A real-time optimization method, system and electronic equipment for a chemical production device

Abstract

The application provides a real-time optimization method, system and electronic equipment of a chemical production device, and relates to the field of data processing. In the method, mechanism data and real-time operation data of the chemical production device in a production process are acquired; a mechanism model is constructed according to the mechanism data, and the mechanism model is used to reflect material transfer, energy balance and chemical reaction inside the chemical production device; a prediction model is constructed according to the real-time operation data, and the prediction model is used to predict a nonlinear relationship between input variables and output variables of the chemical production device; the mechanism model and the prediction model are fused to obtain a fusion model; and the fusion model is used to process target optimization parameters to obtain an optimization result. The technical solution provided by the application facilitates improving the accuracy of real-time optimization of the chemical production device.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a real-time optimization method, system, and electronic equipment for a chemical production plant. Background Technology

[0002] With the acceleration of global industrialization, the chemical industry has occupied an important position in the modern economy. The operating efficiency, economic benefits, and safety of chemical production plants directly affect the competitiveness of enterprises and the level of energy and resource consumption. Against this backdrop, the optimization of chemical production plants has gradually become a research hotspot in the industry.

[0003] Currently, most optimization methods are limited to the application of single models or the combination of linear models with other models through simple weighting. However, the application of single models or simple combinations often has limitations. For example, linear models often neglect nonlinear characteristics and complex operating conditions, resulting in lower accuracy of the optimization results.

[0004] Therefore, there is an urgent need for a real-time optimization method, system, and electronic equipment for chemical production plants. Summary of the Invention

[0005] This application provides a method, system, and electronic equipment for real-time optimization of chemical production plants, which facilitates improving the accuracy of real-time optimization of chemical production plants.

[0006] A first aspect of this application provides a real-time optimization method for a chemical production plant. The method includes: acquiring mechanistic data and real-time operating data of the chemical production plant during the production process; constructing a mechanistic model based on the mechanistic data, the mechanistic model reflecting mass transfer, energy balance, and chemical reactions within the chemical production plant; constructing a prediction model based on the real-time operating data, the prediction model predicting the nonlinear relationship between input and output variables of the chemical production plant; fusing the mechanistic model and the prediction model to obtain a fused model; and processing target optimization parameters using the fused model to obtain optimization results.

[0007] By adopting the above technical solutions, the mechanistic model, based on the theoretical foundations of mass transfer, energy balance, and chemical reactions, can accurately reflect the fundamental principles of chemical processes. This model can precisely describe the core physical and chemical laws governing the process. The predictive model, driven by data, can effectively capture the complex nonlinear relationships between inputs and outputs, providing rapid response and relatively accurate predictions, especially in the face of highly complex and dynamically changing operating conditions. The fusion model combines the physicality of the mechanistic model with the adaptability of the predictive model, taking into account both the fundamental laws of the process and the actual operating conditions during optimization, thereby achieving higher precision optimization and avoiding errors that may occur with a single model under certain operating conditions. In chemical production processes, the operating conditions of the equipment are often dynamically changing, making traditional static optimization methods difficult to cope with. By fusing the mechanistic and predictive models, the system can process real-time operating data and make rapid adjustments based on process changes and dynamic operating conditions. The predictive model, through learning from real-time data, can respond to changes in operating conditions and adjust optimization strategies to ensure optimal operating conditions under different circumstances. While the mechanistic model provides an accurate theoretical framework, it often faces uncertainties in practical applications. Predictive models, through learning from extensive real-world operational data, enhance their adaptability to uncertainties. By fusing these two models, the accuracy of the mechanistic model can be preserved while the predictive model allows for dynamic correction and adjustment, thereby improving system robustness and maintaining good optimization performance even in the face of data noise or missing data. Furthermore, the fused model effectively improves the response speed of real-time optimization by balancing the complexity of the mechanistic model with the computational speed of the predictive model in real-time calculations. During production, when operating conditions change, the system can react quickly, adjusting control strategies and optimizing operating parameters to ensure production efficiency and safety. Therefore, it facilitates the accurate real-time optimization of chemical production plants.

[0008] Optionally, acquiring the mechanism data and real-time operating data of the chemical production unit during the production process specifically includes: determining the physical and chemical principles of the chemical production unit; determining the process behavior of the chemical production unit during the production process based on the physical and chemical principles to obtain the mechanism data; acquiring the raw data of the chemical production unit during the production process in real time; and preprocessing the raw data to obtain the real-time operating data, wherein the preprocessing includes noise reduction, filtering, and normalization.

[0009] By adopting the above technical solutions and defining these fundamental principles, the accuracy of the mechanistic model can be ensured, avoiding unrealistic predictions or optimization results in complex production environments, thereby improving the model's reliability and practical application effectiveness. This capture of process behavior provides a strong theoretical foundation for the optimization model, enabling it to better reflect the dynamic characteristics and variability of the production process and enhance its adaptability to dynamic changes. Real-time acquisition of raw data from the production process allows for dynamic monitoring of the device's operating status. This raw data provides valuable input information for the prediction model, capturing subtle changes in actual production and improving the real-time nature of optimization decisions. Preprocessing the raw data ensures its accuracy and consistency. Denoising and filtering help eliminate random noise caused by equipment errors or environmental fluctuations, making the data more stable and reliable; normalization ensures that data of different dimensions and ranges have the same influence in the model, avoiding inaccuracies in the optimization model due to different data scales. This processed real-time data provides feedback on the actual operation of the mechanistic model, helping to optimize and adjust according to the current actual working conditions, enabling the fusion model to more accurately predict and optimize the system's operating status. By collecting and preprocessing real-time operational data, the optimization process can dynamically adjust optimization decisions based on the latest production data, rather than relying solely on the assumptions of theoretical models. The real-time nature and accuracy of the data enable the optimization model to flexibly respond to changes in the production process, thereby improving the accuracy of the optimization results. Combining mechanistic data with real-time data can overcome the limitations of mechanistic models in the face of uncertainty and complex operating conditions.

[0010] Optionally, constructing a mechanistic model based on the mechanistic data specifically includes: determining the basic reaction relationships between substances within the chemical production unit based on the mechanistic data; determining the feed parameters, output parameters, intermediate product parameters, and by-product parameters of the chemical production unit based on the basic reaction relationships; determining the complex reaction relationships between substances within the chemical production unit based on the mechanistic data; determining the heat transfer parameters, heat loss parameters, and mass transfer parameters of the chemical production unit based on the complex reaction relationships; and generating the mechanistic model under natural constraints based on the feed parameters, output parameters, intermediate product parameters, by-product parameters, and the heat transfer parameters, heat loss parameters, and mass transfer parameters.

