A method and system for optimizing control parameters of a multi-stage reaction kettle

Abstract

The present application belongs to the field of automation control technology, and particularly relates to a control parameter optimization method and system for a multi-stage reaction kettle, which comprises the following steps: collecting key operation data of a series multi-stage reaction kettle at each moment in real time, constructing an initial temperature sequence and an initial valve opening sequence after standardization and adaptive filtering denoising; calculating temperature fluctuation standard deviation and other indexes based on the initial temperature sequence, obtaining temperature deviation degree through normalized weighted fusion, and constructing a reaction nonlinearity strength index by combining the measured temperature and other parameters according to the Arrhenius law; coupling the initial valve opening sequence change rate based on the index, and obtaining a model vector adaptive correction coefficient by combining a smoothing factor and other parameters; correcting the basic step response vector of dynamic matrix control by using the coefficient, solving the optimal valve opening adjustment amount and calculating the execution value, and outputting to the actuator to complete the regulation and control, so as to realize the whole cycle optimization of the control parameters. The present application improves the control accuracy of the multi-stage reaction kettle.

Description

Technical Field

[0001] This invention relates to the field of automation control technology, and in particular to a method and system for optimizing control parameters of a multi-stage reactor. Background Technology

[0002] In the fine chemical and pharmaceutical manufacturing industries, multistage reactors are widely used in continuous production processes. Through multistage series connection, the materials gradually complete the predetermined chemical reaction during the flow process. In order to ensure product quality, improve product conversion rate and purity, and ensure production safety, it is necessary to strictly control the key process parameters such as temperature and pressure of each stage reactor.

[0003] In the prior art, the control of multi-stage reactors usually adopts the conventional dynamic matrix control algorithm. This algorithm first obtains the process model under a single operating condition through step response testing, constructs a predictive control matrix based on the process model, and then solves the optimal control quantity based on the deviation between the actual value and the set value of the real-time collected process parameters and the predictive control matrix. Finally, the optimal control quantity is output to the actuator to realize the adjustment of process parameters.

[0004] However, conventional dynamic matrix control algorithms have significant limitations in actual production conditions of multi-stage reactors. Conventional dynamic matrix control algorithms are usually designed based on a fixed linear step response model obtained under a single operating condition. However, actual chemical reaction processes have strong nonlinear and phased characteristics. When the reaction process in a multi-stage reactor transitions from the heating initiation phase to the intense exothermic phase, the rate of heat release inside the reactor changes exponentially. This leads to significant changes in the temperature change and temperature response time constant caused by changes in the opening of the temperature control valve. If the fixed linear step response model established based on the steady-state reaction phase is still used for predictive control at this time, it will cause a serious deviation between the model's predicted output and the actual output. This deviation will cause critical control problems in the intense reaction phase. Excessive control will lead to temperature overshoot, while insufficient control will lead to insufficient reaction conversion. Both of these will seriously affect product quality and bring production safety hazards. Summary of the Invention

[0005] To address the technical problem that conventional dynamic matrix control algorithms, when adapted to multi-stage reactor production, rely on fixed linear step response models, which cannot match the highly nonlinear and phased process characteristics of chemical reactions, resulting in a significant deviation between the model's predicted output and the actual output, this invention provides a method for optimizing control parameters of multi-stage reactors.

[0006] In a first aspect, the present invention provides a method for optimizing control parameters of a multi-stage reactor, comprising: real-time acquisition of key operating data of a series-connected multi-stage reactor at various times; performing standardization and adaptive filtering denoising processing on the key operating data in sequence; constructing an initial temperature sequence and an initial valve opening sequence based on the filtered output values; calculating the standard deviation of temperature fluctuation, the average gradient of temperature change, and the extreme temperature deviation based on the initial temperature sequence of each stage reactor at each time; obtaining the temperature deviation degree of each stage reactor at each time after normalization and weighted fusion, by combining the Arrhenius law and integrating the temperature deviation degree with the measured temperature, reference temperature, and characteristic temperature constant of the reaction system; constructing a reaction nonlinearity intensity index for each stage reactor at each time based on the reaction nonlinearity intensity index coupled with the rate of change of the initial valve opening sequence, and combining the smoothing factor, sign function, and maximum correction amplitude adjustment weight to obtain the adaptive correction coefficient of the model vector for each stage reactor at each time; correcting the basic step response vector of the dynamic matrix control using the adaptive correction coefficient of the model vector, solving for the optimal valve opening adjustment amount and calculating the valve opening execution value, outputting it to the actuator to complete temperature control, thereby realizing full-cycle rolling optimization of the control parameters of the multi-stage reactor.

[0007] This invention deeply integrates the characteristics of chemical reaction processes with dynamic matrix control algorithms. First, it performs noise reduction and feature extraction on the operating data of multi-stage reactors to construct an intensity index that accurately characterizes the degree of reaction nonlinearity. Then, it couples the dynamic feature generation model of valve opening control with adaptive correction coefficients, achieving real-time condition-based correction of the basic model of dynamic matrix control. This breaks the limitations of conventional dynamic matrix control relying on fixed linear models, ensuring that the optimization of control parameters always matches the strong nonlinearity and phased dynamic changes of the multi-stage reactor process. Simultaneously, through a closed-loop rolling optimization process encompassing data acquisition, feature calculation, model correction, parameter optimization, and control execution, the valve opening control parameters can be adaptively adjusted in real time according to operating conditions, accurately offsetting temperature deviations within the reactor and effectively avoiding problems such as temperature overshoot and insufficient reaction conversion. This improves the accuracy and stability of temperature control in multi-stage reactors, ensures consistent product quality and production process safety in continuous production, and achieves intelligent and adaptive optimization of control parameters for multi-stage reactors.

[0008] Preferably, the adaptive filtering denoising method includes: In the formula, For the first The stage reaction vessel at time The Filtered output values ​​of key operational data; Indicates the first The stage reaction vessel at time The Standardized observation values ​​of key operational data; Indicates the first The stage reaction vessel at time The Filtered output values ​​of key operational data; Indicates time The gain coefficient of the adaptive filter; This is the stage index for a multi-stage reactor. Sampling time, An index for critical runtime data.

[0009] This method constructs an adaptive filtering relationship with dynamic gain coefficients. It uses standardized observation values ​​of key operating data and the filtered output value of the previous moment as the core calculation basis, and dynamically adjusts the value of the filtering gain coefficient to accurately suppress various noises such as environmental interference and equipment sensing in the collected data. At the same time, it preserves the characteristics of process parameter mutations caused by sudden changes in raw material composition and catalyst activity to the greatest extent, effectively avoiding noise-induced distortion of control model input and preventing over-filtering from masking the changes in real process parameters. This ensures the accuracy of subsequent calculations of features such as temperature deviation and reaction nonlinearity intensity, laying a reliable data foundation for the precise optimization of control parameters of multi-stage reactors.

