A method and system for optimizing control of a catalytic cracking desorption column section

Abstract

The present application relates to catalytic cracking technical field, disclose a kind of catalytic cracking desorption tower section optimization control method and system, the method includes: based on historical operation data and catalytic cracking mechanism constructs multivariate dynamic response model, the dynamic relationship between control variable and controlled variable is described;According to the target interval of controlled variable and the constraint interval of control variable, the solution quadratic programming determines the feasible operation interval of control variable and is adjusted limit;Based on the model constructs dynamic matrix control interval controller, interval control performance index is calculated by weighted time integral and deviation integral;Parameter optimization is carried out to prediction time domain, control time domain and weight coefficient using bayesian optimization algorithm, realize the stable control of controlled variable;Real-time acquisition operation data is carried out through OPC communication protocol, and model is updated and adaptive optimization controller parameter is optimized.The present application realizes dynamic correction control behavior, prevents desorption tower temperature fluctuation and liquid level instability phenomenon.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of catalytic cracking, in particular to a catalytic cracking desorption tower interval optimization control method and system. BACKGROUND

[0002] Fluid Catalytic Cracking (FCC for short) technology is one of the most core secondary processing processes in modern petroleum refining industry, which is used to crack heavy oil into light hydrocarbons, so as to improve the yield of gasoline, diesel and liquefied gas and other high value-added products. In the catalytic cracking device, the desorption tower (also known as the fractionating tower) is a key heat separation unit, which mainly functions to separate the mixed components entering the tower through the vapor-liquid equilibrium process, control the content of light and heavy components, and ensure the stability of the quality of subsequent products and the optimization of energy utilization. The stability of the desorption tower directly affects the energy balance, catalyst circulation efficiency and production safety of the entire catalytic cracking system, so it is of great significance to realize high-precision intelligent control of the desorption tower.

[0003] The existing desorption tower control system usually adopts a conventional proportional-integral-derivative (PID) controller or a simple cascade control loop to independently adjust the tower bottom temperature, the tower top temperature and the tower liquid level. Although this type of control method has a simple structure and low implementation cost, for a complex system such as the desorption tower, which has strong nonlinearity, strong coupling between variables, large process lag and multiple constraint characteristics, the traditional PID control is difficult to meet the requirements of high precision and high stability, and it is difficult to maintain stable control of the desorption tower liquid level and the desorption tower bottom temperature under complex constraint conditions.

[0004] Therefore, it is necessary to design a catalytic cracking desorption tower interval optimization control method and system to solve the problems existing in the prior art. SUMMARY

[0005] In view of this, the present application provides a catalytic cracking desorption tower interval optimization control method and system, which aims to solve the problem of difficult stable control of the desorption tower liquid level and the desorption tower bottom temperature.

[0006] In one aspect, the present application provides a catalytic cracking desorption tower interval optimization control method, comprising:

[0007] Based on historical operation data and catalytic cracking process mechanism, a multivariable dynamic response model is constructed, which describes the dynamic response relationship between control variables and controlled variables;

[0008] According to the target control interval of the controlled variable and the operation constraint interval of the control variable, and based on solving a quadratic programming problem, a feasible operation interval of the control variable satisfying the target control interval is calculated, and the adjustment of the control variable is limited according to the feasible operation interval;

[0009] Based on the multivariable dynamic response model, a dynamic matrix control interval controller is constructed, and the weighted time integral and the deviation amplitude integral of the controlled variable outside the target control interval are calculated as an interval control performance index by the dynamic matrix control interval controller;

[0010] Based on a Bayesian optimization algorithm, and with the minimization of the interval control performance index as the target, parameter optimization is performed on the prediction time domain, control time domain and control weight coefficient of the dynamic matrix control interval controller to obtain a controller parameter combination that stabilizes the controlled variable in the target control interval; and the controller parameter combination is applied to the dynamic matrix control interval controller;

[0011] Based on the OPC communication protocol, the measured value of the controlled variable is obtained in real time and the set value of the control variable is calculated;

[0012] The actual operation data of the desorption tower is collected in real time as process feedback data, and the multivariable dynamic response model is updated according to the process feedback data and the Bayesian optimization algorithm is re-executed to obtain an updated dynamic matrix control interval controller.

[0013] Further, the control variable includes the desethanized gasoline out of the desorption tower flow valve opening, the desorption tower bottom reboiler one shell temperature three-way valve opening and the condensed oil into the desorption tower flow valve opening; the controlled variable includes the desorption tower liquid level and the desorption tower bottom temperature.

[0014] Further, when constructing the multivariable dynamic response model based on historical operation data and catalytic cracking process mechanism, it includes:

[0015] The historical operation data is preprocessed, and the preprocessing includes removing signal noise, rejecting abnormal data points and standardizing processing;

[0016] The catalytic cracking process mechanism includes the mass conservation principle and the energy conservation principle, and a simplified mechanism model of the desorption tower is established according to the catalytic cracking process mechanism;

[0017] The simplified mechanism model and the preprocessed historical operation data are determined by a recursive parameter identification method to determine the multivariable dynamic response parameters, and a multivariable dynamic response model is constructed according to the multivariable dynamic response parameters.

[0018] Further, according to the target control interval of the controlled variable and the operation constraint interval of the control variable, and based on solving a quadratic programming problem to calculate the feasible operation interval of the control variable satisfying the target control interval, comprising:

[0019] presetting the target control interval of the controlled variable and the operation constraint interval of the control variable;

[0020] constructing a quadratic programming problem objective function, which is determined according to the size of the control variable adjustment amount and the control effect;

[0021] setting a quadratic programming problem constraint condition, which includes that the controlled variable must be kept in the target control interval and the control variable change cannot exceed the operation constraint interval;

[0022] solving the quadratic programming problem by using an effective set method to obtain the actual safe operation interval of the control variable, which is the feasible operation interval of the control variable satisfying the target control interval.

[0023] Further, based on the multi-variable dynamic response model to construct a dynamic matrix control interval controller, comprising:

[0024] calculating the step response coefficient between the control variable and the controlled variable based on the multi-variable dynamic response model, and constructing a prediction model according to the step response coefficient to predict the change trend of the controlled variable in the future time domain;

[0025] setting an interval control objective function according to that the controlled variable does not produce a penalty when it is in the target control interval, and a corresponding penalty term is produced according to the deviation degree and the duration when the controlled variable exceeds the target control interval;

[0026] setting an interval constraint condition according to the requirement that the controlled variable must be maintained in the target control interval and the limitation that the control variable must satisfy the operation constraint range;

[0027] integrating the prediction model, the interval control objective function and the interval constraint condition to form the dynamic matrix control interval controller.

[0028] Further, through the dynamic matrix control interval controller, when calculating the sum of the weighted time integral and the deviation amplitude integral of the controlled variable outside the target control interval as an interval control performance index, comprising:

[0029] determining the deviation amount of the controlled variable exceeding the target control interval based on the interval control objective function;

[0030] calculating the weighted time integral according to the length of time that the controlled variable exceeds the target control interval and the deviation amount thereof;

[0031] calculating the deviation amplitude integral according to the deviation amount of the controlled variable from the target control interval;

[0032] proportionally combining the weighted time integral and the deviation amplitude integral to form the interval control performance index.

[0033] Further, when the parameters of the dynamic matrix control interval controller are optimized to obtain a controller parameter combination that stabilizes the controlled variable within the target control interval, the method comprises:

[0034] setting a Bayesian optimization target based on the interval control performance index, the Bayesian optimization target being oriented to minimize the interval control performance index;

[0035] determining a feasible range of controller parameter combinations according to the prediction time domain, the control time domain and the control weight coefficient;

[0036] constructing a performance prediction model based on historical evaluation data, the performance prediction model being used to estimate the control effect of different controller parameter combinations;

[0037] repeating the parameter evaluation and model updating process until a convergence condition is met; and determining a controller parameter combination that minimizes the interval control performance index.

[0038] Further, when the measurement value of the controlled variable is obtained in real time and the set value of the control variable is calculated based on the OPC communication protocol, the method comprises:

[0039] establishing a secure communication connection based on the OPC communication protocol; obtaining the measurement values of the desorption column liquid level and the column bottom temperature in real time; inputting the obtained measurement values into the dynamic matrix control interval controller and performing data processing in combination with the current controller parameters to calculate the set value of the control variable.

[0040] Further, when the multivariable dynamic response model is updated according to the process feedback data and the Bayesian optimization algorithm is re-executed to obtain an updated dynamic matrix control interval controller, the method comprises:

[0041] calculating a prediction error according to the process feedback data and the predicted data of the multivariable dynamic response model; comparing the prediction error with a prediction error threshold; when the prediction error is greater than the prediction error threshold, updating the multivariable dynamic response model by using a recursive parameter identification method; re-evaluating an interval control performance index by using the updated multivariable dynamic response model; executing a Bayesian optimization algorithm based on the re-evaluated performance index to obtain an updated controller parameter combination; and applying the updated controller parameter combination to the dynamic matrix control interval controller.

