Chemical process optimization system and method based on serial integration models

Abstract

The invention discloses a chemical process optimization system and method based on serial integration models. The chemical process optimization system comprises: a big data database construction module, which constructs a big data database by receiving historical data from chemical production sites and integrating it with a chemical theory model; a serial integration artificial intelligence training module, which includes a first AI training model based on linear regression and a second AI training model based on nonlinear regression, used for model training during the training phase and for outputting product parameters during the prediction phase based on process parameters and intermediate parameters; and a profit maximization module embedded with a nature-inspired algorithm, which calculates production decisions.

Description

Chemical Process Optimization System and Method Based on Series Integration Model This invention relates to a chemical process optimization system, and more particularly to a chemical process optimization system and method based on a series integration model. In the chemical industry, syngas is a key intermediate in the production of many chemical products, including ammonia, methanol, hydrogen, and fuels, and its industrial applications are extremely widespread. Currently, traditional syngas production processes typically rely on manual experience to adjust operating parameters. However, due to the nonlinear nature of chemical processes, the high dimensionality of operating parameters, and the influence of external factors such as raw material costs and market demand fluctuations, traditional adjustment methods suffer from the following significant problems: Insufficient timeliness: General operators have difficulty grasping the fluctuations in market demand, raw material prices, and utility fluid prices in a timely manner, making it difficult to determine the optimal operating conditions in a timely manner. Accuracy limitations: Chemical processes often involve nonlinear and interactive parameter relationships, and it is impossible to accurately judge the impact of these complex factors on yield and quality based solely on human experience. High resource consumption: Due to the lack of precise condition optimization mechanisms, excessive use of energy such as fuel and steam is often caused, increasing cost expenditures and waste. Lack of intelligent technology application: Traditional methods have not yet widely integrated chemical simulation software and big data analysis technology, and cannot provide real-time and scientific best operation advice. In summary, the inventor, recognizing that the aforementioned deficiencies could be improved, devoted himself to research and applied theoretical principles, and finally proposed an invention with a reasonable design that effectively improves the aforementioned deficiencies. This invention discloses a chemical process optimization system and method based on a series integration model, which is mainly used to improve the deficiencies of existing technologies. One embodiment of the present invention discloses a chemical process optimization system based on a cascaded integrated model, comprising: a big data database construction module, which has a built-in chemical engineering theoretical model and is configured in a database construction operation environment to receive historical data from the chemical production site, and performs data fitting between the historical data and the chemical engineering theoretical model to construct a big data database; a cascaded artificial intelligence training module, which is connected to the big data database and trains the model using training data from the big data database during a training phase; wherein the cascaded artificial intelligence training module includes: a first artificial intelligence training model constructed based on a linear regression algorithm; and a second artificial intelligence training model constructed based on a nonlinear regression algorithm; wherein, the The first artificial intelligence training model receives a process parameter from an external input during a prediction phase and outputs an intermediate parameter; wherein, the second artificial intelligence training model receives the process parameter and further receives the intermediate parameter during the prediction phase, thereby outputting a prediction result; and a benefit maximization module, which embeds a natural heuristic algorithm; wherein, the benefit maximization module is data-connected to the serial artificial intelligence training module, and the benefit maximization module is configured to further import an economic parameter from an external input; the benefit maximization module is configured to use the natural heuristic algorithm, based on the demand output of a product, combined with the economic parameter and the prediction result, and incorporating operational upper and lower limit factors, thereby outputting a production decision. Preferably, the chemical process optimization system further includes: a visualization interface module, which is configured to allow users to input product demand in real time and visualize the production decisions of the benefit maximization module. Preferably, the chemical engineering theoretical model is a chemical engineering simulation software with physical meaning, and is at least one of Aspen Plus, Aspen HYSYS, PRO / II, and Petro-SIM; wherein the historical data is historical production data from the chemical production site, and the training data includes process data, intermediate data, and product data. Preferably, the first artificial intelligence training model employs at least one of Ordinary Least Squares (OLS), Gradient Descent, and Regularized Linear Regression. Preferably, the second artificial intelligence training model employs at least one of the following: Polynomial Regression, Support Vector Regression (SVR), Decision Tree Regression, Random Forest Regression, Neural Network Regression, Nonlinear Least Squares, Gaussian Process Regression (GPR), and XGBoost Regressor. Preferably, the natural heuristic algorithm employs at least one of the