Method and system for generating condition parameters for optimizing simulation of ethylene cracking process

Abstract

This invention provides a method, system, and storage medium for generating conditional parameters to optimize the simulation of an ethylene cracking process. The generation method includes the following steps: acquiring a pre-trained conditional optimization iterative model and an ethylene cracking furnace model, and constructing a conditional parameter generation model; acquiring multiple sets of sample data involving the true values ​​of a first feedstock parameter and product parameters related to the ethylene cracking process; simulating the ethylene cracking process based on the first feedstock parameter and the optimized conditional parameters output by the pre-trained conditional optimization iterative model, and outputting corresponding product parameter output values; adjusting the learning parameters of the conditional parameter generation model based on the true product parameter values ​​and the product parameter output values ​​to train the conditional parameter generation model; and acquiring a second feedstock parameter of the sample to be tested, and inputting the second feedstock parameter into the conditional parameter generation model to generate conditional parameters for optimizing the simulation of the ethylene cracking process.

Description

Technical Field

[0001] This invention relates to the field of chemical process simulation technology, and in particular to a method for generating conditional parameters for optimizing ethylene cracking process simulation, a system for generating conditional parameters for optimizing ethylene cracking process simulation, and a computer-readable storage medium. Background Technology

[0002] In the petrochemical industry, the efficiency and effectiveness of ethylene cracking directly impact the quality and cost of downstream products. Therefore, accurate simulation of the ethylene cracking process is beneficial for improving the yield and quality of ethylene products.

[0003] Currently, the Newton-Raphson method is commonly used to simulate ethylene cracking processes. This method sets initial conditions based on a conjectured value close to zero, and then uses the tangent line of the function to approximate the root. However, setting this initial conjectured value relies heavily on the operator's experience and intuition, resulting in low accuracy. If the initial conjectured value is far from zero, the Newton-Raphson method will fail to converge, and the simulation process requires manual adjustment of the initial conditions (e.g., heat flux, pressure, etc.). Therefore, using the Newton-Raphson model for ethylene cracking simulation suffers from insufficient intelligence and low accuracy in selecting initial conjectured parameters, directly leading to a high number of iterations in the ethylene cracking furnace model. This results in low efficiency and accuracy in the ethylene cracking process simulation, and may even lead to convergence failure.

[0004] In order to overcome the above-mentioned defects in the existing technology, there is an urgent need in the field for a condition parameter generation technology to optimize the simulation of ethylene cracking process, so as to generate and optimize the condition parameters of ethylene cracking process simulation with higher accuracy, thereby improving the efficiency, accuracy and intelligence of ethylene cracking process simulation. Summary of the Invention

[0005] The following provides a brief overview of one or more aspects to offer a basic understanding of them. This overview is not an exhaustive summary of all conceived aspects, nor is it intended to identify key or decisive elements of all aspects, nor to define the scope of any or all aspects. Its sole purpose is to present some concepts of one or more aspects in a simplified form as a prelude to the more detailed descriptions that follow.

[0006] To overcome the aforementioned deficiencies in the existing technology, this invention provides a method for generating conditional parameters for ethylene cracking process simulation, a system for generating conditional parameters for ethylene cracking process simulation, and a computer-readable storage medium, for generating and optimizing conditional parameters for ethylene cracking process simulation with higher precision, thereby improving the efficiency, accuracy, and intelligence of ethylene cracking process simulation.

[0007] Specifically, the method for generating conditional parameters for simulating an optimized ethylene cracking process according to the first aspect of the present invention includes the following steps: acquiring a pre-trained conditional optimization iterative model and an ethylene cracking furnace model, and constructing a conditional parameter generation model; acquiring multiple sets of sample data involving the true values ​​of the first feedstock parameter and product parameter of the ethylene cracking process; inputting the first feedstock parameter from each set of sample data into the conditional parameter generation model and the ethylene cracking furnace model respectively, firstly generating and outputting corresponding conditional parameters based on the first feedstock parameter through the conditional parameter generation model, then inputting the conditional parameters into the conditional optimization iterative model for iterative optimization, and then inputting the optimized conditional parameters output by the conditional optimization iterative model into the ethylene cracking furnace model to simulate the ethylene cracking process based on the first feedstock parameter and the optimized conditional parameters, and outputting corresponding product parameter output values; adjusting the learning parameters of the conditional parameter generation model based on the true values ​​of the product parameters and the product parameter output values ​​to train the conditional parameter generation model; and acquiring the second feedstock parameter of the sample to be tested, and inputting the second feedstock parameter into the conditional parameter generation model to generate conditional parameters for simulating the optimized ethylene cracking process.