[0042] In another aspect, the application further provides a catalytic cracking disengaging tower interval optimization control system for applying the catalytic cracking disengaging tower interval optimization control method, comprising:

[0043] A response model construction module is configured to construct a multivariable dynamic response model based on historical operation data and a catalytic cracking process mechanism, the multivariable dynamic response model describing a dynamic response relationship between a control variable and a controlled variable;

[0044] A control variable limiting module is configured to calculate a feasible operation interval of the control variable satisfying a target control interval of the controlled variable according to an operation constraint interval of the control variable and based on solving a quadratic programming problem, and limit adjustment of the control variable according to the feasible operation interval;

[0045] A controller construction module is configured to construct a dynamic matrix control interval controller based on the multivariable dynamic response model, and calculate a sum of weighted time integral and deviation amplitude integral of the controlled variable outside the target control interval as an interval control performance index by the dynamic matrix control interval controller;

[0046] A parameter combination obtaining module is configured to perform parameter optimization on a prediction time domain, a control time domain and a control weight coefficient of the dynamic matrix control interval controller based on a Bayesian optimization algorithm and with the objective of minimizing an interval control performance index, to obtain a controller parameter combination that stabilizes the controlled variable within the target control interval, and to apply the controller parameter combination to the dynamic matrix control interval controller;

[0047] An OPC communication module is configured to obtain a measurement value of the controlled variable and calculate a set value of the control variable in real time based on an OPC communication protocol;

[0048] A model dynamic updating module is configured to collect actual operation data of the disengaging tower as process feedback data in real time, and update the multivariable dynamic response model and re-execute the Bayesian optimization algorithm to obtain an updated dynamic matrix control interval controller according to the process feedback data.

[0049] Compared with the prior art, the beneficial effects of the present application are that: by constructing a multivariate dynamic response model based on the catalytic cracking process mechanism and historical operation data, the dynamic coupling relationship between the control variables and the controlled variables can be accurately described, the systematic modeling of the complex industrial process is realized, and reliable theoretical support is provided for the intelligent control of the desorption tower. The dynamic matrix control (DMC) interval control strategy is adopted, the prediction model is combined with the constraint conditions, and it is ensured that the controlled variables are always maintained within the preset target interval; when the controlled variables deviate from the interval, a penalty term will be automatically generated according to the deviation amplitude and duration, so as to dynamically correct the control behavior and prevent the desorption tower temperature fluctuation and liquid level instability. Through the Bayesian optimization algorithm with the minimum interval control performance index as the target, the optimal combination of the prediction time domain, the control time domain and the control weight is automatically searched, so that the controller can dynamically self-adjust according to different working conditions, and the self-learning and self-adaptive ability of the desorption tower control system is improved. Through the OPC communication protocol, the real-time interaction of the control system and the field equipment data is realized, the desorption tower operation data can be continuously collected, and the multivariate dynamic response model can be automatically updated according to the feedback information; when the prediction error exceeds the threshold, the model re-identification and parameter re-optimization are automatically triggered, forming a closed-loop control mechanism of "online monitoring-dynamic modeling-parameter optimization-feedback correction". The feasible operation interval is solved through the quadratic programming, it is ensured that the adjustment range of the control variable is always within the safety boundary of the equipment, and abnormal working conditions such as excessive opening and closing of the valve or overheating of the reboiler are prevented. BRIEF DESCRIPTION OF DRAWINGS

[0050] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are for purposes of illustration only and are not intended to limit the present application thereto. Moreover, the use of the same reference symbols in different drawings indicates similar or identical items.

[0051] Figure 1 The flowchart of the catalytic cracking desorption tower interval optimization control method provided for the embodiments of the present application;

[0052] Figure 2 The functional block diagram of the catalytic cracking desorption tower interval optimization control system provided for the embodiments of the present application. DETAILED DESCRIPTION

[0053] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0054] Existing control technologies for catalytic cracking desorption towers primarily rely on traditional PID control. While this control strategy is simple in structure and easy to implement, it has significant limitations under complex operating conditions. PID control is highly dependent on model parameters; when operating conditions fluctuate or feedstock properties change, fixed parameters struggle to adapt in real time to the system's nonlinear and time-varying characteristics, easily leading to control lag or oscillations. PID control lacks adaptive and predictive capabilities, failing to coordinate and optimize the coupling relationships between multiple variables such as temperature, pressure, and component concentration within the tower, often requiring frequent manual intervention to maintain stable operation.

[0055] For example, during the operation of a catalytic cracking unit, if the properties of the feed oil suddenly change, the temperature at the top of the column will rise rapidly. Traditional PID controllers, failing to detect these changes in oil properties in a timely manner, cannot adjust their fixed proportional coefficient and integral / derivative parameters immediately. This results in a lag in the control signal response, causing the top temperature to briefly exceed the safe range. This not only affects product distribution but may also lead to equipment overheating. In such cases, operators typically need to manually correct the control parameters or switch operating modes to restore stability, increasing the operational burden and posing safety hazards.

[0056] For this, please refer to Figure 1 As shown, a method for optimizing the control of the desorption tower section in catalytic cracking includes:

[0057] S100: Based on historical operating data and the mechanism of catalytic cracking process, a multivariate dynamic response model is constructed. The multivariate dynamic response model describes the dynamic response relationship between the control variable and the controlled variable.

[0058] S200: Based on the target control interval of the controlled variable and the operational constraint interval of the control variable, and based on solving the quadratic programming problem, calculate the feasible operational interval of the control variable that satisfies the target control interval, and restrict the adjustment of the control variable according to the feasible operational interval;

[0059] S300: A dynamic matrix control interval controller is constructed based on a multivariable dynamic response model. The dynamic matrix control interval controller calculates the sum of the weighted time integral and the deviation amplitude integral of the controlled variable outside the target control interval as the interval control performance index.

[0060] S400: Based on the Bayesian optimization algorithm, and taking the minimum of the interval control performance index as the goal, the prediction time domain, control time domain and control weight coefficient of the dynamic matrix control interval controller are optimized to obtain the controller parameter combination that makes the controlled variable stable in the target control interval; and the controller parameter combination is applied to the dynamic matrix control interval controller;

[0061] S500: Based on the OPC communication protocol, the measured value of the controlled variable is obtained in real time and the set value of the control variable is calculated;

[0062] S600: Real-time collection of actual operation data of the desorption tower as process feedback data, and updating of the multivariate dynamic response model according to the process feedback data and re-execution of the Bayesian optimization algorithm to obtain an updated dynamic matrix control interval controller.

[0063] Specifically, a multivariate dynamic response model is established based on historical operation data and catalytic cracking mechanism. The historical data is first denoised (e.g. low-pass filtering or wavelet denoising), outlier points are removed (based on statistical threshold or isolated point detection), and standardized, then the desorption tower simplified mechanism is constructed combined with mass conservation and energy conservation, and recursive parameter identification (such as recursive least squares with forgetting factor, forgetting factor is 0.98-0.999range) to obtain a multi-input multi-output (MIMO) dynamic response matrix that can describe the step and time-domain responses of the controlled variables such as column liquid level and column bottom temperature to the control variables such as valve opening and reboiler temperature; a quadratic programming (QP) problem is established according to the target control interval of the controlled variable and the upper and lower limits and action rate constraints of each actuator: the objective function is the sum of the squares of the control increments plus a penalty term related to the deviation of the controlled variable (or energy consumption / operation cost regularization term), and the constraints include that the controlled variable needs to be kept in the target interval, the control variable and its rate of change are bounded, etc. An efficient solver (such as active-set or real-time QP solver based on interior point method) is used to solve in real time to obtain the feasible operation interval of the control variable that meets the target interval and is used to project / limit the control output; an interval controller in the form of dynamic matrix control (DMC) is constructed based on the above multivariable dynamic response matrix: the prediction matrix is composed of step response coefficients, the prediction time domain and control time domain are set, and the interval control objective function is constructed: when the controlled variable is within the target interval, no penalty is counted, and when it exceeds the interval, the deviation amplitude integral and weighted time integral are calculated according to the deviation and the length of time exceeded respectively and combined according to the design weight to obtain the interval control performance index for evaluating the performance of the controller; the interval control performance index is introduced as the optimization target to introduce Bayesian optimization (BO): Gaussian process is used as the surrogate model, expected improvement (EI) or probability improvement (PI) is used as the acquisition function, a number of samples (for example, 5-20 groups) are initialized using Latin hypercube or historical test points, and the optimal combination of prediction time domain, control time domain and control weight is iteratively searched under the limited budget (for example, maximum iteration 50 times or acquisition function threshold convergence), and finally the optimal parameter group is issued and applied to the DMC interval controller; the controller establishes a secure connection through OPC (supporting subscription and event-driven mode) to obtain the measurement values of liquid level, column bottom temperature and the like in real time, uses time stamp synchronization, missing value interpolation and moving average / Kalman filter and the like to filter and detect abnormalities, and inputs the processed data into the controller to calculate the control variable set value, while the output is projected, rate limited and anti-integrated to protect the safety of the actuator; the desorption tower operation data is continuously collected as process feedback, the model prediction error (such as normalized root mean square error, NRMSE) is calculated according to the rolling window (for example, the last N hours or N samples), and when the error exceeds the preset threshold (for example, more than 5% of the historical baseline or more than 3 times the mean square deviation), the recursive identification is automatically triggered to update the model parameters and reevaluate the interval control performance index, and then the Bayesian optimization is called to re-optimize the controller parameters under the limited budget and smoothly switch to the updated control strategy; to ensure adaptive, constraint-compliant and interval-optimal control of the desorption tower under the premise of ensuring safety and real-time performance.

