Hydrogen consumption prediction method and device based on continuous lumped dual-drive model

Abstract

The embodiment of the invention provides a hydrogen consumption prediction method and device based on a continuous lumped dual-drive model, and the method comprises the steps: responding to a condition that a prediction result of a data model pre-constructed based on a recurrent neural network meets a preset precision requirement under a current operation condition, if yes, predicting the hydrogen consumption of the residual oil hydrogenation device by using the data model to obtain a first hydrogen consumption prediction result; and in response to the condition that a prediction result of a data model pre-constructed based on the recurrent neural network does not meet a preset accuracy requirement under the current operation condition, predicting the hydrogen consumption of the residual oil hydrogenation device by using a mechanism model pre-constructed based on multiple preset chemical reactions to obtain a second hydrogen consumption prediction result. According to the hydrogen consumption prediction method and device based on the continuous lumped dual-drive model provided by the embodiment of the invention, the accuracy and robustness of hydrogen consumption prediction of the residual oil hydrogenation device are improved, and the data utilization rate is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of petroleum industry, and in particular to a method and device for predicting hydrogen consumption based on a continuous lumped dual-drive model. Background Art

[0002] The residue oil hydrogenation process in the field of petroleum refining and chemical industry can remove impurities such as sulfur, nitrogen, and metals, and convert the heavy components in the residue oil into lighter components, providing high-quality raw materials for catalytic cracking units. Therefore, the residue oil hydrogenation process is the main means to lighten heavy petroleum and make efficient use of it. Hydrogen, as an important raw material in the residue oil hydrogenation process, comes from new hydrogen in the hydrogen production unit and recycled hydrogen reused after purification. Since the production of hydrogen by the hydrogen production unit and the scheduling of hydrogen by the circulating hydrogen system have a time delay effect, the accurate prediction of the hydrogen consumption of the residue oil hydrogenation unit in advance is the key to improving the production efficiency and benefits of the current residue oil hydrogenation unit.

[0003] With the continuous development of artificial intelligence algorithms and big data in recent years, the petrochemical industry, as one of the pillar industries of energy, is also trying to integrate with new information technology and continuously develop in the direction of intelligence. The hydrogen consumption prediction model of the residue oil hydrotreating unit established by the neural network algorithm in machine learning can make a stable and accurate prediction of the new hydrogen flow required by the residue oil hydrotreating unit at the next moment, which has important guiding significance for the production scheduling of hydrogen production units and the optimization scheduling of circulating hydrogen systems.

[0004] The current prediction models for hydrogen consumption in residue hydrotreating units mainly include: multi-component lumped mechanism model and data-driven model, and these traditional models face many difficulties in residue hydrotreating process. Since the mechanism of residue hydrotreating process is very complex, involving more than a dozen chemical reactions, and the multi-component lumped mechanism model can only be established for a single reaction, the prediction accuracy of the multi-component lumped mechanism model is not high, and it cannot meet the actual industrial operation and prediction needs. Data-driven models mainly include multivariate statistical analysis methods and machine learning algorithms. For multivariate statistical analysis methods such as multivariate variance analysis, principal component analysis, factor analysis, partial least squares method, etc., it is impossible to perform nonlinear fitting on complex chemical processes dominated by nonlinearity; and machine learning methods such as support vector machines and back propagation neural networks cannot handle the potential features in time series data well, and there will be problems such as insufficient data feature mining and low utilization. The above limitations affect the operation accuracy and robustness of the model.

[0005] It can be seen that how to improve the accuracy and robustness of the prediction of hydrogen consumption in the residue oil hydrotreating unit has become a difficult problem that needs to be solved urgently. Summary of the invention

[0006] In view of the defects in the prior art, an embodiment of the present invention provides a method and device for predicting hydrogen consumption based on a continuous lumped dual-drive model.

[0007] An embodiment of the present invention provides a method for predicting hydrogen consumption based on a continuous lumped dual-drive model, comprising: in response to a current operating condition, if a prediction result of a data model pre-constructed based on a recurrent neural network meets a preset accuracy requirement, the data model is used to predict the hydrogen consumption of a residue oil hydrogenation device to obtain a first hydrogen consumption prediction result; in response to a current operating condition, if a prediction result of the data model pre-constructed based on a recurrent neural network does not meet the preset accuracy requirement, the hydrogen consumption of the residue oil hydrogenation device is predicted using a mechanism model pre-constructed based on preset multiple chemical reactions to obtain a second hydrogen consumption prediction result.

[0008] According to an embodiment of the present invention, a hydrogen consumption prediction method based on a continuous lumped dual-drive model is provided, wherein the hydrogen consumption of a residue oil hydrogenation unit is predicted using the data model to obtain a first hydrogen consumption prediction result, including: inputting residue oil hydrogenation operating condition data into the data model to obtain a first product yield prediction result, and obtaining the first hydrogen consumption prediction result according to the first product yield prediction result.

[0009] According to a method for predicting hydrogen consumption based on a continuous lumped dual-drive model provided by an embodiment of the present invention, the residual oil hydrogenation operating condition data includes reactor bed temperature, liquid phase space velocity, residual oil feed composition, residual oil feed flow rate, residual oil feed density and distillation temperature.

[0010] According to an embodiment of the present invention, a method for predicting hydrogen consumption based on a continuous lumped dual-drive model is provided, the method also comprising: based on four preset chemical reactions, representing and generating multiple reaction paths and a reaction network composed of the multiple reaction paths according to the structure of the raw material molecules and the actual boiling point distillation data; constructing a partial differential equation for each divided narrow fraction segment according to the reaction network; wherein the partial differential equation comprises a residue oil hydrogenation reactor flow model and a reaction network model, the parameters of the residue oil hydrogenation reactor flow model comprise the reaction rate, and the parameters of the reaction network model comprise the component concentration of the narrow fraction segment; solving the partial differential equation according to the preset reaction rate and the initial component concentration of the narrow fraction segment to obtain the predicted yield of the product of each narrow fraction segment; when the difference between the actual yield of the product and the predicted yield is greater than a preset threshold, correcting the preset reaction rate, determining the expression of the partial differential equation based on the corrected preset reaction rate, and obtaining a mechanism model based on continuous lumping.

