Residual oil hydrogenation reaction effect analysis method and device

Abstract

The invention discloses a residual oil hydrogenation reaction effect analysis method and device, and relates to the technical field of heavy oil hydrogenation and artificial intelligence, and the method comprises the following steps: obtaining historical reaction data of a fixed bed residual oil hydrogenation reaction and historical effect data corresponding to the historical reaction data; the historical reaction data comprises raw oil property data, process condition data and catalyst grading data, and the historical effect data comprises one or any combination of demetallization rates, carbon residue removal rates and desulfurization rates corresponding to the historical reaction data at different reaction times; constructing a sample data set according to the historical reaction data and the historical effect data corresponding to the historical reaction data; constructing a convolutional neural network model; training a convolutional neural network model by using the sample data set to obtain a residual oil hydrogenation reaction effect analysis model; inputting the to-be-analyzed residual oil hydrogenation reaction data into the residual oil hydrogenation reaction effect analysis model to obtain effect data of the to-be-analyzed residual oil hydrogenation reaction data. The application range of the model can be expanded, and the accuracy of the model is improved.

Description

Technical Field

[0001] The invention relates to the fields of heavy oil hydrogenation and artificial intelligence technology, and in particular to a method and a device for analyzing the effect of residual oil hydrogenation reaction. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] At present, with the increasing heaviness and inferiority of oil resources, as well as the urgent requirements for upgrading the quality of oil products and the increasingly stringent environmental protection regulations, refining enterprises are in urgent need of clean and efficient processing technology for heavy oil. Residue oil hydrogenation is the most effective processing technology for heavy and residual oil raw materials. Through residue oil hydrogenation, most of the metals, sulfur and nitrogen in the residue oil are removed, and the residual carbon value is reduced, which is conducive to its further efficient and clean processing and conversion. The technical combination of fixed-bed residue oil hydrogenation treatment and residue oil catalytic cracking can not only maximize the conversion of residual oil with low utilization value and easy to cause environmental pollution, greatly improve the yield of light oil, but also obtain clean oil products with high added value and high quality. In a sense, the crude oil has been converted 100%, realizing the desire to use up the crude oil in the oil refining process.

[0004] For industrial hydrogenation units, in order to guide production operations flexibly, efficiently and at low cost, the existing technology usually adopts a method of establishing a mathematical model based on reaction kinetics to predict product properties, optimize reaction conditions, predict the unit operation cycle and optimize catalyst loading schemes. However, residue oil is the part of crude oil with the highest boiling point, the largest molecular weight, the most complex composition and the highest impurity content. During the hydrogenation process, residue oil undergoes many reactions such as hydrodesulfurization, demetallization, denitrogenation, carbon removal and cracking. The impurity content, molecular structure characteristics and hydrogenation reaction characteristics of residue oil from different origins and different base crude oils vary greatly. In addition, during the hydrogenation process of residue oil, the catalyst usually deactivates much faster than the distillate oil hydrogenation catalyst due to the deposition of metal impurities and carbon deposition. Therefore, it is very difficult to simulate the residue oil hydrogenation process using a kinetic equation based on the reaction mechanism, and it is difficult to accurately reflect the impact of changes in factors such as raw material properties, catalysts and process conditions on the reaction results. The residue oil hydrogenation kinetic model generally has problems such as narrow adaptability, low precision, poor flexibility and difficulty in fitting and solving model parameters.

[0005] In recent years, with the development of machine learning algorithms, they have been widely used in computer vision, speech recognition, text translation, personalized recommendation and other fields and have achieved great success, providing new ideas for the construction of residue oil hydrogenation reaction prediction models. The modeling method based on machine learning is data-driven, and can automatically extract the deep laws and connections between many influencing factors such as raw material properties, process conditions and catalyst grading and reaction results through historical evaluation data. There is no need to consider complex reaction networks and mechanisms, which reduces the difficulty of model construction, has strong nonlinear fitting ability, and high model prediction accuracy. It can effectively solve the drawbacks and shortcomings of traditional kinetic modeling methods. At present, there have been studies on the method of establishing a residue oil hydrogenation reaction prediction model based on machine learning. However, the existing technology mainly uses artificial neural networks or recurrent neural networks to construct a residue oil hydrogenation reaction prediction model, thereby predicting the residue oil hydrogenation reaction process. These methods have the following problems: 1. The models contain many parameters and require a large amount of calculation, which can easily lead to overfitting; 2. Most of them rely on specific industrial devices to provide modeling data. However, in order to maintain stable operation, the raw material properties and operating conditions of industrial devices usually do not change much, and the catalyst loading and grading ratios are fixed. The universality of the models established based on specific device data is restricted; 3. The residue hydrogenation reaction process cannot be accurately predicted, and the reaction influencing factors considered are relatively few and not comprehensive enough. Summary of the invention

[0006] The embodiment of the present invention provides a method for analyzing the effect of a residue oil hydrogenation reaction, which is used to reduce the complexity of a residue oil hydrogenation reaction effect analysis model, reduce the number of model parameters and the amount of calculation, help alleviate overfitting, and improve the generalization and accuracy of the model. The method includes:

[0007] Obtaining historical reaction data of a fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; the historical reaction data includes raw oil property data, process condition data and catalyst grading data, and the historical effect data includes one or any combination of a demetallization rate, a residual carbon removal rate and a desulfurization rate corresponding to the historical reaction data at different reaction times;

[0008] A sample data set is constructed based on the historical reaction data of the fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data;

[0009] Constructing a convolutional neural network model, training the convolutional neural network model with a sample data set, and obtaining a residual oil hydrogenation reaction effect analysis model; the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format;

[0010] The residual oil hydrogenation reaction data to be analyzed are input into the residual oil hydrogenation reaction effect analysis model to obtain the effect data of the residual oil hydrogenation reaction data to be analyzed.