[0064] wherein the multivariable dynamic response model is represented in discrete state space as:

[0065] ;

[0066] where x(k)∈R^n is the state vector, u(k)∈R^r is the control variable vector, y(k)∈R^m is the controlled variable vector, A∈R^(n×n) is the system state matrix, B∈R^(n×r) is the input matrix, C∈R^(m×n) is the output matrix, and D∈R^(m×r) is the direct transmission matrix;

[0067] The multivariable dynamic response model is parameterized by recursive least squares, and the objective function is:

[0068] ;

[0069] where Θ is the model parameter vector, y(k) is the actual measurement value, is the model prediction value, and λ is the regularization coefficient. The prediction error is minimized and an L2 regularization term is added to prevent overfitting;

[0070] The dynamic matrix control interval controller uses a step response model to construct the prediction equation:

[0071] ;

[0072] where Y(k+1|k)∈R^(m×P) is the predicted value of the controlled variable at the next P time points, is the predicted value of the controlled variable without control effect, G∈R^(mP×rM) is the dynamic matrix composed of the step response coefficients of the multivariable dynamic response model, and ΔU(k)∈R^(r×M) is the control increment at the next M time points;

[0073] The interval control performance index of the dynamic matrix control interval controller is defined as:

[0074] ;

[0075] where y(t) is the controlled variable, y_high and y_low are the upper and lower limits of the target control interval, and w1 and w2 are the weight coefficients;

[0076] Based on the Bayesian optimization algorithm, the prediction time domain P, control time domain M, and control weight coefficient λ of the dynamic matrix control interval controller are parameterized to minimize the interval control performance index. The Bayesian optimization algorithm uses a Gaussian process to establish a proxy model for the objective function J(θ):

[0077] ;

[0078] where θ = [P, M, λ] is the controller parameter vector, μ(θ) is the mean function, and k(θ, θ') is the covariance function.

[0079] The next evaluation point is determined by maximizing the expected improvement acquisition function:

[0080]

[0081] where J_min is the minimum performance index found so far, μ(θ) and σ(θ) are the mean and standard deviation of the Gaussian process at θ, Φ(·) and φ(·) are the cumulative distribution function and probability density function of the standard normal distribution, respectively, and the optimal controller parameter combination is obtained by iterative optimization.

[0082] where the target control interval of the controlled variable refers to the numerical range that the controlled variable (such as the liquid level of the desorption tower, the bottom temperature, etc.) is expected to maintain in a safe and efficient operation, which is usually set according to process requirements, equipment bearing capacity, and product quality standards to ensure stable operation and avoid overrunning. The operating constraint interval of the controlled variable refers to the minimum and maximum range of the control device (such as the opening of the flow valve, the opening of the temperature control valve, etc.) allowed in actual operation, which needs to consider factors such as equipment mechanical limit, safety specification, and energy consumption optimization.

[0083] where the minimization of the interval control performance index is the goal, which means that the degree and duration of the controlled variable deviating from the target interval are minimized through the control strategy, and the weighted integral of the deviation amplitude and deviation time is considered to achieve the optimal precision and maximum efficiency of process control.

[0084] where the setting rules of the prediction time domain, the control time domain, and the control weight coefficient are mainly determined according to the prediction ability of the multivariate dynamic response model and the dynamic characteristics of the control object: the prediction time domain is used to determine the response prediction time length of the future controlled variable, the control time domain is used to set the adjustment frequency or step length of the control action, and the control weight coefficient is used to balance the amplitude of the controlled variable adjustment and the priority of the controlled variable response, so as to achieve the optimal balance between response speed and stability.

[0085] ​​The working principle and process of the present application are as follows: through collecting and preprocessing long-term operation data of a catalytic cracking device, combining the material and energy balance mechanism of the catalytic cracking desorption tower, a multivariate dynamic response model is established. The model takes the tower bottom temperature, tower top pressure, tower kettle liquid level and other key process parameters as controlled variables, takes the reboiler steam flow, condensate flow, valve opening and the like as control variables, and describes the dynamic coupling relationship and time delay characteristics between the two. The model can be constructed by using recursive least squares method, subspace identification or ARX / ARMAX structure model, and through fitting the historical disturbance and response data, a multi-input multi-output dynamic response matrix which can accurately reflect the dynamic characteristics is obtained, providing a basis for subsequent prediction and control. According to the production operation requirements and safety specifications, the target control interval of each controlled variable is determined (such as the tower bottom temperature is controlled within a certain process set range), and the operation constraint interval of the control variable is set in combination with the equipment capacity and safety boundary. Subsequently, by constructing and solving a quadratic programming (QP) problem, the feasible operation interval of the control variable satisfying the target control interval constraint is calculated. The objective function of QP usually includes two parts of minimum deviation of controlled variable and minimum control action, and the constraint condition ensures that the operation is carried out within the safety and equipment limit. The feasible operation interval obtained by solving is used to constrain the adjustment range of the control variable in real time, preventing excessive adjustment or equipment over-limit operation. A dynamic matrix control (DMC) interval controller is constructed based on the foregoing dynamic response model. The controller predicts the trend of the controlled variable at future multiple sampling times using step response characteristics, and calculates the deviation value of the controlled variable outside the target control interval. By weighting and integrating the deviation amplitude and its duration, an interval control performance index is formed, which can comprehensively reflect the comprehensive control effect of the controller in terms of steady-state deviation, dynamic response speed and over-limit time, providing a quantitative evaluation basis for subsequent controller parameter optimization. The Bayesian optimization algorithm is used to intelligently optimize the key parameters of the dynamic matrix controller with the goal of minimizing the interval control performance index. Bayesian optimization describes the mapping relationship between control performance and parameters by establishing a prediction model (such as Gaussian process regression), and dynamically selects new sampling points based on the expected improvement (EI) criterion, so as to find the optimal combination of prediction time domain, control time domain and control weight coefficient in fewer experimental times. The combination of controller parameters obtained after optimization can make the controlled variables of the desorption tower quickly converge to the target interval, and significantly reduce the fluctuation amplitude and energy consumption. Through the real-time data interface based on OPC (OLE for Process Control) communication protocol, the measurement values of the field sensors are continuously collected and transmitted to the control system. The control system calculates the set value of the control variable in real time according to the latest process state and model prediction result, and adjusts through the control valve or actuator. The process has periodic updating, data checking and abnormal detection functions, which can ensure the stability and reliability of the control signal transmission.The running data of the desorption tower, including temperature, pressure, flow rate and other key process parameters, are continuously collected as process feedback information for adaptive updating of the model and the controller. When changes in dynamic characteristics are monitored (such as catalyst activity attenuation or material composition fluctuation), the algorithm will re-identify the multivariate dynamic response model, and based on the latest data, re-execute the Bayesian optimization process to obtain an updated combination of controller parameters, and realize online adaptive adjustment.

[0086] As a preferred embodiment, the scheme of the application is implemented as follows: during the operation of the desorption tower, which is mainly used for desorbing the adsorbed light hydrocarbons and water from the catalyst, the tower bottom temperature, the tower top pressure and the steam flow are the key control parameters affecting the catalyst regeneration quality and energy balance. The historical operation data of the device under different loads and different raw material conditions are collected, including the tower bottom temperature, the tower top pressure, the feed flow, the steam flow, the condensate temperature, etc., and combined with the thermodynamics and mass transfer mechanism of the catalytic cracking desorption process, a multivariate dynamic response model is established. The model describes the dynamic influence law of the steam flow (control variable) change on the tower bottom temperature and the tower top pressure (controlled variable), thereby reflecting the multi-input multi-output characteristics and the coupling relationship between the variables. The operator sets the target control interval of the controlled variable, such as controlling the tower bottom temperature between 250-260℃ to ensure that the residual carbon content in the catalyst is controlled within a reasonable range; according to the equipment safety boundary, the steam flow is constrained within the range of 10-18t / h. By establishing a quadratic programming model, the feasible operation interval of the steam flow is solved under the condition of meeting the target temperature interval requirement, and the control signal is constrained to avoid the increase of energy consumption or equipment impact caused by excessive adjustment. A dynamic matrix control (DMC) interval controller is constructed based on the dynamic model. The controller calculates the predicted trajectory of the tower bottom temperature in the future multiple sampling periods, evaluates the weighted time integral and deviation amplitude integral of the temperature deviation from the target interval, and uses them as the interval control performance index. This index can reflect the comprehensive performance of the controller in terms of temperature deviation elimination speed, out-of-bound amplitude and stability. The Bayesian optimization algorithm is used to automatically optimize the key parameters of the controller, such as prediction time domain, control time domain and weight coefficient. The optimization goal is to minimize the interval control performance index, that is, to achieve the optimal control performance of the desorption tower temperature. For example, in a certain optimization process, it is found that prolonging the prediction time domain and increasing the temperature deviation penalty weight can more quickly suppress temperature fluctuations, so the controller parameters are automatically updated. After optimization, the tower bottom temperature deviation is reduced from ±4℃ to ±1℃, and the tower top pressure fluctuation is reduced by 30%, and the control quality is significantly improved. Through the OPC communication protocol, real-time data interaction with the DCS (Distributed Control System) is realized, and the field measurement signals such as temperature and pressure are continuously collected, and the set value of the steam flow is calculated in real time. The actuator automatically adjusts the steam valve opening according to the set value, realizing the closed-loop precise control of the desorption tower. The running data of the device are continuously collected, and the tower bottom temperature, steam flow and energy consumption changes are monitored. When the change of raw material properties or equipment aging causes the dynamic characteristics of the system to deviate, the algorithm will automatically re-identify the multivariate dynamic response model and re-execute the Bayesian optimization process to obtain a new combination of controller parameters, thereby realizing adaptive control. For example, when the raw material is switched from high-sulfur heavy oil to light distillate oil, the change of heat balance in the tower is automatically identified, and the control parameters are adjusted to ensure that the tower bottom temperature remains stable within the set interval.

