Method and device for predicting octane number of gasoline

Abstract

The invention provides a method and a device for predicting the octane number of gasoline. The method comprises the following steps: measuring the near infrared spectrum of a gasoline sample to be measured, carrying out second-order differential treatment on the near infrared spectrum, and intercepting the absorbance of the near infrared spectrum in a predetermined spectral region; inputting the absorbance into a partial least square analysis model to obtain the following information of the gasoline sample to be detected: 1) an initial prediction octane number, a characteristic prediction value and the absorbance as an information combination; 2) combining the near infrared spectrum score, the initial prediction octane number and the characteristic prediction value as an information combination; and 3) a near infrared spectrum score and a canonical variable, both as an information combination; and inputting the information combination into a corresponding octane number support vector machine prediction model to obtain an octane number prediction value of the gasoline sample to be detected. According to the scheme, other information of the gasoline sample to be detected is added besides the near infrared spectrum of the gasoline to participate in information amplification, so that the prediction accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a method and device for predicting gasoline octane number. Background Art

[0002] Gasoline octane number is an indicator of gasoline anti-knock performance. Its size directly affects the performance of automotive gasoline. The measurement of gasoline octane number is of great significance to the production and use of gasoline. At present, the standard method for determining octane number is the octane number machine bench test method. GB / T 5487-2015 stipulates the determination method of gasoline octane number by research method, and GB / T 503-2016 stipulates the determination method of gasoline octane number by motor method. However, in the production control and use of gasoline, the use of the bench method is subject to certain limitations due to the following factors: (1) The test speed is slow. An octane number measurement takes at least more than one hour, and it is impossible to provide timely feedback for production optimization control; (2) The octane number machine is large in size and requires a special laboratory, which is not conducive to on-site analysis and use; (3) The equipment, operation and maintenance costs are high, and continuous maintenance is required during use; (4) The environmental protection is poor, the sample consumption is large, and the test requires about 500ml of gasoline sample, which has volatilization losses and serious pollution.

[0003] Near infrared spectroscopy has the advantages of fast speed, high precision and low cost. In recent years, it has been widely used to determine the octane number of gasoline, which has produced good social and economic benefits. Near infrared spectroscopy is a method of correlating the spectrum of known samples with the composition or properties, using the multivariate calibration method in chemometrics to establish a calibration model, and then predicting the composition or properties of unknown samples based on the model and the spectrum of unknown samples.

[0004] However, existing methods for predicting gasoline octane number based on near-infrared spectroscopy analysis are all based on the correlation between gasoline near-infrared spectrum and octane number. Various different methods are used to establish regression models to predict octane number. The usual prediction standard deviation is above 0.3, and the accuracy is not high enough. Summary of the invention

[0005] The purpose of the embodiment of the present application is to provide a method and device for predicting gasoline octane number. In addition to the gasoline near-infrared spectrum, the present invention also adds other information of the gasoline sample to be tested (such as gasoline composition information, gasoline physical properties, near-infrared spectrum score, typical variables, etc.) to participate in information amplification, and uses the support vector machine method to establish a regression model to predict gasoline octane number, which can improve the prediction accuracy.

[0006] In order to achieve the above-mentioned purpose, the first aspect of the present application provides a method for predicting the octane number of gasoline, the method comprising: measuring the near-infrared spectrum of the gasoline sample to be tested, performing second-order differential processing on the near-infrared spectrum, and intercepting the absorbance of the near-infrared spectrum in a predetermined spectral region; inputting the absorbance of the gasoline sample to be tested into a partial least squares analysis model to obtain the following information of the gasoline sample to be tested: 1) an initial predicted octane number and a characteristic prediction value, the initial predicted octane number and the characteristic prediction value together with the absorbance as an information combination; 2) a near-infrared spectrum score relative to the octane number, an initial predicted octane number and a characteristic prediction value, the three of which are used as an information combination; or 3) a near-infrared spectrum score relative to the octane number and a typical variable, the two of which are used as an information combination, the typical variable reflects the association between the absorbance of the gasoline sample to be tested in the predetermined spectral region and the octane number and the characteristics of the gasoline sample to be tested, and the information combination is input into an octane number support vector machine prediction model corresponding to the information combination to obtain the octane number prediction value of the gasoline sample to be tested.

[0007] In an embodiment of the present application, the partial least squares analysis model is established by: collecting multiple gasoline samples, measuring their near-infrared spectra, and performing second-order differential processing on the near-infrared spectra to intercept the absorbance of the near-infrared spectra within a predetermined spectral region; measuring the octane number and characteristics of the multiple gasoline samples using standard methods; and based on the absorbance, octane number and characteristics of each of the multiple gasoline samples, using the partial least squares method to establish partial least squares analysis models for predicting the octane number and characteristics of the gasoline samples.

[0008] In an embodiment of the present application, the initial predicted octane number and the near-infrared spectrum score relative to the octane number are obtained by utilizing a partial least squares analysis model for predicting the octane number of a gasoline sample; the characteristic prediction value is obtained by utilizing a partial least squares analysis model for predicting the characteristics of a gasoline sample.

[0009] In an embodiment of the present application, the typical variables of the gasoline sample to be tested are determined by: inputting the absorbance of the gasoline sample to be tested into a typical analysis model to obtain the typical variables of the gasoline sample to be tested, wherein the typical analysis model is established by utilizing the absorbance of near-infrared spectra of multiple gasoline samples in the predetermined spectral region and the octane number and characteristics of the multiple gasoline samples determined by a standard method, and in the process of establishing the typical analysis model, the absorbance of the gasoline sample and the octane number and characteristics are combined to calculate their covariance matrix, and then the matrix is ​​transformed to derive eigenvalues ​​to obtain a set of variables associated with the reaction spectrum and the octane number and characteristics, that is, typical variables.

[0010] In the embodiment of the present application, the octane number support vector machine prediction model is established by utilizing the information combination of multiple gasoline samples and the octane numbers of the multiple gasoline samples determined by a standard method.

[0011] In the embodiment of the present application, the predetermined spectral range is 5500-6350 cm -1 and 7800~8800cm -1 .

[0012] In the embodiments of the present application, the characteristics include one or more of the following: vapor pressure, benzene, olefin and aromatic content.

[0013] In the embodiment of the present application, the octane number support vector machine prediction model adopts an exponential radial basis kernel function.

[0014] A second aspect of the present application provides a device for predicting gasoline octane number, the device comprising: a memory; and a processor, the processor being configured to execute the above-mentioned method for predicting gasoline octane number.

[0015] A third aspect of the present application provides a machine-readable storage medium having instructions stored thereon, which, when executed by a processor, configure the processor to execute the above-mentioned method for predicting gasoline octane number.

[0016] A fourth aspect of the present application provides a computer program product, including a computer program, which implements the above-mentioned method for predicting gasoline octane number when executed by a processor.

[0017] Through the above technical scheme, in addition to the gasoline near-infrared spectrum, other information of the gasoline sample to be tested (such as gasoline composition information (for example, benzene, olefin and aromatic content), gasoline physical properties (for example, vapor pressure), near-infrared spectrum score, typical variables, etc.) is added to participate in information amplification, and the support vector machine method is used to establish a regression model to predict the gasoline octane number, which can improve the prediction accuracy.