[0008] Furthermore, in some embodiments of the present invention, the first raw material parameter and the second raw material parameter respectively include at least one of initial temperature, raw material component distribution, and raw material flow rate; the condition parameter includes the inlet heat flux and / or inlet pressure of the ethylene cracking furnace; and the product parameter includes the outlet temperature and / or outlet pressure indicating the product component distribution of the ethylene cracking furnace.

[0009] Furthermore, in some embodiments of the present invention, the step of constructing the conditional parameter generation model includes: obtaining XGBoost model data packages, the maximum depth of the decision tree, and the minimum weight of the child nodes. Each set of sample data is located at the root node of the decision tree of the XGBoost model; features are constructed using multidimensional data related to the first raw material parameters and the product parameters, and with the goal of minimizing the mean squared error, a greedy algorithm is used to traverse the values ​​of each feature dimension to determine the segmentation value of each feature dimension; the segmentation point of the corresponding feature dimension is determined according to each segmentation value, and the nodes of the decision tree are split according to the segmentation point; and in response to all split child nodes satisfying the conditions of the maximum depth of the decision tree and the minimum weight of the child nodes, the construction of the conditional parameter generation model is determined to be complete.

[0010] Furthermore, in some embodiments of the present invention, the step of training the conditional parameter generation model includes: randomly dividing the multiple sets of sample data according to a preset subsample ratio to obtain a training set and a test set; inputting the first raw material parameter of at least one training sample in the training set into the conditional parameter generation model to determine the predicted value of the conditional parameter; determining the loss function value according to the predicted value of the conditional parameter and its corresponding true value; and updating the learning parameters of the conditional parameter generation model according to a preset optimization objective and the loss function value until the loss function value converges and stabilizes, thereby obtaining a first training conditional parameter generation model to complete the training of the conditional parameter generation model.

[0011] Furthermore, in some embodiments of the present invention, the learning parameters include at least one of the following: maximum tree depth, learning rate, minimum weight condition for child nodes, L1 regularization coefficient, L2 regularization coefficient, proportion of subsamples, and column sampling proportion.

[0012] Furthermore, in some embodiments of the present invention, the sample data further includes a first conditional parameter corresponding to the true values ​​of the first raw material parameter and the product parameter. The step of training the conditional parameter generation model further includes: inputting the first raw material parameter of at least one test sample in the test set into the first training conditional parameter generation model to determine at least one corresponding second conditional parameter; inputting the first conditional parameter and the second conditional parameter of the at least one test sample into the conditional optimization iterative model for iterative optimization to determine the first iteration number and the second iteration number of the conditional optimization iterative model convergence respectively; retraining the conditional parameter generation model using the training sample data in the training set in response to the first iteration number being less than or equal to the second iteration number; and determining that the training of the conditional parameter generation model is complete in response to the first iteration number being greater than the second iteration number.

[0013] Furthermore, in some embodiments of the present invention, the loss function value MSE is expressed as:

[0014]

[0015] Where n is the total number of training samples, y i Let i be the true value of the i-th training sample. Generate the model's predicted value for the i-th training sample based on the conditional parameters.

[0016] Furthermore, in some embodiments of the present invention, the optimization objective is expressed as:

[0017]

[0018] Among them, Gi H is the gradient of the i-th training sample. i Let f(x) be the Hessian matrix of the i-th training sample. i ) represents the predicted value of the i-th training sample by the conditional parameter generation model, and Ω(f) is the regularization term.

[0019] Furthermore, the condition parameter generation system for optimizing ethylene cracking process simulation provided by the second aspect of the present invention includes: a memory storing computer instructions thereon; and a processor connected to the memory and configured to execute the computer instructions stored in the memory to implement the condition parameter generation method for optimizing ethylene cracking process simulation as described in the first aspect of the present invention.

[0020] Furthermore, according to a third aspect of the present invention, a computer-readable storage medium is provided thereon storing computer instructions. When the computer instructions are executed by a processor, the method for generating conditional parameters to optimize the simulation of the ethylene cracking process as described in the first aspect of the present invention is implemented. Attached Figure Description

[0021] The above-described features and advantages of the present invention will be better understood after reading the following detailed description of embodiments of the present disclosure in conjunction with the accompanying drawings. In the drawings, components are not necessarily drawn to scale, and components having similar related characteristics or features may have the same or similar reference numerals.

[0022] To overcome the aforementioned deficiencies in the existing technology, this invention provides a method for generating conditional parameters for ethylene cracking process simulation, a system for generating conditional parameters for ethylene cracking process simulation, and a computer-readable storage medium, for generating and optimizing conditional parameters for ethylene cracking process simulation with higher accuracy, thereby improving the prediction efficiency, control accuracy, and intelligence level of the ethylene cracking process.