[0087] By the technical solution, the multivariable dynamic response model constructed based on historical operation data and a catalytic cracking process mechanism can accurately describe the dynamic coupling relationship between control variables and controlled variables, thereby improving the accuracy of control prediction and avoiding the lag and overshoot phenomenon existing in traditional single variable or fixed parameter control methods; by solving a quadratic programming problem to calculate the feasible operation interval of the control variable and limiting the adjustment range, the increase in energy consumption, equipment impact or operation abnormality caused by excessive operation in the control process is prevented, and the safety and stability are improved; the dynamic matrix control interval controller combines the interval performance index of weighted time integral and deviation amplitude integral, so that the controller can comprehensively consider the amplitude and duration of the controlled variable deviating from the target interval, and accurate adjustment of key indicators such as temperature, flow and liquid level is realized; the adaptive parameter optimization based on the Bayesian optimization algorithm can dynamically adjust the prediction time domain, control time domain and control weight coefficient under different working conditions, so that the controlled variable is always stable in the target interval, and the adaptability and control robustness are improved; the OPC communication protocol is used to realize real-time data interaction with the field control system, so that the fast response and accurate execution of the control instruction are ensured; by collecting process feedback data in real time and updating the multivariable dynamic response model, and iteratively updating the controller parameters combined with the Bayesian optimization algorithm, self-learning and self-optimization of the control strategy can be realized.

[0088] The application further proposes that the control variables include the desethanized gasoline out of the desorption tower flow valve opening, the desorption tower bottom reboiler one shell temperature three-way valve opening and the condensed oil into the desorption tower flow valve opening; the controlled variables include the desorption tower liquid level and the desorption tower bottom temperature.

[0089] Specifically, the control variables mainly include the desethanized gasoline out of the desorption tower flow valve opening, the desorption tower bottom reboiler one shell temperature three-way valve opening and the condensed oil into the desorption tower flow valve opening. Among them, the opening of the desethanized gasoline flow valve directly adjusts the flow of the vapor phase product at the top of the desorption tower, thereby affecting the liquid load in the tower and the tower top pressure; the desorption tower bottom reboiler one shell temperature three-way valve opening is used to adjust the heating energy at the tower bottom, and the temperature balance in the tower is maintained by controlling the reflux temperature and the tower bottom temperature; the condensed oil into the desorption tower flow valve opening controls the liquid reflux amount in the tower, and affects the liquid level height and the mass transfer efficiency. The corresponding controlled variables include the desorption tower liquid level and the tower bottom temperature, wherein the liquid level reflects the liquid phase load in the tower and the operation safety, and the tower bottom temperature directly affects the desethanization effect and the heat balance in the tower.

[0090] The application further proposes that when the multivariable dynamic response model is constructed based on historical operation data and a catalytic cracking process mechanism, the following steps are included:

[0091] The historical operation data is preprocessed, and the preprocessing includes removing signal noise, eliminating abnormal data points and standardizing processing;

[0092] The catalytic cracking process mechanism includes the mass conservation principle and the energy conservation principle, and a simplified mechanism model of the desorption tower is established according to the catalytic cracking process mechanism;

[0093] The simplified mechanism model and the pretreated historical operation data are combined, and a recursive parameter identification method such as recursive least squares or recursive Bayesian estimation is used to identify the multivariable dynamic response parameters.

[0094] Specifically, when constructing the multivariable dynamic response model based on historical operation data and the catalytic cracking process mechanism, the collected historical operation data is systematically pretreated to ensure the accuracy and reliability of model parameter identification. The pretreatment steps include removing signal noise, eliminating random noise interference in the collection process through filtering algorithms such as low-pass filtering or wavelet denoising; eliminating abnormal data points, identifying and eliminating extreme values caused by sensor failure or operation anomalies during the collection process to avoid misleading model training; and standardizing the data to unify different dimensions and orders of magnitude to a comparable numerical range, thereby improving the convergence speed and stability of subsequent parameter identification. According to the mechanism characteristics of the catalytic cracking process, a simplified mechanism model of the desorption tower is established. The mechanism model is based on the mass conservation principle to ensure that the flow and composition of each phase material (vapor phase, liquid phase) in the tower meet the conservation relationship; based on the energy conservation principle, the heat balance and heat transfer process of each operating point in the tower are described to ensure that heat input, heat output and heat accumulation are accurately represented in the model. The simplified mechanism model considers key in-tower process parameters (such as tower top reflux ratio, tower bottom temperature, tower liquid level, etc.) to reduce computational complexity while still accurately reflecting the dynamic characteristics of the tower, providing a basis for subsequent multivariable modeling. The above simplified mechanism model and pretreated historical operation data are combined, and a recursive parameter identification method such as recursive least squares or recursive Bayesian estimation is used to identify the multivariable dynamic response parameters. The recursive parameter identification method can dynamically adjust the model parameters in the process of updating the historical data step by step, so that the model can continuously adapt to changes in operating conditions. The multivariable dynamic response parameters determined by this method can completely describe the dynamic response relationship of the control variables (such as valve opening, heating power) to the controlled variables (such as tower liquid level, tower bottom temperature) in time and amplitude, thereby constructing an accurate and online controllable multivariable dynamic response model.

[0095] As a preferred embodiment, the scheme of the present application is implemented as follows: in order to optimize the operation of the desorption tower in a catalytic cracking unit of a certain refinery, the operation data of the past year are collected, including the key parameters such as the tower top gasoline flow, the tower bottom liquid level, the tower bottom temperature, the reflux rate, and the valve opening. For these historical data, first, preprocessing is performed: through low-pass filtering, the high-frequency noise of the flow sensor and the temperature sensor is removed, and at the same time, extreme data points generated due to equipment maintenance or abnormal operation are eliminated, for example, instantaneous data deviation caused by abnormal increase of the tower bottom temperature on a certain day; then, standardization processing is performed on various variables, so as to unify the flow (unit: ton / hour), the temperature (unit: ℃), and the valve opening (unit: %) to the dimension range of 0 to 1, so as to be used in subsequent model training. According to the catalytic cracking process mechanism, a simplified mechanism model of the desorption tower is established. The model is based on the mass conservation principle, describes the flow balance relationship of the liquid phase and the vapor phase materials in the tower, and ensures that the in-out material quantity of each section of the tower plate or the filler layer is consistent with the reaction product quantity; based on the energy conservation principle, the heat balance in the tower is simulated, including the heat provided by the tower bottom reboiler, the condensation and reflux heat at the tower top, and the influence of the endothermic reaction in the tower on the temperature. For example, the change of the heating power of the tower bottom reboiler will cause the change of the tower bottom temperature and the liquid level, and the dynamic response of the temperature is calculated through the energy balance formula. The simplified mechanism model is combined with the preprocessed historical data, and a recursive parameter identification method (such as recursive least squares) is used to determine the multivariate dynamic response parameters. In actual operation, when the tower bottom valve opening increases by 1%, the model can predict that the liquid level increases by 0.5 meters and the tower bottom temperature increases by 2℃, while considering the dynamic response of the tower top gasoline flow. This recursive updating method can continuously adjust the parameters over time, so that the model can reflect the influence of different load conditions, changes in raw material properties, and seasonal temperature fluctuations on the dynamic behavior in the tower. The constructed multivariate dynamic response model can describe the dynamic response relationship of multiple control variables to the controlled variable, and provide accurate prediction basis for online optimization control and interval control strategy.

[0096] Through the above technical scheme, the present application pre-processes the historical operation data, including removing sensor noise, eliminating abnormal data points, and standardization processing, which can ensure the accuracy and consistency of the input data, and avoid model errors caused by data abnormalities or dimension differences. The simplified mechanism model of the desorption tower is established by using the mass conservation principle and the energy conservation principle of catalytic cracking, so that the model can truly reflect the dynamic distribution law of the materials and energy in the tower, thereby realizing accurate prediction of the key controlled variables. The simplified mechanism model is combined with the preprocessed historical data to determine the multivariate dynamic response parameters by using the recursive parameter identification method, so that the model can be updated in real time and adapt to different load conditions, changes in raw material properties, and seasonal temperature fluctuations, thereby improving the adaptability and prediction accuracy of the model to actual working conditions.

[0097] The application further proposes a target control interval of the controlled variable and an operation constraint interval of the control variable, and includes the following when calculating a feasible operation interval of the control variable satisfying the target control interval based on solving a quadratic programming problem:

[0098] presetting a target control interval of the controlled variable and an operation constraint interval of the control variable;

[0099] constructing a quadratic programming problem objective function, which is determined according to the size of the control variable adjustment amount and the control effect;

[0100] setting a quadratic programming problem constraint condition, which includes that the controlled variable must be kept in the target control interval and the control variable change cannot exceed the operation constraint interval;

[0101] solving the quadratic programming problem by using an active set method to obtain an actual safe operation interval of the control variable, which is a feasible operation interval of the control variable satisfying the target control interval.

