Nonlinear predictive control method for Novolens polypropylene process
By establishing a mechanistic state-space model for the Novolen polypropylene process, the problem of nonlinear modeling in the Novolen polypropylene process was solved, enabling precise nonlinear model predictive control and improving control accuracy and product quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-04-10
AI Technical Summary
Due to its strong nonlinearity, large time delay, time-varying nature, and multivariate coupling characteristics, the Novolen polypropylene process is difficult to adapt to existing process modeling models, resulting in high control difficulty. Existing linear models have decreased accuracy, while nonlinear models have high computational complexity and poor real-time performance, making them difficult to meet the real-time control requirements of industrial sites.
A mechanistic model based on the Novolen polypropylene process was established. By introducing a state-space model and fusing information from the mechanistic model, a mechanistic state-space model was constructed. Predictive controllers were then designed to achieve accurate nonlinear model predictive control of the Novolen polypropylene process.
It improves the control accuracy and product quality of the Novolen polypropylene process, realizes precise nonlinear model predictive control of the Novolen polypropylene process, and solves the adaptation problem of strong nonlinearity, large time delay, time variation and multivariable coupling characteristics.
Smart Images
Figure CN121832275A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of nonlinear predictive control technology for chemical processes, and in particular to a nonlinear predictive control method for the Novolen polypropylene process. Background Technology
[0002] Novolen polypropylene technology is one of the world's mainstream polypropylene production technologies. Its core is gas-phase bulk polymerization, offering advantages such as a short production process, low energy consumption, and high product flexibility. However, this process suffers from significant nonlinearity due to the nonlinear changes in reaction kinetics with temperature and concentration; large time lags (tens of seconds to minutes) in material transport and reaching steady state; time-varying catalyst activity and raw material composition fluctuations; and multivariate coupling characteristics, such as the ethylene feed flow rate affecting both gas-phase ethylene concentration and reactor temperature. Therefore, the Novolen polypropylene process is an extremely complex chemical process, accompanied by strong nonlinear kinetic behavior, making it difficult to describe with simple linear models and resulting in significant challenges in process control.
[0003] In existing technologies, process modeling is the foundation for achieving effective control. Current process modeling mainly falls into two categories:
[0004] First, there are data fitting models, such as ARIMA models and neural network models. These models rely on a large amount of historical data. Although the modeling process is simple, they cannot reveal the internal physicochemical nature of the process and have poor adaptability to changes in operating conditions. When production load, raw material quality, etc. deviate from the range of training data, the model accuracy will drop significantly.
[0005] Secondly, there are mechanistic models, constructed based on fundamental physicochemical laws such as material balance, energy balance, and reaction kinetics. These models accurately reflect the inherent laws of the process, offering strong interpretability and good generalization, and can reveal the essential laws of the process. However, the strong nonlinearity of the Novolen polypropylene process makes it impossible to directly adapt to mature model predictive control technology. While nonlinear model predictive control can theoretically handle nonlinear processes, it requires solving complex nonlinear programming problems, resulting in high computational complexity and poor real-time performance. This makes it difficult to meet the real-time control requirements of industrial sites, limiting its practical application in the Novolen polypropylene process.
[0006] Therefore, there is an urgent need for a nonlinear predictive control method for the Novolen polypropylene process to solve the above-mentioned technical problems. Summary of the Invention
[0007] The purpose of this invention is to provide a nonlinear predictive control method for the Novolen polypropylene process, addressing the technical problem that existing process modeling models are ill-suited to the highly nonlinear, large time-delay, time-varying, and multivariable coupled characteristics of the Novolen polypropylene process. The numerous technical effects of the preferred solutions provided by this invention are detailed below.
[0008] To achieve the above objectives, the present invention provides the following technical solution:
[0009] This invention provides a nonlinear control method for the Novolen polypropylene process, comprising the following steps:
[0010] S1: Based on the internal reaction mechanism of the Novolen polypropylene process, establish the mechanism model corresponding to the controlled loop and obtain the real-time gain of the process;
[0011] S2: Obtain the time constant and lag time parameters of the controlled Novolen polypropylene process model;
[0012] S3: Construct a corresponding mechanistic state-space model using the real-time gain, the time constant, and the lag time;
[0013] S4: Based on the aforementioned mechanistic state-space model, obtain the state prediction of the controlled Novolen polypropylene process;
[0014] S5: Introduce an objective function to obtain the optimal control input increment for the controlled Novolen polypropylene process;
[0015] S6: Optimal control is obtained by calculating the optimal control input increment.