following: Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), and Simulated Annealing (SA). Preferably, the serial artificial intelligence training module is configured in the training phase to train and cross-validate the first artificial intelligence training model and the second artificial intelligence training model respectively using the training data from the big data database. Preferably, in the serial artificial intelligence training module, the process parameters include at least one of the following: light oil feed rate, steam consumption, reaction temperature, reaction pressure, and carbon dioxide reflux rate; wherein, the intermediate parameters include at least one of the reaction rate and reaction conversion rate; wherein, the prediction results include at least one of the product yield and productivity. Preferably, the economic parameters include revenue parameters and expenditure parameters; wherein the revenue parameter is the product selling price, and the expenditure parameter includes raw material prices and energy costs. Preferably, the benefit maximization module analyzes and outputs the production decision based on a profit maximization function or a cost minimization function through a loop iteration method. Another embodiment of the present invention discloses a chemical process optimization method based on a cascaded integrated model, comprising: using a big data database construction module with a built-in chemical theory model in a database construction operation environment, receiving historical data from a chemical production site, and fitting the historical data with the chemical theory model to construct a big data database; connecting a cascaded artificial intelligence training module to the big data database, and training the model using training data from the big data database during a training phase; wherein the cascaded artificial intelligence training module includes a first artificial intelligence training model and a second artificial intelligence training model connected in series; wherein the first artificial intelligence training model is constructed based on a linear regression algorithm, and the... The second artificial intelligence training model is constructed based on a nonlinear regression algorithm; wherein, the first artificial intelligence training model receives an externally input process parameter in a prediction stage and outputs an intermediate parameter; wherein, the second artificial intelligence training model receives the process parameter in the prediction stage and further receives the intermediate parameter, thereby outputting a prediction result; and the serial artificial intelligence training module is connected to a benefit maximization module embedded with a natural heuristic algorithm, and further imports an externally input economic parameter; wherein the benefit maximization module is configured to output a production decision based on the natural heuristic algorithm, according to the demand output of a product, combined with the economic parameter and the prediction result, and incorporating operational upper and lower limit factors. Preferably, the chemical process optimization method further includes: providing users with real-time input of product demand through a visual interface module, and visually presenting the production decisions of the benefit maximization module. This invention utilizes a chemical process optimization system based on a cascaded integrated model, combining big data, artificial intelligence, and optimization algorithms to achieve accurate modeling of chemical processes and maximize economic benefits. The collaborative operation of each module addresses problems in traditional chemical processes such as data gaps, inaccurate modeling, and low decision-making efficiency. To further understand the features and technical content of this invention, please refer to the following detailed description and accompanying drawings. However, these descriptions and drawings are only for illustrating the invention and are not intended to limit the scope of protection of the invention in any way. In the following description, if it is indicated that a specific diagram is referred to or as shown in a specific diagram, it is only to emphasize that most of the relevant content mentioned in the following description appears in that specific diagram, but does not limit the following description to refer only to that specific diagram. The following specific embodiments illustrate the implementation methods disclosed in this invention. Those skilled in the art can understand the advantages and effects of this invention from the content disclosed in this specification. This invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed based on different viewpoints and applications without departing from the concept of this invention. It should be understood that although terms such as "first," "second," and "third" may be used herein to describe various elements or signals, these elements or signals should not be limited by these terms. These terms are primarily used to distinguish one element from another, or one signal from another. Furthermore, the term "or" as used herein may, depending on the specific circumstances, include any combination of one or more of the associated listed items. [Chemical Process Optimization System Based on Series Integration Model] Referring to Figure 1, this embodiment of the invention provides a chemical process optimization system 100 based on a cascaded integrated model, which can be used to provide production decisions for chemical process optimization and maximize economic benefits. The chemical process optimization system 100 of this embodiment can be used to optimize production decisions for synthesis gas chemical processes, but the invention is not limited thereto. To achieve the above objectives, the chemical process optimization system 100 based on the serial integration model of the present invention includes a big data database construction module 1, a serial artificial intelligence training module 2, a benefit maximization module 3, and a visualization interface module 4. [Big Data Database Construction Module] The big data database construction module 1 incorporates a chemical engineering theory model 11. This model is