[0018] Other features and advantages of the embodiments of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present application, but do not constitute a limitation on the embodiments of the present application. In the accompanying drawings:

[0020] Figure 1 A schematic diagram schematically shows a flow chart of a method for predicting gasoline octane number according to an embodiment of the present application; and

[0021] Figure 2The internal structure of the device for predicting gasoline octane number according to an embodiment of the present application is schematically shown. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present application, and are not used to limit the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0023] It should be noted that if the embodiments of the present application involve directional indications (such as up, down, left, right, front, back...), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0024] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present application, the descriptions of "first", "second", etc. are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by this application. The "partial least squares score", "gasoline sample score", and "near infrared spectrum score" mentioned below refer to the same thing and can be used interchangeably. The "near infrared spectrum of the pretreated gasoline sample" and "absorbance of the near infrared spectrum in a predetermined spectral range" mentioned below refer to the same thing and can be used interchangeably.

[0025] Figure 1 FIG. 1 is a flow chart of a method for predicting gasoline octane number in one embodiment. It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0026] In order to achieve the above purpose, Figure 1 As shown, the present application provides a method for predicting gasoline octane number, the method comprising:

[0027] Step 102, measuring the near infrared spectrum of the gasoline sample to be tested, and performing second-order differential processing on the near infrared spectrum to intercept the absorbance of the near infrared spectrum in a predetermined spectral region;

[0028] Step 104, input the absorbance of the gasoline sample to be tested into a partial least squares analysis model to obtain the following information of the gasoline sample to be tested:

[0029] 1) The initial predicted octane number and characteristic predicted value (combining the absorbance of the gasoline sample to be tested with them as an information combination);

[0030] 2) a near infrared spectral score relative to octane number, an initial predicted octane number, and a property prediction value, the three being a combination of information; or

[0031] 3) a near infrared spectrum score and a typical variable relative to the octane number, the two being a combination of information, the typical variable reflecting the correlation between the absorbance of the gasoline sample to be tested in the predetermined spectral region and the octane number and characteristics of the gasoline sample to be tested;

[0032] Step 106: input the information combination into an octane number support vector machine prediction model corresponding to the information combination to obtain a predicted octane number of the gasoline sample to be tested.

[0033] The present invention increases information of the gasoline sample to be tested (such as gasoline composition information (for example, benzene, olefin and aromatic content), gasoline physical properties (for example, vapor pressure), near-infrared spectrum score, typical variables, etc.) in addition to the gasoline near-infrared spectrum to participate in information amplification, and uses the support vector machine method to establish a regression model to predict the gasoline octane number, thereby improving the prediction accuracy.

[0034] In an embodiment of the present application, the partial least squares analysis model is established by: collecting multiple gasoline samples, measuring their near-infrared spectra, and performing second-order differential processing on the near-infrared spectra to intercept the absorbance of the near-infrared spectra within a predetermined spectral region; measuring the octane number and characteristics of the multiple gasoline samples using standard methods; and based on the absorbance, octane number and characteristics of each of the multiple gasoline samples, using the partial least squares method to establish partial least squares analysis models for predicting the octane number and characteristics of the gasoline samples.

[0035] In an embodiment of the present application, the initial predicted octane number and the near-infrared spectrum score relative to the octane number are obtained by utilizing a partial least squares analysis model for predicting the octane number of a gasoline sample; the characteristic prediction value is obtained by utilizing a partial least squares analysis model for predicting the characteristics of a gasoline sample.

[0036] In the above embodiment of the present application, the predetermined spectral range is 5500-6350cm -1 and 7800~8800cm -1 Of course, the present invention is not limited to the 5500-6350cm -1 , 7800~8800cm -1 Spectral area, other spectral areas can also be used according to needs and actual conditions.

[0037] In the above-mentioned embodiment of the present application, determining the typical variables of the gasoline sample to be tested includes: inputting the absorbance of the gasoline sample to be tested into a typical analysis model to obtain the typical variables of the gasoline sample to be tested, wherein the typical analysis model is established by utilizing the absorbance of near-infrared spectra of multiple gasoline samples in the predetermined spectral region and the octane number and characteristics of the multiple gasoline samples determined by a standard method, and in the process of establishing the typical analysis model, the absorbance and octane number and characteristics of the gasoline samples are combined to calculate their covariance matrix, and then the matrix is ​​transformed to derive eigenvalues ​​to obtain a set of variables associated with the reaction spectrum and characteristics, i.e., typical variables.

[0038] In the above embodiments of the present application, the characteristics may include one or more of the following: vapor pressure, benzene, olefin and aromatic content. Of course, the present invention is not limited thereto, and some other characteristics of the gasoline sample to be tested may also be used.

[0039] The following is a detailed description of the above method for predicting gasoline octane number, taking the information combination "Near infrared spectral scores and typical variables relative to octane number" as an example, specifically involving model establishment and use of the established model. The corresponding parts of these descriptions are also applicable to the other two information combinations.

[0040] The method for predicting gasoline octane number provided by the present invention comprises the following steps:

[0041] (1) Collect a certain amount of gasoline sample, measure its near infrared spectrum, and then measure its octane number, vapor pressure, benzene, olefin and aromatic content using standard methods;

[0042] (2) Preprocessing the near infrared spectrum of the gasoline sample, and performing canonical correlation analysis on the near infrared spectrum of the pretreated gasoline sample and the octane number, vapor pressure, benzene, olefin and aromatic content to obtain typical variables;

[0043] (3) Using the partial least squares method, the gasoline octane number prediction analysis model is established to obtain the gasoline sample score;

[0044] (4) The typical variables and partial least squares scores of the gasoline samples are combined and correlated with the octane number obtained by the standard method, and the support vector machine method is used to establish an analytical model for predicting the octane number of the gasoline samples to be tested.

[0045] (5) When predicting the octane number of the gasoline sample to be tested, first measure the near-infrared spectrum of the gasoline sample to be tested and perform preprocessing, then use the typical analysis model established in step (2) to obtain the typical variables of the gasoline sample to be tested, use the partial least squares analysis model established in step (3) to obtain the partial least squares score of the gasoline sample to be tested, combine the two, and substitute them into the octane number support vector machine prediction model in step (4) to obtain the predicted octane number of the gasoline sample to be tested.

[0046] The invention combines typical variables of near infrared spectrum of gasoline samples with partial least square score information and correlates them with octane number, and then uses support vector machine method to establish octane number prediction model to predict the octane number of the sample to be tested.

[0047] Specifically, the following steps may be included:

[0048] (1) Collect gasoline samples, the number of samples is at least 100, measure the near infrared spectrum of each sample, and then measure its octane number, vapor pressure, benzene, olefin and aromatic content using standard methods;

[0049] (2) Perform second-order differential processing on the near-infrared spectrum of the gasoline sample and intercept the 5500-6350 cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0050] (3) The near-infrared spectral absorbance and octane number, vapor pressure, benzene, olefin and aromatic content of the pretreated gasoline sample are subjected to canonical correlation analysis to obtain typical variables and a typical analysis model; the typical variable refers to a set of variables associated with the reaction spectrum and characteristics obtained by calculating the characteristic values ​​by calculating the near-infrared spectral absorbance and octane number, vapor pressure, benzene, olefin and aromatic content through canonical correlation calculation, that is, covariance matrix transformation.

[0051] Canonical correlation analysis is a multivariate statistical analysis method that uses the correlation between comprehensive variable pairs to reflect the overall correlation between two groups of indicators. Its basic principle is: in order to study the relationship between two groups of variables X1, X2, ..., X p , and Y1, Y2, …, Y q The correlation between the two groups of variables is studied by using a method similar to principal component analysis. Several representative variables are selected from the two groups of variables to form representative comprehensive indicators. The correlation between the two groups of comprehensive indicators is replaced by the correlation between the two groups of variables. These comprehensive indicators are called typical variables.