[0011] The embodiment of the present invention also provides a residual oil hydrogenation reaction effect analysis device, which is used to reduce the complexity of the residual oil hydrogenation reaction effect analysis model, reduce the number of model parameters and the amount of calculation, help alleviate overfitting, and improve the generalization and accuracy of the model. The device includes:

[0012] A data acquisition module is used to acquire historical reaction data of a fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; the historical reaction data includes raw oil property data, process condition data and catalyst grading data, and the historical effect data includes one or any combination of the demetallization rate, residual carbon removal rate and desulfurization rate corresponding to the historical reaction data at different reaction times;

[0013] A sample construction module is used to construct a sample data set based on historical reaction data of fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data;

[0014] A model building module is used to build a convolutional neural network model, and use a sample data set to train the convolutional neural network model to obtain a residual oil hydrogenation reaction effect analysis model; the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format;

[0015] The analysis module is used to input the residual oil hydrogenation reaction data to be analyzed into the residual oil hydrogenation reaction effect analysis model to obtain the effect data of the residual oil hydrogenation reaction data to be analyzed.

[0016] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned residual oil hydrogenation reaction effect analysis method when executing the computer program.

[0017] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned residual oil hydrogenation reaction effect analysis method is implemented.

[0018] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned residual oil hydrogenation reaction effect analysis method is implemented.

[0019] In an embodiment of the present invention, a convolutional neural network model is constructed; historical reaction data of a fixed-bed residue oil hydrogenation reaction and corresponding historical effect data thereof are obtained; the historical reaction data include feedstock oil property data, process condition data and catalyst grading data, and the historical effect data include one or any combination of demetallization rate, residual carbon removal rate and desulfurization rate corresponding to the historical reaction data at different reaction times; a sample data set is constructed according to the historical reaction data of the fixed-bed residue oil hydrogenation reaction and the corresponding historical effect data thereof; a convolutional neural network model is trained using the sample data set to obtain a residue oil hydrogenation reaction effect analysis model, the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format; the residue oil hydrogenation reaction data to be analyzed is input into the residue oil hydrogenation reaction effect analysis model to obtain effect data of the residue oil hydrogenation reaction data to be analyzed. Compared with the existing technical solutions of using artificial neural networks or recurrent neural networks to construct a residue oil hydrogenation reaction prediction model, the present invention fully considers the raw oil property data, process condition data and catalyst grading data during the residue oil hydrogenation reaction, and uses the raw oil property data, process condition data and catalyst grading data as input data of the residue oil hydrogenation reaction effect analysis model, and associates them with the reaction effect data. At the same time, a deformation layer is added to the convolutional neural network model, which can convert the input one-dimensional data into a three-dimensional matrix format. In this way, the residue oil hydrogenation reaction effect analysis model constructed by the convolutional neural network model is used to analyze the effect of the residue oil hydrogenation reaction, which can reduce the complexity of the network model, reduce the number of model parameters and the amount of calculation, help alleviate overfitting, and improve the generalization and accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0021] Figure 1 This is a flow chart of a method for analyzing the effect of a residual oil hydrogenation reaction provided in an embodiment of the present invention;

[0022] Figure 2 A structural diagram of a convolutional neural network model provided in an embodiment of the present invention;

[0023] Figure 3 A flow chart of a method for training a convolutional neural network model using sample data to obtain a residual oil hydrogenation reaction effect analysis model in an embodiment of the present invention;

[0024] Figure 4It is a schematic diagram of the absolute error of the desulfurization rate of the model provided in the embodiment of the present invention on the test set;

[0025] Figure 5 A schematic diagram of the absolute error of the residual carbon removal rate of the model provided in an embodiment of the present invention on a test set;

[0026] Figure 6 A schematic diagram of the absolute error of the demetallization rate of the model provided in an embodiment of the present invention on a test set;

[0027] Figure 7 A schematic diagram of a residual oil hydrogenation reaction effect analysis device provided in an embodiment of the present invention;

[0028] Figure 8 A schematic diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0030] In the description of this specification, the terms "include", "including", "have", "contain", etc. are all open terms, which mean including but not limited to. The descriptions with reference to the terms "one embodiment", "a specific embodiment", "some embodiments", "for example", etc. mean that the specific features, structures or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. The order of steps involved in each embodiment is used to schematically illustrate the implementation of the present application, and the order of steps is not limited and can be appropriately adjusted as needed.

[0031] At present, a variety of kinetic models have been developed to simulate the residue hydrogenation process. For example, the measured data of residue hydrogenation industrial units using FZC series catalysts were used, and the empirical kinetic model and catalyst time-varying deactivation model were used to simulate and analyze the measured operation data of three residue hydrogenation industrial units, and a residue hydrogenation deactivation kinetic model was established. Another example is the construction of a residue catalytic hydrodesulfurization (HDS) kinetic model to describe the quantitative relationship between reaction conditions, catalyst deactivation, feedstock properties and other factors and the residue HDS conversion rate.

[0032] However, residue oil is the part of crude oil with the highest boiling point, largest molecular weight, most complex composition and highest impurity content. During the hydrogenation process, residue oil undergoes many reactions such as hydrodesulfurization, demetallization, denitrogenation, carbon removal and cracking. The impurity content, molecular structure characteristics and hydrogenation reaction characteristics of residue oil from different origins and different base crude oils vary greatly. In addition, during the hydrogenation process of residue oil, the catalyst usually deactivates much faster than the distillate oil hydrogenation catalyst due to the deposition of metal impurities and carbon deposition. Therefore, it is very difficult to simulate the residue oil hydrogenation process using kinetic equations based on reaction mechanisms, and it is difficult to accurately reflect the impact of changes in factors such as raw material properties, catalysts and process conditions on the reaction results. Kinetic models for residue oil hydrogenation generally have problems such as narrow adaptability, low accuracy, poor flexibility, and difficulty in fitting and solving model parameters.

[0033] Therefore, with the development of machine learning algorithms, the use of machine learning algorithms in the prediction of residue hydrogenation reactions has been considered. At present, there have been studies on the methods of establishing prediction models for residue hydrogenation reactions based on machine learning, such as the use of long short-term memory neural networks (LSTM) and gated recurrent unit network structures (GRU) to establish a new hydrogen flow prediction model for residue hydrogenation units, which is used to predict the amount of new hydrogen required for the operation of the unit; another example is the use of 200ml small residue hydrogenation unit evaluation data, the application of a fully connected neural network to simulate and analyze the residue hydrogenation process, and the residue hydrogenation law at different space velocities and temperatures is examined.

