CN105911016A - Non-linear modeling method for spectral properties of crude oil - Google Patents
Non-linear modeling method for spectral properties of crude oil Download PDFInfo
- Publication number
- CN105911016A CN105911016A CN201610211481.XA CN201610211481A CN105911016A CN 105911016 A CN105911016 A CN 105911016A CN 201610211481 A CN201610211481 A CN 201610211481A CN 105911016 A CN105911016 A CN 105911016A
- Authority
- CN
- China
- Prior art keywords
- crude oil
- character
- spectra
- sample
- modeling method
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000010779 crude oil Substances 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 title claims abstract description 46
- 230000003595 spectral effect Effects 0.000 title abstract description 9
- 238000001228 spectrum Methods 0.000 claims abstract description 31
- IJGRMHOSHXDMSA-UHFFFAOYSA-N Atomic nitrogen Chemical compound N#N IJGRMHOSHXDMSA-UHFFFAOYSA-N 0.000 claims abstract description 26
- 229910052757 nitrogen Inorganic materials 0.000 claims abstract description 13
- 230000009466 transformation Effects 0.000 claims abstract description 9
- 238000012937 correction Methods 0.000 claims abstract description 8
- 238000012549 training Methods 0.000 claims abstract description 8
- 238000010606 normalization Methods 0.000 claims abstract description 7
- 239000002356 single layer Substances 0.000 claims abstract description 6
- 238000013528 artificial neural network Methods 0.000 claims abstract description 4
- 239000003921 oil Substances 0.000 claims description 12
- 238000002329 infrared spectrum Methods 0.000 claims description 11
- 238000002835 absorbance Methods 0.000 claims description 10
- 230000007935 neutral effect Effects 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 5
- 210000005036 nerve Anatomy 0.000 claims 1
- 238000002834 transmittance Methods 0.000 claims 1
- 238000006243 chemical reaction Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000013459 approach Methods 0.000 description 4
- 238000001514 detection method Methods 0.000 description 3
- 238000004611 spectroscopical analysis Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 239000010410 layer Substances 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 239000003209 petroleum derivative Substances 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
Abstract
The invention discloses a non-linear modeling method for the spectral properties of crude oil. The method is oriented to refinery enterprises and based on a near infrared spectral database and a corresponding crude oil property database. The method comprises the following steps: carrying out spectral pretreatment on crude oil in a spectral database and to-be-detected crude oil through baseline correction, interception and vector normalization; then subjecting all the spectra to single-layer wavelet transformation so as to extract spectral characteristics; subjecting crude oil in the spectral database to neural network training; and predicating the properties of crude oil according to training results and the spectral data of the to-be-detected crude oil. Compared with a common linear modeling method, the method has obviously improved prediction precision in prediction of properties like nitrogen content and mass yield of crude oil.
Description
Technical field
The present invention relates to the oil property context of detection of petrochemical industry, the Nonlinear Modeling of a kind of crude oil spectra character
Method.
Background technology
Currently, near infrared spectrum combines appropriate modeling technique, is widely used in the quick inspection of crude oil and petroleum product property
Survey.These modeling techniques, can be roughly divided into two classes, and a class is linear modeling approach, including multiple linear regression, main constituent
Recurrence, offset minimum binary (PLS) etc. are multiple;Another kind of is non-linear modeling method, with artificial neural network, supports vector
The methods such as machine are representative.
In oil property detects, also exist stronger non-linear between the character such as nitrogen content, mass yield and near infrared spectrum
Relation, if predicting this character only with linear modeling approach, can cause precision of prediction not enough.In view of the foregoing, it is considered to adopt
With non-linear modeling method, it is expected to the forecast error overcoming non-linear factor to bring, improves the precision of prediction of oil property.
Summary of the invention
Owing to there is non-linear relation between crude oil some properties and its near infrared spectrum, use single linear modeling approach, difficult
To meet the Petrochemical Enterprises requirement to oil property precision of analysis.For this problem, the present invention proposes a kind of crude oil spectra
The non-linear modeling method of matter.