[0052] The basic procedure of canonical correlation analysis is to extract one variable from each linear function of the two groups of variables to form a pair. They should be the pair with the maximum correlation coefficient, called the first pair of typical variables. Similarly, the second pair, the third pair, and so on can be calculated. These pairs of variables are uncorrelated with each other.

[0053] (4) Using the partial least squares method, the gasoline octane number prediction analysis model was established to obtain the near infrared spectrum score of the gasoline sample;

[0054] The following is a brief introduction to the PLS algorithm:

[0055] First, the spectral matrix X (n × m) and the concentration matrix Y (n × 1) (the present invention is gasoline octane number) are decomposed as follows. In this algorithm, n is the number of samples, and m is the number of absorbance wavelength points in the characteristic spectral region, that is, the number of sampling points of absorbance in the characteristic spectral region.

[0056]

[0057] Where: t k (n×1) is the score of the kth principal factor of the absorbance matrix X;

[0058] p k (1×m) is the loading of the kth principal factor of the absorbance matrix X;

[0059] u k (n×1) is the score of the kth principal factor of the concentration matrix Y;

[0060] q k (1×1) is the load of the kth principal factor of the concentration matrix Y; f is the number of principal factors. That is, T and U are the score matrices of the X and Y matrices, P and Q are the load matrices of the X and Y matrices, and E X and E Y are the PLS fitting residual matrices for X and Y respectively.

[0061] In the actual PLS algorithm, PLS combines matrix decomposition and regression into one step, that is, the decomposition of the X and Y matrices is performed simultaneously, and the information of Y is introduced into the decomposition process of the X matrix. Before calculating each new principal component, the score T of X is exchanged with the score U of Y, so that the obtained X principal component is directly associated with Y.

[0062] PLS is calculated by the nonlinear iterative partial least squares algorithm (NIPALS) proposed by HWold. The specific algorithm is as follows:

[0063] For the correction process, ignoring the residual matrix E, when the number of main factors is 1, we have:

[0064] For X = tp T , multiply t on the left T Get: p T =t T X / t T t; multiply the right side by p to get: t = Xp / p T p.

[0065] For Y = uq T , multiply u on the left T Get: q T =u T Y / u T u, divide both sides to get q T So: u=Y / q T .

[0066] 1) Find the weight vector w of the absorbance matrix X

[0067] Take a column of the concentration matrix Y (only one column in the present invention) as the starting iteration value of u, replace t with u, and calculate w. The equation is: X = uw T , its solution is: w T =u T X / u T u

[0068] 2) Normalize the weight vector w

[0069] w T =w T / ||w T ||

[0070] 3) Find the factor score t of the absorbance matrix X, and calculate t from the normalized w. The equation is: X = tw T , the solution is: t = Xw / w T w

[0071] 4) Find the load q value of the concentration matrix Y, and use t instead of u to calculate q. The equation is: Y = tq T , whose solution is: q T =t T Y / t T t

[0072] 5) Normalize the load q

[0073] q T =q T / ||q T ||

[0074] 6) Find the factor score u of the concentration matrix Y, from q T Calculate u, the equation is: Y = uq T , the solution is: u = Yq / q T q

[0075] 7) Replace t with this u and return to step 1) to calculate w T , by w T Calculate t 新 , and iterate repeatedly. If t has converged (||t 新 -t 旧 ||≤10 -6 ||t 新 ||), go to step 8) for calculation, otherwise return to step 1).

[0076] 8) Find the load vector p of the absorbance matrix X from the converged t, the equation is: X = tp T , its solution is: p T =t T Y / t T t

[0077] 9) Normalize the load p

[0078] p T =p T / ||p T ||

[0079] 10) Standardize the factor scores of X

[0080] t=t||p||

[0081] The t obtained here is the partial least squares method, and the gasoline octane number prediction analysis model is established to obtain the near-infrared spectrum score of the gasoline sample.

[0082] 11) Normalize the weight vector w

[0083] w=w||p||

[0084] 12) Calculate the intrinsic relationship b between t and u

[0085] b=u T t / t T t

[0086] 13) Calculate the residual matrix E

[0087] E X =X-tp T

[0088] E Y =Y-btq T

[0089] 14) With E, X Instead of X, E Y Replace Y and return to step 1). Similarly, find w, t, p, u, q, and b of the main factors of X and Y. Use the interaction test method to determine the optimal number of main factors f and save w f 、p f ,q f

[0090] The octane number of the gasoline sample to be tested un The prediction process is as follows:

[0091] x un The characteristic spectral absorbance of the gasoline sample to be tested is called the saved w f 、p f ,q f ,

[0092] First calculate the score of the sample to be tested, t un , t un =x un w;

[0093] First calculate the octane number y of the sample to be tested un ;

[0094] y un =b PLS x un ,in

[0095] (5) The typical variables and near infrared spectrum scores of gasoline samples are combined and correlated with the octane number obtained by the standard method. The support vector machine regression method is used to establish an analytical model for predicting the octane number of the gasoline sample to be tested. The input of the support vector machine regression model is: X (typical variables of gasoline samples, near infrared spectrum scores), Y (octane number of gasoline samples), kernel function, regularization coefficient, error precision. Output: number of support vectors, Lagrange multiplier difference, bias term.

[0096] The support vector machine algorithm is briefly introduced as follows: Support vector machine (SVM) is a machine learning algorithm based on statistical learning theory, which was originally generated for pattern recognition problems. With the introduction of ε-insensitive functions, SVM has been extended to nonlinear regression and function approximation, showing good learning performance, which is called support vector regression correction method (SVR).

[0097] Support vector machine regression (SVR) is derived for the regression problem as follows: given the training data D = {(x1, y1), (x2, y2), ..., (x m ,y m )}, hoping to learn a regression model f(x) = w T x+b makes f(x) as close to y as possible, w and b are model parameters.

[0098] For samples (x, y), traditional regression models usually calculate the loss directly based on the difference between the model output f(x) and the true output y. The loss is 0 only when f(x) is exactly the same as y. In contrast, support vector regression (SVR) assumes that we can tolerate a maximum error of ∈ between f(x) and y, and only calculates the loss when the absolute value of the difference between f(x) and y is greater than ∈.

[0099] Therefore, the SVR problem is written as:

[0100]

[0101] C is the regularization constant, l ∈ for:

[0102] If |z|<∈, l∈(z)=0, l∈(z)=|z|-∈;

[0103] Introducing the slack variable ξ i and ξ^ i , rewrite the above formula as:

[0104]

[0105] st f(x i)-y i ≤∈+ξ i ,

[0106]

[0107]

[0108] Introducing the Lagrange multiplier μ i ≥0,μ i ^≥0, α≥0, α^≥0, we get the following Lagrangian function:

[0109]

[0110] Taking partial derivatives we get:

[0111]

[0112]

[0113] C=α i +μ i ,

[0114]

[0115] Substituting the above formula into the aforementioned Lagrangian function, we get the support vector machine regression (SVR) dual problem

[0116]

[0117] The above process satisfies the Carlo-Kuhn-Tucker (KKT) condition:

[0118] α i (f(x i )-y i -∈-ξ i )=0,

[0119]

[0120]

[0121]

[0122] It can be seen that if and only if f(x i )-y i -∈-ξ i =0 when α i can take non-zero values ​​if and only if y i -f(x i )-∈-ξ i ^ = 0 when α ican take non - zero values. In other words, only when the sample does not fall into the band between ∈, the corresponding α i and α i ^can take non - zero values. In addition, the constraints f(x i ) - y i -∈ - ξ i = 0 and y i -f(x i ) - y i -∈ - ξ i = 0 cannot hold simultaneously. So at least one of α i and α i ^is 0.