[0016] Further, in step S1, the mechanism model is a nonlinear model established based on the internal reaction mechanism of the Novolen polypropylene process, and its output is the actual output of the process, while its input is the control input of the process; wherein:
[0017] The internal reaction mechanism includes material balance, energy balance, and reaction kinetics.
[0018] Preferably, in step S1, the real-time gain is obtained by differentiating the mechanism model and changes dynamically with the control input of the controlled Novolen polypropylene process.
[0019] Further, in step S2, obtaining the time constant and lag time parameters includes:
[0020] A step test was performed on the controlled Novolen polypropylene process, and the corresponding step response data were collected.
[0021] The collected step response data is then subjected to dimensionless transformation.
[0022] The time constant and lag time of the controlled Novolen polypropylene process model were calculated using the two-point method.
[0023] Furthermore, the two-point method includes: selecting two time points at which the dimensionless output reaches a preset proportion of the typical response characteristics of the corresponding process, and solving for the time constant and the lag time based on the mathematical relationship between the two time points.
[0024] Further, in step S3, the corresponding mechanistic state-space model is constructed, including:
[0025] A discrete model of the controlled Novolen polypropylene process is established based on real-time gain, time constant, and sampling period.
[0026] Define the output setpoint of the controlled Novolen polypropylene process at time k, calculate the tracking error at that time, and derive the tracking error at time k+1.
[0027] Select a vector containing the tracking error information mentioned above as a state vector, and construct a state space model based on the state vector;
[0028] By introducing transformation variables to simplify the state-space model, the final mechanistic state-space model is obtained.
[0029] Furthermore, the tracking error is the difference between the output set value and the actual output quantity;
[0030] The input matrix of the state-space model is an online time-varying matrix, whose value is updated in real time according to the current control input.
[0031] Further, in step S4, the state prediction includes the following steps:
[0032] Based on the aforementioned mechanistic state-space model, and combining the prediction and control time domains of model predictive control, the state evolution trajectory of the controlled Novolen polypropylene process in the future prediction time domain is obtained through model recursive calculation; the prediction process uses a coefficient matrix derived from the aforementioned mechanistic state-space model.
[0033] Further, in step S5, obtaining the optimal control input increment for the controlled Novolen polypropylene process includes:
[0034] Construct an objective function, which is based on the state vector and includes weighting coefficients for adjusting the control weights of each state parameter;
[0035] Solving for the optimal transformation variable vector includes: when there are no process constraints, solving by taking the first derivative of the objective function; when there are process constraints, solving by using the standard quadratic programming method; the process constraints include control input limits and controlled output limits;
[0036] Based on the nonlinear characteristics of the aforementioned mechanistic state-space model, the optimal transformation variable vector is solved using nonlinear equations to obtain the optimal control input increment for the controlled Novolen polypropylene process.
[0037] Furthermore, it also includes a closed-loop update step, specifically including:
[0038] Based on the real-time operating status of the controlled Novolen polypropylene process, update the real-time gain in step S1 and the mechanism state space model parameters in step S3.
[0039] Repeat steps S4 to S5 to obtain the real-time optimal control input increment for the controlled Novolen polypropylene process.
[0040] Based on this incremental execution step S6, real-time optimal control is obtained and applied to the controlled Novolen polypropylene process.
[0041] The nonlinear control method for the Novolen polypropylene process provided by this invention first establishes a mechanistic model corresponding to the controlled loop based on the internal reaction mechanism of the Novolen polypropylene process. Then, the process model information contained in the mechanistic model is integrated into the state-space model, thereby obtaining the mechanistic state-space model of the Novolen polypropylene process. Finally, based on the obtained nonlinear mechanistic state-space model, predictive derivation and controller design are performed to achieve accurate nonlinear model predictive control of the Novolen polypropylene process. This solves the problem that existing process modeling models are difficult to adapt to the strong nonlinearity, large time delay, time-varying, and multivariable coupling characteristics of the Novolen polypropylene process. Attached Figure Description
[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0043] Figure 1 This is a flowchart of the Novolen polypropylene process nonlinear control method of the present invention;
[0044] Figure 2This is a flowchart illustrating how the time constant and lag time parameters are obtained according to the present invention.