built upon fundamental theories such as mass balance, energy balance, and reaction kinetics. The big data database construction module 1 is configured within a database construction environment, receiving historical data Ph (i.e., historical production data) from the chemical production site. It then fits the historical data Ph with the chemical engineering theory model 11 to construct a physically meaningful big data database 12. The big data database 12 contains training data Pt, and the training data Pt can be used for training the subsequent connected artificial intelligence training module 2. In some embodiments of the present invention, the chemical engineering theoretical model 11 is a chemical engineering simulation software with physical meaning, such as Aspen Plus, Aspen HYSYS, PRO / II, Petro-SIM, etc. The chemical engineering simulation software is supplemented with equations for mass balance, energy balance, and reaction kinetics to simulate production results under different operating conditions, thereby supplementing the lack of measurement conditions in on-site production. The historical data Ph (historical production data) is obtained from the actual operation records of the chemical plant. Taking the synthesis gas production process as an example, the historical data Ph may include: light oil feed rate, steam consumption, reaction temperature, reaction pressure, carbon dioxide reflux rate, and the corresponding product (e.g., synthesis gas) yield or productivity. The training data Pt in the big data database 12 is obtained by fitting the historical data Ph with the chemical engineering theory model 11. In this embodiment, the training data Pt includes: process data, intermediate data, and product data. Taking the synthesis gas production process as an example, the process data can be, for example, light oil feed rate, steam consumption, reaction temperature, reaction pressure, or carbon dioxide reflux rate, etc., process conditions that can be controlled by the operator. The intermediate data are chemical reaction kinetic data during the reaction process, such as reaction rate or reaction conversion rate. Additionally, the product data can be, for example, the synthesis gas yield or production rate. It is worth noting that the intermediate data may not be available from historical production data from the field; however, in this embodiment of the invention, it can be obtained from the chemical engineering theory model 11 through data fitting, but is not limited to this. In one embodiment of the present invention, in order to ensure the prediction accuracy of the subsequent connected artificial intelligence training module 2, the big data database construction module 1 can further perform a data verification and cleaning operation, including: detecting and removing outliers in historical data Ph, supplementing missing values ​​in historical data Ph or estimating and replacing them in a statistical manner, and filtering samples in historical data Ph that do not conform to physical meaning or production context. However, the present invention is not limited to performing the above-mentioned data verification and cleaning operation. The big data database construction module 1 of this invention constructs a large data database 12 covering multiple scenarios and possessing physical meaning, which can serve as the training basis for the subsequent connected artificial intelligence training module 2. By combining chemical engineering theory with historical experience, the reliability and physical operability of the model can be significantly improved. In other words, since actual production sites do not directly record reaction data under various conditions (e.g., insufficient partial yield or reaction data), this invention embodiment can supplement the above-mentioned deficiencies by initially constructing a big data database through theoretical simulation. On hardware devices, the Big Data Database Construction Module 1 is suitable for servers or workstations with high-performance computing capabilities, such as multi-core processors, large-capacity memory (at least 64GB), and high-performance storage devices (such as solid-state drives or distributed storage systems). It also supports running large-scale data processing frameworks to process historical data from chemical production sites and perform data fitting to construct a physically meaningful big data database. [Connected AI Training Module] Please refer to Figure 1. The serial artificial intelligence training module 2 includes a first artificial intelligence training model 21 and a second artificial intelligence training model 22 that are connected in series. The first artificial intelligence training model 21 is a linear regression artificial intelligence model, and the second artificial intelligence training model 22 is a nonlinear regression artificial intelligence model. In some embodiments of the present invention, the first artificial intelligence training model 21 and the second artificial intelligence training model 22 may be, for example, two-stage artificial intelligence training models based on Python, but the present invention is not limited thereto. The first artificial intelligence training model 21 is constructed based on the linear regression algorithm. It can be implemented using the Ordinary Least Squares (OLS), Gradient Descent, and Regularized Linear Regression, or a combination of these methods, to adapt to different data characteristics and model requirements. The second artificial intelligence training model 22 is constructed based on a nonlinear regression algorithm. It can employ polynomial regression, support vector regression (SVR), decision tree regression, random forest regression, neural network regression, nonlinear least squares, Gaussian process regression (GPR), gradient boosting (XGBoost Regressor), or a combination of these methods to adapt to different nonlinear data patterns and improve the accuracy and applicability of predictions. The serial artificial intelligence training module 2 is configured in a training stage, receiving training data Pt (including process data, intermediate data, and product data) provided by the big data database 12 of the big data database construction module 1, and undergoing separate