A method for identifying upper dilution concentration limit of high density crude oil in near infrared spectroscopy analysis

Abstract

The application discloses a method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectrum analysis. The method is aimed at the problem that improper selection of the dilution concentration can reduce the detection precision of the near-infrared spectrum of the crude oil, especially the problem that too high concentration can cause spectrum nonlinearity. The method is based on a linear mixing model to remove the spectral contribution of the diluent and restore the spectrum of the crude oil. In the iteration process, the optimal inner point set is obtained, so as to determine the upper limit of the dilution ratio concentration. The method can identify the upper limit of the dilution concentration, thereby improving the detection precision of the near-infrared spectrum of the crude oil, being beneficial to subsequent analysis of the properties of the crude oil and processing, and having important value for improving the economic benefits of a refining enterprise.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of crude oil processing of refining enterprises, and particularly to a method for identifying upper limit of dilution concentration of high-density crude oil in near-infrared spectrum analysis. BACKGROUND

[0002] There are many types of crude oil with great differences in properties handled by refining enterprises in China. When analyzing the properties of high-density crude oil by using near-infrared spectrum technology, it is usually necessary to dilute the crude oil with solvents such as toluene to improve the flowability of the crude oil sample in the spectrum pretreatment system, so as to facilitate the near-infrared spectrum detection of the properties of the crude oil.

[0003] Although dilution treatment is beneficial to the near-infrared spectrum detection of the crude oil, excessive dilution may have adverse effects. In the process of reducing the near-infrared spectrum of the crude oil, excessive diluent may cause nonlinearity in the spectrum, affecting the accuracy of the spectrum detection and reducing the detection accuracy of the properties of the crude oil.

[0004] After a new high-density crude oil is purchased from the international market by a domestic refining enterprise, it is generally necessary to determine the upper limit of the dilution concentration. In the early stage, we used the 3σ method to determine the upper limit of the dilution concentration (see the invention patent “Method for reconstructing near-infrared spectrum of heavy crude oil”, application number: 202410786075.0), but in actual application, this method may be affected by outliers, especially when the number of samples is small or the sample distribution is uneven, this method may sometimes fail to accurately reflect the true situation of the data, thereby leading to result deviation. Therefore, how to find and approach the reasonable upper limit of the dilution concentration while ensuring the flowability of the crude oil sample has become an important problem to be solved in the near-infrared spectrum property analysis of high-density crude oil by refining enterprises. SUMMARY

[0005] The present application discloses a method for identifying the upper limit of the dilution concentration of high-density crude oil in near-infrared spectrum analysis, which uses a linear mixing model to remove the spectral contribution of the diluent in the crude oil mixture, constructs a crude oil reduction spectrum dataset, and establishes a response model by iteratively selecting a data subset to obtain an optimal inner point set, thereby determining the upper limit of the dilution ratio concentration. The method has the following steps:

[0006] 1) Dilute the crude oil to be identified with toluene diluent at different concentration ratios, change the concentration ratio of the diluent by 0.05, obtain N groups of diluted mixtures, and record the near-infrared spectrum of the mixtures as A m , respectively measure the near-infrared spectrum A m,i , i = 1, 2,..., N, and measure the near-infrared spectrum A d of the diluent;

[0007] 2) Remove the near-infrared spectrum Am,i The diluent spectral contribution in the mixture is reduced from the near infrared spectrum of the crude oil in the i-th group, A h,i , as follows:

[0008]

[0009] where A h,i is the absorbance of the crude oil reduced from the i-th group of mixtures at wave number point j, A m,i (j) is the absorbance of the mixture at wave number point j, A d,i (j) is the absorbance of the diluent at wave number point j, w i is the diluent concentration ratio, and the spectral data information A h,i of each group of crude oils is reduced to form a data set S;

[0010] 3) Perform polynomial fitting baseline correction on the crude oil spectrum in S;

[0011] 4) Randomly select n groups of crude oil spectral data from S to form a subset S1, where n≥2, and n=3 is generally taken;

