A method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with oil property calculation capabilities

Abstract

The present invention belongs to the technical field of molecular dynamics models, and specifically relates to a method for constructing a molecular-level catalytic cracking process reaction kinetics model that incorporates oil property calculation capabilities. Verified by industrial data, the model provided by the present invention shows a high degree of agreement between the calculated values ​​for catalytic cracking product yield, product group composition content, product typical molecular content, and gasoline and diesel distillation ranges and industrial values. The relative errors of gasoline density and octane number, as well as diesel density and cetane number, compared with industrial data are all less than 2.2%, demonstrating that the molecular-level catalytic cracking process reaction kinetics model that incorporates oil property calculation capabilities has high reliability and can predict product composition and properties at the molecular level.

Description

Technical Field

[0001] The present invention belongs to the technical field of molecular dynamics models, and in particular relates to a method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with an oil property calculation function. Background Art

[0002] As a critical non-renewable resource, achieving efficient utilization of petroleum has become a research hotspot in the refining and chemical industry. By establishing reaction kinetic models for petroleum processing and predicting the impact of process conditions on product distribution and properties, guidance can be provided for optimizing industrial plants at a relatively low cost.

[0003] Because the traditional multi-lumped reaction kinetics model has too rough a division of raw materials and products, it is difficult to provide refineries with accurate and detailed product distribution and property information, which limits the efficient utilization of petroleum resources. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method for constructing a molecular-level catalytic cracking process reaction kinetic model combined with the oil property calculation function, which can predict the product composition distribution and properties of petroleum catalytic cracking at the molecular level.

[0005] The present invention provides a method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with an oil property calculation function, comprising the following steps:

[0006] (1) Based on the structure-guided lumping method, a data set for training a computational model of petroleum molecular properties is constructed; based on the data set, a computational model of petroleum molecular properties is constructed using an artificial neural network algorithm;

[0007] (2) Based on the petroleum molecular property calculation model and in combination with the mixing rules from molecular properties to oil product properties, a calculation model for oil product properties is constructed;

[0008] The oil product property calculation model includes a gasoline distillation range calculation model, a gasoline density calculation model, an octane number property calculation model, a diesel distillation range calculation model, a diesel density calculation model and a cetane number property calculation model;

[0009] (3) Based on the structure-guided lumping method, combined with the composition and property analysis data of the catalytic cracking feedstock oil, a simulated annealing algorithm and the oil property calculation model are used to obtain the molecular composition matrix of the catalytic cracking feedstock oil;

[0010] (4) formulating catalytic cracking reaction rules based on the structure-guided lumping method and the reaction mechanism of catalytic cracking, and then generating multiple reaction paths and a catalytic cracking reaction network composed of the multiple reaction paths according to the reaction rules and the molecular composition matrix of the catalytic cracking feedstock oil;

[0011] (5) converting the catalytic cracking reaction network into a set of reaction kinetic differential equations based on the reaction rate constant, solving the set of reaction kinetic differential equations using the improved Runge-Kutta method, and obtaining a molecular-level catalytic cracking process reaction kinetic model;

[0012] The reaction rate constant is calculated by formula I:

[0013]

[0014] Among them, k B is a constant, k B =1.38×10 -23 J / K;

[0015] h is Planck's constant;

[0016] R is the ideal gas constant;

[0017] T is the reaction temperature, K;

[0018] ΔS is the entropy change of the reaction, J / (mol·K);

[0019] ΔE is the energy barrier of the reaction, kcal / mol;

[0020] P is the reaction pressure, MPa;

[0021] (6) Combining the molecular-level catalytic cracking process reaction kinetics model and the oil property calculation model, a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function is constructed.

[0022] Preferably, the calculation model of petroleum molecular properties is constructed as follows:

[0023] Based on the structure-guided lumping method, a dataset was constructed for training computational models of petroleum molecular properties.

[0024] According to the data set, the optimal number of hidden layers and the optimal number of neurons in the hidden layer are obtained through the k-fold cross validation algorithm in the artificial neural network algorithm;

[0025] According to the optimal number of hidden layers and the optimal number of neurons in the hidden layers, a calculation model of petroleum molecular properties is constructed through the neural network algorithm.

[0026] Preferably, the molecular composition matrix of the catalytic cracking feedstock oil is obtained as follows:

[0027] Determining seed molecules based on composition information and property analysis data of catalytic cracking feedstock oil, and then performing structural vector expansion on the seed molecules by adding side chains to construct a catalytic cracking feedstock oil molecular matrix;

[0028] According to the composition information and property analysis data of crude oil and the structure vector of each molecule in the catalytic cracking feedstock oil molecular matrix, a simulated annealing algorithm is used to perform multi-objective optimization to calculate the content of each structure vector in the catalytic cracking feedstock oil molecular matrix to obtain the catalytic cracking feedstock oil molecular composition matrix.

[0029] Preferably, the reaction rules include reactant selection rules and product generation rules.

[0030] Preferably, the method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function is:

[0031] The product molecular composition matrix is ​​calculated through the molecular-level catalytic cracking process reaction kinetics model. The product molecular composition matrix is ​​input into the oil property calculation model to obtain the product properties.

[0032] Preferably, the k-fold cross validation algorithm includes:

[0033] Divide the dataset into k subsets, use the data from each subset as a validation set, and use the remaining k-1 subsets as training sets, thereby obtaining k models.

