A polymerization reactor reaction prediction and optimization system

Abstract

The application relates to the field of reaction prediction, and discloses a polymerization kettle reaction prediction and optimization system which is used for realizing high-precision prediction and autonomous optimization of a polymerization reaction process. The system comprises the following steps: obtaining and processing data from a polymerization reaction history database, generating a standardized input data set, extracting a mechanism index set based on a polymerization reaction mechanism model, constructing a mechanism prediction model based on coupling of fractional order dynamics and non-equilibrium thermodynamics, generating a conversion rate prediction curve and a reaction time length prediction value, taking the prediction curve and the prediction value as optimization targets, adopting a gradient analysis mathematical optimization method to generate an optimized formula and a working condition parameter set, and converting the optimized formula and the working condition parameter set into execution instructions to implement production adjustment. The application realizes continuous accumulation and self-improvement of system knowledge, and improves the accuracy and efficiency of polymerization kettle reaction prediction and optimization.

Description

Technical Field

[0001] This invention relates to the field of reaction prediction, and more particularly to a system for predicting and optimizing polymerization reactor reactions. Background Technology

[0002] Polymerization reactors are the core process in polymer synthesis, and their reaction efficiency, product quality, and energy consumption directly affect the economic and environmental aspects of production. As the chemical industry moves towards intelligent and refined processes, higher demands are placed on the accurate prediction and optimized control of polymerization processes. Traditional polymerization process control relies heavily on experience and static models, making it difficult to cope with complex conditions such as raw material fluctuations, reaction nonlinearity, and multivariate coupling. This leads to problems such as unstable conversion rates, excessively long reaction times, high energy consumption, and inconsistent product quality.

[0003] In recent years, data-driven process optimization technology has been widely used in the chemical industry. In existing technologies, polymerization process optimization systems usually establish a black-box mapping relationship between input and output through historical data, but they lack consideration of reaction mechanisms, have poor extrapolation capabilities, and have low accuracy in predicting operating conditions that have not been experienced.

[0004] The shortcomings of existing technologies are mainly reflected in:

[0005] The failure to organically combine reaction mechanism parameters with big data analysis resulted in weak model generalization ability;

[0006] Ignoring fractional-order kinetics and heat-mass transfer coupling effects, it is impossible to accurately describe the memory characteristics and non-equilibrium processes of polymerization reactions;

[0007] Without a multi-objective collaborative optimization mechanism, it is difficult to achieve both high conversion rates and short response times simultaneously.

[0008] Therefore, we propose a polymerization reactor reaction prediction and optimization system to solve the above problems. Summary of the Invention

[0009] This invention provides a polymerization reactor reaction prediction and optimization system for achieving high-precision prediction and autonomous optimization of the polymerization reaction process.

[0010] The first aspect of this invention provides a polymerization reactor reaction prediction and optimization system, comprising: an acquisition module for acquiring and processing raw material addition data, additive dosage data, and reaction condition data to generate a standardized input dataset; an extraction module for extracting initiation efficiency parameters, dispersion stability parameters, and thermodynamic state parameters based on the standardized input dataset to form a set of mechanistic indicators; a construction module for using the set of mechanistic indicators as input variables to construct a mechanistic prediction model, determining model coefficients through parameter identification, and generating a conversion rate prediction curve and a reaction time prediction value; and a setting module for optimizing the raw material ratio and operating parameters with the conversion rate prediction curve and reaction time prediction value as optimization targets to generate a future formula parameter set and a future operating condition parameter set.

[0011] Optionally, in a first implementation of the first aspect of the present invention, the method includes: extracting raw material addition records, additive addition timing data, and reactor operating condition monitoring data from multiple data tables in a historical database of polymerization reactions to generate an original multi-source data set; performing integrity checks and outlier detection on the original multi-source data set, filtering the data based on the physical probability range, and generating a valid dataset; performing dimensional analysis on each parameter in the valid dataset, identifying the physical dimension type of each parameter, generating a parameter dimension classification table, and generating a dimensionless parameter set based on the parameter dimension classification table; performing combined operations on related parameters in the dimensionless parameter set to generate a dimensionless composite parameter set; and organizing the dimensionless composite parameter set according to time series and reaction batches to generate a standardized input dataset.

[0012] Optionally, in a second implementation of the first aspect of the present invention, the method includes: organically integrating the set of initiation efficiency parameters, the set of dispersion stability parameters, the set of thermodynamic state parameters, and the set of transport characteristic parameters, establishing the correlation between the parameters, and forming a set of mechanism indicators.

[0013] Optionally, in a third implementation of the first aspect of the present invention, the initiator concentration, temperature history, and impurity content data in the standardized input dataset are analyzed to calculate the effective free radical concentration and initiation rate parameters, generating an initiation efficiency parameter set; the dispersant characteristics, stirring conditions, and phase interface parameters in the standardized input dataset are processed to evaluate the dispersed phase stability and particle distribution characteristics, generating a dispersion stability parameter set; the temperature gradient, pressure change, and energy balance data in the standardized input dataset are analyzed to calculate the system driving force and thermodynamic flow, generating a thermodynamic state parameter set; and the viscosity change, diffusion coefficient, and heat transfer characteristic data in the standardized input dataset are processed to evaluate the transport resistance and efficiency, generating a transport characteristic parameter set.

[0014] Optionally, in the fourth implementation of the first aspect of the present invention, the method includes: designing a basic model framework containing fractional differential terms and non-equilibrium thermodynamic coupling terms, and generating a model structure description file; introducing fractional differential operators into the model structure description file to establish a kinetic equation with memory effect, and generating a fractional kinetic model; introducing heat-mass transfer coupling terms into the fractional kinetic model to establish a synergistic relationship between energy and mass transfer, and generating a set of coupled model equations; determining the undetermined coefficients in the model using a physical constraint-based numerical optimization method based on the set of mechanistic indices and experimental data of the reaction process, and generating a calibrated set of prediction model parameters; substituting the calibrated set of prediction model parameters into the set of coupled model equations, performing numerical solution calculations, and generating a conversion rate prediction curve and a reaction time prediction value.

[0015] Optionally, in the fifth implementation of the first aspect of the present invention, the method includes: extracting the final conversion rate value based on the conversion rate prediction curve, combining it with the reaction time prediction value, establishing a composite objective function with the goal of maximizing the final conversion rate and minimizing the reaction time, and generating a multi-objective optimization problem statement; determining the feasible range of raw material ratio and operating parameters according to equipment operation limitations and product quality requirements, and establishing a set of multiple constraints; analyzing the sensitivity of the composite objective function to each decision variable, and generating a complete gradient information matrix; using the gradient information matrix as the search direction, performing iterative optimization calculations under the constraints of the set of multiple constraints, and generating a preliminary optimized parameter solution; substituting the preliminary optimized parameter solution into the mechanism prediction model for verification calculation, and generating a future formula parameter set and a future operating condition parameter set.

[0016] Optionally, in a sixth implementation of the first aspect of the present invention, a composite objective function is established with the goal of maximizing the final conversion rate and minimizing the reaction time, wherein the composite objective value is... :

[0017] ;

[0018] in, This is the conversion rate weighting coefficient; The reaction time weighting coefficient satisfies α+β= 1; This represents the actual conversion rate. Target conversion rate; This refers to the actual reaction time. The target reaction time.