[0123] The SVR solution is as follows:

[0124] The samples that can make (α i ^ - α i )≠0 in the above formula are the support vectors of SVR, and they fall outside the band between ∈. At this time, the support vectors of SVR are only a part of the training samples, that is, its solution is still sparse, which is different from the fact that the SVM model is only related to the support vectors.

[0125] It can be seen from the KKT conditions that for each sample, (C - α i )ξ i = 0 and α i (f(x i ) - y i -∈ - ξ i ) = 0. Then, after obtaining α i , if 0 < α i < C, then there must be ξ i = 0, and then

[0126]

[0127] In practice, a more robust method is adopted to select multiple or all samples that satisfy 0 < α i < C, solve for b, and then take the average value.

[0128] Similarly, the kernel function can also be introduced. Replace x with φ(x) to obtain the final model as:

[0129]

[0130] SVR realizes non - linear regression by introducing the kernel function K(x i , x j ). There are three commonly used kernel functions: radial basis kernel function, polynomial kernel function, and sigmoid kernel function.

[0131] This patent uses an exponential radial basis kernel function (ERBF) in the operation during the support vector machine model building process, and combines the typical variables (n×p) obtained by canonical correlation analysis of gasoline samples and the scores (n×q) obtained by partial least squares regression as input variables (n×(p+q)), where n is the number of gasoline samples modeled and p is the number of typical variables.

[0132] The parameters to be set when modeling the support vector machine regression method are the error precision ε, the regularization coefficient C, and the width σ of the radial basis kernel function. According to the measurement error requirements of the actual prediction properties, the error precision ε, the regularization coefficient C, and the width σ of the radial basis kernel function are set by optimizing the modeling samples.

[0133] (6) When predicting the octane number of the gasoline sample to be tested, the near-infrared spectrum of the gasoline sample to be tested is first measured and the second-order differential processing is performed to intercept the 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0134] (7) Then, the typical analysis model established in step (3) is used to obtain the typical variables of the gasoline sample to be tested, and the partial least squares analysis model established in step (4) is used to obtain the near-infrared spectrum score of the gasoline sample to be tested. The gasoline typical variables and near-infrared spectrum scores are combined and substituted into the octane number support vector machine prediction model in step (5) to obtain the predicted octane number of the gasoline sample to be tested. When the support vector machine method is used for prediction, the input is: X (typical variables of the modeled gasoline sample, near-infrared spectrum score), Y (octane number of the modeled gasoline sample), vX (typical variables of the modeled gasoline sample, near-infrared spectrum score), kernel function, Lagrange multiplier difference, and bias term. Output: octane number of the gasoline sample to be tested.

[0135] The following is a more detailed implementation process:

[0136] 1.1 Gasoline sample collection

[0137] At least 100 finished gasoline samples are collected, one part of which is used as modeling samples to determine the research octane number, vapor pressure, benzene, aromatics and olefin content using standard methods, and the other part is used as test samples to determine the research octane number using standard methods for final comparison with the predicted octane number.

[0138] The method for determining the research octane number of gasoline may be the gasoline octane number determination method (research method) (GB / T 5487—2015).

[0139] The method for determining the vapor pressure of gasoline is the Determination of the vapor pressure of petroleum products-Reid method (GB / T 8017-2012).

[0140] The method for determining the benzene, olefin and aromatic content of gasoline is Determination of hydrocarbon group composition and benzene in light petroleum fractions and products-Multidimensional gas chromatography (GB / T 30519-2014).

[0141] 1.2 Modeling gasoline sample near infrared spectrum collection

[0142] The near infrared spectra of the modeling gasoline samples were collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectral range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0143] 1.3 Preprocessing of gasoline sample near infrared spectra

[0144] The near-infrared spectrum of the modeled gasoline sample was processed by second-order differential analysis, and the wavelength range from 5500 to 6350 cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0145] 1.4 Establishment of typical analysis model

[0146] Based on the near infrared spectrum and octane number, vapor pressure, benzene, olefin and aromatic content of the gasoline sample obtained in 1.3, canonical correlation analysis was performed to obtain canonical variables;

[0147] 1.5 Establishment of partial least squares analysis model

[0148] Based on the near infrared spectrum of the gasoline sample obtained in 1.3 and the standard method for determining the octane number, a partial least squares analysis model for predicting the gasoline octane number was established to obtain the near infrared spectrum score of the gasoline sample;

[0149] 1.6 Establishment of octane number analysis model for samples to be tested

[0150] The typical variables of the gasoline samples obtained in 1.4 and the near-infrared spectrum scores of the gasoline samples obtained in 1.5 were combined and correlated with the octane number obtained by the standard method, and the support vector machine method was used to establish an analytical model for predicting the octane number of the gasoline samples to be tested.

[0151] 1.7 Near infrared spectrum collection of gasoline sample to be tested

[0152] The near infrared spectrum of the gasoline sample to be tested was collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0153] 1.8 Pretreatment of gasoline samples by near infrared spectrum

[0154] The near infrared spectrum of the gasoline sample to be tested was processed by second-order differential, and the 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0155] 1.9 Obtaining Typical Variables of Gasoline Samples to be Tested

[0156] Based on the typical analysis model established in 1.4, the typical variables of the gasoline sample to be tested are obtained.

[0157] 1.10 Obtaining the near infrared spectrum score of the gasoline sample to be tested

[0158] The partial least squares analysis model established in 1.5 is used to obtain the near-infrared spectrum score of the gasoline sample to be tested.

[0159] 1.11 Prediction of the final octane number of the gasoline sample to be tested

[0160] The typical variables of the gasoline sample to be tested obtained in 1.9 and the near infrared spectrum scores of the gasoline sample to be tested obtained in 1.10 are combined and substituted into the octane number support vector machine prediction model in step (5) to obtain the final octane number prediction value of the gasoline sample to be tested.

[0161] 1.12 Evaluation of octane number prediction accuracy

[0162] Compare the predicted octane number of the gasoline sample to be tested obtained in 1.11 with the octane number obtained by the standard method in 1.1, and calculate the standard error of prediction (SEP) and the mean absolute error (MAE).

[0163]

[0164]

[0165] y i,预测值 is the predicted value of the octane number of the i-th gasoline sample to be tested, y i,实测值 is the octane number of the i-th gasoline sample to be tested measured by the standard method, and n is the number of gasoline samples to be tested used to calculate the standard deviation and the mean absolute deviation.

[0166] The following specific examples are given to demonstrate the prediction effect of this case.

[0167] Example 1 (Near Infrared Spectrum Score + Typical Variables)

[0168] 2.1 Gasoline sample collection

[0169] 216 desulfurized gasoline samples from the S Zorb unit were collected, 170 of which were used as modeling samples and their research octane number, vapor pressure, benzene, aromatics and olefin content were determined using standard methods. 46 of them were used as test samples and their research octane number was determined using standard methods for comparison with the predicted octane number.

[0170] The method for determining the research octane number of gasoline is the gasoline octane number determination method (research method) (GB / T 5487-2015).

[0171] The method for determining the vapor pressure of gasoline is the Determination of the vapor pressure of petroleum products-Reid method (GB / T 8017-2012).

[0172] The method for determining the benzene, olefin and aromatic content of gasoline is Determination of hydrocarbon group composition and benzene in light petroleum fractions and products-Multidimensional gas chromatography (GB / T 30519-2014).