[0011] According to a method for predicting hydrogen consumption based on a continuous lumped dual-drive model provided in an embodiment of the present invention, the four chemical reactions include hydrocracking reaction, hydrosaturation reaction, hydrodesulfurization reaction and hydrodenitrogenation reaction.

[0012] According to an embodiment of the present invention, a method for predicting hydrogen consumption based on a continuous lumped dual-drive model is provided, in which the hydrogen consumption of the residue oil hydrogenation unit is predicted by using a mechanism model pre-constructed based on a plurality of preset chemical reactions to obtain a second hydrogen consumption prediction result, including: bringing the initial component concentrations of each narrow fraction segment into the partial differential equation, using the Runge-Kutta numerical calculation method to simultaneously solve the partial differential equation, obtaining a second product yield prediction result of each narrow fraction segment that changes with time, and obtaining the second hydrogen consumption prediction result according to the second product yield prediction result.

[0013] An embodiment of the present invention also provides a hydrogen consumption prediction device based on a continuous lumped dual-drive model, comprising: a data model prediction module, used for: in response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, then using the data model to predict the hydrogen consumption of the residue oil hydrogenation device to obtain a first hydrogen consumption prediction result; a mechanism model prediction module, used for: in response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, then using a mechanism model pre-constructed based on preset multiple chemical reactions to predict the hydrogen consumption of the residue oil hydrogenation device to obtain a second hydrogen consumption prediction result.

[0014] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the hydrogen consumption prediction method based on the continuous lumped dual-drive model as described in any one of the above are implemented.

[0015] An embodiment of the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the hydrogen consumption prediction method based on the continuous lumped dual-drive model as described in any one of the above are implemented.

[0016] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any of the above-mentioned methods for predicting hydrogen consumption based on the continuous lumped dual-drive model.

[0017] The hydrogen consumption prediction method and device based on the continuous lumped dual-drive model provided by the embodiment of the present invention predicts the hydrogen consumption of the residue oil hydrogenation unit by using the data model in response to the current operating condition, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, and obtains a first hydrogen consumption prediction result. In response to the current operating condition, if the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, the hydrogen consumption of the residue oil hydrogenation unit is predicted by using a mechanism model pre-constructed based on preset multiple chemical reactions, and a second hydrogen consumption prediction result is obtained, thereby improving the accuracy and robustness of the hydrogen consumption prediction of the residue oil hydrogenation unit and improving data utilization. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0019] Figure 1 This is one of the flow charts of the method for predicting hydrogen consumption based on the continuous lumped dual-drive model provided in an embodiment of the present invention;

[0020] Figure 2 It is a schematic diagram of the data processing flow of the recurrent neural network;

[0021] Figure 3 is a schematic diagram of a parallel hybrid model in a hydrogen consumption prediction method based on a continuous lumped dual-drive model provided in an embodiment of the present invention;

[0022] Figure 4 It is a schematic diagram of fitting results in volume fraction calculation in the hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in an embodiment of the present invention;

[0023] Figure 5 It is a schematic diagram of a residual oil hydrogenation reaction network in a hydrogen consumption prediction method based on a continuous lumped dual-drive model provided in an embodiment of the present invention;

[0024] Figure 6 This is the second flow chart of the method for predicting hydrogen consumption based on the continuous lumped dual-drive model provided in an embodiment of the present invention;

[0025] Figure 7 It is a structural schematic diagram of a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided in an embodiment of the present invention;

[0026] Figure 8It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] Figure 1 FIG. 1 is one of the flow charts of the method for predicting hydrogen consumption based on the continuous lumped dual-drive model provided in the embodiment of the present invention. Figure 1 As shown, the method includes:

[0029] Step S1: In response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, the data model is used to predict the hydrogen consumption of the residue oil hydrotreating unit to obtain a first hydrogen consumption prediction result.

[0030] The recurrent neural network (RNN) model has a strong ability to process sequence data. In the network model, the entire sequence is processed through weight sharing, and the data on a long time series is generalized. Therefore, the recurrent neural network model is more suitable for the dynamic model of the residue hydrotreating unit.

[0031] Figure 2 This is a schematic diagram of the data processing flow of the recurrent neural network. Figure 2 As shown in the figure, a recurrent neural network architecture contains an input layer, an output layer and an intermediate hidden layer. It also has a loop itself. This loop allows information to be passed from one step in the network to the next step. The network state at the previous moment will affect the network state at the next moment, and it is required to have an input at every moment, but not necessarily an output at every moment.

[0032] in,

[0033] Input unit: {x0,…x t-1 ,x t ,x t+1 ,…},x t Represents the input at time t.

[0034] Hidden units: {s0,…s t-1 ,s t ,s t+1 ,…}, output s t The calculation formula is s t =f(Uxt +Ws t-1 ), where f is the nonlinear activation function, and the ReLu function is selected here.

[0035] Output unit: {o0,…o t-1 ,o t ,o t+1 ,…}, output o t =softmax(Vs t ), the softmax function is a normalized exponential function so that each element ranges between 0 and 1 and the sum of all elements is 1.

[0036] U, V, W represent network weights.

[0037] The recurrent neural network is pre-trained based on the residual oil hydrotreating operation condition data to obtain a data model. In actual use, the data model pre-built based on the recurrent neural network is used to predict the hydrogen consumption of the residual oil hydrotreating unit to obtain the first hydrogen consumption prediction result. The data model can model the effects and processes ignored in the mechanism model through its nonlinear fitting, effectively improving the prediction performance of the hybrid model.

[0038] Since the accuracy of data model prediction is related to operating conditions (including reactor bed temperature, reactor pressure and pressure drop, hydrogen-to-oil ratio, and liquid phase space velocity), if the operating conditions during prediction are different from those during data model training, the accuracy of the model will be affected.

[0039] If the prediction result of the data model pre-built based on the recurrent neural network meets the preset accuracy requirement under the current operating conditions, the data model is used to predict the hydrogen consumption of the residue oil hydrotreating unit to obtain a first hydrogen consumption prediction result. The first hydrogen consumption prediction result is used as the final prediction result and output.