[0011] By adopting the above technical solutions, the fundamental reaction relationships between substances are the core of constructing the mechanistic model, helping to determine how reactants are transformed into products. Through these relationships, reaction kinetic equations can be established, further deriving information such as the rate and conversion rate of each reaction. Based on these reaction relationships, by determining feed parameters, output parameters, and parameters for intermediate products and byproducts, the material flow and changes throughout the entire production process can be accurately simulated. This not only helps in understanding the conversion efficiency of the reaction but also provides support for optimizing product quality, yield, and other aspects. Heat transfer, heat loss, and mass transfer parameters are key factors describing the energy and mass exchange processes in chemical plants. Heat and mass transfer in chemical production processes are often the result of multiple factors working together. By defining and calculating these parameters, it can be ensured that the mechanistic model not only describes the chemical process of the reaction but also considers the thermal effects and mass transfer efficiency of the equipment, which is crucial for optimizing energy efficiency, increasing reaction rates, and controlling reaction temperatures. By modeling complex reaction relationships, more realistic chemical reaction processes can be simulated, which not only improves the accuracy of the model but also ensures that optimization and control strategies remain effective under varying operating conditions. When generating mechanistic models, these constraints are strictly followed to ensure that the models do not produce unrealistic results. For example, mass in a chemical reaction cannot be created or destroyed out of thin air, and heat transfer processes cannot violate the first law of thermodynamics. By adhering to these natural constraints, the generated mechanistic models are closer to actual operations, avoiding unreasonable predictions found in theoretical models.

[0012] Optionally, constructing a prediction model based on the real-time operating data specifically includes: acquiring historical operating data of the chemical production unit; determining the real-time operating data and the historical operating data as input variables, processing them using a support vector machine algorithm to obtain the output variables; acquiring the input variables and output variables corresponding to each of the multiple training processes, evaluating the training process using an evaluation algorithm, and obtaining the prediction model.

[0013] By adopting the above technical solutions, real-time operational data reflects the current state of the production process, while historical operational data provides a record of past operating conditions. Combining the two allows for a more comprehensive understanding and prediction of the behavior of chemical production units. Real-time data provides immediate feedback on current operations, while historical data provides long-term behavioral patterns under different operating conditions; combining the two helps capture more comprehensive system characteristics. Support Vector Machines (SVMs), through the use of kernel functions, can effectively handle nonlinear relationships, thus providing more accurate predictions under complex operating conditions. By acquiring input and output variables from multiple training processes and using evaluation algorithms to assess the training process, it can be ensured that the model reaches its optimal state during training. Evaluation algorithms typically help select the most suitable parameters, avoiding overfitting and underfitting problems, making the final prediction model more accurate and robust. The evaluation process helps detect and optimize model performance, further improving its predictive accuracy and stability. Through evaluation, potential problems in model training can be identified, such as overly complex or unsuitable assumptions for the current dataset, allowing for model adjustments to better adapt to the production environment. This process ensures that the model not only performs well on the training set but also possesses strong predictive capabilities in real-world applications. By acquiring data in real time and applying the predictive model, immediate feedback and forecasts can be provided to chemical production processes. This real-time nature allows for dynamic adjustments to the production process, optimizing operating conditions and preventing malfunctions or inefficiencies.

[0014] Optionally, fusing the mechanistic model and the prediction model to obtain a fused model specifically includes: constructing a population containing the mechanistic model and the prediction model, the population comprising multiple individuals, each individual representing a combination structure of a model; performing differential mutation on a target individual to obtain a mutated individual, the target individual being any one of the multiple individuals; performing a crossover operation on the target individual and the mutated individual to obtain a candidate individual; comparing the fitness value corresponding to the candidate individual with the fitness value corresponding to the target individual; if it is determined that the fitness value corresponding to the candidate individual is greater than or equal to the fitness value corresponding to the target individual, then replacing the target individual with the candidate individual and entering the next generation population; repeating the differential mutation, the crossover operation, and the comparison until a preset number of iterations is reached or the fitness value converges to obtain the fused model.

[0015] By employing the aforementioned technical solution and constructing a population containing both mechanistic and predictive models, the method utilizes various model combinations as individuals. Each individual represents a different model structure, allowing for the exploration of possible model combinations and fully exploiting the fusion potential of mechanistic and predictive models. Differential mutation generates new individuals by mutating the target individual. This mutation introduces new structures and parameters, increasing population diversity. The mutation process enables the algorithm to escape local optima, explore a broader solution space, and enhance the adaptability and robustness of the fusion model. Through crossover, the target individual and mutated individuals exchange some structures, generating candidate individuals. This operation helps retain the characteristics of excellent individuals while combining the advantages of different individuals, potentially leading to a better model structure. By comparing the fitness values ​​of candidate and target individuals, the differential evolution algorithm can select a more suitable model combination by optimizing the objective function. Individuals with higher fitness values ​​indicate better model performance in real-world tasks and are therefore more aligned with the optimization objective. This evaluation mechanism ensures that each iteration progresses towards the optimal fusion model. Fitness evaluation can help algorithms optimize models in a targeted manner, gradually improving the prediction accuracy and reliability of fusion models.

[0016] Optionally, the step of processing the target optimization parameters through the fusion model to obtain the optimization result specifically includes: determining the target input variable and the target output variable from the target optimization parameters; inputting the target input variable into the fusion model to obtain a feature variable, the feature variable including a feature vector; determining the positional relationship between the feature vector and the hyperplane of the support vector machine model corresponding to the fusion model; calculating a first probability value and a second probability value based on the positional relationship, the first probability value representing the probability value corresponding to the feature vector on the first side of the hyperplane, the second probability value representing the probability value corresponding to the feature vector on the second side of the hyperplane, the first side and the second side being two opposite sides of the hyperplane; if it is determined that the feature vector is located on the first side and the first probability value is greater than or equal to a preset threshold, then the feature variable is determined to be the target output variable, and the optimization result is determined to be normal; if it is determined that the feature vector is located on the second side and the second probability value is greater than or equal to the preset threshold, then the feature variable is determined not to be the target output variable, and the optimization result is determined to be abnormal.

[0017] By adopting the above technical solutions, and determining the target input and output variables from the target optimization parameters, the system can focus on key parameters in the optimization process, avoiding interference from irrelevant data, thereby improving the accuracy of the optimization results. By inputting the target input variables into the fusion model, the advantages of the mechanistic model and the predictive model can be fully utilized to obtain more accurate and comprehensive feature variables, ensuring that the optimization results are more in line with actual production needs. The construction of the hyperplane enables the support vector machine to effectively distinguish different categories in high-dimensional space, thereby achieving more efficient decision-making. The support vector machine can accurately perform classification and regression. Through the segmentation of the hyperplane, the optimization results are clearly classified in the model, providing a high-precision judgment basis. By setting a preset threshold, the system can automatically filter out optimization results that meet the standards and exclude abnormal cases. The setting of the threshold provides a risk control mechanism, ensuring that only optimization results that meet certain conditions are considered normal, further enhancing the reliability of the decision. By judging the positional relationship between the feature vector and the hyperplane, abnormal cases of optimization results can be automatically identified. For example, when the optimization result is located on the other side of the hyperplane and the probability value does not reach the preset threshold, the system will automatically judge the optimization result as abnormal. This feature can effectively prevent human error and reduce the possibility of misjudgment.