[0019] Optionally, in the seventh implementation of the first aspect of the present invention, an allocation module is further included, configured to obtain execution instructions based on the future formula parameter set and the future operating condition parameter set: converting the future formula parameter set and the future operating condition parameter set into a control instruction format recognizable by the equipment to generate a preliminary control instruction set; performing security verification and feasibility assessment on the preliminary control instruction set to generate a control instruction set; prioritizing the execution of the security-verified control instruction set to generate an ordered control instruction sequence; packaging the ordered control instruction sequence into a data packet of a specific format and transmitting it to the polymerization reactor control system via an industrial communication network; monitoring the execution status of the polymerization reactor control system in response to the received instructions in real time, collecting feedback on key parameters of the actual production process, and generating an instruction execution status report.

[0020] Optionally, in the eighth implementation of the first aspect of the present invention, an update module is further included, used for real-time data acquisition, real-time optimization of the parameter set, and knowledge base update: real-time acquisition of temperature, pressure, viscosity, and component concentration data during the reaction process to generate a real-time process state dataset; comparative analysis of the real-time process state dataset with the conversion rate prediction curve to calculate the dynamic deviation between the actual reaction process and the prediction model, generating a process deviation index set; online correction of the parameters of the mechanism prediction model based on the process deviation index set to generate an updated model parameter set; re-execution of the optimization calculation process using the updated model parameter set to generate a real-time optimized parameter set; and rationality judgment and necessary adjustments to the real-time optimized parameter set based on the polymerization reaction expert experience knowledge base to generate a final control parameter set.

[0021] The mechanism of this invention is as follows: A data processing system based on first principles and completely independent of machine learning is constructed. Through the mechanism-driven transformation of multi-source industrial data into dimensionless physical parameters, a dynamic model that integrates fractional differential and non-equilibrium thermodynamics is established.

[0022] Beneficial effects: Combining dimensionless parameters based on the polymerization reaction mechanism makes the generated dimensionless composite parameter set more closely match the actual physical process of the reaction, and can more accurately reflect the essential characteristics of the reaction, avoiding model deviations caused by improper data processing;

[0023] Introducing fractional differential operators to construct kinetic equations can more realistically reflect the reaction process. By introducing heat-mass transfer coupling terms, the synergistic and restrictive relationship between energy and mass transfer in the reaction process can be accurately characterized.

[0024] Sensitivity analysis by generating gradient information matrix provides a clear search direction for optimization algorithm, enabling it to find Pareto optimal solution set quickly and efficiently under multiple complex constraints, avoiding optimization getting trapped in local optima, and achieving the best balance between economic benefits and production efficiency;

[0025] Organically integrating parameter sets extracted from different dimensions and establishing correlations between parameters helps to understand the reaction mechanism at the system level and provides rich information support for building more accurate prediction models. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of one embodiment of the polymerization reactor reaction prediction and optimization system of the present invention;

[0027] Figure 2 This is a schematic diagram of another embodiment of the polymerization reactor reaction prediction and optimization system of the present invention;

[0028] Figure 3 This is a schematic diagram of one embodiment of the polymerization reactor reaction prediction and optimization device in this invention. Detailed Implementation

[0029] This invention provides a polymerization reactor reaction prediction and optimization system for achieving high-precision prediction and autonomous optimization of the polymerization reaction process. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the polymerization reactor reaction prediction and optimization system of the present invention includes:

[0031] 101. Acquisition module, used for data standardization processing: acquire raw material addition data, auxiliary agent dosage data and reaction condition data from the historical database of polymerization reaction, perform quality verification and normalization processing on the data, and generate a standardized input dataset;

[0032] It is understood that the executing entity of this invention can be a polymerization reactor reaction prediction and optimization device, or it can be a terminal or a server; the specific implementation is not limited here. This embodiment of the invention will be described using a server as an example.

[0033] It should be noted that, taking the PVC (polyvinyl chloride) polymerization process of a chemical company as an example, it is necessary to extract the production data of the most recent 100 batches from the historical database, covering three types of raw data: raw material addition, auxiliary agent dosage, and reaction conditions.

[0034] Data acquisition content, raw material addition data: including the amount of fresh vinyl chloride monomer (VCM) added per batch (unit: tons, range: 10-50 tons), the amount of recovered vinyl chloride monomer (VCM) added (unit: tons, range: 2-15 tons), the amount of deionized water (tons), the amount of dispersant (hydroxypropyl methylcellulose) used (kg), and the amount of initiator (diethylhexyl peroxide dicarbonate) used (kg). For example, the data for a certain batch is VCM=28.5 tons, water=15.2 tons, dispersant=85 kg, and initiator=50 kg.

[0035] Additive dosage data: including the amount of terminator (bisphenol A), chain transfer agent (mercaptoethanol), and buffer (sodium phosphate) added (all in kg), and the concentration (mol / L) of pH adjuster (sodium hydroxide): terminator = 2.1 kg, chain transfer agent = 0.8 kg, buffer = 3.5 kg, pH adjuster concentration = 0.15 mol / L.

[0036] Reaction operating data: including initial reactor temperature (°C, range: 45-75°C), pressure (MPa), stirring rate (rpm), reaction time (h), and cooling water flow rate (m³ / h). 3 Time-series data such as ( / h) were recorded, with initial temperature = 62.5℃, pressure = 0.85MPa, and stirring speed = 120rpm.

[0037] The quality verification rules are as follows: records with empty or invalid (negative) key fields (VCM dosage, reaction temperature, initiator dosage) are removed; reasonable ranges are set according to process constraints (initiator dosage should be between 30-100kg, and temperature should be between 45-75℃), and records exceeding the range are marked as abnormal and removed; the matching between the total amount of material fed and the reactor volume is checked (the total material volume should not exceed 80% of the reactor volume), and the continuity of the timestamps of the time series data is verified.

[0038] The validated data is then subjected to Min-Max normalization, scaling each indicator to the [0,1] range: Numerical fields (dosage, temperature, pressure): ;where X min and X maxBased on historical data (minimum VCM dosage is 10 tons, maximum is 50 tons).

[0039] Category fields (additive type): converted to one-hot encoding. Time series data (temperature change curve): segmented by time step, and the average value within each segment is normalized.

[0040] The output is a standardized dataset. The final structured dataset contains the following fields, as shown in Table 1 below:

[0041] Table 1 Standardized Dataset

[0042]

[0043] This dataset can be directly used for subsequent metric extraction and model building.

[0044] 102. Extraction module, used for index extraction: Based on the polymerization reaction mechanism model, the standardized input dataset is subjected to physical mechanism analysis to extract initiation efficiency parameters, dispersion stability parameters and thermodynamic state parameters, forming a set of mechanism indices;

[0045] It should be noted that, taking a 70m³ PVC production plant as an example... 3 Using the polymerization reactor as the object, 100 batches of standardized production data (including raw material dosage, auxiliary agent dosage, reaction temperature and pressure time series data, etc.) were used to extract three types of key parameters based on the polymerization reaction mechanism model, providing input for subsequent prediction models.