[0023] Figure 1 A flowchart illustrating a method for generating condition parameters for simulating an optimized ethylene cracking process, according to some embodiments of the present invention, is shown.

[0024] Figure 2 A schematic diagram illustrating the principle of a method for generating conditional parameters for simulating an optimized ethylene cracking process, provided by some embodiments of the present invention, is shown.

[0025] Figure 3A A schematic diagram of the target shooting effect under fixed conditions provided by some embodiments of the present invention is shown.

[0026] Figure 3B A schematic diagram of an iterative path for fixed condition parameters provided according to some embodiments of the present invention is shown.

[0027] Figure 4A A schematic diagram of the target shooting effect generated by the conditional parameter generation model provided according to some embodiments of the present invention is shown.

[0028] Figure 4B A schematic diagram of the iterative path of condition parameters generated by the condition parameter generation model provided according to some embodiments of the present invention is shown. Detailed Implementation

[0029] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Although the description of the present invention is presented in conjunction with preferred embodiments, this does not mean that the features of the invention are limited to these embodiments. On the contrary, the purpose of describing the invention in conjunction with embodiments is to cover other options or modifications that may be derived based on the claims of the present invention. To provide a thorough understanding of the invention, many specific details will be included in the following description. The invention may also be implemented without using these details. Furthermore, to avoid confusion or obscuring the focus of the invention, some specific details will be omitted in the description.

[0030] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0031] Furthermore, the terms "upper," "lower," "left," "right," "top," "bottom," "horizontal," and "vertical" used in the following description should be understood as the orientations shown in the relevant paragraphs and accompanying drawings. These relative terms are for illustrative purposes only and do not imply that the described apparatus must be manufactured or operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0032] It is understood that although terms such as "first," "second," and "third" may be used herein to describe various components, regions, layers, and / or parts, these components, regions, layers, and / or parts should not be limited by these terms, and these terms are only used to distinguish different components, regions, layers, and / or parts. Therefore, the first components, regions, layers, and / or parts discussed below may be referred to as second components, regions, layers, and / or parts without departing from some embodiments of the present invention.

[0033] As mentioned above, the Newton-Raphson method is commonly used to simulate ethylene cracking processes. This method sets initial conditions based on a conjectured value close to zero, and then uses the tangent line of the function to approximate the root. However, setting this initial conjectured value relies heavily on the operator's experience and intuition, resulting in low accuracy. If the initial conjectured value is far from zero, the Newton-Raphson method will fail to converge, and the simulation process requires manual adjustment of the initial conditions (e.g., heat flux, pressure, etc.). Therefore, using the Newton-Raphson model for ethylene cracking simulation suffers from insufficient intelligence and low accuracy in selecting initial conjectured parameters, directly leading to a high number of iterations in the ethylene cracking furnace model. This results in low efficiency and accuracy in the ethylene cracking process simulation, and may even lead to convergence failure.

[0034] To overcome the aforementioned deficiencies in the existing technology, this invention provides a method for generating conditional parameters for ethylene cracking process simulation, a system for generating conditional parameters for ethylene cracking process simulation, and a computer-readable storage medium, for generating and optimizing conditional parameters for ethylene cracking process simulation with higher precision, thereby improving the efficiency, accuracy, and intelligence of ethylene cracking process simulation.

[0035] In some non-limiting embodiments, the method for generating conditional parameters for simulating the optimized ethylene cracking process provided in the first aspect of the present invention can be implemented based on the system for generating conditional parameters for simulating the optimized ethylene cracking process provided in the second aspect of the present invention. Specifically, the system for generating conditional parameters for simulating the optimized ethylene cracking process is configured with a memory and a processor. The memory includes, but is not limited to, the computer-readable storage medium provided in the third aspect of the present invention, on which computer instructions are stored. The processor is connected to the memory and configured to execute the computer instructions stored in the memory to implement the method for generating conditional parameters for simulating the optimized ethylene cracking process as described in the first aspect of the present invention.

[0036] The following will describe the working principle of the above-mentioned system for generating conditional parameters for ethylene cracking process simulation, using examples of methods for generating optimized ethylene cracking process simulation conditions. Those skilled in the art will understand that these examples of methods for generating optimized ethylene cracking process simulation conditions are merely non-limiting embodiments provided by this invention, intended to clearly demonstrate the main concepts of the invention and provide specific solutions convenient for public implementation, rather than limiting all functions or operating methods of the system. Similarly, these methods for generating optimized ethylene cracking process simulation conditions are also merely non-limiting embodiments provided by this invention, and do not limit the executing entities or execution order of the steps in these methods.

