Construction method and device for soft measurement model of initial boiling point of heavy naphtha and application

Abstract

The invention relates to the technical field of petroleum initial boiling point detection, in particular to a construction method and device of a soft measurement model of a heavy naphtha initial boiling point and application. According to the method, firstly, auxiliary variables related to the heavy naphtha initial boiling point in the refining and chemical industry are decomposed into trend signals and periodic signals based on a variational mode decomposition method, trend features and periodic features are extracted respectively, and then a double-space feature encoder constructed based on a graph attention model learns the dynamic interaction relation of all the auxiliary variables; extracting spatial-temporal characteristics of the trend characteristics and the periodic characteristics, and finally constructing a soft measurement model of the initial boiling point of the heavy naphtha according to the dual spatial-temporal characteristics of the trend characteristics and the periodic characteristics. According to the soft measurement model constructed by the method, the dynamic interaction relationship between the auxiliary variables is fully considered, the robustness of the model is improved, the generalization ability is enhanced, and the prediction precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of petroleum initial boiling point detection, and particularly to a construction method, device and application of a soft measurement model for the initial boiling point of heavy naphtha. Background Art

[0002] Naphtha is one of the petroleum products, also known as chemical light oil, which is a light oil used as a chemical raw material processed from crude oil or other raw materials. Generally, it contains 55.4% of straight-chain alkanes, 30.3% of monocyclic alkanes, 2.4% of bicyclic alkanes, 11.7% of alkylbenzenes, 0.1% of benzene, 0.1% of indane and tetralin, with an average molecular weight of 114 and an explosion limit of 1.2% - 6.0%. It is mainly used as a chemical raw material. Predicting the initial boiling point of heavy naphtha is of great significance for increasing the output of naphtha and realizing product quality control.

[0003] In the related art, soft measurement models based on data-driven are used to predict the initial boiling point of heavy naphtha, including multiple linear regression, multi-layer perceptron, convolutional neural network, etc.

[0004] In the prior art, in the actual industrial process, the relationship between different variables may be complex and non-linear, and the data-driven soft measurement model cannot accurately capture this complex relationship, resulting in poor performance of the constructed soft measurement model. Summary of the Invention

[0005] In view of the above problems, the present invention provides a construction method, device and application of a soft measurement model for the initial boiling point of heavy naphtha, which is used to solve the technical problem of poor performance of the data-driven soft measurement model.

[0006] In a first aspect, an embodiment of the present invention provides a construction method of a soft measurement model for the initial boiling point of heavy naphtha, including: obtaining auxiliary variable information related to the initial boiling point of heavy naphtha; decomposing the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method; respectively extracting the features of the trend signal and the periodic signal to obtain corresponding trend features and periodic features; respectively extracting the spatio-temporal features of the trend features and the periodic features based on the dual-space feature encoder of the graph attention model; splicing the two spatio-temporal features to obtain a joint spatio-temporal feature; constructing a soft measurement model for predicting the initial boiling point of heavy naphtha based on the joint spatio-temporal feature.

[0007] Further, the auxiliary variables include: the top temperature of the petroleum fractionation tower, the top pressure, the heat flow rate of the reboiler at the bottom of the tower, the side stream flow rate, the bottom temperature and the bottom outlet temperature.

[0008] Further, the step of respectively extracting the features of the trend signal and the periodic signal includes: respectively extracting the features of the trend signal and the periodic signal based on the long short-term memory network model.

[0009] Further, decomposing the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method includes:

[0010] Utilizing the variational mode decomposition method to decompose the auxiliary variable information into a trend signal and a periodic signal, the expression is as follows:

[0011] [x tp , x tt = VMD(x t )

[0012] where x t = (x t1 , x t2 , x t3 , x t4 , x t5 , x t6 ), represents the auxiliary variable information, x t1 is the top temperature, x t2 is the top pressure, x t3 is the heat flow rate of the reboiler at the bottom of the column, x t4 is the side stream flow rate, x t5 is the bottom temperature, x t6 is the outlet temperature at the bottom of the column, x tt represents the trend signal, x tp represents the periodic signal; VMD represents the variational mode decomposition;

[0013] The trend signal is used to represent the change trend of the raw material input or product output data in the chemical process, and the periodic signal is used to represent the signal whose amplitude changes repeatedly with time.

[0014] Further, the dual-space feature encoder based on the graph attention model respectively extracts the spatio-temporal features of the trend feature and the periodic feature, including: determining the first dynamic interaction relationship value between any two auxiliary variables based on the trend features corresponding to each auxiliary variable, and determining the spatio-temporal feature corresponding to the trend feature according to the first dynamic interaction relationship value; determining the second dynamic interaction relationship value between any two auxiliary variables based on the periodic features corresponding to each auxiliary variable, and determining the spatio-temporal feature corresponding to the periodic feature according to the second dynamic interaction relationship value.