[0034] It can be seen that the existing technology mainly uses artificial neural networks or recurrent neural networks to construct a residue oil hydrogenation reaction prediction model, thereby predicting the residue oil hydrogenation reaction process. These methods have the following problems: 1. The model contains many parameters, the amount of calculation is large, and it is easy to overfit; 2. Most of them rely on specific industrial devices to provide modeling data, and in order to maintain stable operation, the raw material properties and operating conditions of industrial devices usually do not change much, and the catalyst loading and grading ratio are fixed. The universality of the model established by the specific device data is restricted; 3. It cannot accurately predict the residue oil hydrogenation reaction process, and the reaction influencing factors considered are relatively few and not comprehensive enough.

[0035] In view of the shortcomings of the prior art, the embodiments of the present invention propose a method and device for analyzing the effect of residue oil hydrogenation reaction, which can reduce the complexity of the network model, reduce the number of model parameters and the amount of calculation, help alleviate overfitting, and improve the generalization and accuracy of the model.

[0036] Figure 1 A flow chart of a method for analyzing the effect of a residual oil hydrogenation reaction provided in an embodiment of the present invention, the method comprising the following steps:

[0037] Step 101, obtaining historical reaction data of a fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; the historical reaction data includes feed oil property data, process condition data and catalyst grading data, and the historical effect data includes one or any combination of a demetallization rate, a residual carbon removal rate and a desulfurization rate corresponding to the historical reaction data at different reaction times;

[0038] Step 102, constructing a sample data set according to historical reaction data of fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data;

[0039] Step 103, constructing a convolutional neural network model, using the sample data set to train the convolutional neural network model, and obtaining a residual oil hydrogenation reaction effect analysis model; the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format;

[0040] Step 104, inputting the residue oil hydrogenation reaction data to be analyzed into the residue oil hydrogenation reaction effect analysis model to obtain effect data of the residue oil hydrogenation reaction data to be analyzed.

[0041] In an embodiment of the present invention, a convolutional neural network model is constructed; historical reaction data of a fixed-bed residue oil hydrogenation reaction and corresponding historical effect data thereof are obtained; the historical reaction data include feedstock oil property data, process condition data and catalyst grading data, and the historical effect data include one or any combination of demetallization rate, residual carbon removal rate and desulfurization rate corresponding to the historical reaction data at different reaction times; a sample data set is constructed according to the historical reaction data of the fixed-bed residue oil hydrogenation reaction and the corresponding historical effect data thereof; a convolutional neural network model is trained using the sample data set to obtain a residue oil hydrogenation reaction effect analysis model, the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format; the residue oil hydrogenation reaction data to be analyzed is input into the residue oil hydrogenation reaction effect analysis model to obtain effect data of the residue oil hydrogenation reaction data to be analyzed. Compared with the existing technical solutions of using artificial neural networks or recurrent neural networks to construct a residue oil hydrogenation reaction prediction model, the present invention fully considers the raw oil property data, process condition data and catalyst grading data during the residue oil hydrogenation reaction, and uses the raw oil property data, process condition data and catalyst grading data as input data of the residue oil hydrogenation reaction effect analysis model, and associates them with the reaction effect data. At the same time, a deformation layer is added to the convolutional neural network model, which can convert the input one-dimensional data into a three-dimensional matrix format. In this way, the residue oil hydrogenation reaction effect analysis model constructed by the convolutional neural network model is used to analyze the effect of the residue oil hydrogenation reaction, which can reduce the complexity of the network model, reduce the number of model parameters and the amount of calculation, help alleviate overfitting, and improve the generalization and accuracy of the model.

[0042] Combine the following Figure 1 The method for analyzing the effect of the residue oil hydrogenation reaction shown is described in detail.

[0043] In the above step 101, historical reaction data and corresponding historical effect data in the database system of the fixed-bed residue oil hydrogenation evaluation device and the laboratory management system LIMS may be obtained.

[0044] Among them, the historical reaction data may include crude oil property data, process condition data and catalyst grading data, and the historical effect data is impurity removal performance data, which may include one or any combination of the demetallization rate, residual carbon removal rate and desulfurization rate corresponding to the historical reaction data at different reaction times.

[0045] In one embodiment, the above-mentioned crude oil property data may include one or any combination of the specific gravity, total nitrogen content, total sulfur content, residual carbon content, viscosity, carbon and hydrogen content (C / H content), and metal content of the crude oil.

[0046] In one embodiment, the process condition data may include one or any combination of pressure, temperature, space velocity, and the ratio of hydrogen to oil (hydrogen / oil ratio).

[0047] In one embodiment, the catalyst grading data may include the proportions of different catalysts.

[0048] In the above step 102, a sample data set may be constructed based on historical reaction data of the fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data.

[0049] In one embodiment, before step 102, the following steps may also be included:

[0050] Preprocessing the historical reaction data of the fixed-bed residue oil hydrogenation reaction and the corresponding historical effect data, wherein the preprocessing includes: one or any combination of missing data processing, duplicate data elimination, inconsistent data correction, and abnormal data elimination;

[0051] Step 102 may specifically include:

[0052] Analyze the historical reaction data after pretreatment and screen the characteristic data that affect the catalyst effect;

[0053] A sample data set is constructed based on the characteristic data that affects the catalyst effect and its corresponding historical effect data.

[0054] In specific implementation, after obtaining the historical reaction data and its corresponding historical effect data, the historical data can also be preprocessed and then saved in a unified data format. Specifically, the preprocessing can be: missing data processing, duplicate data elimination, inconsistent data correction, abnormal data elimination, one or any combination thereof.

[0055] Specifically, the missing data processing process can be as follows: for impurity removal performance data, during the fixed bed residue oil hydrogenation reaction, the impurity removal performance data at several time points will be measured, and the impurity removal performance data at similar time points will change relatively little, so the last observation value before the missing time point can be used for filling. For process condition data, since the process conditions of the fixed bed residue oil hydrogenation reaction are relatively stable, the process condition data of different operation cycles are relatively close, and the median or average value can be used for filling. For crude oil property data, since the property data of different crude oils are quite different, the existing data can be used to fit the model, and the model can be used to predict the missing values ​​and fill them. When the sample size is small and the correlation between the attributes is not strong, the median or average value can also be used for filling. . In this way, it can ensure that the missing data filled are more real and reliable, providing favorable data support for the subsequent effect analysis of the residue oil hydrogenation reaction.