[0046] Initiation efficiency parameters were extracted, including the initiator decomposition rate constant: Based on the initiator half-life model (the half-life of diethylhexyl peroxide dicarbonate is 1.5 hours at 50°C), combined with reaction temperature and time-series data, the actual decomposition rate constant was calculated, typically ranging from 0.02 to 0.08 min. -1 Free radical generation rate: Calculated by multiplying the initiator concentration by the decomposition rate; a typical value is 1.5 × 10⁻⁶. -5 mol·L -1 ·s -1 Induction period duration: The time from the start of feeding to the first drop in pressure, averaging 25 minutes (range: 20–30 minutes).

[0047] Dispersion stability parameters extracted: Interfacial tension: calculated based on the relationship between dispersant (hydroxypropyl methylcellulose) concentration and aqueous phase viscosity, with a standard value of 35–40 mN / m. Critical particle size: estimated based on dispersant concentration and stirring rate (120 rpm), with a typical particle size range of 100–150 μm. Foam stability index: calculated based on the amount of terminator added and foam height data at discharge, with a value of 0.85 (range 0.7–1.0).

[0048] Thermodynamic state parameters were extracted, and the peak exothermic reaction rate was calculated as follows: based on the conversion curve and the heat of reaction (propylene polymerization heat is 2172 kJ / kg), the peak exothermic rate was 820 kW. The overall heat transfer coefficient, calculated based on the jacket water temperature difference, flow rate, and internal temperature gradient, was 277 W / (m³) after reactor cleaning. 2 ·K), after scaling, it drops to 176W / (m 2 • K). Critical explosive polymerization temperature threshold: 85℃ calculated using a thermal equilibrium model (exceeding this temperature may trigger a runaway reaction).

[0049] The mechanism indicator set is output, and the above parameters are integrated into a structured indicator set, as shown in Table 2 below:

[0050] Table 2 Set of Mechanism Indicators

[0051]

[0052] 103. Construction module for predictive model establishment: Using the set of mechanistic indicators as input variables, construct a mechanistic predictive model based on the coupling of fractional-order kinetics and non-equilibrium thermodynamics, determine the model coefficients through parameter identification, and generate conversion rate prediction curves and reaction time prediction values;

[0053] It should be noted that, taking a 70m³ PVC production plant as an example... 3 Using the prepolymerization reactor as the object, a mechanistic model is constructed using the set of mechanistic indicators (including initiation efficiency parameters, dispersion stability parameters, and thermodynamic state parameters) extracted in the aforementioned steps. This model can predict the conversion rate over time and the total reaction time, providing a basis for optimized control.

[0054] Model inputs and outputs; input variables: key parameters in the mechanistic index set, including the initiator decomposition rate constant (0.05 min). -1 ), free radical generation rate (1.5×10) -5 mol·L -1 ·s -1 ), interfacial tension (38 mN / m), peak exothermic rate (820 kW), overall heat transfer coefficient (277 W / (m²·K)), etc. Output objectives: Generate a conversion rate prediction curve (time-conversion rate relationship) and a reaction time prediction value (time required to reach the target conversion rate).

[0055] Fractional-order kinetic modeling was employed, with the reaction rate equation describing the memory effect and non-exponential decay characteristics of chain growth and termination during free radical polymerization using fractional derivatives. The conversion rate was expressed as a fractional differential with an order of 0.85 (determined through fitting historical data), reflecting the sub-diffusion behavior of the reaction rate. Parameter identification was performed based on 100 batches of historical data, using the least squares method to identify model coefficients. The chain growth rate constant was identified as 0.12 L·mol⁻¹. -1 ·s -1 The chain transition coefficient is 0.008, and the fractional order is 0.85.

[0056] Non-equilibrium thermodynamic coupling, thermal effect integration: combining the peak exothermic reaction rate (820kW) with the in-vessel heat transfer coefficient (277W / (m²)). 2 •K) coupling is used to construct an energy balance equation and predict the impact of transient changes in reaction temperature on the conversion rate. Critical state early warning: A thermodynamic non-equilibrium criterion (critical burst polymerization temperature threshold of 85℃) is introduced, and dynamic adjustment is triggered when the predicted temperature approaches this value.

[0057] Prediction results output, conversion rate curve: The model outputs a predicted curve of conversion rate changing over time (example data is shown in the table below); Response time: The predicted time required to reach the target conversion rate (10%) is 95 minutes (actual production statistics show 90-100 minutes), see Table 3 below:

[0058] Table 3 Prediction Results Output

[0059]

[0060] By comparing actual production data (10 batches of validation data), the average absolute error between the predicted conversion rate curve and the measured value is less than 0.5%, and the reaction time prediction error is within ±3 minutes. This model can be directly used for real-time monitoring and endpoint control of the prepolymerization reaction.

[0061] 104. Setting module for optimization calculation: Taking the conversion rate prediction curve and reaction time prediction value as optimization targets, a mathematical optimization method based on gradient analysis is used to optimize the raw material ratio and operating parameters, and generate a future formula parameter set and a future operating condition parameter set.

[0062] It should be noted that the conversion rate prediction curve (the predicted reaction time is 285 minutes when the target conversion rate is 90%) and the predicted reaction time value obtained in the above steps are used to optimize the raw material ratio and operating parameters with the core objective of shortening the reaction time and ensuring the conversion rate.

[0063] Optimization objectives and constraints: core objectives: reaction time ≤ 275 minutes (currently 285 minutes); conversion rate ≥ 90% (currently 90%); resin apparent density stable at 0.51–0.53 g / mL (historical fluctuation range 0.48–0.51 g / mL).

[0064] Constraints: Raw material dosage range (VCM: 10–50 tons), total initiator limit (30–100 kg), safe temperature range (45–75℃), stirring speed range (50–130 rpm).

[0065] Decision variables and initial values, raw material ratio variables: VCM dosage (initial 28.5 tons), deionized water volume (initial 15.2 tons), initiator EHP and CNP compound ratio (initial EHP content 70%), dispersant hydroxypropyl methylcellulose dosage (initial 85 kg).

[0066] Operating parameter variables: initial reaction temperature (initial 62.5℃), stirring rate (initial 120 rpm), cooling water flow rate (initial 25 m³ / h). 3 / h).

[0067] Gradient analysis optimization method, sensitivity analysis: By calculating the partial derivatives of the objective function (reaction time + conversion rate deviation) with respect to each variable, the key influencing factors are determined. The gradient value of the initiator compound ratio on the reaction rate is -0.15 (a negative value indicates that increasing the CNP ratio can accelerate the reaction), and the gradient value of the stirring rate on the particle dispersion is +0.08.

[0068] Iterative optimization: Adjust parameters along the negative gradient direction: Initiator blending: Reduce EHP content from 70% to 60% (CNP has a short half-life, increasing the initial reaction rate); Stirring rate: Reduce from 120 rpm to 50 rpm (Reduce mechanical shear force and reduce fine particle generation); Initial temperature: Increase from 62.5℃ to 65.0℃ (Accelerate initiator decomposition); Recalculate the objective function after each adjustment until the termination condition is met (change <1%).

[0069] Future formulation parameters: VCM = 30.0 tons, deionized water = 16.0 tons, EHP / CNP = 60% / 40%, dispersant = 80 kg. Future operating parameters: initial temperature = 65.0℃, stirring speed = 50 rpm (increased to 80 rpm in the later stages of the reaction), cooling water flow rate = 28 m³ / h. 3 / h.