[0034] The performance of the k models is evaluated by taking the sum of the root mean square errors of the validation sets in the k models, and then the number of hidden layers and neurons of the model is determined; the formula for calculating the root mean square error is Formula II:

[0035]

[0036] Where N is the number of samples in the validation set, Yexp,i is the experimental data of the property of sample i, and Ypre,i is the calculated data of the property of sample i.

[0037] Preferably, the petroleum molecular property calculation model includes prediction models of petroleum molecule boiling point, critical temperature, critical volume, critical pressure, density, octane number and cetane number respectively.

[0038] Preferably, the method for obtaining the composition and property analysis data of the catalytic cracking feedstock oil is:

[0039] The composition and property analysis data of catalytic cracking feed oil were obtained through column chromatography separation and nuclear magnetic resonance analysis.

[0040] The present invention also provides a method for predicting catalytic cracking products and their properties using a molecular-level catalytic cracking process reaction kinetics model obtained by the construction method and combined with oil property calculation function, comprising the following steps:

[0041] The molecular composition matrix of the catalytic cracking feedstock oil is input into the molecular-level catalytic cracking process reaction kinetics model obtained by the construction method. After the molecular-level catalytic cracking process reaction kinetics model is run, the composition distribution and properties of the catalytic cracking products are output.

[0042] The present invention also provides a computer-readable storage medium having computer instructions stored thereon. When the computer instructions are executed, the steps of the method for constructing a reaction kinetic model of a molecular-level catalytic cracking process are implemented.

[0043] The present invention provides a method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with the function of calculating oil properties. Based on the reaction mechanism, a molecular-level catalytic cracking process reaction kinetics model combined with the function of calculating oil properties is constructed. The model has high reliability, conforms to the reaction laws of the catalytic cracking process, and can predict the product composition and properties at the molecular level.

[0044] The present invention provides a method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with an oil product property calculation function, comprising the following steps: (1) constructing a data set for training a petroleum molecular property calculation model based on a structure-guided lumping method; constructing a petroleum molecular property calculation model based on the data set by using an artificial neural network algorithm; (2) constructing an oil product property calculation model based on the petroleum molecular property calculation model by combining a mixing rule from molecular properties to oil product properties; (3) constructing an oil product property calculation model based on a structure-guided lumping method by combining composition and property analysis data of catalytic cracking feedstock oil by using a simulated annealing algorithm and The oil product property calculation model is used to obtain the molecular composition matrix of the catalytic cracking feed oil; (4) based on the structure-guided lumping method and the reaction mechanism of catalytic cracking, the reaction rules of catalytic cracking are formulated, and then according to the reaction rules and the molecular composition matrix of the catalytic cracking feed oil, multiple reaction paths and a catalytic cracking reaction network composed of multiple reaction paths are generated; (5) combined with the reaction rate constant, the catalytic cracking reaction network is converted into a reaction kinetic differential equation system, which is solved using the improved Runge-Kutta method to obtain a molecular-level catalytic cracking process reaction kinetic model; the reaction rate constant is calculated by formula I: Among them, k B is a constant, k B =1.38×10 -23J / K; h is Planck's constant; R is the ideal gas constant; T is the reaction temperature, K; ΔS is the entropy change of the reaction, J / (mol·K); ΔE is the energy barrier of the reaction, kcal / mol; P is the reaction pressure, MPa. (6) Combining the molecular-level catalytic cracking process reaction kinetics model and the oil property calculation model, a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function is constructed. Through industrial data verification, the model provided by the present invention has a high degree of consistency with the industrial values ​​for the catalytic cracking product yield, the group composition content in the product, the typical molecular content in the product, and the gasoline and diesel distillation range. The relative errors of the gasoline density and octane number and the diesel density and cetane number indicators with the industrial data are all less than 2.2%, indicating that the molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function has high reliability and can predict the product composition and properties at the molecular level. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a process flow chart of the catalytic cracking unit of Example 1;

[0046] Figure 2 22 structural units and their meanings in the structure-guided lumping method of Example 1;

[0047] Figure 3 This is a schematic diagram of the k-fold cross validation algorithm of Example 1;

[0048] Figure 4 This is a diagram showing the framework for developing the petroleum molecular property computational model of Example 1;

[0049] Figure 5 This is a comparison chart of the calculated values ​​and experimental values ​​of the petroleum molecular property calculation model in Example 1;

[0050] Figure 6 This is a block diagram of the simulated annealing algorithm calculation in Example 1;

[0051] Figure 7 This is a calculation block diagram of the molecular-level catalytic cracking reaction kinetic model of Example 1;

[0052] Figure 8 This is a comparison chart of the calculated and industrial values ​​of group composition content in gasoline and diesel of Example 1;

[0053] Figure 9 The figure is a comparison chart of the calculated and industrial values ​​of the gasoline and diesel distillation ranges of Example 1;

[0054] Figure 10 This is a graph showing the effect of reaction temperature on gasoline properties in Example 1;

[0055] Figure 11This is a diagram showing the effect of reaction temperature on diesel properties in Example 1. DETAILED DESCRIPTION

[0056] The present invention provides a method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with an oil property calculation function, comprising the following steps:

[0057] (1) Based on the structure-guided lumping method, a data set for training a petroleum molecular property calculation model is constructed; based on the data set, a petroleum molecular property calculation model is constructed using an artificial neural network algorithm;

[0058] (2) Based on the calculation model of petroleum molecular properties and combined with the mixing rules from molecular properties to oil properties, a calculation model of oil properties is constructed;

[0059] (3) Based on the structure-guided lumping method, combined with the composition and property analysis data of FCC feedstock, the molecular composition matrix of FCC feedstock is obtained through the simulated annealing algorithm and the oil property calculation model;

[0060] (4) Based on the structure-guided lumping method and the reaction mechanism of catalytic cracking, the reaction rules of catalytic cracking are formulated. Then, according to the reaction rules and the molecular composition matrix of catalytic cracking feedstock oil, multiple reaction paths and a catalytic cracking reaction network composed of multiple reaction paths are generated;

[0061] (5) combining the reaction rate constants, converting the catalytic cracking reaction network into a set of reaction kinetic differential equations, solving the set of reaction kinetic differential equations using the improved Runge-Kutta method, and obtaining a molecular-level catalytic cracking process reaction kinetic model;

[0062] The reaction rate constant is calculated by formula I:

[0063]

[0064] Among them, k B is a constant, k B =1.38×10 -23 J / K;

[0065] h is Planck's constant;

[0066] R is the ideal gas constant;

[0067] T is the reaction temperature, K;

[0068] ΔS is the entropy change of the reaction, J / (mol·K);

[0069] ΔE is the energy barrier of the reaction, kcal / mol;

[0070] P is the reaction pressure, MPa.

[0071] (6) Combining the molecular-level catalytic cracking process reaction kinetics model and the oil property calculation model, a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function is constructed.

[0072] The present invention constructs a data set for training a petroleum molecular property calculation model based on a structure-guided lumping method; and constructs a petroleum molecular property calculation model based on the data set and an artificial neural network algorithm.

[0073] In the present invention, the step of constructing a petroleum molecular property calculation model is preferably:

[0074] A digital description of catalytic cracking feedstock oil was performed based on 22 structural units in the structure-guided lumping method, and a dataset was constructed for training computational models of oil molecular properties.

[0075] Constructing a petroleum molecular property calculation model based on a k-fold cross validation algorithm and the data set;

[0076] According to the optimal number of hidden layers and the optimal number of neurons in the hidden layer, a calculation model of petroleum molecular properties is constructed using the neural network (ANN) algorithm.

[0077] Figure 2 22 structural units and their meanings in the structure-guided lumping method of the embodiment, as shown in FIG. Figure 2 As shown in the figure, a petroleum molecule can be represented by a structure vector consisting of 22 structural units in one row, specifically: A6, A4 and A2 are used to describe the aromatic ring skeleton structure of aromatic hydrocarbon molecules; N6, N5, N4, N3, N2 and N1 are used to describe the cycloalkane ring skeleton structure of cycloalkane molecules; R, br, me and IH are used to describe the skeleton structure of alkanes, alkenes and side chain methyl groups on the ring; AA is used to describe the connection between nuclei; NS, NN and NO are used to describe the sulfur, nitrogen and oxygen atoms between CC, respectively; RS, RN and RO are used to describe the sulfur, nitrogen and oxygen atoms between CH, respectively; AN is used to describe the nitrogen atom on the aromatic ring skeleton; KO is used to describe the oxygen atom connected to the carbon atom with a double bond structure.

[0078] In the present invention, the step of constructing a petroleum molecular property calculation model based on the k-fold cross validation algorithm in the artificial neural network algorithm and the data set is preferably:

[0079] Divide the data set into k subsets, use the data of each subset as a validation set, and use the remaining k-1 subsets as training sets to obtain k models;

[0080] The performance of the model was evaluated by taking the sum of the root mean square error (RMSE) of the validation set in the k models to obtain the optimal number of hidden layers and the optimal number of neurons in the hidden layer of the k models;

[0081] The root mean square error is calculated according to formula II:

[0082]

[0083] In formula II, N is the number of samples in the validation set, Y exp,i is the experimental data of the properties of sample i, Y pre,i Calculate data for the properties of sample i;

[0084] In the present invention, the k-fold cross-validation algorithm is used to find the optimal number of hidden layers and the optimal number of neurons in the hidden layer of the model, which can improve the reliability of the model and prevent the model from overfitting.

[0085] In the present invention, the petroleum molecular property calculation model preferably includes prediction models of molecular boiling point, critical temperature, critical volume, critical pressure, density, octane number and cetane number.

[0086] After obtaining the petroleum molecular property calculation model, the present invention constructs an oil product property calculation model based on the petroleum molecular property calculation model and in combination with the mixing rule from molecular properties to oil product properties.

[0087] In the present invention, the mixing principle is taken as an example of the oil properties of gasoline and diesel. The distillation range of gasoline and diesel (oil property) is obtained by adding the molecular volume within the distillation range; the density of gasoline and diesel (oil property) is obtained by calculating the density of a single molecule according to the volume additivity.

[0088] The gasoline octane number is calculated according to formula III;

[0089]

[0090] In formula III, v i is the volume fraction of molecule i, ON i is the octane number of molecule i, β i is an adjustable parameter, representing whether molecule i is positively or negatively correlated with gasoline octane number. P For the interaction between molecules.