[0040] Step S2: In response to the fact that the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement under the current operating conditions, the hydrogen consumption of the residue oil hydrogenation unit is predicted using a mechanism model pre-constructed based on a preset plurality of chemical reactions to obtain a second hydrogen consumption prediction result.

[0041] If, under the current operating conditions, the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, the hydrogen consumption of the residue oil hydroprocessing unit is predicted using the mechanism model pre-constructed based on the preset multiple chemical reactions, a second hydrogen consumption prediction result is obtained, and the second hydrogen consumption prediction result is output. The mechanism model of the embodiment of the present invention is pre-constructed based on the preset multiple chemical reactions of the residue oil hydroprocessing unit, thereby improving the prediction accuracy.

[0042] Figure 3 Schematic diagram of a parallel hybrid model in a method for predicting hydrogen consumption based on a continuous lumped dual-drive model provided in an embodiment of the present invention. Figure 3 As shown, considering the model prediction accuracy and computational complexity, the embodiment of the present invention adopts a parallel hybrid model fusion method. If the prediction result of the data model meets the preset accuracy requirement, the data model is called for simulation calculation, and the first hydrogen consumption prediction result is output. When the accuracy of the prediction result of the data model does not meet the requirement, the mechanism model is called for simulation calculation, and the second hydrogen consumption prediction result is output. Figure 3 In the data model, the functional relationship is expressed as f(X k ), the functional relationship of the mechanism model is expressed as g(X k ), the fusion function is expressed as h(f(X k ),g(X k )). Fusion function h(f(X k ),g(X k )) means that if the prediction result of the data model meets the preset accuracy requirement, the data model is called for simulation prediction; if the prediction result of the data model does not meet the preset accuracy requirement, the mechanism model is called for simulation prediction.

[0043] When judging whether the prediction accuracy of the data model meets the accuracy requirements, since the hydrogen consumption (hydrogen consumption) can be calculated by the product yield, it is possible to judge whether the prediction result of hydrogen consumption meets the preset accuracy requirements based on whether the product yield meets the preset accuracy requirements; or the hydrogen consumption is calculated based on the product yield, and the prediction result of hydrogen consumption meets the preset accuracy requirements to make a direct judgment; or it is also possible to consider that the hydrogen consumption prediction result of the data model meets the preset accuracy requirements when both the product yield and the hydrogen consumption meet the preset accuracy requirements. Among them, the model prediction error calculation formula is: The preset accuracy requirement can be set to a model prediction error that is less than or equal to a preset value, such as 20%. If the data model prediction error is greater than 20%, the preset accuracy requirement is not met and the mechanism model is called as the simulation prediction model.

[0044] The hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in an embodiment of the present invention predicts the hydrogen consumption of the residue oil hydrogenation unit by using the data model in response to the current operating condition, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, and obtains a first hydrogen consumption prediction result. In response to the current operating condition, if the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, the hydrogen consumption of the residue oil hydrogenation unit is predicted by using a mechanism model pre-constructed based on preset multiple chemical reactions, and a second hydrogen consumption prediction result is obtained, thereby improving the accuracy and robustness of the hydrogen consumption prediction of the residue oil hydrogenation unit and improving data utilization.

[0045] According to an embodiment of the present invention, a hydrogen consumption prediction method based on a continuous lumped dual-drive model is provided, wherein the hydrogen consumption of a residue oil hydrogenation unit is predicted using the data model to obtain a first hydrogen consumption prediction result, including: inputting residue oil hydrogenation operating condition data into the data model to obtain a first product yield prediction result, and obtaining the first hydrogen consumption prediction result according to the first product yield prediction result.

[0046] A data-driven model (data model) based on recurrent neural network (RNN) is constructed to collect and acquire the distributed control system (DCS) data, manufacturing execution system (MES) data, and LIMS (laboratory Information Management System) analysis data of the residue oil hydrotreating unit according to the preset sampling interval (e.g. every 10 minutes).

[0047] The collected residual oil hydrotreating operation data include:

[0048] Residue oil hydrogenation feed: feed flow, feed composition, density, viscosity, sulfur content, nitrogen content, distillation temperature, and four-component composition properties (paraffin group composition, cycloparaffin group composition, aromatic hydrocarbon group composition and asphaltene composition).

[0049] Operating parameters of the residue oil hydrotreating unit: temperature of each bed in the reactor, reactor pressure and pressure drop, hydrogen-to-oil ratio, and liquid phase space velocity.

[0050] The correlation analysis of various parameters of the collected residual oil hydroprocessing operation data is carried out and key variables (key variables include bed temperature, hydrogen consumption, liquid phase space velocity, residual oil feed composition, distillation temperature, density and flow rate) are selected as the input of the data model for predicting hydrogen consumption and the yield of each component. The product yield is used as the output. The predicted value of hydrogen consumption can be calculated through the predicted value of product yield, thereby establishing a real-time prediction model with strong nonlinear time series correlation. If physical property parameters (such as sulfur content and nitrogen content) are also to be predicted, the corresponding physical property parameters must also be used as output.

[0051] The key variable data in the original working condition data of residue oil hydroprocessing are processed by data outliers, smoothing and noise reduction, steady-state analysis and normalization, and then used as training data set to train recurrent neural network (RNN) to obtain the data model. Among them, the original data of residue oil hydroprocessing are collected and time series processed; the labeled sample data set D is obtained, and the data of the sample data set D is normalized using the min-max normalization method to eliminate the influence of different dimensions.

[0052] The normalized training data set D is used as the input data of the modeling process, and a recurrent neural network model is constructed using the processed training data set D. The recurrent neural network (RNN) is trained using the Levenberg-Marquardt regression algorithm to obtain the residue oil hydrogenation model (data model) of the deep recurrent neural network RNN.

[0053] An example of training a data model is as follows:

[0054] (I) For a sample set with 1160 samples, a sample set X and a label set Y are defined, and the ratio of training and testing samples is m. In this embodiment, m=0.7 is set. According to the real-time prediction demand of the hydrogen flow rate in the next 10 minutes of the residual oil hydrogenation process, the input step size is set to t1=3 and the output step size is set to t2=1 in this embodiment.