[0010] Preferably, the method for obtaining the gain coefficient of the adaptive filter includes: In the formula, Indicates time The gain coefficient of the adaptive filter; Indicates the first The stage reaction vessel at time The Standardized observation values ​​of key operational data; Indicates the first The stage reaction vessel at time The Filtered output values ​​of key operational data; This represents the preset error sensitivity threshold parameter; Based on the natural constant An exponential function with base 1; This is the stage index for a multi-stage reactor. The sampling time.

[0011] Preferably, the method for obtaining the nonlinear intensity index of the response includes: In the formula, For the first The stage reaction vessel at time The nonlinear intensity index of the response; For the first The stage reaction vessel at time The degree of temperature deviation; For the first The stage reaction vessel at time The measured temperature; This is the reference temperature under the nominal operating conditions of a multi-stage reactor. The characteristic temperature constant of the reaction system; Based on the natural constant An exponential function with base 1; This is the stage index for a multi-stage reactor. The sampling time.

[0012] This method constructs a calculation formula for the reaction nonlinearity intensity index by combining the Arrhenius law, deeply integrating the degree of temperature deviation at the data level with the temperature change characteristics at the physical level of chemical reactions. It utilizes an exponential function to characterize the exponential growth law of the reaction rate with temperature, enabling the constructed reaction nonlinearity intensity index to accurately quantify the degree to which the reaction process in the reactor deviates from the linear nominal model at different times and stages. This achieves a precise mapping from a simple representation of temperature data fluctuations to the physical essence of reaction nonlinearity. This index not only reflects the impact of temperature deviation on reaction nonlinearity but also closely matches the actual process characteristics of chemical reactions, providing a precise process physics basis for the subsequent generation of adaptive correction coefficients for model vectors. This ensures that model correction is highly matched with the actual operating conditions of the reactor, fundamentally solving the shortcomings of conventional control methods that rely solely on data-level regulation and ignore the physical characteristics of the reaction.

[0013] Preferably, the method for obtaining the rate of change of the initial valve opening sequence includes: In the formula, It is the first The stage reaction vessel at time The rate of change of the initial valve opening sequence; , The first The stage reaction vessel at time The filtered output values ​​for the heating valve opening and cooling valve opening, respectively. , They are time points The filtered output values ​​for the opening degree of the heating valve and the opening degree of the cooling valve; It is the rate of change of the opening degree of the heating valve. These are its weighting coefficients; It is the rate of change of the cooling valve opening. These are its weighting coefficients; The sampling time interval; This is the stage index for a multi-stage reactor. The sampling time.

[0014] Preferably, obtaining the adaptive correction coefficients of the model vector for each stage of the reactor at each time step includes: In the formula, For the first The stage reaction vessel at time The adaptive correction coefficients for the model vector; For the first The stage reaction vessel at time The nonlinear intensity index of the response; For the first The stage reaction vessel at time The rate of change of the initial valve opening sequence; This is the preset smoothing factor; It is a symbolic function; Adjust the weights to the preset maximum correction range; This is the stage index for a multi-stage reactor. The sampling time.

[0015] This method uses an adaptive correction coefficient calculation formula that integrates process and control characteristics into a model vector. It couples the nonlinear intensity index of the reaction with the rate of change of the initial valve opening sequence. Simultaneously, it introduces a smoothing factor, a sign function, and a maximum correction amplitude adjustment weight to achieve dual constraints and precise control of the correction coefficient. The smoothing factor effectively avoids the problem of zero denominator and adjusts the coefficient's response sensitivity to changes in operating conditions. The sign function accurately judges the trend of system heat gain changes caused by exothermic or endothermic reactions based on the degree of temperature deviation. The maximum correction amplitude adjustment weight limits the coefficient fluctuation range, ensuring that the generated correction coefficient fluctuates smoothly and within bounded ranges around 1. This approach allows the correction coefficient to accurately match the nonlinear process characteristics of the reactor and adapt to the real-time dynamic control of valve openings, achieving condition-based adaptive correction of the dynamic matrix control model vector. This breaks the adaptation limitations of conventional fixed models, providing a model foundation that fits actual operating conditions for subsequent precise optimization of control parameters, ensuring the rationality of model correction and the stability of the control system.

[0016] Preferably, the method for correcting the basic step response vector using the adaptive correction coefficient of the model vector includes: scaling and correcting the basic step response vector of the dynamic matrix control using the adaptive correction coefficient of the model vector of each stage reactor at the corresponding time, to obtain the real-time step response vector of each stage reactor at the corresponding time; the basic step response vector is a fixed value calibrated by step response test under the stable operating condition of the multi-stage reactor.

[0017] Preferably, the step of solving for the optimal valve opening adjustment and calculating the valve opening execution value includes: combining the optimization coefficient, the real-time step response vector, and the deviation between the reference temperature and the measured temperature to solve for the optimal valve opening adjustment for each stage of the reactor at the corresponding time; then superimposing the optimal valve opening adjustment to the current valve opening of each stage of the reactor at the corresponding time; and obtaining the final valve opening execution value for the heating valve opening and the cooling valve opening, respectively; wherein the optimization coefficient is a constant calibrated according to the process characteristics of the multi-stage reactor and the control capability of the actuator; and the current valve opening is the filtered output value of the initial valve opening sequence.

[0018] Preferably, the full-cycle rolling optimization of the control parameters of the multi-stage reactor includes: re-collecting key operating data of the multi-stage reactor at the next sampling time, and sequentially performing all operations such as standardization, adaptive filtering and denoising, feature index calculation, model vector adaptive correction coefficient generation, control parameter optimization, and regulation execution; updating the control parameters in real time according to the new operating data, so that the control parameters always adapt to the dynamic changes in the reaction conditions of the multi-stage reactor, thereby realizing the full-cycle adaptive rolling optimization of the control parameters.

[0019] This method constructs a closed-loop optimization mechanism that iterates throughout the entire sampling process. Using the temperature control at the previous sampling moment as a node, it re-collects operating data at the next sampling moment and fully executes the entire process of data processing, feature calculation, coefficient generation, parameter optimization, and control execution. This allows the control parameters to be updated and iteratively optimized in real time as the reaction conditions of the multi-stage reactors change dynamically, ensuring that the control parameters are always precisely matched with the current nonlinear reaction conditions. This avoids the drawbacks of fixed control parameters being unable to adapt to dynamic changes in operating conditions. Through this full-cycle rolling optimization approach, a continuous closed loop of data acquisition, parameter optimization, control execution, and data re-acquisition is formed, enabling adaptive adjustment of the control parameters of the multi-stage reactors throughout the entire production cycle. This ensures that the accuracy and stability of temperature control are always online, solving the technical problem of mismatch between conventional fixed model control and dynamic nonlinear conditions from the perspective of the entire process cycle. This improves the process stability and product quality consistency of continuous production in multi-stage reactors.