[0091] The diesel cetane number is calculated according to Formula IV:

[0092]

[0093] In formula IV, v i is the volume fraction of molecule i, CN iis the octane number of molecule i, β i It is an adjustable parameter, which represents whether the molecule i is positively or negatively correlated with the diesel cetane number.

[0094] The oil product property calculation model preferably includes a gasoline distillation range calculation model, a gasoline density calculation model, an octane number property calculation model, a diesel distillation range calculation model, a diesel density calculation model and a cetane number property calculation model.

[0095] After obtaining the oil property calculation model, based on the structure-guided lumping method, combined with the composition and property analysis data of catalytic cracking feedstock oil, the molecular composition matrix of catalytic cracking feedstock oil is obtained through the simulated annealing algorithm and the oil property calculation model.

[0096] In the present invention, the composition and property analysis data of the catalytic cracking feedstock oil are obtained by column chromatography separation method and nuclear magnetic resonance analysis.

[0097] In the present invention, the objective function of the simulated annealing algorithm is calculated by formula V:

[0098]

[0099] In formula V, subscript c represents the calculated value, subscript r represents the actual value, d represents the density, C, H, S, N, O represent the content of carbon, hydrogen, sulfur, nitrogen, and oxygen, P represents the content of alkane molecules, and A i 、N i represents the content of molecules with i-membered aromatic rings and i-membered cycloalkane rings, Y represents the yield, and α represents the weight factor of each indicator.

[0100] The present invention establishes reaction rules for catalytic cracking based on a structure-guided lumping method and the reaction mechanism of catalytic cracking. Multiple reaction paths and a catalytic cracking reaction network composed of multiple reaction paths are generated based on the reaction rules and the molecular composition matrix of the catalytic cracking feedstock oil. The catalytic cracking reaction network is converted into a set of reaction kinetic differential equations using reaction rate constants, which are solved using an improved Runge-Kutta method to obtain a molecular-level catalytic cracking process reaction kinetic model.

[0101] In the present invention, the reaction rate constant is calculated by formula I:

[0102]

[0103] Among them, k B is a constant, k B =1.38×10 -23 J / K;

[0104] h is Planck's constant;

[0105] R is the ideal gas constant;

[0106] T is the reaction temperature, K;

[0107] ΔS is the entropy change of the reaction, J / (mol·K);

[0108] ΔE is the energy barrier of the reaction, kcal / mol;

[0109] P is the reaction pressure, MPa.

[0110] In the present invention, the entropy change and energy barrier are obtained by using the transition state search method in the Dmol3 module of Materials Studio software.

[0111] The present invention combines a molecular-level catalytic cracking process reaction kinetics model with an oil property calculation model to construct a molecular-level catalytic cracking process reaction kinetics model combined with an oil property calculation function.

[0112] In the present invention, the constructed molecular composition matrix of the catalytic cracking feedstock oil is input into the molecular-level catalytic cracking reaction kinetics model for calculation to obtain the product yield, group composition content, and typical molecule content (see Table 5). Combined with the oil property calculation model, the main property indicators of the product (distillation range, density, octane number properties, and cetane number properties) are calculated.

[0113] In the embodiment of the present invention, the products are preferably gasoline and diesel.

[0114] The present invention also provides a computer-readable storage medium having computer instructions stored thereon. When the computer instructions are executed, the steps of the method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function are implemented.

[0115] The present invention also provides a method for predicting catalytic cracking products and their properties using the molecular-level catalytic cracking process reaction kinetics model obtained by the construction method, comprising the following steps:

[0116] The molecular composition matrix of the catalytic cracking feedstock oil is input into the molecular-level catalytic cracking process reaction kinetics model obtained by the construction method. After the molecular-level catalytic cracking process reaction kinetics model is run, the composition distribution and properties of the catalytic cracking products are output.

[0117] The present invention also provides a computer-readable storage medium having computer instructions stored thereon. When the computer instructions are executed, the steps of the method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function are implemented.

[0118] In order to further illustrate the present invention, the scheme of the present invention is described in detail below with reference to the embodiments, but they should not be understood as limiting the scope of protection of the present invention.

[0119] Example 1

[0120] Figure 1 The catalytic cracking unit process flow diagram of the embodiment is as follows: Figure 1 As shown, the feedstock for the catalytic cracking process is a mixture of hydrogenated wax oil and recycled oil. After preheating, the feedstock enters a catalytic cracking riser reactor for reaction. After the reaction, the catalytic cracking products are separated into gas fraction, diesel fraction, and slurry oil fraction in a fractionating tower. The gas fraction distilled from the top of the fractionating tower is separated into dry gas, liquefied gas, and gasoline in a separator. The diesel fraction distilled from the middle of the fractionating tower is stripped in a stripper to produce catalytic diesel. The slurry oil fraction is distilled from the bottom of the fractionating tower.

[0121] As Figure 1 The process flow chart of the catalytic cracking unit shown is an embodiment, and the steps of the method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function proposed by the present invention are described in detail.

[0122] S1: Based on the structure-guided lumping method, a data set for training a petroleum molecular property calculation model is constructed; based on the data set, a petroleum molecular property calculation model is constructed using an artificial neural network algorithm.

[0123] Specifically, the team digitally describes catalytic cracking feedstock oil using 22 structural units in a structure-guided lumping method, and constructs a dataset for training computational models of oil molecular properties.