[0055] (II) Since too large a model depth will result in a large number of parameters, which will bring huge calculation and convergence difficulties, the model depth n is selected from a positive integer greater than or equal to 2. In this embodiment, the RNN model depth n is set to 2, the number of hidden nodes of each layer of RNN is h1=15, h2=40, and the activation function is ReLu. In order to prevent overfitting, in this embodiment, the number of iterative training times of the model training process is set to Tmax=800, and the number of batch samples is set to K=200.

[0056] (III) The Levenberg-Marquardt algorithm is used to train the recurrent neural network RNN, and the MSE is less than 10 -5 The trained neural network RNN ​​data-driven model (data model).

[0057] When the hydrogen consumption of the residue oil hydrogenation unit is predicted using the trained data model to obtain a first hydrogen consumption prediction result, the residue oil hydrogenation operating condition data is input into the data model to obtain a first product yield prediction result, and the first hydrogen consumption prediction result is obtained based on the first product yield prediction result.

[0058] The hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in an embodiment of the present invention obtains a first product yield prediction result by inputting the residual oil hydrogenation operating condition data into the data model, and obtains a first hydrogen consumption prediction result based on the first product yield prediction result, thereby realizing hydrogen consumption prediction based on the residual oil hydrogenation operating condition data and the data model, and further improving the accuracy and robustness of hydrogen consumption prediction.

[0059] According to a method for predicting hydrogen consumption based on a continuous lumped dual-drive model provided by an embodiment of the present invention, the residual oil hydrogenation operating condition data includes reactor bed temperature, liquid phase space velocity, residual oil feed composition, residual oil feed flow rate, residual oil feed density and distillation temperature.

[0060] When training the data model and using the data model to predict hydrogen consumption, the residue oil hydrotreating operating condition data input into the model are the key parameter data obtained through correlation analysis, including reactor bed temperature, liquid phase space velocity, residue oil feed composition, residue oil feed flow rate, residue oil feed density and distillation temperature.

[0061] The hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in the embodiment of the present invention further improves the accuracy and robustness of the hydrogen consumption prediction by reasonably determining the residual oil hydrogenation operation condition data of the input data model.

[0062] According to an embodiment of the present invention, a method for predicting hydrogen consumption based on a continuous lumped dual-drive model is provided, the method also comprising: based on four preset chemical reactions, representing and generating multiple reaction paths and a reaction network composed of the multiple reaction paths according to the structure of the raw material molecules and the actual boiling point distillation data; constructing a partial differential equation for each divided narrow fraction segment according to the reaction network; wherein the partial differential equation comprises a residue oil hydrogenation reactor flow model and a reaction network model, the parameters of the residue oil hydrogenation reactor flow model comprise the reaction rate, and the parameters of the reaction network model comprise the component concentration of the narrow fraction segment; solving the partial differential equation according to the preset reaction rate and the initial component concentration of the narrow fraction segment to obtain the predicted yield of the product of each narrow fraction segment; when the difference between the actual yield of the product and the predicted yield is greater than a preset threshold, correcting the preset reaction rate, determining the expression of the partial differential equation based on the corrected preset reaction rate, and obtaining a mechanism model based on continuous lumping.

[0063] Since the residue hydrogenation reactions mainly include the following: hydrodesulfurization reaction, hydrodenitrogenation reaction, aromatic saturation reaction and hydrocracking reaction; each of the above reactions is a complex reaction system composed of a series of parallel and sequential reactions. Based on the preset reaction rules, according to the actual boiling point distillation data of the residue feed and the structure of the raw material molecules (four-component composition properties), multiple reaction paths and reaction networks are represented and generated. Taking advantage of the fact that petroleum composition has a continuous distribution and obeys the skew-normal statistical distribution law, the petroleum fraction composition is divided into a narrow fraction segment at a preset temperature interval (such as every 10 degrees Celsius, and this temperature interval is taken as an example below), and a partial differential equation is constructed for each narrow fraction segment according to the hydrogenation reaction network. The partial differential equation is solved according to the content of each narrow fraction and the preset reaction rate constant to obtain the predicted yield of the product of each narrow fraction segment. When the difference between the actual yield and the predicted yield of the product is greater than a preset threshold, the preset reaction rate is corrected, and the expression of the partial differential equation is determined based on the corrected preset reaction rate to obtain a mechanism model based on continuous lumping. According to the mechanism model, the yield of each product and the consumption of hydrogen can be obtained.

[0064] A specific step of constructing the mechanism model is as follows:

[0065] (I) The residual oil feed fraction is cut into narrow fractions of 10 degrees Celsius, the true boiling point TBP (TrueBoiling Point) data is used as the ordinate, and the distillate volume fraction is used as the abscissa, and a cubic polynomial is used for fitting to obtain a cubic fitting function. The volume fraction is obtained by inputting the TBP temperature of a narrow fraction component of 10 degrees Celsius, and the narrow fraction volume fraction data is obtained by subtracting the volume fractions cut at every 10 degrees Celsius, and then the volume fraction is converted into a mass fraction. The mass fraction is related to the component concentration, and the component concentration can be calculated based on the mass fraction.

[0066] Figure 4 FIG. 1 is a schematic diagram of the fitting results in the volume fraction calculation in the hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in an embodiment of the present invention. Figure 4 The VR, VGO, and LCO narrow fraction data fitting results are shown, where VR (Vacuum Residue) refers to the oil remaining after conventional distillation under high temperature and pressure in petroleum processing. VGO (Vacuum Gas Oil) refers to the heavier oil component separated from VR during vacuum distillation. LCO (Light Cycle Oil) refers to the light oil obtained in the catalytic cracking unit after the light fraction contacts the regenerated catalyst.

[0067] (II) Based on the continuous distribution of petroleum components, a residue hydrogenation reaction network is constructed. Figure 5As shown, it includes hydrocracking reaction, hydrosaturation reaction, hydrodesulfurization reaction and hydrodenitrogenation reaction. The column vector represents the narrow fraction cut at 10 degrees Celsius, and the row vector represents the different compositions classified by structure, where P represents the paraffin family composition, N represents the cycloalkane family composition, and A represents the aromatic family composition. Since nitrogen and sulfur are mainly concentrated on the aromatic ring, AN represents the nitrogen on the aromatic ring and AS represents the sulfur on the aromatic ring.