[0015] Further, the calculation formula for the first dynamic interaction relationship value or the second dynamic relationship value is as follows:

[0016]

[0017] The calculation formula for the dual spatio-temporal feature corresponding to the trend feature or the periodic feature is as follows:

[0018]

[0019] Among them, e i,j represents the first dynamic interaction relationship value or the second dynamic interaction relationship value between the i-th auxiliary variable and the j-th auxiliary variable. N(i) represents the set of auxiliary variables. respectively represent the trend characteristics or periodic characteristics of the i-th, j-th, and k-th auxiliary variables. k ∈ N(i) means that k takes all the auxiliary variables in the set of auxiliary variables. W represents the mapping weight of the corresponding trend characteristics or periodic characteristics of the auxiliary variables. a T represents the transpose of the preset weight vector. LeakyReLU represents the activation function. h′ i represents the spatio-temporal characteristics of the trend characteristics or periodic characteristics. e i,i represents the first dynamic interaction relationship value or the second dynamic interaction relationship value between the i-th auxiliary variable itself. j ∈ N(i) means that j takes the other auxiliary variables in the set of auxiliary variables except the i-th auxiliary variable.

[0020] Further, the soft measurement model for predicting the initial boiling point of heavy naphtha constructed based on the joint spatio-temporal characteristics includes: learning the dynamic mapping relationship between the joint spatio-temporal characteristics and the initial boiling point of heavy naphtha based on Transformer to establish the soft measurement model.

[0021] In a second aspect, an embodiment of the present invention provides an application method of a soft measurement model for the initial boiling point of heavy naphtha. The soft measurement model constructed by using the construction method of the soft measurement model for the initial boiling point of heavy naphtha described in the first aspect is used to measure the initial boiling point of heavy naphtha, including: collecting the current auxiliary variable information related to the initial boiling point of heavy naphtha; inputting the current auxiliary variable information into the soft measurement model, and outputting the initial boiling point information of heavy naphtha.

[0022] In a third aspect, an embodiment of the present invention provides a construction device for a soft measurement model of the initial boiling point of heavy naphtha, including: an information acquisition module for acquiring auxiliary variable information related to the initial boiling point of heavy naphtha; a signal decomposition module for decomposing the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method; a first extraction module for respectively extracting the characteristics of the trend signal and the periodic signal to obtain corresponding trend characteristics and periodic characteristics; a second extraction module for respectively extracting the spatio-temporal characteristics of the trend characteristics and the periodic characteristics based on the dual-space feature encoder of the graph attention model; a feature splicing module for splicing the two spatio-temporal characteristics to obtain joint spatio-temporal characteristics; a model construction module for constructing a soft measurement model for predicting the initial boiling point of heavy naphtha based on the joint spatio-temporal characteristics.

[0023] Fourth aspect, an application device of the soft sensor model for the initial boiling point of heavy naphtha according to an embodiment of the present invention uses the soft sensor model constructed by the construction method of the soft sensor model for the initial boiling point of heavy naphtha described in the first aspect to measure the initial boiling point of heavy naphtha. The device includes: an acquisition module for acquiring current auxiliary variable information related to the initial boiling point of heavy naphtha; a prediction module for inputting the current auxiliary variable information into the soft sensor model and outputting the initial boiling point information of heavy naphtha.

[0024] Fifth aspect, an embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus; the memory is used to store a computer program; when the processor executes the program stored on the memory, it implements the steps of the construction method of the soft sensor model for the initial boiling point of heavy naphtha described in the first aspect or the application method of the soft sensor model for the initial boiling point of heavy naphtha described in the second aspect.

[0025] Sixth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the construction method of the soft sensor model for the initial boiling point of heavy naphtha described in the first aspect or the application method of the soft sensor model for the initial boiling point of heavy naphtha described in the second aspect.

[0026] The present invention at least has the following beneficial effects:

[0027] The present invention first decomposes the auxiliary variables related to the initial boiling point of heavy naphtha in the refining industry into trend signals and periodic signals based on the variational mode decomposition method, and extracts trend features and periodic features respectively. Then, the dual-space feature encoder constructed based on the graph attention model learns the dynamic interaction relationships of each auxiliary variable, extracts the spatio-temporal features of the trend features and periodic features. Finally, a soft sensor model for the initial boiling point of heavy naphtha is constructed according to the dual spatio-temporal features of the trend features and periodic features. The soft sensor model constructed by the present invention fully considers the dynamic interaction relationships between auxiliary variables, improves the robustness of the model, enhances the generalization ability, and improves the prediction accuracy.

[0028] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention can be realized and obtained through the structures pointed out in the specification and the drawings. Description of the Drawings

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0030] Figure 1 Flow chart of the construction method of the soft sensor model for the initial boiling point of heavy naphtha provided by the embodiment of the present invention;

[0031] Figure 2 Flow chart of the production process of heavy naphtha in the hydrocracking unit provided by the embodiment of the present invention;

[0032] Figure 3 Flow block diagram of the construction method of the soft sensor model for the initial boiling point of heavy naphtha provided by the embodiment of the present invention;

[0033] Figure 4 For Figure 1 Detailed flow chart of step S104 in the illustrated embodiment;

[0034] Figure 5 Flow chart of the application method of the soft sensor model for the initial boiling point of heavy naphtha provided by the embodiment of the present invention;

[0035] Figure 6 Structure diagram of the construction device of the soft sensor model for the initial boiling point of heavy naphtha provided by the embodiment of the present invention;