[0124] According to the data set, the optimal number of hidden layers and the optimal number of neurons in the hidden layer are obtained by using a k-fold cross validation algorithm in an artificial neural network algorithm;

[0125] According to the optimal number of hidden layers and the optimal number of neurons in the hidden layer, the gasoline distillation range calculation model, gasoline density calculation model, octane number property calculation model, diesel distillation range calculation model, diesel density calculation model and cetane number property calculation model are constructed respectively through the neural network algorithm.

[0126] Figure 2 A diagram showing 22 structural units and their meanings in the structure-guided lumping method of an embodiment.

[0127] Figure 3 The schematic diagram of the k-fold cross validation algorithm of the embodiment is as follows: Figure 3As shown in the figure, the k-fold cross-validation algorithm is used to find the optimal number of hidden layers and the optimal number of neurons in the hidden layer. First, the data set is divided into k subsets, and each subset data is used as a validation set. The remaining k-1 subsets of data are used as training sets, and k models are obtained. Secondly, the performance of the model is evaluated by adding the average of the root mean squared error (RMSE) of the validation set in the k models, and the optimal structural parameters of the model are determined. See Table 1; the calculation formula of RMSE is:

[0128]

[0129] Where N is the number of samples in the validation set, Yexp,i is the experimental data of the property of sample i, and Ypre,i is the calculated data of the property of sample i.

[0130] Table 1 Optimal parameters and performance of the calculation model for petroleum molecular properties

[0131]

[0132]

[0133] Figure 4 The framework diagram for developing the petroleum molecular property calculation model of the embodiment is as follows: Figure 4 As shown in the figure, based on the optimal model parameters obtained by the k-fold cross-validation algorithm, the calculation models of various molecular properties are trained, and finally the prediction models of molecular boiling point, critical temperature, critical volume, critical pressure, density, octane number and cetane number are obtained.

[0134] Figure 5 1 is a comparison chart of the calculated values ​​and experimental values ​​of the petroleum molecular property calculation model of the embodiment, as shown in FIG. Figure 5 As shown in the figure, the calculated values ​​of the petroleum molecular properties calculation model are in good agreement with the experimental values ​​and the prediction error is small, indicating that the petroleum molecular properties calculation model is reliable, accurate and rigorous.

[0135] Step S2: Based on the petroleum molecular property calculation model and combined with the mixing rules from molecular properties to oil properties, an oil property calculation model (gasoline and diesel distillation range, density and octane number property calculation model) is constructed, and then the reliability of the oil property calculation model is verified using the verification data set.

[0136] The distillation range of gasoline is obtained by adding the molecular volumes within the distillation range, and the density of gasoline is calculated based on the density of individual molecules according to the volume additivity.

[0137] The formula for calculating gasoline octane number is:

[0138]

[0139] Among them, v i is the volume fraction of molecule i, ON i is the octane number of molecule i, β i is an adjustable parameter, representing whether molecule i is positively or negatively correlated with gasoline octane number. P For the interaction between molecules.

[0140] The distillation range of diesel is obtained by adding the molecular volumes within the distillation range, and the density of diesel is calculated based on the density of a single molecule according to volume additivity.

[0141] The calculation formula for diesel cetane number is:

[0142]

[0143] Among them, v i is the volume fraction of molecule i, CN i is the octane number of molecule i, β i It is an adjustable parameter, which represents whether the molecule i is positively or negatively correlated with the diesel cetane number.

[0144] S3: Based on the structure-guided lumping method, combined with the composition and property analysis data of FCC feedstock, the molecular composition matrix of FCC feedstock is obtained through the simulated annealing algorithm and the oil property calculation model.

[0145] Specifically, referring to the national standard "Determination of Four Components of Petroleum Asphalt (NB / SH / T 0509-2010)", the FCC feedstock is separated by column chromatography to obtain three sub-components: saturates, aromatics, and gums. By analyzing and testing the FCC feedstock and its three sub-components, the elemental content, average molecular weight, and distillation range distribution properties of the FCC feedstock are obtained, as well as the hydrocarbon composition information of the saturates and aromatics, and the average molecular structure information of the gums.

[0146] Based on the analysis and detection data, 61 seed molecules were selected for the construction of the catalytic cracking feedstock oil molecular composition matrix. The method of adding side chain methylene (-CH2-) to the seed molecules was adopted. Combined with the calculation data of the petroleum molecule boiling point calculation model and the catalytic cracking feedstock oil distillation range analysis data, a catalytic cracking feedstock oil molecular matrix with 3308 rows and 22 columns containing 3308 structure vectors was constructed.

[0147] Based on the analytical data of composition and properties, a simulated annealing algorithm was used for multi-objective optimization to calculate the content of each structural vector in the molecular matrix of catalytic cracking feedstock oil. The molecular composition matrix of catalytic cracking feedstock oil was obtained, and the reliability of the matrix calculation data was verified using industrial plant data.

[0148] The objective function of the simulated annealing algorithm is:

[0149]

[0150] Wherein, subscript c represents the calculated value, subscript r represents the actual value, d represents density, C, H, S, N, O represent the content of carbon, hydrogen, sulfur, nitrogen, and oxygen, P represents the content of alkane molecules, and A i , Ni represents the content of molecules with i-membered aromatic rings and i-membered cycloalkane rings, Y represents the yield, and α represents the weight factor of each indicator.