[0018] Optionally, the method further includes: receiving an optimization request uploaded by a user through a user device, the optimization request including the target optimization parameters; and optimizing the target optimization parameters according to the optimization request.

[0019] By adopting the above technical solution, users can directly participate in the optimization process by uploading optimization requests through their devices. This approach allows users to proactively request optimizations based on their own needs and actual production conditions, enhancing the system's interactivity and flexibility. Users can submit new optimization goals at any time based on real-time data and changes in operating conditions, increasing the controllability of the production process. By receiving uploaded optimization requests, cumbersome interaction steps between users and the system are avoided, allowing the optimization process to start more quickly and efficiently. This saves users time and improves work efficiency. Users can upload optimization requests based on specific production needs or encountered problems. Each optimization request can target different production goals, thereby effectively improving production efficiency. The system processes the target optimization parameters based on the user's uploaded requests, enabling precise responses to current production conditions.

[0020] A second aspect of this application provides a real-time optimization system for a chemical production plant. The real-time optimization system includes an acquisition module and a processing module. The acquisition module is used to acquire mechanistic data and real-time operating data of the chemical production plant during the production process. The processing module is used to construct a mechanistic model based on the mechanistic data, the mechanistic model reflecting the mass transfer, energy balance, and chemical reactions within the chemical production plant. The processing module is also used to construct a prediction model based on the real-time operating data, the prediction model predicting the nonlinear relationship between the input and output variables of the chemical production plant. The processing module is further used to fuse the mechanistic model and the prediction model to obtain a fused model. The processing module is also used to process the target optimization parameters through the fused model to obtain optimization results.

[0021] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, and both the user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method described above.

[0022] A fourth aspect of this application provides a computer-readable storage medium storing instructions that, when executed, perform the method described above.

[0023] In summary, one or more technical solutions provided in this application have at least the following technical effects or advantages:

[0024] Mechanistic models, based on the theoretical foundations of mass transfer, energy balance, and chemical reactions, accurately reflect the fundamental principles of chemical processes. These models can precisely describe the core physical and chemical laws governing the process. Predictive models, driven by data, effectively capture the complex nonlinear relationships between inputs and outputs, providing rapid response and relatively accurate predictions, especially in the face of highly complex and dynamically changing operating conditions. Fusion models combine the physicality of mechanistic models with the adaptability of predictive models, taking into account both the fundamental laws of the process and actual operating conditions during optimization, thus achieving higher precision optimization and avoiding errors that can occur with single models under certain operating conditions. In chemical production processes, the operating conditions of equipment are often dynamically changing, making traditional static optimization methods difficult to handle. By fusing mechanistic and predictive models, the system can process real-time operating data and quickly adjust according to process changes and dynamic operating conditions. Predictive models, through learning from real-time data, can respond to changes in operating conditions and adjust optimization strategies to ensure optimal operating conditions under various circumstances. While mechanistic models provide an accurate theoretical framework, they often face uncertainties in practical applications. Predictive models, through learning from extensive real-world operational data, enhance their adaptability to uncertainties. By fusing these two models, the accuracy of the mechanistic model can be preserved while the predictive model allows for dynamic correction and adjustment, thereby improving system robustness and maintaining good optimization performance even in the face of data noise or missing data. Furthermore, the fused model effectively improves the response speed of real-time optimization by balancing the complexity of the mechanistic model with the computational speed of the predictive model in real-time calculations. During production, when operating conditions change, the system can react quickly, adjusting control strategies and optimizing operating parameters to ensure production efficiency and safety. Therefore, it facilitates the accurate real-time optimization of chemical production plants. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating a real-time optimization method for a chemical production apparatus provided in an embodiment of this application.

[0026] Figure 2 This is another schematic diagram of a real-time optimization method for a chemical production plant provided in an embodiment of this application.

[0027] Figure 3 This is a schematic diagram of a real-time optimization system for a chemical production plant provided in an embodiment of this application.

[0028] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0029] Explanation of reference numerals in the attached figures: 31. Acquisition module; 32. Processing module; 41. Processor; 42. Communication bus; 43. User interface; 44. Network interface; 45. Memory. Detailed Implementation

[0030] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0031] In the description of the embodiments in this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Specifically, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a concrete manner.

[0032] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0033] With the rapid advancement of global industrialization, the chemical industry has gradually occupied a crucial position in the modern economic system. As one of the foundational industries supporting various sectors globally, the operational efficiency, economic benefits, and safety of chemical production facilities directly determine a company's market competitiveness and are also related to the rational utilization of energy resources and environmental protection. Against this backdrop, the optimization of chemical production facilities has gradually become one of the core issues in the chemical industry's research field. How to improve the overall efficiency of production facilities and reduce energy consumption and waste emissions through effective optimization methods has become a key focus for major chemical companies.

[0034] However, current research and application of optimization methods for chemical production plants are mostly limited to the use of single models, or simply combining different types of models through simple weighting. For example, many optimization methods still rely on linear models. Although linear models can provide fast solutions for certain simple operating conditions, their applicability is often limited by the nonlinear characteristics and complex operating conditions in actual production. Furthermore, when faced with complex production environments, the accuracy of their optimization results is relatively low.

[0035] To address the aforementioned technical problems, this application provides a real-time optimization method for chemical production plants, referring to... Figure 1 , Figure 1 This is a flowchart illustrating a real-time optimization method for a chemical production plant, provided in an embodiment of this application. The real-time optimization method is applied to a server and includes steps S110 to S150, as follows:

[0036] S110. Obtain mechanism data and real-time operation data of chemical production equipment during the production process.

[0037] Specifically, a server is a computer system or cluster that monitors the entire production process of a chemical production unit. Mechanistic data refers to data reflecting physical and chemical processes such as mass transfer, heat transfer, and chemical reactions within a chemical production unit. This data is derived through experiments, theoretical derivations, or calculations using physical models and is used to describe the basic operating mechanisms within the system. Mechanistic data reveals the behavior of the production unit under different operating conditions, including raw material conversion, reaction rates, and the flow of matter and heat. For example, reaction temperature, pressure, flow rate, and reactant and product concentrations are all mechanistic data. They reflect the fundamental characteristics of processes such as chemical reactions, heat transfer, and mass conversion within the reactor. Real-time operational data refers to process parameters and equipment status information collected in real time from the production unit. This data is acquired based on monitoring devices such as sensors and instruments and dynamically reflects the actual operating conditions during the production process. Real-time operational data reflects the real-time operating status of equipment, the actual execution of the process, and the health status of the system. For example, equipment operating temperature, pressure sensor data, flow meter readings, and energy consumption data are all real-time operational data. They can display the current process conditions of the production unit in real time and help maintenance personnel determine whether the unit is operating normally.