[0070] Expected results: The reaction time is shortened to 270 minutes (reduced by 15 minutes), the conversion rate remains at 90%, the apparent density is stabilized at 0.52 g / mL, and the residual vinyl chloride content is reduced to 2 μg / kg (superior grade standard ≤5 μg / kg).

[0071] 105. Allocation module, used for control instruction generation: converts the future formula parameter set and future operating condition parameter set into execution instructions, and outputs them to the polymerization reactor control system to implement production adjustments.

[0072] It should be noted that, based on the optimized calculations of the future formula parameter set (VCM dosage 30.0 tons, EHP / CNP initiator compound ratio 60% / 40%, dispersant dosage 80 kg) and the future operating condition parameter set (initial temperature 65.0℃, stirring speed 50 rpm later increased to 80 rpm, cooling water flow rate 28 m³ / h), these parameters are converted into instructions that can be executed by the DCS (Distributed Control System) and ESS (Emergency Safety System), thereby realizing automated closed-loop control of production adjustments.

[0073] Command conversion rules and formula parameter conversion: Raw material addition command: Convert VCM dosage of 30.0 tons to DCS metering tank discharge valve opening command (opening 85%, duration 12 minutes). Additive addition command: Convert dispersant 80kg to metering pump stroke command (stroke frequency 35Hz, duration 45 seconds). Initiator compounding command: Convert EHP / CNP=60% / 40% to the linkage command of the feed valves of the two initiator storage tanks (EHP tank valve opening 60%, CNP tank valve opening 40%).

[0074] Operating Parameter Conversion: Temperature Control Command: Converts the initial temperature of 65.0℃ to cascade control parameters for the hot water valve and circulating water valve (main controller setpoint 65.0℃, secondary controller jacket water temperature setpoint 62.0℃). Agitation Rate Command: Converts the agitation rate of 50rpm (later 80rpm) to the inverter output frequency command (initially 25Hz, increasing to 40Hz in the later stages of the reaction). Cooling Water Flow Command: Converts the 28m³ / h cooling water flow rate... 3 / h is converted to a control valve opening command (opening degree 70%).

[0075] Integrated safety constraints and emergency termination conditions: If the reaction temperature exceeds 70℃ or the pressure exceeds 1.4MPa, the ESS termination agent addition command is triggered (automatically opening the termination agent valve and injecting 2.5kg of organic amine termination agent). Venting control: If the pressure continues to rise to 1.45MPa, the venting procedure is automatically started (venting regulating valve opening to 50%, shut-off valve fully open).

[0076] Command output and execution: The generated command set is transmitted to the DCS operator station via the OPC protocol. Specific execution steps include: Batch initialization: Issuing the batch number (BATCH-20240904-06) and synchronizing it to all metering tanks and reactors. Sequential control: Executing commands according to the optimized timing sequence (VCM and water added first, dispersant added after 90 seconds, initiator added after 300 seconds). Real-time monitoring: The DCS provides real-time feedback on temperature, pressure, and stirring current data. If deviations exceed limits (temperature fluctuation ±0.3℃), the valve opening is automatically fine-tuned.

[0077] In this embodiment of the invention, not only are various types of data (raw material addition, auxiliary agent dosage, reaction conditions) covered, but detailed quality verification rules are also formulated for different data characteristics to ensure that the acquired data is of high quality and uniform in format, providing a reliable foundation for subsequent indicator extraction and model construction, and improving the accuracy and stability of the entire system;

[0078] Based on the polymerization reaction mechanism model, three key parameters—initiation efficiency, dispersion stability, and thermodynamic state—are extracted from a standardized dataset to form a set of mechanism indices. For each parameter category, the extraction methods and rationale are elaborated in detail, delving into the physicochemical essence behind the data. This ensures that the extracted indices accurately reflect the key characteristics of the polymerization reaction, providing more fundamental and accurate input for the prediction model and enhancing its ability to explain and predict the reaction process.

[0079] Fractional-order kinetics can more accurately describe the sub-diffusion behavior of the reaction rate, while non-equilibrium thermodynamic coupling considers the influence of the heat of reaction on the conversion rate. The combination of the two enables the model to more accurately predict the conversion rate change curve over time and the reaction time, providing a more reliable basis for optimal control.

[0080] Using the conversion rate prediction curve and reaction time prediction as optimization objectives, a gradient-based mathematical optimization method is employed to optimize the raw material ratio and operating parameters. Key influencing factors are identified through sensitivity analysis, and parameters are iteratively adjusted along the negative gradient direction until the termination condition is met. This method can quickly and accurately find the optimal parameter combination, ensuring stable conversion rate and product quality while shortening reaction time, thereby improving production efficiency and product competitiveness.

[0081] The optimized formula and operating condition parameter set are converted into instructions that can be executed by DCS and ESS. Detailed instruction conversion rules are formulated, including formula parameter conversion and operating condition parameter conversion. Safety constraints, emergency termination conditions and venting control are integrated to realize automated closed-loop control of production adjustment, improve the stability and safety of the production process, reduce human intervention and operational errors, and reduce production costs and safety risks.

[0082] Please see Figure 2 Another embodiment of the polymerization reactor reaction prediction and optimization system of the present invention includes:

[0083] 201. Acquisition module, used for data standardization processing: acquire raw material addition data, auxiliary agent dosage data and reaction condition data from the historical database of polymerization reaction, perform quality verification and normalization processing on the data, and generate a standardized input dataset;

[0084] Specifically, the data acquisition steps are as follows: Raw material dosage records, additive dosage time series data, and reactor operating condition monitoring data are extracted from multiple data tables in the historical database of polymerization reactions to generate a raw multi-source data set; the data verification steps involve performing integrity checks and outlier detection on the raw multi-source data set, filtering the data based on the physical probability range, and generating a validated valid dataset; the physical dimension analysis steps involve performing dimensional analysis on each parameter in the valid dataset, identifying the physical dimension type of each parameter, and generating a parameter dimension classification table; the dimensionless processing steps involve selecting appropriate feature parameters to perform dimensionless calculations on each parameter according to the parameter dimension classification table, generating a dimensionless parameter set; the composite parameter generation steps involve combining and calculating related parameters in the dimensionless parameter set based on the polymerization reaction mechanism to generate a dimensionless composite parameter set with clear physical meaning; and the dataset assembly steps involve organizing the dimensionless composite parameter set according to time series and reaction batches to generate the final standardized input dataset.

[0085] It should be noted that the following is a data standardization example using the styrene polymerization reaction as an example: The following raw data (example values) were extracted from the historical database: Raw material dosage: styrene monomer (1000 kg), initiator benzoyl peroxide (5 kg); Additive timing data: dispersant polyvinyl alcohol (added at minute 0, concentration 0.5%); Reactor conditions: temperature (gradual increase from 70℃ to 85℃), pressure (0.8 MPa), stirring rate (200 rpm);

[0086] Data validation and integrity check: Detect missing values ​​(temperature records at a certain time point are missing and filled using linear interpolation); Outlier detection: Remove data that exceeds the physical range (pressure drops suddenly to negative values ​​or exceeds the equipment's upper limit of 2.0 MPa).