[0045] Figure 3 This is a flowchart of the process of constructing the corresponding mechanistic state-space model according to the present invention;
[0046] Figure 4 This is a flowchart illustrating the optimal control input increment for obtaining the controlled Novolen polypropylene process according to the present invention. Detailed Implementation
[0047] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be described in detail below. Obviously, the described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0048] Model predictive control (MDC), as a widely used intelligent control technology, has demonstrated significant advantages in handling constrained chemical processes with large time delays. For linear process models, controller design is usually quite straightforward. However, for highly nonlinear mechanistic models, the lack of general formulas in model prediction derivations makes it impossible to directly design corresponding predictive controllers. Therefore, combining nonlinear mechanistic models with MDC technology has become a challenge.
[0049] Compared to transfer function models, state-space models are better suited for describing nonlinear process systems and are more convenient for subsequent controller design. Combining mechanistic and state-space models, designing predictive controllers based on the resulting mechanistic-state-space model, and then applying it to the Novolen polypropylene process loop will significantly improve the control accuracy of the polypropylene loop, thereby further enhancing the yield and quality of polypropylene products.
[0050] To address this issue, the present invention provides a nonlinear predictive control method for the Novolen polypropylene process. It tackles the challenge of designing predictive control based on the Novolen polypropylene process mechanism model by introducing a state-space model for fusion. This allows the state-space model to incorporate information from the mechanism model, thereby improving the accuracy of the state-space model's representation of the Novolen polypropylene process. Ultimately, this also allows the corresponding model predictive controller to achieve better control performance, thus enhancing the overall accuracy of the Novolen polypropylene production process.
[0051] like Figure 1 As shown in the figure, this embodiment provides a nonlinear predictive control method for the Novolen polypropylene process based on a mechanistic state-space model, which specifically includes the following steps:
[0052] S1: Based on the internal reaction mechanism of the Novolen polypropylene process, establish the mechanism model corresponding to the controlled loop and obtain the real-time gain of the process;
[0053] S2: Obtain the time constant and lag time parameters of the controlled Novolen polypropylene process model;
[0054] S3: Construct a corresponding mechanistic state-space model using the real-time gain, the time constant, and the lag time;
[0055] S4: Based on the aforementioned mechanistic state-space model, obtain the state prediction of the controlled Novolen polypropylene process;
[0056] S5: Introduce an objective function to obtain the optimal control input increment for the controlled Novolen polypropylene process;
[0057] S6: Optimal control is obtained by calculating the optimal control input increment.
[0058] This nonlinear control method for the Novolen polypropylene process first establishes a mechanistic model corresponding to the controlled loop based on the internal reaction mechanism of the Novolen polypropylene process. Then, the process model information contained in the mechanistic model is integrated into the state-space model to obtain the mechanistic state-space model of the Novolen polypropylene process. Finally, based on the obtained nonlinear mechanistic state-space model, prediction derivation and controller design are performed to achieve accurate nonlinear model predictive control of the Novolen polypropylene process.
[0059] Specifically, in step S1, the mechanistic model is a nonlinear model established based on the internal reaction mechanism of the Novolen polypropylene process. Its output is the actual output of the process, and its input is the control input of the process. The internal reaction mechanism includes material balance, energy balance, and reaction kinetics. The real-time gain is obtained by differentiating the mechanistic model and changes dynamically with the control input of the controlled Novolen polypropylene process.
[0060] In this embodiment, a mechanistic model formula corresponding to the controlled loop is established based on the internal reaction mechanism of the Novolen polypropylene production process:
[0061] y(k)=M(u(k))
[0062] Where y(k) is the output of the controlled Novolen polypropylene process, and u(k) is the input of the corresponding controlled polypropylene process. M is the formula for the nonlinear mechanism model of the controlled Novolen polypropylene process.
[0063] For the above mechanistic model formula, the real-time gain of the controlled Novolen polypropylene process can be calculated by the following formula.
[0064] K(k)=M′(u(k))
[0065] Where K(k) is the real-time gain of the controlled polypropylene process. M′ is the first derivative of the nonlinear mechanism model formula. From this formula, it can be seen that the real-time gain K(k) of the controlled Novolen polypropylene process at time k varies according to the process input value u(k).