[0038] In one possible implementation, acquiring mechanistic data and real-time operational data of a chemical production unit during the production process specifically includes: determining the physical and chemical principles of the chemical production unit; determining the process behavior of the chemical production unit during the production process based on the physical and chemical principles to obtain mechanistic data; acquiring raw data of the chemical production unit during the production process in real time; and preprocessing the raw data to obtain real-time operational data, wherein the preprocessing includes noise reduction, filtering, and normalization.

[0039] Specifically, physical principles refer to the fundamental physical laws involved in chemical production plants. Examples include fluid dynamics, the laws of thermodynamics, and heat and mass transfer processes. These principles describe phenomena such as the flow of matter and the transfer of heat within the plant. Chemical principles refer to the chemical reaction processes within the plant, including reaction kinetics, equilibrium, and catalytic reactions. Through these chemical principles, we can understand how raw materials are transformed into products and how chemical reactions occur. Based on these physical and chemical principles, a mechanistic model of the chemical production plant can be established, describing the reaction processes, heat transfer, and mass transfer within the plant. For example, in a reactor, how reactants are transformed into products, how heat is transferred through heat transfer devices, and how matter flows through pipes. Mechanistic data, derived from these principles through theoretical calculations or experimental data, reflects the fundamental behaviors in the production process, such as reaction rates, temperature changes, and pressure distribution. Raw data refers to process data collected in real time by monitoring equipment such as sensors and instruments; these are typically unprocessed raw signals. Examples include temperature sensor data inside the reactor, pressure values ​​read by pressure sensors, liquid or gas flow rates collected by flow meters, and reactant concentrations recorded by material concentration sensors. These data are digital signals that reflect the current state of the production process in real time, but they come directly from the production site and may contain noise or irregular fluctuations.

[0040] Raw data may be affected by equipment errors, environmental factors, and other interferences, resulting in noise. Denoising refers to eliminating these interferences through techniques such as filtering and averaging. For example, the moving average method can be used to eliminate rapid fluctuations in temperature sensor data. Filtering removes unwanted frequency components from the data, retaining information meaningful for system analysis. For example, a low-pass filter can be used to remove high-frequency noise from temperature sensor data. To enable comparison or processing of raw data from different sources or with different dimensions under the same standard, normalization is required. Normalization transforms data with different dimensions into a unified order of magnitude; for example, temperature data, flow data, and pressure data are converted into standardized values ​​for easier subsequent processing.

[0041] S120. Based on the mechanistic data, a mechanistic model is constructed. The mechanistic model is used to reflect the mass transfer, energy balance and chemical reactions inside the chemical production plant.

[0042] Specifically, a mechanistic model is a mathematical model built upon mechanistic data. It quantitatively describes the processes of mass transfer, energy balance, and chemical reactions within a device. It does not rely solely on historical data or empirical rules, but rather derives the system's behavior by combining physical and chemical principles. Mass transfer refers to the process by which matter moves from one location or phase to another during chemical production. For example, in distillation, volatile components in a liquid evaporate into the gas phase and then condense back into a liquid. The mechanistic model details various parameters in this process, such as gas-liquid balance, mass migration rate, and concentration distribution. Energy balance refers to how various forms of energy, such as thermal and mechanical energy, flow and transform within a system during chemical processes. In a reactor, heat is input into the system through heaters, while some heat flows out through cooling devices; heat is also released or absorbed during the reaction. In the mechanistic model, the energy balance equations describe heat flow, heat loss, and temperature changes during this process, helping to optimize the efficiency of heating and cooling systems. A chemical reaction is the process by which reactants are converted into products, usually accompanied by the release or absorption of heat. In reaction vessels or other reaction equipment, the rate, conversion rate, and byproduct formation of chemical reactions are closely related to factors such as temperature, pressure, and reactant concentration. Mechanistic models describe the changes in various variables during the reaction process by considering factors such as reaction kinetics and reaction rate constants.

[0043] In one possible implementation, a mechanistic model is constructed based on mechanistic data, specifically including: determining the basic reaction relationships between substances within the chemical production unit based on the mechanistic data; determining the feed parameters, output parameters, intermediate product parameters, and by-product parameters of the chemical production unit based on the basic reaction relationships; determining the complex reaction relationships between substances within the chemical production unit based on the mechanistic data; determining the heat transfer parameters, heat loss parameters, and mass transfer parameters of the chemical production unit based on the complex reaction relationships; and generating a mechanistic model under natural constraints based on the feed parameters, output parameters, intermediate product parameters, by-product parameters, heat transfer parameters, heat loss parameters, and mass transfer parameters.

[0044] Specifically, fundamental reaction relationships refer to the simple, basic reaction equations or pathways between substances in a chemical reaction. For example, in a chemical reaction, two or more substances react with each other to produce products. Mechanistic models help describe how the reaction occurs through these fundamental reaction relationships. Feed parameters are the characteristics of the substances entering the reaction system, such as concentration, flow rate, and temperature. Mechanistic models can determine the input conditions for different reactants. Product parameters are the characteristics of the products formed by the reaction, including concentration and flow rate. Intermediate product parameters are the concentrations of temporary substances generated during the reaction. Byproduct parameters are the concentrations of unwanted additional substances generated during the reaction. For example, in an ethylene cracking reaction, feed parameters might include the flow rate, temperature, and pressure of ethylene; product parameters might include the concentrations and flow rates of acetylene and hydrogen; and byproducts might be undesirable low-molecular-weight hydrocarbons or carbon black.

[0045] Many chemical reactions involve not only simple basic reaction relationships but also complex interactions and reaction pathways between multiple substances. These complex reaction relationships involve multi-step reactions and variations in reaction rates. These complex relationships are derived through more in-depth experimental data from mechanistic data, reaction kinetic models, and reactor design. For example, in the ethylene cracking reaction, in addition to the basic reaction, there may be multiple reaction pathways, and the reaction rate may be strongly influenced by factors such as temperature and pressure. Mechanistic models need to consider these complex reactions, such as side reactions and pyrolysis reactions in ethylene cracking.

[0046] In chemical reactions, the input and output of heat directly affect the reaction rate and efficiency. Mechanistic models, by describing heat transfer, help predict the temperature distribution within the reactor. Heat loss parameters refer to the heat lost by the reactor and its components during operation. These heat losses affect the energy efficiency of the entire reaction system. In chemical reactions, substances need to be transferred between different phases, such as gas and liquid, solid and liquid. Mass transfer parameters help us understand how substances migrate between different regions during the reaction. For example, in the ethylene cracking reaction, the reactor temperature is a critical parameter. Mechanistic models need to consider how heat is transferred from the heater to the reactants and how to maintain a constant temperature within the reactor. Simultaneously, mass transfer processes may occur within the reactor, such as gas diffusion and flow.