[0087] Physical dimension analysis, classification parameters: mass type: raw material dosage (kg), additive concentration (%); thermodynamic type: temperature (K), pressure (Pa); kinetic type: stirring rate (rpm), reaction time (s).

[0088] Dimensionless processing, using Min-Max normalization to map parameters to the [0,1] range: Temperature normalization: (Example value: 75℃→0.5); Pressure normalization: (Example value: 0.8MPa→1.0);

[0089] Composite parameter generation, based on the reaction mechanism, constructs dimensionless combined parameters: initiation efficiency coefficient = (initiator concentration × temperature normalized value) / impurity content; dispersion stability index = (dispersant concentration × stirring rate normalized value) / interfacial tension parameter;

[0090] The dataset was assembled by organizing the data according to reaction batch (Batch#20230901) and time series (every 5-minute interval), generating a structured input dataset, the format of which is shown in Table 4 below:

[0091] Table 4 Organizational Data

[0092]

[0093] The final output is a standardized input dataset, which is used for subsequent extraction of mechanistic indicators.

[0094] 202. Extraction module, used for index extraction: Based on the polymerization reaction mechanism model, the standardized input dataset is subjected to physical mechanism analysis to extract initiation efficiency parameters, dispersion stability parameters and thermodynamic state parameters, forming a set of mechanism indices;

[0095] Specifically, the initiation efficiency analysis steps are as follows: Based on initiator decomposition kinetics and chain initiation mechanisms, the initiator concentration, temperature history, and impurity content data in the standardized input dataset are analyzed to calculate the effective free radical concentration and initiation rate parameters, generating an initiation efficiency parameter set; the dispersion stability analysis steps are as follows: Based on interfacial chemistry and colloidal stability theory, the dispersant characteristics, stirring conditions, and phase interface parameters in the standardized input dataset are processed to evaluate the dispersion phase stability and particle distribution characteristics, generating a dispersion stability parameter set; the thermodynamic state analysis steps are as follows: Applying the principles of non-equilibrium thermodynamics, the temperature gradient, pressure change, and energy balance data in the standardized input dataset are analyzed to calculate the system driving force and thermodynamic flow, generating a thermodynamic state parameter set; the transport characteristic analysis steps are as follows: Based on momentum, mass, and energy transport theories, the viscosity change, diffusion coefficient, and heat transfer characteristic data in the standardized input dataset are processed to evaluate transport resistance and efficiency, generating a transport characteristic parameter set; the index integration step is as follows: The initiation efficiency parameter set, dispersion stability parameter set, thermodynamic state parameter set, and transport characteristic parameter set are organically integrated to establish the correlation between parameters and form a complete set of mechanistic indexes.

[0096] It should be noted that, based on the decomposition kinetics of the initiator benzoyl peroxide, the initiator concentration (0.05 mol / L), temperature history (70℃→85℃ heating curve), and impurity content (water content <0.01%) in the standardized dataset were analyzed. The effective free radical concentration (2.1×10⁻⁶) was calculated using the Arrhenius equation. -4mol / L) and initiation rate parameter (1.8 × 10⁻⁶) -3 s -1 This generates a set of initiation efficiency parameters, including free radical generation rate and half-life indices.

[0097] Based on the theory of colloidal stability, the concentration of the dispersant polyvinyl alcohol (0.5%), stirring rate (200 rpm), and interfacial tension data (35 mN / m) were processed. The Hamaker constant (4.2 × 10⁻⁶) was calculated. -20 The dispersion stability (sedimentation rate < 0.1% / h) and particle size distribution (D50 = 1.2 μm) were evaluated using the J) and critical coagulation concentration, and a set of dispersion stability parameters, including Zeta potential and flocculation risk index, were generated.

[0098] Applying the principles of nonequilibrium thermodynamics, the temperature gradient (temperature difference between jacket and core ±0.5℃), pressure fluctuation (0.8±0.05MPa), and energy balance data (heating power to heat loss ratio) of the reactor were analyzed. The entropy production rate (0.12W / K) and thermodynamic flow (reaction heat release rate 85kJ / min) of the system were calculated, generating a set of thermodynamic state parameters, including Gibbs free energy change and dissipation function.

[0099] Based on mass and heat transfer theory, this study addresses viscosity changes (from an initial 10 cP to a final 3500 cP) and diffusion coefficients (monomer diffusion coefficient 2.7 × 10⁻⁶). -10 m 2 / s) and heat transfer efficiency (overall heat transfer coefficient 450W / m 2 •K). Evaluate transmission resistance (Nuser number Nu=18.5) and mixed efficiency (Schmidt number Sc=1200) to generate a set of transmission characteristic parameters, including Reynolds number and Prandtl number.

[0100] The above parameter set is organically integrated to establish the relationship between initiation efficiency and thermodynamic state (the sensitivity coefficient of temperature to free radical concentration is 0.15), the coupling relationship between dispersion stability and transport characteristics (the inhibition factor of viscosity on particle settling rate is 0.8), and finally form a set of mechanistic indicators containing 12 core indicators for subsequent prediction model construction.

[0101] By analyzing physical mechanisms in multiple dimensions, the raw data is transformed into quantifiable mechanistic indicators, providing an input basis for dynamic prediction.

[0102] 203. Construction module for predictive model establishment: Using the set of mechanistic indicators as input variables, construct a mechanistic predictive model based on the coupling of fractional-order kinetics and non-equilibrium thermodynamics, determine the model coefficients through parameter identification, and generate conversion rate prediction curves and reaction time prediction values;

[0103] Specifically, the model structure design steps are as follows: Based on the principles of mass and energy conservation in polymerization reactions, a basic model framework containing fractional differential terms and non-equilibrium thermodynamic coupling terms is designed, generating a model structure description file; Fractional operator introduction step: Based on the memory characteristics and historical path dependence of the reaction process, fractional differential operators are introduced into the model structure description file to establish kinetic equations with memory effects, generating a fractional kinetic model; Thermodynamic coupling processing step: Based on non-equilibrium thermodynamics theory, heat-mass transfer coupling terms are introduced into the fractional kinetic model to establish a synergistic relationship between energy and mass transfer, generating a set of coupled model equations; Parameter identification step: Using the set of mechanistic indices and experimental data of the reaction process, a numerical optimization method based on physical constraints is used to determine the undetermined coefficients in the model, generating a calibrated set of prediction model parameters; Prediction generation step: Substituting the calibrated set of prediction model parameters into the set of coupled model equations, numerical solutions are performed to generate a predicted curve of conversion rate versus time and a predicted value of the time required for the reaction to reach a specified conversion rate.

[0104] It should be noted that the basic model framework is constructed based on the principles of mass and energy conservation. Taking styrene polymerization as an example, the model includes a monomer mass balance equation (describing the styrene consumption rate) and a reaction heat balance equation (covering the exothermic polymerization and jacket heat transfer). The model framework specifically incorporates fractional derivative terms to describe the reaction kinetic memory effect, and non-equilibrium thermodynamic coupling terms to express the influence of temperature gradients on mass transfer.

