CN103134763A - Method for predicting crude oil density by infrared spectroscopy - Google Patents

Abstract

A method for predicting crude oil density by infrared spectroscopy comprises the following steps: (1) collecting various crude oil samples, determining the densities of the crude oil samples by a standard method, establishing a crude oil density matrix Y; (2) determining infrared spectra of the collected various crude oil samples at different temperatures, performing first-order or second order differential processing, establishing a three dimensional spectrum matrix X (I*J*K) based on absorbances at characteristic spectrum zones of 650-1810 cm-1 and 2750-3100 cm-1, wherein I is the number of the crude oil samples, J is the wavelength point number of the characteristic spectrum zones, and K is the temperature change value, establishing a correction model with the crude oil density matrix established by the standard method by using a multidimensional partial least squares method; (3) determining the infrared spectra of a crude oil sample to be determined at different temperatures under a same condition as the collected crude oil samples, performing first-order or second order differential processing, establishing a three dimensional spectrum matrix Xun based on absorbances at characteristic spectrum zones of 650-1810 cm-1 and 2750-3100 cm-1, substituting Xun into the correction model established in step (2) to obtain the density of the crude oil sample to be determined. The method is rapid in analysis speed, high in testing accuracy, and good in repeatability.

Description

Method for predicting crude oil density by infrared spectroscopy

Technical Field

The invention relates to a method for predicting crude oil density by using a spectrum, in particular to a method for predicting the density of a crude oil sample by using the infrared spectrum of the crude oil sample.

Background

At present, the prices of crude oil produced and traded in global petroleum trading markets vary greatly, and the density of the crude oil is an important reason for determining the price of the crude oil. In terms of processing links, crude oil imported from refineries in China is large in amount and complex in type, and most refineries process blended crude oil. The method for rapidly detecting the density of the crude oil has important significance for optimizing feeding, blending and improving enterprise efficiency.

There are many methods for measuring the density of crude oil, and a densitometer method, a pycnometer method, a U-shaped vibrating tube method (SH/T0604-2000 crude oil and petroleum product densitometer method) and the like are commonly used, but the methods have slow measuring speed and complicated steps and cannot meet the requirement of quick evaluation. In recent years, with the development of spectroscopic technology and chemometrics, molecular spectroscopic analysis technology, especially near infrared spectroscopy (NIR), has been widely used in rapid analysis of oil products due to its advantages of high testing speed, high precision, simple operation, suitability for on-line analysis, etc. Compared with the near infrared spectrum only containing hydrogen radical frequency doubling and combined frequency information, the intermediate infrared spectrum (MIR) contains more molecular functional group information, but is more used for qualitative analysis of molecular structures and difficult to be used for quantitative analysis of oil products due to the inconvenience of using a traditional infrared spectrum instrument and a measurement mode. With the continuous improvement of the manufacturing level of instruments, particularly the performance of measuring accessories, the analysis technology of infrared spectroscopy combined with chemometrics is gradually accepted and applied by people, wherein for the quantitative analysis of liquid, Attenuated Total Reflection (ATR) is a convenient and rapid measuring method, and the method is combined with a multivariate calibration method to be used for rapidly measuring the physicochemical properties of dark heavy oil products, such as the four-component content of residual oil, the residual carbon measurement and the like.

Sena et al in "N-way PLS applied to a biological assay of an acetyl salicylic acid, paracasetic and caffeine" { Journal of Pharmaceutical and biological Analysis, Issue 34, Pages 27-34(2004) }, use a multidimensional partial least squares method to model corrections of acetyl salicylic acid, acetaminophen, caffeine with pH as a conditional variable, and the prediction results show that the predicted standard deviation (RESET) is lower than the deviation using the partial least squares method.

Disclosure of Invention

The invention aims to provide a method for measuring crude oil density by infrared spectroscopy, which has the advantages of high analysis speed, accurate test and good repeatability.