[0047] Natural constraints refer to the operating conditions and physical limitations of the equipment during actual production. Examples include the reactor's maximum temperature, pressure, and flow rate limits. Based on known feed parameters, output parameters, intermediate product parameters, complex reaction relationships, heat transfer, and mass transfer parameters, a complete mechanistic model is generated under actual operating conditions and equipment limitations. For instance, in ethylene cracking, the reactor operates at a temperature between 800-1000°C, and the pressure is maintained within a certain range, such as under high pressure. The mechanistic model needs to incorporate these constraints to ensure that it reflects the reaction behavior and results under these conditions.

[0048] S130. Based on real-time operating data, a prediction model is constructed. The prediction model is used to predict the nonlinear relationship between the input and output variables of the chemical production unit.

[0049] Specifically, real-time operational data refers to data collected in real time during the actual operation of a chemical production plant, including dynamic data measured by various sensors and instruments. This data includes information reflecting the state of the production process, such as feed flow rate, temperature, pressure, concentration, and reaction rate. Real-time operational data is crucial for predictive models because it reflects the actual operating conditions of the plant. For example, assuming a reactor, real-time operational data might include: Feed flow rate: 50 m³ / s. 3 / h, reactor temperature: 400°C, pressure: 20 bar, gas concentration: 80%, ethylene: 10%, 10% other impurities.

[0050] Predictive models, in particular, use machine learning algorithms to infer future states or outputs based on existing real-time operational data. By learning from extensive historical data, they identify complex patterns or relationships between inputs and outputs. During this process, the model automatically adjusts its parameters to minimize prediction errors. The primary function of this model is to predict how output variables in a production process, such as product concentration and yield, change with input variables, such as feedstock flow rate and reaction temperature. Nonlinear relationships refer to relationships between input and output variables that are not simple linear but possess a degree of complexity. For example, the relationship between temperature and product reaction rates may not be linear; the yield may increase rapidly after the temperature rises to a certain critical value, while it may plateau at a certain temperature. Many variable relationships in chemical production processes are often nonlinear because they may be influenced by the interaction of multiple factors and reaction conditions.

[0051] For example, in chemical reactions, the relationship between reaction temperature and reaction rate is non-linear. As temperature increases, the reaction rate may initially accelerate, but above a certain temperature threshold, the rate may plateau or even experience thermal runaway. Predictive models need to learn from historical data. These non-linear input variables are controllable or measurable factors in chemical production plants, such as feedstock flow rate, temperature, and pressure. Output variables are the results of the reaction process, such as product concentration, yield, and energy efficiency. By building predictive models, the models can predict corresponding output values ​​based on inputs from real-time operational data.

[0052] In one possible implementation, a prediction model is constructed based on real-time operating data, specifically including: acquiring historical operating data of the chemical production unit; determining real-time operating data and historical operating data as input variables, processing them using a support vector machine algorithm to obtain output variables; acquiring the input and output variables corresponding to each of the multiple training processes, evaluating the training process using an evaluation algorithm, and obtaining the prediction model.

[0053] Specifically, historical operating data refers to data collected over a past period regarding the chemical production unit. This data includes feed flow rate, temperature, pressure, reactor concentration, product quality, etc., and can be used to analyze performance under different conditions during production. Input variables include real-time operating data and historical operating data, which will serve as input to the support vector machine (SVM) model to help the model predict future states or outputs. SVM is a machine learning algorithm used for classification and regression analysis. In this embodiment, SVM is used for regression analysis, that is, learning the relationship between input and output from the input data to predict the output variables of the chemical production unit, such as product concentration and yield.

[0054] Support Vector Machines (SVMs) find an optimal hyperplane in a high-dimensional space to accurately distinguish or predict the categories of input data. In regression problems, SVMs learn the relationship between input and output variables, generating a predictive model that can accurately predict future outputs. For example, in chemical reactor applications, SVMs can be used to predict product concentrations or reaction rates based on historical and real-time data. Input variables might include reaction temperature and flow rate, while the output variable is product concentration. Through regression analysis, the model can learn how temperature and flow rate affect product concentration. Building a predictive model requires multiple training iterations of the SVM. Each training iteration uses different input variables and learns based on the corresponding output variables. The goal of training is to allow the SVM to gradually find the optimal mapping between inputs and outputs. For example, the input variables might differ in each training iteration; some training might use temperature and flow rate as inputs, while others might incorporate more factors such as reactor pressure or substance concentration. After each training iteration, the model calculates the corresponding output, such as product concentration, and compares it with the actual results to optimize the model.

[0055] In addition, evaluation algorithms are used to verify the accuracy and reliability of the support vector machine (SVM) model. These include cross-validation and mean squared error, which measure the performance of the predictive model on different training data. The evaluation process helps determine whether the trained model can accurately predict on new data. If the model's evaluation results are good, it indicates that it can accurately predict the relationship between input and output variables. For example, suppose cross-validation during training shows that the SVM model based on historical and real-time data can accurately predict the product concentration in a reactor. When the evaluation results show that the model can predict new production data, the model can be confirmed as effective. Finally, the SVM, after multiple training and evaluations, will generate a predictive model. This model can predict output variables based on new real-time and historical data. The model's output can be used to guide the optimization of actual production processes. For example, this predictive model can be used for real-time monitoring of chemical production plants. When new real-time data is input into the model, it can predict the upcoming product concentration, helping operators adjust production parameters in a timely manner, thereby optimizing the production process.

[0056] S140. The mechanistic model and the prediction model are fused to obtain a fused model.

[0057] Specifically, by combining mechanistic and predictive models, the server can leverage the strengths of both to improve the accuracy and stability of predictions. Mechanistic models provide a physical understanding of the plant's behavior, while predictive models capture nonlinear and time-varying characteristics based on data-driven approaches. The fusion of these two models helps to more comprehensively describe the operational patterns of chemical plants, overcoming the limitations of a single model. For example, optimization algorithms, such as genetic algorithms and differential evolution algorithms, can automatically adjust the outputs of the mechanistic and predictive models, ensuring that the final fused result best reflects the actual behavior of the plant.

[0058] In one possible implementation, the mechanistic model and the prediction model are fused to obtain a fused model. Specifically, this includes: constructing a population containing both the mechanistic model and the prediction model, the population comprising multiple individuals, each representing a combination of models; performing differential mutation on the target individual to obtain mutated individuals, the target individual being any one of the multiple individuals; performing a crossover operation on the target individual and the mutated individuals to obtain candidate individuals; comparing the fitness values ​​of the candidate individuals and the target individual; if the fitness value of the candidate individual is greater than or equal to the fitness value of the target individual, then replacing the target individual with the candidate individual and entering the next generation of the population; repeating the differential mutation, crossover operation, and comparison until a preset number of iterations is reached or the fitness values ​​converge to obtain the fused model.