[0151] Figure 6 The block diagram of the simulated annealing algorithm of the embodiment is as follows: Figure 6 As shown, a function containing the difference between analytical data and statistical data on the distillation range distribution, elemental content, density, and hydrocarbon composition content is set as the objective function F(x). By continuously optimizing the parameters of each skewed distribution density function, the objective function value is continuously reduced. When the objective function value is less than 0.5, the wax oil molecular composition matrix is ​​considered to be suitable for describing the composition of the feedstock at the molecular level. By inputting the distillation range distribution, hydrocarbon composition content, elemental content, and density data of the catalytic cracking feedstock and running the program, a 3308-row × 23-column molecular composition matrix of the catalytic cracking feedstock is obtained.

[0152] Step S4: Based on the structure-guided lumping method and the reaction mechanism of catalytic cracking, the reaction rules of catalytic cracking are formulated, and then according to the reaction rules and the molecular composition matrix of the catalytic cracking feedstock oil, multiple reaction paths and a catalytic cracking reaction network composed of the multiple reaction paths are generated.

[0153] Specifically, the reaction rules consist of two parts: reactant selection rules and product generation rules. The reactant selection rules are used to select molecules that can undergo a specific reaction from the raw material molecule composition matrix, while the product generation rules are used to generate new molecules obtained through the reaction. This example uses a structure vector containing 22 structural units in one row to describe the molecules in the reaction system. The selection of reactant molecules and the generation of product molecules are achieved by determining and changing the values ​​of the structural units.

[0154] A total of 17 major categories and 96 minor categories of reaction rules were developed based on the compositional classification of alkanes, alkenes, cycloalkanes, aromatics, sulfides, nitrides, and oxides in the reaction system. The 17 major categories of reaction rules are shown in Table 2. After obtaining the reaction rules, the constructed molecular composition matrix of the FCC feedstock was used as input. Combined with the reaction rules, a FCC process reaction network consisting of approximately 80,000 reactions was obtained.

[0155] Table 2 17 major reaction rules

[0156] Family composition Reaction type Alkanes Carbon chain cracking and isomerization reactions Olefins Carbon chain cracking reactions, isomerization reactions, aromatization reactions and hydrogenation reactions Cycloalkanes Side chain cracking, dealkylation, ring opening, and dehydrogenation Aromatics Side chain cracking reaction, dealkylation reaction, hydrogenation saturation reaction and coking reaction sulfide Hydrodesulfurization reaction Nitride Hydrodenitrogenation reaction oxides Hydrodeoxygenation reaction

[0157] S5: Combined with the reaction rate constant, the catalytic cracking reaction network is converted into a set of reaction kinetic differential equations, which are solved using the improved Runge-Kutta method to obtain a molecular-level catalytic cracking process reaction kinetic model.

[0158] Specifically: The transition state search method in the Dmol3 module of Materials Studio software was used to obtain the entropy change and energy barrier of different reactions, and to calculate the reaction rate constants of different types of reactions.

[0159] The reaction rate constant is calculated as:

[0160]

[0161] Among them, k B is a constant, k B =1.38×10 -23 J / K; h is Planck's constant; R is the ideal gas constant; T is the reaction temperature, K; ΔS is the entropy change of the reaction, J / (mol·K); ΔE is the energy barrier of the reaction, kcal / mol; P is the reaction pressure, MPa

[0162] Figure 7 The calculation block diagram of the molecular-level catalytic cracking reaction kinetic model of the embodiment is as follows: Figure 7 As shown in the figure, the reaction network is combined with the reaction rate constant to establish a molecular-level catalytic cracking reaction kinetic model. Specifically, the total catalytic cracking reaction time is divided into n reaction intervals, and the improved Runge-Kutta method is used to solve the reaction process. In each reaction interval, the raw material molecular composition matrix is ​​input, and the reaction network is generated according to the reaction rules. Combined with the reaction rate constant, the reaction kinetic differential equations in each reaction interval are solved to obtain the reaction product molecular composition matrix of the reaction interval.

[0163] The product molecular composition matrix of the previous reaction interval is combined with the remaining reactant molecular composition matrix to serve as the reaction feedstock for the next reaction interval. The above process is repeated 20 times to obtain the final reaction product molecular composition matrix of the catalytic cracking reaction process.

[0164] The present invention also cuts the catalytic cracking products into dry gas, liquefied gas, gasoline, diesel, slurry oil and coke, and uses industrial device data to verify the reliability of the model calculation data.

[0165] The reaction kinetics model is combined with the actual production equipment of the factory, and the corresponding product partitioning rules are designed according to the product segmentation situation. This paper takes the 800,000 tons / year wax oil catalytic cracking unit of a refinery of PetroChina as the research object. The main operating parameters of the reactor in the catalytic cracking unit are shown in Table 3.

[0166] Table 3 Main operating parameters of the reactor in the catalytic cracking unit

[0167] project Numerical Inlet temperature / ℃ 545.0 Outlet temperature / ℃ 501.0 Agent-oil ratio 7.32 Residence time / s 4.01

[0168] The constructed molecular composition matrix of the catalytic cracking feedstock oil was input into the molecular-level catalytic cracking reaction kinetics model for calculation to obtain the product yield, group composition content and typical molecule content (see Table 5). Combined with the oil property calculation model, the main property indicators of the product were calculated.