[0102] Specifically, according to the target control interval of the controlled variable and the operation constraint interval of the control variable, first, the interval ranges of each controlled variable and control variable need to be preset. Taking a desorption tower as an example, the controlled variables such as the tower liquid level can be set in the range of 1.5-2.0 meters, and the tower bottom temperature can be set in the range of 250-260 DEG C; the control variables such as the de-ethane gasoline flow valve opening, the heavy boiler one shell temperature three-way valve opening and the condensed oil into the tower flow valve opening are set with minimum and maximum values according to the equipment design and operation experience to ensure the operation safety. After the interval ranges are determined, the objective function of the quadratic programming problem is constructed, which considers the control variable adjustment amplitude and the control effect. The objective function forms a comprehensive evaluation index by squaring and weighting the adjustment amplitude of each control variable within a certain time and combining its contribution to the deviation of the controlled variable from the target interval, aiming to achieve fast and effective control and minimize the energy consumption or equipment wear caused by excessive adjustment. The constraint condition of the quadratic programming problem needs to be set, mainly including two aspects: one is that the controlled variable must be kept in the target control interval to ensure that the tower liquid level and the tower bottom temperature and other key parameters run in the safe and efficient operation interval; the other is that the change amount of the control variable cannot exceed the operation constraint interval to prevent the valve or pump from moving beyond the mechanical limit or causing system instability. In order to solve the above quadratic programming problem, the application adopts the active set method. By iteratively judging whether the constraint condition is activated, the active constraint set is dynamically updated, so as to quickly converge to the optimal solution satisfying all constraint conditions. The final result is the actual safe operation interval of the control variable, that is, the range in which the control variable can be safely adjusted under the premise of ensuring that the controlled variable is stable in the target interval.

[0103] As a preferred embodiment, the scheme of the present application is implemented as follows: taking the desorption tower of a catalytic cracking unit as an example, the target control interval of the controlled variable and the operation constraint interval of the control variable are preset. The target control interval of the desorption tower liquid level can be set to 1.6-1.9 meters, and the target control interval of the tower bottom temperature is set to 255-260℃, so as to ensure the stable operation of the desorption tower and the product quality. The control variable such as the ethane removal gasoline flow valve opening can be limited to 20%-80%, the heavy boiler one-shell temperature three-way valve opening is limited to 10%-70%, and the condensed oil into tower flow valve opening is limited to 15%-75%, which ensures that the valves and equipment are operated within a safe range and avoids overload or mechanical damage. Based on these settings, a quadratic programming problem is constructed. The objective function of the quadratic programming problem comprehensively considers the adjustment range of the control variable and its influence on the deviation of the controlled variable from the target interval. For example, if the tower liquid level is too high, the liquid level can be pulled back to the target interval by appropriately adjusting the ethane removal gasoline flow valve and the condensed oil into tower flow valve, and the objective function will be weighted and punished for the large-scale opening of the valve, so as to prevent frequent adjustment from causing equipment wear or increased energy consumption. When setting the constraint conditions, it is strictly ensured that the controlled variable does not exceed the target control interval, and the adjustment range of the control variable does not exceed the preset operation constraint interval. By solving the quadratic programming problem by using the effective set method, it is iteratively judged which constraint conditions are in the active state, and the actual feasible operation interval of each control variable is dynamically determined on the premise of ensuring safety and control targets. For example, at a certain time point, it is calculated that the ethane removal gasoline flow valve can be safely adjusted in the range of 35%-60%, the three-way valve in the range of 20%-55%, and the condensed oil into tower valve in the range of 30%-50%, which are the actual safe operation intervals.

[0104] Through the above technical scheme, the present application ensures that the desorption tower liquid level and the tower bottom temperature fluctuate minimally within the target interval, improves the operation stability and process consistency; the actual safe operation interval provides reliable constraints for the subsequent dynamic matrix controller, so that the controller will not appear out-of-limit operation in the execution process.

[0105] The present application further proposes that when a dynamic matrix control interval controller is constructed based on a multivariate dynamic response model, the following steps are included:

[0106] The step response coefficients between the control variable and the controlled variable are calculated based on the multivariate dynamic response model, and a prediction model is constructed according to the step response coefficients to predict the trend of the controlled variable in the future time domain;

[0107] The interval control objective function is set according to that no penalty is generated when the controlled variable is within the target control interval, and corresponding penalty terms are generated according to the deviation and duration when the controlled variable exceeds the target control interval;

[0108] According to the requirement that the controlled variable must be maintained in the target control interval and the limitation that the control variable must meet the operating constraint range, an interval constraint condition is set;

[0109] The prediction model, the interval control objective function and the interval constraint condition are integrated to form a dynamic matrix control interval controller.

[0110] Specifically, according to the multivariable dynamic response model, step response coefficients between the control variable and the controlled variable are calculated. These step response coefficients can describe the response characteristics of the controlled variable at different time steps when the control variable is adjusted by one unit, thereby establishing a prediction model of the change trend of the controlled variable in the future time domain. Through the prediction model, the fluctuation of key controlled variables such as liquid level and tower bottom temperature in future operation can be estimated in advance, providing a basis for subsequent control decisions. When constructing the interval control objective function, whether the controlled variable is in the target control interval is taken as the core judgment standard. If the controlled variable remains in the target control interval, no penalty term is generated, and the controller will not make additional intervention to the current operation; if the controlled variable exceeds the target interval, a penalty term is calculated according to the exceeding amplitude and duration. This design ensures the sensitivity of the controller to the deviation, which can timely adjust the control variable for correction, so as to quickly pull the controlled variable back to the target interval. The deviation amplitude is considered in the objective function, and the deviation duration is also comprehensively considered, so that the controller can balance between control accuracy and stability. The setting of the interval constraint condition is an important link to ensure the control safety and equipment reliability. The controlled variable must be strictly maintained in the predetermined target interval to avoid unsafe states such as liquid overflow, low liquid level or tower bottom overheating, and at the same time, the control variable must meet the operating constraint range to ensure that the valve opening and temperature control operation are in the safe interval allowed by the machinery and process. By integrating the prediction model, the interval control objective function and the constraint condition, a dynamic matrix control interval controller is formed. The controller can dynamically adjust the control variable in the future prediction time domain to minimize the risk of the controlled variable deviating from the target interval, and realize accurate, stable and safe interval control of the liquid level and tower bottom temperature of the desorption tower.

[0111] As a preferred embodiment, the scheme of the present application is implemented as follows: for example, in the control of the liquid level of a desorption tower in a certain refinery, a dynamic matrix control interval controller (DMC-IC) based on a multivariate dynamic response model can accurately manage the liquid level and the bottom temperature of the tower. When the controller is running, the step response coefficients between the control variables (such as the desethanized gasoline outflow valve opening degree of the desorption tower, the shell pass temperature three-way valve opening degree of the desorption tower bottom reboiler, and the condensed oil inflow valve opening degree of the desorption tower) and the controlled variables (the liquid level and the bottom temperature) are calculated using historical operation data and mechanism models. Through these step response coefficients, the controller can predict the trend of the changes of the liquid level and the bottom temperature in the future time domain, for example, estimate that the liquid level may rise from 1.7 meters to 1.95 meters in the next ten minutes. On this basis, the controller establishes an interval control objective function: when the liquid level is maintained within the target interval of 1.6 meters to 1.9 meters, no penalty term is generated; if the liquid level exceeds the target interval, for example, reaches 1.95 meters, a penalty value is calculated according to the exceeding amplitude (0.05 meters) and the duration. The objective function not only considers the deviation amplitude, but also combines the deviation duration, so that the controller can make slight adjustments for short-term fluctuations and take more active control measures for persistent deviations. The interval constraint conditions ensure the safety and reliability of the control operation: the liquid level must be kept between 1.6 meters and 1.9 meters, the bottom temperature of the tower must be controlled within the range of 280-300°C, and the valve opening degrees of the control variables must be within the operating limits allowed by the equipment. When the predicted liquid level exceeds the upper limit, the controller will reduce the desethanized gasoline flow valve opening degree in advance, while adjusting the condensed oil inflow valve and the reboiler temperature three-way valve opening degree, to ensure that the liquid level returns smoothly to the safe interval. Through the integration of the prediction model, the interval objective function and the constraint conditions, the dynamic matrix control interval controller can realize continuous feedforward-feedback combined control, so that the controlled variables remain in a safe and stable interval during the entire operation process.

[0112] Through the above technical scheme, by calculating the step response coefficients between the control variables and the controlled variables, the controller can accurately predict the trend of the changes of the liquid level and the bottom temperature in the future time domain, realizing prospective regulation. The interval control objective function quantitatively punishes the situation of exceeding the target interval by combining the deviation amplitude and the deviation duration, so that the controller can not only respond to instantaneous fluctuations, but also take active adjustment measures for persistent deviations, ensuring that the controlled variables are stable within the set interval. The interval constraint conditions ensure the safety and feasibility of the operation of the control variables.

[0113] The present application further proposes that when the dynamic matrix control interval controller calculates the sum of the weighted time integral and the deviation amplitude integral of the controlled variable outside the target control interval as the interval control performance index, it includes:

[0114] determining the deviation of the controlled variable from the target control interval based on the interval control objective function;

[0115] According to the length of time that the controlled variable exceeds the target control interval and the amount of deviation thereof, a weighted time integral is calculated;

[0116] According to the amount of deviation of the controlled variable from the target control interval, a deviation magnitude integral is calculated;

[0117] The weighted time integral and the deviation magnitude integral are proportionally combined to form an interval control performance index.