[0036] Figure 7 Structure diagram of the measuring device for the initial boiling point of heavy naphtha provided by the embodiment of the present invention;

[0037] Figure 8 Hardware structure diagram of the electronic device provided by the embodiment of the present invention. Specific embodiments

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0039] Naphtha is one of the petroleum products, also known as chemical light oil and crude gasoline. It is produced by processing other raw materials and is mainly used as a chemical raw material. Naphtha can be classified into light naphtha, heavy naphtha, and solvent naphtha according to its properties. With the continuous development of the petrochemical industry in recent years and the increasing demand for downstream chemical raw materials such as ethylene, the output of naphtha needs to be further increased. Predicting the initial boiling point of heavy naphtha is of great significance for increasing production and achieving product quality control.

[0040] The auxiliary variable features extracted by the traditional soft sensor model for the initial boiling point of heavy naphtha are entangled with each other and have strong coupling, resulting in the problem of being difficult to adapt to complex and changeable operating conditions. There are mainly two categories of its soft sensor modeling methods: process mechanism analysis and data-driven soft sensor models. Among them, process mechanism mainly uses principles such as material balance, energy balance, and chemical reaction kinetics. Through the mechanism analysis of the process object, the physical connection between the unmeasurable main variable and the measurable auxiliary variable is found, so as to realize the soft measurement of a certain parameter. For a process with a relatively clear process mechanism, this method can construct a soft sensor model with good performance. However, for a complex industrial process with insufficient mechanism research and not fully understood, it is difficult to establish a suitable mechanism model. The data-driven soft sensor model automatically learns and constructs a model by analyzing historical data, and has stronger flexibility and adaptability. Because it can be applied to highly nonlinear and severely uncertain systems, it provides an effective way to solve the soft measurement problem of process parameters in complex systems. Common data-driven soft sensor models include multiple linear regression, multi-layer perceptron, convolutional neural network, etc. However, in actual industrial processes, the relationship between different variables may be complex and nonlinear, and the data-driven soft sensor model may not be able to accurately capture this complex relationship, resulting in poor performance of the soft sensor model. In summary, the features extracted by the traditional soft sensor model do not fully consider the dynamic interaction relationship between auxiliary variables, resulting in poor robustness, weak generalization ability, and insufficient prediction accuracy of the model.

[0041] In view of the above technical problems, the technical concept of the present invention is as follows: Considering that the refining industry data is a typical time series data with periodicity and trend, and at the same time, each auxiliary variable will affect each other during the production process. Therefore, the embodiments of the present invention mainly start from two aspects: the extraction of periodic trend features and the construction of the interaction relationship between auxiliary variables, and propose a soft sensor model based on trend-cycle double spatio-temporal features, that is, first decompose the refining industry data into trend and periodic signals based on the variational mode decomposition method, and extract trend features and periodic features respectively; then construct a double-space feature encoder based on GAT to learn the dynamic interaction relationship of auxiliary variables and extract the double spatio-temporal features of trend and cycle; finally, learn the dynamic mapping relationship between the double spatio-temporal features and the initial boiling point of heavy naphtha to establish a soft sensor model.

[0042] The present invention provides a method, apparatus and application for constructing a soft sensor model of the initial boiling point of heavy naphtha, including a method for constructing a soft sensor model of the initial boiling point of heavy naphtha, an application method of the soft sensor model of the initial boiling point of heavy naphtha, a device for constructing the soft sensor model of the initial boiling point of heavy naphtha, an application device of the soft sensor model of the initial boiling point of heavy naphtha, an electronic device and a computer-readable storage medium.

[0043] Explanation of the terms involved in the present invention:

[0044] Variational Mode Decomposition (VMD for short): It is a signal decomposition and estimation method. In the process of obtaining decomposition components, this method determines the frequency center and bandwidth of each component by iteratively searching for the optimal solution of the variational model, so as to adaptively achieve the frequency-domain dissection of the signal and the effective separation of each component.

[0045] Graph attention networks (GAT for short): A new type of neural network architecture based on graph-structured data, which uses hidden self-attention layers to solve the deficiencies of previous methods based on graph convolution or its approximations. By stacking layers, nodes can participate in the features of neighbors, and different weights can be (implicitly) assigned to different nodes in the neighborhood without any costly matrix operations (such as inversion) and without prior knowledge of the graph structure.

[0046] Long Short-Term Memory (LSTM) is a type of recurrent neural network designed specifically to address the long-term dependence problem existing in general recurrent neural networks (RNNs). All RNNs have a chained form of repeating neural network modules. In a standard RNN, this repeating structural module has a very simple structure, such as a tanh layer.

[0047] Soft sensor: It organically combines the knowledge of the production process and applies computer technology to important variables that are difficult to measure or temporarily cannot be measured. By selecting other variables that are easy to measure and establishing a certain mathematical relationship to infer or estimate, it replaces the function of hardware with software. Applying soft sensor technology to achieve on-line detection of element component content is not only economical and reliable, but also has a rapid dynamic response, can continuously give the element component content during the extraction process, and is easy to achieve the control of product quality.