[0020] Secondly, the present invention provides a control parameter optimization system for a multi-stage reactor, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned control parameter optimization method for a multi-stage reactor is implemented.

[0021] By adopting the above technical solution, the above-mentioned method for optimizing the control parameters of a multi-stage reactor is generated into a computer program and stored in a memory for loading and execution by a processor. This allows for the creation of a terminal device based on the memory and processor, facilitating its use.

[0022] The beneficial effects of this invention are as follows: It effectively solves the technical problem that conventional dynamic matrix control algorithms rely on fixed linear step response models and cannot match the strong nonlinearity and staged process characteristics of multi-stage reactor chemical reactions. Adaptive filtering and noise reduction ensure the validity of the operating data. Combined with the reaction nonlinearity intensity index constructed using Arrhenius's law, it achieves precise quantification of the reaction nonlinearity. Furthermore, by coupling the dynamically generated model vector adaptive correction coefficients for valve opening control, it can perform real-time condition-based correction of the dynamic matrix control model, ensuring that the optimal valve opening control parameters always adapt to the dynamic operating conditions of the reactor. Simultaneously, through full-cycle rolling optimization to form closed-loop control, it improves the accuracy and stability of temperature control in multi-stage reactors, effectively avoiding problems such as temperature overshoot and insufficient reaction conversion, improving the purity and conversion rate of reaction products, and ensuring the process stability and product quality consistency of continuous production. Moreover, this method does not require large-scale modification of existing equipment, has flexible parameter calibration, and is highly practical. It can be widely applied to multi-stage reactor production control in industries such as fine chemicals and pharmaceutical manufacturing, combining technological innovation with engineering practicality. Attached Figure Description

[0023] Figure 1 This is a flowchart illustrating a method for optimizing control parameters of a multi-stage reactor according to the present invention;

[0024] Figure 2 This is a schematic diagram illustrating the adaptive filtering effect of the first-stage reactor in this invention;

[0025] Figure 3 This is a schematic diagram illustrating the degree of temperature deviation in the first-stage reactor of the present invention;

[0026] Figure 4 This is a schematic diagram illustrating the nonlinearity intensity index of the first-stage reactor in this invention;

[0027] Figure 5 This is a schematic diagram illustrating the adaptive correction coefficients of the model vector of the first-stage reactor in this invention;

[0028] Figure 6 This is a schematic diagram illustrating the temperature control effect of the multi-stage reactor in this invention;

[0029] Figure 7 This is a schematic diagram illustrating the comparison of the average temperature control error between this method and the traditional DMC method. Detailed Implementation

[0030] 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, not all, of the embodiments of the present invention. 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.

[0031] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] This invention discloses a method for optimizing control parameters of a multi-stage reactor, referring to... Figure 1 This includes steps S1 to S5:

[0033] S1. Real-time acquisition of key operating data of the multi-stage reactor at each moment, and preprocessing of the data to obtain the initial temperature sequence and initial valve opening sequence of the multi-stage reactor at each moment.

[0034] It should be noted that the core purpose of preprocessing is to effectively remove environmental interference noise, equipment sensing noise, and bus transmission noise introduced during the acquisition and transmission of key operating data. At the same time, it preserves to the greatest extent the parameter mutation characteristics caused by factors such as sudden changes in raw material composition, sudden changes in catalyst activity, and switching of process conditions during the operation of multi-stage reactors. This avoids noise causing distortion of the control model input and prevents excessive filtering from masking the real process parameter changes, thus ensuring the accuracy and effectiveness of subsequent control parameter optimization.

[0035] Specifically, through a distributed industrial fieldbus, according to a preset sampling frequency... Synchronous reading of serial Key operating data for the multi-stage reactor includes: the reaction temperature, jacket inlet and outlet temperature difference, and the opening degree of the heating valve and cooling valve at each moment. The heating valve opening degree refers to the degree of opening of the valve in the heating medium delivery pipeline, and the cooling valve opening degree refers to the degree of opening of the valve in the cooling medium delivery pipeline. Both are core parameters for controlling the reaction temperature within the multi-stage reactor, and their changes directly affect the heat exchange efficiency and the trend of reaction temperature change within the reactor. Taking the first stage as an example... The stage reaction vessel at time Taking the operational status as an example for analysis, the data is collected sequentially according to the order of... The stage reaction vessel at time reaction temperature Temperature difference between jacket inlet and outlet and the opening degree of the heating valve at the previous moment. and cooling valve opening Based on the first The stage reaction vessel at time reaction temperature Temperature difference between jacket inlet and outlet , constituting time The The original temperature sequence of the first-stage reactor; based on the first stage... The stage reaction vessel at time Heating valve opening and cooling valve opening , constituting time The The original valve opening sequence of the stage reactor; among which, This is the stage index for a multi-stage reactor, with a value range of [value range missing]. positive integers, The total number of stages in a multi-stage reactor, sampling frequency Based on the process dynamics and control requirements of multi-stage reactors, as exemplified, hertz, .

[0036] Furthermore, all the acquired raw temperature sequences and raw valve opening sequences were standardized, and the first... The stage reaction vessel at time The Standardized observation values ​​of key operational data Standardization is performed to eliminate the dimensional differences between different key operational data, ensuring the uniformity and effectiveness of subsequent adaptive filtering calculations. Subsequently, an adaptive filtering denoising method is used to filter the standardized observation values. This method differs from the fixed filter coefficient setting of conventional filtering methods. It can dynamically adjust the filtering strategy according to the actual change characteristics of the data, achieving the dual goals of noise suppression and parameter mutation preservation.

[0037] Specifically, the first value is obtained based on standardized observation values. The stage reaction vessel at time The The filtered output values ​​of key operational data are given by the adaptive filtering formula as follows:

[0038] ;

[0039] In the formula, For the first The stage reaction vessel at time The Filtered output values ​​of key operational data; Indicates the first The stage reaction vessel at time The Standardized observation values ​​of key operational data; Indicates the first The stage reaction vessel at time The Filtered output values ​​of key operational data; Indicates time The gain coefficient of the adaptive filter.

[0040] Wherein, the gain coefficient of the adaptive filter It is not a fixed constant, but is determined by the prediction error between the standardized observation values ​​of the key operational data at the current moment and the filtered output value at the previous moment, specifically calculated using the following exponential function:

[0041] ;

[0042] In the formula, Indicates time The gain coefficient of the adaptive filter; Indicates the first The stage reaction vessel at time The Standardized observation values ​​of key operational data; Indicates the first The stage reaction vessel at time The Filtered output values ​​of key operational data; This indicates a preset error sensitivity threshold parameter, which is a fixed value calibrated based on the process characteristics of the multi-stage reactor. For example, this method sets... When the standardized observed values ​​deviate significantly from the historical filtered output values, it indicates a real abrupt change in process parameters; in this case, the gain coefficient... When the gain coefficient approaches 1, the filtering process will follow the current observation more closely, enabling the system to respond quickly to parameter changes; when the deviation is small, it indicates that the data fluctuation is caused by noise, and the gain coefficient will be lower. As the value approaches zero, the filtering process retains more historical filtering states to effectively suppress noise interference. Through the above adaptive filtering calculation, the filtered output value of each key operating data of each stage reactor at each time can be output. Based on the filtered output value of the reaction temperature and jacket inlet / outlet temperature difference of each stage reactor at each time in the key operating data, the initial temperature sequence of each stage reactor at each time is constructed. Based on the filtered output value of the heating valve opening and cooling valve opening of each stage reactor at the previous time in the key operating data, the initial valve opening sequence of each stage reactor at each time is constructed.