[0066] like Figure 2 As shown, in step S2, the time constant and lag time parameters are obtained, which specifically includes the following steps:
[0067] S21: Perform a step test on the controlled Novolen polypropylene process and collect the corresponding step response data;
[0068] S22: Perform dimensionless transformation on the collected step response data;
[0069] S23: The time constant T and lag time τ of the controlled Novolen polypropylene process model are calculated using the two-point method.
[0070] The two-point method includes: selecting two time points at which the dimensionless output reaches a preset proportion of the typical response characteristics of the corresponding process, and solving for the time constant and the lag time based on the mathematical relationship between the two time points.
[0071] The dimensionless conversion formula for the step response data of the Novolen polypropylene process and the formulas for calculating the time constant T and lag time τ of the controlled Novolen polypropylene process model are as follows:
[0072]
[0073] T = 2(k2 - k1)
[0074] τ=2k1-k2
[0075] in, These represent the dimensionless form and steady-state value of the output y(k) of the Novolen polypropylene process, respectively. In the two-point method, the value chosen is the one that satisfies the dimensionless polypropylene process output... and The two points.
[0076] like Figure 3 As shown, in step S3, the corresponding mechanism state-space model is constructed, which specifically includes the following steps:
[0077] S31: Based on real-time gain, time constant and sampling period, establish a discrete model of the controlled Novolen polypropylene process;
[0078] S32: Define the output setpoint of the controlled Novolen polypropylene process at time k, calculate the tracking error at that time, and derive the tracking error at time k+1;
[0079] S33: Select a vector containing the tracking error information mentioned above as a state vector, and construct a state space model based on the state vector;
[0080] S34: Introduce transformation variables to simplify the state-space model and obtain the final mechanistic state-space model.
[0081] The tracking error is the difference between the output setpoint and the actual output. The input matrix of the state-space model is an online time-varying matrix, whose value is updated in real time according to the current control input.
[0082] In this embodiment, specifically, the corresponding mechanistic state-space model is constructed using the real-time gain expression, time constant T, and lag time τ of the controlled Novolen polypropylene process.
[0083] First, the corresponding discrete equation model for the Novolen polypropylene process can be obtained:
[0084] y(k+1)=αy(k)+K(kd)(1-α)u(kd)
[0085] in, d=τ / T s T s This refers to the sampling period for the polypropylene process.
[0086] Taking Δ on both sides of the above discrete equation model, we can obtain
[0087] Δy(k+1)=αΔy(k)+K(kd)(1-α)Δu(kd)
[0088] Secondly, assuming the output setpoint of the Novolen polypropylene process at time k is c, the tracking error at time k is...
[0089] e(k)=y(k)-c
[0090] Furthermore, the tracking error at time k+1 is
[0091] e(k+1) = e(k) + Δe(k+1)
[0092] =e(k)+αΔy(k)+K(kd)(1-α)Δu(kd)
[0093] Here, the state vector is selected as...
[0094] x(k) = [Δy(k), e(k)] Τ
[0095] The corresponding state-space model is
[0096] x(k+1)=Ax(k)+B(kd)Δu(kd)
[0097] in,
[0098]
[0099] This shows that the input matrix B(kd) is time-varying online, and its value depends on u(kd).
[0100] Further achievable
[0101]
[0102] in,
[0103]
[0104] In step S4, the state prediction includes the following steps: based on the mechanistic state-space model, combined with the prediction time domain and control time domain of the model predictive control, the state evolution trajectory of the controlled Novolen polypropylene process in the future prediction time domain is obtained through model recursive calculation; the prediction process uses a coefficient matrix derived from the mechanistic state-space model.
[0105] Specifically, based on the state-space model, the state prediction of the Novolen polypropylene process can be obtained as follows:
[0106] X(k+d)=A m x(k)+B m θ(k)
[0107] in,
[0108]
[0109] P and M represent the prediction time domain and control time domain of model predictive control, respectively.
[0110] like Figure 4 As shown, in step S5, the optimal control input increment for the controlled Novolen polypropylene process is obtained, which specifically includes the following steps:
[0111] S51: Construct an objective function, which is based on the state vector and includes weighting coefficients for adjusting the control weights of each state parameter;
[0112] S52: Solving for the optimal transformation variable vector includes: when there are no process constraints, solving the objective function by taking the first derivative; when there are process constraints, solving the objective function using the standard quadratic programming method; the process constraints include control input limits and controlled output limits;
[0113] S53: Based on the nonlinear characteristics of the aforementioned mechanism state-space model, the optimal transformation variable vector is solved using nonlinear equations to obtain the optimal control input increment of the controlled Novolen polypropylene process.