[0048] Figure 1The figure is a schematic flow chart of a method for constructing a soft sensor model for the initial boiling point of heavy naphtha provided by an embodiment of the present invention. The execution subject is a device for constructing a soft sensor model for the initial boiling point of heavy naphtha, or an electronic device deployed with a device for constructing a soft sensor model for the initial boiling point of heavy naphtha. As Figure 1 shown, the method for constructing a soft sensor model for the initial boiling point of heavy naphtha includes:

[0049] Step S101, obtain auxiliary variable information related to the initial boiling point of heavy naphtha.

[0050] Figure 2 The figure is a production flow chart of heavy naphtha in a hydrocracking unit provided by an embodiment of the present invention. Multiple sensors can be installed on the hydrocracking unit to collect auxiliary variable information related to the initial boiling point of heavy naphtha, and the auxiliary variable information is based on time series data.

[0051] In some embodiments, the auxiliary variables include at least one of the following: the top temperature of the petroleum naphtha fractionation tower, the top pressure, the heat flow rate of the reboiler at the bottom of the tower, the bypass flow rate, the bottom temperature, and the bottom outlet temperature.

[0052] For example, the auxiliary variables are shown in Table 1, where the number is the tag number of the corresponding sensor for the auxiliary variable.

[0053] Table 1

[0054]

[0055]

[0056] Step S102, decompose the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method.

[0057] Among them, the trend signal is the change trend of raw material input or product output data in the chemical process, and the periodic signal is a signal whose amplitude changes repeatedly with time. In this step, the variational mode decomposition method is used to decompose the auxiliary variable information into a trend signal and a periodic signal, and the expression is as follows:

[0058] [x tp , x tt = VMD(x t ) (1)

[0059] Among them, x t = (x t1 , x t2 , x t3 , x t4 , x t5 , x t6 ), represents the auxiliary variable information (raw data) obtained in step S101, x t1is the top tower temperature, x t2 is the top tower pressure, x t3 is the heat flow rate of the reboiler at the bottom of the tower, x t4 is the side stream flow rate, x t5 is the bottom tower temperature, x t6 is the outlet temperature at the bottom of the tower, x tt represents the trend signal, x tp represents the periodic signal.

[0060] Step S103: Extract the features of the trend signal and the periodic signal respectively to obtain the corresponding trend features and periodic features.

[0061] In some embodiments, step S103 includes: extracting the features of the trend signal and the periodic signal respectively based on the long short-term memory network model. Specifically, the trend signal and the periodic signal are respectively fed into the long short-term memory unit for feature extraction to obtain the trend features and the periodic features. The expressions are as follows:

[0062] h tp = LSTM(x tp ) (2)

[0063] h tt = LSTM(x tt ) (3)

[0064] where, h tp represents the trend features, h tt represents the periodic features. Further, the implementation details of LSTM are as follows:

[0065]

[0066]

[0067]

[0068]

[0069]

[0070]

[0071] where, t represents the time, represents the input value of the LSTM at the current time, h t-1 represents the output value of the LSTM at the previous time, i t represents the input gate of the long short-term memory unit, W i represents the weight matrix of the input gate, represents concatenating two vectors into a longer vector, b iis the bias term of the forget gate, f t represents the forget gate of the long short-term memory unit, W f denotes the weight matrix of the forget gate, b f is the bias term of the forget gate, c t-1 represents the state of the memory unit at the previous moment, c t represents the state of the memory unit at the current moment, W C denotes the weight matrix of the memory unit, b C is the bias term of the memory unit, c t represents the state of the memory unit at the current moment, g t represents the cell update state of the long short-term memory unit, o t represents the output gate of the long short-term memory unit, W to represents the weight matrix of the input value of the output gate, W ho represents the output value h at the previous moment of the output gate t-1 of the weight matrix, tanh() and σ() represent activation functions, represents the multiplication operation. In this embodiment, represents the input signal of the LSTM, that is, the above-mentioned trend signal or periodic signal. If the input is a periodic signal x tp , then the features extracted by the LSTM are periodic features h tp , if the input is a trend signal x tt , the trend features h are extracted tt .

[0072] Step S104: The spatio-temporal features of the trend feature and the periodic feature are respectively extracted by the dual-space feature encoder based on the graph attention model.

[0073] On the basis of obtaining the trend feature and the periodic feature, the dual-space feature encoder based on GAT further extracts the spatial interaction features of the trend feature and the periodic feature to form dual spatio-temporal interaction features. The expression is as follows:

[0074] h′ tp = GAT(h tp ) (10)

[0075] h′ tt = GAT(h tt ) (11)

[0076] where h′ tp represents the spatio-temporal feature corresponding to the periodic feature, and h′ tt represents the spatio-temporal feature corresponding to the trend feature.

[0077] Step S105: The two spatio-temporal features are concatenated to obtain the joint spatio-temporal feature.

[0078] Specifically, the spatio-temporal features corresponding to the trend and periodic features are concatenated, and the expression is as follows:

[0079]

[0080] where cat is the concatenation function, represents the combined spatio-temporal features after concatenation.

[0081] Step S106: Construct a soft-sensing model for predicting the initial boiling point of heavy naphtha based on the combined spatio-temporal features.