[0037] Please refer to the reference for details. Figures 1-2 , Figure 1 A flowchart illustrating a method for generating condition parameters for simulating an optimized ethylene cracking process, according to some embodiments of the present invention, is shown. Figure 2 A schematic diagram illustrating the principle of a method for generating conditional parameters for simulating an optimized ethylene cracking process, provided by some embodiments of the present invention, is shown.

[0038] like Figures 1-2 As shown, the method for generating conditional parameters for ethylene cracking process simulation can first perform step S1: obtain the pre-trained conditional optimization iterative model and ethylene cracking furnace model, and construct the conditional parameter generation model.

[0039] This ethylene cracking furnace model, built upon a free radical mechanism model, is a computational model that describes and predicts the chemical transformation, energy exchange, and hydrodynamic behavior during the cracking reaction. By incorporating various complex reactions during the cracking process, the model accurately simulates the physical and chemical processes within the cracking furnace, including the thermal decomposition of hydrocarbon molecules and the generation and consumption of free radicals. This provides users with a powerful tool to optimize operating conditions, increase ethylene yield, reduce energy consumption, and minimize byproduct generation.

[0040] In some embodiments, the present invention can utilize this ethylene cracking furnace model to simulate the reaction process on input feedstock parameters using optimized condition parameters output by a conditional optimization iterative model as initial conditions, thereby obtaining the output values ​​of product parameters within the cracking furnace, and simultaneously obtaining the product composition distribution results of the feedstock. Here, the feedstock parameters include at least one of initial temperature, feedstock composition distribution, and feedstock flow rate, and may even include furnace tube shape depending on the specific simulation conditions. The condition parameters may include the inlet heat flux and / or inlet pressure of the ethylene cracking furnace. The product parameters include the outlet temperature and / or outlet pressure, indicating the product composition distribution of the ethylene cracking furnace. Furthermore, this ethylene cracking furnace model can not only predict the cracking furnace performance under given initial conditions, but also help design and adjust process parameters to adapt to changes in feedstock or meet specific product requirements.

[0041] The conditional optimization iterative model can adopt the Newton-Raphson model, the core of which is an iterative method for solving nonlinear equations. It is used to iteratively optimize the conditional parameters output by the conditional parameter generation model by subtracting the ratio of the function value to its derivative from the current estimate in each iteration.

[0042] This condition parameter generation model can learn from a large amount of historical data and generate initial conditions for the ethylene cracking process (e.g., inlet heat flux and inlet pressure), thereby reducing the number of iterations in the condition optimization iterative model and improving the efficiency, accuracy, and intelligence of iterative convergence and ethylene cracking process simulation.

[0043] Specifically, constructing the conditional parameter generation model can be achieved by executing steps S11 to S14:

[0044] S11: Obtain the XGBoost model data package, the maximum depth of the decision tree, and the minimum weight of each child node. Each set of sample data is located at the root node of the XGBoost model's decision tree. S12: Construct features using multi-dimensional data related to the first raw material parameters and product parameters. Using a greedy algorithm to minimize the mean squared error, traverse the values ​​of each feature dimension to determine the segmentation value for each feature dimension. S13: Determine the segmentation point for each feature dimension based on the segmentation value, and split each node of the decision tree according to the segmentation point. S14: If all split child nodes satisfy the conditions for the maximum depth of the decision tree and the minimum weight of each child node, the construction of the conditional parameter generation model is considered complete.

[0045] XGBoost is a classic ensemble learning model widely used in prediction tasks involving complex data. The first set of feedstock and product parameters involves multidimensional data including feedstock composition distribution, initial temperature, feedstock flow rate, outlet temperature, and outlet pressure. Initial temperature is a critical input parameter for the cracking furnace, affecting the efficiency of the cracking process and the distribution of products.

[0046] Furthermore, for each possible split point, the greedy algorithm calculates the mean squared error (MSE) of the split data subset with respect to the target output (in this case, inlet heat flux and inlet pressure), and selects the split point that minimizes the MSE as the optimal split point. Since all parameters in this feedstock parameter set influence the selection of the optimal split point, to identify the features and split points that minimize the MSE, this conditional parameter generation model uses a greedy algorithm to traverse all features constructed from the multidimensional data and their possible split values ​​to select the split point. For example, feedstock flow rate may relate to heat load and material residence time, which can affect pyrolysis efficiency and product selectivity. During model training, the prediction error under different flow conditions can be analyzed to select the split point that minimizes the error to the greatest extent.

[0047] For each potential split point, the algorithm calculates the Mean Squared Effect (MSE) of the two resulting data subsets and selects the features and split points that maximize the reduction in MSE for node splitting. This process considers not only individual features but also the interactions between features and how they collectively affect the model's predictive performance. In this way, the model can identify the most influential features and split points, prioritizing them in subsequent iterations to improve the model's prediction accuracy and efficiency.