The invention provides a method for predicting crude oil density by infrared spectroscopy, which comprises the following steps:

(1) collecting various crude oil samples, measuring the density of the crude oil samples by a standard method, establishing a crude oil density matrix Y,

(2) measuring the infrared spectra of the collected crude oil samples at different temperatures, performing first-order or second-order differential treatment, and taking 650-1810 cm^-1And 2750-3100 cm^-1The absorbances of the characteristic spectrum region form a three-dimensional spectrum matrixX(I × J × K), wherein I is the number of crude oil samples, J is the number of wavelength points in a characteristic spectrum region, K is the number of temperature changes, and a multi-dimensional partial least square method is adopted to establish a correction model with a crude oil density matrix established by a standard method,

(3) measuring the infrared spectrum of the crude oil sample to be detected at different temperatures under the same condition as the collected crude oil sample, performing first-order or second-order differential treatment, and taking 650-1810 cm^-1And 2750-3100 cm^-1The absorbances of the characteristic spectrum region form a three-dimensional spectrum matrixX ^unAnd (3) substituting the density of the crude oil sample to be detected into the correction model established in the step (2).

The method adopts an Attenuated Total Reflection (ATR) measurement mode, measures the infrared spectrum of a crude oil sample at different temperatures, performs differential processing on the spectrum, selects a proper infrared spectrum characteristic spectrum area, correlates the absorbance corresponding to the characteristic spectrum area with the crude oil density measured by a standard method, establishes a correction model through multiple regression analysis, and predicts the density of the crude oil sample to be measured by the absorbance of the crude oil sample in the selected characteristic spectrum area through the correction model. The method can provide data for formulating the oil refining processing scheme and optimizing the production conditions in time.

Detailed Description

The method selects an infrared spectral region with good correlation with the crude oil density, namely the wave number is 650-1810 cm^-1And 2750-3100 cm^-1The wave band interval of the model is used as a characteristic spectrum area, the absorbance of the crude oil sample measured at different temperatures in the characteristic spectrum area is selected, then the absorbance of the characteristic spectrum area obtained by various crude oils at different temperatures is correlated with the density measured by a standard method, a prediction model is established, and the density of the crude oil sample is predicted by the absorbance measured at different temperatures of the prediction model and an unknown crude oil sample. The method is quick and accurate.

The infrared spectrum is generated by the step transition of the vibrational transition of the molecule. Conventionally, the wavelength is usually 2500 to 25000nm (wave number is 4000 to 400 cm)^-1) The spectral region is called as middle infrared (infrared for short) region, and the wavelength is 780-2500 nm (wave number 12820-4000 cm)^-1) The spectral region of (a) is called the near infrared region. The fundamental frequency of molecular vibration of most organic compounds and many inorganic compounds appears in the infrared region, which is very effective for qualitative analysis and composition analysis of organic structures. Quantification of crude oil Density in this region due to the Presence of fingerprint regionIs more convincing than near infrared spectrum. The wave number is the number of waves contained in a unit centimeter, and the wave number is the reciprocal of the wavelength.

The scanning range for measuring the infrared spectrum of the crude oil sample is 4000-400 cm^-1。

The method for establishing the calibration model comprises the steps of firstly selecting crude oil samples of different types, such as crude oil with different production areas, different genera and different viscosities, then measuring the density of the crude oil samples by using a standard method, and then correlating the absorbance and the density of the crude oil samples in a characteristic spectrum area to establish the calibration model. The standard method for measuring the density of the crude oil used in the step (1) of the invention is SH/T0604-2000, namely a U-shaped vibrating tube method. The larger the number of selected crude oil samples, the more accurate and reliable the model is built. However, in order to reduce the workload in the actual operation, an appropriate number of samples capable of covering all possible predicted values are generally selected, and the number of the crude oil samples of different types is preferably 200-300.