[0173] 2.2 Modeling gasoline sample near infrared spectrum acquisition

[0174] The near infrared spectra of 170 modeling gasoline samples were collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectral range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0175] 2.3 Preprocessing of gasoline sample near infrared spectra for modeling

[0176] The near infrared spectrum of the modeled gasoline sample was processed with a second-order differential window of 25 points, intercepting 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region.

[0177] 2.4 Establishment of partial least squares analysis model

[0178] Based on the near infrared spectrum of the gasoline sample obtained in 2.3 and the standard method for determining the octane number, a partial least squares analysis model for predicting the gasoline octane number was established to obtain the near infrared spectrum score of the gasoline sample, and the first 15 scores (170×15) were taken. The specific partial least squares model establishment method refers to CN201410307163 "A method for predicting the hydrogen sulfide content in desulfurization amine liquid by near infrared spectroscopy".

[0179] 2.5 Establishment of typical analysis model

[0180] Based on the near-infrared spectral absorbance of the gasoline sample obtained in 2.3 and the octane number, vapor pressure, benzene, olefin and aromatic content obtained in 2.1, a series of typical variables were obtained by canonical correlation analysis, and the first 6 typical variables (170×6) were taken. The canonical correlation analysis process is detailed in "Research on Regional Short-term Climate Prediction Using Canonical Correlation Analysis" (Shanxi Meteorology. 2001, (02), 9:12).

[0181] 2.6 Establishment of octane number analysis model for samples to be tested

[0182] The 15 scores of the gasoline samples obtained in 2.4 and the 6 typical variables obtained in 2.5 were combined (170×21), and correlated with the octane number obtained by the standard method, and the support vector machine method was used to establish an analytical model for predicting the octane number of the gasoline samples to be tested. The specific vector machine model establishment method refers to "Research on Establishing a Universal Near Infrared Calibration Model for Finished Gasoline by Support Vector Regression" (Journal of Analysis and Testing, 2008, 27(6): 619-622).

[0183] 2.7 Near infrared spectrum collection of gasoline samples to be tested

[0184] The near infrared spectra of 46 gasoline samples to be tested were collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0185] 2.8 Pretreatment of gasoline samples by near infrared spectroscopy

[0186] The near infrared spectrum of the gasoline sample to be tested was processed with a second-order differential window width of 25 points, and the 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region.

[0187] 2.9 Obtaining the Near Infrared Spectrum Score of the Gasoline Sample to be Tested

[0188] The partial least squares analysis model established in 2.4 is used to obtain the near-infrared spectrum score of the gasoline sample to be tested, and the first 15 scores (46×15) are taken.

[0189] 2.10 Obtaining typical variables of gasoline samples to be tested

[0190] Based on the typical analysis model established in 2.5, the typical variables of the gasoline sample to be tested are obtained, and the first 6 typical variables are taken.

[0191] 2.11 Prediction and accuracy evaluation of the final octane number of the gasoline sample to be tested

[0192] The 15 scores of the gasoline samples to be tested obtained in 2.9 and the 6 typical variables of the gasoline samples to be tested obtained in 2.10 are combined (46×21) and substituted into the octane number support vector machine prediction model established in 2.6 to obtain the final octane number prediction value of the gasoline samples to be tested (46×1). The predicted octane number of the gasoline samples to be tested is compared with the octane number obtained by the standard method in 2.1, and the standard deviation SEP (standard error of prediction) and the mean absolute error MAE (mean absolute error) are calculated, as shown in Table 1.

[0193] Table 1 Comparison of predicted and measured octane numbers of the tested samples

[0194]

[0195]

[0196] Compared with Example 1, Comparative Example 1 is based solely on the correlation between the partial least squares score of the near infrared spectrum and the octane number, and uses the support vector machine method to establish an octane number prediction model to predict the octane number of the sample to be tested. The specific steps are as follows.

[0197] (1) Gasoline sample collection, as described in 2.1.

[0198] (2) Collection of near-infrared spectra of modeled gasoline samples, as described in 2.1.

[0199] (3) Preprocessing of near-infrared spectra of gasoline samples for modeling is the same as described in 2.3.

[0200] (4) The partial least squares analysis model is established, as described in 2.4.

[0201] (5) Establishment of octane number analysis model for the sample to be tested,

[0202] The 15 scores of the modeled gasoline samples obtained in step (4) are correlated with the octane numbers obtained by the standard method, and a support vector machine method is used to establish an analytical model for predicting the octane number of the gasoline samples to be tested.

[0203] (6) Collection of near infrared spectrum of the gasoline sample to be tested, as described in 2.7.

[0204] (7) Preprocessing of the near infrared spectrum of the gasoline sample to be tested is the same as described in 2.8.

[0205] (8) Obtain the near infrared spectrum score of the gasoline sample to be tested, as described in 2.9.

[0206] (9) Prediction and accuracy evaluation of the final octane number of the gasoline sample to be tested

[0207] Substitute the 15 scores (46×15) of the gasoline samples to be tested obtained in step (8) into the octane number support vector machine prediction model established in step (5) to obtain the final octane number prediction value (46×1) of the gasoline samples to be tested. Compare the predicted octane number of the gasoline samples to be tested with the octane number obtained by the standard method in step (1), and calculate the standard deviation SEP (standard error of prediction) and the mean absolute error MAE (mean absolute error), as shown in Table 2.

[0208] Table 2 Comparison of predicted and measured octane number of samples to be tested (Comparative Example 1)

[0209]

[0210]

[0211]

[0212] Compared with the comparative example 1, in which the octane number prediction model was established based on the correlation between the partial least squares score of the near infrared spectrum and the octane number using the support vector machine method, and the octane number prediction standard deviation was 0.233, the present invention combines the typical variables of the near infrared spectrum of the gasoline sample with the partial least squares score and the correlation with the octane number, and establishes a prediction octane number model using the support vector machine method. The octane number prediction standard deviation in the embodiment is 0.216, which improves the accuracy of octane number prediction.

[0213] The following describes the information combination "absorbance, initial predicted octane number and characteristic predicted value". The prediction method using this information combination may include the following steps:

[0214] (1) Collect a certain amount of gasoline sample, measure its near infrared spectrum, and then measure its octane number, vapor pressure, benzene, olefin and aromatic content using standard methods;

[0215] (2) Based on the near-infrared spectrum and octane number, vapor pressure, benzene, olefins and aromatics content of pretreated gasoline, the partial least squares method was used to establish analytical models for predicting gasoline octane number, vapor pressure, benzene, olefins and aromatics respectively;

[0216] (3) Based on the near-infrared spectrum of pretreated gasoline and the combined information of octane number, vapor pressure, benzene, olefin and aromatic content determined by standard methods and the correlation with octane number, a support vector machine method was used to establish an analytical model for predicting the octane number of the gasoline sample to be tested.

[0217] (4) When predicting the octane number of the gasoline sample to be tested, firstly measure the near infrared spectrum of the gasoline sample to be tested and perform pretreatment; then use the analysis model established in step (2) to obtain the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample to be tested;

[0218] (5) Combining the near infrared spectrum of the pretreated gasoline sample with the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample obtained in step (4), and substituting them into the octane number support vector machine prediction model in step (4) to obtain the predicted octane number of the gasoline sample.

[0219] In addition to the near-infrared spectrum of gasoline, this scheme adds gasoline characteristic information (such as vapor pressure, benzene, olefin and aromatic content) for information amplification, correlates it with the octane number, and then uses the support vector machine method to establish an octane number prediction model to predict the octane number of the sample to be tested.