[0012] 5) Establish a linear model M1 for the spectral data in S1:

[0013] A h (j) = A h,0 (j) + c(j) * w

[0014] where A h,0 (j) represents the true absorbance at wave number point j, and c(j) * w represents the absorbance deviation at wave number point j, c(j) and A h,0 (j) are obtained by least squares, j=1,2,...,J, J is the number of wave number points, and the model is expressed in vector form as A h = A h,0 + C * w, C = [c(1), c(2),..., c(J)] T ;

[0015] 6) Use the model M1 to predict all spectral samples in the data set S, and obtain the predicted spectral data A h,pre , calculate the error E = ||A h,pre -A h || between the predicted spectrum A h and the reduced spectrum A h,pre , and if E > E thr , it is determined that the current spectrum is an in-point of the model M1, otherwise it is determined to be an out-point, and E thr is the error threshold, which is obtained by the following formula:

[0016]

[0017] wherein the coefficient a is 2.5-4.5;

[0018] 7) Spectra of all inliers form inlier set Indicates the consistency set of S1, and the optimal model is M best , corresponding to the optimal inlier set If it is the first iteration, M best = M1, Otherwise, compare With The number of inliers, if The number of inliers is greater than Update M best And

[0019] 8) Repeat 4) to 7) until the iteration number k≥K, or the inlier ratio exceeds the threshold value t, and the optimal inlier set is considered at this time The highest concentration ratio of spectral data in w best,max , wherein the inlier ratio threshold value t is 85%-95%, and the maximum iteration number K is obtained by the following formula:

[0020]

[0021] In the formula, p is the probability that the randomly selected point in the iteration process is an inlier, which is generally 0.95, and ω is the inlier ratio in the data set, which is estimated by using an adaptive iteration method, i.e., the inlier ratio of the current model is used to replace the actual inlier ratio in each iteration, so as to estimate the maximum iteration number, and the maximum iteration number K is updated after each iteration.

[0022] Beneficial effects:

[0023] The application discloses a method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectral analysis. The method can improve the detection accuracy of near-infrared spectra of crude oil, and is beneficial to subsequent analysis and processing of the properties of crude oil, which has important value for improving the economic benefits of refining enterprises. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 It is a flowchart of the method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectral analysis according to the application;

[0025] Figure 2 It is the near-infrared spectrum of toluene diluent and each group of dilution mixtures in the embodiment of the application.

[0026] Figure 3 Near infrared spectrum of the reduced Geno crude oil in the embodiment of the present application;

[0027] Figure 4 Iterative process of the embodiment of the present application;

[0028] Figure 5 Each group of error results corresponding to the optimal model after the iteration in the embodiment of the present application. DETAILED DESCRIPTION

[0029] The present application will be further described below in combination with the drawings and specific embodiments. The specific operation process illustrates the implementation effect of the present method in identifying the upper limit of the dilution concentration of crude oil. The present embodiment is implemented on the premise of the technical solution of the present application, but the protection scope of the present application is not limited to the following embodiments.

[0030] The present application takes a certain refining enterprise as an example. The enterprise recently purchased Geno crude oil from Africa. The crude oil has a relatively high density, about 890 kg / cm 3 (20℃). In order to accurately measure the near infrared spectrum of Geno crude oil for property analysis, the present application is used to dilute the crude oil and further identify the upper limit of the dilution concentration of the crude oil. The method steps are shown in Figure 1 , and the specific process is as follows:

[0031] 1) The Geno crude oil to be identified is diluted with different concentration ratios of toluene solvent. The concentration ratio range is taken as 0.15-0.85, and the concentration ratio of the diluent is changed by 0.05 increments to obtain N groups of diluted mixtures. For the present embodiment, N=15, the toluene concentration and the increment of each group are shown in Table 1, and the near infrared spectrum of the mixture is denoted as A m , the near infrared spectrum of toluene is denoted as A m,i , i=1, 2,..., N, and the near infrared spectrum of toluene is denoted as A d . The measured near infrared spectra of the mixture and toluene are shown in Figure 2 .