[0056] Correction of inconsistent data can be done by unifying various information units of historical data collected from different sources, correcting data that may have incorrect values, etc.

[0057] Abnormal data refers to data that deviates far from other data. Abnormal data can be eliminated by quantile method. According to the quantile method, the value range of abnormal data x can be:

[0058] x>Q 3 +1.5×IQR, or x 1 -1.5×IQR.

[0059] Where: Q 3 is the upper quartile, Q 1 is the lower quartile, and IQR is the difference between the upper and lower quartiles;

[0060] The abnormal data are identified and eliminated according to the above-mentioned value range of the abnormal data x.

[0061] ​In a specific implementation, the pre-processed historical reaction data can be analyzed, and the Pearson correlation coefficient method, variance selection method and prior knowledge can be used to screen the characteristic data that affects the catalyst effect (impurity removal performance) in the historical reaction data, and the characteristic data that affects the catalyst effect is used as the input data of the convolutional neural network model, and the historical effect data is used as the output data of the convolutional neural network model, that is, the specific gravity of the crude oil, the total nitrogen content, the total sulfur content, the residual carbon content, the viscosity, the C / H content, the metal content, the pressure, the temperature, the space velocity, the hydrogen / oil ratio, and the catalyst grading. One or more combinations thereof are used as the input data of the convolutional neural network model, and one or more combinations of the demetallization rate, the residual carbon removal rate and the demetallization rate are used as the output data of the convolutional neural network model.

[0062] After determining the model input features and output data, the characteristic data and historical effect data that affect the catalyst effect are standardized to construct a sample data set.

[0063] Specifically, the standardization method may be a Z-score standardization method:

[0064]

[0065] Where x is the original data, x ′ is the standardized data, μ is the data mean, and σ is the data standard deviation.

[0066] In the above step 103, a convolutional neural network model is constructed, and the convolutional neural network model is trained using a sample data set to obtain a residual oil hydrogenation reaction effect analysis model.

[0067] The convolutional neural network model may include a deformation layer; the deformation layer may convert the input one-dimensional data into a three-dimensional matrix format. In one embodiment, the convolutional neural network model may further include an input layer, a convolution layer, a pooling layer, and a fully connected layer;

[0068] Among them, the input layer is used to receive input data; the convolution layer is used to extract features in the input data; the pooling layer is used to reduce the dimension of the extracted features; and the fully connected layer is used to map the extracted features into effect data.

[0069] In the specific implementation, a residual oil hydrogenation reaction effect analysis model based on convolutional neural network is built. The structure of the constructed convolutional neural network model includes input layer, deformation layer, convolution layer, pooling layer and fully connected layer, among which: the input layer is used to accept the input residual oil hydrogenation reaction data; the deformation layer is used to convert the shape of the input one-dimensional data into the same three-dimensional matrix format as the picture; the convolution layer is used to extract hidden features in the input data; the pooling layer is used to reduce the dimension of the extracted features and reduce the amount of calculation; the fully connected layer is used to map the extracted features to the catalyst decontamination performance, and the number of neurons is consistent with the number of decontamination property data expected to be predicted.

[0070] For example, Figure 2 The structural diagram of the convolutional neural network model provided by the embodiment of the present invention is as follows: Figure 2 As shown, the reaction conditions (process condition data), raw material properties (raw material oil property data), and catalyst grading are used as the input of the model, and the catalyst decontamination performance (effect data) is used as the output of the model. Figure 2 First, the input vector is mapped to a high-dimensional vector through a fully connected layer; since the data of the convolutional neural network must be in the format of a three-dimensional matrix, it is necessary to convert the one-dimensional vector into a three-dimensional tensor (three-dimensional matrix) through a deformation layer, and extract high-level features through a convolution layer, and then convert the feature map extracted by the convolution layer into a one-dimensional vector through a pooling layer, and convert it into a predicted de-impurity performance through a fully connected layer.

[0071] After building the convolutional neural network as described above, the network parameters need to be randomly initialized, and then the parameters of the convolutional neural network model are adjusted using the sample data set.

[0072] In one embodiment, Figure 3 As shown, step 103 may specifically include:

[0073] Step 301, dividing the sample data into a training set, a validation set and a test set according to a preset ratio;

[0074] Step 302, setting training hyperparameters, wherein the training hyperparameters include the type of optimizer, the learning rate of the optimizer, the learning rate decay coefficient, the weight decay coefficient, the number of training rounds, the batch size, the random inactivation probability, the number of convolutional neural network layers, and the number of neurons in each layer of the convolutional neural network;

[0075] Step 303, training the convolutional neural network model using the training set;

[0076] Step 304, using the validation set to determine the error of the convolutional neural network model, and stopping the model training in advance when the error of the convolutional neural network model no longer decreases;

[0077] Step 305, using the test set to test the generalization performance of the convolutional neural network model, when the generalization performance does not meet the preset conditions, adjusting the training hyperparameters, and re-training the convolutional neural network model using the training set; when the generalization performance meets the preset conditions, a residue oil hydrogenation reaction effect analysis model is obtained.

[0078] In step 301, the sample data can be randomly divided into a training set, a validation set and a test set according to a preset ratio (8:1:1). The training set is used for model training, the validation set is used to check the state of the model during the training process, and the test set is used to evaluate the generalization performance of the model.

[0079] In step 302, training hyperparameters are set, and the model is trained for multiple rounds of iterations using the divided training set. The training hyperparameters include the type of optimizer, the learning rate of the optimizer, the learning rate decay coefficient, the weight decay coefficient, the number of training rounds, the batch size, the random inactivation probability, the number of neural network layers, and the number of neurons in each layer of the neural network.

[0080] In step 303, the convolutional neural network model may be trained using the training set.

