Construction method and control device of regression prediction model in oil refining and chemical process

Through the European distance screening and nonlinear dimensionality reduction algorithm processing, the problem of uneven sample distribution in the construction of small and medium-sized data set regression prediction model in petrochemical industry is solved, and a regression prediction model with better generalization ability is constructed, avoiding overfitting.

CN119004934APending Publication Date: 2024-11-22CHINA PETROLEUM & CHEMICAL CORP +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202310559206.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-17
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

In the petrochemical field, when the random division method is used to construct regression prediction models of small data sets, it often leads to uneven distribution of samples in the training set and test set, resulting in overfitting of the regression prediction models.

Method used

By calculating the Euclidean distance between every two sample vectors in the dataset matrix, filtering out the samples and dividing them into training sets and test sets, ensuring that the samples are distributed evenly, and processing the dataset with a nonlinear dimensionality reduction algorithm to construct Regression prediction model with better generalization ability.

Benefits of technology

A uniform division of the training sample set and test set samples of small data sets is achieved, and a regression prediction model with better generalization ability is constructed, avoiding overfitting.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004934A_ABST
    Figure CN119004934A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a construction method and a control device for a regression prediction model in oil refining and chemical engineering processes, and belongs to the technical field of machine learning modeling in petrochemical engineering processes. The method comprises the steps of determining a corresponding machine learning model according to a to-be-predicted process task; acquiring a data set comprising m feature parameters and one continuous target variable; dividing a training sample set and a test sample set according to the acquired data set; and training the machine learning model through the training sample set, and verifying the machine learning model through the test sample set to obtain a regression prediction model of the to-be-predicted process task. According to the embodiment of the invention, uniform division of training sample set and test set samples can be carried out on a small data set comprising a plurality of feature parameters so as to construct a regression prediction model with good generalization ability.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning modeling in petrochemical processes, and in particular to a method for constructing a regression prediction model in an oil refining and chemical process and a control device. Background Art

[0002] In the process of petroleum refining, organic synthesis, industrial catalysis and catalyst preparation in the field of petrochemicals, machine learning can be used to predict parameters such as product yield and product properties in the process. The regression model used to predict continuous variables is an important task in machine learning. According to the process of machine learning modeling, the data set must first be divided. For example, the data set is divided into a training set, a validation set and a test set in a ratio of 8:1:1. The model is trained on the training set, the hyperparameters of the model are optimized through the validation set, and the generalization ability of the model is evaluated on the test set. For learning tasks with a small number of samples, the data set can be divided into a training set and a test set in a ratio of 3:1. The hyperparameters of the model are optimized through cross-validation calculations on the training set, thereby replacing the role of the validation set. According to the assumptions of the machine learning model, the samples of the training set and the test set must meet the conditions of uniform distribution, so that the constructed model has sufficient extrapolation and generalization capabilities. For the regression model task of petrochemical industry, there are many categories of feature data, but the value range of each feature is narrow, and usually only dozens of valid samples can be collected for modeling. For such a small data set, the division of training set and test set samples directly determines the quality of the model.

[0003] At present, for the construction of regression prediction models, the data set is generally divided by random partitioning. Different data set partitioning results can be obtained by disrupting the sample order and changing the setting of random number seeds. This method is suitable for large data sets. For small data sets, the random partitioning method often leads to uneven distribution of training set and test set samples, causing overfitting problems in the regression prediction model. Summary of the invention

[0004] The purpose of an embodiment of the present invention is to provide a method for constructing a regression prediction model in oil refining and chemical processes. The method for constructing a regression prediction model in oil refining and chemical processes is used to solve the problem that random division of small data sets often leads to uneven distribution of training set and test set samples, thereby causing overfitting of the regression prediction model.