[0032] Table 1: Toluene concentration of each group

[0033]

[0034]

[0035] 2) The diluent spectrum contribution in A m is removed, and the near infrared spectrum of the crude oil A h is reduced, which is as follows:

[0036]

[0037] where A h (j) is the absorbance of the crude oil at wave number point j, A m (j) is the absorbance of the mixture at wave number point j, A d (j) is the absorbance of the diluent at wave number point j, w is the proportion of toluene concentration, and each group of crude oil spectral data information A h,i is obtained by reducing the mixture from different dilution ratios. Figure 3

[0038] 3) The near infrared spectral range is selected as 4000-4800 cm -1 , and the baseline correction of the crude oil spectrum in S is performed by polynomial fitting to eliminate the influence of baseline drift.

[0039] 4) Randomly select n groups of crude oil spectral data from S to form a subset S1, where n≥2, and n=3 in this embodiment.

[0040] 5) A linear model M1 is established for the spectral data in S1:

[0041] A h (j) = A h,0 (j) + c(j)*w

[0042] where A h,0 (j) represents the true absorbance at wave number point j, c(j)*w represents the absorbance deviation at wave number point j, and c(j) and A h,0 (j) are obtained by least squares, j=1, 2,..., J, J is the number of wave number points, J=426 in this embodiment, and the model is expressed in vector form as A h = A h,0 +C*w, C=[c(1), c(2),..., c(J)] T ;

[0043] 6) All spectral samples in the data set S are predicted using the model M1, and the predicted spectral data is A h,pre , the error E between the predicted spectrum A h,pre and the reduced spectrum A h is calculated as E=||A h -A h,pre ||, and if E>E thr , the current spectrum is determined as an in-point of the model M1, otherwise it is determined as an out-point. E thr is the error threshold, which is obtained by the following formula:

[0044]

[0045] ​where the coefficient a takes 2.5-4.5, and in this embodiment, the coefficient is 3.8;

[0046] 7) Spectra corresponding to all inliers form inlier set denotes the consistency set of S1, and the optimal model is denoted as M best , corresponding to the optimal inlier set If it is the first iteration, M best = M1, Otherwise, compare with the number of inliers, if the number of inliers is greater than update M best and

[0047] 8) Repeat 4) to 7) until the iteration number k≥K, or the inlier ratio exceeds the threshold value t, and the optimal inlier set is considered to be obtained at this time The highest concentration ratio of spectral data in the formula is the upper limit of dilution of the current diluent and crude oil, denoted as w best,max , wherein the inlier ratio threshold value t is 93%, and the maximum iteration number K is obtained by the following formula:

[0048]

[0049] In the formula, the inlier probability p is 0.95, and ω is the inlier ratio in the data set. An adaptive iteration method is used to estimate the prior value, that is, the inlier ratio of the current model is used to replace the actual inlier ratio in each iteration, so as to estimate the maximum iteration number. The maximum iteration number K is updated after each iteration.

[0050] The iteration process is shown in Figure 4 It can be seen that the iteration ends when k=9, and the optimal model M best has an inlier ratio of t=0.80. The error between the predicted spectrum and the reduced spectrum using M best is shown in Figure 5 The errors of groups 13, 14 and 15 exceed the threshold value, and it is determined that the upper limit of the dilution concentration of the toluene dilution of the Geno crude oil is 0.70.

[0051] To further demonstrate the effectiveness of the method, the concentration upper limit of toluene-diluted Geno crude oil is obtained using the method, and the properties of Geno crude oil in groups with a concentration less than the upper limit are detected. The measured properties are consistent with the results of manual laboratory tests, and the deviations of key properties such as naphtha yield and acid value are less than 5%. For groups exceeding the concentration upper limit, the properties of Geno crude oil are also detected, and the measured properties are compared with the test results. There is a phenomenon that the property deviation is greater than 5% and increases with the increase of concentration. This shows that the method has good effectiveness and can accurately identify the upper limit of the concentration of high-density crude oil dilution.