To verify the accuracy of the calibration model, crude oil samples whose density is determined by standard methods are preferably divided into calibration and validation sets. And (4) establishing a correction model by using the correction set samples, wherein the number of the correction set samples is larger than that of the verification set samples. The calibration set samples are representative and their densities should cover the densities of all predicted crude oil samples. The verification set sample is composed of samples randomly drawn from the collected samples, and the verification set sample is used as an unknown sample to verify the accuracy of the correction model. The number of validation set samples was small, about 1/3 for the total number of crude oil samples tested.

After the density of the crude oil sample is measured by a standard method, the infrared spectrum of the crude oil sample is measured by an infrared spectrometer, and then the absorbance of a characteristic spectrum region is subjected to first-order or second-order differential processing to eliminate interference.

The method of the invention measures the absorbance of each crude oil sample at different temperatures, the temperature for measuring the infrared spectrum of the crude oil is 30-60 ℃, and the temperature interval changed in each measurement is 5-10 ℃. Then, a three-dimensional spectrum matrix is established by using the absorbance of the crude oil in the characteristic spectrum region measured at different temperaturesXI.e. the absorbance matrix, the size of which is determined by the number of samples used for modeling, the number of wavelength points in the characteristic spectral region and the number of temperature changes.

Following the invention, the absorbance matrixXThe method associated with the density matrix Y using the N-PLS algorithm is briefly described as follows:

the principle of the N-PLS algorithm is to use a three-dimensional matrixX(I.times.JXK) was decomposed into a trilinear model:

where t is the score vector, w^JAnd w^KFor the corresponding two load vectors, F is the number of main factors, e_ijkIs a residual matrix. Like the conventional PLS, the N-PLS decomposes the spectrum array and decomposes the concentration array (density matrix Y in the invention), and determines the number of main factors by interactive verification by combining two decomposition processes through iteration. The specific algorithm is as follows:

establishing a prediction model:

X(I multiplied by J multiplied by K) is an absorbance matrix, I is the number of crude oil samples used for modeling, J is the number of wavelength points in a characteristic spectrum region, namely the number of sampling points of absorbance in the characteristic spectrum region, and K is the number of temperature changes. Y (I x 1) is the concentration matrix of the modeled sample, and the method of the invention is the density matrix Y.

(1) Will be provided withXExpanded into a two-dimensional matrix X₀(I × JK), namely sequentially splicing the absorbances measured at different temperatures according to the temperature change sequence to form a two-dimensional matrix X₀(I×JK)；

(2) Determining the maximum value of the main factor number, and selecting the main factor number F as 1 one by one;

(3) calculating the Z (JXK) matrix, Z_f＝X_f-1 ^Ty；

(4) Singular value decomposition of the Z matrix, [ wk, s, wj]＝svd(Z_f)

Let w^K＝wk(:，1)，w^J＝wj(:，1)；

(5) Computing $w_f=w^K\⊗w^J;$ (JK×1)

(6) Calculating t_f＝X_f-1w_f；(I×1)

(7) Calculating q_f＝y_f-1 ^Tt_f；(1×1)

(8) Calculating u_f＝y_f-1q_f；(I×1)

(9) Computing $b_f={\left({T_f}^T{T}_f\right)}^{-1}{T_f}^T{u}_f,$ Wherein T is_f＝[t₁，...，t_f]；

(10) Let X_f＝X_f-1-t_fw_f， $y_f=y_{f-1}-T_f{b}_f{q_f}^T;$

(11) And f is f +1, returning to the step (3), and sequentially obtaining the score and the load of each cycle. Then, the sum of squares of Prediction Residuals (PRESS) is plotted using the number of main factors, called PRESS graph, the lowest point of the corresponding PRESS graph is the optimal number of main factors,

wherein y is^*And (4) obtaining a predicted value for interactive verification, wherein y is an actual measured value, and n is the total number of the calibration set samples.

(12) Preservation of w_f，b_fAnd q is_fTo an unknown sample matrixX ^unA concentration value (density in the present invention) prediction is performed.