[0005] In order to achieve the above-mentioned purpose, an embodiment of the present invention provides a method for constructing a regression prediction model in an oil refining and chemical process, and the method for constructing a regression prediction model in an oil refining and chemical process comprises: determining a corresponding machine learning model according to a process task to be predicted; obtaining a data set including m feature parameters and 1 continuous target variable, wherein the m feature parameters are associated with the input value of the machine learning model, and the 1 continuous target variable is associated with the output value of the machine learning model; constructing a corresponding data set matrix according to the acquired data set; taking each row or each column of the data set matrix as a sample vector, respectively calculating the Euclidean distance between each two sample vectors; determining the two sample vectors with the largest Euclidean distance as the first sample data; and second sample data, and add the first sample data and the second sample data to a training sample set; for the remaining sample vectors in the data set matrix, screen out new sample data by calculating the Euclidean distance with each sample data in the current sample training set, and add the new sample data to the training sample set; repeat the above screening of the remaining sample vectors in the data set matrix until the number of sample data in the training sample set reaches a preset number, and use the remaining sample vectors in the data set matrix as sample data of a test sample set; and train the machine learning model through the training sample set, and verify the machine learning model through the test sample set to obtain a regression prediction model for the process task to be predicted.

[0006] Optionally, the method of screening out new sample data by calculating the Euclidean distance with each sample data in the current sample training set includes: obtaining the sample vectors with the smallest Euclidean distance with each sample data in the current sample training set respectively, and determining the sample vector with the largest Euclidean distance as the new sample data.

[0007] Optionally, the machine learning model is constructed using one of the XGBoost algorithm, random forest regression and gradient boosted tree regression GBDT algorithm.

[0008] Optionally, after constructing the corresponding data set matrix, the method for constructing a regression prediction model for oil refining and chemical processes further includes: standardizing the data of the data set matrix.

[0009] Optionally, the method for constructing a regression prediction model for oil refining and chemical processes further includes: performing dimensionality reduction calculations on the standardized data set matrix using a nonlinear dimensionality reduction algorithm.

[0010] Optionally, the nonlinear dimensionality reduction algorithm includes one of principal local linear embedding LLE, isometric mapping ISOMAP, t-distributed stochastic neighbor embedding t-SNE, consistent manifold approximation and projection UMAP.

[0011] Optionally, each row or column of the data set matrix is ​​taken as a sample vector, the number of samples in the data set matrix is ​​the number n of sample vectors, and the ratio of the number of samples in the training sample set and the test sample set is determined to be 3:1, then the preset number of training sample sets is 3 / 4n.

[0012] An embodiment of the present invention also provides a control device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned method for constructing a regression prediction model for oil refining and chemical processes.

[0013] An embodiment of the present invention further provides a machine-readable storage medium having instructions stored thereon, the instructions enabling a machine to execute the above-mentioned method for constructing a regression prediction model in oil refining and chemical processes.

[0014] An embodiment of the present invention also provides a device, which includes at least one processor, and at least one memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the above-mentioned method for constructing a regression prediction model in oil refining and chemical processes.

[0015] Through the above technical solution, the embodiment of the present invention can evenly divide the training sample set and the test set sample for a small data set including multiple feature parameters to build a regression prediction model with good generalization ability. The embodiment of the present invention can treat the regression prediction model including multiple feature parameters as a continuous manifold embedded in a high-dimensional space, merge the feature parameters and the continuous target variables to build a complete data set matrix, perform nonlinear dimensionality reduction calculations on the spatial morphology of the data set matrix to restore the manifold of the machine learning model in its intrinsic dimension, and then combine the calculation of the Euclidean distance between each sample on the low-dimensional manifold to achieve uniform sampling of samples. Based on this, the data set is evenly divided into a training set and a test set to build a regression prediction model with better generalization ability.

[0016] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0018] Figure 1It is a flow chart of a method for constructing a regression prediction model in oil refining and chemical processes provided by an embodiment of the present invention;

[0019] Figure 2 is a schematic diagram of a process of dividing a data set provided by an embodiment of the present invention; and

[0020] Figure 3 It is a schematic diagram of the structure of a device provided in an embodiment of the present invention.