[0118] Specifically, in the process of calculating the interval control performance index by the dynamic matrix control interval controller, first, the amount of deviation of the controlled variable (such as the liquid level of the desorption tower or the bottom temperature of the tower) when exceeding the target control interval is determined by the interval control target function. This amount of deviation not only reflects the difference between the controlled variable and the set interval, but also provides the basis data required for subsequent integral calculation. For example, when the liquid level is 0.5 meters higher than the upper limit, the amount of deviation is 0.5 meters; when the bottom temperature of the tower is 2°C lower than the lower limit, the amount of deviation is 2°C. According to the duration of the controlled variable deviating from the target interval and the deviation magnitude, a weighted time integral (WTI) is calculated. This integral is accumulated by multiplying the deviation at each time point by the time weight, thereby reflecting the persistence and severity of the deviation. For example, if the liquid level exceeds the upper limit for 10 minutes, the deviation of each minute is 0.2 meters, 0.3 meters, 0.4 meters, etc., then the weighted time integral is calculated according to the deviation and the time length of each minute, emphasizing the influence of long-term deviation on control performance. The deviation magnitude integral (DMI) is only accumulated according to the magnitude of deviation from the target interval, without considering the time factor. It is mainly used to evaluate the magnitude of deviation in the short term, to ensure that the controller can still maintain the controlled variable in the safe interval when facing rapid fluctuations. For example, the bottom temperature of the tower instantaneously decreases by 3°C, and this deviation is immediately included in the deviation magnitude integral, prompting the controller to respond quickly. The weighted time integral and the deviation magnitude integral are combined according to a predetermined proportion to form a comprehensive interval control performance index. This index comprehensively reflects the magnitude and duration of the deviation, providing a quantitative basis for controller optimization.

[0119] As a preferred embodiment, the scheme of the application is implemented as follows: assuming that the target level control interval of the desorption tower is 3.0-3.5 meters, and the target control interval of the tower bottom temperature is 280-285℃. During an operation, the liquid level exceeds the upper limit to 3.7 meters due to feed fluctuation, and lasts for 5 minutes, while the tower bottom temperature drops to 277℃ for a short time due to the reboiler load change, and lasts for 2 minutes. First, the deviation amount of the liquid level exceeding the upper limit is determined as 0.2 meters, and the deviation amount of the tower bottom temperature below the lower limit is determined as 3℃, respectively, by using the interval control target function. The weighted time integral (WTI) is calculated according to the deviation time and amplitude: the liquid level WTI is 0.2×5=1 meter·minute, since the liquid level deviation of 0.2 meters lasts for 5 minutes; and the tower bottom temperature WTI is 3×2=6℃·minute, since the tower bottom temperature deviation of 3℃ lasts for 2 minutes. The integral reflects the degree of the deviation existing for a long time, so that the controller gives a higher weight to the long-term deviation. At the same time, the deviation amplitude integral (DMI) is calculated only according to the deviation amplitude: the liquid level DMI is 0.2 meters, and the tower bottom temperature DMI is 3℃, which are used to quantify the severity of the short-term deviation. The WTI and DMI are combined in a set proportion to form the interval control performance index. For example, if the weight proportion is set as 0.6:0.4, then the liquid level comprehensive index is 1×0.6+0.2×0.4=0.68, and the tower bottom temperature comprehensive index is 6×0.6+3×0.4=4.8. Through the index, the control system can quantify the severity of the controlled variable deviating from the target interval in each operation, thereby providing a clear basis for optimizing the control parameters, adjusting the valve opening or the reboiler heating power.

[0120] Through the above technical scheme, the application identifies the deviation amount of the controlled variable exceeding the target control interval based on the interval control target function. The weighted time integral is calculated according to the deviation amount and the deviation duration, which reflects the influence of the deviation existing for a long time, so that the control system can identify the potential threat of the long-term deviation to the process stability. The deviation amplitude integral is calculated according to the deviation amplitude, which quantifies the severity of the instantaneous deviation, ensuring that the controller can also respond quickly when there is a large fluctuation in the short term. The weighted time integral and the deviation amplitude integral are proportionally combined according to the set weight to form a comprehensive interval control performance index, which reflects the overall deviation of the controlled variable outside the target interval.

[0121] The application further proposes parameter optimization of the prediction time domain, the control time domain and the control weight coefficient of the dynamic matrix control interval controller, to obtain a controller parameter group that stabilizes the controlled variable in the target control interval, including:

[0122] The Bayesian optimization target is set based on the interval control performance index, and the Bayesian optimization target takes minimizing the interval control performance index as the optimization direction;

[0123] The feasible range of the controller parameter combination is determined according to the prediction time domain, the control time domain and the control weight coefficient.

[0124] construct a performance prediction model based on historical evaluation data, the performance prediction model being used to estimate control effects of different controller parameter combinations;

[0125] repeating the parameter evaluation and model updating process until a convergence condition is met; and determining a controller parameter combination that minimizes the interval control performance index.

[0126] Specifically, the prediction horizon, control horizon, and control weight coefficient of the dynamic matrix control interval controller are optimized to obtain a set of optimal controller parameters that can maintain the controlled variables (such as the desorption column liquid level and the column bottom temperature) within the target control interval. The objective function of Bayesian optimization is set based on the aforementioned interval control performance index (the combined value of weighted time integral and deviation amplitude integral) to minimize the performance index, ensuring that the optimization process can prioritize improving the stability and control accuracy of the controlled variables. According to the physical meaning and engineering limitations of the prediction horizon, control horizon, and control weight coefficient, the feasible range of controller parameter combinations is determined in advance, for example, the prediction horizon cannot exceed the lag time of the controlled variable response characteristics, the control horizon should cover appropriate sampling periods, and the control weight coefficient needs to balance the control priorities among multiple variables. A performance prediction model is constructed using historical evaluation data, which refers to actual operating data and corresponding control effect evaluation results collected based on existing controller parameter combinations during the operation of the desorption column or catalytic cracking unit. This model can predict the response effect of the controller under different parameter combinations, including the ability of the controlled variables to maintain the target interval and the smoothness of the control variable adjustment. Based on this, an iterative process of parameter combination evaluation and model updating is repeatedly performed, and the Bayesian optimization gradually converges to the optimal parameter combination. In each iteration, the prediction model is used to estimate the control effect of the new parameter combination, and the combination that is most likely to improve the interval control performance index is selected. When the optimization process meets the convergence condition, a set of optimal controller parameters for the prediction horizon, control horizon, and control weight coefficient is determined, enabling the dynamic matrix control interval controller to minimize the amplitude and duration of the controlled variables deviating from the target control interval during actual operation.

[0127] As a preferred embodiment, the scheme of the present application is implemented as follows: the aforementioned interval control performance index (including the weighted time integral and the deviation amplitude integral of the controlled variable exceeding the target control interval) is taken as the objective function of Bayesian optimization, and the optimization direction is to minimize the index. When the desorption column liquid level fluctuates for a short time or the tower bottom temperature is slightly higher than the set upper limit, the optimization target is to reduce the liquid level fluctuation amplitude and shorten the time for the tower bottom temperature to return to the target interval by adjusting the controller parameters, thereby improving the overall control performance. According to the operating characteristics of the desorption column, the feasible range of the controller parameters is preset: the prediction time domain needs to cover the typical response time of the liquid level and the tower bottom temperature, for example, 10-30 seconds; the control time domain should match the adjustment period of the valve and the reboiler, for example, 5-15 seconds; the control weight coefficient needs to reasonably allocate the priority of each controlled variable in the multivariable control, for example, the liquid level accounts for 70% and the tower bottom temperature accounts for 30%. The performance prediction model is constructed using historical operating data and previous controller evaluation results, the response effect of the controlled variable under different parameter combinations is predicted by simulation, for example, the maximum deviation and the regression time of the liquid level under the operation fluctuation are predicted under a certain parameter combination. Based on the prediction model, the iterative process of parameter combination evaluation and model updating is repeatedly performed, and the parameter combination most likely to improve the interval control performance index is selected for testing and simulation in each iteration. When the Bayesian optimization iteration converges, a set of optimal prediction time domain, control time domain and control weight coefficient combination is determined, so that the liquid level and the tower bottom temperature of the desorption column can be stably maintained within the preset target interval, and the adjustment action of the controlled variables such as the de-ethane gasoline out-of-column flow valve, the reboiler three-way valve and the condensed oil into-column flow valve is smooth and safe.

[0128] Through the above technical scheme, the present application takes the interval control performance index as the objective function of Bayesian optimization, and the optimization direction is to minimize the index, thereby ensuring that the deviation amplitude of the controlled variable in the target control interval is the smallest and the duration is the shortest. Based on the actual running characteristics and equipment response time of the desorption column, the feasible range of the prediction time domain, the control time domain and the control weight coefficient is preset, so that the optimization search space is scientific and reasonable. The performance prediction model is constructed using historical operating data and the performance evaluation results of previous parameter combinations, the control effect that may be produced by different parameter combinations is simulated and predicted, and the optimal candidate combination is selected. By repeatedly performing parameter evaluation and model updating iteration, the optimal solution is gradually approached until the optimization converges, and finally the obtained controller parameter combination can not only stably maintain the liquid level and the tower bottom temperature within the target control interval, but also consider the smoothness and safety of the control action.

[0129] The present application further proposes that when the measurement value of the controlled variable is obtained in real time and the set value of the control variable is calculated based on the OPC communication protocol, it includes:

[0130] The safety communication connection is established based on the OPC communication protocol, measurement values of the desorption tower liquid level and the tower bottom temperature are acquired in real time, and the acquired measurement values are input into a dynamic matrix control interval controller to perform data processing in combination with current controller parameters to calculate a set value of a control variable.