[0048] Please continue to refer to this. Figure 1 and Figure 2 After constructing the condition parameter generation model, the condition parameter generation system can execute steps S2 to S3:

[0049] S2: Obtain multiple sets of sample data involving the true values ​​of the first raw material parameters and product parameters in the ethylene cracking process; S3: Input the first raw material parameters from each set of sample data into the condition parameter generation model and the ethylene cracking furnace model respectively. First, the condition parameter generation model generates and outputs the corresponding condition parameters based on the first raw material parameters. Then, the condition parameters are input into the condition optimization iteration model for iterative optimization. Finally, the optimized condition parameters output by the condition optimization iteration model are input into the ethylene cracking furnace model to simulate the ethylene cracking process based on the first raw material parameters and optimized condition parameters, and output the corresponding product parameter output values.

[0050] Specifically, in the process of acquiring sample data, the condition parameter generation system can use the ethylene cracking furnace model to perform multiple simulation calculations of the ethylene cracking process, and record in detail the initial temperature, inlet heat flux, inlet pressure, feed component distribution, feed flow rate and furnace tube shape and other feed parameters and condition parameters used in each simulation calculation, as well as the corresponding calculated product parameters such as outlet temperature and outlet pressure, thereby obtaining multiple sets of sample data of the true values ​​of the first feed parameters and product parameters involved in the ethylene cracking process.

[0051] In this embodiment using liquefied petroleum gas (LPG) as feedstock, the simulated initial temperature ranges from 700K to 1200K; the feedstock flow rate ranges from 1.5kg / h to 3kg / h; and the simulated inlet heat flux ranges from 3e8W / m³. 2 Up to 4e8W / m 2 The simulated inlet pressure ranges from 2.5 MPa to 3 MPa; its key components include ethane, propane, n-butane, and isobutane. To accommodate different combinations of LPG, this condition parameter generation system uses the Dirichlet distribution method to extract feedstock characteristics, ensuring that the sum of the mass fractions of the four components—ethane, propane, n-butane, and isobutane—is always equal to 1, thereby constructing a four-dimensional uniform feature distribution space.

[0052] Subsequently, during the training of the conditional parameter generation model, this invention can input the first raw material parameter from each group of sample data into the conditional parameter generation model and the ethylene cracking furnace model respectively. First, the conditional parameter generation model generates and outputs the corresponding conditional parameters based on the first raw material parameter. Then, the conditional parameters are input into the conditional optimization iteration model for iterative optimization. Finally, the optimized conditional parameters output by the conditional optimization iteration model are input into the ethylene cracking furnace model to simulate the ethylene cracking process based on the first raw material parameter and the optimized conditional parameters, and output the corresponding product parameter values. Please continue to refer to... Figure 1 and Figure 2 The conditional parameter generation system can then execute step S4: adjust the learning parameters of the conditional parameter generation model based on the actual values ​​of the product parameters and the output values ​​of the product parameters, in order to train the conditional parameter generation model.

[0053] Specifically, when training the conditional parameter generation model, the method for generating conditional parameters for simulating the optimized ethylene cracking process can execute steps S41 to S44:

[0054] S41: Randomly divide multiple sets of sample data according to a preset subsample ratio (e.g., 7:3) to obtain training and test sets. S42: Input the first raw material parameter of at least one training sample from the training set into the conditional parameter generation model to determine the predicted value of the conditional parameter. S43: Determine the loss function value based on the predicted value of the conditional parameter and its corresponding true value. S44: Update the learning parameters of the conditional parameter generation model according to the preset optimization objective and the loss function value until the loss function value converges and stabilizes, obtaining the first training conditional parameter generation model to complete the training of the conditional parameter generation model.

[0055] In this way, by inferring the true value of the conditional parameter from the true value of the product parameter, and then comparing it with the predicted value of the conditional parameter, we can evaluate whether the predicted result is close to the true value.

[0056] Specifically, the data collection process needs to be as comprehensive and accurate as possible, as this directly affects the training effect of the subsequent machine learning model. After collecting 800 sets of simulation calculation data, the model training phase begins.

[0057] The loss function value MSE can be expressed as:

[0058]

[0059] Where n is the total number of training samples, y i Let i be the true value of the i-th training sample. Generate the model's predicted value for the i-th training sample based on the conditional parameters.

[0060] The optimization objective can be expressed as:

[0061]

[0062] Among them, G i H is the gradient of the i-th training sample. i Let f(x) be the Hessian matrix of the i-th training sample. i ) represents the predicted value of the model for the i-th training sample, where Ω(f) is the conditional parameter and Ω(f) is the regularization term.