[0081] In one embodiment, step 303 may specifically include:

[0082] Input the characteristic data affecting the catalyst effect in the training set into the convolutional neural network model, and obtain the predicted effect data of the convolutional neural network model through network forward propagation;

[0083] The mean square error loss function is used to calculate the error between the predicted effect data and the historical effect data in the training set;

[0084] The error between the predicted effect data and the historical effect data in the training set is used to adjust the parameters of the convolutional neural network model through the error back propagation algorithm.

[0085] In specific implementation, the specific training process may include: inputting the feature data in the training set into the convolutional neural network model, obtaining the prediction effect data of the model through network forward propagation, calculating the model prediction error using the mean square error loss function, and updating the model parameters through the error back propagation algorithm.

[0086] In step 304, the validation set may be used to monitor the training process of the convolutional neural network model.

[0087] In specific implementation, the error of the convolutional neural network model can be calculated based on the validation set. When the error no longer decreases, the training of the convolutional neural network model needs to be stopped in advance, even if the number of training times of the convolutional neural network model has not reached the number of training rounds. In this way, overfitting of the convolutional neural network model can be avoided.

[0088] In step 305, the generalization performance of the model is tested using the test set. If the model performance fails to meet the expected requirements (preset conditions), the training hyperparameters are adjusted, and then the model is retrained and evaluated. If the model performance meets the expected requirements, the model training ends.

[0089] In step 104, the residue oil hydrogenation reaction data to be analyzed may be input into a residue oil hydrogenation reaction effect analysis model to obtain effect data of the residue oil hydrogenation reaction data to be analyzed.

[0090] During specific implementation, the residue oil hydrogenation reaction data to be analyzed are processed to meet the model input requirements, including unit conversion, data scaling and other processing measures; the processed residue oil hydrogenation reaction data to be analyzed are input into the residue oil hydrogenation reaction effect analysis model, and the analysis results, i.e., the effect data of the residue oil hydrogenation reaction data to be analyzed, are read.

[0091] In summary, the above-mentioned residual oil hydrogenation reaction effect analysis method includes the following beneficial effects:

[0092] 1. It is helpful to realize the accurate analysis and prediction of the catalyst decontamination performance (effect data of residual oil hydrogenation reaction). Specifically: when analyzing the catalyst decontamination performance, the influence of various factors such as raw oil properties, process conditions, catalyst grading, etc. is fully considered. The raw oil properties, process conditions and catalyst grading are used as model input variables, and are associated with the catalyst decontamination performance. At the same time, a convolutional neural network is used to build a prediction model. The convolutional neural network is one of the representative algorithms of deep learning. It is usually composed of a convolution layer, a pooling layer, and a fully connected layer. Since the convolutional neural network model is usually used to process three-dimensional data such as images, therefore, there is currently no use of the model to process one-dimensional data such as residual oil hydrogenation reaction. The embodiment of the present invention adds a deformation layer to the convolutional neural network model, which can convert the input one-dimensional data into a three-dimensional matrix format. In this way, the prediction model is constructed using a convolutional neural network. Compared with other neural networks, its particularity mainly lies in two aspects: weight sharing and sparse connection. Weight sharing enables the convolutional layer of the convolutional neural network to share the parameters of the convolution kernel; sparse connection is not like a traditional neural network, where each neuron in the upper layer is connected to all neurons in the lower layer, but rather, the neurons in the upper layer are connected to some neurons in the lower layer. The role of these two characteristics is to reduce the complexity of the network model, reduce the number of model parameters and the amount of calculation, help alleviate overfitting, and improve the generalization and accuracy of the model.

[0093] 2. It helps to reduce the workload of residual oil hydroprocessing analysis and save a lot of manpower and material resources. Specifically, in the embodiment of the present invention, only a large amount of historical data information is needed to predict the reaction performance of the new catalyst, thereby reducing the workload of manual analysis experiments and reducing research and development costs.

[0094] In order to more clearly understand the above-mentioned method for analyzing the effect of residue oil hydrogenation reaction, a specific embodiment and the comparative effect of this embodiment and the prior art are described below.

[0095] Embodiment 1:

[0096] Step 1: Obtain historical evaluation data in the database system of the fixed-bed residue oil hydrogenation evaluation device and the laboratory management system LIMS (i.e., obtain historical reaction data of the fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data). In this embodiment, a total of 442 residue oil hydrogenation evaluation experimental data were collected, and the collected data information included specific gravity, total nitrogen content, total sulfur content, residual carbon content, viscosity, C / H content, metal content, pressure, temperature, space velocity, hydrogen / oil ratio, catalyst grading, desulfurization rate, demetallization rate and residual carbon removal rate of the feed oil.

[0097] Step 2: Correct the historical evaluation data. Manually correct the unreasonable data in the historical evaluation data, and unify the units of each data item of the historical evaluation data to generate corrected historical evaluation data.

[0098] Step 3: Preprocess the historical evaluation data corrected in step 2. First, delete the samples with outliers and missing data in the historical evaluation data. After processing, 361 historical evaluation data are obtained for the subsequent model construction. Then the missing values ​​in the historical evaluation data are processed, and the two attributes of C / H and viscosity with a high proportion of missing data are eliminated and not used as model input. At the same time, for the attribute of pressure, since the pressure values ​​corresponding to all samples are the same, they are also eliminated and not used as model input. 15 features are selected as model input features, and desulfurization rate, demetallization rate and residual carbon removal rate are selected as model outputs. Then the input features are standardized and the output labels are scaled to between 0-1 to prevent gradient explosion, stabilize the model training process, and accelerate model convergence. After the data preprocessing is completed, the data set is divided into three parts: training set, validation set and test set in a ratio of 8:1:1. The training set includes 288 samples, the validation set contains 37 samples, and the test set contains 37 samples.

[0099] Step 4: First, build a convolutional neural network model. The structure of the convolutional neural network model constructed in this embodiment is as follows: Figure 2As shown. Then, the model is trained using the divided training set. During model training, the mean square error loss function is selected to measure the gap between the model prediction value and the actual value (real effect data, that is, the historical effect data in the training set), and the model parameters are updated through the back propagation algorithm. During the model training process, the validation set is used to monitor the model training process. When the model error on the test set decreases, the model parameters are saved. After the training is completed, the generalization performance of the model is tested using the test set. If the model performance meets the expected requirements, the model training is stopped. Otherwise, the model structure, training hyperparameters, etc. are adjusted, and the model is retrained until the model effect on the test set meets the expected requirements.