[0021] Description of Reference Numerals

[0022] 30 device 301 processor

[0023] 302 memory 303 bus DETAILED DESCRIPTION

[0024] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0025] Figure 1 This is a flow chart of a method for constructing a regression prediction model in an oil refining and chemical process provided by an embodiment of the present invention. Please refer to Figure 1 , the method for constructing a regression prediction model in oil refining and chemical processes may include the following steps:

[0026] Step S110: Determine the corresponding machine learning model according to the process task to be predicted.

[0027] The process task of prediction refers to the use of machine learning to predict parameters such as product yield and product properties during the processes of petroleum refining, organic synthesis, industrial catalysis and catalyst preparation in the petrochemical industry.

[0028] Preferably, the machine learning model can be constructed using one of the XGBoost algorithm, random forest regression, and gradient boosted tree regression GBDT algorithms.

[0029] Step S120: Obtain a data set including m feature parameters and 1 continuous target variable, wherein the m feature parameters are associated with the input values ​​of the machine learning model, and the 1 continuous target variable is associated with the output value of the machine learning model.

[0030] Among them, the continuous target variable can be, for example, the predicted value of the process task to be predicted, and the selection of characteristic parameters is associated with the process task to be predicted and the continuous target variable. For example, the predicted process task is the prediction of the particle size distribution properties of the catalyst particles during the pilot preparation of the FCC catalyst, then one continuous target variable can be the particle size distribution of the catalyst particles, and the characteristic parameters can include the composition and properties of the catalyst, such as the mass fraction of kaolin, the mass fraction of pseudo-boehmite, the mass fraction of aluminum sol, the mass fraction of molecular sieves, solid content, acid-aluminum mass ratio, and can also include the process conditions of spray drying, such as the inlet temperature of the drying tower, the outlet temperature, the feed pump speed, and the relative speed of the atomizing wheel. For the predicted process task, it is necessary to obtain the data of the above m characteristic parameters, m=10.

[0031] Step S130: Divide the acquired data set into a training sample set and a test sample set.

[0032] As described above, the embodiment of the present invention evenly divides a small data set including multiple feature parameters into a training sample set and a test sample set to construct a regression prediction model with good generalization ability. The number of samples in the small data set is, for example, less than 100, and the feature parameter of each sample is, for example, greater than 4.

[0033] Please refer to Figure 1 and Figure 2 , step S130 may include the following steps:

[0034] Step S131: construct a corresponding data set matrix according to the acquired data set.

[0035] For example, the acquired data set includes n samples, each sample contains m feature parameters and 1 continuous target variable, then the data set matrix XY can be expressed as (x ij ), where i = 1 to n, j = 1 to (m+1). Where m feature parameters and a continuous target variable indicate the columns of the data set matrix (i.e., the matrix has n rows and m+1 columns), or m feature parameters and a continuous target variable indicate the rows of the data set matrix. Then XY can be expressed as (x ij ), where i=1~(m+1), j=1~n. In the following examples, m feature parameters and a continuous target variable are used to indicate the columns of the data set matrix, and the explanation of the embodiments of the present invention will not be repeated hereafter.

[0036] Preferably, after step S131, the method for constructing a regression prediction model in oil refining and chemical processes may further include: standardizing the data of the data set matrix.

[0037] Taking an example, the data matrix XY is subjected to data standardization to eliminate the influence of the dimensions and numerical values ​​of different feature parameters and continuous target variables on the samples. For example, for each data of each feature parameter in each column of the data matrix XY, standardization can be performed by subtracting the mean value of the column and then dividing it by the standard deviation.

[0038] Step S132: taking each row or each column of the data set matrix as a sample vector, respectively calculating the Euclidean distance between every two sample vectors.

[0039] The Euclidean metric, also known as the Euclidean distance, refers to the true distance between two points in an m-dimensional space, or the natural length of a vector (i.e., the distance from the point to the origin). In the embodiment of the present invention, each row of the data set matrix is ​​preferably used as a sample vector, and the data set is divided by the Euclidean distance between the sample vectors.