[0068] Figure 5 It is a schematic diagram of a residue oil hydrogenation reaction network in a hydrogen consumption prediction method based on a continuous lumped dual-drive model provided in an embodiment of the present invention.

[0069] (III) According to the residue hydrogenation reaction network obtained in step (II), a partial differential equation is constructed for each narrow fraction segment according to the hydrogenation reaction network. The constructed partial differential equation includes a residue hydrogenation reactor flow model and a reaction network model. The residue hydrogenation reactor flow model is used as a reactor flow simulation equation, and an axial center diffusion model can be used.

[0070] The axial center diffusion model formula is:

[0071] Where C represents the component concentration of each narrow fraction, t represents time, and D e represents the diffusion coefficient, z represents the axial length of the reactor, u represents the fluid flow rate, and R represents the source term.

[0072] For different reactions, the source terms of the axial center diffusion model are: hydrocracking reaction source term R1, hydrogenation saturation reaction source terms R2 and R3, hydrodesulfurization reaction source term R4, and hydrodenitrogenation reaction source term R5. Among them:

[0073] The formula for source term R1 is: R1 = -k i C i +∑ j＞i P ij k j C j ;

[0074] Among them, k i represents the cracking reaction coefficient of each narrow fraction, C i represents the concentration of each narrow fraction i, P ij represents the narrow fraction cracking distribution coefficient, k j represents the narrow-cut cracking reaction coefficient of component j being heavier than component i, C j Represents the concentration of narrow fraction j where component j is heavier than component i.

[0075] The formula for source terms R2 and R3 is: R2 = -k N-P C i , R3=-k A-N Ci ;

[0076] k N-P represents the cracking reaction rate of cycloalkanes into paraffins by hydrocracking, k A-N It indicates the cracking reaction rate of aromatic hydrocarbons into cycloalkanes by hydrocracking.

[0077] The source term R4 formula is:

[0078] k AS-A It represents the reaction rate of converting aromatic sulfur into aromatic hydrocarbons by hydrodesulfurization. Indicates the concentration of components whose actual boiling point increases by 50°C in narrow fractions, TBP min Indicates initial distillation point, TBP max represents the highest distillation point, and θ represents the actual boiling point narrow distillation segment.

[0079] The source term R5 formula is:

[0080] k AN-N represents the reaction rate of the conversion of aromatic nitrogen into cycloalkanes by hydrodenitrogenation, It indicates the concentration of the component whose actual boiling point of the narrow fraction increases by 100°C.

[0081] The reaction network model is used to simulate the conversion process between PNA in the hydrogenation process and the conversion process between different fractions in the cracking process, including the following differential equations:

[0082]

[0083] Among them, θ represents the actual boiling point narrow fraction, t represents time, c P (θ,t) represents the concentration of each narrow fraction of the paraffin component, R P,out (θ, t) represents the disappearance of each narrow cut cracking part of the paraffin component, R P,in (θ, t) represents the fraction generated by cracking each narrow fraction of the paraffin component, R N-P (θ, t) represents the portion of cycloalkanes converted to paraffins by hydrogenation saturation.

[0084]

[0085] Among them, c N (θ,t) represents the concentration of each narrow fraction of the cycloparaffin component, R N,out (θ, t) represents the disappearance of each narrow cut cracking part of the cycloalkane component, R N,in (θ, t) represents the fraction generated by cracking each narrow fraction of the cycloparaffin component, R N-P (θ,t) represents the part of cycloalkanes converted to paraffins by hydrogenation and saturation, RA-N (θ,t) represents the portion of aromatic hydrocarbons converted to cycloalkanes by hydrogenation and saturation, θ +100 Indicates that the actual boiling point increases by 100°C, R AN-N (θ +1 ,t) represents the portion converted into cycloalkanes by hydrodenitrogenation of aromatic nitrogen.

[0086]

[0087] Among them, c A (θ,t) represents the concentration of each narrow fraction of aromatic hydrocarbon components, R A,out (θ, t) represents the disappearance of each narrow cut cracking part of aromatic hydrocarbon components, R A,in (θ, t) represents the fraction generated by the cracking of each narrow fraction of the aromatic hydrocarbon component, R A-N (θ,t) represents the portion of aromatic hydrocarbons converted to cycloalkanes by hydrogenation and saturation, R AS- (θ +50 ,t) represents the portion converted into aromatics by hydrodesulfurization of aromatic sulfur.

[0088]

[0089] Among them, c AN (θ,t) represents the concentration of each narrow fraction of aromatic nitrogen components, R AN-N (θ, t) represents the portion of cycloalkanes converted by hydrodenitrogenation of aromatic nitrogen.

[0090]

[0091] Among them, c AS (θ,t) represents the concentration of each narrow fraction of aromatic sulfur components, R AS-A (θ, t) represents the portion of aromatic sulfur converted to aromatics by hydrodesulfurization.

[0092] The partial differential equations are solved according to the preset reaction rate and the initial component concentration of the narrow fraction segment to obtain the predicted yield of the product of each narrow fraction segment. When the difference between the actual yield of the product and the predicted yield is greater than a preset threshold, the preset reaction rate is corrected, and the expression of each partial differential equation is determined based on the corrected preset reaction rate to obtain a mechanism model based on continuous lumping composed of the above partial differential equations.

[0093] The hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in the embodiment of the present invention represents and generates multiple reaction paths and a reaction network composed of multiple reaction paths based on the structure of the raw material molecules and the actual boiling point distillation data based on four preset chemical reactions, constructs a partial differential equation for each divided narrow fraction segment according to the reaction network, solves the partial differential equation according to the preset reaction rate and the initial component concentration of the narrow fraction segment, and obtains the predicted yield of the product of each narrow fraction segment. When the difference between the actual yield of the product and the predicted yield is greater than a preset threshold, the preset reaction rate is corrected, and the expression of the partial differential equation is determined based on the corrected preset reaction rate to obtain a mechanism model based on continuous lumping, thereby achieving accurate construction of the mechanism model.