[0220] The specific implementation of this information combination is as follows:

[0221] 3.1 Gasoline sample collection

[0222] At least 100 gasoline samples were collected, one part of which was used as modeling samples to determine the research octane number, vapor pressure, benzene, aromatics and olefin content using standard methods, and the other part was used as test samples to determine the research octane number using standard methods for final comparison with the predicted octane number.

[0223] The method for determining the research octane number of gasoline is the gasoline octane number determination method (research method) (GB / T 5487-2015).

[0224] The method for determining the vapor pressure of gasoline is the Determination of the vapor pressure of petroleum products-Reid method (GB / T 8017-2012).

[0225] The method for determining the benzene, olefin and aromatic content of gasoline is Determination of hydrocarbon group composition and benzene in light petroleum fractions and products-Multidimensional gas chromatography (GB / T 30519-2014).

[0226] 3.2 Modeling gasoline sample near infrared spectrum acquisition

[0227] The near infrared spectra of the modeling gasoline samples were collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectral range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0228] 3.3 Preprocessing of gasoline sample near infrared spectra for modeling

[0229] The near-infrared spectrum of the modeled gasoline sample was processed by 25-point second-order differential, intercepting 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0230] 3.4 Initial analysis model establishment

[0231] Based on the near-infrared spectrum and octane number, vapor pressure, benzene, olefins and aromatics content of the pretreated modeled gasoline sample, the partial least squares method was used to establish analysis models for predicting gasoline octane number, vapor pressure, benzene, olefins and aromatics respectively; the specific partial least squares model establishment method refers to CN201410307163 "A method for predicting the hydrogen sulfide content in desulfurization amine liquid by near-infrared spectroscopy".

[0232] 3.5 Establishment of octane number analysis model for samples to be tested

[0233] Based on the near infrared spectrum of the pre-treated modeled gasoline sample and the combined information of octane number, vapor pressure, benzene, olefin and aromatic content determined by standard methods, the support vector machine method is used to establish an analytical model for predicting the octane number of the gasoline sample to be tested. The specific vector machine model establishment method is referred to "Research on Establishing a Universal Near Infrared Calibration Model for Finished Gasoline by Support Vector Regression" (Journal of Analysis and Testing, 2008, 27(6): 619-622).

[0234] 3.6 Near infrared spectrum collection of gasoline samples to be tested

[0235] The near infrared spectrum of the gasoline sample to be tested was collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0236] 3.7 Pretreatment of gasoline samples by near infrared spectroscopy

[0237] The near infrared spectrum of the gasoline sample to be tested was processed by 25-point second-order differential, and the 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0238] 3.8 Initial prediction of near-infrared properties of gasoline samples to be tested

[0239] Based on the near infrared spectrum pretreatment of the gasoline sample to be tested after pretreatment in 3.7 and the analysis model established in 3.4, the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content of the gasoline sample to be tested are obtained;

[0240] 3.9 Prediction of the final octane number of the gasoline sample to be tested

[0241] The near infrared spectrum of the gasoline to be tested after pretreatment in 3.7 and the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content of the gasoline sample to be tested obtained in 3.8 are combined and substituted into 3.5 to obtain the support vector machine analysis model to obtain the predicted octane number of the gasoline sample to be tested.

[0242] 3.10 Evaluation of octane number prediction accuracy

[0243] Compare the predicted octane number of the gasoline sample to be tested obtained in 3.9 with the octane number obtained by the standard method in 3.1, and calculate the standard deviation SEP (standard error of prediction) and the mean absolute error MAE (mean absolute error).

[0244]

[0245]

[0246] y i,预测值 is the predicted value of the octane number of the i-th gasoline sample to be tested, y i,实测值 is the octane number of the i-th gasoline sample to be tested measured by the standard method, and n is the number of gasoline samples to be tested used to calculate the standard deviation and the mean absolute deviation.

[0247] The following specific examples are given to demonstrate the prediction effect of this case.

[0248] Example 2 (absorbance + initial predicted octane number + characteristic predicted value)

[0249] 4.1 Gasoline sample collection

[0250] 216 desulfurized gasoline samples from the S Zorb unit were collected, 170 of which were used as modeling samples and their research octane number, vapor pressure, benzene, aromatics and olefin content were determined using standard methods. 46 of them were used as test samples and their research octane number was determined using standard methods for comparison with the predicted octane number.

[0251] 4.2 Modeling gasoline sample near infrared spectrum acquisition

[0252] The near infrared spectra of the modeling gasoline samples were collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectral range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0253] 4.3 Preprocessing of gasoline sample near infrared spectra for modeling

[0254] The near-infrared spectrum of the modeled gasoline sample was processed by 25-point smoothing second-order differential, and the 5500-6350 cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0255] 4.4 Initial analysis model establishment

[0256] Based on the near-infrared spectra and octane number, vapor pressure, benzene, olefins and aromatics content of the pretreated modeled gasoline samples, the partial least squares method was used to establish analytical models for predicting gasoline octane number, vapor pressure, benzene, olefins and aromatics.

[0257] 4.5 Establishment of octane number analysis model for samples to be tested

[0258] Based on the near infrared spectrum of the pre-processed gasoline sample and the combined information of octane number, vapor pressure, benzene, olefin and aromatic content measured by standard method, the support vector machine method is used to establish an analytical model for predicting the octane number of the gasoline sample to be tested. Support vector machine model parameters: erbf radial basis kernel function, penalty coefficient C is 200.

[0259] 4.6 Near infrared spectrum collection of gasoline samples to be tested

[0260] The near infrared spectrum of the gasoline sample to be tested was collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0261] 4.7 Pretreatment of gasoline samples by near infrared spectroscopy

[0262] The near infrared spectrum of the gasoline sample to be tested was processed by 25-point second-order differential, and the 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0263] 4.8 Initial prediction of near-infrared properties of gasoline samples to be tested

[0264] Based on the near infrared spectrum pretreatment of the gasoline sample to be tested after pretreatment in 3.7 and the analysis model established in 4.4, the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample to be tested are obtained;

[0265] 4.9 Prediction of the final octane number of the gasoline sample to be tested

[0266] The near infrared spectrum of the pretreated gasoline to be tested in 4.7 and the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample to be tested obtained in 4.8 are combined and substituted into the support vector machine analysis model obtained in 4.5 to obtain the predicted octane number of the gasoline sample to be tested. The predicted octane number of the gasoline sample to be tested is compared with the octane number obtained by the standard method in 4.1, and the standard deviation SEP (standard error of prediction) and the mean absolute error MAE (mean absolute error) are calculated, as shown in Table 3.

[0267] Table 3 Comparison of predicted and measured octane numbers of the tested samples

[0268]

[0269]

[0270] Compared with Example 2, Comparative Example 2 is based solely on the correlation between near infrared spectrum and octane number, and uses the support vector machine method to establish an octane number prediction model to predict the octane number of the sample to be tested. The specific steps are as follows.

[0271] (1) Gasoline sample collection, as described in 4.1.

[0272] (2) Collection of near-infrared spectra of modeled gasoline samples, as described in 4.2.

[0273] (3) Preprocessing of near-infrared spectra of gasoline samples for modeling is the same as described in 4.3.

[0274] (4) Establishment of octane number analysis model for samples to be tested

[0275] Based on the near infrared spectrum of the pre-processed modeled gasoline sample and the octane number measured by the standard method, the support vector machine method is used to establish an analytical model for predicting the octane number of the gasoline sample to be tested. Support vector machine model parameters: erbf radial basis kernel function, penalty coefficient C is 200.

[0276] (5) Collection of near infrared spectrum of the gasoline sample to be tested is the same as described in 4.6.