[0105] To address the path dependence of styrene polymerization, a Caputo-type fractional differential operator (derivative order set to 0.85) is introduced into the kinetic equation. This operator characterizes the influence of historical concentrations on the current reaction rate through a weighting function, enabling the model to accurately reflect the delayed effect of initiator decomposition and the memory characteristics of free radical chain growth.

[0106] Based on nonequilibrium thermodynamics, a heat-mass transfer coupling term is introduced into the fractional-order kinetic model to establish the correlation equation between the temperature gradient (the difference between the reaction core region (85℃) and the wall region (82℃)) and the monomer diffusion flux. The coupling coefficient is determined by minimizing the entropy yield, which characterizes the driving effect of energy flow on mass transfer.

[0107] Using a set of mechanistic indices (initiation efficiency parameter 0.92, dispersion stability index 1.15) and historical reaction data (conversion-time curves for 10 batches), a physical constraint optimization algorithm was employed for parameter identification. Key identified parameters included: fractional derivative coefficient (0.15), coupling strength factor (0.42), and activation energy (85 kJ / mol). After optimization, the average relative error between the model and experimental data was reduced to 3.2%.

[0108] Substituting the calibrated parameter set into the coupled model, the numerical solution yields the prediction results. For the baseline condition (starting at 70℃ and rapidly increasing to 85℃), the model outputs: 95% conversion rate is achieved in 215 minutes, and the reaction curve deviates from the experimental data by less than 4%. Simultaneously, the predicted reaction time can be shortened to 195 minutes under optimal conditions, providing a quantitative basis for optimization.

[0109] 204. Setting module for optimization calculation: Taking the conversion rate prediction curve and reaction time prediction value as optimization targets, a mathematical optimization method based on gradient analysis is used to optimize the raw material ratio and operating parameters, and generate a future formula parameter set and a future operating condition parameter set.

[0110] Specifically, the optimization objective setting steps are as follows: Based on the conversion rate prediction curve, the final conversion rate value is extracted. Combined with the reaction time prediction value, a composite objective function is established with the goal of maximizing the final conversion rate and minimizing the reaction time, generating a multi-objective optimization problem statement. The constraint establishment steps are as follows: Based on equipment operation limitations and product quality requirements, the feasible range of raw material ratios and operating parameters is determined, establishing a set of multiple constraints including safety constraints, quality constraints, and equipment constraints. The sensitivity analysis steps are as follows: Using gradient calculation technology based on the adjoint equation method, the sensitivity of the composite objective function to each decision variable is analyzed, generating a complete gradient information matrix. The iterative optimization calculation steps are as follows: Using the gradient information matrix as the search direction, under the constraints of the multiple constraint set, iterative optimization calculations are performed using a sequential quadratic programming algorithm to generate preliminary optimization parameter solutions. The feasibility verification steps are as follows: Substituting the preliminary optimization parameter solutions into the mechanism prediction model for verification calculations, evaluating its feasibility and robustness in actual production, and finally generating an implementable future formula parameter set and future operating condition parameter set.

[0111] It should be noted that a composite objective function is established based on the conversion rate curve (final conversion rate 92%) and the predicted reaction time (215 minutes) output by the prediction model:

[0112] ;

[0113] The conversion rate weighting coefficient is a constant between 0 and 1, representing the importance of the conversion rate X in the overall target. It is subjectively set by process engineers or decision-makers based on production needs and economic benefits. α is the reaction time weighting coefficient, a constant between 0 and 1, representing the importance of reaction time T in the overall objective. It is set together with α and satisfies α+β= 1. Actual conversion rate, the percentage of raw materials converted into the target product during a chemical reaction. This is a key process performance indicator that directly affects yield and raw material costs. The target conversion rate is the ideal maximum conversion rate (95%). This is the actual reaction time, the total time (in minutes) required to complete the reaction and achieve the expected conversion rate. This is an efficiency indicator that directly affects equipment utilization and energy consumption. The target reaction time is the shortest reaction time that is ideally achieved (200 minutes).

[0114] Where α = 0.7 (conversion rate weight), β = 0.3 (duration weight), and X is the actual conversion rate. =95% (target conversion rate), T is the actual reaction time. =200 minutes (target duration). This function quantifies the dual objectives of improving conversion rates and shortening duration.

[0115] Constraints are set according to equipment limitations and product quality requirements: Safety constraints: upper limit of reaction temperature 90℃ (explosion-proof polymerization), upper limit of pressure 1.2MPa; Quality constraints: polymer molecular weight must be maintained at 100,000-150,000 g / mol (grade requirements); Equipment constraints: stirring speed range 150-250 rpm (mechanical limitation), initiator dosage range 4-6 kg (metering pump capacity).

[0116] The gradient of the objective function with respect to decision variables was calculated using the adjoint equation method: For every 1°C increase in reaction temperature, the conversion rate increases by 0.8% but the reaction time decreases by 2 minutes; an increase of 0.1 kg in initiator dosage increases the conversion rate by 0.5% but the molecular weight decreases by 5000 g / mol; an increase of 10 rpm in stirring speed improves the dispersion stability index by 0.15 but increases energy consumption by 5%.

[0117] Starting with the initial operating conditions (temperature 85℃, initiator 5kg, stirring 200rpm), a sequential quadratic programming algorithm was used for 5 rounds of iteration: the 3rd round yielded a preliminary solution: temperature 88℃, initiator 5.3kg, stirring 220rpm; under this solution, the predicted conversion rate was 94.5%, the reaction time was 205 minutes, and the molecular weight was 138000g / mol;

[0118] The preliminary solution was substituted into the mechanism prediction model for verification: the actual conversion rate reached 94.2% (error 0.3%), the reaction time was 207 minutes; the molecular weight of 135000 g / mol met the mass constraint, and the temperature and pressure were within the safe range; the final output optimized parameter set was: initiator dosage 5.3 kg, temperature curve (0-60 min to 88℃ and maintain), and stirring speed 220 rpm.

[0119] 205. Allocation module, used for control instruction generation: converting the future formula parameter set and future operating condition parameter set into execution instructions, and outputting them to the polymerization reactor control system to implement production adjustments;

[0120] Specifically, the control command encoding step involves: based on the communication protocol and interface specifications of the polymerization reactor control system, converting the future formula parameter set and future operating condition parameter set into a control command format recognizable by the equipment, generating a preliminary control command set; the safety constraint verification step involves: according to the safe operating range and process limitations of the polymerization reactor equipment, performing safety verification and feasibility assessment on the preliminary control command set, generating a safety-verified control command set; the command priority sorting step involves: according to the importance and urgency of the production process, prioritizing the safety-verified control command set, generating an ordered control command sequence; the command packaging and transmission step involves: according to the data receiving requirements of the polymerization reactor control system, packaging the ordered control command sequence into a data packet of a specific format and transmitting it to the polymerization reactor control system through an industrial communication network; the execution status monitoring step involves: real-time monitoring of the execution status of the polymerization reactor control system in response to received commands, collecting feedback on key parameters of the actual production process, and generating a command execution status report; and the closed-loop adjustment step involves: based on the deviation between the actual production data and the expected target in the command execution status report, dynamically adjusting the parameter settings of subsequent control commands to achieve closed-loop optimization control of the production process.