[0040] Preferably, the method for constructing a regression prediction model for oil refining and chemical processes may further include: performing dimensionality reduction calculation on the standardized data set matrix by using a nonlinear dimensionality reduction algorithm.

[0041] The nonlinear dimensionality reduction algorithm preferably used in the embodiment of the present invention includes one of principal local linear embedding LLE, isometric mapping ISOMAP, t-distributed stochastic neighbor embedding t-SNE, consistent manifold approximation and projection UMAP.

[0042] The oil refining and chemical tasks to be predicted in the embodiment of the present invention are processes such as petroleum refining, organic synthesis, industrial catalysis and catalyst preparation for petrochemical industry. For such machine learning tasks, there are many categories of feature data, but the value range of each feature is narrow, and usually only dozens of valid samples can be collected for modeling. In high-dimensional space (multi-feature parameters), the Euclidean distance may fail, affecting the division of the data set. Therefore, the embodiment of the present invention preferably performs dimensionality reduction calculation on the data set matrix.

[0043] By way of example, by using one of the above nonlinear dimensionality reduction algorithms, the data set matrix XY is reduced in dimensionality, so that the number of columns of the data set matrix XY is reduced from m+1 to d (d is determined by the selected nonlinear dimensionality reduction algorithm), and a new data set matrix XY-R can be obtained, the size of which is n rows and d columns. Each row of the data set matrix XY-R is used as a sample vector, and the Euclidean distance between each two sample vectors is calculated.

[0044] Step S133: determine the two sample vectors with the largest Euclidean distance as the first sample data and the second sample data, and add the first sample data and the second sample data to the training sample set.

[0045] Continuing with the above example, for the data set matrix XY-R, after calculating the Euclidean distance between every two sample vectors, the identifiers (ID) of the two sample vectors with the largest Euclidean distance are determined, and the first sample data S1 and the second sample data S2 are determined corresponding to the sample vectors of the data set matrix XY.

[0046] Step S134: for the remaining sample vectors in the data set matrix, new sample data are screened out by calculating the Euclidean distance with each sample data in the current sample training set, and the new sample data are added to the training sample set.

[0047] Preferably, step S134 may include: respectively obtaining sample vectors with the smallest Euclidean distance to each sample data in the current sample training set, and determining the sample vector with the largest Euclidean distance as the new sample data.

[0048] Continuing with the above example, for the remaining n-2 sample vectors of the data set matrix XY-R, find the sample vectors with the smallest Euclidean distance to the first sample data S1 and the second sample data S2 (for example, the sample vector with the smallest Euclidean distance Z1 to S1 is S 13 , the sample vector with the smallest Euclidean distance Z2 from S2 is S 23 ), and the sample vector S with the larger Euclidean distance (for example, Z1 is greater than Z2) is 13 The sample vector S3 corresponding to the data set matrix XY is used as the third sample data S3 and added to the training sample set. Next, from the remaining n-3 sample vectors of the data set matrix XY-R, the sample vectors (e.g., S1, S2, S3) with the smallest Euclidean distance to the selected sample data (e.g., S1, S2, S3) are found. 14 , S 24 , S 34 ), the sample vector with the larger Euclidean distance corresponding to the sample vector S4 of the data set matrix XY is taken as the new sample data.

[0049] Step S135: Repeat the above screening of the remaining sample vectors in the data set matrix until the number of sample data in the training sample set reaches a preset number, and use the remaining sample vectors in the data set matrix as sample data for the test sample set.

[0050] Following the above example, step S134 is repeated until the number of sample data in the training sample set reaches a preset number n. train When the remaining nn in the data set matrix XY is train The sample vectors are used as the sample data of the test sample set.

[0051] Preferably, each row or column of the data set matrix is ​​taken as a sample vector, then the number of samples in the data set matrix is ​​the number n of sample vectors, and the ratio of the number of samples in the training sample set and the test sample set is determined to be 3:1, then the preset number of training sample sets is 3 / 4n.

[0052] Step S140: training the machine learning model using the training sample set, and verifying the machine learning model using the test sample set, so as to obtain a regression prediction model for the process task to be predicted.