The present invention uses following technical scheme:
The present invention, based on crude oil near-infrared spectrogram, first carries out Pretreated spectra, including baseline to library of spectra crude oil and crude oil to be measured
Correction, intercepting and vector normalization, carry out monolayer wavelet transformation to spectrogram the most again, extract the low frequency coefficient after wavelet transformation
And reconstruct.The library of spectra crude data handled well is inputted as neutral net, the character (nitrogen content of each spectrogram correspondence oil sample
Or mass yield) export as network.Crude oil to be measured can be carried out character prediction after having trained.
Preferably, this method chooses 6400cm-1And 9200cm-12 wave number points are as two basic points of baseline correction.Pass through following formula
Calculate:
In formula, xiFor the crude oil wave number near infrared spectrum district;kxi+ b was 6400cm-1And 9200cm-1The straight line side of 2
Journey, wherein k is this straight slope, and b is this Linear intercept;yiRepresent that former spectrogram is in wave number xiUnder absorbance;Represent base
After line correction, spectrogram is in wave number xiUnder absorbance.
Test finds, crude oil near-infrared spectrogram all contains much noise at low frequency and high frequency region.It is preferred, therefore, that this method is cut
Take 4000cm-1~4800cm-1Spectrogram in wave number section models.
Preferably, this method, when spectrogram is carried out vector normalization, uses following formula to calculate
In formula, XijRefer to i-th sample absorbance under wave number j;Refer to the absorbance values of i-th sample;M is ripple
The number of several points;Xij *Represent the absorbance under wave number j of the i-th sample after vector normalization.
Preferably, the present invention uses back propagation (BP) neural network algorithm of band momentum term to set up model.
For removing spectral noise further, use single order Daubechies small echo that spectrogram is carried out monolayer wavelet transform, obtain
Low frequency coefficient and high frequency coefficient.Give up high frequency coefficient, use low frequency coefficient to reconstruct again to remove noise.Spectroscopic data after reconstruct
Input as BP neutral net.
Preferably, in this method, the object function of BP neutral net is set to:
In formula, n is the oil sample number of Sample Storehouse;Refer to the character trained values of i-th sample in Sample Storehouse;yiRefer in Sample Storehouse
The character actual value of i-th sample.
Preferably, the character that this method is predicted includes nitrogen content and mass yield.
The present invention is for the Appreciation gist root-mean-square error of final result, i.e. RMSE.Root-mean-square error in engineering survey by extensively
General employing, it one group is measured in especially big or the reflection of special little error is very sensitive, therefore, it is possible to reflect the essence of measurement well
Degree, and can comprehensively weigh a group observations with the deviation between true value.RMSE is calculated by following formula:
In formula, np is the number of crude oil to be measured;Refer to i-th former oil properties predictive value to be measured;xiRefer to i-th crude oil to be measured
Character actual value.RMSE value is the least, illustrates that the degree of accuracy of prediction is the highest, it was predicted that effect is the best.
Beneficial effect:
Detection method provided by the present invention, based on crude oil near infrared spectrum, uses wavelet transformation to combine nerual network technique, it is achieved
The quick detection of crude oil non-linear nature.Compared with general linear modeling approach, this method can quickly and more Accurate Prediction is former
Oil nature, the crude oil on-line blending contributing to Petrochemical Enterprises controls, and then improves the economic benefit of enterprise.
Accompanying drawing explanation
Fig. 1 oil property to be measured fast prediction procedural block diagram.
Fig. 2 (a) is that crude oil sample JLPEC_SRS_101210_0000_3 to be measured is at 4000cm-1~4800cm-1In wave number section
Near infrared spectrum spectrogram.
Fig. 2 (b) is that crude oil sample JLPEC_SUB_100810_1410_3 to be measured is at 4000cm-1~4800cm-1In wave number section
Near infrared spectrum spectrogram.
Fig. 2 (c) is that crude oil sample MPEC02_NAP_140726_0900_140815_1117_3 to be measured exists
4000cm-1~4800cm-1Near infrared spectrum spectrogram in wave number section.
Fig. 2 (d) is that crude oil sample MPEC02_SRA_130801_1200_140326_1321_3 to be measured exists
4000cm-1~4800cm-1Near infrared spectrum spectrogram in wave number section.
Specific implementation process
Below in conjunction with the accompanying drawings and case study on implementation, detailed calculating process and concrete operations flow process are given.Library of spectra in embodiment is adopted
With international INTERTEK crude oil spectra storehouse (using other standards crude oil spectra storehouse can also reach identical result), it is somebody's turn to do
Library of spectra includes nearly 1300 kinds of crude oil spectra data all over the world.The present embodiment enters under premised on technical solution of the present invention
Row is implemented, but protection scope of the present invention is not limited to following embodiment.