[0059] Specifically, a population refers to a set of possible model combinations. Each individual represents a combination of a mechanistic model and a predictive model. For example, you can combine different weights of the mechanistic and predictive models to create different individuals. Each individual in the population represents a potential solution. For example, suppose the mechanistic model describes temperature changes in a reactor, while the predictive model predicts the concentration of products in the reactor using historical data. An individual in the population might represent a combination of the mechanistic and predictive models, with the mechanistic model having a weight of 70% and the predictive model a weight of 30%. The server performs differential mutation on each target individual to generate a variant individual. Differential mutation refers to creating new candidate solutions by adding the differences between other individuals in the population to the target individual. Variant individuals may make certain adjustments to the model structure or parameters. Suppose the target individual has a mechanistic model weight of 70% and a predictive model weight of 30%. Differential mutation can generate a new variant individual by adding the combined weight differences of other individuals, such as an individual having a mechanistic model weight of 60% and a predictive model weight of 40%, to the target individual; for example, a mechanistic model weight of 75% and a predictive model weight of 25%.

[0060] Secondly, the server performs a crossover operation on the target individual and the mutated individual to generate a candidate individual. This crossover operation combines some features of the target individual and the mutated individual to form a new individual. For example, some parameters of the mechanistic model can be alternately copied from the target individual and the mutated individual. Assuming that the target individual and the mutated individual have different values ​​for certain parameters, such as the reaction rate constant and temperature change, the crossover operation can combine the temperature parameter of the target individual with the reaction rate constant of the mutated individual to generate a new candidate individual. Each individual has a fitness value, which measures its performance in solving the problem. The fitness value is calculated using an objective function, such as minimizing error or maximizing prediction accuracy. By comparing the fitness value of the candidate individual with that of the target individual, it is determined whether to replace the target individual with the candidate individual. Assuming the fitness value reflects the model's prediction error, the fitness value of the candidate individual is 0.01, and the fitness value of the target individual is 0.03. Since the candidate individual has a smaller error, it will replace the target individual.

[0061] Next, if the fitness value of a candidate individual is greater than or equal to that of the target individual, the candidate individual replaces the target individual and enters the next generation of the population. This indicates that the combined structure is better. If, after evaluation, the candidate individual is found to have more accurate prediction results and higher fitness, then the candidate individual is introduced into the population, replacing the original target individual. This process of differential mutation, crossover, and fitness comparison is repeated until a preset number of iterations is reached or the fitness value converges, meaning the model's prediction accuracy no longer significantly improves. For example, through multiple iterations, updating the individuals in the population each time makes the model combination increasingly accurate until sufficient accuracy is achieved or the optimization stopping condition is met, such as reaching the upper limit of the number of iterations or the error reaching an acceptable level. Finally, after multiple iterations of optimization, the best individuals obtained, i.e., the combination with the optimal fitness value, will constitute the final fusion model. Assuming that after multiple iterations, the optimal model combination is a combination of the mechanistic model and the predictive model, with the mechanistic model having a weight of 80% and the predictive model having a weight of 20%, this optimized combined model can more accurately predict the product concentration of a chemical reactor.

[0062] S150. By using a fusion model, the target optimization parameters are processed to obtain the optimization results.

[0063] Specifically, target optimization parameters refer to the specific parameters that need to be optimized in a chemical production plant. These may be input parameters, such as feed flow rate and temperature, or output parameters, such as product concentration and reaction efficiency, or overall performance indicators of the plant, such as energy efficiency, safety, and economy. For example, in a certain chemical production process, target optimization parameters might be reaction temperature, reaction time, feed flow rate, and product concentration. The purpose of optimizing these parameters is to increase product yield, optimize resource utilization, or reduce energy consumption. Processing target optimization parameters means using a fusion model to simulate and calculate different operating conditions to find the optimal operating point. The calculation results of the fusion model will provide predicted values ​​for each possible operating condition, and then, based on the optimization objective, such as minimizing energy consumption, maximizing yield, or improving product quality, the best parameter configuration will be selected. For example, in a chemical production process, a server can use a fusion model to simulate the impact of different feed flow rates, temperatures, and other parameters on product quality and calculate the optimization results under different parameter combinations. Through processing the target optimization parameters, the server will eventually obtain a set of optimal solutions, i.e., the optimization results. The optimization results can be the optimal values ​​of certain operating variables or the best operating scheme under certain process conditions. For example, optimization results may show that under specific reaction conditions, such as a temperature of 300°C and a feed flow rate of 100 L / min, the product concentration of the chemical reaction is maximized, while energy efficiency and reaction rate are optimal.

[0064] In one possible implementation, the target optimization parameters are processed through a fusion model to obtain the optimization result. Specifically, this includes: determining the target input variable and target output variable from the target optimization parameters; inputting the target input variable into the fusion model to obtain feature variables, including feature vectors; determining the positional relationship between the feature vectors and the hyperplane of the support vector machine model corresponding to the fusion model; calculating a first probability value and a second probability value based on the positional relationship, where the first probability value represents the probability value corresponding to the feature vector on the first side of the hyperplane, and the second probability value represents the probability value corresponding to the feature vector on the second side of the hyperplane, the first side and the second side being two opposite sides of the hyperplane; if it is determined that the feature vector is located on the first side and the first probability value is greater than or equal to a preset threshold, then the feature variable is determined to be the target output variable, and the optimization result is determined to be normal; if it is determined that the feature vector is located on the second side and the second probability value is greater than or equal to the preset threshold, then the feature variable is determined to be not the target output variable, and the optimization result is determined to be abnormal.

[0065] Specifically, the server inputs the target input variables into the fusion model. The fusion model combines a mechanistic model and a predictive model, processing these variables and outputting feature variables. Feature variables are processed variables that contain the potential relationship between the input variables and the output results. For example, if the target input variables are reaction temperature and flow rate, the fusion model might output feature variables representing the predicted reaction results under specific conditions. An eigenvector is a multidimensional representation of the feature variables, containing the various features calculated from the input variables by the fusion model. The support vector machine (SVM) model distinguishes different categories of data by constructing a hyperplane. The hyperplane divides the data space into two regions, each corresponding to a category. During this process, the output of the fusion model is compared with the hyperplane of the SVM model. This hyperplane is used to determine whether the target input variables lead to the expected optimization result.