[0043] Furthermore, based on the adaptive filtering processing operation, a schematic diagram of the adaptive filtering effect of each stage of the reactor is obtained; for example, Figure 2This diagram illustrates the adaptive filtering effect of the first-stage reactor in this invention. The horizontal axis represents time (in seconds), signifying the time series of data acquisition. The vertical axis represents normalized temperature, a dimensionless value obtained after standardizing the data in the original temperature sequence. By comparing the shape differences between the noisy curve corresponding to the original temperature sequence and the filtered curve corresponding to the filtered output value, the smoothing and denoising effect of the adaptive filtering is reflected. Compared to the original noisy curve, the filtered curve shows significantly suppressed high-frequency noise and smoother fluctuations, while fully preserving the overall trend and parameter mutation characteristics of the temperature signal. This verifies the effectiveness of the adaptive filtering denoising method of this invention. This method can effectively remove various types of noise from the reaction temperature data while well preserving the core change characteristics of the original temperature signal, thereby improving the overall quality of key operating data and laying a reliable data foundation for the accuracy of subsequent control parameter optimization.

[0044] S2. Based on the standard deviation of the fluctuation, the average gradient of temperature change, and the extreme temperature deviation of the initial temperature sequence of each stage reactor at each time point, the degree of temperature deviation of each stage reactor at each time point is obtained.

[0045] It should be noted that the temperature response deviation characteristic can quantify the dynamic response characteristics and fluctuation degree of temperature in multi-stage reactors as valve opening is adjusted. It is directly related to the heat release rate and heat exchange efficiency of the reactor and is the core characteristic basis for subsequent matching and adaptation control models and optimization of control parameters. Through this characteristic, the dynamic temperature change law of reactors of different times and different stages can be accurately characterized, making up for the deficiency of conventional control methods that ignore the spatiotemporal differences of temperature response characteristics, and providing data support for the condition-based optimization of control parameters.

[0046] Based on this, by calculating the time-domain fluctuation statistics of the initial temperature sequence and the temperature change gradient of adjacent time points, a multi-dimensional temperature response deviation characteristic index is constructed to comprehensively characterize the temperature fluctuation amplitude, change rate and stability of each stage reactor at each time point, thereby obtaining the temperature response deviation characteristics of each stage reactor at each time point.

[0047] Specifically, setting For the length of the sliding window, the first... The stage reaction vessel at time The initial temperature sequence is denoted as ,in, For the first The stage reaction vessel at time The filtered output value, This is the index of the initial temperature sequence, with a value range of... positive integers, The sampling time interval, The preset sampling frequency; using this initial temperature sequence Based on this, three core characteristic indicators are calculated sequentially: standard deviation of temperature fluctuation, average gradient of temperature change, and deviation of temperature extreme values. The specific calculation method is as follows: First, the initial temperature sequence is obtained. The standard deviation of is used as the first The stage reaction vessel at time Temperature fluctuation standard deviation Secondly, calculate the initial temperature sequence. The difference between the maximum and minimum values ​​of the internal filter output is then divided by the initial temperature sequence. The average value of the quotient is used as the quotient of the first quotient. The stage reaction vessel at time Temperature extreme deviation Finally, calculate the first... The stage reaction vessel at time The average gradient of temperature change is given by the following formula:

[0048] ;

[0049] In the formula, It is the first The stage reaction vessel at time The average gradient of temperature change; For the first The stage reaction vessel at time The filtered output value; For the first The stage reaction vessel at time The filtered output value; The length of the sliding window. Take a positive integer greater than or equal to 2, for example, ; This is the index of the initial temperature sequence, with a value range of... positive integers, The sampling time interval, This is the preset sampling frequency.

[0050] The temperature fluctuation standard deviation is used to characterize the overall fluctuation amplitude of the initial temperature sequence within the sliding window. The larger the temperature fluctuation standard deviation, the more drastic the overall temperature fluctuation in the reactor at that time, and the worse the temperature operation stability, and vice versa. The temperature extreme value deviation is used to characterize the degree of extreme temperature fluctuation of the initial temperature sequence within the sliding window. The larger the temperature extreme value deviation, the more obvious the extreme temperature fluctuation in the reactor at that time, and the more likely the reactor will experience local overheating or undertemperature process problems. The average temperature change gradient is used to characterize the average temperature change rate of the initial temperature sequence within the sliding window. The larger the average temperature change gradient, the faster the dynamic response of the temperature in the reactor at that time, and vice versa.

[0051] Furthermore, regarding the first The stage reaction vessel at time standard deviation of fluctuation Average gradient of temperature change Temperature extreme deviation Perform sigmoid normalization and obtain the first sigmoid normalization result through weighted fusion. The stage reaction vessel at time The degree of temperature deviation is related by the following formula:

[0052] ;

[0053] In the formula, It is the first The stage reaction vessel at time The degree of temperature deviation indicates that the greater the temperature deviation, the more drastic the temperature fluctuation and the more unstable the dynamic response within the reactor at that moment, and the more significant the overall deviation from the process set temperature. They are the first The stage reaction vessel at time standard deviation of fluctuation Average gradient of temperature change Temperature extreme deviation The weighting coefficients; the weighting coefficients All are non-negative constants calibrated based on the process characteristics of multi-stage reactors, and satisfy the following conditions: Used to characterize the standard deviation of fluctuation Average gradient of temperature change Temperature extreme deviation The weighting of the influence of the degree of temperature deviation; where the weighting coefficient of the standard deviation of fluctuation is... It can be appropriately increased, for example, the setting , , It prioritizes highlighting the impact of overall temperature fluctuations on the degree of deviation, adapting to the core temperature control needs of multi-stage reactors; the specific value of the weighting coefficient can be flexibly adjusted according to the actual production process and product quality requirements.

[0054] Specifically, the calculations at each stage of the reaction vessel at each moment are obtained. The standard deviation of fluctuation, the average gradient of temperature change, and the temperature extreme value deviation are obtained to obtain the degree of temperature deviation of each stage of the reactor at each time.