training and cross-validation by the first artificial intelligence training model 21 and the second artificial intelligence training model 22. Accordingly, the serial artificial intelligence training module 2 can be connected with the big data database 12 of the big data database construction module 1 to accurately model and predict production conditions. In one embodiment of the present invention, the second artificial intelligence training model 22 further includes execution calibration hyperparameters, such as XGBoost tree depth and learning rate, but the present invention is not limited thereto. Furthermore, the first artificial intelligence training model 21 of the serial artificial intelligence training module 2 receives a process parameter P1 from an external input in a prediction stage, and performs calculations based on the model constructed from the training data Pt using a linear regression algorithm, thereby outputting an intermediate parameter P2. In some embodiments of the present invention, taking syngas as an example, the externally input process parameters P1 are, for example, light oil feed rate, steam consumption, reaction temperature, reaction pressure, or carbon dioxide reflux rate, which are process conditions that can be controlled by the operator. The intermediate parameter P2 is chemical reaction kinetic data during the reaction process, which may be, for example, the reaction rate or the reaction conversion rate. The first artificial intelligence training model 21 can quickly handle partially linear parameter relationships. That is, the first artificial intelligence training model 21 can quickly calculate the intermediate parameter P2 based on the process parameter P1, which can provide a preliminary reference for the subsequent second artificial intelligence training model 22 (nonlinear model) and reduce the complex computational load required by the second artificial intelligence training model 22. The second artificial intelligence training model 22 also receives the process parameters P1 (such as light oil feed rate, steam consumption, reaction temperature, reaction pressure, or carbon dioxide reflux rate) from external input during the prediction stage, and further receives the intermediate parameters P2 (such as reaction rate or reaction conversion rate) output by the first artificial intelligence training model 21. Then, it performs in-depth learning and calculation based on the model constructed by the training data Pt using a nonlinear regression algorithm, and finally outputs a prediction result P3 (such as product yield or productivity). The second artificial intelligence training model 22 can handle high-dimensional nonlinear data and has a strong tolerance for missing values ​​and noise, resulting in higher prediction accuracy. Overall, the first artificial intelligence training model 21 is based on linear regression algorithms (such as ordinary least squares, gradient descent, normalization methods, etc.), and can handle data patterns where the input and output data have a clear linear relationship. The second artificial intelligence training model 22 is based on nonlinear regression algorithms (such as support vector regression, decision tree regression, random forest regression, neural network regression, etc.), and is used to capture more complex nonlinear relationships between input and output. The serial artificial intelligence training module 2 in this embodiment of the invention consists of a first artificial intelligence training model 21 (linear regression model) and a second artificial intelligence training model 22 (nonlinear regression model), which work together in the training phase and the prediction phase, respectively. During the training phase, the AI ​​training module 2 receives training data Pt provided by the big data database construction module 1, including process data, intermediate data and product data. After training and cross-validation of the two models, the model construction is completed. In the prediction phase, the first AI training model 21 processes the partially linear process parameters P1 (such as light oil feed rate and steam consumption) and quickly calculates the intermediate parameters P2 (such as reaction rate), providing them to the second AI training model 22 to reduce the latter's computational burden. The second AI training model 22 processes high-dimensional nonlinear data based on a nonlinear regression algorithm, ultimately outputting the prediction result P3 (such as yield or output rate). It has tolerance for noise and missing data and improves prediction accuracy. This architecture combines the advantages of linear and nonlinear models, effectively achieving accurate modeling and optimization of chemical processes. The architecture described above in this invention demonstrated high accuracy in experiments, with R² exceeding 0.95, indicating that the model has a high explanatory power for data variation. The mean absolute percentage error (MAPE) is less than 0.2%, showing minimal deviation between the predicted results and actual data, providing reliable decision support for the optimization of chemical processes. It should be noted that R 2 It is a statistical indicator used to measure the explanatory power of a regression model for data variation. Its value ranges from 0 to 1. The closer the value is to 1, the stronger the explanatory power of the model for the target data and the higher its fit. In an embodiment of the present invention, R 2 R is used to evaluate the predictive power of cascaded AI training models. 