[0063] Furthermore, the learning parameters may include at least one of the following: the maximum depth of the tree, the learning rate, the minimum weight condition for child nodes, the L1 regularization coefficient, the L2 regularization coefficient, the proportion of subsamples, and the column sampling proportion.

[0064] The values ​​of these learning parameters need to be carefully tuned. Initial tuning relies on domain knowledge and experience with similar problems, including setting a reasonable starting range for parameters such as the maximum tree depth and learning rate. For example, a smaller learning rate and a larger maximum tree depth are generally suitable for complex models, but may require more training time. This step helps to quickly narrow down the search space and provides a good starting point for grid search. After this initial experience-based tuning, a grid search method is used to further optimize the hyperparameters. Grid search finds the optimal parameter settings by traversing all possible combinations in our defined hyperparameter grid. Each set of parameters is used to train a new model, and its performance is evaluated through cross-validation. Finally, the parameter combination that performs best on the validation set is selected as the optimal learning parameters to ensure that the model performs optimally on the test set. In some embodiments, the method for generating conditional parameters for simulating the ethylene cracking process can set the maximum tree depth to 6, indicating that each tree has a maximum of 6 layers; the learning rate to 0.3, which is a parameter controlling the step size of weight updates in each iteration; the minimum weight of child nodes to 1, which is a parameter used to avoid overfitting; the L1 and L2 regularization coefficients to 10 and 2, respectively, which are used to control model complexity and prevent overfitting; and the proportion of random sampling of training samples and the proportion of feature sampling when constructing each tree to 0.7, i.e., training set:test set = 7:3, which is also an important means to prevent overfitting.

[0065] Furthermore, the XGBoost algorithm utilizes the gradient (i.e., the first derivative) of the loss function with respect to the model's predictions and the Hessian matrix (i.e., the second derivative) to guide parameter updates. This means that the update condition is to reduce the total loss of the current model across all training samples. In some embodiments of the invention, the update can be performed according to the following steps:

[0066] a) Calculate the gradient and Hessian matrix: For mean squared error, first calculate the gradient of the prediction error (the first derivative of the error) and the Hessian matrix (the second derivative of the error) for each sample. These derivatives express the magnitude of the gap between the current model prediction and the true value, as well as the curvature of the loss function.

[0067] b) Update steps:

[0068] i. Gradient descent: Uses gradient information to determine the direction of parameter updates, that is, adjusts the parameters in the direction that can reduce the loss the fastest.

[0069] ii. Learning rate: A preset learning rate is used to control the magnitude of each update step to prevent the update from being too large and overshooting the optimal solution.

[0070] iii. Regularization considerations: When updating parameters, L1 and L2 regularization terms are also considered to further control model complexity and prevent overfitting.

[0071] c) Specific update formula: Considering the gradient and Hessian matrix of the MSE, for each leaf node of each tree, the parameter update is actually achieved by optimizing the objective function that incorporates the gradient and Hessian matrix information. This process involves calculating an optimal score at each leaf node based on the gradient and Hessian matrix values ​​of all samples at that node.

[0072] Furthermore, in some embodiments, the present invention may preferably use at least one test sample from the aforementioned test set to evaluate the model performance after training the conditional parameter generation model. Specifically, in embodiments of the present invention, the sample data further includes a first conditional parameter corresponding to the true values ​​of the first raw material parameter and the product parameter. The conditional parameter generation method may also preferably include the following steps S441 to S444:

[0073] S441: Input the first raw material parameter of at least one test sample in the test set into the first training conditional parameter generation model to determine at least one corresponding second conditional parameter. S442: Input the first and second conditional parameters of at least one test sample into the conditional optimization iterative model for iterative optimization to determine the first and second iterations for convergence of the conditional optimization iterative model, respectively. S443: In response to the first iteration count being less than or equal to the second iteration count, retrain the conditional parameter generation model using the training sample data in the training set. S444: In response to the first iteration count being greater than the second iteration count, determine that the training of the conditional parameter generation model is complete.

[0074] Therefore, by comparing the model's predicted inlet pressure and inlet heat flux with the actual values ​​in the validation set, the model's prediction accuracy can be obtained. When the model's prediction performance fails to meet expectations, the model's hyperparameter range is readjusted, and hyperparameter optimization, training, and evaluation are performed again until the model's prediction performance meets the requirements.

[0075] Subsequently, in some embodiments of the present invention, after the construction and optimization of the condition parameter generation model are completed, the method for generating condition parameters to optimize the simulation of the ethylene cracking process can continue to execute step S5: obtaining the second feedstock parameters of the sample to be tested, and inputting the second feedstock parameters into the condition parameter generation model to generate the condition parameters for optimizing the simulation of the ethylene cracking process. Here, the second feedstock parameters may include at least one of initial temperature, feedstock component distribution, and feedstock flow rate, and may even include furnace tube shape depending on the specific simulation conditions.

