reaction heat, and the equivalent heat transfer coefficient is estimated by recursive least squares method. 3.The fine intelligent control method for temperature of chemical reaction kettle based on adaptive search according to claim 1, characterized in that, The model update comprises: Based on the reaction heat and the equivalent heat transfer coefficient, the heat transfer coefficient and the heat release term in the prediction model are corrected to obtain the updated prediction model; The future change value of the reactor temperature is calculated in the given prediction time domain through the updated prediction model. 4.The fine intelligent control method for temperature of chemical reaction kettle based on adaptive search according to claim 1, characterized in that, The predictive control solving comprises: Taking the updated prediction model as the control prediction model, taking the reactor temperature as the controlled variable and the jacket valve opening or the refrigerant inlet temperature as the control variable; Setting temperature upper and lower limit constraint, temperature rising rate constraint and control variable value range constraint in the prediction time domain; According to the constraint condition, the optimization problem is constructed, and the control input candidate meeting the constraint condition is solved in the control time domain.

5. The fine chemical reactor temperature intelligent control method based on adaptive search according to claim 1, characterized in that, The stage identification comprises: Obtaining operation data of the reaction process, and judging that the reaction is in a warming-up stage, a constant-temperature stage or a cooling-down stage based on the operation data; When it is judged that the reaction is in the warming-up stage, selecting a temperature setting trajectory parameter and a prediction time domain as search variables; When it is judged that the reaction is in the constant-temperature stage, selecting a weight coefficient as a search variable; When it is judged that the reaction is in the cooling-down stage, selecting a control time domain and a temperature setting trajectory parameter as search variables. 6.The fine intelligent control method for temperature of chemical reaction kettle based on adaptive search according to claim 1, characterized in that, The control execution comprises: issuing the control instruction screened based on the safety constraint to an execution mechanism; the execution mechanism comprises a valve for adjusting jacket fluid or a heat exchange device for adjusting refrigerant temperature; the control instruction is applied to the reaction kettle through the execution mechanism. 7.The fine intelligent control method for temperature of chemical reaction kettle based on adaptive search according to claim 1, characterized in that, In the control execution, the operation data is monitored; when the operation data is monitored to be out of a preset range, the control parameter is switched to a safety parameter set, and a control instruction is generated based on the safety parameter set and applied to the reaction kettle.

8. An intelligent control system for fine chemical reactor temperature based on adaptive search, characterized in that, A fine chemical reaction kettle temperature intelligent control method based on adaptive search for any one of claims 1-7, the system comprises: a data acquisition module, which acquires reaction kettle operation data, including kettle temperature, jacket inlet and outlet temperature, fluid flow and valve opening signal; a heat balance calculation module, which establishes a heat balance equation based on the operation data and calculates reaction heat and equivalent heat transfer coefficient; a model updating module, which corrects the prediction model based on the heat balance equation by the reaction heat and the equivalent heat transfer coefficient to generate a prediction value of future change of the reaction kettle temperature; a predictive control solving module, which constructs an optimization problem according to the prediction result and the set temperature range, warming-up rate range and valve opening range, and solves to obtain a control instruction candidate and transfer it to the adaptive search module; an adaptive search module, which performs adaptive search on the control instruction candidate, determines the parameters to be optimized, generates a new parameter candidate and feeds it back to the predictive control solving module for re-solution; a safety constraint module, which establishes a safety guardrail and a trust domain based on the generated new parameter candidate according to the temperature range, the warming-up rate range and the control instruction range, screens the parameter candidate, and outputs the control instruction meeting the constraint condition; a stage identification module, which monitors the operation data and judges that the reaction process is in a warming-up stage, a constant-temperature stage or a cooling-down stage, and transfers the stage determination result to the adaptive search module to determine the corresponding search variable; a control execution module, which issues the control instruction to the execution mechanism to adjust the jacket valve or the refrigerant temperature, so as to realize temperature adjustment of the reaction kettle; an abnormality rollback module, which monitors the operation data during the control execution, switches the control parameter to the safety parameter set when the operation data is out of the preset range, generates a new control instruction and transfers it to the control execution module.

Images