[0277] (6) Pretreatment of the near infrared spectrum of the gasoline sample to be tested is the same as described in 4.7.

[0278] (7) Prediction and accuracy evaluation of the final octane number of the gasoline sample to be tested

[0279] Substitute the near-infrared spectrum of the gasoline to be tested after pretreatment in 4.7 into the support vector machine analysis model obtained in step (4) to obtain the predicted octane number of the gasoline sample to be tested. Compare the predicted octane number of the gasoline sample to be tested with the octane number obtained by the standard method in step (1), and calculate the standard deviation SEP (standard error of prediction) and the mean absolute error MAE (mean absolute error), as shown in Table 4.

[0280] Table 4 Comparison of predicted and measured octane number of samples to be tested (Comparative Example 2)

[0281]

[0282]

[0283] It can be seen from the comparison results that the present invention adds gasoline composition data in addition to gasoline near-infrared spectrum for information amplification, and the standard deviation of octane number prediction is reduced from 0.231 to 0.182, thereby improving the accuracy of octane number prediction, with obvious effect.

[0284] The following describes the information combination "near infrared spectrum score, initial predicted octane number and characteristic predicted value". The prediction method using this information combination may include the following steps:

[0285] (1) Collect a certain amount of gasoline sample, measure its near infrared spectrum, and then measure its octane number, vapor pressure, benzene, olefin and aromatic content using standard methods;

[0286] (2) Based on the pre-treated near-infrared spectrum and octane number, vapor pressure, benzene, olefins and aromatics content, the partial least squares method was used to establish analytical models for predicting gasoline octane number, vapor pressure, benzene, olefins and aromatics respectively;

[0287] (3) Based on the pre-processed near-infrared spectrum and the octane number determined by the standard method, a partial least squares model was established to obtain the gasoline sample score and the combined information of octane number, vapor pressure, benzene, olefin and aromatic content obtained by the standard method, and a support vector machine method was used to establish an analytical model for predicting the octane number of the gasoline sample to be tested.

[0288] (4) When predicting the octane number of the gasoline sample to be tested, firstly measure the near infrared spectrum of the gasoline sample to be tested and perform preprocessing; then, use the analysis model established in step (2) to obtain the score and the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample to be tested;

[0289] (5) The scores of the gasoline samples to be tested and the predicted octane number and the predicted values ​​of vapor pressure, benzene, olefin and aromatic content are combined and substituted into the octane number support vector machine prediction model in step (4) to obtain the predicted octane number of the gasoline samples to be tested.

[0290] In addition to the gasoline near-infrared spectrum score, this scheme adds gasoline characteristic information (such as vapor pressure, benzene, olefin and aromatic content) for information amplification, correlates it with the octane number, and then uses the support vector machine method to establish an octane number prediction model to predict the octane number of the sample to be tested.

[0291] The specific implementation of this information combination is as follows:

[0292] 5.1 Gasoline Sample Collection

[0293] At least 100 gasoline samples were collected, one part of which was used as modeling samples to determine the research octane number, vapor pressure, benzene, aromatics and olefin content using standard methods, and the other part was used as test samples to determine the research octane number using standard methods for final comparison with the predicted octane number.

[0294] The method for determining the research octane number of gasoline is the gasoline octane number determination method (research method) (GB / T 5487-2015).

[0295] The method for determining the vapor pressure of gasoline is the Determination of the vapor pressure of petroleum products-Reid method (GB / T 8017-2012).

[0296] The method for determining the benzene, olefin and aromatic content of gasoline is Determination of hydrocarbon group composition and benzene in light petroleum fractions and products-Multidimensional gas chromatography (GB / T 30519-2014).

[0297] 5.2 Modeling gasoline sample near infrared spectrum acquisition

[0298] The near infrared spectra of the modeling gasoline samples were collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0299] 5.3 Preprocessing of gasoline sample near infrared spectra for modeling

[0300] The near-infrared spectrum of the modeled gasoline sample was processed by 25-point second-order differential, intercepting 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0301] 5.4 Initial analysis model establishment

[0302] Based on the near-infrared spectrum and octane number, vapor pressure, benzene, olefins and aromatics content of the pretreated modeled gasoline sample, the partial least squares method was used to establish analysis models for predicting gasoline octane number, vapor pressure, benzene, olefins and aromatics respectively; the specific partial least squares model establishment method refers to CN201410307163 "A method for predicting the hydrogen sulfide content in desulfurization amine liquid by near-infrared spectroscopy".

[0303] 5.5 Establishment of octane number analysis model for samples to be tested

[0304] Based on the near infrared spectrum of the treated gasoline sample and the octane number measured by the standard method, the gasoline sample score obtained by the partial least squares model and the combined information of the octane number, vapor pressure, benzene, olefin and aromatic content measured by the standard method are established, and the support vector machine method is used to establish an analytical model for predicting the octane number of the gasoline sample to be tested. The specific vector machine model establishment method refers to "Research on Establishing a Universal Near Infrared Calibration Model for Finished Gasoline by Support Vector Regression" (Journal of Analysis and Testing, 2008, 27(6): 619-622).

[0305] 5.6 Near-infrared spectrum collection of gasoline samples to be tested

[0306] The near infrared spectrum of the gasoline sample to be tested was collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0307] 5.7 Pretreatment of gasoline samples by near infrared spectroscopy

[0308] The near infrared spectrum of the gasoline sample to be tested was processed by 25-point second-order differential, and the 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0309] 5.8 Initial prediction of near-infrared properties of gasoline samples to be tested

[0310] Based on the near infrared spectrum pretreatment of the gasoline sample to be tested after pretreatment in 5.7 and the analysis model established in 5.4, the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample to be tested are obtained;

[0311] 5.9 Prediction of the final octane number of the gasoline sample to be tested

[0312] The analysis model established in 5.4 is used to obtain the score of the gasoline sample to be tested, which is combined with the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample to be tested obtained in 5.8, and substituted into 5.5 to obtain the support vector machine analysis model to obtain the predicted octane number of the gasoline sample to be tested.

[0313] 5.10 Evaluation of octane number prediction accuracy

[0314] Compare the predicted octane number of the gasoline sample to be tested obtained in 5.9 with the octane number obtained by the standard method in 5.1, and calculate the standard deviation SEP (standard error of prediction) and the mean absolute error MAE (mean absolute error).

[0315]

[0316]

[0317] y i,预测值 is the predicted value of the octane number of the i-th gasoline sample to be tested, y i,实测值 is the octane number of the i-th gasoline sample to be tested measured by the standard method, and n is the number of gasoline samples to be tested used to calculate the standard deviation and the mean absolute deviation.

[0318] The following specific examples are given to demonstrate the prediction effect of this case.

[0319] Example 3 (Near infrared spectrum score + initial predicted octane number + characteristic prediction value)

[0320] 6.1 Gasoline Sample Collection

[0321] 216 desulfurized gasoline samples from the S Zorb unit were collected, 170 of which were used as modeling samples and their research octane number, vapor pressure, benzene, aromatics and olefin content were determined using standard methods. 46 of them were used as test samples and their research octane number was determined using standard methods for comparison with the predicted octane number.

[0322] 6.2 Modeling gasoline sample near infrared spectrum acquisition

[0323] The near infrared spectra of the modeling gasoline samples were collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0324] 6.3 Preprocessing of gasoline sample near infrared spectra for modeling

[0325] The near-infrared spectrum of the modeled gasoline sample was processed by 25-point second-order differential, intercepting 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0326] 6.4 Initial analysis model establishment

[0327] Based on the near-infrared spectra and octane number, vapor pressure, benzene, olefins and aromatics content of the pretreated modeled gasoline samples, the partial least squares method was used to establish analytical models for predicting gasoline octane number, vapor pressure, benzene, olefins and aromatics.