[0066] The server determines the optimization result by judging the position of the feature vector on either side of the hyperplane of the support vector machine model. A first probability value represents the probability that the feature vector is on the first side of the hyperplane, and a second probability value represents the probability that it is on the second side. These two probability values ​​represent whether the feature vector meets the expected optimization result. If the feature vector is on the first side of the hyperplane and the first probability value is greater than or equal to a preset threshold, the system considers the optimization result normal. In this case, the feature variable is considered to meet the expected target, and the system confirms the optimization result is successful. If the feature vector is on the second side of the hyperplane and the second probability value is greater than or equal to the preset threshold, the optimization result is considered abnormal. This means that the input parameters have failed to achieve the expected optimization effect, which may lead to poor output or an inefficient production process.

[0067] In one possible implementation, refer to Figure 2 , Figure 2 Another schematic flowchart of a real-time optimization method for a chemical production apparatus provided in this application embodiment includes steps S210 to S220, which are as follows: S210: Receive an optimization request uploaded by a user through a user device, the optimization request including target optimization parameters; S220: Optimize the target optimization parameters according to the optimization request.

[0068] Specifically, user equipment refers to computers, consoles, or other terminal devices in chemical plants or laboratories through which users can issue optimization requests. An optimization request is a request made by a user to the server, indicating their desire to optimize certain process parameters to improve production efficiency, save costs, improve product quality, or achieve other goals. In an optimization request, the target optimization parameter is the variable or indicator that the user wishes to optimize. Depending on the different goals of chemical production, target optimization parameters may include: reaction temperature, pressure, flow rate, feedstock concentration, product yield, energy efficiency, cost, etc. For example, a user may want to maximize product yield by adjusting reaction temperature and flow rate, or reduce energy consumption without affecting quality. After receiving the request, the server will perform optimization calculations based on the target optimization parameters provided by the user. This process may involve adjustments to the chemical production equipment or updates to control strategies. Optimization methods may include finding the optimal solution based on mechanistic models, predictive models, machine learning algorithms, or through simulation, experimental design, etc. For example, if a user requests to increase yield while ensuring product quality, the server may analyze the interrelationships of parameters such as reaction temperature, pressure, and flow rate to determine the most suitable operating conditions and then generate a control strategy.

[0069] This application also provides a real-time optimization system for a chemical production plant, referring to... Figure 3 , Figure 3 This is a schematic diagram of a real-time optimization system for a chemical production plant provided in an embodiment of this application. The real-time optimization system is a server, which includes an acquisition module 31 and a processing module 32. The acquisition module 31 acquires mechanistic data and real-time operating data of the chemical production plant during the production process. The processing module 32 constructs a mechanistic model based on the mechanistic data, reflecting the mass transfer, energy balance, and chemical reactions within the chemical production plant. The processing module 32 also constructs a prediction model based on the real-time operating data, predicting the nonlinear relationship between the input and output variables of the chemical production plant. The processing module 32 fuses the mechanistic model and the prediction model to obtain a fused model. Finally, the processing module 32 processes the target optimization parameters using the fused model to obtain the optimization result.

[0070] In one possible implementation, the acquisition module 31 acquires the mechanism data and real-time operating data of the chemical production unit during the production process, specifically including: the processing module 32 determining the physical and chemical principles of the chemical production unit; the processing module 32 determining the process behavior of the chemical production unit during the production process based on the physical and chemical principles, and obtaining mechanism data; the acquisition module 31 acquiring the raw data of the chemical production unit during the production process in real time; and the processing module 32 preprocessing the raw data to obtain real-time operating data, the preprocessing including noise reduction, filtering, and normalization.

[0071] In one possible implementation, the processing module 32 constructs a mechanism model based on the mechanism data, specifically including: the processing module 32 determining the basic reaction relationships between substances within the chemical production unit based on the mechanism data; the processing module 32 determining the feed parameters, output parameters, intermediate product parameters, and by-product parameters of the chemical production unit based on the basic reaction relationships; the processing module 32 determining the complex reaction relationships between substances within the chemical production unit based on the mechanism data; the processing module 32 determining the heat transfer parameters, heat loss parameters, and mass transfer parameters of the chemical production unit based on the complex reaction relationships; and the processing module 32 generating a mechanism model under natural constraints based on the feed parameters, output parameters, intermediate product parameters, by-product parameters, heat transfer parameters, heat loss parameters, and mass transfer parameters.

[0072] In one possible implementation, the processing module 32 constructs a prediction model based on real-time operating data, specifically including: the acquisition module 31 acquiring historical operating data of the chemical production unit; the processing module 32 determining real-time operating data and historical operating data as input variables, processing them using a support vector machine algorithm to obtain output variables; and the acquisition module 31 acquiring the input and output variables corresponding to each of the multiple training processes, evaluating the training process using an evaluation algorithm to obtain a prediction model.

[0073] In one possible implementation, the processing module 32 fuses the mechanistic model and the prediction model to obtain a fused model. Specifically, the processing module 32 constructs a population containing the mechanistic model and the prediction model, the population including multiple individuals, each individual representing a combination structure of a model; the processing module 32 performs differential mutation on the target individual to obtain mutated individuals, the target individual being any one of the multiple individuals; the processing module 32 performs a crossover operation on the target individual and the mutated individual to obtain candidate individuals; the processing module 32 compares the fitness value corresponding to the candidate individual with the fitness value corresponding to the target individual; if the processing module 32 determines that the fitness value corresponding to the candidate individual is greater than or equal to the fitness value corresponding to the target individual, then the processing module 32 replaces the target individual with the candidate individual and enters the next generation of the population; the processing module 32 repeats the differential mutation, crossover operation, and comparison until a preset number of iterations is reached or the fitness value converges to obtain the fused model.

[0074] In one possible implementation, the processing module 32 processes the target optimization parameters through a fusion model to obtain optimization results. Specifically, the processing module 32 determines the target input variable and the target output variable from the target optimization parameters; the processing module 32 inputs the target input variable into the fusion model to obtain feature variables, which include feature vectors; the processing module 32 determines the positional relationship between the feature vectors and the hyperplane of the support vector machine model corresponding to the fusion model; based on the positional relationship, the processing module 32 calculates a first probability value and a second probability value, where the first probability value represents the probability value corresponding to the feature vector on the first side of the hyperplane, and the second probability value represents the probability value corresponding to the feature vector on the second side of the hyperplane, the first side and the second side being two opposite sides of the hyperplane; if the processing module 32 determines that the feature vector is located on the first side and the first probability value is greater than or equal to a preset threshold, then the feature variable is determined to be the target output variable, and the optimization result is determined to be normal; if the processing module 32 determines that the feature vector is located on the second side and the second probability value is greater than or equal to the preset threshold, then the feature variable is determined to be not the target output variable, and the optimization result is determined to be abnormal.

[0075] In one possible implementation, the acquisition module 31 receives an optimization request uploaded by the user through the user equipment, the optimization request including target optimization parameters; the processing module 32 optimizes the target optimization parameters according to the optimization request.