[0053] Accordingly, the embodiment of the present invention can evenly divide the training sample set and the test set sample for a small data set including multiple feature parameters to construct a regression prediction model with good generalization ability. The embodiment of the present invention can treat the regression prediction model including multiple feature parameters as a continuous manifold embedded in a high-dimensional space, merge the feature parameters and the continuous target variables to construct a complete data set matrix, perform nonlinear dimensionality reduction calculations on the spatial morphology of the data set matrix to restore the manifold of the machine learning model in its intrinsic dimension, and then combine the calculation of the Euclidean distance between each sample on the low-dimensional manifold to achieve uniform sampling of samples. Accordingly, the data set is evenly divided into a training set and a test set to construct a regression prediction model with better generalization ability.

[0054] Next, the embodiments of the present invention are further explained through application examples. For example, 37 sets of valid data were collected for the pilot preparation process of FCC catalysts. The characteristic parameters of the data set are as described above, for example, including the composition and properties of the catalyst, such as kaolin mass fraction, pseudo-boehmite mass fraction, aluminum sol mass fraction, molecular sieve mass fraction, solid content, acid-aluminum mass ratio, and spray drying process conditions, such as the inlet temperature, outlet temperature, feed pump speed, and relative speed of the atomizing wheel of the drying tower, a total of 10 characteristic parameters, and 1 continuous target variable is the volume fraction V (0-149) of the catalyst particle size in the range of 0-149 μm. This data set is used as the research object.

[0055] Example 1: 1) Divide the above data set into a training set and a test set in a ratio of 3:1, and the preset number of training sample sets is 28. 2) According to the acquired data set, construct the corresponding data set matrix, merge the feature parameters and the continuous target variable to obtain the data set matrix. 3) The data of the data set matrix is ​​standardized, and then the data set matrix is ​​reduced in dimension by, for example, the ISOMAP algorithm, and the number of columns (d=9) of the reduced-dimensional data set matrix XY is determined by minimizing the reconstruction error. 4) The Euclidean distance between sample vectors is calculated on the reduced-dimensional data set matrix XY-R, and the two samples with the largest Euclidean distance (corresponding to the samples of the data set matrix XY) are first selected as the first sample data and the second sample data, and then the samples closest to the selected samples (with the smallest Euclidean distance) and with the largest Euclidean distance are added to the training sample set one by one, until the number of samples in the training sample set reaches the set requirement, that is, the sample data of the training sample set reaches 28, and the remaining 9 samples of the data set matrix XY are used as the sample data of the test sample set.

[0056] 5) Using the XGBoost algorithm to build a machine learning model, the machine learning model is trained based on the above training sample set, and the machine learning model is verified by the test sample set to obtain a regression prediction model of the volume fraction V (0-149).

[0057] 6) The above data sets were divided by random partitioning, Kennard-Stone algorithm and SPXY algorithm respectively, and XGBoost models for predicting volume fraction V (0-149) were constructed on the corresponding training sets in turn. The goodness of fit R of the training sample set and the test sample set were compared. 2 The partitioning effect of the data set is determined by the prediction standard deviation RMSE, as shown in Table 1.

[0058] Table 1 Model calculation results of Example 1 and corresponding comparative examples

[0059]

[0060] As can be seen from Table 1, for different data set division methods, the goodness of fit R2 of the training sample set is greater than 0.91, and the prediction standard deviation RMSE is also relatively small; for the test sample set, the model goodness of fit R2 obtained by the other three methods is very small, indicating that the prediction deviation is large and the model has serious overfitting, while the goodness of fit R2 of the regression prediction model obtained according to the embodiment of the present invention can reach 0.818, and the RMSE value is also the smallest, with good generalization ability. The above results show that the embodiment of the present invention can evenly divide the training sample set and the test set samples for a small data set including multiple feature parameters to construct a regression prediction model with good generalization ability.