[0055] Furthermore, based on the processing operations related to the degree of temperature deviation, a schematic diagram of the temperature deviation degree of each stage of the reactor is obtained; for example, Figure 3 This diagram illustrates the temperature deviation of the first-stage reactor in this invention. The horizontal axis represents time in seconds, indicating the time series of data acquisition. The vertical axis represents the degree of temperature deviation, a dimensionless value obtained by normalizing and weighting the standard deviation of fluctuation, the average gradient of temperature change, and the extreme deviation of temperature after calculating the initial temperature series. The fluctuation characteristics of the curves in the diagram show that when the temperature fluctuation within the reactor intensifies, the rate of change accelerates, or extreme deviations occur, the temperature deviation value increases synchronously, as seen in the peak regions around 2 seconds, 4 seconds, and 8 seconds. When the temperature operation is relatively stable, the temperature deviation remains at a low level, such as the flat regions from 0 seconds to 1 second and from 6 seconds to 7 seconds. This characteristic indicates that the index can accurately quantify the fluctuation amplitude, rate of change, and stability of the temperature within the reactor at different times, providing a reliable characteristic basis for the subsequent construction of nonlinear intensity indices.

[0056] S3. Based on the temperature deviation of each stage of the reactor at each time point, construct the nonlinear intensity index of the reaction at each stage of the reactor at each time point.

[0057] It should be noted that the temperature deviation level only quantifies the temperature data fluctuation characteristics of the multi-stage reactor, without relating it to the physical root causes of the chemical reaction. The nonlinearity of the reaction process is the core physical factor causing temperature deviation. When the reaction enters a stage of intense exothermic or endothermic reaction, the heat of reaction superimposed on external heat exchange alters the system's thermal gain, causing the reaction process to deviate from the linear nominal model. This step transforms the temperature deviation level at the data level into a reaction nonlinearity intensity index characterizing the physical characteristics of the chemical reaction scenario. It accurately quantifies the degree to which the reaction process deviates from the linear nominal model, providing a core physical basis for subsequent matching and adaptation of the dynamic matrix control model and optimization of control parameters. This overcomes the shortcomings of conventional control methods that rely solely on data-level regulation and ignore the physical characteristics of the reaction.

[0058] Based on this, and combining the core logic of the Arrhenius law that the reaction rate increases exponentially with increasing temperature, the degree of temperature deviation of each stage of the reactor at each time is used to integrate the measured temperature of the reactor, the reference temperature, and the characteristic temperature constant of the reaction system to construct a reaction nonlinearity intensity index. This achieves the combination of data characteristics and reaction physical properties, and accurately characterizes the degree of reaction nonlinearity of reactors at different times and different stages.

[0059] Specifically, with the first The stage reaction vessel at time Based on the degree of temperature deviation, combined with the time of the reaction vessel The measured temperature, the reference temperature under nominal operating conditions, and the characteristic temperature constant of the reaction system are used to construct the first... The stage reaction vessel at time The nonlinear intensity index of the response is expressed by the following formula:

[0060]

[0061] In the formula, For the first The stage reaction vessel at time The nonlinearity intensity index is used to quantify the degree to which the reaction process in the reactor deviates from the linear nominal model at a given moment. The larger the value, the more significant the nonlinear characteristics of the reaction process, the greater the deviation of the system thermal gain from the initial linear nominal model, and the more intense the chemical reaction in the reactor. For the first The stage reaction vessel at time The degree of temperature deviation, with a value range of... between; For the first The stage reaction vessel at time The measured temperature, i.e., the filtered output value of the reaction temperature at that moment. ; The reference temperature is the nominal operating temperature of the multi-stage reactor, while the fixed value is the temperature calibrated according to the process settings of the multi-stage reactor. It is the characteristic temperature constant of the reaction system, which is positively correlated with the activation energy of the reaction system. It is a fixed value calibrated according to the type of chemical reaction and material properties in the multi-stage reactor. Based on the natural constant An exponential function with base 1.

[0062] Among them, the The stage reaction vessel at time Measured temperature The filtered output value of the reaction temperature at that moment in step S1 is used directly. To ensure the consistency and accuracy of data sources; reference temperature under nominal operating conditions. The process temperature is set for the stable reaction stage of a multi-stage reactor, for example, according to the requirements of a fine chemical esterification reaction process. The characteristic temperature constant of the reaction system Calibration is calculated based on the reaction activation energy; for example, setting... The specific value can be flexibly adjusted according to the actual chemical reaction type, material composition and process requirements in the multi-stage reactor.

[0063] It should be noted that the reaction nonlinearity intensity index is obtained by multiplying the temperature deviation degree by the temperature exponential term. The greater the temperature deviation in the reactor, the more significant the deviation between the actual temperature response and the linear nominal model. At the same time, the higher the measured temperature is than the reference temperature, the more the reaction rate increases exponentially according to the Arrhenius law. The stronger the superposition effect of reaction heat and external heat exchange, the more significant the change in system thermal gain. Ultimately, this leads to an exponential increase in the reaction nonlinearity intensity index, which precisely matches the physical characteristic that the degree of nonlinearity in chemical reactions increases sharply with increasing temperature.

[0064] Furthermore, by sequentially traversing all stages and all sampling times of the multi-stage reactor, the temperature deviation and measured temperature of each stage of the reactor at each time are substituted into the above relationship to complete the construction of the nonlinear intensity index of the reaction at each stage of the reactor at each time.

[0065] Specifically, based on the processing operation of the reaction nonlinearity intensity index, a schematic diagram of the reaction nonlinearity intensity index for each stage of the reactor is obtained; for example, Figure 4 This diagram illustrates the nonlinear intensity index of the first-stage reactor in this invention. The horizontal axis represents time (in seconds), indicating the time series of data acquisition. The vertical axis represents the nonlinear intensity, a quantitative index obtained by integrating temperature deviation, measured temperature, reference temperature, and the characteristic temperature constant of the reaction system. The curves in the diagram show that as the temperature deviation increases, the nonlinear intensity index increases significantly, peaking around 8 seconds, corresponding to a greater deviation of the reaction process from the linear nominal model. Conversely, when the temperature is relatively stable, the index remains at a lower level. This characteristic indicates that the index can accurately correlate temperature fluctuations with the nonlinear characteristics of the reaction, effectively quantifying the degree of nonlinearity of the reaction process at different times. This provides a core basis for generating adaptive correction coefficients for the subsequent model vector, closely aligning with the physical characteristics of the process.

[0066] S4. Based on the nonlinear intensity index of the reaction at each stage of the reactor at each time point and the rate of change of the initial valve opening sequence, generate the adaptive correction coefficient of the model vector for each stage of the reactor at each time point.