2 A value exceeding 0.95 indicates that the model has high accuracy and excellent adaptability to chemical process data, and can effectively support production decisions. Mean Absolute Percentage Error (MAPE) is a metric used to measure the accuracy of a forecasting model. It represents the average percentage error between the predicted and actual values. The smaller the MAPE value, the smaller the deviation between the model's predictions and the actual values, and the higher the model's accuracy. In this embodiment of the invention, the MAPE is less than 0.2%, indicating that the prediction bias of the tandem integrated model is extremely low, and it can reliably predict the yield of the syngas process, providing higher reliability for the optimization of the production process. In one implementation, if R² is too small or MAPE is too high, the serial artificial intelligence training module 2 will return to the training phase and continue training until R² and MAPE reach the ideal values. On the hardware side, the cascaded AI training module 2 is suitable for running in environments that support hardware-accelerated computing, such as servers equipped with high-performance graphics processing units (GPUs) or dedicated hardware accelerators. This module can run in computing environments that support deep learning and data modeling frameworks, enabling efficient computation of linear and nonlinear regression models. This hardware environment can significantly improve training speed and prediction accuracy, thereby achieving precise modeling and optimization of chemical processes. [Efficiency Maximization Module] Please refer to Figure 1. The benefit maximization module 3 is embedded with a nature-inspired algorithm. This algorithm simulates the mechanisms or phenomena in nature to seek the optimal solution of chemical process parameters in order to maximize economic benefits. In some embodiments of the present invention, the natural heuristic algorithm may be a genetic algorithm (GA), a particle swarm optimization (PSO) algorithm, an ant colony optimization (ACO) algorithm, or a simulated annealing (SA) algorithm. The benefit maximization module 3 is connected to the serial artificial intelligence training module 2 to synchronize and utilize the data and intelligent model of the serial artificial intelligence training module 2, and receive the data predicted and output by the serial artificial intelligence training module 2, such as prediction result P3 (e.g., prediction result of the yield or production rate of syngas products under different operating conditions). The benefit maximization module 3 is further configured to import an externally input economic parameter P4. The economic parameter P4 may include, for example, a revenue parameter and an expenditure parameter. The revenue parameter may be, for example, the product selling price, and the expenditure parameter may include, for example, raw material prices and energy costs (e.g., steam costs, electricity costs, fuel costs). The efficiency maximization module 3 is configured to use the natural heuristic algorithm to calculate a production decision Pf based on the product's demand output, combined with economic parameters P4 and data from the connected artificial intelligence training module 2 (such as prediction results P3 under different operating conditions), and taking into account the upper and lower limits of operation. The production decision (Pf) includes suggested process operating conditions, such as light oil feed rate, steam consumption, reaction temperature, reaction pressure, and carbon dioxide reflux rate, to maximize profits or minimize costs while meeting production demand. In one embodiment of the present invention, the benefit maximization module 3 takes the objective function (such as the profit maximization function or the cost minimization function) as its core and analyzes the optimal chemical process operating conditions through a loop iterative RC method. It should be noted that, in this embodiment of the invention, "incorporating operational upper and lower limit factors" means that when the benefit maximization module 3 uses a natural heuristic algorithm to solve for the optimal production decision, it is necessary to simultaneously consider the safe range and the theoretical range of each process parameter in actual production. The operational upper and lower limit factors originate from historical production data output by the front-end big data database construction module 1, equipment operation manuals, technical specifications, or pre-set limiting parameters by on-site operators. For example, reactor operating temperature, pressure, light oil feed rate, steam consumption, and CO2... For example, the return flow rate will have clear maximum and minimum limits to avoid exceeding the equipment safety limits or violating the chemical design requirements. In implementation, these upper and lower limits can be directly input as constraints into the benefit maximization module and incorporated into the objective function calculation along with economic parameters (such as product selling price, raw material and energy costs) and prediction results (such as product output or yield). When searching for the optimal solution, the natural heuristic algorithm uses these upper and lower limits to restrict the range of candidate solutions, ensuring that the process operation parameters obtained in each iteration fall within a safe and feasible range. If the parameter combinations generated by the algorithm exceed the predetermined range, an adjustment mechanism can be used to bring them back to a reasonable range for correction. In addition to the inherent operating limits of the equipment, factors affecting the upper and lower limits of operation may also include environmental regulations, product quality requirements, and other actual conditions at the production site, such as reactant purity and equipment aging. Therefore, by incorporating these constraints, the system can provide a safe and practically feasible production decision-making solution while maximizing market demand and economic benefits. Overall, the benefit maximization model 3 starts with the prediction results of the cascade ensemble model, and further combines economic parameters P4 (such as raw material prices, energy costs, and product selling prices) to calculate the optimal operating parameters using natural heuristic algorithms (such as particle swarm optimization and genetic algorithms) to achieve the goal of maximizing economic benefits. This model uses a loop approach, repeatedly substituting the prediction results of the cascade ensemble model into the loop to dynamically adjust and optimize production conditions. For example, users can input the required output into the system. The benefit maximization module 3, based on the required output, and taking into account the predicted results P3 and economic parameters P4 under different operating conditions, outputs the production decision