[0094] According to a method for predicting hydrogen consumption based on a continuous lumped dual-drive model provided in an embodiment of the present invention, the four chemical reactions include hydrocracking reaction, hydrosaturation reaction, hydrodesulfurization reaction and hydrodenitrogenation reaction.

[0095] The embodiment of the present invention improves the accuracy and robustness of the mechanism model in predicting hydrogen consumption by constructing a mechanism model based on the main reactions in the residue hydrogenation reaction, namely, the hydrocracking reaction, the hydrogenation saturation reaction, the hydrodesulfurization reaction and the hydrodenitrogenation reaction.

[0096] According to an embodiment of the present invention, a method for predicting hydrogen consumption based on a continuous lumped dual-drive model is provided, in which the hydrogen consumption of the residue oil hydrogenation unit is predicted by using a mechanism model pre-constructed based on a plurality of preset chemical reactions to obtain a second hydrogen consumption prediction result, including: bringing the initial component concentrations of each narrow fraction segment into the partial differential equation, using the Runge-Kutta numerical calculation method to simultaneously solve the partial differential equation, obtaining a second product yield prediction result of each narrow fraction segment that changes with time, and obtaining the second hydrogen consumption prediction result according to the second product yield prediction result.

[0097] After obtaining the mechanism model, the hydrogen consumption of the residue oil hydrogenation unit is predicted using the mechanism model pre-constructed based on preset multiple chemical reactions. When obtaining the second hydrogen consumption prediction result, the initial component concentration of each narrow fraction segment is substituted into the partial differential equation, and the partial differential equation is solved simultaneously using the Runge-Kutta numerical calculation method to obtain the final component concentration of each narrow fraction segment that changes with time. The second product yield prediction result of each narrow fraction segment is obtained based on the final component concentration of each narrow fraction segment, and the second hydrogen consumption prediction result is obtained based on the second product yield prediction result. At the same time, the changes in the physical properties of the sulfur content and nitrogen content in the narrow fraction over time are also obtained.

[0098] The hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in an embodiment of the present invention, by bringing the initial component concentrations of each narrow fraction segment into the partial differential equation, and using the Runge-Kutta numerical calculation method to simultaneously solve the partial differential equation, obtains the second product yield prediction results of each narrow fraction segment that changes with time, and obtains the second hydrogen consumption prediction results based on the second product yield prediction results, thereby further improving the accuracy and robustness of the hydrogen consumption prediction of the mechanism model.

[0099] Figure 6 This is the second flow chart of the method for predicting hydrogen consumption based on the continuous lumped dual-drive model provided by the embodiment of the present invention. Figure 6 As shown, the method includes:

[0100] (1) Collect and process residual oil hydrotreating data;

[0101] Obtain feedstock composition information, including feed composition, density, sulfur content, nitrogen content, actual boiling point distillation range temperature, four-component composition, as well as product yield and hydrogen consumption.

[0102] (2) Establish a residue oil hydrogenation mechanism model based on continuous lumping;

[0103] Based on the continuous lumped model, the mechanism model of the residue oil hydrogenation unit is constructed, including the flow model of the residue oil hydrogenation reactor, the reaction network model, the kinetic parameter correction model (referring to the Runge-Kutta numerical calculation method), etc. The established mechanism model is used to predict different feed compositions and feed flow rates (feed composition and feed flow rate are used to calculate the properties and boiling points of the four components) as well as the yield and properties of each component of the reaction product under operating conditions.

[0104] Specifically include:

[0105] Based on the four preset reaction rules, multiple reaction paths and a reaction network composed of multiple reaction paths are represented and generated according to the structure of the raw material molecules and the actual boiling point distillation data, where the reaction rules include the changes in PNA and boiling point of each narrow fraction in the corresponding reaction path.

[0106] For each narrow fraction segment, a partial differential equation is constructed according to the reaction network, and the partial differential equation is solved according to the preset reaction rate constant and the narrow fraction content to obtain the yield and hydrogen consumption of each fraction.

[0107] The partial differential equation is solved according to the narrow fraction content and the preset reaction rate, and the preset reaction rate is corrected according to the difference between the yield of each product and the actual reaction product to obtain a mechanism model based on continuous lumping.

[0108] (3) Establish a prediction model based on recurrent neural network;

[0109] The residue oil hydrotreating operating data was processed with data outliers, smoothed and denoised, and subjected to steady-state analysis. The sample data was normalized using the min-max standardization method.

[0110] The recurrent neural network is trained using the sorted data. The recurrent neural network model is optimized by adjusting the hidden layer model depth n, the number of training times T, and the number of batch samples K to obtain the data model.

[0111] (4) Data-mechanism dual-driven model fusion;

[0112] A parallel hybrid model fusion method is adopted, and the yield of each component is selected as the key indicator of model accuracy. When the key points predicted by the data model meet the set accuracy requirements, the data model is called for simulation calculation; when the accuracy does not meet the requirements, the mechanism model is called for simulation calculation.

[0113] The hydrogen consumption prediction method based on the continuous lumped dual-drive model provided in the embodiment of the present invention has the following advantages:

[0114] (1) The method for constructing a continuous lumped dual-drive model for residue hydrogenation based on a recurrent neural network provided in an embodiment of the present invention is based on a continuous lumped modeling method. According to different process reaction mechanisms, the method can realize the conversion tracking of narrow fractions and different family compositions in the reaction process, provide a clear explanation of the reaction process of each unit in the refining process, and realize the prediction of product yield, hydrogen consumption and product properties (product physical property parameters).

[0115] (2) The data model of the embodiment of the present invention is based on a recurrent neural network structure, so that the current output is not only related to the current input, but also to the previous output. The previous information can be memorized and applied to the calculation of the current output. Therefore, it is suitable for the dynamic prediction of the yield of each component and the hydrogen consumption demand of the residue oil hydrotreating unit.

[0116] (3) The parallel hybrid model fusion method adopted in the embodiment of the present invention integrates the process mechanism and flow information, which can effectively reduce the complexity of the model and greatly improve the model performance; the mechanism model and the data model work in parallel, and the outputs of the mechanism model and the data model are combined by a custom fusion function. The data model can model the effects and processes ignored in the mechanism model through nonlinear fitting, effectively improving the prediction performance of the hybrid model.