[0114] Specifically, the following objective function is introduced for the Novolen polypropylene controlled process.
[0115]
[0116] Here, Q represents the weighting coefficients of the state vector. The objective function does not include weighting for the control input increment. A similar effect can be achieved by constraining the output increment Δy of the Novolen polypropylene process in the state prediction vector by adjusting the first few terms in Q.
[0117] The optimal transformation variable vector θ(k) can be obtained by differentiating the above objective function or by solving a standard quadratic programming problem under constraints. For example, under differentiation...
[0118] θ(k)=-(B m Τ QB m ) -1 B m Τ QA m X(k+d)
[0119] After obtaining the optimal θ(k), the first term cannot be directly applied to the system; it needs to be applied according to the formula. Solving the nonlinear equations will yield the final optimal Novolen polypropylene process control input increment Δu(k).
[0120] This embodiment also includes a closed-loop update step, specifically including:
[0121] Based on the real-time operating status of the controlled Novolen polypropylene process, update the real-time gain in step S1 and the mechanistic state-space model parameters in step S3; repeat steps S4 to S5 to obtain the real-time optimal control input increment for the controlled Novolen polypropylene process; based on this increment, execute step S6 to obtain the real-time optimal control and apply it to the controlled Novolen polypropylene process.
[0122] After obtaining the optimal control input increment Δu(k), the Novolen polypropylene process control input u(k) = u(k-1) + Δu(k) can be calculated and written into the system to complete closed-loop intelligent control. Repeating the calculation in step S5 in subsequent cycles will allow the optimal Novolen polypropylene process control input to be obtained in real time.
[0123] This nonlinear predictive control method for the Novolen polypropylene process incorporates the mechanistic model of the Novolen polypropylene process into a state-space model, ultimately obtaining a state-mechanistic model applicable to the complex nonlinear Novolen polypropylene process. Based on this state-mechanistic model, a state-mechanistic model predictive controller for the polypropylene process is further derived, ultimately achieving more precise nonlinear control of the Novolen polypropylene process and further improving the production accuracy and product quality of the nonlinear Novolen polypropylene process.
[0124] The following example uses the controlled loop of gas phase ethylene concentration in the Novolen polypropylene copolymerization process, where the controlled variable and the control variable are gas phase ethylene concentration and ethylene feed flow rate, respectively.
[0125] Step S1: Based on the internal reaction mechanism of the gas-phase ethylene concentration control process in Novolen polypropylene production, establish the corresponding gas-phase ethylene concentration control mechanism model formula:
[0126] y(k)=lu(k) n
[0127] Where y(k) is the gaseous ethylene concentration, and u(k) is the ethylene feed flow rate. l and n are the relevant reaction coefficients of the process mechanism model where the gaseous ethylene concentration is controlled.
[0128] For the mechanism model formula of the gas phase ethylene concentration control process in the production of Novolen polypropylene mentioned above, its real-time gain can be calculated by the following formula.
[0129] K(k)=nlu(k) n-1
[0130] Where K(k) is the real-time gain of the gas-phase ethylene concentration control process. From this formula, it can be seen that the real-time gain K(k) of the gas-phase ethylene concentration control process at time k varies according to the ethylene feed flow rate u(k).
[0131] Step S2: Calculate the time constant and lag time parameters of the gas-phase ethylene concentration control process model, specifically:
[0132] S21: Conduct a step test on the gas phase ethylene concentration control process in the production of Novolen polypropylene, and collect data on the change in gas phase ethylene concentration after adjusting the ethylene feed flow rate.
[0133] S22: The collected data on changes in gaseous ethylene concentration are processed by dimensionless conversion, and then the time constant T and lag time τ of the gaseous ethylene concentration control process model in Novolen polypropylene production are calculated using the two-point method.
[0134] The formulas for dimensionless conversion of gaseous ethylene concentration change data and for calculating the time constant T and lag time τ of the gaseous ethylene concentration control process model in Novolen polypropylene production are as follows:
[0135]
[0136] T = 2(k2 - k1)
[0137] τ=2k1-k2
[0138] in, These represent the dimensionless form and steady-state value of the gaseous ethylene concentration y(k), respectively. In the two-point method, the value is taken as the dimensionless gaseous ethylene concentration. and The two points.