As it is shown in figure 1, as a example by the character of nitrogen content is predicted, specific implementation process is as follows:
1) gather crude oil sample to be measured, totally 4, scan through near infrared spectrometer, obtain the near infrared spectrum of crude oil to be measured
Figure, as shown in Fig. 2 (a), 2 (b), 2 (c), 2 (d).
2) spectrogram of crude oil to be measured and library of spectra crude oil is carried out pretreatment.
3) use the dwt function in Matlab workbox that crude oil to be measured and library of spectra crude oil are carried out monolayer wavelet transformation,
Low-frequency spectrum data after conversion, re-use upcoef function and are reconstructed the low frequency coefficient after conversion.
4) by the library of spectra crude data input BP neutral net after conversion, the nitrogen content value of library of spectra crude oil is defeated as training
Go out.BP network structure selects single hidden layer, the number of hidden nodes 25, output node number 1.Parameter is provided that iteration maximum week
Issue 100, learning rate 0.01, factor of momentum 0.1, training objective 0.08.Table 1 is object function E during training
Situation of change.
The situation of change of object function E during table 1 training
Iterations | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
E | 0.1568 | 0.0882 | 0.1241 | 0.1385 | 0.1290 | 0.0920 | 0.0729 |
5) network weight after training and the spectroscopic data to be measured after threshold matrix, and conversion is utilized to calculate the property of crude oil to be measured
Matter value.Table 2 is that the nitrogen content character of crude oil to be measured predicts the outcome.
The nitrogen content character of table 2 crude oil to be measured predicts the outcome
In order to contrast, the library of spectra spectroscopic data after conversion be set up PLS model, obtains fitting coefficient, utilize this matching system
Crude oil spectra data to be measured after number, and conversion calculate the nitrogen content of crude oil to be measured.Table 3 is crude oil to be measured under PLS method
Nitrogen content character predict the outcome.
Under table 3 PLS method, the nitrogen content character of crude oil to be measured predicts the outcome
Contrast table 2 and table 3, it is found that compared to the result of PLS modeling and forecasting, use the crude oil nitrogen of this method prediction to contain
The forecast error of amount all decreases, it was predicted that root-mean-square error RMSE of result is reduced to 0.0869 from 0.1171.
Present case is also adopted by same method and predicts the crude oil to be measured mass yield (W_200-250) at 200 DEG C~250 DEG C, and
Contrast has been made with PLS modeling result.Table 4 is that the W_200-250 character of crude oil to be measured predicts the outcome.
The W_200-250 character of table 4 crude oil to be measured predicts the outcome
Table 5 is that the W_200-250 character of crude oil to be measured under PLS method predicts the outcome.
Under table 5 PLS method, the W_200-250 character of crude oil to be measured predicts the outcome
Contrast table 4 and table 5, compared to the result of PLS modeling and forecasting, the quality at using this method to predict 200 DEG C~250 DEG C
The forecast error of yield has reduced, and root-mean-square error RMSE is reduced to 1.22 from 1.68.
Above example is merely to illustrate technical scheme, rather than limiting the scope of the invention, although with reference to relatively
The present invention has been made to explain by good embodiment, it will be understood by those within the art that, can be to the technology of the present invention
Scheme is modified or equivalent, without deviating from the spirit and scope of technical solution of the present invention.
Claims (8)
1. the non-linear modeling method of a crude oil spectra character, it is characterised in that the method comprises the steps:
1) based on crude oil near infrared light spectrogram, first library of spectra crude oil and crude oil to be measured being carried out Pretreated spectra, described spectrum is pre-
Process includes baseline correction, intercepting and vector normalization;
2) spectrogram after processing is carried out monolayer wavelet transformation, extract the low frequency coefficient after wavelet transformation and reconstruct;
3) the library of spectra crude data handled well being inputted as neutral net, former oil properties exports as network, and training completes
After crude oil to be measured can be carried out character prediction.