[0117] It should be noted that the multiple preferred implementations given in this embodiment can be freely combined under the premise that they do not conflict with each other in logic or structure, and the present invention does not limit this.

[0118] The following describes a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided in an embodiment of the present invention. The hydrogen consumption prediction device based on a continuous lumped dual-drive model described below and the hydrogen consumption prediction method based on a continuous lumped dual-drive model described above can be referred to each other.

[0119] Figure 7 Schematic diagram of the structure of a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided by an embodiment of the present invention. Figure 7 As shown, the device includes a data model prediction module 10 and a mechanism model prediction module 20, wherein: the data model prediction module 10 is used for: in response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, then the data model is used to predict the hydrogen consumption of the residue oil hydrogenation device to obtain a first hydrogen consumption prediction result; the mechanism model prediction module 20 is used for: in response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, then the hydrogen consumption of the residue oil hydrogenation device is predicted using a mechanism model pre-constructed based on preset multiple chemical reactions to obtain a second hydrogen consumption prediction result.

[0120] The hydrogen consumption prediction device based on the continuous lumped dual-drive model provided by the embodiment of the present invention predicts the hydrogen consumption of the residue oil hydrogenation device by using the data model in response to the current operating condition, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, and obtains a first hydrogen consumption prediction result. In response to the current operating condition, if the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, the hydrogen consumption of the residue oil hydrogenation device is predicted by using a mechanism model pre-constructed based on preset multiple chemical reactions, and a second hydrogen consumption prediction result is obtained, thereby improving the accuracy and robustness of the hydrogen consumption prediction of the residue oil hydrogenation device and improving data utilization.

[0121] According to a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided in an embodiment of the present invention, the data model prediction module 10 is used to predict the hydrogen consumption of a residue oil hydrogenation device using the data model to obtain a first hydrogen consumption prediction result. The data model prediction module 10 is specifically used to: input the residue oil hydrogenation operating condition data into the data model to obtain a first product yield prediction result, and obtain the first hydrogen consumption prediction result according to the first product yield prediction result.

[0122] The hydrogen consumption prediction device based on the continuous lumped dual-drive model provided in an embodiment of the present invention obtains a first product yield prediction result by inputting the residual oil hydrogenation operating condition data into the data model, and obtains a first hydrogen consumption prediction result according to the first product yield prediction result, thereby realizing hydrogen consumption prediction based on the residual oil hydrogenation operating condition data and the data model, and further improving the accuracy and robustness of hydrogen consumption prediction.

[0123] According to a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided in an embodiment of the present invention, the residual oil hydrogenation operation condition data includes reactor bed temperature, liquid phase space velocity, residual oil feed composition, residual oil feed flow rate, residual oil feed density and distillation temperature.

[0124] The hydrogen consumption prediction device based on the continuous lumped dual-drive model provided in the embodiment of the present invention further improves the accuracy and robustness of hydrogen consumption prediction by reasonably determining the residual oil hydrogenation operation condition data of the input data model.

[0125] According to a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided by an embodiment of the present invention, the mechanism model prediction module 20 is also used to: based on four preset chemical reactions, represent and generate multiple reaction paths and a reaction network composed of the multiple reaction paths according to the structure of the raw material molecules and the actual boiling point distillation data; for each divided narrow fraction segment, construct a partial differential equation according to the reaction network; wherein the partial differential equation includes a residue oil hydrogenation reactor flow model and a reaction network model, the parameters of the residue oil hydrogenation reactor flow model include the reaction rate, and the parameters of the reaction network model include the component concentration of the narrow fraction segment; solve the partial differential equation according to the preset reaction rate and the initial component concentration of the narrow fraction segment to obtain the predicted yield of the product of each narrow fraction segment; when the difference between the actual yield of the product and the predicted yield is greater than a preset threshold, correct the preset reaction rate, determine the expression of the partial differential equation based on the corrected preset reaction rate, and obtain a mechanism model based on continuous lumping.

[0126] The hydrogen consumption prediction device based on the continuous lumped dual-drive model provided in the embodiment of the present invention, based on four preset chemical reactions, represents and generates multiple reaction paths and a reaction network composed of multiple reaction paths according to the structure of the raw material molecules and the actual boiling point distillation data, constructs a partial differential equation for each divided narrow fraction segment according to the reaction network, solves the partial differential equation according to the preset reaction rate and the initial component concentration of the narrow fraction segment, and obtains the predicted yield of the product of each narrow fraction segment. When the difference between the actual yield of the product and the predicted yield is greater than a preset threshold, the preset reaction rate is corrected, and the expression of the partial differential equation is determined based on the corrected preset reaction rate, so as to obtain a mechanism model based on continuous lumping, thereby realizing the accurate construction of the mechanism model.

[0127] According to a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided in an embodiment of the present invention, the four chemical reactions include hydrocracking reaction, hydrosaturation reaction, hydrodesulfurization reaction and hydrodenitrogenation reaction.

[0128] The embodiment of the present invention improves the accuracy and robustness of the mechanism model in predicting hydrogen consumption by constructing a mechanism model based on the main reactions in the residue hydrogenation reaction, namely, the hydrocracking reaction, the hydrogenation saturation reaction, the hydrodesulfurization reaction and the hydrodenitrogenation reaction.

[0129] According to a hydrogen consumption prediction device based on a continuous lumped dual-drive model provided by an embodiment of the present invention, the mechanism model prediction module 20 is used to predict the hydrogen consumption of the residue oil hydrogenation device by using a mechanism model pre-constructed based on preset multiple chemical reactions to obtain a second hydrogen consumption prediction result. It is specifically used to: bring the initial component concentration of each narrow fraction segment into the partial differential equation, use the Runge-Kutta numerical calculation method to solve the partial differential equation simultaneously, obtain the second product yield prediction result of each narrow fraction segment that changes with time, and obtain the second hydrogen consumption prediction result according to the second product yield prediction result.

[0130] The hydrogen consumption prediction device based on the continuous lumped dual-drive model provided in an embodiment of the present invention, by bringing the initial component concentrations of each narrow fraction segment into the partial differential equation, and using the Runge-Kutta numerical calculation method to simultaneously solve the partial differential equation, obtains the second product yield prediction results of each narrow fraction segment that changes with time, and obtains the second hydrogen consumption prediction results based on the second product yield prediction results, thereby further improving the accuracy and robustness of the hydrogen consumption prediction of the mechanism model.