with the best process parameters, including the best raw material ratio, operating temperature, pressure, etc., so that the process can maximize profits and minimize costs while meeting the required output. The aforementioned efficiency maximization module 3 is suitable for operation in hardware environments with high-performance computing capabilities, including multi-core, high-frequency processors to support a large number of loop calculations and optimization operations. For further improvements in computational efficiency, a graphics processing unit (GPU) or a dedicated hardware accelerator can be used, making it particularly suitable for computationally demanding algorithms such as particle swarm optimization and genetic algorithms. This module can be deployed on local servers, enterprise data centers, or cloud platforms that support high-performance computing, thus flexibly addressing the economic optimization needs of various scenarios. It should be noted that the aforementioned benefit-maximizing module 3 must be subject to constraints, such as upper and lower operating limits, to consider actual equipment operating limitations (e.g., safe operating ranges for temperature and pressure). Product demand must reach a specific output or purity. Environmental regulations must also consider restrictions imposed by carbon dioxide emissions or safety standards. By using the aforementioned benefit maximization module 3, the present invention can perform accurate and real-time cost-benefit analysis on parameters of the syngas process (such as light oil feed rate and CO₂ reflux rate) to further improve overall chemical production profits or reduce energy consumption. [Visual Interface Module] Please refer to Figure 1. The visualization interface module 4 can present the production decision Pf of the benefit maximization module 3 (such as the suggested light oil feed rate, steam consumption, reaction temperature, reaction pressure, carbon dioxide reflux rate, etc. process operation parameters). More specifically, the visualization interface module 4 can also be defined as a real-time economic navigation module, which can integrate all calculations into a single interface (such as a company internal webpage or APP) using Python or other open-source tools. This allows on-site operators or operational decision-makers to input product demand and price information in real time, enabling the system to automatically calculate and return optimal operating parameters and profit forecasts. This helps save decision-making time and reduce trial-and-error costs. More specifically, the chemical process optimization system 100 of this embodiment can ultimately present the integrated data processing and optimization results of the cascaded integrated model through a visual interface, allowing users to observe the model output in real time and operate it easily. This interface provides input of operating parameters; for example, users can input product demand and economic parameters, such as raw material prices. The interface transmits this information to the benefit maximization module 3 for real-time calculation. Multi-scenario simulation provides simulated operating conditions under various scenarios (such as high or low demand, high or low prices), and displays output, cost, and profit under different scenarios using bar charts, radar charts, or heat maps. Optimal operating conditions are provided in real-time through a display or control system interface, such as a suggested reduction of 3% in light oil usage and 2% in steam usage, allowing for adjustments to actual production conditions. This visualization interface effectively reduces human error and decision-making time during adjustments in chemical production, making the technological achievements of this invention easier to implement on-site in the factory. [Taking the syngas process as an example] The following examples illustrate how this invention can be applied to the syngas process to improve efficiency: The process example is a recombination reaction. The main raw material is methane (CH4). 4) with steam (H 2O), producing syngas (CO mixed with H2O) 2). A typical reaction formula is as follows: CH 4+ H 2O ↔ CO + 3H 2 (Main Reaction) CH 4+ CO 2↔ 2CO + 2H 2 CO 2+ H 2↔ CO + H 2O Input the following process parameters P1 variables into the first artificial intelligence training model 21 (linear regression model): light oil feed rate, steam consumption, reaction temperature, reaction pressure, CO₂ reflux flow rate, etc. The first artificial intelligence training model 21 outputs the predicted intermediate parameter P2 (i.e., reaction rate). The intermediate parameter P2 (reaction rate) from the first stage output and the process parameters P1 from the first stage input are input into the second artificial intelligence training model 22 (a nonlinear regression model such as XGBoost Regressor). These parameters include: light oil feed rate, steam consumption, reaction temperature, pressure, and CO. 2. Return flow rate). The second artificial intelligence training model 22 outputs the prediction result P3 (e.g., syngas production). The profit maximization module (embedded PSO) uses the objective function: maximize profit(x) = (product selling price × syngas production) - (raw material cost + energy cost), and takes into account constraints such as: reaction temperature and pressure must not exceed safety limits, and production must meet customer needs. The results showed that, compared to production conditions with the same amount of light oil, this model could increase syngas production by 2%. Compared to production conditions with the same syngas production, it could reduce light oil consumption by 2% and steam consumption by 3%. As can be seen from the above implementation examples, the present invention can grasp the changes in market and process parameters (such as changes in raw material prices) in real time, so as to achieve the goal of reducing costs and increasing profits. [A Chemical Process Optimization Method Based on a Series Integration Model] Please refer to Figure 2. This embodiment of the invention also provides a chemical process optimization method based on a series integration model, which includes steps S110, S120, S130 and S140. Step S110 includes: using a big