[0082] In some embodiments, step S106 includes: learning the dynamic mapping relationship between the combined spatio-temporal features and the initial boiling point of heavy naphtha based on Transformer to establish the soft-sensing model. Specifically, Transformer learns the dynamic mapping relationship between the dual spatio-temporal features and the initial boiling point of heavy naphtha, and the established soft-sensing model is as follows:.

[0083]

[0084]

[0085] where Q represents the query vector, K represents the key vector of the information to be queried, and V represents the vector of the information to be queried, is the predicted initial boiling point of heavy naphtha. In this embodiment, the self-attention mechanism is used, so Q, K, and V are all set to the combined spatio-temporal features after concatenation

[0086] Figure 3 is the flowchart of a method for constructing a soft-sensing model for the initial boiling point of heavy naphtha provided by an embodiment of the present invention. As Figure 3 shown, the input variable X, that is, the auxiliary variable information (raw data) related to the initial boiling point of heavy naphtha; the raw data is decomposed into a trend signal and a periodic signal based on VMD; the trend features corresponding to the trend signal and the periodic features corresponding to the periodic signal are respectively extracted through LSTM; the spatio-temporal features corresponding to the trend features and the periodic features are extracted by a dual-space feature encoder based on GAT; the two spatio-temporal features are concatenated based on the cat function; finally, Transformer is used to learn the dynamic mapping relationship between the dual spatio-temporal features and the initial boiling point of heavy naphtha to establish a soft-sensing model.

[0087] The construction method of the soft sensor model for the initial boiling point of heavy naphtha provided by the embodiments of the present invention first decomposes the auxiliary variables related to the initial boiling point of heavy naphtha in the refining industry into trend signals and periodic signals based on the variational mode decomposition method, and extracts trend features and periodic features respectively. Then, the dual-space feature encoder constructed based on the graph attention model learns the dynamic interaction relationships of each auxiliary variable, and extracts the spatio-temporal features of the trend features and periodic features. Finally, a soft sensor model for the initial boiling point of heavy naphtha is constructed according to the dual spatio-temporal features of the trend features and periodic features. That is, the soft sensor model constructed by the embodiments of the present invention fully considers the dynamic interaction relationships between auxiliary variables, improves the robustness of the model, enhances the generalization ability, and improves the prediction accuracy.

[0088] On the basis of the foregoing embodiments, Figure 4 For Figure 1 A detailed flowchart of step S104 in the illustrated embodiment is as Figure 4 As shown, step S104 includes:

[0089] Step S1041: Determine the first dynamic interaction relationship value between any two auxiliary variables based on the trend features corresponding to each auxiliary variable, and determine the spatio-temporal features corresponding to the trend features according to the first dynamic interaction relationship value.

[0090] Step S1042: Determine the second dynamic interaction relationship value between any two auxiliary variables based on the periodic features corresponding to each auxiliary variable, and determine the spatio-temporal features corresponding to the periodic features according to the second dynamic interaction relationship value.

[0091] Specifically, the calculation formulas for the first dynamic interaction relationship value and the second dynamic interaction relationship value are as follows:

[0092]

[0093] The calculation formula for the dual spatio-temporal features corresponding to the trend features or periodic features is as follows

[0094]

[0095] Where, e i,j Represents the first dynamic interaction relationship value or the second dynamic interaction relationship value between the i-th auxiliary variable and the j-th auxiliary variable, which can form a dynamic interaction relationship matrix of the auxiliary variables. N(i) represents the set of auxiliary variables, Represent the trend features or periodic features of the i-th, j-th, and k-th auxiliary variables respectively. k∈N(i) means that k takes all the auxiliary variables in the set of auxiliary variables. W represents the mapping weight of the trend features or periodic features of the corresponding auxiliary variables. a T Represents the transpose of the preset weight vector. LeakyReLU represents the activation function. h′ iSpatio-temporal features representing trend features or periodic features, e i,i Represents the first dynamic interaction relationship value or the second dynamic interaction relationship value between the i-th auxiliary variables themselves, and j∈N(i) indicates that j takes values as other auxiliary variables in the auxiliary variable set except the i-th auxiliary variable.

[0096] Among them, W is a learnable weight that is automatically learned and determined according to the data during the training process, a T The specific value of is set according to the experience of those skilled in the art, and the present invention does not limit this.

[0097] On the basis of the foregoing embodiments, the dual-space feature encoder constructed based on GAT learns the dynamic interaction relationships between the trend features of each auxiliary variable and the dynamic interaction relationships between the periodic features of each auxiliary variable, and further calculates the dual spatio-temporal features of trends and periods according to the dynamic interaction relationships, fully considering the dynamic interaction relationships between the auxiliary variables, and solving the problems that the auxiliary variables extracted by the traditional soft sensor model of the initial boiling point of heavy naphtha are entangled with each other and have strong coupling, and it is difficult to adapt to complex and changeable operating conditions.