Claims

1. A method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis, characterized in that Based on the linear mixed model, the spectral contribution of the diluent in the crude oil mixture is removed, and a crude oil reduction spectral dataset is constructed. Then, by iteratively selecting data subsets and building a model, the optimal interior point set is obtained to determine the upper limit of the dilution ratio concentration. The following steps are involved: 1) Use diluents of different concentration ratios to dilute the crude oil to be identified, and obtain N groups of diluted mixtures. The near-infrared spectrum of the mixture is recorded as A m , respectively measure the near infrared spectrum A of each group m,i , i=1,2,…,N, and measure the near infrared spectrum of the diluent A d ; 2) Remove the near infrared spectrum A of the i-th group of mixtures m,i The diluent spectrum contribution in the crude oil is restored to the near-infrared spectrum A of this group h,i , all groups of crude oil near infrared spectra A h,i Composition reduction spectrum A h , as follows: Among them A h,i (j) is the absorbance of the crude oil obtained from the mixture of group i at wave number point j, A m,i (j) is the absorbance of the mixture at wave number point j, A d,i (j) is the absorbance of the diluent at wave number point j, w i is the diluent concentration ratio of the i-th group of mixtures, and the crude oil spectral data A of each group is restored. h,i , forming the data set S; 3) Perform polynomial fitting baseline correction on the crude oil spectrum in S; 4) Randomly select n sets of crude oil spectral data from S to form subset S1; 5) Establish a linear model M1 for the spectral data in S1: A 1 h (j)=A 1 h,0 (j)+c(j)*w Where A 1 h (j) represents the absorbance of the crude oil obtained from the reduction in subset S1 at wave number point j, A 1 h,0 (j) represents the true absorbance of crude oil at wave number point j in subset S1, c(j) and A 1 h,0 (j) is obtained by least squares, j = 1, 2, ..., J, J is the number of wave points, and the model is represented by a vector A 1 h =A 1 h,0 +C*w, C=[c(1),c(2),...,c(J)] T , w is the diluent concentration ratio; 6) Use model M1 to predict all spectral samples in the data set S, and the predicted spectral data is A h,pre , calculated and predicted spectrum A h,pre The reduction spectrum A obtained in step 2) h The error E=||A h -A h,pre ||, E>E thr The current spectrum is considered as an interior point of model M1, otherwise it is considered as an exterior point. thr is the error threshold; 7) The spectra corresponding to all interior points constitute the interior point set Represents the consistency set of S1, and the optimal model is M best , corresponding to the optimal interior point set If this is the first iteration, then M best =M1, Otherwise compare and The number of interior points of The number of interior points is greater than Then update M best and 8) Repeat 4) to 7) until the number of iterations k ≥ K, or the proportion of inliers exceeds the threshold t, and the optimal inlier set is considered at this time The highest concentration ratio of the spectrum data is the dilution limit of the current diluent and crude oil, denoted as w best,max .

2. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 1, characterized in that Toluene was selected as the diluent, and the concentration ratio of the diluent was changed by an adjustment variable of 0.

05.

3. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 1, characterized in that The number of spectra in the subset n≥2.

4. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 3, characterized in that The number of spectra in the subset is n=3.

5. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 1, characterized in that Error threshold E thr It is obtained by the following formula: The coefficient a is between 2.5 and 4.

5.

6. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 1, characterized in that The maximum number of iterations K is obtained by the following formula: Where p is the probability that all randomly selected points in the iteration process are inliers, ω is the proportion of inliers in the dataset, n is the number of spectra in the subset, and the maximum number of iterations K is updated after each iteration.

7. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 6, characterized in that p is taken as 0.

95.

8. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 6, characterized in that An adaptive iterative method is used to estimate ω, that is, the inlier ratio of the current model is used to replace the actual inlier ratio in each iteration, thereby estimating the maximum number of iterations.

9. The method for identifying the upper limit of dilution concentration of high-density crude oil in near-infrared spectroscopy analysis according to claim 1, characterized in that The inlier ratio threshold t is set between 85% and 95%.

Images