[0131] Figure 8 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 8As shown, the electronic device may include: a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820 and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logic instructions in the memory 830 to execute the hydrogen consumption prediction method based on the continuous lumped dual-drive model, the method comprising: in response to the current operating condition, the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, then the hydrogen consumption of the residual oil hydrogenation unit is predicted using the data model to obtain a first hydrogen consumption prediction result; in response to the current operating condition, the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, then the hydrogen consumption of the residual oil hydrogenation unit is predicted using a mechanism model pre-constructed based on preset multiple chemical reactions to obtain a second hydrogen consumption prediction result.

[0132] In addition, the logic instructions in the above-mentioned memory 830 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0133] On the other hand, an embodiment of the present invention further provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the hydrogen consumption prediction method based on the continuous lumped dual-drive model provided by the above-mentioned methods, the method including: in response to the current operating conditions, the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, then the data model is used to predict the hydrogen consumption of the residue oil hydrogenation unit to obtain a first hydrogen consumption prediction result; in response to the current operating conditions, the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, then the hydrogen consumption of the residue oil hydrogenation unit is predicted using a mechanism model pre-constructed based on preset multiple chemical reactions to obtain a second hydrogen consumption prediction result.

[0134] On the other hand, an embodiment of the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the hydrogen consumption prediction method based on the continuous lumped dual-drive model provided by the above-mentioned methods, the method comprising: in response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, then the data model is used to predict the hydrogen consumption of the residue oil hydrogenation unit to obtain a first hydrogen consumption prediction result; in response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, then the hydrogen consumption of the residue oil hydrogenation unit is predicted using a mechanism model pre-constructed based on preset multiple chemical reactions to obtain a second hydrogen consumption prediction result.

[0135] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0136] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0137] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting hydrogen consumption based on a continuous lumped dual-drive model, characterized in that: include: In response to the current operating condition, the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, the hydrogen consumption of the residue oil hydrotreating unit is predicted using the data model to obtain a first hydrogen consumption prediction result; In response to the current operating conditions, the prediction result of the data model pre-constructed based on the recurrent neural network does not meet the preset accuracy requirement, the hydrogen consumption of the residue oil hydrogenation unit is predicted using a mechanism model pre-constructed based on preset multiple chemical reactions to obtain a second hydrogen consumption prediction result.

2. The method for predicting hydrogen consumption based on the continuous lumped dual-drive model according to claim 1 is characterized in that: The method of using the data model to predict the hydrogen consumption of the residue oil hydrogenation unit to obtain a first hydrogen consumption prediction result includes: The residual oil hydrogenation operation condition data is input into the data model to obtain the first product yield prediction result, and the first hydrogen consumption prediction result is obtained based on the first product yield prediction result.

3. The method for predicting hydrogen consumption based on the continuous lumped dual-drive model according to claim 2 is characterized in that: The residual oil hydrogenation operation condition data include reactor bed temperature, liquid phase space velocity, residual oil feed composition, residual oil feed flow rate, residual oil feed density and distillation temperature.

4. The method for predicting hydrogen consumption based on the continuous lumped dual-drive model according to claim 1 is characterized in that: The method further comprises: Based on the four preset chemical reactions, multiple reaction paths and a reaction network composed of the multiple reaction paths are represented and generated according to the structure of the raw material molecules and the actual boiling point distillation data; For each divided narrow fraction segment, a partial differential equation is constructed according to the reaction network; wherein the partial differential equation includes a residual oil hydrogenation reactor flow model and a reaction network model, the parameters of the residual oil hydrogenation reactor flow model include a reaction rate, and the parameters of the reaction network model include component concentrations of the narrow fraction segment; Solving the partial differential equation according to a preset reaction rate and an initial component concentration of the narrow fraction segment to obtain a predicted yield of the product of each narrow fraction segment; When the difference between the actual yield of the product and the predicted yield is greater than a preset threshold, the preset reaction rate is corrected, and the expression of the partial differential equation is determined based on the corrected preset reaction rate to obtain a mechanism model based on continuous lumping.

5. The method for predicting hydrogen consumption based on the continuous lumped dual-drive model according to claim 1 is characterized in that: The four chemical reactions include hydrocracking reaction, hydrosaturation reaction, hydrodesulfurization reaction and hydrodenitrogenation reaction.

6. The method for predicting hydrogen consumption based on the continuous lumped dual-drive model according to claim 1 is characterized in that: The method of predicting the hydrogen consumption of the residue oil hydrogenation unit by using a pre-constructed mechanism model based on a plurality of preset chemical reactions to obtain a second hydrogen consumption prediction result includes: By bringing the initial component concentration of each narrow cut segment into the partial differential equation and using the Runge-Kutta numerical calculation method to solve the partial differential equation simultaneously, the second product yield prediction result of each narrow cut segment that changes with time is obtained, and the second hydrogen consumption prediction result is obtained based on the second product yield prediction result.

7. A hydrogen consumption prediction device based on a continuous lumped dual-drive model, characterized in that: include: A data model prediction module is used to: in response to the current operating conditions, if the prediction result of the data model pre-constructed based on the recurrent neural network meets the preset accuracy requirement, use the data model to predict the hydrogen consumption of the residue oil hydrotreating unit to obtain a first hydrogen consumption prediction result; The mechanism model prediction module is used to: in response to the prediction result of the data model pre-constructed based on the recurrent neural network under the current operating conditions not meeting the preset accuracy requirement, use the mechanism model pre-constructed based on the preset multiple chemical reactions to predict the hydrogen consumption of the residue oil hydrogenation unit to obtain a second hydrogen consumption prediction result.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the method for predicting hydrogen consumption based on the continuous lumped dual-drive model as described in any one of claims 1 to 6 are implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for predicting hydrogen consumption based on a continuous lumped dual-drive model as claimed in any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method for predicting hydrogen consumption based on a continuous lumped dual-drive model as claimed in any one of claims 1 to 6 are implemented.

Images