[0100] Step 5: Use the model obtained in step 4 to analyze the impurity removal performance of the residue oil hydrogenation catalyst to be predicted (the effect data of the residue oil hydrogenation reaction data to be analyzed).

[0101] Verification results: Figure 4 , Figure 5 and Figure 6 The absolute error diagrams of the desulfurization rate, carbon removal rate and metal removal rate of the model in this embodiment on the test set are shown respectively. The effect analysis method of the residual oil hydrogenation reaction provided by the embodiment of the present invention is repeated five times, and the average values ​​of the evaluation indicators of the five operations are shown in Table 1.

[0102] Comparative Example 1 (Prior Art):

[0103] Comparative Example 1 is a residue oil hydrogenation model based on semi-supervised deep GRU. By performing time series processing on the original data, an unlabeled sample training set A and a labeled sample training set A* are obtained; the sample data is normalized using the min-max normalization method to eliminate the impact of different dimensions; the normalized sample data sets A and A* are both input data for the modeling process; a deep GRU model is constructed using the unlabeled sample data set A processed as above, and the deep GRU model is pre-trained layer by layer in an unsupervised manner using the GRU autoencoder according to the greedy principle; the deep GRU model of the same structure is initialized with the parameters obtained from the pre-training, and the labeled data set A* is used for secondary training. After supervised learning fine-tuning, the final residue oil hydrogenation model based on the semi-supervised deep GRU is obtained.

[0104] Using the same training set, validation set, and test set as in Example 1, the data was preprocessed according to the method of Comparative Example 1 to construct an analysis model. The errors of the evaluation indicators from five runs are shown in Table 1.

[0105] Comparative Example 2 (Prior Art):

[0106] Comparative Example 2 is a residue oil hydrogenation reaction kinetic model based on endoplasmic reticulum-like membrane calculation, and the implementation steps include: 1) obtaining actual input and output data of the residue oil hydrogenation process from the refinery, and for the same set of sampled input data, the sum of the absolute values ​​of the errors between the estimated output and the actual output of the model is used as the objective function; 2) according to the structure of cells, the function of the endoplasmic reticulum and the cell differentiation phenomenon in nature, abstracting the parameter estimation method based on endoplasmic reticulum-like membrane calculation and its corresponding bionic optimization rules; 3) setting various parameters of the algorithm; 4) running the estimation of parameters in the residue oil hydrogenation reaction kinetic model based on endoplasmic reticulum-like membrane calculation.

[0107] Using the same data set as in Example 1, according to the method of Comparative Example 2, the sum of the absolute values ​​of the errors between the estimated output and the actual output was used as the objective function, and the parameters in the residue oil hydrogenation reaction kinetic model were estimated based on the endoplasmic reticulum-like membrane calculation. The errors of the evaluation indicators of the analytical model are shown in Table 1.

[0108] Comparative Example 3 (Prior Art):

[0109] Comparative Example 3 is the optimization modeling of the residue oil hydrogenation process of the variable search domain P system, and the implementation steps include: 1) obtaining the input and output sampling data of the residue oil hydrogenation process through field operation or experiment, and taking the sum of the squares of the relative errors between the estimated output of the residue oil hydrogenation process and the actual output sampling data as the objective function; 2) based on the contraction and expansion mechanism of the biological cell membrane and the selection and transport function, abstracting the optimization modeling method of the residue oil hydrogenation process of the P system with contraction and expansion mechanism and its corresponding bionic rules; 3) setting the algorithm operation parameters; 4) running the variable search domain P system optimization algorithm to estimate the unknown parameters in the residue oil hydrogenation process model, and obtaining the estimated values ​​of the unknown parameters in the model by minimizing the objective function, and substituting the estimated values ​​into the residue oil hydrogenation process model to form a mathematical model.

[0110] Using the same data set as in Example 1, according to the method of Comparative Example 3, the sum of the absolute values ​​of the errors between the estimated output and the actual output was used as the objective function, and the variable search domain P system optimization algorithm was run to estimate the unknown parameters in the residue hydrotreating process model. The errors of the prediction model evaluation indicators are shown in Table 1.

[0111] Table 1. Error of evaluation index of prediction model

[0112]

[0113]

[0114] It can be seen from Table 1 that the difference between the predicted value and the actual value of the model in this embodiment is small, and the linear correlation is good; while the model errors of Comparative Examples 1, 2, and 3 are large, and their linear correlation is poor.

[0115] The present invention also provides a residual oil hydrogenation reaction effect analysis device, as described in the following embodiments. Since the principle of the device to solve the problem is similar to the residual oil hydrogenation reaction effect analysis method, the implementation of the device can refer to the implementation of the residual oil hydrogenation reaction effect analysis method, and the repeated parts will not be repeated.

[0116] like Figure 7 FIG. 1 is a schematic diagram of a residual oil hydrogenation reaction effect analysis device provided by an embodiment of the present invention, and the device may include:

[0117] The data acquisition module 701 is used to acquire historical reaction data of the fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; the historical reaction data includes raw oil property data, process condition data and catalyst grading data, and the historical effect data includes one or any combination of the demetallization rate, residual carbon removal rate and desulfurization rate corresponding to the historical reaction data at different reaction times;

[0118] A sample construction module 702 is used to construct a sample data set according to the historical reaction data of the fixed bed residue oil hydrogenation reaction and its corresponding historical effect data;

[0119] The model building module 703 is used to build a convolutional neural network model, and train the convolutional neural network model using a sample data set to obtain a residual oil hydrogenation reaction effect analysis model; the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format;

[0120] The analysis module 704 is used to input the residue oil hydrogenation reaction data to be analyzed into the residue oil hydrogenation reaction effect analysis model to obtain the effect data of the residue oil hydrogenation reaction data to be analyzed.

[0121] In one embodiment, the crude oil property data may include one or any combination of the specific gravity, total nitrogen content, total sulfur content, residual carbon content, viscosity, carbon and hydrogen content, and metal content of the crude oil.