[0076] Furthermore, in some embodiments of the present invention, those skilled in the art can sequentially combine the above-mentioned condition parameter generation model, condition optimization iteration model, and ethylene cracking furnace model, and use this combined model to simulate the ethylene cracking process in order to predict the corresponding ethylene cracking products.

[0077] Specifically, the second raw material parameters of the sample to be tested are input into the conditional parameter generation model and the ethylene cracking furnace model, respectively. First, the conditional parameter generation model generates and outputs corresponding conditional parameters based on the second raw material parameters. Then, these conditional parameters are input into the conditional optimization iterative model for iterative optimization. Next, the optimized conditional parameters output by the conditional optimization iterative model are input into the ethylene cracking furnace model to simulate the ethylene cracking process based on the second raw material parameters and the optimized conditional parameters, and output corresponding product parameter values. Finally, the ethylene cracking products are determined based on the product parameter output values. These product parameters include the outlet temperature and / or outlet pressure, which indicate the product composition distribution of the ethylene cracking furnace.

[0078] Please refer to further information. Figures 3A to 4B , Figure 3A A schematic diagram of the target shooting effect under fixed conditions provided by some embodiments of the present invention is shown. Figure 3B A schematic diagram of an iterative path for fixed condition parameters provided according to some embodiments of the present invention is shown. Figure 4A A schematic diagram of the target shooting effect generated by the conditional parameter generation model provided according to some embodiments of the present invention is shown. Figure 4B A schematic diagram of the iterative path of condition parameters generated by the condition parameter generation model provided according to some embodiments of the present invention is shown.

[0079] like Figures 3A-3BAs shown, this conditional optimization iterative model uses fixed conditional parameters (inlet heat flux and inlet pressure) selected based on empirical guesses as input values ​​and performs 100 Newton-Raphson iterations to obtain the optimized conditional parameters. In the figure, the farther each discrete point is from the origin (0, 0), the further it deviates from the true value of the conditional parameters; the numbers marked on each discrete point indicate the number of system iterations. Correspondingly, multiple iterations are required for the outlet temperature and outlet pressure to reach convergence. Green dots represent the iteration start points, red dots represent the iteration end points, and blue lines represent the iteration paths.

[0080] like Figures 4A-4B As shown, this conditional optimization iterative model uses the conditional parameters (inlet heat flux and inlet pressure) generated by the conditional parameter generation model as input values, and performs Newton-Raphson iterative calculations to obtain the optimized conditional parameters. In the figure, the farther each discrete point is from the origin (0, 0), the further it deviates from the true value of the conditional parameters; the numbers marked on each discrete point indicate the number of system iterations. Correspondingly, multiple iterations are required for the outlet temperature and outlet pressure to reach convergence. Green dots represent the iteration start points, red dots represent the iteration end points, and blue lines represent the iteration paths.

[0081] As shown in the figure, when the conditional optimization iterative model uses fixed conditional parameters (inlet heat flux and inlet pressure) selected based on empirical guesses as input values, the outlet temperature and pressure are far from the origin (0, 0), and the number of iterations is high. In contrast, the conditional optimization iterative model uses conditional parameters (inlet heat flux and inlet pressure) generated by the conditional parameter generation model as input values. These outlet temperature and pressure are closer to the origin (0, 0), and the number of iterations is low. Therefore, the conditional parameters generated by the conditional parameter generation model provide conditional parameters closer to the true values ​​than the fixed conditional parameters selected based on empirical guesses.

[0082] In summary, the method for generating conditional parameters for ethylene cracking process simulation, the system for generating conditional parameters for ethylene cracking process simulation, and the computer-readable storage medium provided by this invention can be used to generate and optimize conditional parameters for ethylene cracking process simulation with higher precision, thereby improving the efficiency, accuracy, and intelligence of ethylene cracking process simulation.

[0083] Although the methods described above are illustrated and depicted as a series of actions for the sake of simplicity, it should be understood and appreciated that these methods are not limited by the order of the actions, as some actions may occur in a different order and / or concurrently with other actions from the illustrations and descriptions herein or not illustrated and described herein but which may be understood by those skilled in the art, according to one or more embodiments.