[0061] Example 2: 1) 1) Divide the above data set into training set and test set in a ratio of 3:1, then the preset number of training sample sets is 28. 2) According to the acquired data set, construct the corresponding data set matrix, merge the feature parameters and the continuous target variable to obtain the data set matrix. 3) The data of the data set matrix is ​​standardized, and then the sample matrix is ​​reduced in dimension by the LLE algorithm, and the number of columns (d=5) of the reduced-dimensional data set matrix XY is determined by minimizing the reconstruction error; 4) The Euclidean distance between sample vectors is calculated on the reduced-dimensional data set matrix XY-R, and the two samples with the largest Euclidean distance (corresponding to the samples of the data set matrix XY) are first selected as the first sample data and the second sample data, and then the samples closest to the selected samples (with the smallest Euclidean distance) and with the largest Euclidean distance are added to the training sample set one by one, until the number of samples in the training sample set reaches the set requirement, that is, the sample data of the training sample set reaches 28, and the remaining 9 samples of the data set matrix XY are used as the sample data of the test sample set.

[0062] 5) Using the GBDT algorithm to build a machine learning model, the machine learning model is trained based on the above training sample set, and the machine learning model is verified by the test sample set to obtain a regression prediction model of the volume fraction V (0-149).

[0063] 6) The above data sets were divided by random partitioning, Kennard-Stone algorithm and SPXY algorithm respectively, and XGBoost models for predicting volume fraction V (0-149) were constructed on the corresponding training sets in turn. The goodness of fit R of the training sample set and the test sample set were compared. 2 The partitioning effect of the data set is determined by the prediction standard deviation RMSE, as shown in Table 2.

[0064] Table 2 Model calculation results of Example 2 and corresponding comparative examples

[0065]

[0066] As can be seen from Table 2, for different data set division methods, the goodness of fit R2 of the training sample set is greater than 0.98, and the prediction standard deviation RMSE is also relatively small; while for the test set, the model goodness of fit R2 obtained by the other three methods is very small, indicating that the prediction deviation is large and the model is seriously overfitted, while the goodness of fit R2 of the model obtained according to the embodiment of the present invention can reach 0.766, and the RMSE value is also the smallest, with good generalization ability. The above results show that the embodiment of the present invention can evenly divide the training sample set and the test set samples for a small data set including multiple feature parameters to construct a regression prediction model with good generalization ability.

[0067] Through examples 1) and 2), the embodiment of the present invention can compare the goodness of fit R of the training sample set and the test sample set. 2 The prediction standard deviation RMSE is used to determine the partitioning effect of the data set. If the partitioning effect is not good, the data set matrix XY can be reduced in dimension by reselecting the nonlinear dimensionality reduction algorithm, and the data set can be re-partitioned to obtain the training sample set and the test sample set to build a regression prediction model with good generalization ability.

[0068] Through examples 1) and 2), compared with the traditional random partitioning and uniform sampling methods based on high-dimensional feature space, the embodiment of the present invention is more suitable for constructing a regression prediction model for predicting continuous target variables on a small data set with a small number of samples and a large number of features. The embodiment of the present invention is based on the manifold of the regression prediction model, eliminates the influence of high-dimensional features through nonlinear dimensionality reduction, and achieves uniform partitioning of training set and test set samples. The regression prediction model constructed on this basis has better generalization ability and strong practicality.

[0069] An embodiment of the present invention also provides a control device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for constructing a regression prediction model for oil refining and chemical processes described in steps S110-S140.

[0070] An embodiment of the present invention further provides a machine-readable storage medium having instructions stored thereon, the instructions enabling a machine to execute the method for constructing a regression prediction model for oil refining and chemical processes described in steps S110-S140.

[0071] Please refer to Figure 3 , the embodiment of the present invention provides a device, the device includes at least one processor 301, and at least one memory 302 and bus 303 connected to the processor 301; wherein the processor 301 and the memory 302 communicate with each other through the bus 303; the processor 301 is used to call the program instructions in the memory 302 to execute the flow diagram of the method for constructing a regression prediction model in the refining and chemical process described in steps S110-S140. The device 30 of the embodiment of the present invention can be a server, PC, PAD, mobile phone, etc.