[0098] Figure 5 This is a schematic flowchart of the application method of the soft sensor model for the initial boiling point of heavy naphtha provided by the embodiments of the present invention. It is applied to the soft sensor model constructed based on the construction method of the soft sensor model for the initial boiling point of heavy naphtha described above. Its execution subject is a measuring device for the initial boiling point of heavy naphtha, or an electronic device deployed with a measuring device for the initial boiling point of heavy naphtha. As Figure 5 shown, the application method of the soft sensor model for the initial boiling point of heavy naphtha includes:

[0099] Step S501, collect the current auxiliary variable information related to the initial boiling point of heavy naphtha.

[0100] Step S502, input the current auxiliary variable information into the soft sensor model, and output the initial boiling point information of heavy naphtha.

[0101] Specifically, when it is necessary to measure the initial boiling point of heavy naphtha, the current auxiliary variable information related to the initial boiling point of heavy naphtha is collected, including the current top temperature, the current top pressure, the heat flow rate of the reboiler at the bottom of the column, the current bypass flow rate, the current bottom temperature, and the current bottom outlet temperature; the obtained current auxiliary variable information is input into the soft sensor model, and the current auxiliary variable information is decomposed into a trend signal and a periodic signal based on the variational mode decomposition method, the characteristics of the trend signal and the periodic signal are respectively extracted to obtain the corresponding trend characteristics and periodic characteristics; the spatio-temporal characteristics of the trend characteristics and the periodic characteristics are respectively extracted by the dual-space feature encoder based on the graph attention model, the two spatio-temporal characteristics are spliced to obtain the joint spatio-temporal characteristics, and the joint spatio-temporal characteristics are input into the trained Transformer model to obtain the information of the initial boiling point of heavy naphtha.

[0102] The application method of the soft sensor model for the initial boiling point of heavy naphtha provided in this embodiment can predict the information of the initial boiling point of heavy naphtha with higher accuracy by inputting the currently collected auxiliary variable information related to the initial boiling point of heavy naphtha into the soft sensor model constructed by using the construction method of the soft sensor model for the initial boiling point of heavy naphtha described above.

[0103] Figure 6 The following is a schematic structural diagram of a construction device for a soft sensor model of the initial boiling point of heavy naphtha provided by an embodiment of the present invention, as Figure 6 shown, the construction device includes:

[0104] An information acquisition module 601, configured to acquire auxiliary variable information related to the initial boiling point of heavy naphtha; a signal decomposition module 602, configured to decompose the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method; a first extraction module 603, configured to respectively extract the characteristics of the trend signal and the periodic signal to obtain the corresponding trend characteristics and periodic characteristics; a second extraction module 604, configured to respectively extract the spatio-temporal characteristics of the trend characteristics and the periodic characteristics by a dual-space feature encoder based on the graph attention model; a feature splicing module 605, configured to splice the two spatio-temporal characteristics to obtain joint spatio-temporal characteristics; a model construction module 605, configured to construct a soft sensor model for predicting the initial boiling point of heavy naphtha based on the joint spatio-temporal characteristics.

[0105] In some embodiments, the auxiliary variables include at least one of the following: the top temperature of the petroleum naphtha fractionation column, the top pressure, the heat flow rate of the reboiler at the bottom of the column, the bypass flow rate, the bottom temperature, and the bottom outlet temperature.

[0106] In some embodiments, the second extraction module 604 is specifically configured to: determine a first dynamic interaction relationship value between any two auxiliary variables based on the trend features corresponding to the auxiliary variables, and determine the spatio-temporal features corresponding to the trend features according to the first dynamic interaction relationship value; determine a second dynamic interaction relationship value between any two auxiliary variables based on the periodic features corresponding to the auxiliary variables, and determine the spatio-temporal features corresponding to the periodic features according to the second dynamic interaction relationship value.

[0107] In some embodiments, the calculation formula for the first dynamic interaction relationship value or the second dynamic relationship value is as follows:

[0108]

[0109] The calculation formula for the double spatio-temporal features corresponding to the trend features or the periodic features is as follows:

[0110]

[0111] where, e i,j represents the dynamic relationship value between the i-th auxiliary variable and the j-th auxiliary variable, N(i) represents the set of auxiliary variables, represents the trend feature or the periodic feature of the i-th auxiliary variable, represents the trend feature or the periodic feature of the j-th auxiliary variable, represents the trend feature or the periodic feature of the K-th auxiliary variable, respectively represent the trend features or the periodic features of the i-th, j-th, and k-th auxiliary variables, k∈N(i) means that k takes all the auxiliary variables in the set of auxiliary variables, W represents the mapping weight of the corresponding trend feature or the periodic feature of the auxiliary variable, a T represents the transpose of the preset weight vector, LeakyReLU represents the activation function, h′ i represents the spatio-temporal features of the trend feature or the periodic feature, e i,i represents the first dynamic interaction relationship value or the second dynamic interaction relationship value between the i-th auxiliary variable itself, and j∈N(i) means that j takes the other auxiliary variables in the set of auxiliary variables except the i-th auxiliary variable.

[0112] In some embodiments, the first extraction module 603 is specifically configured to: respectively extract the features of the trend signal and the periodic signal based on the long short-term memory network model.

[0113] In some embodiments, the model construction module 605 is specifically configured to: learn the dynamic mapping relationship between the joint spatio-temporal features and the initial boiling point of heavy naphtha based on Transformer to establish the soft measurement model.