data database construction module with a built-in chemical engineering theory model in a database construction operation environment, receiving historical data from the chemical production site, and fitting the historical data with the chemical engineering theory model to construct a big data database. Step S120 includes: connecting a serial artificial intelligence training module to the big data database, and training the model using training data from the big data database during a training phase; wherein the serial artificial intelligence training module includes a first artificial intelligence training model and a second artificial intelligence training model connected in series; wherein the first artificial intelligence training model is constructed based on a linear regression algorithm, and the second artificial intelligence training model is constructed based on a nonlinear regression algorithm; wherein the first artificial intelligence training model receives an externally input process parameter during a prediction phase and outputs an intermediate parameter; wherein the second artificial intelligence training model receives the process parameter during the prediction phase and further receives the intermediate parameter, thereby outputting a prediction result. Step S130 includes: receiving the prediction result with a benefit maximization module embedded with a natural heuristic algorithm, and further importing an externally input economic parameter; wherein the benefit maximization module is configured to output a production decision based on the demand output of a product, combined with the economic parameter and the prediction result, and incorporating upper and lower limit factors of operation, through the natural heuristic algorithm. Step S140 includes: providing users with real-time input of product demand through a visual interface module, and visually presenting the production decisions of the benefit maximization module. [Beneficial Effects of the Examples] The technical solution of this invention utilizes a chemical process optimization system based on a cascaded integrated model, combining big data, artificial intelligence, and optimization algorithms to achieve accurate modeling and maximize economic benefits in chemical processes. This system comprises a big data database construction module, a cascaded artificial intelligence training module, a benefit maximization module, and a visualization interface module. These modules work collaboratively to address issues such as data gaps, inaccurate modeling, and low decision-making efficiency in traditional chemical processes. First, the big data database construction module combines chemical engineering theoretical models with historical data from chemical production sites to build a physically meaningful big data database, providing a highly reliable data foundation for subsequent modeling. Secondly, the cascaded AI training module, through a cascaded architecture of linear and nonlinear regression models, fully leverages the rapid computational advantages of linear models and the high accuracy of nonlinear models to achieve precise modeling and prediction of complex chemical process conditions and reactions. This module can dynamically handle the relationship between process parameters, intermediate parameters, and prediction results, significantly improving the model's prediction accuracy. The efficiency maximization module calculates the optimal process operating conditions by embedding natural heuristic algorithms and combining the prediction results output by the cascaded artificial intelligence model with external economic parameters (such as raw material costs, energy consumption, and product selling prices), so that the chemical process can maximize profits or minimize costs while meeting demand output. In addition, the module takes into account the upper and lower limits of operation, environmental regulations and safety constraints to ensure the feasibility and compliance of the results. Finally, the visualization interface module presents the system's decision results intuitively in multiple formats, supporting on-site operation and multi-scenario simulation, thereby improving decision-making efficiency and accuracy. This invention optimizes chemical processes, providing improvements for enhancing chemical production efficiency and economic benefits. The above description is only a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Therefore, all equivalent technical changes made based on the description and drawings of the present invention are included within the protection scope of the present invention. 100: Chemical Process Optimization System Based on Serial Integration Model 1: Big Data Database Construction Module 11: Chemical Theory Model 12: Big Data Database 2: Serial Artificial Intelligence Training Module 21: First Artificial Intelligence Training Model 22: Second Artificial Intelligence Training Model 3: Efficiency Maximization Module 4: Visual Interface Module Ph: Historical Data Pt: Training Data P1: Process Parameters P2: Intermediate Parameters P3: Prediction Results P4: Economic Parameters Pf: Production Decision RC: Loop Iteration S110: Step S110 S120: Step S120 S130: Step S130 S140: Step S140 Figure 1 is a schematic diagram of the chemical process optimization system according to an embodiment of the present invention. Figure 2 is a flowchart of the chemical process optimization method according to an embodiment of the present invention. 100: Chemical Process Optimization System Based on Series Integration Model 1: Big Data Database Construction Module 11: Chemical Engineering Theoretical Model 12: Big Data Database 2: Connected AI training modules 21: First Artificial Intelligence Training Model 22: Second AI Training Model 3: Module for maximizing benefits 4: Visual Interface Module Ph: Historical Data Pt: Training data P1: Process parameters P2: Intermediate parameters P3: Prediction Results P4: Economic Parameters Pf: Production Decisions RC: Loop Iteration

Claims