The non-linear modeling method of a kind of crude oil spectra character the most according to claim 1, it is characterised in that described baseline correction
Two basic points choose 6400cm-1And 9200cm-12 wave number points, baseline correction is calculated by following formula:
In formula, xiFor the crude oil wave number near infrared spectrum district;kxi+ b was 6400cm in spectrogram-1And 9200cm-1Two wave numbers
The linear equation of point, wherein k is this straight slope, and b is this Linear intercept;yiRepresent that former spectrogram is in wave number xiUnder absorbance;After representing baseline correction, spectrogram is in wave number xiUnder absorbance.
The non-linear modeling method of a kind of crude oil spectra character the most according to claim 1, it is characterised in that described non-linear build
Mould method intercepts 4000cm-1~4800cm-1Interior spectrogram models.
The non-linear modeling method of a kind of crude oil spectra character the most according to claim 1, it is characterised in that spectrogram is carried out vector
During normalization, calculated by following formula:
In formula, XijRefer to i-th sample absorbance under wave number j;Refer to the absorbance values of i-th sample;M is
The number of wave number point;Xij *Represent the absorbance under wave number j of the i-th sample after vector normalization.
The non-linear modeling method of a kind of crude oil spectra character the most according to claim 1, it is characterised in that this method uses one
Rank Daubechies wavelet function realizes monolayer wavelet transformation, to extract the low frequency coefficient after wavelet transformation.
The non-linear modeling method of a kind of crude oil spectra character the most according to claim 1, it is characterised in that the method uses and drives
The reverse transmittance nerve network algorithm of quantifier sets up model, and network is output as the property data of crude oil.
The non-linear modeling method of a kind of crude oil spectra character the most according to claim 6, it is characterised in that the method is using instead
When Propagation Neural Network algorithm calculates the property data of crude oil, object function E is set to:
In formula, n is the oil sample number of Sample Storehouse;Refer to the character trained values of i-th sample in Sample Storehouse;yiRefer in Sample Storehouse i-th
The character actual value of individual sample.
The non-linear modeling method of a kind of crude oil spectra character the most according to claim 1, it is characterised in that described former oil properties
Data include nitrogen content data and mass yield data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610211481.XA CN105911016A (en) | 2016-04-06 | 2016-04-06 | Non-linear modeling method for spectral properties of crude oil |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610211481.XA CN105911016A (en) | 2016-04-06 | 2016-04-06 | Non-linear modeling method for spectral properties of crude oil |
Publications (1)
Publication Number | Publication Date |
---|---|
CN105911016A true CN105911016A (en) | 2016-08-31 |
Family
ID=56744836
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610211481.XA Pending CN105911016A (en) | 2016-04-06 | 2016-04-06 | Non-linear modeling method for spectral properties of crude oil |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105911016A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110987866A (en) * | 2019-12-19 | 2020-04-10 | 汉谷云智(武汉)科技有限公司 | Gasoline property evaluation method and device |
CN113433088A (en) * | 2021-06-25 | 2021-09-24 | 南京富岛信息工程有限公司 | Fine monitoring method for oil mixing section of crude oil long-distance pipeline |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101403689A (en) * | 2008-11-20 | 2009-04-08 | 北京航空航天大学 | Nondestructive detection method for physiological index of plant leaf |
CN102374975A (en) * | 2010-08-19 | 2012-03-14 | 中国石油化工股份有限公司 | Method for predicting physical property data of oil product by using near infrared spectrum |
CN102507516A (en) * | 2011-09-28 | 2012-06-20 | 江南大学 | Method for detecting food pigment by combination of fluorescence spectroscopy and artificial neural network |
CN102590211A (en) * | 2011-01-11 | 2012-07-18 | 郑州大学 | Method for utilizing spectral and image characteristics to grade tobacco leaves |
CN102830087A (en) * | 2011-09-26 | 2012-12-19 | 武汉工业学院 | Method for quickly identifying food waste oils based on near infrared spectroscopy |
CN104020135A (en) * | 2014-06-18 | 2014-09-03 | 中国科学院重庆绿色智能技术研究院 | Calibration model establishing method based on near infrared spectrum |