In the above calculation method, the symbols

Kronecker product (Kronecker product) representing the matrix, the Kronecker product of matrix a (I × J) and matrix C (M × N) is expressed as:

(II) predicting sample density

Spectral matrix for an unknown crude oil sampleX ^un(1 × J × K), the prediction result is calculated by the following steps:

(1) will be provided withX ^unExpanded into a two-dimensional matrix X^un ₀(1 XJK), namely sequentially splicing the absorbances measured at different temperatures according to the temperature condition change sequence,form a two-dimensional matrix X^un ₀(1×JK)；

(2) Call saved w_f，b_fAnd q is_f；

(3) Calculating t_f＝X^un _f-1w_f，X_f＝X^un _f-1-t_fw_f，f＝1，...，F；

(4) Computing $y_{\mathrm{pred}}=\underset{f=1}{\overset{F}{\mathrm{\Σ}}}{T}_f{b}_f{q_f}^T,$ Wherein T is_f＝[t₁，...，t_f]。

The method is suitable for the prediction analysis of the crude oil density, and can predict the density of crude oil samples in main production areas in the world.

The present invention is illustrated in detail below by way of examples, but the present invention is not limited thereto.

Each crude oil sample collected in the examples was subjected to density measurement according to SH/T0604-2000 method and used for establishing a calibration model.

The repeatability of the SH/T0604-2000 method on the measurement result is expressed as follows: the difference between the results of the two repeated tests cannot exceed 0.0004g/cm³The expression for reproducibility results is: the difference between the results of measurements made on the same sample by different operators in different laboratories cannot exceed 0.0015g/cm³。

Example 1

And establishing a crude oil density infrared spectrum correction model and verifying.

(1) Determination of crude oil Density Using Standard methods

280 crude oil samples were collected from each main production zone, their densities were determined by the SH/T0604-2000 method, and a representative sample 236 was collected to make up the calibration set.

(2) Establishing a calibration model using calibration set samples

The infrared spectra of the calibration set samples were measured using a Thermo Nicolet-6700 Fourier transform infrared spectrometer. The measuring accessory is a 45-degree ZnSe ATR variable temperature crystal pool manufactured by Thermo company, and the temperature change range is 30-60 ℃.

The measuring method comprises the following steps: and pouring the test sample into a ZnSe ATR sample absorption cell, heating to 30 ℃, and performing spectrum scanning by taking air as a reference after 2 minutes, wherein the scanning times are 64 times. The spectrum scanning range is 650-4000 cm^-1Resolution of 8cm^-1。

And (3) heating to 40 ℃, 50 ℃ and 60 ℃ in sequence, stabilizing for 2 minutes, and then performing infrared spectrum scanning, wherein the determination time of each sample is 20 minutes. Performing first-order differential processing on the spectrum, and taking 650-1810 cm^-1And 2750-3100 cm^-1Absorbance in the characteristic spectrum region constitutes a three-dimensional absorbance matrixX(I × J × K), where I is the number of calibration set samples, J is the number of wavelength points, K is the number of temperature conditions, where K is 4 and J is 1510.

The density measured by SH/T0604-2000 method corresponding to each crude oil sample is used to form a density matrix Y, and the matrix X and the matrix Y are related by a multidimensional partial least squares method (N-PLS) to establish a crude oil density correction model. The minimum value of the sum of squared Prediction Residuals (PRESS) is calculated to determine the number of the best primary factors of the N-PLS to be 7, and relevant statistical parameters used for establishing the model are shown in Table 1.

Wherein, $\mathrm{RMSEP}=\sqrt{\frac{\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{(y_{i,\mathrm{actual}}-y_{i,\mathrm{predicted}})}^2}{m-1}}$

$R^2=1-\frac{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_{i,\mathrm{actual}}-y_{i,\mathrm{predicted}})}^2}{\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(y_{i,\mathrm{actual}}-{\stackrel{\‾}{y}}_{\mathrm{actual}})}^2}$

in the above formula, m is the total number of samples in the verification set, n is the total number of samples in the calibration set, y_i，actualIs a measured value of a standard method, y_{i，predicted}Is a predicted value.