[0067] It should be noted that conventional dynamic matrix control algorithms use fixed model vectors to solve for control quantities, which cannot adapt to the dynamic changes in nonlinear intensity during multi-stage reactor reactions. This leads to a mismatch between the model vectors and the actual operating conditions, resulting in deviations in the predicted output. This step uses the reaction nonlinear intensity index as the core physical basis and integrates the initial valve opening sequence change rate as an actuator-side control feature to generate smooth and bounded adaptive correction coefficients for the model vectors. This enables real-time adaptation and adjustment of the internal model vectors of the dynamic matrix control, allowing the model vectors to follow the dynamic changes in reaction nonlinearity and valve control status. This solves the mismatch problem between the fixed model and dynamic operating conditions at the model level, providing a model basis adapted to the current operating conditions for subsequent solutions to the optimal control quantity, and ensuring the real-time performance and accuracy of control parameter optimization.

[0068] Based on this, and in conjunction with the stability requirements of the control system, the process nonlinearity characteristics characterized by the reaction nonlinearity intensity index are used to couple the dynamic control of the actuator end reflected by the rate of change of the initial valve opening sequence. By constructing a calculation formula with a smoothing factor, a sign function, and an amplitude weight, an adaptive correction coefficient of the model vector that smoothly fluctuates around the nominal model coefficient is generated. This ensures that the correction coefficient responds quickly to changes in operating conditions while avoiding sudden changes that could lead to instability of the control system.

[0069] Specifically, with the first The stage reaction vessel at time The reaction nonlinearity intensity index is the core, and the reaction vessel is integrated at time... The initial valve opening sequence change rate, combined with the smoothing factor, sign function, and maximum correction magnitude adjustment weight, is used to construct the first... The stage reaction vessel at time The adaptive correction coefficients for the model vector are expressed as follows:

[0070] ;

[0071] In the formula, For the first The stage reaction vessel at time The adaptive correction coefficient of the model vector is used to adjust the internal model vector of the dynamic matrix control in real time. Its value fluctuates smoothly around 1, representing the scaling ratio of the actual temperature change caused by the change of unit control quantity. For the first The stage reaction vessel at time The nonlinear intensity index of the reaction characterizes the degree to which the reaction process deviates from the linear nominal model; For the first The stage reaction vessel at time The rate of change of the initial valve opening sequence characterizes the real-time control dynamics of the valve opening. The preset smoothing factor is a fixed constant greater than 0, used to prevent the denominator from being zero and to adjust the sensitivity of the correction coefficient to changes in operating conditions. The sign function is determined by the degree of temperature deviation. As input, determine the direction of temperature deviation and characterize the trend of system thermal gain change caused by exothermic or endothermic reaction; The preset maximum correction range adjustment weight is a fixed constant between 0 and 1, used to limit the maximum variation of the correction coefficient and ensure the stability of the control system.

[0072] Among them, smoothing factor Based on the calibration of the process dynamic characteristics of the multi-stage reactor, an exemplary setting is provided. Maximum correction magnitude adjustment weight The settings are configured according to the stability requirements of the control system, as exemplary. Both values ​​can be flexibly adjusted according to actual production processes and control requirements; sign function The rule for determining the value is: when When the temperature deviation exceeds its preset threshold, it indicates that the temperature deviation is caused by increased exothermic reaction, the system thermal gain increases, and the sign function takes the value of 1; when When the temperature deviation is less than its preset threshold, it indicates that the temperature deviation is caused by increased endothermic reaction, the system thermal gain decreases, and the sign function takes a value of -1; when When the value equals the preset threshold, the sign function takes the value of 0. The preset threshold is calibrated according to the degree of temperature deviation under stable operating conditions of the multi-stage reactor, and is set to 0.2 for example.

[0073] Furthermore, the first The stage reaction vessel at time Initial valve opening sequence rate of change The calculation method is as follows: Based on the initial valve opening sequence at that moment, calculate the rate of change of the heating valve opening and the cooling valve opening at adjacent moments within the sliding window, and take the weighted sum of their absolute values ​​as the rate of change of the initial valve opening sequence. For example, the relationship is as follows:

[0074] ;

[0075] In the formula, , The first The stage reaction vessel at time The filtered output values ​​for the heating valve opening and cooling valve opening, respectively. , These are the previous sampling times, i.e. The filtered output values ​​of the heating valve opening and cooling valve opening at any given time. , These are the weighting coefficients for the rate of change of the opening degree of the heating and cooling valves, respectively, and are non-negative constants that satisfy... Exemplary settings , ; This represents the sampling time interval.

[0076] It should be noted that multiplying the reaction nonlinearity intensity index by the rate of change of the initial valve opening sequence achieves coupling between process nonlinearity characteristics and actuator-side control characteristics, allowing the correction coefficient to not only respond to the nonlinear changes of the reaction itself but also adapt to the control state of the valve opening; smoothing factor The introduction of this avoids the case where the denominator is zero, and also makes the change of the correction coefficient smoother; sign function The direction of temperature deviation is used to determine the trend of heat gain change, enabling the correction coefficient to be adjusted to the model vector to adapt to different operating conditions of heat release or heat absorption; the maximum correction amplitude is adjusted by weighting. The fluctuation range of the correction coefficient is limited to prevent the control system from becoming unstable due to excessive correction. Ultimately, the correction coefficient is smooth and bounded around 1, which ensures both the adaptability of the model vector to dynamic conditions and the stability of the control system.

[0077] Furthermore, by sequentially traversing all stages and all sampling times of the multi-stage reactor, the reaction nonlinearity intensity index and the rate of change of the initial valve opening sequence of each stage reactor at each time are substituted into the above relationship. Combined with the values ​​of the sign function, smoothing factor and maximum correction amplitude adjustment weight, the adaptive correction coefficient of the reactor model vector at each time is generated.

[0078] Specifically, based on the processing operation of the model vector adaptive correction coefficients, a schematic diagram of the model vector adaptive correction coefficients for each stage of the reactor is obtained; for example, Figure 5 This diagram illustrates the adaptive correction coefficients of the model vector for the first-stage reactor in this invention. The horizontal axis represents time (in seconds), indicating the time series of data acquisition. The vertical axis represents the correction coefficients, with the dashed line indicating the baseline value of 1.0 for the nominal model. The curve's fluctuation characteristics show that the correction coefficients smoothly fluctuate around the baseline value of 1.0, without significant abrupt changes, thus preventing control system instability. Furthermore, the correction coefficients dynamically adjust according to operating conditions: when the nonlinearity index of the reaction increases, the correction coefficients synchronously deviate from the baseline value, achieving condition-based scaling of the dynamic matrix control model; while when the reaction conditions are relatively stable, the correction coefficients converge towards the baseline value. This characteristic indicates that the correction coefficients possess both rapid response capabilities to nonlinear conditions and ensure system stability through constraint mechanisms, providing adaptable and reliable parameter support for the subsequent accurate correction of the control model.

[0079] S5. Based on the adaptive correction coefficient of the reactor model vector at each time step, optimize the control parameters of the multi-stage reactor and perform regulation.