[0114] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working process and corresponding beneficial effects of the device for constructing the soft measurement model of the initial boiling point of heavy naphtha described above can refer to the corresponding process in the foregoing method examples, and will not be elaborated here.

[0115] Figure 7 FIG. is a schematic structural diagram of a measuring device for the initial boiling point of heavy naphtha provided by an embodiment of the present invention, which is applied to the soft measurement model constructed based on the foregoing method for constructing the soft measurement model of the initial boiling point of heavy naphtha. As Figure 7 shown, the measuring device includes:

[0116] An acquisition module 701, configured to acquire current auxiliary variable information related to the initial boiling point of heavy naphtha; a prediction module 702, configured to input the current auxiliary variable information into the soft measurement model and output the initial boiling point information of heavy naphtha.

[0117] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working process and corresponding beneficial effects of the measuring device for the initial boiling point of heavy naphtha described above can refer to the corresponding process in the foregoing method examples, and will not be elaborated here.

[0118] Figure 8 FIG. is a schematic hardware structure diagram of an electronic device provided by an embodiment of the present invention. As Figure 8 shown, the electronic device includes: a processor 801, a communication interface 802, a memory 803, and a communication bus 804. Among them, the processor 801, the communication interface 802, and the memory 803 communicate with each other through the communication bus 804.

[0119] The memory 803 is used to store a computer program;

[0120] In an embodiment of the present application, when the processor 801 executes the program stored on the memory 803, it implements the steps of the method for constructing the soft measurement model of the initial boiling point of heavy naphtha or the application method of the soft measurement model of the initial boiling point of heavy naphtha provided by any one of the foregoing method embodiments.

[0121] For the electronic device provided by the embodiment of the present application, its implementation principle and technical effects are similar to those of the above embodiment, and will not be elaborated here.

[0122] The above-mentioned memory 803 may be an electronic memory such as a flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. The memory 603 has a storage space for program codes for executing any of the method steps in the above-mentioned method. For example, the storage space for program codes may include respective program codes for implementing each of the steps in the above-mentioned method. These program codes may be read from or written into one or more computer program products. These computer program products include program code carriers such as hard disks, optical discs (CDs), memory cards, or floppy disks. Such computer program products are usually portable or fixed storage units. The storage unit may have a storage segment or storage space, etc., arranged similarly to the memory 803 in the above-mentioned electronic device. The program codes may be compressed in an appropriate form, for example. Generally, the storage unit includes a program for executing the method steps according to the embodiments of the present application, that is, codes that can be read by a processor such as 801, and when these codes are run by the electronic device, they cause the electronic device to execute each of the steps in the method described above.

[0123] Embodiments of the present application also provide a computer-readable storage medium. A computer program is stored on the above-mentioned computer-readable storage medium, and when the computer program is executed by a processor, it implements the steps of the method for constructing a soft measurement model of the initial boiling point of heavy naphtha or the method for applying a soft measurement model of the initial boiling point of heavy naphtha as described above.

[0124] The computer-readable storage medium may be included in the device / apparatus described in the above embodiments; or it may exist separately without being assembled into the device / apparatus. The above-mentioned computer-readable storage medium carries one or more programs, and when the one or more programs are executed, they implement the method according to the embodiments of the present application.

[0125] According to the embodiments of the present application, the computer-readable storage medium may be a non-volatile computer-readable storage medium, and may include, for example, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In the present application, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or combined with an instruction execution system, device, or apparatus.

[0126] It should be noted that in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising said element.

[0127] Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they may still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a soft measurement model for the initial boiling point of heavy naphtha, characterized in that, The method includes: Obtaining auxiliary variable information related to the initial boiling point of heavy naphtha; Decomposing the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method; Extracting the features of the trend signal and the periodic signal respectively to obtain corresponding trend features and periodic features; Extracting the spatio-temporal features of the trend features and the periodic features respectively based on the dual-space feature encoder of the graph attention model; Concatenating the two spatio-temporal features to obtain a joint spatio-temporal feature; Constructing a soft sensor model for predicting the initial boiling point of heavy naphtha based on the joint spatio-temporal feature.

2. The method for constructing a soft sensor model for the initial boiling point of heavy naphtha according to claim 1, wherein The auxiliary variable information includes: the top temperature of the petroleum naphtha fractionation tower, the top pressure, the heat flow rate of the reboiler at the bottom of the tower, the bypass flow rate, the bottom temperature and the bottom outlet temperature.

3. The method for constructing a soft sensor model for the initial boiling point of heavy naphtha according to claim 2, wherein Decomposing the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method includes: Using the variational mode decomposition method to decompose the auxiliary variable information into a trend signal and a periodic signal, and the expression is as follows: [x tp , x tt = VMD(x t ) where x t =(x t1 , x t2 , x t3 , x t4 , x t5 , x t6 ), representing auxiliary variable information, x t1 is the top temperature, x t2 is the top pressure, x t3 is the heat flow rate of the reboiler at the bottom of the column, x t4 is the bypass flow rate, x t5 is the bottom temperature, x t6 is the outlet temperature at the bottom of the column, x tt represents the trend signal, x tp represents the periodic signal; VMD represents variational mode decomposition; The trend signal is used to represent the change trend of the raw material input or product output data in the chemical process, and the periodic signal is used to represent the signal whose amplitude changes repeatedly with time.