[0122] In one embodiment, the process condition data may include one or any combination of pressure, temperature, space velocity, and ratio of hydrogen to oil.

[0123] In one embodiment, the catalyst grading data may include the proportions of different catalysts.

[0124] In one embodiment, a pre-processing module may also be included, which is used before the sample construction module constructs the sample data set according to the historical reaction data of the fixed bed residue oil hydrogenation reaction and its corresponding historical effect data:

[0125] Preprocessing the historical reaction data of the fixed-bed residue oil hydrogenation reaction and the corresponding historical effect data, wherein the preprocessing includes: one or any combination of missing data processing, duplicate data elimination, inconsistent data correction, and abnormal data elimination;

[0126] Sample building blocks that can be used to:

[0127] Analyze the historical reaction data after pretreatment and screen the characteristic data that affect the catalyst effect;

[0128] A sample data set is constructed based on the characteristic data that affects the catalyst effect and its corresponding historical effect data.

[0129] In one embodiment, the convolutional neural network model may further include an input layer, a convolutional layer, a pooling layer, and a fully connected layer;

[0130] Among them, the input layer is used to receive input data; the convolution layer is used to extract features in the input data; the pooling layer is used to reduce the dimension of the extracted features; and the fully connected layer is used to map the extracted features into effect data.

[0131] In one embodiment, the model building module can be specifically used to:

[0132] Divide the sample data into training set, validation set and test set according to the preset ratio;

[0133] Setting training hyperparameters, wherein the training hyperparameters include the type of optimizer, the learning rate of the optimizer, the learning rate decay coefficient, the weight decay coefficient, the number of training rounds, the batch size, the random inactivation probability, the number of convolutional neural network layers, and the number of neurons in each layer of the convolutional neural network;

[0134] Use the training set to train the convolutional neural network model;

[0135] Use the validation set to determine the error of the convolutional neural network model, and stop model training in advance when the error of the convolutional neural network model no longer decreases;

[0136] The generalization performance of the convolutional neural network model is tested using the test set. When the generalization performance does not meet the preset conditions, the training hyperparameters are adjusted and the convolutional neural network model is retrained using the training set. When the generalization performance meets the preset conditions, a residue oil hydrogenation reaction effect analysis model is obtained.

[0137] In one embodiment, the model building module can also be used to:

[0138] Input the characteristic data affecting the catalyst effect in the training set into the convolutional neural network model, and obtain the predicted effect data of the convolutional neural network model through network forward propagation;

[0139] The mean square error loss function is used to calculate the error between the predicted effect data and the historical effect data in the training set;

[0140] The error between the predicted effect data and the historical effect data in the training set is used to adjust the parameters of the convolutional neural network model through the error back propagation algorithm.

[0141] The embodiment of the present invention also provides a computer device, such as Figure 8 As shown, it is a schematic diagram of a computer device in an embodiment of the present invention, wherein the computer device 800 includes a memory 810, a processor 820, and a computer program 830 stored in the memory 810 and executable on the processor 820, and when the processor 820 executes the computer program 830, the above-mentioned residual oil hydrogenation reaction effect analysis method is implemented.

[0142] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned residual oil hydrogenation reaction effect analysis method is implemented.

[0143] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned residual oil hydrogenation reaction effect analysis method is implemented.

[0144] In an embodiment of the present invention, a convolutional neural network model is constructed; historical reaction data of a fixed-bed residue oil hydrogenation reaction and corresponding historical effect data thereof are obtained; the historical reaction data include feedstock oil property data, process condition data and catalyst grading data, and the historical effect data include one or any combination of demetallization rate, residual carbon removal rate and desulfurization rate corresponding to the historical reaction data at different reaction times; a sample data set is constructed according to the historical reaction data of the fixed-bed residue oil hydrogenation reaction and the corresponding historical effect data thereof; a convolutional neural network model is trained using the sample data set to obtain a residue oil hydrogenation reaction effect analysis model; the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format; the residue oil hydrogenation reaction data to be analyzed is input into the residue oil hydrogenation reaction effect analysis model to obtain effect data of the residue oil hydrogenation reaction data to be analyzed. Compared with the existing technical solutions of using artificial neural networks or recurrent neural networks to construct a residue oil hydrogenation reaction prediction model, the present invention fully considers the raw oil property data, process condition data and catalyst grading data during the residue oil hydrogenation reaction, and uses the raw oil property data, process condition data and catalyst grading data as input data of the residue oil hydrogenation reaction effect analysis model, and associates them with the reaction effect data. At the same time, a deformation layer is added to the convolutional neural network model, which can convert the input one-dimensional data into a three-dimensional matrix format. In this way, the residue oil hydrogenation reaction effect analysis model constructed by the convolutional neural network model is used to analyze the effect of the residue oil hydrogenation reaction, which can reduce the complexity of the network model, reduce the number of model parameters and the amount of calculation, help alleviate overfitting, and improve the generalization and accuracy of the model.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0147] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0149] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for analyzing the effect of residual oil hydrogenation reaction, characterized in that: include: Obtaining historical reaction data of a fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; the historical reaction data includes raw oil property data, process condition data and catalyst grading data, and the historical effect data includes one or any combination of a demetallization rate, a residual carbon removal rate and a desulfurization rate corresponding to the historical reaction data at different reaction times; A sample data set is constructed based on the historical reaction data of the fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; Constructing a convolutional neural network model, training the convolutional neural network model with a sample data set, and obtaining a residual oil hydrogenation reaction effect analysis model; the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format; The residual oil hydrogenation reaction data to be analyzed are input into the residual oil hydrogenation reaction effect analysis model to obtain the effect data of the residual oil hydrogenation reaction data to be analyzed.

2. The method according to claim 1, characterized in that The raw oil property data include one or any combination of the specific gravity, total nitrogen content, total sulfur content, residual carbon content, viscosity, carbon and hydrogen content, and metal content of the raw oil.

3. The method according to claim 1, characterized in that The process condition data include one or any combination of pressure, temperature, space velocity, and ratio of hydrogen to oil.

4. The method according to claim 1, characterized in that The catalyst grading data include the proportions of different catalysts.