[0139] Step S3: Construct the corresponding state-space model of the gas phase ethylene concentration control process mechanism using the obtained real-time gain expression, time constant T, and lag time τ of the gas phase ethylene concentration control process.
[0140] First, the corresponding discrete equation model for the gas-phase ethylene concentration control process can be obtained:
[0141] y(k+1)=αy(k)+K(kd)(1-α)u(kd)
[0142] in, d=τ / T s T s This refers to the sampling cycle for the gas-phase ethylene concentration control process.
[0143] Taking Δ on both sides of the discrete equation model for the gas phase ethylene concentration control process in the Novolen polypropylene production process described above, we can obtain...
[0144] Δy(k+1)=αΔy(k)+K(kd)(1-α)Δu(kd)
[0145] Secondly, assuming the setpoint for the gaseous ethylene concentration at time k is c, the tracking error for the gaseous ethylene concentration at time k is:
[0146] e(k)=y(k)-c
[0147] Furthermore, the tracking error of the gaseous ethylene concentration at time k+1 is:
[0148] e(k+1) = e(k) + Δe(k+1)
[0149] =e(k)+αΔy(k)+K(kd)(1-α)Δu(kd)
[0150] Here, the state vector is selected as...
[0151] x(k) = [Δy(k), e(k)] Τ
[0152] The corresponding state-space model for the gas-phase ethylene concentration control process in Novolen polypropylene production is as follows:
[0153] x(k+1)=Ax(k)+B(kd)Δu(kd)
[0154] in,
[0155]
[0156] This shows that the input matrix B(kd) of the state-space model for the gas-phase ethylene concentration control process is online and time-varying, and its value depends on the ethylene feed flow rate u(kd) at the corresponding time.
[0157] Further achievable
[0158]
[0159] in
[0160]
[0161] Step S4: Based on the above state-space model of the gas-phase ethylene concentration control process, the state prediction of the gas-phase ethylene concentration control process in Novolen polypropylene production can be obtained as follows:
[0162] X(k+d)=A m x(k)+B m θ(k)
[0163] in,
[0164]
[0165] P and M represent the prediction time domain and control time domain of the predictive controller for the gas phase ethylene concentration control process model, respectively.
[0166] Step S5: Introduce the following objective function for the gas phase ethylene concentration control process in Novolen polypropylene production.
[0167]
[0168] Here, Q is the weighting coefficient of the state vector of the gas phase ethylene concentration control process. The objective function does not include weighting for the change in ethylene feed flow rate. Here, a similar effect can be achieved by constraining the change in gas phase ethylene concentration Δy in the state prediction vector by the first few terms in Q.
[0169] By differentiating the above objective function or solving a standard quadratic programming problem under constraints, the optimal transformation variable vector θ(k) for controlling the gas phase ethylene concentration in Novolen polypropylene production can be obtained. For example, under differentiation...
[0170] θ(k)=-(B m Τ QB m ) -1 B m Τ QA m X(k+d)
[0171] After obtaining the optimal θ(k) for the gas-phase ethylene concentration control process, the first term cannot be directly applied to the gas-phase ethylene concentration control process. Here, it is necessary to follow the formula... Solving the nonlinear equations will yield the optimal ethylene feed flow rate change Δu(k) for the final required gas phase ethylene concentration control process.
[0172] Step S6: After obtaining the optimal ethylene feed flow rate change Δu(k), the optimal ethylene feed flow rate u(k) = u(k-1) + Δu(k) can be entered into the gas phase ethylene concentration control process to complete closed-loop intelligent control. Repeating the calculation in step S5 in subsequent cycles will allow real-time acquisition of the optimal ethylene feed flow rate in the gas phase ethylene concentration control process during Novolen polypropylene production.
[0173] This embodiment establishes a nonlinear model predictive control method for the polypropylene process based on the mechanistic state-space model by means of establishing a mechanism model of the Novolen polypropylene process, incorporating mechanism information, transforming the mechanism state-space model, and designing a state-space model predictive controller, thereby achieving more accurate optimization control of the Novolen polypropylene process.