[0328] 6.5 Establishment of octane number analysis model for samples to be tested

[0329] Based on the partial least squares analysis model for octane number established in 6.4, the first 15 scores of gasoline samples and the combined information of octane number, vapor pressure, benzene, olefin and aromatic content determined by the standard method are obtained. The support vector machine method is used to establish an analysis model for predicting the octane number of the gasoline sample to be tested. Support vector machine model parameters: erbf radial basis kernel function, penalty coefficient C is 200.

[0330] 6.6 Near-infrared spectrum collection of gasoline samples to be tested

[0331] The near infrared spectrum of the gasoline sample to be tested was collected on an FTPA2000-260 online Fourier transform near infrared spectrometer; InGaAs detector; scanning spectrum range 10000~5000cm -1 ; Resolution 8cm -1 , scan times 64 times.

[0332] 6.7 Pretreatment of gasoline samples by near infrared spectroscopy

[0333] The near infrared spectrum of the gasoline sample to be tested was processed by 25-point second-order differential, and the 5500-6350cm -1 , 7800~8800cm -1 Absorbance in the spectral region;

[0334] 6.8 Initial prediction of near-infrared properties of gasoline samples to be tested

[0335] Based on the near infrared spectrum pretreatment of the gasoline sample to be tested after pretreatment in 6.7 and the analysis model established in 6.4, the initial predicted octane number and vapor pressure, benzene, olefin and aromatic content predicted values ​​of the gasoline sample to be tested are obtained;

[0336] 6.9 Prediction of the final octane number of the gasoline sample to be tested

[0337] The first 15 scores of the gasoline samples to be tested were obtained by using the analysis model established in 6.5. The scores were combined with the initial predicted octane number and the predicted values ​​of vapor pressure, benzene, olefin and aromatic content of the gasoline samples to be tested obtained in 6.8, and substituted into the support vector machine analysis model in 6.5 to obtain the predicted octane number of the gasoline samples to be tested. The predicted octane number of the gasoline samples to be tested was compared with the octane number obtained by the standard method in 6.1, and the standard deviation SEP (standard error of prediction) and mean absolute error MAE (mean absolute error) were calculated, as shown in Table 5.

[0338] Table 5 Comparison of predicted and measured octane numbers of the tested samples

[0339]

[0340]

[0341] Compared with the comparative example 1, which is based solely on the correlation between the near infrared spectrum partial least squares score and the octane number, and uses the support vector machine method to establish an octane number prediction model with a standard deviation of 0.233, the present invention adds composition data in addition to the gasoline near infrared spectrum score information for information amplification, thereby improving the accuracy of octane number prediction. The standard deviation of octane number prediction in the embodiment of the present invention is 0.192, which is obvious.

[0342] In one embodiment, Figure 2 As shown, a device for predicting gasoline octane number is provided, including a processor and a memory, wherein the processor executes a program stored in the memory to implement corresponding functions.

[0343] The processor includes a kernel, and the kernel calls the corresponding program unit from the memory. One or more kernels can be set, and the method for predicting gasoline octane number is realized by adjusting kernel parameters.

[0344] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0345] Specifically, the above-mentioned device for predicting gasoline octane number includes a processor A01, a network interface A02, a display screen A04, an input device A05 and a memory (not shown in the figure) connected through a system bus. Among them, the processor A01 is used to provide computing and control capabilities. The memory includes an internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A06. The network interface A02 is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor A01, the above-mentioned method for predicting gasoline octane number is implemented. The display screen A04 can be a liquid crystal display screen or an electronic ink display screen, and the input device A05 can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0346] An embodiment of the present application provides a storage medium on which a program is stored. When the program is executed by a processor, the above-mentioned method for predicting gasoline octane number is implemented.

[0347] An embodiment of the present application provides a processor, which is used to run a program, wherein the program executes the above-mentioned method for predicting gasoline octane number when running.

[0348] 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.

[0349] 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.

[0350] 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.

[0351] 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.

[0352] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0353] 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.

[0354] Computer readable media include permanent and non-permanent, removable and non-removable media, and 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.

[0355] 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.

[0356] 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 predicting gasoline octane number, characterized in that: The method comprises: Determine the near infrared spectrum of the gasoline sample to be tested, perform second-order differential processing on the near infrared spectrum, and intercept the absorbance of the near infrared spectrum in a predetermined spectral region; The absorbance of the gasoline sample to be tested is input into the partial least squares analysis model to obtain the following information of the gasoline sample to be tested: 1) an initial predicted octane number and a property predicted value, wherein the initial predicted octane number and the property predicted value together with the absorbance serve as an information combination; 2) a near infrared spectral score relative to octane number, an initial predicted octane number, and a property prediction value, the three being a combination of information; or 3) a near infrared spectrum score and a typical variable relative to the octane number, the two being a combination of information, the typical variable reflecting the correlation between the absorbance of the gasoline sample to be tested in the predetermined spectral region and the octane number and characteristics of the gasoline sample to be tested, The information combination is input into an octane number support vector machine prediction model corresponding to the information combination to obtain an octane number prediction value of the gasoline sample to be tested.

2. The method according to claim 1, characterized in that The partial least squares analysis model is established by: Collecting a plurality of gasoline samples, measuring their near-infrared spectra, performing second-order differential processing on the near-infrared spectra, and intercepting the absorbance of the near-infrared spectra in a predetermined spectral region; Determining the octane number and characteristics of the plurality of gasoline samples using standard methods; as well as Based on the absorbance, octane number and characteristics of the plurality of gasoline samples, a partial least squares analysis model for predicting the octane number of the gasoline sample and a partial least squares analysis model for predicting the characteristics of the gasoline sample are respectively established using the partial least squares method.

3. The method according to claim 2, characterized in that The initial predicted octane number and the near infrared spectrum score relative to the octane number are obtained by using a partial least squares analysis model for predicting the octane number of a gasoline sample; The property prediction values ​​are obtained by using a partial least squares analysis model for predicting the properties of the gasoline sample.

4. The method according to claim 1, characterized in that: The typical variables of the gasoline sample to be tested are determined by: The absorbance of the gasoline sample to be tested is input into a typical analysis model to obtain typical variables of the gasoline sample to be tested, wherein the typical analysis model is established by utilizing the absorbance of near-infrared spectra of multiple gasoline samples in the predetermined spectral region and the octane numbers and characteristics of the multiple gasoline samples measured by a standard method.

5. The method according to claim 1, characterized in that The octane number support vector machine prediction model is established by using the information combination of multiple gasoline samples and the octane numbers of the multiple gasoline samples measured using a standard method.

6. The method according to any one of claims 1 to 5, characterized in that The predetermined spectral range is 5500-6350 cm -1 and 7800~8800cm -1 .

7. The method according to any one of claims 1 to 5, characterized in that The properties include one or more of: vapor pressure, benzene, olefins, and aromatics content.

8. The method according to any one of claims 1 to 5, characterized in that The octane number support vector machine prediction model adopts an exponential radial basis kernel function.

9. A device for predicting gasoline octane number, the device comprising: Memory; and A processor, wherein the processor is configured to execute the method for predicting gasoline octane number according to any one of claims 1 to 8.

10. A machine-readable storage medium having instructions stored thereon, characterized in that: When the instruction is executed by a processor, the processor is configured to execute the method for predicting gasoline octane number according to any one of claims 1 to 8.

Images