4. The method for constructing a soft sensor model for the initial boiling point of heavy naphtha according to claim 1, wherein Extracting the features of the trend signal and the periodic signal includes: Extracting the features of the trend signal and the periodic signal respectively based on the long short-term memory network model.

5. The method for constructing a soft sensor model for the initial boiling point of heavy naphtha according to claim 1, wherein Extracting the spatio-temporal features of the trend features and the periodic features respectively based on the dual-space feature encoder of the graph attention model includes: Determining the first dynamic interaction relationship value between any two auxiliary variables based on the trend features corresponding to each auxiliary variable, and determining the spatio-temporal feature corresponding to the trend features according to the first dynamic interaction relationship value; Determining the second dynamic interaction relationship value between any two auxiliary variables based on the periodic features corresponding to each auxiliary variable, and determining the spatio-temporal feature corresponding to the periodic features according to the second dynamic interaction relationship value.

6. The method for constructing a soft sensor model for the initial boiling point of heavy naphtha according to claim 5, wherein The calculation formula of the first dynamic interaction relationship value or the second dynamic relationship value is as follows: The calculation formula of the spatio-temporal feature corresponding to the trend feature or the periodic feature is as follows: where, e i，j represents the first dynamic interaction relationship value or the second dynamic interaction relationship value between the i-th auxiliary variable and the j-th auxiliary variable, e i，i represents the first dynamic interaction relationship value or the second dynamic interaction relationship value between the i-th auxiliary variable itself, exp represents the exponential with the natural constant e as the base, N(i) represents the set of auxiliary variables, respectively represent the trend feature or the periodic feature of the i-th, j-th, and k-th auxiliary variables, k ∈ N(i) means that k takes values as all auxiliary variables in the set of auxiliary variables, W represents the mapping weight of the trend feature or the periodic feature of the corresponding auxiliary variable, a T represents the transpose of the preset weight vector, LeakyReLU represents the activation function, h i represents the spatio-temporal feature of the trend feature or the periodic feature, j ∈ N(i) means that j takes values as other auxiliary variables in the set of auxiliary variables except the i-th auxiliary variable.

7. The method for constructing a soft sensor model for the initial boiling point of heavy naphtha according to any one of claims 1-6, wherein Constructing a soft sensor model for predicting the initial boiling point of heavy naphtha based on the joint spatio-temporal feature includes: Learning the dynamic mapping relationship between the joint spatio-temporal feature and the initial boiling point of heavy naphtha based on Transformer to establish the soft sensor model.

8. Application method of soft sensor model for initial boiling point of heavy naphtha, characterized in that, The soft sensor model constructed by the construction method of the soft sensor model for the initial boiling point of heavy naphtha according to any one of claims 1-7, which is used to measure the initial boiling point of heavy naphtha, includes: Collect the current auxiliary variable information related to the initial boiling point of heavy naphtha; Input the current auxiliary variable information into the soft sensor model, and output the initial boiling point information of heavy naphtha.

9. Device for constructing soft measurement model of initial boiling point of heavy naphtha, characterized in that, It includes: An information acquisition module, which is used to acquire the auxiliary variable information related to the initial boiling point of heavy naphtha; A signal decomposition module, which is used to decompose the auxiliary variable information into a trend signal and a periodic signal based on the variational mode decomposition method; A first extraction module, which is used to extract the features of the trend signal and the periodic signal respectively, and obtain the corresponding trend features and periodic features; A second extraction module, which is used to extract the spatio-temporal features of the trend features and the periodic features respectively based on the dual-space feature encoder of the graph attention model; A feature splicing module, which is used to splice the two spatio-temporal features to obtain a joint spatio-temporal feature; A model construction module, which is used to construct a soft sensor model for predicting the initial boiling point of heavy naphtha based on the joint spatio-temporal feature.

10. Application device of soft measurement model for initial boiling point of heavy naphtha, characterized in that, The soft sensor model constructed by the construction method of the soft sensor model for the initial boiling point of heavy naphtha according to any one of claims 1-7, which is used to measure the initial boiling point of heavy naphtha, includes: A collection module, which is used to collect the current auxiliary variable information related to the initial boiling point of heavy naphtha; A prediction module, which is used to input the current auxiliary variable information into the soft sensor model, and output the initial boiling point information of heavy naphtha.

11. An electronic device, characterized in that, It includes a processor, a communication interface, a memory and a communication bus. Among them, the processor, the communication interface and the memory complete the communication with each other through the communication bus; The memory stores a computer program; The processor is used to implement the construction method of the soft sensor model for the initial boiling point of heavy naphtha according to any one of claims 1-7 or the application method of the soft sensor model for the initial boiling point of heavy naphtha according to claim 8 when executing the computer program stored on the memory.

12. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the construction method of the soft sensor model for the initial boiling point of heavy naphtha according to any one of claims 1-7 or the application method of the soft sensor model for the initial boiling point of heavy naphtha according to claim 8.

Images