[0121] It should be noted that the following is a specific embodiment of the "control instruction generation" step, taking the styrene polymerization reaction as an example:

[0122] The optimized parameter set (initiator dosage 5.3 kg, temperature curve linearly increasing to 88℃ from 0-60 min, stirring speed 220 rpm) was converted into equipment-recognizable commands. Based on the Ethernet / IP protocol supported by the polymerization reactor control system, the temperature parameter was encoded as a floating-point data packet (TSET=88.0), the stirring speed as an integer command (AGITSPD=220), and the initiator dosage was converted into a volume control command (INIV=5.3), generating a preliminary set of 12 commands.

[0123] Safety verification of the commands: the temperature command of 88℃ did not exceed the equipment's upper limit of 90℃; the stirring rate of 220rpm was within the motor's allowable range (150-250rpm); the initiator dosage of 5.3kg was within the metering pump capacity range (4-6kg); all commands passed safety verification and did not trigger any equipment protection limits.

[0124] Based on the criticality of the production process, the priority order is as follows: Priority 1: Temperature control command (directly affects reaction safety and rate); Priority 2: Initiator addition command (affects molecular weight distribution); Priority 3: Stirring rate command (affects dispersion stability); and an ordered command sequence is generated: Temperature setting → Initiator addition → Stirring adjustment.

[0125] Ordered instructions are packaged into explicit message data packets (including timestamps, instruction IDs, and parameter values) according to the Ethernet / IP protocol and transmitted to the polymerization reactor PLC controller via an industrial switch. The data packets include a header checksum (CRC32) to ensure integrity, and the transmission period is set to 100ms.

[0126] Real-time monitoring of command execution status: actual temperature 87.9℃ (0.1% deviation from set value); actual initiator dosage 5.28kg (0.4% deviation); stirring speed 219rpm (0.5% deviation); generating execution status report and marking all commands as "normal execution".

[0127] The system dynamically adjusts subsequent instructions based on real-time feedback data: if the temperature deviation is consistently lower than the set value by 0.2℃, it automatically compensates by 0.3℃ and generates the adjustment instruction "TSET=88.3"; if the initiator addition deviation is within the tolerance range, it maintains the original parameters to achieve closed-loop optimization control based on the actual response, ensuring that production accurately tracks the optimization target.

[0128] 206. Update Module: Used for real-time data acquisition, real-time parameter set optimization, and knowledge base updates. Through a sensor network installed on the polymerization reactor, it collects real-time data on temperature, pressure, viscosity, and component concentration during the reaction process, generating a real-time process state dataset. Dynamic Deviation Analysis Step: The real-time process state dataset is compared with the conversion rate prediction curve to calculate the dynamic deviation between the actual reaction process and the prediction model, generating a process deviation index set. Model Parameter Correction Step: Based on the process deviation index set, a mechanism-based adaptive parameter adjustment algorithm is used to correct the parameters of the mechanism prediction model online, generating an updated model parameter set. Real-time Optimization Recalculation Step: Using the updated model parameter set, the optimization calculation process is re-executed to generate a real-time optimized parameter set adapted to the current reaction state. Expert Rule Intervention Step: Based on the polymerization reaction expert experience knowledge base, the real-time optimized parameter set is judged for rationality and necessary adjustments are made, generating a final control parameter set verified by experts. Knowledge Base Update Step: Key data, deviation analysis results, and parameter adjustment records from this optimization process are saved to the expert experience knowledge base, enabling continuous accumulation and self-improvement of system knowledge.

[0129] It should be noted that data is collected in real time through a sensor network installed in the polymerization reactor:

[0130] Temperature: Four PT100 temperature sensors (upper, middle, lower parts of the reactor and jacket), sampling frequency 1 time / second, accuracy ±0.1℃ (monitored core temperature 88.2℃, jacket temperature 85.5℃); Pressure: Piezoresistive pressure sensor, range 0-2MPa, real-time value 0.82MPa; Viscosity: Online rotational viscometer, monitoring range 10-5000cP, current value 1250cP; Concentration: Near-infrared spectrometer for real-time analysis of styrene monomer concentration (current value 42.5%); Generate a process status dataset containing 15 parameters, updated every 5 seconds.

[0131] Compare the real-time dataset with the predicted curve: Conversion rate deviation: Actual value 78.2% vs. predicted value 80.5% (deviation -2.3%); Temperature rise rate: Actual 0.35℃ / min vs. predicted 0.32℃ / min (deviation +9.4%); Viscosity change: Actual 1250 cP vs. predicted 1180 cP (deviation +5.9%); Generate a process deviation index set containing 7 indicators, marking the reaction rate as too fast and the heat transfer efficiency as too low.

[0132] An adaptive adjustment algorithm is adopted based on the deviation index: the fractional derivative coefficient is lowered (from 0.85 to 0.82); the heat-mass coupling coefficient is corrected (from 0.42 to 0.38); the activation energy parameter is updated (from 85 kJ / mol to 87 kJ / mol); and an updated model parameter set is generated, which improves the matching degree between the predicted curve and the actual curve to over 95%.

[0133] The updated model was re-optimized: the initiator dosage was slightly adjusted from 5.3 kg to 5.25 kg; the temperature setting was adjusted from 88℃ to 87.5℃; the stirring speed was increased from 220 rpm to 225 rpm; a real-time optimized parameter set adapted to the current reaction state was generated, and the predicted reaction time could be shortened to 202 minutes.

[0134] Reasonableness verification was performed based on the expert knowledge base: confirming that the temperature adjustment was not below the safety threshold (>70℃); verifying that the stirring rate adjustment was within the equipment limit (<250rpm); assessing the impact of molecular weight (expected to increase from 135,000 to 138,000g / mol); generating the final set of control parameters: initiator 5.25kg, temperature 87.5℃, stirring 225rpm.

[0135] Save the key data from this optimization to the expert knowledge base: record the deviation analysis results (temperature sensitivity coefficient 0.15→0.18); save the parameter adjustment records (fractional coefficient adjustment impact score 8.5 / 10); archive the actual response curve (final conversion rate 94.8%, duration 204 minutes); and achieve continuous accumulation of system knowledge.

[0136] In this invention embodiment, a complete processing flow including data verification, dimensional analysis, dimensionless transformation, and composite parameter generation is constructed, solving the problem of multi-source heterogeneous data fusion. Verified through a styrene polymerization example, the system can effectively handle missing values ​​(temperature record interpolation) and outliers (negative pressure rejection), and establish a parameter dimensional classification table, providing a high-quality data foundation for subsequent analysis. Breaking through the limitations of traditional kinetic models, a physically meaningful hybrid model is constructed by combining fractional differential operators (memory characteristics) with non-equilibrium thermodynamic coupling terms (energy-mass transport synergy). Taking styrene polymerization as an example, the model accurately characterizes the initiator decomposition delay effect through the Caputo operator, and the coupling coefficient is determined based on the principle of minimizing entropy yield. A dynamic optimization system is constructed through real-time data acquisition and adaptive adjustment of model parameters. The continuous updating mechanism of the expert knowledge base enables the system to have self-learning capabilities; by recording the influence score of fractional coefficient adjustments, the control strategy can be continuously optimized. Through the deep integration of physical mechanisms and data-driven approaches, not only is accurate prediction and dynamic optimization of the polymerization reaction process achieved, but also a sustainably evolving intelligent control platform is constructed.