[0169] The above calculated indicators were compared with industrial data to verify the accuracy of the molecular-level gas oil catalytic cracking reaction process model. A comparison of the calculated catalytic cracking product yields with industrial values ​​is shown in Table 4.

[0170] Table 4 Comparison of calculated and industrial yields of catalytic cracking products

[0171]

[0172]

[0173] As shown in Table 4, the product yields calculated based on the molecular-level catalytic cracking reaction kinetics model are in good agreement with the industrial values, with an absolute error of less than 0.9 wt%.

[0174] Figure 8 The figure is a comparison chart of the calculated and industrial values ​​of the group composition content in gasoline and diesel of the embodiment, as shown in FIG. Figure 8 As shown in the figure, the group composition content in gasoline and diesel calculated based on the molecular-level catalytic cracking reaction kinetics model is in good agreement with the industrial value.

[0175] The molecular-level catalytic cracking reaction kinetics model can be used to calculate the typical molecular content in catalytic cracking products. The comparison between the calculated values ​​and industrial values ​​of the typical molecular content in the product liquefied gas, gasoline and diesel is shown in Table 5:

[0176] Table 5 Comparison of calculated and industrial values ​​of typical molecular contents in product liquefied gas, gasoline and diesel

[0177]

[0178]

[0179] As shown in Table 5, the typical molecular contents in liquefied gas, gasoline and diesel calculated based on the molecular-level catalytic cracking reaction kinetics model are highly consistent with the industrial values, with an absolute error within 0.7 wt%.

[0180] Step S6: Combining the reaction kinetics model and the oil property calculation model, a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function is constructed to predict the product composition distribution and properties.

[0181] Specifically, the molecular composition matrix of gasoline and diesel is calculated through the reaction kinetics model of the molecular-level catalytic cracking process, and the properties of gasoline and diesel can be obtained by inputting the gasoline and diesel molecular composition matrix into the oil property calculation model.

[0182] Figure 9 The gasoline and diesel distillation range calculated values ​​and industrial values ​​of the embodiment are compared in FIG. Figure 9 As shown, the gasoline and diesel distillation range distributions calculated based on the molecular-level catalytic cracking process reaction kinetics model combined with oil product property calculation capabilities are consistent with industrial data. A comparison of the calculated values ​​and industrial values ​​for the density and octane number of the product gasoline, and the density and cetane number of the product diesel, is shown in Table 6:

[0183] Table 6 Comparison of calculated values ​​and industrial values ​​of density and octane number of product gasoline, density and cetane number of product diesel

[0184]

[0185] As shown in Table 6, the relative errors of the density and octane number of the gasoline product, as well as the density and cetane number of the diesel product, calculated based on the molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function, and the industrial data are all less than 2.2%.

[0186] By verifying and comparing indicators such as product yield, product group composition content, typical molecular content, and product gasoline and diesel properties calculated by the molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function with industrial data, it is shown that the molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function established by the present invention has high reliability and can predict product composition and properties at the molecular level.

[0187] The present invention also explores the effect of reaction temperature on the distribution of group composition in gasoline and diesel, and examines the change pattern of the properties of alkanes, alkenes, cycloalkanes and aromatics in gasoline and diesel with relative molecular weight, exploring the effect of reaction temperature on the properties of gasoline and diesel from the perspective of molecular composition.

[0188] Figure 10 The effect of reaction temperature on gasoline properties in the embodiment is shown in FIG. Figure 10As shown, as the reaction temperature increases, the depth of the cracking reaction further intensifies, further cracking the macromolecules in the gasoline, diesel, and slurry products to produce more small molecules, thereby increasing the content of small molecules in the gasoline product. Therefore, overall, at a given distillation temperature, the fraction volume of gasoline components increases with increasing reaction temperature. Because higher reaction temperatures favor cracking and aromatization reactions and hinder hydrogen transfer reactions, the content of aromatics and olefins in gasoline increases, while the content of cycloalkanes and alkanes decreases with increasing reaction temperature. The density of aromatic molecules is generally higher than that of cycloalkanes, and the density of olefin molecules is generally higher than that of alkanes. The octane numbers of aromatic and olefin molecules are generally higher than those of cycloalkanes and alkanes. Therefore, as the reaction temperature increases, the density and octane number of the gasoline product increase.

[0189] Figure 11 The effect of reaction temperature on diesel properties is shown in FIG. Figure 11 As shown, as the reaction temperature increases, the aromatic content in diesel increases, while the cycloalkanes and alkanes decrease. This is because higher reaction temperatures favor the highly endothermic cracking and aromatization reactions. The density of aromatic molecules in diesel is generally higher than that of other hydrocarbon molecules, and the density of aromatic molecules increases with relative molecular weight. The cetane number of cycloalkanes is generally higher than that of aromatics, and the cetane number of alkanes is generally higher than that of olefins. Therefore, increasing the reaction temperature increases the density of the diesel product and decreases its cetane number.

[0190] In summary, the molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function established in the present invention conforms to the reaction laws of the catalytic cracking process.

[0191] Although the above embodiment provides a detailed description of the present invention, it is only a part of the embodiments of the present invention, not all of the embodiments. Other embodiments can be obtained based on this embodiment without creativity, and these embodiments all fall within the scope of protection of the present invention.