[0080] It should be noted that this step uses the adaptive correction coefficients of the model vector obtained in step S4 as the core to correct the core model parameters of the dynamic matrix control in real time, thereby solving for the optimal control parameters that adapt to the current nonlinear reaction conditions. At the same time, it completes the execution and rolling optimization of the control parameters, and finally realizes the adaptive optimization and control of the control parameters of the multi-stage reactor. This effectively solves the problem of mismatch between control parameters and dynamic conditions caused by the fixed model in conventional dynamic matrix control, ensures the accuracy and stability of temperature control of the multi-stage reactor, avoids process problems such as temperature overshoot and insufficient reaction conversion rate, and improves product quality and production safety.

[0081] Specifically, with the first The stage reaction vessel at time Model vector adaptive correction coefficient Taking this as an example, the basic step response vector of the dynamic matrix control is scaled and corrected according to the operating conditions using this correction coefficient to obtain the real-time step response vector. The relationship is as follows:

[0082] ;

[0083] In the formula, For the first The stage reaction vessel at time The real-time step response vector is the corrected dynamic matrix control core model parameter, which is used to accurately match the system thermal gain under the current nonlinear operating condition and solve the problem of mismatch between the conventional fixed model and the dynamic operating condition. For the first The stage reaction vessel at time The adaptive correction coefficient of the model vector is directly used from the calculation results of step S4 to perform working condition scaling correction on the basic step response vector. Its value fluctuates smoothly around 1 to adapt to the dynamic changes in the degree of nonlinearity of the response. It is the basic step response vector for dynamic matrix control, a fixed value calibrated by step response test under stable operating conditions of multi-stage reactor, and serves as the basic benchmark for model parameter correction.

[0084] Furthermore, with the first The stage reaction vessel at time Model vector adaptive correction coefficient Taking this as an example, based on the real-time step response vector obtained from the above correction, and combined with the deviation between the process setting reference temperature and the measured temperature inside the vessel, the optimal valve opening adjustment amount adapted to the current operating conditions is solved, and the relationship is as follows:

[0085] ;

[0086] In the formula, For the first The stage reaction vessel at time The optimal valve opening adjustment is the core control parameter after this optimization, used to adjust the current valve opening to offset the temperature deviation and adapt to the current nonlinear operating condition. To optimize the coefficients, calibration is performed based on the process characteristics, temperature control sensitivity, and actuator control capabilities of the multi-stage reactor. The coefficients are then flexibly set according to the actual production process and control requirements of the multi-stage reactor. An example setting is provided below. It is used to adjust the sensitivity of the control parameters to avoid excessive valve opening adjustment leading to temperature overshoot, or insufficient adjustment to effectively offset temperature deviation. The process temperature is set for the stable reaction stage of the multi-stage reactor, for example, 85°C; For the first The stage reaction vessel at time The measured temperature, i.e., the filtered output value of the reaction temperature at that moment. .

[0087] Specifically, with the first The stage reaction vessel at time Model vector adaptive correction coefficient Taking this as an example, the optimal valve opening adjustment obtained above is superimposed on the current valve opening to obtain the final valve opening execution value. The relationship is as follows:

[0088] ;

[0089] In the formula, For the first The stage reaction vessel at time The valve opening execution value is the final control parameter output to the actuator, used to directly regulate the opening degree of the heating valve and cooling valve; For the first The stage reaction vessel at time The current valve opening, i.e., the filtered output value of the initial valve opening sequence, specifically includes the first... The stage reaction vessel at time Filtered output values ​​of heating valve opening and cooling valve opening , .

[0090] It should be noted that the first The stage reaction vessel at time When calculating the valve opening value, it is necessary to separately filter the output values ​​for the heating valve opening and the cooling valve opening. , The system performs separate calculations, that is, it solves for the optimal opening adjustment of the heating valve and the cooling valve separately, and then superimposes them onto the current opening of the heating and cooling valves to obtain their respective valve opening execution values. These values ​​are then output to the corresponding actuators to achieve coordinated and precise control of heating and cooling, meeting the core requirements of temperature control in multi-stage reactors.

[0091] Furthermore, by sequentially traversing all stages and sampling times of the multi-stage reactor, the adaptive correction coefficient of the model vector for each stage of the reactor at each time and the measured temperature are substituted into the above three relationships to solve for the optimal opening adjustment amount and opening execution value of the heating valve and cooling valve of each stage of the reactor. Then, the opening execution value is output to the corresponding actuator, which precisely adjusts the opening degree of the heating and cooling valves according to the execution value, thereby controlling the heat exchange efficiency inside the reactor, realizing precise control of the reaction temperature inside the reactor, offsetting temperature deviation, and adapting to the current nonlinear reaction conditions.

[0092] Specifically, at the next sampling moment, the complete process from steps S1 to S5 is repeated. Based on the newly collected key operating data, the degree of temperature deviation, the nonlinearity index of the reaction, and the adaptive correction coefficient of the model vector are recalculated. Then, based on the new parameters, the valve opening adjustment amount and execution value are continuously optimized to ensure that the control parameters always adapt to the dynamic changes of the reaction conditions of the multi-stage reactor. This achieves full-cycle adaptive optimization and precise control of the control parameters of the multi-stage reactor, effectively solving problems such as temperature overshoot and insufficient reaction conversion rate caused by conventional fixed model control, and ensuring the stability of product production quality and the safety of the production process.

[0093] Furthermore, based on the above-mentioned control operations, a schematic diagram of the temperature control effect of each stage of the reactor is obtained; for example, Figure 6This diagram illustrates the temperature control effect of the multi-stage reactor in this invention. The horizontal axis represents time (in seconds), indicating the time series of data acquisition. The vertical axis represents temperature (in °C). The dashed line represents the process-set reference temperature of 85 °C. Different colored curves correspond to the actual temperature changes of the first, second, and third-stage reactors. The diagram clearly demonstrates the control logic: the temperature curves of each reactor consistently fluctuate around the reference temperature. This is a direct result of correcting the basic step response vector through adaptive correction coefficients of the model vector, solving for the optimal valve opening adjustment, and implementing coordinated control. When the temperature deviates from the reference value, the real-time corrected step response vector will... By matching the current nonlinear operating conditions, the system outputs appropriate valve adjustments, and rapidly offsets deviations through coordinated control of heating and cooling valves. For example, after 2 seconds, all curves quickly converge to the reference temperature. Simultaneously, the full-cycle rolling optimization mechanism ensures that control parameters continuously adapt to changes in operating conditions, and the temperature fluctuations of each reactor remain within a small range in subsequent periods, without temperature overshoot or long-term deviation. This characteristic indicates that this method can achieve precise and coordinated temperature control of multi-stage reactors, effectively matching the nonlinear operating conditions of the reaction, ensuring the temperature stability of the reaction process, and providing reliable temperature control support for improving product quality and production safety.