[0084] The prior description of this disclosure is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to this disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not intended to be limited to the examples and designs described herein, but should be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating conditional parameters to optimize the simulation of ethylene cracking process, characterized in that, Includes the following steps: A pre-trained conditional optimization iterative model and ethylene cracking furnace model are obtained, and a conditional parameter generation model is constructed. The steps for constructing the conditional parameter generation model include: obtaining XGBoost model data packages, the maximum depth of the decision tree, and the minimum weight of the child nodes, wherein each set of sample data is located at the root node of the decision tree of the XGBoost model; constructing features based on the multidimensional data involving the first raw material parameters and the product parameters, and determining the segmentation value of each feature dimension by traversing the values ​​of each feature dimension using a greedy algorithm with the goal of minimizing the mean square error; determining the segmentation point of the corresponding feature dimension according to each segmentation value, and splitting each node of the decision tree according to the segmentation point; and determining that the construction of the conditional parameter generation model is complete when all the split child nodes satisfy the conditions of the maximum depth of the decision tree and the minimum weight of the child nodes. Obtain multiple sets of sample data involving the true values ​​of the first feedstock parameters and product parameters related to the ethylene cracking process; The first raw material parameters in each group of sample data are respectively input into the condition parameter generation model and the ethylene cracking furnace model. First, the condition parameter generation model generates and outputs the corresponding condition parameters based on the first raw material parameters. Then, the condition parameters are input into the condition optimization iteration model for iterative optimization. Finally, the optimized condition parameters output by the condition optimization iteration model are input into the ethylene cracking furnace model to simulate the ethylene cracking process based on the first raw material parameters and the optimized condition parameters, and output the corresponding product parameter output values. Based on the actual values ​​and output values ​​of the product parameters, the learning parameters of the conditional parameter generation model are adjusted to train the conditional parameter generation model; and The second raw material parameters of the sample to be tested are obtained and input into the condition parameter generation model to generate condition parameters for optimizing the simulation of the ethylene cracking process.

2. The condition parameter generation method as described in claim 1, characterized in that, The first raw material parameter and the second raw material parameter respectively include at least one of initial temperature, raw material composition distribution, and raw material flow rate. The condition parameter includes the inlet heat flux and / or inlet pressure of the ethylene cracking furnace. The product parameter includes the outlet temperature and / or outlet pressure indicating the product composition distribution of the ethylene cracking furnace.

3. The condition parameter generation method as described in claim 1, characterized in that, The steps for training the conditional parameter generation model include: The multiple sets of sample data are randomly divided according to a preset sub-sample ratio to obtain a training set and a test set; Input the first raw material parameter of at least one training sample in the training set into the conditional parameter generation model to determine the predicted value of the conditional parameter; Based on the predicted values ​​of the condition parameters and their corresponding true values, the loss function value is determined; and Based on the preset optimization objective and the loss function value, the learning parameters of the conditional parameter generation model are updated until the loss function value converges and stabilizes, thereby obtaining the first training conditional parameter generation model to complete the training of the conditional parameter generation model.

4. The condition parameter generation method as described in claim 3, characterized in that, The learning parameters include at least one of the following: maximum tree depth, learning rate, minimum weight condition for child nodes, L1 regularization coefficient, L2 regularization coefficient, proportion of subsamples, and column sampling proportion.

5. The condition parameter generation method as described in claim 3, characterized in that, The sample data also includes a first conditional parameter corresponding to the true values ​​of the first raw material parameter and the product parameter. The step of training the conditional parameter generation model further includes: The first raw material parameter of at least one test sample in the test set is input into the first training condition parameter generation model to determine at least one corresponding second condition parameter. The first conditional parameter and the second conditional parameter of the at least one test sample are respectively input into the conditional optimization iterative model for iterative optimization, so as to determine the first iteration number and the second iteration number for the convergence of the conditional optimization iterative model. In response to the first iteration number being less than or equal to the second iteration number, the conditional parameter generation model is retrained using the training sample data in the training set; and If the first iteration number is greater than the second iteration number, it is determined that the training of the condition parameter generation model is complete.

6. The condition parameter generation method as described in claim 3, characterized in that, The loss function value It is represented as: in, The total number of training samples. For the first The true value of each training sample. Generate a model for the conditional parameters for the first... The predicted value of each training sample.

7. The condition parameter generation method as described in claim 3, characterized in that, The optimization objective is expressed as: in, For the first The gradient of each training sample. For the first The Hessian matrix of each training sample. Generate a model for the conditional parameters for the first... Predicted values ​​for each training sample. This is a regularization term.

8. A system for generating conditional parameters to optimize the simulation of ethylene cracking processes, characterized in that, include: Memory, on which computer instructions are stored; as well as A processor, connected to the memory, and configured to execute computer instructions stored in the memory to implement the method for generating conditional parameters for simulating an optimized ethylene cracking process as described in any one of claims 1 to 7.

9. A computer-readable storage medium storing computer instructions thereon, characterized in that, When the computer instructions are executed by the processor, the method for generating conditional parameters for optimizing the simulation of the ethylene cracking process as described in any one of claims 1 to 7 is implemented.

Images