[0076] It should be noted that the system provided in the above embodiments is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0077] This application also provides an electronic device, with reference to... Figure 4 , Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 41, at least one network interface 44, a user interface 43, a memory 45, and at least one communication bus 42.

[0078] The communication bus 42 is used to enable communication between these components.

[0079] The user interface 43 may include a display screen and a camera. Optionally, the user interface 43 may also include a standard wired interface and a wireless interface.

[0080] Among them, the network interface 44 may optionally include a standard wired interface or a wireless interface (such as a WI-FI interface).

[0081] The processor 41 may include one or more processing cores. The processor 41 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 45, and by calling data stored in the memory 45. Optionally, the processor 41 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 41 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 41 and may be implemented as a separate chip.

[0082] The memory 45 may include random access memory (RAM) or read-only memory. Optionally, the memory 45 may include a non-transitory computer-readable storage medium. The memory 45 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 45 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 45 may also be at least one storage system located remotely from the aforementioned processor 41. Figure 4 As shown, the memory 45, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a real-time optimization method for a chemical production plant.

[0083] exist Figure 4In the electronic device shown, the user interface 43 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 41 can be used to call an application program stored in the memory 45 for a real-time optimization method of a chemical production device. When executed by one or more processors, the electronic device executes one or more methods as described in the above embodiments.

[0084] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0085] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0086] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0087] In the embodiments provided in this application, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between systems or units may be electrical or other forms.

[0088] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0089] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0090] 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 device (CMD). Based on this understanding, the technical solution of this application, 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 memory 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 of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0091] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for real-time optimization of a chemical production plant, characterized in that, The method comprises: acquiring mechanism data and real-time operation data of a chemical production device in a production process; constructing a mechanism model according to the mechanism data, the mechanism model being used to reflect material transfer, energy balance and chemical reaction inside the chemical production device; constructing a prediction model according to the real-time operation data, the prediction model being used to predict a nonlinear relationship between input variables and output variables of the chemical production device; fusing the mechanism model and the prediction model to obtain a fusion model; processing a target optimization parameter through the fusion model to obtain an optimization result; the mechanism model is constructed according to the mechanism data, specifically comprising: determining basic reaction relationships between materials inside the chemical production device according to the mechanism data; determining feed parameters, output parameters, intermediate product parameters and byproduct parameters of the chemical production device based on the basic reaction relationships; determining complex reaction relationships between materials inside the chemical production device according to the mechanism data; determining heat transfer parameters, heat loss parameters and mass transfer parameters of the chemical production device based on the complex reaction relationships; generating the mechanism model according to the feed parameters, the output parameters, the intermediate product parameters and the byproduct parameters, and the heat transfer parameters, the heat loss parameters and the mass transfer parameters under natural constraint conditions; the prediction model is constructed according to the real-time operation data, specifically comprising: acquiring historical operation data of the chemical production device; determining the real-time operation data and the historical operation data as the input variables, processing the input variables through a support vector machine algorithm to obtain the output variables; acquiring corresponding input variables and output variables of each training process, evaluating the training process through an evaluation algorithm to obtain the prediction model; the mechanism model and the prediction model are fused to obtain the fusion model, specifically comprising: constructing a population comprising the mechanism model and the prediction model, the population comprising a plurality of individuals, each individual representing a combination structure of a model; differential mutation is performed on a target individual to obtain a mutated individual, the target individual being any one of the individuals; a crossover operation is performed on the target individual and the mutated individual to obtain a candidate individual; a comparison is made between an adaptability value corresponding to the candidate individual and an adaptability value corresponding to the target individual; if it is determined that the adaptability value corresponding to the candidate individual is greater than or equal to the adaptability value corresponding to the target individual, the candidate individual replaces the target individual and enters a next generation population; the differential mutation, the crossover operation and the comparison are repeated until a preset iteration number is reached or adaptability value convergence is completed, so as to obtain the fusion model.

2. The real-time optimization method of a chemical production plant according to claim 1, characterized by, the mechanism data and the real-time operation data of the chemical production device in the production process are acquired, specifically comprising: determining physical principles and chemical principles of the chemical production device; determining process behaviors of the chemical production device in the production process based on the physical principles and the chemical principles to obtain the mechanism data; Real-time acquisition of original data of the chemical production device in the production process; The original data is pre-processed to obtain the real-time operation data, and the pre-processing includes denoising, filtering and normalization processing.

3. The real-time optimization method of a chemical production plant according to claim 1, characterized by, The fusion model is used to process the target optimization parameter to obtain an optimization result, specifically including: Determining a target input variable and a target output variable from the target optimization parameter; The target input variable is input into the fusion model to obtain a feature variable, and the feature variable includes a feature vector; Judging the positional relationship between the feature vector and the hyperplane of the support vector machine model corresponding to the fusion model; Based on the positional relationship, a first probability value and a second probability value are calculated, the first probability value representing the probability value corresponding to the feature vector on the first side of the hyperplane, and the second probability value representing the probability value corresponding to the feature vector on the second side of the hyperplane, the first side and the second side being two opposite sides of the hyperplane; If it is determined that the feature vector is located on the first side and the first probability value is greater than or equal to a preset threshold, it is determined that the feature variable is the target output variable, and it is determined that the optimization result is normal. If it is determined that the feature vector is located on the second side and the second probability value is greater than or equal to the preset threshold, it is determined that the feature variable is not the target output variable, and it is determined that the optimization result is abnormal.

4. The real-time optimization method of a chemical production plant according to claim 1, characterized by, The method further includes: Receiving an optimization request uploaded by a user through a user device, the optimization request including the target optimization parameter; Optimizing the target optimization parameter according to the optimization request.

5. A real-time optimization system for a chemical production plant for performing the real-time optimization method for a chemical production plant according to claim 1, characterized in that The real-time optimization system includes an acquisition module (31) and a processing module (32), wherein, The acquisition module (31) is configured to acquire mechanism data and real-time operation data of a chemical production device in a production process; The processing module (32) is configured to construct a mechanism model based on the mechanism data, the mechanism model being used to reflect the mass transfer, energy balance and chemical reaction inside the chemical production device; The processing module (32) is further configured to construct a prediction model based on the real-time operation data, the prediction model being used to predict the nonlinear relationship between the input variable and the output variable of the chemical production device; The processing module (32) is further configured to fuse the mechanism model and the prediction model to obtain a fusion model; The processing module (32) is further configured to process a target optimization parameter through the fusion model to obtain an optimization result.

6. An electronic device, comprising: The electronic device includes a processor (41), a memory (45), a user interface (43) and a network interface (44), the memory (45) is used to store instructions, the user interface (43) and the network interface (44) are used to communicate with other devices, and the processor (41) is used to execute the instructions stored in the memory (45) to make the electronic device execute the method of any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores instructions which, when executed, perform the method of any one of claims 1 to 4.

Images