[0094] Specifically, based on the above operations, a comparative diagram of the average temperature control error between this method and the traditional DMC method is obtained, see [link to diagram]. Figure 7 The graph shows the number of reactor stages (1, 2, and 3) on the horizontal axis and the average temperature control error (°C) on the vertical axis. One type of bar represents the average error of this method, while the other represents the average error of the traditional DMC method. The data shows that for a 1-stage reactor, the average error of this method is approximately 0.8°C, while the traditional DMC method reaches 3.5°C; for a 2-stage reactor, the average error of this method is approximately 1.2°C, while the traditional DMC method reaches 4.2°C; and for a 3-stage reactor, the average error of this method is approximately 1.0°C, while the traditional DMC method reaches 3.8°C. The significant difference in error between the two methods clearly demonstrates that this method, through adaptive model vector correction and full-cycle rolling optimization, significantly improves the temperature control accuracy of multi-stage reactors. It effectively solves the problem of large errors caused by the mismatch between the traditional fixed model and dynamic operating conditions, highlighting the obvious advantage of this method in temperature control accuracy.

[0095] This invention also discloses a control parameter optimization system for a multi-stage reactor, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a control parameter optimization method for a multi-stage reactor according to the present invention.

[0096] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

Claims

1. A method for optimizing control parameters of a multi-stage reactor, characterized in that, include: Real-time acquisition of key operating data at various moments of a multi-stage reactor connected in series; standardization and adaptive filtering denoising processing of the key operating data in sequence; and construction of initial temperature sequence and initial valve opening sequence based on the filtered output value. Based on the initial temperature sequence of each stage of the reactor at each time point, the standard deviation of temperature fluctuation, the average gradient of temperature change, and the temperature extreme value deviation are calculated. After normalization and weighted fusion, the degree of temperature deviation of each stage of the reactor at each time point is obtained. Combining the Arrhenius law, the degree of temperature deviation is used to fuse the measured temperature, the reference temperature, and the characteristic temperature constant of the reaction system to construct the reaction nonlinearity intensity index of each stage of the reactor at each time point. Based on the reaction nonlinearity intensity index coupled with the initial valve opening sequence change rate, and combined with the smoothing factor, sign function and maximum correction amplitude adjustment weight, the adaptive correction coefficient of the model vector of each stage reactor at each time is obtained. By using adaptive correction coefficients to correct the basic step response vector of dynamic matrix control, the optimal valve opening adjustment is solved, and the valve opening execution value is calculated. This value is then output to the actuator to complete temperature control, thereby achieving full-cycle rolling optimization of control parameters for multi-stage reactors. Methods for obtaining the nonlinear intensity index of the response include: In the formula, For the first The stage reaction vessel at time The nonlinear intensity index of the response; For the first The stage reaction vessel at time The degree of temperature deviation; For the first The stage reaction vessel at time The measured temperature; This is the reference temperature under the nominal operating conditions of a multi-stage reactor. The characteristic temperature constant of the reaction system; Based on the natural constant An exponential function with base 1; This is the stage index for a multi-stage reactor. Sampling time; Obtain the adaptive correction coefficients of the model vector for each stage of the reactor at each time step, including: In the formula, For the first The stage reaction vessel at time The adaptive correction coefficients for the model vector; For the first The stage reaction vessel at time The rate of change of the initial valve opening sequence; This is the preset smoothing factor; It is a symbolic function; Adjust the weights to the preset maximum correction range.

2. The method for optimizing control parameters of a multi-stage reactor according to claim 1, characterized in that, The adaptive filtering denoising method includes: ； In the formula, For the first The stage reaction vessel at time The Filtered output values ​​of key operational data; Indicates the first The stage reaction vessel at time The Standardized observation values ​​of key operational data; Indicates the first The stage reaction vessel at time The Filtered output values ​​of key operational data; Indicates time The gain coefficient of the adaptive filter; This is the stage index for a multi-stage reactor. Sampling time, An index for critical runtime data.

3. The method for optimizing control parameters of a multi-stage reactor according to claim 2, characterized in that, The method for obtaining the gain coefficient of the adaptive filter includes: ； In the formula, Indicates time The gain coefficient of the adaptive filter; Indicates the first The stage reaction vessel at time The Standardized observation values ​​of key operational data; Indicates the first The stage reaction vessel at time The Filtered output values ​​of key operational data; This represents the preset error sensitivity threshold parameter; Based on the natural constant An exponential function with base 1; This is the stage index for a multi-stage reactor. The sampling time.

4. The method for optimizing control parameters of a multi-stage reactor according to claim 1, characterized in that, The method for obtaining the rate of change of the initial valve opening sequence includes: ； In the formula, It is the first The stage reactor at time The rate of change of the initial valve opening sequence; , The first The stage reactor at time The filtered output values ​​for the heating valve opening and cooling valve opening, respectively. , They are time points The filtered output values ​​for the opening degree of the heating valve and the opening degree of the cooling valve; It is the rate of change of the opening degree of the heating valve. These are its weighting coefficients; It is the rate of change of the cooling valve opening. These are its weighting coefficients; The sampling time interval; This is the stage index for a multi-stage reactor. The sampling time.

5. The method for optimizing control parameters of a multi-stage reactor according to claim 1, characterized in that, The method for correcting the basic step response vector using adaptive correction coefficients of the model vector includes: The basic step response vector of the dynamic matrix control is scaled and corrected by the adaptive correction coefficient of the model vector of each stage reactor at the corresponding time, so as to obtain the real-time step response vector of each stage reactor at the corresponding time; the basic step response vector is a fixed value calibrated by step response test under the stable operating condition of multi-stage reactor.

6. The method for optimizing control parameters of a multi-stage reactor according to claim 1, characterized in that, The process of finding the optimal valve opening adjustment amount and calculating the valve opening execution value includes: By combining the optimization coefficients, the real-time step response vector, and the deviation between the reference temperature and the measured temperature, the optimal valve opening adjustment amount for each stage of the reactor at the corresponding time is obtained. Then, the optimal valve opening adjustment amount is superimposed on the current valve opening of each stage of the reactor at the corresponding time, and the final valve opening execution value is obtained for the heating valve opening and the cooling valve opening respectively. The optimization coefficients are constants calibrated according to the process characteristics of the multi-stage reactor and the control capability of the actuator, and the current valve opening is the filtered output value of the initial valve opening sequence.

7. The method for optimizing control parameters of a multi-stage reactor according to claim 1, characterized in that, The method for achieving full-cycle rolling optimization of control parameters for multi-stage reactors includes: re-collecting key operating data of the multi-stage reactors at the next sampling time, and sequentially performing all operations such as standardization, adaptive filtering and denoising, feature index calculation, model vector adaptive correction coefficient generation, control parameter optimization, and regulation execution. The control parameters are updated in real time based on the new operating data, so that the control parameters always adapt to the dynamic changes in the reaction conditions of the multi-stage reactors, thereby achieving full-cycle adaptive rolling optimization of control parameters.

8. A control parameter optimization system for a multi-stage reactor, characterized in that, include: A processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a method for optimizing control parameters of a multi-stage reactor according to any one of claims 1-7.

Images