[0137] Figure 3 This is a schematic diagram of a polymerization reactor reaction prediction and optimization device 300 provided in an embodiment of the present invention. The device can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the polymerization reactor reaction prediction and optimization device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the polymerization reactor reaction prediction and optimization device 300.

[0138] The polymerization reactor reaction prediction and optimization device 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, MacOSX, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3The illustrated structure of the polymerization reactor reaction prediction and optimization equipment does not constitute a limitation on the polymerization reactor reaction prediction and optimization equipment. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0139] The present invention also provides a polymerization reactor reaction prediction and optimization device, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the polymerization reactor reaction prediction and optimization system in the above embodiments.

[0140] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the polymerization reactor reaction prediction and optimization system.

[0141] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0142] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0143] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A system for predicting and optimizing polymerization reactor reactions, characterized in that, The polymerization reactor reaction prediction and optimization system includes: The acquisition module is used to acquire and process raw material addition data, auxiliary agent dosage data, and reaction condition data to generate a standardized input dataset. The extraction module is used to extract priming efficiency parameters, dispersion stability parameters, and thermodynamic state parameters based on the standardized input dataset to form a set of mechanistic indicators; The module is used to construct a mechanism prediction model based on the coupling of fractional-order kinetics and non-equilibrium thermodynamics by taking the set of mechanism indicators as input variables, determining the model coefficients through parameter identification, and generating conversion rate prediction curves and reaction time prediction values. The setting module is used to process the raw material ratio and operating parameters with the conversion rate prediction curve and reaction time prediction value as optimization targets, and generate a future formula parameter set and a future operating condition parameter set.

2. The polymerization reactor reaction prediction and optimization system according to claim 1, characterized in that, The acquisition module is used for: Extract raw material addition records, additive addition timing data, and reactor operating condition monitoring data from multiple data tables in the historical database of polymerization reactions to generate a raw multi-source data set; The original multi-source dataset is subjected to integrity checks and outlier detection, and the data is filtered based on the range of physical probability to generate a valid dataset. Dimensional analysis is performed on each parameter in the effective dataset to identify the physical dimension type of each parameter, generate a parameter dimension classification table, and generate a dimensionless parameter set based on the parameter dimension classification table; The relevant parameters in the dimensionless parameter set are combined and calculated to generate a dimensionless composite parameter set. The dimensionless composite parameter set is organized according to time series and reaction batches to generate a standardized input dataset.

3. The polymerization reactor reaction prediction and optimization system according to claim 2, characterized in that, The extraction module is used to organically integrate the set of initiation efficiency parameters, the set of dispersion stability parameters, the set of thermodynamic state parameters, and the set of transport characteristic parameters, establish the correlation between the parameters, and form a set of mechanistic indicators.

4. The polymerization reactor reaction prediction and optimization system according to claim 3, characterized in that, In the extraction module, the initiator concentration, temperature history, and impurity content data in the standardized input dataset are analyzed to calculate the effective free radical concentration and initiation rate parameters, and to generate an initiation efficiency parameter set. Process the dispersant properties, stirring conditions and phase interface parameters in the standardized input dataset, evaluate the stability of the dispersed phase and particle distribution characteristics, and generate a set of dispersion stability parameters; Analyze the temperature gradient, pressure change and energy balance data in the standardized input dataset, calculate the system driving force and thermodynamic flow, and generate a set of thermodynamic state parameters; Process viscosity variation, diffusion coefficient and heat transfer characteristic data in the standardized input dataset, evaluate transport resistance and efficiency, and generate a set of transport characteristic parameters.

5. The polymerization reactor reaction prediction and optimization system according to claim 3, characterized in that, Build modules are used for: Design a basic model framework that includes fractional differential terms and non-equilibrium thermodynamic coupling terms, and generate a model structure description file; Fractional differential operators are introduced into the model structure description file to establish dynamic equations with memory effect and generate a fractional dynamic model. A heat-mass transfer coupling term is introduced into the fractional-order kinetic model to establish the synergistic relationship between energy and mass transfer, generating a set of coupled model equations; Based on the set of mechanistic indicators and experimental data of the reaction process, a numerical optimization method based on physical constraints is used to determine the undetermined coefficients in the mechanistic prediction model and generate a calibrated prediction model parameter set. The calibrated prediction model parameter set is substituted into the coupled model equation set for numerical solution calculation, generating conversion rate prediction curves and reaction time prediction values.

6. The polymerization reactor reaction prediction and optimization system according to claim 5, characterized in that, The settings module is used for: Based on the conversion rate prediction curve, the final conversion rate value is extracted. Combined with the reaction time prediction value, a composite objective function is established with the goal of maximizing the final conversion rate and minimizing the reaction time, generating a multi-objective optimization problem statement. Based on equipment operation limitations and product quality requirements, determine the feasible range of raw material ratios and operating parameters, and establish a set of multiple constraint conditions; Analyze the sensitivity of the composite objective function to each decision variable and generate a complete gradient information matrix; Using the gradient information matrix as the search direction, iterative optimization calculations are performed under the constraints of the multiple constraint set to generate preliminary optimization parameter solutions; The preliminary optimized parameter solution is substituted into the mechanism prediction model for verification calculation, generating a future formula parameter set and a future operating condition parameter set.

7. The polymerization reactor reaction prediction and optimization system according to claim 6, characterized in that, In the configuration module, a composite objective function is established with the goal of maximizing the final conversion rate and minimizing the reaction time. The composite objective value is... : ； in, This is the conversion rate weighting coefficient; The reaction time weighting coefficient satisfies α+β= 1; This represents the actual conversion rate. Target conversion rate; This refers to the actual reaction time. The target reaction time.

8. The polymerization reactor reaction prediction and optimization system according to claim 6, characterized in that, It also includes an allocation module, used to obtain execution instructions based on the future recipe parameter set and the future operating condition parameter set: The future formula parameter set and future operating condition parameter set are converted into a control instruction format that the equipment can recognize, and a preliminary control instruction set is generated. The initial set of control instructions is subjected to security verification and feasibility assessment to generate a new set of control instructions. The control instruction set is prioritized and sorted to generate an ordered sequence of control instructions; The ordered control instruction sequence is packaged into a data packet of a specific format and transmitted to the polymerization reactor control system via an industrial communication network; The system monitors the execution status of received commands in real time, collects feedback on key parameters of the actual production process, and generates a command execution status report.

9. The polymerization reactor reaction prediction and optimization system according to claim 1, characterized in that, It also includes an update module for real-time data collection, real-time optimization of parameter sets, and updating of the knowledge base. Real-time acquisition of temperature, pressure, viscosity, and component concentration data during the reaction process generates a real-time process status dataset; The real-time process status dataset is compared and analyzed with the conversion rate prediction curve to calculate the dynamic deviation between the actual reaction process and the prediction model, and to generate a process deviation index set. Based on the process deviation index set, the parameters of the mechanism prediction model are corrected online to generate an updated model parameter set; Using the updated model parameter set, the optimization calculation process is re-executed to generate a real-time optimization parameter set; Based on the knowledge base of polymerization reaction experts, the real-time optimization parameter set is judged for reasonableness and adjusted as necessary to generate the final control parameter set.

Images