(3) Verifying accuracy of correction model

Randomly selecting 44 component verification sets of the crude oil sample with the density determined by the standard method in the step (1), measuring the infrared spectrum of the crude oil sample at 30 ℃, 40 ℃, 50 ℃ and 60 ℃, performing first-order differential treatment, and taking 650-1810 cm^-1And 2750-3100 cm^-1Absorbance in the characteristic spectrum region constitutes a three-dimensional spectrum matrixX ^unAnd then expanded into a two-dimensional matrix X^un ₀And substituting the corrected model to obtain the density predicted value of the crude oil sample. The relevant statistical parameters of the verification set are shown in table 1, and the comparison result of the predicted value and the SH/T0604-2000 measured value is shown in table 2.

TABLE 1

TABLE 2

Example 2

Collecting crude oil samples according to the method of example 1, measuring crude oil sample density by standard method, measuring infrared spectrum of each sample at 30 deg.C, 40 deg.C, 50 deg.C and 60 deg.C, except performing second order differential treatment on the spectrum, and collecting 650-1810 cm^-1And 2750-3100 cm^-1Forming a three-dimensional spectrum matrix by absorbance subjected to second-order differential processing in the characteristic spectrum regionXThen, the three-dimensional spectrum matrix is formed by correlating the density matrix Y with an N-PLS method to establish a correction model and then forming a three-dimensional spectrum matrix by the absorbance of a characteristic spectrum area of a verification set sample subjected to second-order differential processingX ^unAnd substituting the corrected model to obtain the density predicted value of the crude oil sample. The relevant statistical parameters of the calibration set and the verification set are shown in Table 3, and the comparison result between the predicted value of the verification set sample and the measured value of the SH/T0604-2000 method is shown in Table 4.

TABLE 3

TABLE 4

Claims (7)

1. A method for predicting crude oil density by infrared spectroscopy comprises the following steps:

(1) collecting various crude oil samples, measuring the density of the crude oil samples by a standard method, establishing a crude oil density matrix Y,

(2) measuring the infrared spectra of the collected crude oil samples at different temperatures, performing first-order or second-order differential treatment, and taking 650-1810 cm^-1And 2750-3100 cm^-1The absorbances of the characteristic spectrum region form a three-dimensional spectrum matrixX(I × J × K), wherein I is the number of crude oil samples, J is the number of wavelength points in the characteristic spectrum region, and K is the temperatureThe degree change number and the crude oil density matrix established by the standard method adopt a multidimensional partial least square method to establish a correction model,

(3) measuring the infrared spectrum of the crude oil sample to be detected at different temperatures under the same condition as the collected crude oil sample, performing first-order or second-order differential treatment, and taking 650-1810 cm^-1And 2750-3100 cm^-1The absorbances of the characteristic spectrum region form a three-dimensional spectrum matrixX ^unAnd (3) substituting the density of the crude oil sample to be detected into the correction model established in the step (2).

2. The method of claim 1, wherein the infrared spectrum of the crude oil sample is measured by a scan ranging from 4000 to 400cm^-1。

3. The method according to claim 1, wherein the standard method for determining the density of the crude oil sample in step (1) is SH/T0604-2000.

4. The method of claim 1 wherein the collected samples of each crude oil are divided into a calibration set and a validation set, the number of samples in the calibration set being greater than the number of samples in the validation set, the calibration model being established using the samples in the calibration set, and the accuracy of the calibration model being verified using the samples in the validation set.

5. The method of claim 4, wherein the density of the calibration set of samples covers the density of all predicted crude oil samples.

6. The method of claim 4, wherein the validation set of samples consists of randomly drawn samples of the collected samples.

7. The method of claim 1, wherein the temperature at which the infrared spectrum of the crude oil sample is measured is 30 to 60 ℃ and the temperature interval for each measurement is 5 to 10 ℃.