[0174] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A nonlinear control method for the Novolen polypropylene process, characterized in that, Includes the following steps: S1: Based on the internal reaction mechanism of the Novolen polypropylene process, establish the mechanism model corresponding to the controlled loop and obtain the real-time gain of the process; S2: Obtain the time constant and lag time parameters of the controlled Novolen polypropylene process model; S3: Construct a corresponding mechanistic state-space model using the real-time gain, the time constant, and the lag time; S4: Based on the aforementioned mechanistic state-space model, obtain the state prediction of the controlled Novolen polypropylene process; S5: Introduce an objective function to obtain the optimal control input increment for the controlled Novolen polypropylene process; S6: Optimal control is obtained by calculating the optimal control input increment.
2. The nonlinear control method for the Novolen polypropylene process according to claim 1, characterized in that, In step S1, the mechanism model is a nonlinear model established based on the internal reaction mechanism of the Novolen polypropylene process. Its output is the actual output of the process, and its input is the control input of the process; wherein: The internal reaction mechanism includes material balance, energy balance, and reaction kinetics.
3. The nonlinear control method for the Novolen polypropylene process according to claim 2, characterized in that, In step S1, the real-time gain is obtained by differentiating the mechanism model and changes dynamically with the control input of the controlled Novolen polypropylene process.
4. The nonlinear control method for the Novolen polypropylene process according to claim 1, characterized in that, In step S2, obtaining the time constant and lag time parameters includes: A step test was performed on the controlled Novolen polypropylene process, and the corresponding step response data were collected. The collected step response data is then subjected to dimensionless transformation. The time constant and lag time of the controlled Novolen polypropylene process model were calculated using the two-point method.
5. The nonlinear control method for the Novolen polypropylene process according to claim 4, characterized in that, The two-point method includes: selecting two time points at which the dimensionless output reaches a preset proportion of the typical response characteristics of the corresponding process, and solving for the time constant and the lag time based on the mathematical relationship between the two time points.
6. The nonlinear control method for the Novolen polypropylene process according to any one of claims 1-5, characterized in that, In step S3, the corresponding mechanistic state-space model is constructed, including: A discrete model of the controlled Novolen polypropylene process is established based on real-time gain, time constant, and sampling period. Define the output setpoint of the controlled Novolen polypropylene process at time k, calculate the tracking error at that time, and derive the tracking error at time k+1. Select a vector containing the tracking error information mentioned above as a state vector, and construct a state space model based on the state vector; By introducing transformation variables to simplify the state-space model, the final mechanistic state-space model is obtained.
7. The nonlinear control method for the Novolen polypropylene process according to claim 6, characterized in that: The tracking error is the difference between the output set value and the actual output value. The input matrix of the state-space model is an online time-varying matrix, whose value is updated in real time according to the current control input.
8. The nonlinear control method for the Novolen polypropylene process according to claim 6, characterized in that, In step S4, the state prediction includes the following steps: Based on the aforementioned mechanistic state-space model, and combining the prediction and control time domains of model predictive control, the state evolution trajectory of the controlled Novolen polypropylene process in the future prediction time domain is obtained through model recursive calculation; the prediction process uses a coefficient matrix derived from the aforementioned mechanistic state-space model.
9. The nonlinear control method for the Novolen polypropylene process according to claim 6, characterized in that: In step S5, the optimal control input increment for the controlled Novolen polypropylene process is obtained, including: Construct an objective function, which is based on the state vector and includes weighting coefficients for adjusting the control weights of each state parameter; Solving for the optimal transformation variable vector includes: when there are no process constraints, solving by taking the first derivative of the objective function; when there are process constraints, solving by using the standard quadratic programming method; the process constraints include control input limits and controlled output limits; Based on the nonlinear characteristics of the aforementioned mechanistic state-space model, the optimal transformation variable vector is solved using nonlinear equations to obtain the optimal control input increment for the controlled Novolen polypropylene process.
10. The nonlinear control method for the Novolen polypropylene process according to claim 1, characterized in that: It also includes closed-loop update steps, specifically including: Based on the real-time operating status of the controlled Novolen polypropylene process, update the real-time gain in step S1 and the mechanism state space model parameters in step S3. Repeat steps S4 to S5 to obtain the real-time optimal control input increment for the controlled Novolen polypropylene process. Based on this incremental execution step S6, real-time optimal control is obtained and applied to the controlled Novolen polypropylene process.