[0072] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0073] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes 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 generate 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.

[0074] 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.

[0075] 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 The steps for the functions specified in one or more boxes.

[0076] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0077] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0078] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0079] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0080] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for constructing a regression prediction model in oil refining and chemical processes, It is characterized in that The method for constructing a regression prediction model for oil refining and chemical processes includes: Determine the corresponding machine learning model based on the process task to be predicted; Acquire a data set including m feature parameters and 1 continuous target variable, wherein the m feature parameters are associated with input values ​​of the machine learning model, and the 1 continuous target variable is associated with an output value of the machine learning model; According to the acquired data set, a corresponding data set matrix is ​​constructed; Taking each row or each column of the data set matrix as a sample vector, respectively calculating the Euclidean distance between each two sample vectors; Determine the two sample vectors with the largest Euclidean distance as first sample data and second sample data, and add the first sample data and the second sample data to a training sample set; For the remaining sample vectors in the data set matrix, new sample data are screened out by calculating the Euclidean distance with each sample data in the current sample training set, and the new sample data are added to the training sample set; Repeat the above screening of the remaining sample vectors in the data set matrix until the number of sample data in the training sample set reaches a preset number, and use the remaining sample vectors in the data set matrix as sample data of the test sample set; and The machine learning model is trained by the training sample set, and the machine learning model is verified by the test sample set to obtain a regression prediction model for the process task to be predicted.

2. The method for constructing a regression prediction model for oil refining and chemical processes according to claim 1, It is characterized in that The new sample data is screened out by calculating the Euclidean distance with each sample data in the current sample training set, including: The sample vectors with the smallest Euclidean distance to each sample data in the current sample training set are obtained respectively, and the sample vector with the largest Euclidean distance is determined as the new sample data.

3. The method for constructing a regression prediction model for oil refining and chemical processes according to claim 1, It is characterized in that The machine learning model is constructed using one of the XGBoost algorithm, random forest regression and gradient boosted tree regression (GBDT) algorithms.

4. The method for constructing a regression prediction model for oil refining and chemical processes according to claim 1, It is characterized in that After constructing the corresponding data set matrix, the method for constructing a regression prediction model in oil refining and chemical processes further includes: The data in the data set matrix is ​​standardized.

5. The method for constructing a regression prediction model for oil refining and chemical processes according to claim 4, It is characterized in that The method for constructing a regression prediction model in oil refining and chemical processes also includes: The dimension reduction calculation is performed on the standardized data set matrix through nonlinear dimension reduction algorithm.

6. The method for constructing a regression prediction model for oil refining and chemical processes according to claim 5, It is characterized in that The nonlinear dimensionality reduction algorithm includes one of principal local linear embedding LLE, isometric mapping ISOMAP, t-distributed stochastic neighbor embedding t-SNE, consistent manifold approximation and projection UMAP.

7. The method for constructing a regression prediction model for oil refining and chemical processes according to claim 1, It is characterized in that Each row or column of the data set matrix is ​​taken as a sample vector, and the number of samples of the data set matrix is ​​the number n of the sample vectors. It is determined that the ratio of the number of samples in the training sample set to that in the test sample set is 3:1, and the preset number of the training sample set is 3 / 4n.

8. A control device, It is characterized in that The control device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the computer program to implement the method for constructing a regression prediction model in oil refining and chemical processes according to any one of claims 1 to 7.

9. A machine-readable storage medium, It is characterized in that The machine-readable storage medium stores instructions, which enable the machine to execute the method for constructing a regression prediction model in oil refining and chemical processes according to any one of claims 1-7.

10. A device, It is characterized in that The device includes at least one processor, and at least one memory and a bus connected to the processor; Among them, the processor and memory communicate with each other through the bus; The processor is used to call program instructions in the memory to execute the method for constructing a regression prediction model in the oil refining and chemical process as described in any one of claims 1-7.