5. The method according to claim 1, characterized in that According to the historical reaction data of fixed bed residue oil hydrogenation reaction and its corresponding historical effect data, before constructing the sample data set, it also includes: Preprocessing the historical reaction data of the fixed-bed residue oil hydrogenation reaction and the corresponding historical effect data, wherein the preprocessing includes: one or any combination of missing data processing, duplicate data elimination, inconsistent data correction, and abnormal data elimination; Based on the historical reaction data of fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data, a sample data set is constructed, including: Analyze the historical reaction data after pretreatment and screen the characteristic data that affect the catalyst effect; A sample data set is constructed based on the characteristic data that affects the catalyst effect and its corresponding historical effect data.

6. The method according to claim 1, characterized in that The convolutional neural network model also includes an input layer, a convolution layer, a pooling layer and a fully connected layer; Among them, the input layer is used to receive input data; the convolution layer is used to extract features in the input data; the pooling layer is used to reduce the dimension of the extracted features; and the fully connected layer is used to map the extracted features into effect data.

7. The method according to claim 6, characterized in that The convolutional neural network model is trained using the sample data set to obtain a residual oil hydrogenation reaction effect analysis model, including: Divide the sample data into training set, validation set and test set according to the preset ratio; Setting training hyperparameters, wherein the training hyperparameters include the type of optimizer, the learning rate of the optimizer, the learning rate decay coefficient, the weight decay coefficient, the number of training rounds, the batch size, the random inactivation probability, the number of convolutional neural network layers, and the number of neurons in each layer of the convolutional neural network; Use the training set to train the convolutional neural network model; Use the validation set to determine the error of the convolutional neural network model, and stop model training in advance when the error of the convolutional neural network model no longer decreases; The generalization performance of the convolutional neural network model is tested using the test set. When the generalization performance does not meet the preset conditions, the training hyperparameters are adjusted and the convolutional neural network model is retrained using the training set. When the generalization performance meets the preset conditions, a residue oil hydrogenation reaction effect analysis model is obtained.

8. The method according to claim 7, characterized in that The convolutional neural network model is trained using the training set, including: Input the characteristic data affecting the catalyst effect in the training set into the convolutional neural network model, and obtain the predicted effect data of the convolutional neural network model through network forward propagation; The mean square error loss function is used to calculate the error between the predicted effect data and the historical effect data in the training set; The error between the predicted effect data and the historical effect data in the training set is used to adjust the parameters of the convolutional neural network model through the error back propagation algorithm.

9. A residual oil hydrogenation reaction effect analysis device, characterized in that: include: A data acquisition module is used to acquire historical reaction data of a fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; the historical reaction data includes raw oil property data, process condition data and catalyst grading data, and the historical effect data includes one or any combination of the demetallization rate, residual carbon removal rate and desulfurization rate corresponding to the historical reaction data at different reaction times; A sample construction module is used to construct a sample data set based on historical reaction data of fixed-bed residue oil hydrogenation reaction and its corresponding historical effect data; A model building module is used to build a convolutional neural network model, and use a sample data set to train the convolutional neural network model to obtain a residual oil hydrogenation reaction effect analysis model; the convolutional neural network model includes a deformation layer; the deformation layer is used to convert the input one-dimensional data into a three-dimensional matrix format; The analysis module is used to input the residual oil hydrogenation reaction data to be analyzed into the residual oil hydrogenation reaction effect analysis model to obtain the effect data of the residual oil hydrogenation reaction data to be analyzed.

10. The device according to claim 9, characterized in that The raw oil property data include one or any combination of the specific gravity, total nitrogen content, total sulfur content, residual carbon content, viscosity, carbon and hydrogen content, and metal content of the raw oil.

11. The device according to claim 9, characterized in that The process condition data include one or any combination of pressure, temperature, space velocity, and ratio of hydrogen to oil.

12. The device according to claim 9, characterized in that The catalyst grading data include the proportions of different catalysts.

13. The device according to claim 9, characterized in that It also includes a pre-processing module, which is used before the sample construction module constructs the sample data set according to the historical reaction data of the fixed bed residue oil hydrogenation reaction and its corresponding historical effect data: Preprocessing the historical reaction data of the fixed-bed residue oil hydrogenation reaction and the corresponding historical effect data, wherein the preprocessing includes: one or any combination of missing data processing, duplicate data elimination, inconsistent data correction, and abnormal data elimination; Sample building blocks for: Analyze the historical reaction data after pretreatment and screen the characteristic data that affect the catalyst effect; A sample data set is constructed based on the characteristic data that affects the catalyst effect and its corresponding historical effect data.

14. The device according to claim 9, characterized in that The convolutional neural network model also includes an input layer, a convolution layer, a pooling layer and a fully connected layer; Among them, the input layer is used to receive input data; the convolution layer is used to extract features in the input data; the pooling layer is used to reduce the dimension of the extracted features; and the fully connected layer is used to map the extracted features into effect data.

15. The device according to claim 14, characterized in that Model building modules, specifically for: Divide the sample data into training set, validation set and test set according to the preset ratio; Setting training hyperparameters, wherein the training hyperparameters include the type of optimizer, the learning rate of the optimizer, the learning rate decay coefficient, the weight decay coefficient, the number of training rounds, the batch size, the random inactivation probability, the number of convolutional neural network layers, and the number of neurons in each layer of the convolutional neural network; Use the training set to train the convolutional neural network model; Use the validation set to determine the error of the convolutional neural network model, and stop model training in advance when the error of the convolutional neural network model no longer decreases; The generalization performance of the convolutional neural network model is tested using the test set. When the generalization performance does not meet the preset conditions, the training hyperparameters are adjusted and the convolutional neural network model is retrained using the training set. When the generalization performance meets the preset conditions, a residue oil hydrogenation reaction effect analysis model is obtained.

16. The device according to claim 15, characterized in that Model building modules are also used to: Input the characteristic data affecting the catalyst effect in the training set into the convolutional neural network model, and obtain the predicted effect data of the convolutional neural network model through network forward propagation; The mean square error loss function is used to calculate the error between the predicted effect data and the historical effect data in the training set; The error between the predicted effect data and the historical effect data in the training set is used to adjust the parameters of the convolutional neural network model through the error back propagation algorithm.

17. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

18. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

19. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images