[0131] Specifically, a safe and reliable communication connection is established between the control system and the field instrument or the control terminal through the OPC communication protocol to ensure the real-time and integrity of data transmission, and an encryption or access permission control mechanism is adopted to prevent data from being tampered with or lost; measurement data of key controlled variables of the desorption tower are acquired in real time, including the liquid level height acquired by the liquid level sensor and the tower bottom temperature acquired by the temperature sensor, and the acquired data is processed through filtering, denoising and time synchronization to ensure that the data input into the controller is accurate and reliable; the processed measurement values are input into the dynamic matrix control interval controller in combination with current controller parameters (including a prediction time domain, a control time domain and a control weight coefficient), the controller calculates the change trend of the controlled variable in the future time period by using a prediction model, and determines the optimal set value of each control variable (such as the desethanized gasoline out of the desorption tower flow valve opening, the reboiler three-way valve opening and the condensed oil into the desorption tower flow valve opening) through an interval control algorithm to ensure that the controlled variable is stable in the target control interval; and the calculated control variable set value is issued to the field actuator through the OPC protocol to realize closed-loop control.

[0132] As a preferred embodiment, the scheme of the application is implemented as follows: for example, in the desorption tower control process of a certain catalytic cracking device, a safe communication connection is established with the field controller and the sensor through the OPC communication protocol to ensure the real-time and reliability of data transmission, and the communication channel is encrypted and permission controlled to prevent unauthorized data access. Key controlled variables at the top and bottom of the desorption tower are acquired in real time, wherein the liquid level sensor continuously measures the liquid height in the tower, and the temperature sensor acquires the reboiler temperature at the tower bottom in real time. The acquired original measurement data is first processed through filtering and denoising, and is time-synchronized to eliminate the influence of sensor delay or sudden fluctuation. The processed liquid level and temperature data are input into the dynamic matrix control interval controller in combination with current controller parameters, including a prediction time domain, a control time domain and a control weight coefficient, the controller analyzes the change trend of the controlled variable in the future time period based on a prediction model, and calculates the optimal set value of the desethanized gasoline out of the desorption tower flow valve opening, the reboiler one-shell temperature three-way valve opening and the condensed oil into the desorption tower flow valve opening to keep the liquid level and the tower bottom temperature in the preset target control interval. The calculated control variable set value is issued to the actuator through the OPC protocol to realize closed-loop control.

[0133] By the technical solution, the application establishes a safe OPC communication connection, realizes efficient and reliable data transmission between process control and field sensors and actuators, ensures data integrity and protection, avoids abnormality caused by misoperation or network attack. The measurement values of the desorption tower liquid level and the tower bottom temperature are obtained in real time, the liquid level fluctuation and temperature change in the tower can be sensed in time, so that the abnormal fluctuation in the production process can be responded quickly. The obtained measurement values are input into the dynamic matrix control interval controller, and the data is processed in combination with the current controller parameters, so that the optimal set values of the desorption tower flow valve opening degree, the desorption tower bottom reboiler one-shell temperature three-way valve opening degree and the condensed oil into the desorption tower flow valve opening degree can be calculated accurately, so that the desorption tower liquid level and the tower bottom temperature can be kept stable in the preset target interval.

[0134] The application further proposes that when the multivariate dynamic response model is updated according to the process feedback data and the Bayesian optimization algorithm is re-executed to obtain the updated dynamic matrix control interval controller, the following are included:

[0135] The prediction error is calculated according to the process feedback data and the prediction data of the multivariate dynamic response model; the prediction error is compared with the prediction error threshold; when the prediction error is greater than the prediction error threshold, the recursive parameter identification method is used to update the multivariate dynamic response model; the interval control performance index is re-evaluated by using the updated multivariate dynamic response model; the Bayesian optimization algorithm is executed based on the re-evaluated performance index to obtain the updated controller parameter combination; and the updated controller parameter combination is applied to the dynamic matrix control interval controller.

[0136] Specifically, real-time operation data of the desorption column is collected, including liquid level, column bottom temperature and actual execution state of related control variables. These process feedback data are compared with the predicted data of the previously constructed multivariate dynamic response model, and the prediction error of each controlled variable at the current time is calculated to quantify the deviation between the model prediction and the actual operation. The calculated prediction error is compared with the preset prediction error threshold, and when the error exceeds the threshold, a recursive parameter identification method is triggered to incrementally update the multivariate dynamic response model to dynamically adjust the model parameters so that they more accurately reflect the actual dynamic behavior of the desorption column. The updated multivariate dynamic response model is used to reevaluate the interval control performance indicators, including the deviation amplitude and duration weighted integral of the liquid level and column bottom temperature out of the target control interval, and through this evaluation the effectiveness of the current control strategy can be quantified. Based on the performance indicators reevaluated, the prediction time domain, control time domain and control weight coefficient of the dynamic matrix control interval controller are optimized using the Bayesian optimization algorithm to find the controller parameter combination that can minimize the interval control performance indicators. The optimized controller parameters are applied to the dynamic matrix control interval controller to achieve adaptive updating of the controller, so that the desorption column can still maintain the liquid level and column bottom temperature within the target control interval under complex operating conditions.

[0137] Through the above technical solution, the actual operation data of the desorption column liquid level, column bottom temperature and control variables collected in real time are compared with the predicted values of the multivariate dynamic response model, the prediction error is calculated, and compared with the preset error threshold. When the prediction error exceeds the threshold, a recursive parameter identification method is automatically triggered to update the multivariate dynamic response model, so that the model can more accurately reflect the dynamic change law of the current column materials and heat. The updated model is used to reevaluate the interval control performance indicators, including the amplitude, duration and weighted integral of the controlled variables out of the target control interval, thereby quantifying the effectiveness of the current control strategy. Based on these performance evaluation indicators, the Bayesian optimization algorithm is executed again to optimize the prediction time domain, control time domain and control weight coefficient of the dynamic matrix control interval controller, and obtain the updated controller parameter combination that can minimize the interval control performance indicators. After applying the optimized parameters to the controller, accurate regulation of the liquid level and column bottom temperature is achieved, so that they can be maintained within the target control interval under complex operating conditions, reducing the risk of operation fluctuation, improving production efficiency, and enhancing the response capability to operating condition disturbances and emergencies.

[0138] In another preferred manner based on the above, referring to Figure 2 The embodiment provides a catalytic cracking desorption column interval optimization control system for applying the catalytic cracking desorption column interval optimization control method described above, which comprises:

[0139] The response model construction module is configured to construct a multivariate dynamic response model based on historical operation data and a catalytic cracking process mechanism, the multivariate dynamic response model describing a dynamic response relationship between a control variable and a controlled variable;

[0140] The control variable limiting module is configured to limit adjustment of the control variable according to a target control interval of the controlled variable and an operation constraint interval of the control variable, and based on calculation of a quadratic programming problem to obtain a feasible operation interval of the control variable satisfying the target control interval, and according to the feasible operation interval.

[0141] The controller construction module is configured to construct a dynamic matrix control interval controller based on the multivariate dynamic response model, and calculate a weighted time integral and a deviation amplitude integral of the controlled variable outside the target control interval as an interval control performance index through the dynamic matrix control interval controller.

[0142] The parameter combination acquisition module is configured to perform parameter optimization on a prediction time domain, a control time domain and a control weight coefficient of the dynamic matrix control interval controller based on a Bayesian optimization algorithm and with minimization of the interval control performance index as a target, to obtain a controller parameter combination that stabilizes the controlled variable in the target control interval, and apply the controller parameter combination to the dynamic matrix control interval controller.

[0143] The OPC communication module is configured to obtain a measurement value of the controlled variable and calculate a set value of the control variable in real time based on an OPC communication protocol.

[0144] The model dynamic updating module is configured to collect actual operation data of the desorption tower as process feedback data in real time, and update the multivariate dynamic response model and re-execute the Bayesian optimization algorithm to obtain an updated dynamic matrix control interval controller according to the process feedback data.

[0145] In summary, by constructing the multivariable dynamic response model based on the catalytic cracking process mechanism and historical operation data, the dynamic coupling relationship between the control variables and the controlled variables can be accurately described, the systematic modeling of the complex industrial process is realized, and reliable theoretical support is provided for the intelligent control of the desorption tower. By adopting the dynamic matrix control (DMC) interval control strategy, the combination of the prediction model and the constraint condition is used to ensure that the controlled variables are always maintained within the preset target interval; when the controlled variables deviate from the interval, a penalty term is automatically generated according to the deviation amplitude and the duration, so as to dynamically correct the control behavior and prevent the temperature fluctuation and the liquid level instability of the desorption tower. Through the Bayesian optimization algorithm with the minimum interval control performance index as the target, the optimal combination of the prediction time domain, the control time domain and the control weight is automatically searched, so that the controller can be dynamically self-adjusted according to different working conditions, and the self-learning and self-adaptive ability of the desorption tower control system is improved. Through the OPC communication protocol, the real-time interaction of the control system and the field equipment is realized, the operation data of the desorption tower can be continuously collected, and the multivariable dynamic response model can be automatically updated according to the feedback information; when the prediction error exceeds the threshold, the model re-identification and parameter re-optimization are automatically triggered, forming a closed-loop control mechanism of “online monitoring-dynamic modeling-parameter optimization-feedback correction”. The feasible operation interval is solved through the quadratic programming, so that the adjustment range of the control variable is always within the safety boundary of the equipment, and abnormal working conditions such as excessive opening and closing of the valve or overheating of the reboiler are prevented.