Claims

1. A method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with oil property calculation function, characterized in that: The following steps are involved: (1) Based on the structure-guided lumping method, a data set for training a computational model of petroleum molecular properties is constructed; based on the data set, a computational model of petroleum molecular properties is constructed using an artificial neural network algorithm; (2) Based on the petroleum molecular property calculation model and in combination with the mixing rules from molecular properties to oil product properties, a calculation model for oil product properties is constructed; The oil product property calculation model includes a gasoline distillation range calculation model, a gasoline density calculation model, an octane number property calculation model, a diesel distillation range calculation model, a diesel density calculation model and a cetane number property calculation model; (3) Based on the structure-guided lumping method, combined with the composition and property analysis data of the catalytic cracking feedstock oil, a simulated annealing algorithm and the oil property calculation model are used to obtain the molecular composition matrix of the catalytic cracking feedstock oil; (4) formulating catalytic cracking reaction rules based on the structure-guided lumping method and the reaction mechanism of catalytic cracking, and then generating multiple reaction paths and a catalytic cracking reaction network composed of the multiple reaction paths according to the reaction rules and the molecular composition matrix of the catalytic cracking feedstock oil; (5) converting the catalytic cracking reaction network into a set of reaction kinetic differential equations based on the reaction rate constant, solving the set of reaction kinetic differential equations using the improved Runge-Kutta method, and obtaining a molecular-level catalytic cracking process reaction kinetic model; The reaction rate constant is calculated by formula I: Among them, k B is a constant, k B =1.38×10 -23 J / K; h is Planck's constant; R is the ideal gas constant; T is the reaction temperature, K; ΔS is the entropy change of the reaction, J / (mol·K); ΔE is the energy barrier of the reaction, kcal / mol; P is the reaction pressure, MPa; (6) Combining the molecular-level catalytic cracking process reaction kinetics model and the oil property calculation model, a molecular-level catalytic cracking process reaction kinetics model combined with the oil property calculation function is constructed.

2. The construction method according to claim 1, characterized in that The calculation model of petroleum molecular properties is constructed as follows: Based on the structure-guided lumping method, a dataset was constructed for training computational models of petroleum molecular properties. According to the data set, the optimal number of hidden layers and the optimal number of neurons in the hidden layer are obtained through the k-fold cross validation algorithm in the artificial neural network algorithm; According to the optimal number of hidden layers and the optimal number of neurons in the hidden layers, a calculation model of petroleum molecular properties is constructed through the neural network algorithm.

3. The construction method according to claim 1, characterized in that The molecular composition matrix of the catalytic cracking feedstock oil is obtained as follows: Determining seed molecules based on composition information and property analysis data of catalytic cracking feedstock oil, and then performing structural vector expansion on the seed molecules by adding side chains to construct a catalytic cracking feedstock oil molecular matrix; According to the composition information and property analysis data of crude oil and the structure vector of each molecule in the catalytic cracking feedstock oil molecular matrix, a simulated annealing algorithm is used to perform multi-objective optimization to calculate the content of each structure vector in the catalytic cracking feedstock oil molecular matrix to obtain the catalytic cracking feedstock oil molecular composition matrix.

4. The construction method according to claim 1, characterized in that The reaction rules include reactant selection rules and product generation rules.

5. The construction method according to claim 1, characterized in that The method for constructing a molecular-level catalytic cracking process reaction kinetics model combined with oil property calculation function is: The product molecular composition matrix is ​​calculated through the molecular-level catalytic cracking process reaction kinetics model. The product molecular composition matrix is ​​input into the oil property calculation model to obtain the product properties.

6. The construction method according to claim 2, characterized in that The k-fold cross validation algorithm includes: Divide the dataset into k subsets, use the data from each subset as a validation set, and use the remaining k-1 subsets as training sets, thereby obtaining k models. The performance of the k models is evaluated by taking the sum of the root mean square errors of the validation sets in the k models, and then the number of hidden layers and neurons of the model is determined; the formula for calculating the root mean square error is Formula II: Where N is the number of samples in the validation set, Yexp,i is the experimental data of the property of sample i, and Ypre,i is the calculated data of the property of sample i.

7. The construction method according to claim 1, characterized in that The petroleum molecule property calculation model includes prediction models of petroleum molecule boiling point, critical temperature, critical volume, critical pressure, density, octane number and cetane number respectively.

8. The construction method according to claim 3, characterized in that: The method for obtaining the composition and property analysis data of the catalytic cracking feedstock oil is: The composition and property analysis data of catalytic cracking feed oil were obtained through column chromatography separation and nuclear magnetic resonance analysis.

9. A method for predicting catalytic cracking products and their properties using a molecular-level catalytic cracking process reaction kinetics model combined with oil property calculation function obtained by the construction method according to any one of claims 1 to 8, comprising the following steps: The molecular composition matrix of the catalytic cracking feedstock oil is input into the molecular-level catalytic cracking process reaction kinetics model obtained by the construction method. After the molecular-level catalytic cracking process reaction kinetics model is run, the composition distribution and properties of the catalytic cracking products are output.

10. A computer-readable storage medium having computer instructions stored thereon, wherein when the computer instructions are executed, the steps of any one of 1 to 6 of the method for constructing a molecular-level catalytic cracking process reaction kinetics model are implemented.

Images