CN104990893A (en) * | 2015-06-24 | 2015-10-21 | 南京富岛信息工程有限公司 | Gasoline octane number detecting method based on similar discriminance |
-
2016
- 2016-04-06 CN CN201610211481.XA patent/CN105911016A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101403689A (en) * | 2008-11-20 | 2009-04-08 | 北京航空航天大学 | Nondestructive detection method for physiological index of plant leaf |
CN102374975A (en) * | 2010-08-19 | 2012-03-14 | 中国石油化工股份有限公司 | Method for predicting physical property data of oil product by using near infrared spectrum |
CN102590211A (en) * | 2011-01-11 | 2012-07-18 | 郑州大学 | Method for utilizing spectral and image characteristics to grade tobacco leaves |
CN102830087A (en) * | 2011-09-26 | 2012-12-19 | 武汉工业学院 | Method for quickly identifying food waste oils based on near infrared spectroscopy |
CN102507516A (en) * | 2011-09-28 | 2012-06-20 | 江南大学 | Method for detecting food pigment by combination of fluorescence spectroscopy and artificial neural network |
CN104020135A (en) * | 2014-06-18 | 2014-09-03 | 中国科学院重庆绿色智能技术研究院 | Calibration model establishing method based on near infrared spectrum |
CN104990893A (en) * | 2015-06-24 | 2015-10-21 | 南京富岛信息工程有限公司 | Gasoline octane number detecting method based on similar discriminance |
Non-Patent Citations (2)
Title |
---|
李肃义等: "小波变换与神经网络融合法在油页岩近红外光谱分析中的应用", 《光谱学与光谱分析》 * |
陆婉珍: "《现代近红外光谱分析技术》", 31 January 2007, 北京:中国石化出版社 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110987866A (en) * | 2019-12-19 | 2020-04-10 | 汉谷云智(武汉)科技有限公司 | Gasoline property evaluation method and device |
CN113433088A (en) * | 2021-06-25 | 2021-09-24 | 南京富岛信息工程有限公司 | Fine monitoring method for oil mixing section of crude oil long-distance pipeline |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110135612B (en) | Method for monitoring capacity of material supply Shang Sheng and early warning abnormality based on electricity consumption analysis | |
CN107590565B (en) | Method and device for constructing building energy consumption prediction model | |
Di et al. | A four-stage hybrid model for hydrological time series forecasting | |
CN101984340B (en) | Method for identifying flow type of soft grain two-phase turbulence based on artificial intelligence | |
CN104020135B (en) | Calibration model modeling method based near infrared spectrum | |
CN103364364B (en) | Crude oil property rapid detection method based on recombination prediction technology | |
CN108647807B (en) | River flow prediction method | |
CN111783363B (en) | Ionized layer prediction method based on SSA and RBF neural network model | |
CN106407648A (en) | Rapid batch forecast method for key property of gasoline | |
CN112465264A (en) | Food safety risk grade prediction method and device and electronic equipment | |
CN110555570B (en) | Intelligent prediction method and device for gas concentration of mine limited space disaster | |
CN107688863A (en) | The short-term wind speed high accuracy combination forecasting method that adaptive iteration is strengthened | |
CN104573371A (en) | Optimal wavelet packet transformation and AR-SVM based non-stable wind speed prediction method | |
CN107632010B (en) | Method for quantifying steel sample by combining laser-induced breakdown spectroscopy | |
CN105911016A (en) | Non-linear modeling method for spectral properties of crude oil | |
CN111291481B (en) | Bayesian model-based structure early warning analysis method | |
CN109632942B (en) | Inversion method of pipeline defect size based on ensemble learning | |
Li et al. | Energy consumption data prediction analysis based on EEMD-ARMA model | |
CN102539370B (en) | Filtering method for fourier transform infrared spectrum online analysis of multi-component gas | |
CN104361409A (en) | Irrigation control method and system based on crop drought combined prediction model | |
Marin | Dynamical bounds for Sturmian Schrödinger operators | |
Stolzer et al. | Data mining methods applied to flight operations quality assurance data: a comparison to standard statistical methods | |
Sun et al. | Underdetermined blind source separation of pipeline leak vibration signals based on empirical mode decomposition and joint approximate diagonalization of eigenmatrices | |
Xu et al. | The new importance measures based on vector projection for multivariate output: application on hydrological model | |
Yue et al. | Rolling bearing fault diagnosis method based on ELMD hybrid feature extraction and wavelet neural network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160831 |
|
RJ01 | Rejection of invention patent application after publication |