[0146] Those skilled in the art will appreciate that embodiments of the application can be supplied as a method, a system, or a computer program product. Accordingly, the application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the application can be embodied in the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage media, etc.) having computer usable program code embodied thereon.

[0147] The application is described with reference to the flowchart and / or block diagram illustrations of the methods, apparatus (systems) and computer program products according to embodiments of the application. It should be understood that each flow and / or block in the flowchart and / or block diagram illustrations, and a combination of flows and / or blocks in the flowchart and / or block diagram illustrations, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, create means for implementing the functions specified in the flowchart and / or block diagram illustrations. Figure 1 The means for carrying out the functions specified in the flowchart and / or block diagram illustrations. Figure 1 The means for carrying out the functions specified in the flowchart and / or block diagram illustrations.

[0148] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 The functions of a flow or multiple flows and / or a block or multiple blocks in conjunction with the disclosed aspects. Figure 1 The functions of a flow or multiple flows and / or a block or multiple blocks in conjunction with the disclosed aspects.

[0149] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow Figure 1 The functions of a flow or multiple flows and / or a block or multiple blocks in conjunction with the disclosed aspects. Figure 1 The functions of a flow or multiple flows and / or a block or multiple blocks in conjunction with the disclosed aspects.

[0150] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, rather than limit the present application. Although the present application has been described in detail with reference to the above-mentioned embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced without departing from the spirit and scope of the present application, and any modification or equivalent replacement should be covered in the protection scope of the claims of the present application.

Claims

1. A method for optimizing the control of the desorption tower section in catalytic cracking, characterized in that, include: Based on historical operating data and the mechanism of catalytic cracking, a multivariate dynamic response model is constructed, which describes the dynamic response relationship between the control variable and the controlled variable. Based on the target control interval of the controlled variable and the operational constraint interval of the controlled variable, and based on solving the quadratic programming problem, the feasible operational interval of the controlled variable that satisfies the target control interval is calculated, and the adjustment of the controlled variable is restricted according to the feasible operational interval; A dynamic matrix control interval controller is constructed based on the multivariable dynamic response model. The sum of the weighted time integral and the deviation amplitude integral of the controlled variable outside the target control interval is calculated by the dynamic matrix control interval controller as the interval control performance index. Based on the Bayesian optimization algorithm, and with the goal of minimizing the interval control performance index, the parameters of the prediction time domain, control time domain, and control weight coefficients of the dynamic matrix control interval controller are optimized to obtain the controller parameter combination that stabilizes the controlled variable within the target control interval; the controller parameter combination is then applied to the dynamic matrix control interval controller. Based on the OPC communication protocol, the measured values ​​of the controlled variables are acquired in real time and the set values ​​of the control variables are calculated. The actual operating data of the desorption tower is collected in real time as process feedback data, and the multivariate dynamic response model is updated based on the process feedback data, and the Bayesian optimization algorithm is re-executed to obtain the updated dynamic matrix control interval controller. When constructing a multivariate dynamic response model based on historical operating data and the mechanism of catalytic cracking, the following are included: The historical operation data is preprocessed, including removing signal noise, eliminating outlier data points, and standardization. The catalytic cracking process mechanism includes the principles of mass conservation and energy conservation. A simplified mechanism model of the desorption tower is established based on the catalytic cracking process mechanism. The simplified mechanism model and the preprocessed historical operation data are used to determine the multivariate dynamic response parameters using a recursive parameter identification method, and a multivariate dynamic response model is constructed based on the multivariate dynamic response parameters. When calculating the feasible operational interval of the control variable that satisfies the target control interval based on the target control interval and the operational constraint interval of the control variable, and based on solving a quadratic programming problem, the calculation includes: The target control interval and the operational constraint interval of the controlled variable are preset; Construct a quadratic programming problem objective function, which is determined based on the magnitude of the control variable adjustment and the control effect; Set constraints for the quadratic programming problem, including that the controlled variable must be kept within the target control interval and that the change of the controlled variable cannot exceed the operational constraint interval. The quadratic programming problem is solved using the effective set method to obtain the actual safe operating interval of the control variable, which is the feasible operating interval of the control variable that satisfies the target control interval. When constructing a dynamic matrix control interval controller based on the multivariable dynamic response model, the following are included: The step response coefficient between the control variable and the controlled variable is calculated based on the multivariate dynamic response model, and the change trend of the controlled variable in the future time domain is predicted based on the step response coefficient to construct a prediction model. The interval control objective function is set so that no penalty is generated when the controlled variable is within the target control interval, and a corresponding penalty term is generated based on the degree and duration of the deviation when the controlled variable exceeds the target control interval. Based on the requirement that the controlled variable must be maintained within the target control interval and the limitation that the control variable must meet the operational constraint range, interval constraint conditions are set. The prediction model, the interval control objective function, and the interval constraints are integrated to form the dynamic matrix control interval controller. When calculating the sum of the weighted time integral and the deviation amplitude integral of the controlled variable outside the target control interval as an interval control performance index using the dynamic matrix control interval controller, the calculation includes: The deviation of the controlled variable from the target control interval is determined based on the interval control objective function. The weighted time integral is calculated based on the length of time the controlled variable exceeds the target control interval and its deviation. Calculate the integral of the deviation magnitude based on the amount by which the controlled variable deviates from the target control range; The weighted time integral and the deviation amplitude integral are proportionally combined to form the interval control performance index; When optimizing the parameters of the prediction time domain, control time domain, and control weight coefficients of the dynamic matrix control interval controller to obtain the controller parameter set that stabilizes the controlled variable within the target control interval, the process includes: Based on the interval control performance index, a Bayesian optimization objective is set, with the optimization direction being to minimize the interval control performance index; The feasible range of controller parameter combinations is determined based on the prediction time domain, control time domain, and control weight coefficients. A performance prediction model is constructed based on historical evaluation data. The performance prediction model is used to estimate the control effect of different combinations of controller parameters. Repeat the parameter evaluation and model update process until the convergence condition is met; and determine the controller parameter combination that minimizes the performance index of the interval control.

2. The method for optimizing and controlling the desorption tower section of a catalytic cracking reactor according to claim 1, characterized in that, The controlled variables include the opening degree of the valve for the flow rate of deethaned gasoline out of the desorber, the opening degree of the three-way valve for the shell side temperature of the reboiler at the bottom of the desorber, and the opening degree of the valve for the flow rate of condensate oil into the desorber; the controlled variables include the liquid level in the desorber and the temperature at the bottom of the desorber.

3. The method for optimizing and controlling the desorption tower section of a catalytic cracking reactor according to claim 2, characterized in that, When acquiring measured values ​​of controlled variables and calculating setpoints of control variables in real time based on the OPC communication protocol, the process includes: A secure communication connection is established based on the OPC communication protocol; the measured values ​​of the desorption tower liquid level and bottom temperature are acquired in real time; the acquired measured values ​​are input into the dynamic matrix control interval controller, and data processing is performed in combination with the current controller parameters to calculate the set value of the control variable.

4. The method for optimizing and controlling the desorption tower section of a catalytic cracking reactor according to claim 3, characterized in that, When updating the multivariate dynamic response model based on the process feedback data and re-executing the Bayesian optimization algorithm to obtain the updated dynamic matrix control interval controller, the process includes: The prediction error is calculated based on the process feedback data and the prediction data of the multivariate dynamic response model; the prediction error is compared with the prediction error threshold; when the prediction error is greater than the prediction error threshold, the multivariate dynamic response model is updated using a recursive parameter identification method; the updated multivariate dynamic response model is then used to re-evaluate the interval control performance index; a Bayesian optimization algorithm is executed based on the re-evaluated performance index to obtain an updated controller parameter combination; and the updated controller parameter combination is applied to the dynamic matrix control interval controller.

5. A catalytic cracking desorption tower interval optimization control system, used to apply the catalytic cracking desorption tower interval optimization control method as described in any one of claims 1-4, characterized in that, include: The response model construction module is configured to construct a multivariate dynamic response model based on historical operating data and the mechanism of catalytic cracking process. The multivariate dynamic response model describes the dynamic response relationship between the control variable and the controlled variable. The control variable limiting module is configured to calculate the feasible operating range of the control variable that satisfies the target control range based on the target control range of the controlled variable and the operating constraint range of the control variable, and to limit the adjustment of the control variable based on the feasible operating range. The controller construction module is configured to construct a dynamic matrix control interval controller based on the multivariable dynamic response model, and to calculate the sum of the weighted time integral and the deviation amplitude integral of the controlled variable outside the target control interval as the interval control performance index through the dynamic matrix control interval controller; The parameter combination acquisition module is configured to perform parameter optimization on the prediction time domain, control time domain, and control weight coefficients of the dynamic matrix control interval controller based on the Bayesian optimization algorithm and with the goal of minimizing the interval control performance index, to obtain the controller parameter combination that stabilizes the controlled variable within the target control interval; and apply the controller parameter combination to the dynamic matrix control interval controller. The OPC communication module is configured to acquire the measured values ​​of the controlled variable and calculate the setpoint values ​​of the control variable in real time based on the OPC communication protocol. The model dynamic update module is configured to collect the actual operating data of the desorption tower in real time as process feedback data, and update the multivariate dynamic response model based on the process feedback data and re-execute the Bayesian optimization algorithm to obtain the updated dynamic matrix control interval controller.

Images