1. A chemical process optimization system based on a cascade integration model includes: A big data database construction module, with a built-in chemical engineering theory model, is configured in a database construction operation environment to receive historical data from chemical production sites and fit the historical data with the chemical engineering theory model to construct a big data database. A serial artificial intelligence training module is connected to the big data database and, during a training phase, trains the model using training data from the big data database. The serial artificial intelligence training module includes: a first artificial intelligence training model constructed based on a linear regression algorithm; and a second artificial intelligence training model constructed based on a nonlinear regression algorithm. The first artificial intelligence training model receives an externally input process parameter during a prediction phase and outputs an intermediate parameter. The second artificial intelligence training model, during the prediction phase, receives the process parameter and further receives the intermediate parameter, thereby calculating and outputting a prediction result. A benefit maximization module is embedded with a natural heuristic algorithm. The benefit maximization module is data-connected to the serial artificial intelligence training module. The benefit maximization module is configured to further import an externally input economic parameter; the benefit maximization module is configured to use the natural heuristic algorithm to calculate and output a production decision based on the demand output of a product, combined with the economic parameter and the prediction result, and incorporating operational upper and lower limits. The chemical process optimization system as described in claim 1 further includes: A visual interface module, configured to allow users to input product demand in real time, and visually presents the production decisions of the benefit-maximizing module. The chemical process optimization system as described in claim 1, wherein the chemical theoretical model is a physically meaningful chemical simulation software, and is at least one of Aspen Plus, Aspen HYSYS, PRO / II, and Petro-SIM; wherein, The historical data refers to historical production data from chemical production sites, and the training data includes process data, intermediate data, and product data. The chemical process optimization system as described in claim 3, wherein the first artificial intelligence training model employs at least one of Ordinary Least Squares (OLS), Gradient Descent, and Regularized Linear Regression. The chemical process optimization system as described in claim 4, wherein the second artificial intelligence training model employs at least one of the following: polynomial regression, support vector regression (SVR), decision tree regression, random forest regression, neural network regression, nonlinear least squares, Gaussian process regression (GPR), and gradient boosting (XGBoost Regressor). The chemical process optimization system as described in claim 5, wherein the natural heuristic algorithm employs at least one of the following: Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), and Simulated Annealing (SA). As described in claim 6, in the chemical process optimization system, the serial artificial intelligence training module is configured in the training phase to train and cross-validate the first artificial intelligence training model and the second artificial intelligence training model respectively using the training data from the big data database. The chemical process optimization system as described in claim 1, wherein... In the cascaded artificial intelligence training module, the process parameters include at least one of the following: light oil feed rate, steam consumption, reaction temperature, reaction pressure, and carbon dioxide reflux rate; wherein the intermediate parameters include at least one of the following: reaction rate and reaction conversion rate; wherein the prediction results include at least one of the following: product yield and yield. The chemical process optimization system as described in claim 8, wherein the economic parameters include revenue parameters and expenditure parameters; wherein the revenue parameter is the product selling price, and the expenditure parameters include raw material prices and energy costs. The chemical process optimization system as described in claim 1, wherein... The efficiency maximization module analyzes and outputs the production decision based on a profit maximization function or a cost minimization function through a loop iteration method. A chemical process optimization method based on a tandem ensemble model includes: A big data database construction module with a built-in chemical engineering theory model is used in a database construction operation environment to receive historical data from chemical production sites and fit the historical data with the chemical engineering theory model to construct a big data database. A serial AI training module is connected to the big data database, and in a training phase, the model is trained using training data from the big data database. The serial AI training module includes a first AI training model and a second AI training model connected in series. The first AI training model is constructed based on a linear regression algorithm, and the second AI training model is constructed based on a nonlinear regression algorithm. In a prediction phase, the first AI training model receives an externally input process parameter and outputs an intermediate parameter. In the prediction phase, the second AI training model receives the process parameter and further receives the intermediate parameter, thereby outputting a prediction result. The serial AI training module is also connected to a benefit maximization module embedded with a natural heuristic algorithm, and further imports an externally input economic parameter. The benefit maximization module is configured to use the natural heuristic algorithm to determine the production decision based on the product's demand output, the economic parameter, the prediction result, and operational upper and lower limits. The chemical process optimization method as described in claim 11 further includes: A visual interface module allows users to input product demand in real time and visualizes the production decisions of the benefit-maximizing module.

Images